diff --git a/README.md b/README.md index 66dba22..0a09b1a 100644 --- a/README.md +++ b/README.md @@ -67,5 +67,6 @@ mamba activate netrans | 单目标跟踪 | ostrack | [链接](./docs/ostrack.md) | | 分类 | resnet18 | [链接](./docs/resnet18.md) | | 关键点检测 | superpoint | [链接](./docs/superpoint.md) | +| 人体姿态估计 | openpose | [链接](./docs/openpose.md) | > *作者 {{liangliangou}}* diff --git a/docs/openpose.md b/docs/openpose.md new file mode 100644 index 0000000..bfefdf7 --- /dev/null +++ b/docs/openpose.md @@ -0,0 +1,69 @@ +# openpose示例 + +## 概述 + +- 模型类型：openpose 人体姿态估计模型，OpenPose 是由卡耐基梅隆大学等机构联合提出的人体姿态估计开源框架，多阶段、多分支网络结构常采用“Part Affinity Fields (PAFs)”+关键点热力图的思想来实现多人姿态的检测与关联。 +- 代码来源： + +## 输入、输出 + +- 输入、输出信息 + + ```text + INPUT:0 + DATA_FORMAT:UINT8 + NUM_OF_DIMENSION:4 + SIZES_OF_DIMENSION:1272 256 1 1 0 0 + QUANT_FORMAT:NONE + OUTPUT:0 + DATA_FORMAT:FP32 + NUM_OF_DIMENSION:4 + SIZES_OF_DIMENSION:53 32 19 1 0 0 + QUANT_FORMAT:NONE + OUTPUT:1 + DATA_FORMAT:FP32 + NUM_OF_DIMENSION:4 + SIZES_OF_DIMENSION:53 32 38 1 0 0 + QUANT_FORMAT:NONE + OUTPUT:2 + DATA_FORMAT:FP32 + NUM_OF_DIMENSION:4 + SIZES_OF_DIMENSION:53 32 19 1 0 0 + QUANT_FORMAT:NONE + OUTPUT:3 + DATA_FORMAT:FP32 + NUM_OF_DIMENSION:4 + SIZES_OF_DIMENSION:53 32 38 1 0 0 + QUANT_FORMAT:NONE + ``` + +## 后处理流程 + +- 非极大值抑制 + 从热力图中定位候选关键点的位置（非极大值点） +- 关键点对齐与骨架候选（候选边）产生 + 基于两端点的坐标，形成潜在的骨架连线候选 +- 基于 PAf 的连接评分与筛选 + 筛选出高置信度且两端点未被其它连接“占用”的骨架候选 +- 构建姿态条目 + 把筛选出的高置信度骨架连接组装成一个或多个完整的姿态条目 +- 多人姿态聚类与去重 + 在图像中可能出现多个人，需将检测到的骼架分配到不同个体 + +## 执行与性能 + +- 执行 + + ```bash + python openpose.py + ``` + +- 推理时间 + 板卡推理时间约 60.35ms + +- 输出fe的相似度 + 板卡的输出2和pc端输出2的fe_mean、fe_max分别为 0.00037104913 0.14568077 + 板卡的输出3和pc端输出3的fe_mean、fe_max分别为 0.00013199929 0.093690544 +- 输出结果 +结果图保存路径 ../resource/openpose/result.jpg +![本地示例图](../resource/openpose/result.jpg) diff --git a/resource/openpose/0.jpg b/resource/openpose/0.jpg new file mode 100644 index 0000000..61b6e24 Binary files /dev/null and b/resource/openpose/0.jpg differ diff --git a/resource/openpose/1.jpg b/resource/openpose/1.jpg new file mode 100644 index 0000000..050c9de Binary files /dev/null and b/resource/openpose/1.jpg differ diff --git a/resource/openpose/dataset.txt b/resource/openpose/dataset.txt new file mode 100644 index 0000000..26bbc2e --- /dev/null +++ b/resource/openpose/dataset.txt @@ -0,0 +1 @@ +0.jpg diff --git a/resource/openpose/openpose.data b/resource/openpose/openpose.data new file mode 100644 index 0000000..1177908 Binary files /dev/null and b/resource/openpose/openpose.data differ diff --git a/resource/openpose/openpose.json b/resource/openpose/openpose.json new file mode 100644 index 0000000..7b83750 --- /dev/null +++ b/resource/openpose/openpose.json @@ -0,0 +1,2903 @@ +{ + "MetaData": { + "Name": "torch_jit", + "NetransVersion": "6", + "Platform": "tensorflow", + "Org_Platform": "onnx" + }, + "Layers": { + "attach_onnx//Concat_345/out0_0": { + "name": "attach_onnx//Concat_345/out0", + "op": "output", + "parameters": {}, + "inputs": [ + "@attach_onnx//Concat_345/out0_0_netrans_mark_perm_125:out0" + ], + "outputs": [ + "out0" + ] + }, + "attach_onnx//Concat_348/out0_1": { + "name": "attach_onnx//Concat_348/out0", + "op": "output", + "parameters": {}, + "inputs": [ + "@attach_onnx//Concat_348/out0_1_netrans_mark_perm_124:out0" + ], + "outputs": [ + "out0" + ] + }, + "attach_397/out0_2": { + "name": "attach_397/out0", + "op": "output", + "parameters": {}, + "inputs": [ + "@attach_397/out0_2_netrans_mark_perm_123:out0" + ], + "outputs": [ + "out0" + ] + }, + "attach_400/out0_3": { + "name": "attach_400/out0", + "op": "output", + "parameters": {}, + "inputs": [ + "@attach_400/out0_3_netrans_mark_perm_122:out0" + ], + "outputs": [ + "out0" + ] + }, + "400_4": { + "name": "400", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 38, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8:out0" + ], + "outputs": [ + "out0" + ] + }, + "397_5": { + "name": "397", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 19, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9:out0" + ], + "outputs": [ + "out0" + ] + }, + "onnx//Concat_348_6": { + "name": "onnx//Concat_348", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 38, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10:out0" + ], + "outputs": [ + "out0" + ] + }, + "onnx//Concat_345_7": { + "name": "onnx//Concat_345", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 19, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8": { + "name": "refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9": { + "name": "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10": { + "name": "initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11": { + "name": "initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12": { + "name": "refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13": { + "name": "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14": { + "name": "initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15": { + "name": "initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.4/Add_output_0_16": { + "name": "refinement_stages.0/trunk/trunk.4/Add_output_0", + "op": "add", + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0", + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17": { + "name": "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18": { + "name": "initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19": { + "name": "initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20": { + "name": "refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21": { + "name": "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 2, + "pad_w": 2, + "dilation": [ + 1, + 2, + 2, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 2, + 2, + 2, + 2 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22": { + "name": "initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23": { + "name": "refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.3/Add_output_0_26:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24": { + "name": "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25": { + "name": "initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.3/Add_output_0_26": { + "name": "refinement_stages.0/trunk/trunk.3/Add_output_0", + "op": "add", + "inputs": [ + "@refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0", + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27": { + "name": "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28": { + "name": "initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29": { + "name": "refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30": { + "name": "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31": { + "name": "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 2, + "pad_w": 2, + "dilation": [ + 1, + 2, + 2, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 2, + 2, + 2, + 2 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34:out0" + ], + "outputs": [ + "out0" + ] + }, + "initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32": { + "name": "initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@cpm/conv/conv.1/Relu_output_0_35:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33": { + "name": "refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.2/Add_output_0_36:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34": { + "name": "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/conv/conv.1/Relu_output_0_35": { + "name": "cpm/conv/conv.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@cpm/conv/conv.0/Conv_output_0_38:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.2/Add_output_0_36": { + "name": "refinement_stages.0/trunk/trunk.2/Add_output_0", + "op": "add", + "inputs": [ + "@refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0", + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37": { + "name": "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/conv/conv.0/Conv_output_0_38": { + "name": "cpm/conv/conv.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@cpm/Add_output_0_42:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39": { + "name": "refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40": { + "name": "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41": { + "name": "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 2, + "pad_w": 2, + "dilation": [ + 1, + 2, + 2, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 2, + 2, + 2, + 2 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/Add_output_0_42": { + "name": "cpm/Add_output_0", + "op": "add", + "inputs": [ + "@cpm/align/align.1/Relu_output_0_45:out0", + "@cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43": { + "name": "refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.1/Add_output_0_47:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44": { + "name": "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/align/align.1/Relu_output_0_45": { + "name": "cpm/align/align.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@cpm/align/align.0/Conv_output_0_49:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46": { + "name": "cpm/trunk/trunk.2/trunk.2.3/Elu_output_0", + "op": "elu", + "parameters": { + "alpha": 1.0 + }, + "inputs": [ + "@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.1/Add_output_0_47": { + "name": "refinement_stages.0/trunk/trunk.1/Add_output_0", + "op": "add", + "inputs": [ + "@refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0", + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48": { + "name": "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/align/align.0/Conv_output_0_49": { + "name": "cpm/align/align.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.11/model.11.5/Relu_output_0_54:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50": { + "name": "cpm/trunk/trunk.2/trunk.2.2/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51": { + "name": "refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52": { + "name": "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53": { + "name": "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 2, + "pad_w": 2, + "dilation": [ + 1, + 2, + 2, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 2, + 2, + 2, + 2 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.11/model.11.5/Relu_output_0_54": { + "name": "model/model.11/model.11.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.11/model.11.3/Conv_output_0_58:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55": { + "name": "cpm/trunk/trunk.2/trunk.2.1/Elu_output_0", + "op": "elu", + "parameters": { + "alpha": 1.0 + }, + "inputs": [ + "@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56": { + "name": "refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.0/Add_output_0_60:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57": { + "name": "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.11/model.11.3/Conv_output_0_58": { + "name": "model/model.11/model.11.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.11/model.11.2/Relu_output_0_62:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59": { + "name": "cpm/trunk/trunk.2/trunk.2.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 128, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.0/Add_output_0_60": { + "name": "refinement_stages.0/trunk/trunk.0/Add_output_0", + "op": "add", + "inputs": [ + "@refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0", + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61": { + "name": "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.11/model.11.2/Relu_output_0_62": { + "name": "model/model.11/model.11.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.11/model.11.0/Conv_output_0_67:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63": { + "name": "cpm/trunk/trunk.1/trunk.1.3/Elu_output_0", + "op": "elu", + "parameters": { + "alpha": 1.0 + }, + "inputs": [ + "@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64": { + "name": "refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65": { + "name": "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66": { + "name": "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 2, + "pad_w": 2, + "dilation": [ + 1, + 2, + 2, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 2, + 2, + 2, + 2 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.11/model.11.0/Conv_output_0_67": { + "name": "model/model.11/model.11.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 512, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.10/model.10.5/Relu_output_0_71:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68": { + "name": "cpm/trunk/trunk.1/trunk.1.2/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69": { + "name": "refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@Concat_output_0_73:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70": { + "name": "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.10/model.10.5/Relu_output_0_71": { + "name": "model/model.10/model.10.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.10/model.10.3/Conv_output_0_75:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72": { + "name": "cpm/trunk/trunk.1/trunk.1.1/Elu_output_0", + "op": "elu", + "parameters": { + "alpha": 1.0 + }, + "inputs": [ + "@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:out0" + ], + "outputs": [ + "out0" + ] + }, + "Concat_output_0_73": { + "name": "Concat_output_0", + "op": "concat", + "parameters": { + "dim": 3 + }, + "inputs": [ + "@cpm/conv/conv.1/Relu_output_0_35:out0", + "@onnx//Concat_345_7:out0", + "@onnx//Concat_348_6:out0" + ], + "outputs": [ + "out0" + ] + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74": { + "name": "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.10/model.10.3/Conv_output_0_75": { + "name": "model/model.10/model.10.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.10/model.10.2/Relu_output_0_81:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76": { + "name": "cpm/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 128, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77": { + "name": "cpm/trunk/trunk.0/trunk.0.3/Elu_output_0", + "op": "elu", + "parameters": { + "alpha": 1.0 + }, + "inputs": [ + "@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78": { + "name": "cpm/trunk/trunk.0/trunk.0.2/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79": { + "name": "cpm/trunk/trunk.0/trunk.0.1/Elu_output_0", + "op": "elu", + "parameters": { + "alpha": 1.0 + }, + "inputs": [ + "@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:out0" + ], + "outputs": [ + "out0" + ] + }, + "cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80": { + "name": "cpm/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 128, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@cpm/align/align.1/Relu_output_0_45:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.10/model.10.2/Relu_output_0_81": { + "name": "model/model.10/model.10.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.10/model.10.0/Conv_output_0_82:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.10/model.10.0/Conv_output_0_82": { + "name": "model/model.10/model.10.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 512, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.9/model.9.5/Relu_output_0_83:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.9/model.9.5/Relu_output_0_83": { + "name": "model/model.9/model.9.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.9/model.9.3/Conv_output_0_84:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.9/model.9.3/Conv_output_0_84": { + "name": "model/model.9/model.9.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.9/model.9.2/Relu_output_0_85:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.9/model.9.2/Relu_output_0_85": { + "name": "model/model.9/model.9.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.9/model.9.0/Conv_output_0_86:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.9/model.9.0/Conv_output_0_86": { + "name": "model/model.9/model.9.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 512, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.8/model.8.5/Relu_output_0_87:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.8/model.8.5/Relu_output_0_87": { + "name": "model/model.8/model.8.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.8/model.8.3/Conv_output_0_88:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.8/model.8.3/Conv_output_0_88": { + "name": "model/model.8/model.8.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.8/model.8.2/Relu_output_0_89:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.8/model.8.2/Relu_output_0_89": { + "name": "model/model.8/model.8.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.8/model.8.0/Conv_output_0_90:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.8/model.8.0/Conv_output_0_90": { + "name": "model/model.8/model.8.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 512, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.7/model.7.5/Relu_output_0_91:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.7/model.7.5/Relu_output_0_91": { + "name": "model/model.7/model.7.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.7/model.7.3/Conv_output_0_92:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.7/model.7.3/Conv_output_0_92": { + "name": "model/model.7/model.7.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.7/model.7.2/Relu_output_0_93:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.7/model.7.2/Relu_output_0_93": { + "name": "model/model.7/model.7.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.7/model.7.0/Conv_output_0_94:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.7/model.7.0/Conv_output_0_94": { + "name": "model/model.7/model.7.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 512, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 2, + "pad_w": 2, + "dilation": [ + 1, + 2, + 2, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 2, + 2, + 2, + 2 + ] + }, + "inputs": [ + "@model/model.6/model.6.5/Relu_output_0_95:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.6/model.6.5/Relu_output_0_95": { + "name": "model/model.6/model.6.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.6/model.6.3/Conv_output_0_96:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.6/model.6.3/Conv_output_0_96": { + "name": "model/model.6/model.6.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 512, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.6/model.6.2/Relu_output_0_97:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.6/model.6.2/Relu_output_0_97": { + "name": "model/model.6/model.6.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.6/model.6.0/Conv_output_0_98:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.6/model.6.0/Conv_output_0_98": { + "name": "model/model.6/model.6.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 256, + "padding": "VALID", + "bias": true, + "group_number": 256, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.5/model.5.5/Relu_output_0_99:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.5/model.5.5/Relu_output_0_99": { + "name": "model/model.5/model.5.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.5/model.5.3/Conv_output_0_100:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.5/model.5.3/Conv_output_0_100": { + "name": "model/model.5/model.5.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 256, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.5/model.5.2/Relu_output_0_101:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.5/model.5.2/Relu_output_0_101": { + "name": "model/model.5/model.5.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.5/model.5.0/Conv_output_0_102:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.5/model.5.0/Conv_output_0_102": { + "name": "model/model.5/model.5.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 256, + "padding": "VALID", + "bias": true, + "group_number": 256, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.4/model.4.5/Relu_output_0_103:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.4/model.4.5/Relu_output_0_103": { + "name": "model/model.4/model.4.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.4/model.4.3/Conv_output_0_104:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.4/model.4.3/Conv_output_0_104": { + "name": "model/model.4/model.4.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 256, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.4/model.4.2/Relu_output_0_105:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.4/model.4.2/Relu_output_0_105": { + "name": "model/model.4/model.4.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.4/model.4.0/Conv_output_0_106:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.4/model.4.0/Conv_output_0_106": { + "name": "model/model.4/model.4.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 128, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 2, + "stride_w": 2, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.3/model.3.5/Relu_output_0_107:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.3/model.3.5/Relu_output_0_107": { + "name": "model/model.3/model.3.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.3/model.3.3/Conv_output_0_108:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.3/model.3.3/Conv_output_0_108": { + "name": "model/model.3/model.3.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.3/model.3.2/Relu_output_0_109:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.3/model.3.2/Relu_output_0_109": { + "name": "model/model.3/model.3.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.3/model.3.0/Conv_output_0_110:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.3/model.3.0/Conv_output_0_110": { + "name": "model/model.3/model.3.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 128, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.2/model.2.5/Relu_output_0_111:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.2/model.2.5/Relu_output_0_111": { + "name": "model/model.2/model.2.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.2/model.2.3/Conv_output_0_112:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.2/model.2.3/Conv_output_0_112": { + "name": "model/model.2/model.2.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 128, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.2/model.2.2/Relu_output_0_113:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.2/model.2.2/Relu_output_0_113": { + "name": "model/model.2/model.2.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.2/model.2.0/Conv_output_0_114:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.2/model.2.0/Conv_output_0_114": { + "name": "model/model.2/model.2.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 64, + "padding": "VALID", + "bias": true, + "group_number": 64, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 2, + "stride_w": 2, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.1/model.1.5/Relu_output_0_115:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.1/model.1.5/Relu_output_0_115": { + "name": "model/model.1/model.1.5/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.1/model.1.3/Conv_output_0_116:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.1/model.1.3/Conv_output_0_116": { + "name": "model/model.1/model.1.3/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 64, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 1, + "ksize_w": 1, + "stride_h": 1, + "stride_w": 1, + "pad_h": 0, + "pad_w": 0, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 0, + 0, + 0, + 0 + ] + }, + "inputs": [ + "@model/model.1/model.1.2/Relu_output_0_117:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.1/model.1.2/Relu_output_0_117": { + "name": "model/model.1/model.1.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.1/model.1.0/Conv_output_0_118:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.1/model.1.0/Conv_output_0_118": { + "name": "model/model.1/model.1.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 32, + "padding": "VALID", + "bias": true, + "group_number": 32, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 1, + "stride_w": 1, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.0/model.0.2/Relu_output_0_119:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.0/model.0.2/Relu_output_0_119": { + "name": "model/model.0/model.0.2/Relu_output_0", + "op": "relu", + "inputs": [ + "@model/model.0/model.0.0/Conv_output_0_120:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.0/model.0.0/Conv_output_0_120": { + "name": "model/model.0/model.0.0/Conv_output_0", + "op": "convolution", + "parameters": { + "padding_mode": "CONSTANT", + "weights": 32, + "padding": "VALID", + "bias": true, + "group_number": 1, + "regularize": false, + "ksize_h": 3, + "ksize_w": 3, + "stride_h": 2, + "stride_w": 2, + "pad_h": 1, + "pad_w": 1, + "dilation": [ + 1, + 1, + 1, + 1 + ], + "pad_method": "padding_const", + "pad": [ + 1, + 1, + 1, + 1 + ] + }, + "inputs": [ + "@model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm_126:out0" + ], + "outputs": [ + "out0" + ] + }, + "input.1_121": { + "name": "input.1", + "op": "input", + "parameters": { + "size": "", + "channels": 1, + "shape": [ + 1, + 3, + 256, + 424 + ], + "is_scalar": false, + "type": "float32" + }, + "inputs": [], + "outputs": [ + "out0" + ] + }, + "attach_400/out0_3_netrans_mark_perm_122": { + "name": "attach_400/out0_3_netrans_mark_perm", + "op": "permute", + "parameters": { + "perm": [ + 0, + 3, + 1, + 2 + ] + }, + "inputs": [ + "@400_4:out0" + ], + "outputs": [ + "out0" + ] + }, + "attach_397/out0_2_netrans_mark_perm_123": { + "name": "attach_397/out0_2_netrans_mark_perm", + "op": "permute", + "parameters": { + "perm": [ + 0, + 3, + 1, + 2 + ] + }, + "inputs": [ + "@397_5:out0" + ], + "outputs": [ + "out0" + ] + }, + "attach_onnx//Concat_348/out0_1_netrans_mark_perm_124": { + "name": "attach_onnx//Concat_348/out0_1_netrans_mark_perm", + "op": "permute", + "parameters": { + "perm": [ + 0, + 3, + 1, + 2 + ] + }, + "inputs": [ + "@onnx//Concat_348_6:out0" + ], + "outputs": [ + "out0" + ] + }, + "attach_onnx//Concat_345/out0_0_netrans_mark_perm_125": { + "name": "attach_onnx//Concat_345/out0_0_netrans_mark_perm", + "op": "permute", + "parameters": { + "perm": [ + 0, + 3, + 1, + 2 + ] + }, + "inputs": [ + "@onnx//Concat_345_7:out0" + ], + "outputs": [ + "out0" + ] + }, + "model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm_126": { + "name": "model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm", + "op": "permute", + "parameters": { + "perm": [ + 0, + 2, + 3, + 1 + ] + }, + "inputs": [ + "@input.1_121:out0" + ], + "outputs": [ + "out0" + ] + } + }, + "quantize_info": {} +} \ No newline at end of file diff --git a/resource/openpose/openpose.onnx b/resource/openpose/openpose.onnx new file mode 100644 index 0000000..e47c2b5 Binary files /dev/null and b/resource/openpose/openpose.onnx differ diff --git a/resource/openpose/openpose_asymmetric_affine.quantize b/resource/openpose/openpose_asymmetric_affine.quantize new file mode 100644 index 0000000..4dc5287 --- /dev/null +++ b/resource/openpose/openpose_asymmetric_affine.quantize @@ -0,0 +1,1924 @@ +# !!!This file disallow TABs!!! +version: 2 +quantize_parameters: + '@attach_onnx//Concat_345/out0_0:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + scale: 0.005909114144742489 + zero_point: 81 + '@attach_onnx//Concat_348/out0_1:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + scale: 0.005909114144742489 + zero_point: 81 + '@attach_397/out0_2:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0012764930725098 + min_value: -0.022866368293762207 + scale: 0.004016246646642685 + zero_point: 6 + '@attach_400/out0_3:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.068938136100769 + min_value: -0.6061154007911682 + scale: 0.006568837445229292 + zero_point: 92 + '@400_4:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.068938136100769 + min_value: -0.6061154007911682 + scale: 0.006568837445229292 + zero_point: 92 + '@397_5:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0012764930725098 + min_value: -0.022866368293762207 + scale: 0.004016246646642685 + zero_point: 6 + '@onnx//Concat_348_6:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + scale: 0.005909114144742489 + zero_point: 81 + '@onnx//Concat_345_7:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + scale: 0.005909114144742489 + zero_point: 81 + '@refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.8483352661132812 + min_value: 0.0 + scale: 0.007248373702168465 + zero_point: 0 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.1169047355651855 + min_value: 0.0 + scale: 0.008301586844027042 + zero_point: 0 + '@initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.43290501832962036 + min_value: 0.0 + scale: 0.001697666710242629 + zero_point: 0 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.493980348110199 + min_value: 0.0 + scale: 0.0019371778471395373 + zero_point: 0 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.8483352661132812 + min_value: 0.0 + scale: 0.007248373702168465 + zero_point: 0 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.1169047355651855 + min_value: 0.0 + scale: 0.008301586844027042 + zero_point: 0 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.43290501832962036 + min_value: 0.0 + scale: 0.001697666710242629 + zero_point: 0 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.493980348110199 + min_value: 0.0 + scale: 0.0019371778471395373 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.317978382110596 + min_value: 0.0 + scale: 0.02085481770336628 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.195100784301758 + min_value: 0.0 + scale: 0.020372943952679634 + zero_point: 0 + '@initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.3359031677246094 + min_value: 0.0 + scale: 0.005238836165517569 + zero_point: 0 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.3359031677246094 + min_value: 0.0 + scale: 0.005238836165517569 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1060172319412231 + min_value: 0.0 + scale: 0.004337322432547808 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.195100784301758 + min_value: 0.0 + scale: 0.020372943952679634 + zero_point: 0 + '@initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.7721197009086609 + min_value: 0.0 + scale: 0.003027920378372073 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1060172319412231 + min_value: 0.0 + scale: 0.004337322432547808 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9883601665496826 + min_value: 0.0 + scale: 0.007797490805387497 + zero_point: 0 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.7721197009086609 + min_value: 0.0 + scale: 0.003027920378372073 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/Add_output_0_26:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.966081976890564 + min_value: 0.0 + scale: 0.007710125297307968 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9883601665496826 + min_value: 0.0 + scale: 0.007797490805387497 + zero_point: 0 + '@initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.6254621744155884 + min_value: 0.0 + scale: 0.0024527928326278925 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.2146551609039307 + min_value: 0.0 + scale: 0.004763353615999222 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9338730573654175 + min_value: 0.0 + scale: 0.007583816070109606 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9338730573654175 + min_value: 0.0 + scale: 0.007583816070109606 + zero_point: 0 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.6254621744155884 + min_value: 0.0 + scale: 0.0024527928326278925 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.2146551609039307 + min_value: 0.0 + scale: 0.004763353615999222 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9390138387680054 + min_value: 0.0 + scale: 0.00760397594422102 + zero_point: 0 + '@cpm/conv/conv.1/Relu_output_0_35:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: 0.0 + scale: 0.004036176949739456 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/Add_output_0_36:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.489450693130493 + min_value: 0.0 + scale: 0.009762551635503769 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9390138387680054 + min_value: 0.0 + scale: 0.00760397594422102 + zero_point: 0 + '@cpm/conv/conv.0/Conv_output_0_38:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: 0.0 + scale: 0.004036176949739456 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.7065311670303345 + min_value: 0.0 + scale: 0.006692279130220413 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.8692786693572998 + min_value: 0.0 + scale: 0.007330504711717367 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.8692786693572998 + min_value: 0.0 + scale: 0.007330504711717367 + zero_point: 0 + '@cpm/Add_output_0_42:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.6703611016273499 + min_value: 0.0 + scale: 0.0026288670487701893 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.7065311670303345 + min_value: 0.0 + scale: 0.006692279130220413 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.845794439315796 + min_value: 0.0 + scale: 0.007238409481942654 + zero_point: 0 + '@cpm/align/align.1/Relu_output_0_45:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.6703611016273499 + min_value: 0.0 + scale: 0.0026288670487701893 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/Add_output_0_47:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.222336530685425 + min_value: 0.0 + scale: 0.008715045638382435 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.845794439315796 + min_value: 0.0 + scale: 0.007238409481942654 + zero_point: 0 + '@cpm/align/align.0/Conv_output_0_49:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.6703611016273499 + min_value: 0.0 + scale: 0.0026288670487701893 + zero_point: 0 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.7675563097000122 + min_value: 0.0 + scale: 0.0069315931759774685 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.1082682609558105 + min_value: 0.0 + scale: 0.008267718367278576 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.1082682609558105 + min_value: 0.0 + scale: 0.008267718367278576 + zero_point: 0 + '@model/model.11/model.11.5/Relu_output_0_54:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0608251094818115 + min_value: 0.0 + scale: 0.0041600982658565044 + zero_point: 0 + '@cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.7675563097000122 + min_value: 0.0 + scale: 0.0069315931759774685 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.027811050415039 + min_value: 0.0 + scale: 0.007952200248837471 + zero_point: 0 + '@model/model.11/model.11.3/Conv_output_0_58:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0608251094818115 + min_value: 0.0 + scale: 0.0041600982658565044 + zero_point: 0 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/Add_output_0_60:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.873496174812317 + min_value: 0.0 + scale: 0.007347043603658676 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.027811050415039 + min_value: 0.0 + scale: 0.007952200248837471 + zero_point: 0 + '@model/model.11/model.11.2/Relu_output_0_62:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 9.764772415161133 + min_value: 0.0 + scale: 0.038293223828077316 + zero_point: 0 + '@cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.722118616104126 + min_value: 0.0 + scale: 0.0028318376280367374 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.801229476928711 + min_value: 0.0 + scale: 0.00706364493817091 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.801229476928711 + min_value: 0.0 + scale: 0.00706364493817091 + zero_point: 0 + '@model/model.11/model.11.0/Conv_output_0_67:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 9.764772415161133 + min_value: 0.0 + scale: 0.038293223828077316 + zero_point: 0 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.722118616104126 + min_value: 0.0 + scale: 0.0028318376280367374 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.3806896209716797 + min_value: 0.0 + scale: 0.009336037561297417 + zero_point: 0 + '@model/model.10/model.10.5/Relu_output_0_71:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0497770309448242 + min_value: 0.0 + scale: 0.004116772674024105 + zero_point: 0 + '@cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@Concat_output_0_73:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + scale: 0.005909114144742489 + zero_point: 81 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.3806896209716797 + min_value: 0.0 + scale: 0.009336037561297417 + zero_point: 0 + '@model/model.10/model.10.3/Conv_output_0_75:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0497770309448242 + min_value: 0.0 + scale: 0.004116772674024105 + zero_point: 0 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.033857811242341995 + min_value: -0.056717514991760254 + scale: 0.000355197349563241 + zero_point: 160 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.033857811242341995 + min_value: -0.05838948115706444 + scale: 0.00036175409331917763 + zero_point: 161 + '@model/model.10/model.10.2/Relu_output_0_81:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.415395498275757 + min_value: 0.0 + scale: 0.009472139179706573 + zero_point: 0 + '@model/model.10/model.10.0/Conv_output_0_82:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.415395498275757 + min_value: 0.0 + scale: 0.009472139179706573 + zero_point: 0 + '@model/model.9/model.9.5/Relu_output_0_83:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.130396842956543 + min_value: 0.0 + scale: 0.0044329287484288216 + zero_point: 0 + '@model/model.9/model.9.3/Conv_output_0_84:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.130396842956543 + min_value: 0.0 + scale: 0.0044329287484288216 + zero_point: 0 + '@model/model.9/model.9.2/Relu_output_0_85:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.489227533340454 + min_value: 0.0 + scale: 0.00976167619228363 + zero_point: 0 + '@model/model.9/model.9.0/Conv_output_0_86:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.489227533340454 + min_value: 0.0 + scale: 0.00976167619228363 + zero_point: 0 + '@model/model.8/model.8.5/Relu_output_0_87:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1125234365463257 + min_value: 0.0 + scale: 0.004362836945801973 + zero_point: 0 + '@model/model.8/model.8.3/Conv_output_0_88:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1125234365463257 + min_value: 0.0 + scale: 0.004362836945801973 + zero_point: 0 + '@model/model.8/model.8.2/Relu_output_0_89:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 15.23180866241455 + min_value: 0.0 + scale: 0.0597325824201107 + zero_point: 0 + '@model/model.8/model.8.0/Conv_output_0_90:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 15.23180866241455 + min_value: 0.0 + scale: 0.0597325824201107 + zero_point: 0 + '@model/model.7/model.7.5/Relu_output_0_91:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0009398460388184 + min_value: 0.0 + scale: 0.003925254102796316 + zero_point: 0 + '@model/model.7/model.7.3/Conv_output_0_92:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0009398460388184 + min_value: 0.0 + scale: 0.003925254102796316 + zero_point: 0 + '@model/model.7/model.7.2/Relu_output_0_93:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.2520294189453125 + min_value: 0.0 + scale: 0.02059619314968586 + zero_point: 0 + '@model/model.7/model.7.0/Conv_output_0_94:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.2520294189453125 + min_value: 0.0 + scale: 0.02059619314968586 + zero_point: 0 + '@model/model.6/model.6.5/Relu_output_0_95:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.047563910484314 + min_value: 0.0 + scale: 0.004108093678951263 + zero_point: 0 + '@model/model.6/model.6.3/Conv_output_0_96:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.047563910484314 + min_value: 0.0 + scale: 0.004108093678951263 + zero_point: 0 + '@model/model.6/model.6.2/Relu_output_0_97:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 26.618968963623047 + min_value: 0.0 + scale: 0.10438811033964157 + zero_point: 0 + '@model/model.6/model.6.0/Conv_output_0_98:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 26.618968963623047 + min_value: 0.0 + scale: 0.10438811033964157 + zero_point: 0 + '@model/model.5/model.5.5/Relu_output_0_99:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.6952513456344604 + min_value: 0.0 + scale: 0.006648044567555189 + zero_point: 0 + '@model/model.5/model.5.3/Conv_output_0_100:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.6952513456344604 + min_value: 0.0 + scale: 0.006648044567555189 + zero_point: 0 + '@model/model.5/model.5.2/Relu_output_0_101:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.9576454162597656 + min_value: 0.0 + scale: 0.011598609387874603 + zero_point: 0 + '@model/model.5/model.5.0/Conv_output_0_102:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.9576454162597656 + min_value: 0.0 + scale: 0.011598609387874603 + zero_point: 0 + '@model/model.4/model.4.5/Relu_output_0_103:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.42368483543396 + min_value: 0.0 + scale: 0.005583077669143677 + zero_point: 0 + '@model/model.4/model.4.3/Conv_output_0_104:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.42368483543396 + min_value: 0.0 + scale: 0.005583077669143677 + zero_point: 0 + '@model/model.4/model.4.2/Relu_output_0_105:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 27.753128051757812 + min_value: 0.0 + scale: 0.10883579403162003 + zero_point: 0 + '@model/model.4/model.4.0/Conv_output_0_106:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 27.753128051757812 + min_value: 0.0 + scale: 0.10883579403162003 + zero_point: 0 + '@model/model.3/model.3.5/Relu_output_0_107:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.396226644515991 + min_value: 0.0 + scale: 0.009396967478096485 + zero_point: 0 + '@model/model.3/model.3.3/Conv_output_0_108:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.396226644515991 + min_value: 0.0 + scale: 0.009396967478096485 + zero_point: 0 + '@model/model.3/model.3.2/Relu_output_0_109:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 3.3708269596099854 + min_value: 0.0 + scale: 0.013218929059803486 + zero_point: 0 + '@model/model.3/model.3.0/Conv_output_0_110:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 3.3708269596099854 + min_value: 0.0 + scale: 0.013218929059803486 + zero_point: 0 + '@model/model.2/model.2.5/Relu_output_0_111:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9996494054794312 + min_value: 0.0 + scale: 0.007841762155294418 + zero_point: 0 + '@model/model.2/model.2.3/Conv_output_0_112:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.9996494054794312 + min_value: 0.0 + scale: 0.007841762155294418 + zero_point: 0 + '@model/model.2/model.2.2/Relu_output_0_113:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.853857040405273 + min_value: 0.0 + scale: 0.022956302389502525 + zero_point: 0 + '@model/model.2/model.2.0/Conv_output_0_114:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.853857040405273 + min_value: 0.0 + scale: 0.022956302389502525 + zero_point: 0 + '@model/model.1/model.1.5/Relu_output_0_115:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.880601406097412 + min_value: 0.0 + scale: 0.023061182349920273 + zero_point: 0 + '@model/model.1/model.1.3/Conv_output_0_116:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.880601406097412 + min_value: 0.0 + scale: 0.023061182349920273 + zero_point: 0 + '@model/model.1/model.1.2/Relu_output_0_117:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 34.547969818115234 + min_value: 0.0 + scale: 0.13548223674297333 + zero_point: 0 + '@model/model.1/model.1.0/Conv_output_0_118:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 34.547969818115234 + min_value: 0.0 + scale: 0.13548223674297333 + zero_point: 0 + '@model/model.0/model.0.2/Relu_output_0_119:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.835350751876831 + min_value: 0.0 + scale: 0.007197454106062651 + zero_point: 0 + '@model/model.0/model.0.0/Conv_output_0_120:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.835350751876831 + min_value: 0.0 + scale: 0.007197454106062651 + zero_point: 0 + '@input.1_121:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.49609375 + min_value: -0.5 + scale: 0.00390625 + zero_point: 128 + '@attach_400/out0_3_netrans_mark_perm_122:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.068938136100769 + min_value: -0.6061154007911682 + scale: 0.006568837445229292 + zero_point: 92 + '@attach_397/out0_2_netrans_mark_perm_123:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0012764930725098 + min_value: -0.022866368293762207 + scale: 0.004016246646642685 + zero_point: 6 + '@attach_onnx//Concat_348/out0_1_netrans_mark_perm_124:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + scale: 0.005909114144742489 + zero_point: 81 + '@attach_onnx//Concat_345/out0_0_netrans_mark_perm_125:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + scale: 0.005909114144742489 + zero_point: 81 + '@model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm_126:out0': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.49609375 + min_value: -0.5 + scale: 0.00390625 + zero_point: 128 + '@400_4:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.48149600625038147 + min_value: -0.6190285682678223 + scale: 0.004315782338380814 + zero_point: 143 + '@400_4:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 67178.4530937176 + min_value: -67178.453125 + scale: 3.128240496153011e-05 + zero_point: 0 + '@397_5:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.4067230224609375 + min_value: -0.3356895446777344 + scale: 0.002911421936005354 + zero_point: 115 + '@397_5:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 51903.43747583058 + min_value: -51903.4375 + scale: 2.4169421521946787e-05 + zero_point: 0 + '@onnx//Concat_348_6:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.9801963567733765 + min_value: -1.1255085468292236 + scale: 0.008257665671408176 + zero_point: 136 + '@onnx//Concat_348_6:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 30105.066392231238 + min_value: -30105.06640625 + scale: 1.401876397721935e-05 + zero_point: 0 + '@onnx//Concat_345_7:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.9145514965057373 + min_value: -0.691943347454071 + scale: 0.006299979984760284 + zero_point: 110 + '@onnx//Concat_345_7:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 26208.281237795818 + min_value: -26208.28125 + scale: 1.2204181984998285e-05 + zero_point: 0 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.5439048409461975 + min_value: -0.5195902585983276 + scale: 0.004170569125562906 + zero_point: 125 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 186780.51553802355 + min_value: -186780.515625 + scale: 8.69764553499408e-05 + zero_point: 0 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.3730616569519043 + min_value: -0.40091922879219055 + scale: 0.0030352191533893347 + zero_point: 132 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 135933.43743670106 + min_value: -135933.4375 + scale: 6.329893949441611e-05 + zero_point: 0 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.510176420211792 + min_value: -0.7053191065788269 + scale: 0.004766649566590786 + zero_point: 148 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 53626.3085687783 + min_value: -53626.30859375 + scale: 2.4971695893327705e-05 + zero_point: 0 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.3717966079711914 + min_value: -0.5089793801307678 + scale: 0.0034540235064923763 + zero_point: 147 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 38858.85154440494 + min_value: -38858.8515625 + scale: 1.8095062841894105e-05 + zero_point: 0 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.2983776330947876 + min_value: -0.5611649751663208 + scale: 0.0033707553520798683 + zero_point: 166 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 21918.03123979362 + min_value: -21918.03125 + scale: 1.0206378647126257e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.863423764705658 + min_value: -1.2211904525756836 + scale: 0.0081749577075243 + zero_point: 149 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 136889.53118625586 + min_value: -136889.53125 + scale: 6.3744155340828e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.34477129578590393 + min_value: -0.6421072483062744 + scale: 0.0038701118901371956 + zero_point: 166 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 64078.86715766095 + min_value: -64078.8671875 + scale: 2.9839047783752903e-05 + zero_point: 0 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.321830153465271 + min_value: -0.537034809589386 + scale: 0.003368097823113203 + zero_point: 159 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 17740.890616738754 + min_value: -17740.890625 + scale: 8.261245966423302e-06 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.7087544798851013 + min_value: -1.297533631324768 + scale: 0.00786779634654522 + zero_point: 165 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 73283.24215337483 + min_value: -73283.2421875 + scale: 3.4125168895116075e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.6242436170578003 + min_value: -0.88043612241745 + scale: 0.005900704767554998 + zero_point: 149 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 96355.05464263118 + min_value: -96355.0546875 + scale: 4.4868818804388865e-05 + zero_point: 0 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.29398736357688904 + min_value: -0.5265582799911499 + scale: 0.0032178261317312717 + zero_point: 164 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 27890.906237012285 + min_value: -27890.90625 + scale: 1.298771530855447e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.40234020352363586 + min_value: -0.46787166595458984 + scale: 0.003412595484405756 + zero_point: 137 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 71544.78902918437 + min_value: -71544.7890625 + scale: 3.3315638575004413e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.7640163898468018 + min_value: -1.2688026428222656 + scale: 0.007971839047968388 + zero_point: 159 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 81545.72652452732 + min_value: -81545.7265625 + scale: 3.797268800553865e-05 + zero_point: 0 + '@cpm/conv/conv.0/Conv_output_0_38:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.3061813414096832 + min_value: -0.25823432207107544 + scale: 0.002213394967839122 + zero_point: 117 + '@cpm/conv/conv.0/Conv_output_0_38:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 12495.608392618778 + min_value: -12495.6083984375 + scale: 5.818721092509804e-06 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.5446580052375793 + min_value: -0.6891565322875977 + scale: 0.0048384880647063255 + zero_point: 142 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 75211.22652747705 + min_value: -75211.2265625 + scale: 3.502295658108778e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.48921284079551697 + min_value: -0.9916524291038513 + scale: 0.005807314533740282 + zero_point: 171 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 108686.32026188899 + min_value: -108686.3203125 + scale: 5.061101182946004e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.5472211837768555 + min_value: -0.5824781656265259 + scale: 0.004430193454027176 + zero_point: 131 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 63668.792939101906 + min_value: -63668.79296875 + scale: 2.9648092095158062e-05 + zero_point: 0 + '@cpm/align/align.0/Conv_output_0_49:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.3047269582748413 + min_value: -0.4415889382362366 + scale: 0.0029267289210110903 + zero_point: 151 + '@cpm/align/align.0/Conv_output_0_49:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 26146.64451907452 + min_value: -26146.64453125 + scale: 1.2175480151199735e-05 + zero_point: 0 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2147483647.0 + min_value: -2147483648.0 + scale: 1.0 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.46347150206565857 + min_value: -0.6541740298271179 + scale: 0.00438292371109128 + zero_point: 149 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 74848.15621514611 + min_value: -74848.15625 + scale: 3.485388879198581e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.5771108865737915 + min_value: -0.49492335319519043 + scale: 0.00420405576005578 + zero_point: 118 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 66330.14840661261 + min_value: -66330.1484375 + scale: 3.088738230871968e-05 + zero_point: 0 + '@model/model.11/model.11.3/Conv_output_0_58:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1146647930145264 + min_value: -0.9429869055747986 + scale: 0.008069222792983055 + zero_point: 117 + '@model/model.11/model.11.3/Conv_output_0_58:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 663565.0621910035 + min_value: -663565.0625 + scale: 0.00030899656121619046 + zero_point: 0 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.03312468156218529 + min_value: -0.03255603834986687 + scale: 0.00025757146067917347 + zero_point: 126 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 553130.4997424285 + min_value: -553130.5 + scale: 0.00025757146067917347 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.0274142026901245 + min_value: -0.6097877025604248 + scale: 0.0064203995279967785 + zero_point: 95 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 95570.7499554964 + min_value: -95570.75 + scale: 4.450359847396612e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.4058336615562439 + min_value: -0.3717547357082367 + scale: 0.0030493661761283875 + zero_point: 122 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 61136.707002781004 + min_value: -61136.70703125 + scale: 2.846899769792798e-05 + zero_point: 0 + '@model/model.11/model.11.0/Conv_output_0_67:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 8.021288871765137 + min_value: -6.133103370666504 + scale: 0.05550742149353027 + zero_point: 110 + '@model/model.11/model.11.0/Conv_output_0_67:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 490724.5622714886 + min_value: -490724.5625 + scale: 0.00022851143148727715 + zero_point: 0 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2147483647.0 + min_value: -2147483648.0 + scale: 1.0 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.094415545463562 + min_value: -1.0963801145553589 + scale: 0.008591355755925179 + zero_point: 128 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 109021.9530742327 + min_value: -109021.953125 + scale: 5.076730303699150e-05 + zero_point: 0 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1496243476867676 + min_value: -1.0743796825408936 + scale: 0.00872158445417881 + zero_point: 123 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 53038.78903780189 + min_value: -53038.7890625 + scale: 2.469811079208739e-05 + zero_point: 0 + '@model/model.10/model.10.3/Conv_output_0_75:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.49037596583366394 + min_value: -0.5690826177597046 + scale: 0.004154739435762167 + zero_point: 137 + '@model/model.10/model.10.3/Conv_output_0_75:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 84512.64839814573 + min_value: -84512.6484375 + scale: 3.935426866519264e-05 + zero_point: 0 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.03976425528526306 + min_value: -0.031856559216976166 + scale: 0.00028086593374609947 + zero_point: 113 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 603154.9997191341 + min_value: -603155.0 + scale: 0.00028086593374609947 + zero_point: 0 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0 + min_value: 0.0 + scale: 1.0 + zero_point: 0 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 762780.4996448027 + min_value: -762780.5 + scale: 0.000355197349563241 + zero_point: 0 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.0470142588019371 + min_value: -0.03391284495592117 + scale: 0.00031736117671243846 + zero_point: 107 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1791.646361470387 + min_value: -1791.6463623046875 + scale: 8.34300351471029e-07 + zero_point: 0 + '@model/model.10/model.10.0/Conv_output_0_82:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 6.671187877655029 + min_value: -4.681501865386963 + scale: 0.04452035203576088 + zero_point: 105 + '@model/model.10/model.10.0/Conv_output_0_82:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 423817.8123026445 + min_value: -423817.8125 + scale: 0.00019735554815270007 + zero_point: 0 + '@model/model.9/model.9.3/Conv_output_0_84:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.44361624121665955 + min_value: -0.5646821856498718 + scale: 0.003954111132770777 + zero_point: 143 + '@model/model.9/model.9.3/Conv_output_0_84:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 82890.18746140125 + min_value: -82890.1875 + scale: 3.859875141642987e-05 + zero_point: 0 + '@model/model.9/model.9.0/Conv_output_0_86:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.577437400817871 + min_value: -5.746422290802002 + scale: 0.04440729320049286 + zero_point: 129 + '@model/model.9/model.9.0/Conv_output_0_86:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 416057.31230625825 + min_value: -416057.3125 + scale: 0.00019374178373254836 + zero_point: 0 + '@model/model.8/model.8.3/Conv_output_0_88:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.5306500792503357 + min_value: -0.5236139297485352 + scale: 0.00413436908274889 + zero_point: 127 + '@model/model.8/model.8.3/Conv_output_0_88:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 530335.1247530435 + min_value: -530335.125 + scale: 0.0002469565370120108 + zero_point: 0 + '@model/model.8/model.8.0/Conv_output_0_90:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 6.483071327209473 + min_value: -7.160202980041504 + scale: 0.05350303649902344 + zero_point: 134 + '@model/model.8/model.8.0/Conv_output_0_90:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 450999.499789987 + min_value: -450999.5 + scale: 0.00021001300774514675 + zero_point: 0 + '@model/model.7/model.7.3/Conv_output_0_92:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.587432861328125 + min_value: -0.7095430493354797 + scale: 0.005086179822683334 + zero_point: 140 + '@model/model.7/model.7.3/Conv_output_0_92:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 224961.67177024405 + min_value: -224961.671875 + scale: 0.00010475594172021374 + zero_point: 0 + '@model/model.7/model.7.0/Conv_output_0_94:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.470149517059326 + min_value: -6.129836082458496 + scale: 0.04549013823270798 + zero_point: 135 + '@model/model.7/model.7.0/Conv_output_0_94:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 401316.90606312227 + min_value: -401316.90625 + scale: 0.0001868777471827343 + zero_point: 0 + '@model/model.6/model.6.3/Conv_output_0_96:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.5349634289741516 + min_value: -0.6652888655662537 + scale: 0.004706871695816517 + zero_point: 141 + '@model/model.6/model.6.3/Conv_output_0_96:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1055147.7495086584 + min_value: -1055147.75 + scale: 0.0004913414595648646 + zero_point: 0 + '@model/model.6/model.6.0/Conv_output_0_98:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 3.534395694732666 + min_value: -4.335607528686523 + scale: 0.030862757936120033 + zero_point: 140 + '@model/model.6/model.6.0/Conv_output_0_98:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 440614.218544823 + min_value: -440614.21875 + scale: 0.00020517698430921882 + zero_point: 0 + '@model/model.5/model.5.3/Conv_output_0_100:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.582099437713623 + min_value: -0.6120613217353821 + scale: 0.00468298327177763 + zero_point: 131 + '@model/model.5/model.5.3/Conv_output_0_100:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 116642.9218206839 + min_value: -116642.921875 + scale: 5.431609315564855e-05 + zero_point: 0 + '@model/model.5/model.5.0/Conv_output_0_102:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 9.850687980651855 + min_value: -9.850107192993164 + scale: 0.07725801318883896 + zero_point: 127 + '@model/model.5/model.5.0/Conv_output_0_102:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 926290.1870686625 + min_value: -926290.1875 + scale: 0.00043133748113177717 + zero_point: 0 + '@model/model.4/model.4.3/Conv_output_0_104:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.8545854091644287 + min_value: -0.880989670753479 + scale: 0.006806176621466875 + zero_point: 129 + '@model/model.4/model.4.3/Conv_output_0_104:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1590760.6242592444 + min_value: -1590760.625 + scale: 0.0007407556404359639 + zero_point: 0 + '@model/model.4/model.4.0/Conv_output_0_106:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.777940034866333 + min_value: -2.856468915939331 + scale: 0.02209572121500969 + zero_point: 129 + '@model/model.4/model.4.0/Conv_output_0_106:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 445887.9997923672 + min_value: -445888.0 + scale: 0.00020763278007507324 + zero_point: 0 + '@model/model.3/model.3.3/Conv_output_0_108:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1593574285507202 + min_value: -1.4911694526672363 + scale: 0.010394223965704441 + zero_point: 143 + '@model/model.3/model.3.3/Conv_output_0_108:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 295065.3436125995 + min_value: -295065.34375 + scale: 0.0001374005078105256 + zero_point: 0 + '@model/model.3/model.3.0/Conv_output_0_110:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 8.075460433959961 + min_value: -7.511605262756348 + scale: 0.061125747859478 + zero_point: 123 + '@model/model.3/model.3.0/Conv_output_0_110:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1029360.9995206664 + min_value: -1029361.0 + scale: 0.0004793335683643818 + zero_point: 0 + '@model/model.2/model.2.3/Conv_output_0_112:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 1.1482876539230347 + min_value: -1.6704888343811035 + scale: 0.0110540259629488 + zero_point: 151 + '@model/model.2/model.2.3/Conv_output_0_112:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 544944.4997462404 + min_value: -544944.5 + scale: 0.0002537595573812723 + zero_point: 0 + '@model/model.2/model.2.0/Conv_output_0_114:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 31.821802139282227 + min_value: -32.08101272583008 + scale: 0.25059929490089417 + zero_point: 128 + '@model/model.2/model.2.0/Conv_output_0_114:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 12410556.994220885 + min_value: -12410557.0 + scale: 0.005779115948826075 + zero_point: 0 + '@model/model.1/model.1.3/Conv_output_0_116:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 2.193585157394409 + min_value: -1.7845814228057861 + scale: 0.015600653365254402 + zero_point: 114 + '@model/model.1/model.1.3/Conv_output_0_116:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 4538945.997886389 + min_value: -4538946.0 + scale: 0.002113611437380314 + zero_point: 0 + '@model/model.1/model.1.0/Conv_output_0_118:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 5.203607082366943 + min_value: -3.283339023590088 + scale: 0.033282142132520676 + zero_point: 99 + '@model/model.1/model.1.0/Conv_output_0_118:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 514422.5935104533 + min_value: -514422.59375 + scale: 0.00023954668722581118 + zero_point: 0 + '@model/model.0/model.0.0/Conv_output_0_120:weight': + qtype: u8 + quantizer: asymmetric_affine + rounding: rtne + max_value: 0.9015139937400818 + min_value: -0.923520565032959 + scale: 0.007156998384743929 + zero_point: 129 + '@model/model.0/model.0.0/Conv_output_0_120:bias': + qtype: i32 + quantizer: asymmetric_affine + rounding: rtne + max_value: 60037.25387829298 + min_value: -60037.25390625 + scale: 2.7957024940405972e-05 + zero_point: 0 +customized_quantize_layers: {} diff --git a/resource/openpose/openpose_dynamic_fixed_point-16.quantize b/resource/openpose/openpose_dynamic_fixed_point-16.quantize new file mode 100644 index 0000000..83b4b10 --- /dev/null +++ b/resource/openpose/openpose_dynamic_fixed_point-16.quantize @@ -0,0 +1,1684 @@ +# !!!This file disallow TABs!!! +version: 2 +quantize_parameters: + '@attach_onnx//Concat_345/out0_0:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@attach_onnx//Concat_348/out0_1:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@attach_397/out0_2:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0012764930725098 + min_value: -0.022866368293762207 + fl: 14 + '@attach_400/out0_3:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.068938136100769 + min_value: -0.6061154007911682 + fl: 14 + '@400_4:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.068938136100769 + min_value: -0.6061154007911682 + fl: 14 + '@397_5:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0012764930725098 + min_value: -0.022866368293762207 + fl: 14 + '@onnx//Concat_348_6:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@onnx//Concat_345_7:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.8483352661132812 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.1169047355651855 + min_value: 0.0 + fl: 13 + '@initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.46164512634277344 + min_value: 0.0 + fl: 16 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5643914937973022 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.8483352661132812 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.1169047355651855 + min_value: 0.0 + fl: 13 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.46164512634277344 + min_value: 0.0 + fl: 16 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5643914937973022 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.317978382110596 + min_value: 0.0 + fl: 12 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.195100784301758 + min_value: 0.0 + fl: 12 + '@initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.3359031677246094 + min_value: 0.0 + fl: 14 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.3359031677246094 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1060172319412231 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.195100784301758 + min_value: 0.0 + fl: 12 + '@initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7721197009086609 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1060172319412231 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.8911335468292236 + min_value: 0.0 + fl: 13 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7721197009086609 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.3/Add_output_0_26:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.966081976890564 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.8911335468292236 + min_value: 0.0 + fl: 13 + '@initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7146145105361938 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2146551609039307 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.9338730573654175 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.9338730573654175 + min_value: 0.0 + fl: 14 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7146145105361938 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2146551609039307 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.9390138387680054 + min_value: 0.0 + fl: 14 + '@cpm/conv/conv.1/Relu_output_0_35:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@refinement_stages.0/trunk/trunk.2/Add_output_0_36:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.489450693130493 + min_value: 0.0 + fl: 13 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.9390138387680054 + min_value: 0.0 + fl: 14 + '@cpm/conv/conv.0/Conv_output_0_38:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.7065311670303345 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.9933781623840332 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.9933781623840332 + min_value: 0.0 + fl: 14 + '@cpm/Add_output_0_42:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.6703611016273499 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.7065311670303345 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.845794439315796 + min_value: 0.0 + fl: 14 + '@cpm/align/align.1/Relu_output_0_45:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.6703611016273499 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.1/Add_output_0_47:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.539104700088501 + min_value: 0.0 + fl: 13 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.845794439315796 + min_value: 0.0 + fl: 14 + '@cpm/align/align.0/Conv_output_0_49:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.6703611016273499 + min_value: 0.0 + fl: 15 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.8849024772644043 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.4087772369384766 + min_value: 0.0 + fl: 13 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.4087772369384766 + min_value: 0.0 + fl: 13 + '@model/model.11/model.11.5/Relu_output_0_54:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1312521696090698 + min_value: 0.0 + fl: 14 + '@cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.8849024772644043 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.027811050415039 + min_value: 0.0 + fl: 13 + '@model/model.11/model.11.3/Conv_output_0_58:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1312521696090698 + min_value: 0.0 + fl: 14 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.0/Add_output_0_60:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.873496174812317 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.027811050415039 + min_value: 0.0 + fl: 13 + '@model/model.11/model.11.2/Relu_output_0_62:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 9.764772415161133 + min_value: 0.0 + fl: 11 + '@cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.722118616104126 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.801229476928711 + min_value: 0.0 + fl: 14 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.801229476928711 + min_value: 0.0 + fl: 14 + '@model/model.11/model.11.0/Conv_output_0_67:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 9.764772415161133 + min_value: 0.0 + fl: 11 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.722118616104126 + min_value: 0.0 + fl: 15 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.3806896209716797 + min_value: 0.0 + fl: 13 + '@model/model.10/model.10.5/Relu_output_0_71:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2916451692581177 + min_value: 0.0 + fl: 14 + '@cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@Concat_output_0_73:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.3806896209716797 + min_value: 0.0 + fl: 13 + '@model/model.10/model.10.3/Conv_output_0_75:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2916451692581177 + min_value: 0.0 + fl: 14 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.033857811242341995 + min_value: -0.056717514991760254 + fl: 19 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.033857811242341995 + min_value: -0.05838948115706444 + fl: 19 + '@model/model.10/model.10.2/Relu_output_0_81:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.415395498275757 + min_value: 0.0 + fl: 13 + '@model/model.10/model.10.0/Conv_output_0_82:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.415395498275757 + min_value: 0.0 + fl: 13 + '@model/model.9/model.9.5/Relu_output_0_83:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2915217876434326 + min_value: 0.0 + fl: 14 + '@model/model.9/model.9.3/Conv_output_0_84:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2915217876434326 + min_value: 0.0 + fl: 14 + '@model/model.9/model.9.2/Relu_output_0_85:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.654484748840332 + min_value: 0.0 + fl: 13 + '@model/model.9/model.9.0/Conv_output_0_86:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.654484748840332 + min_value: 0.0 + fl: 13 + '@model/model.8/model.8.5/Relu_output_0_87:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2711007595062256 + min_value: 0.0 + fl: 14 + '@model/model.8/model.8.3/Conv_output_0_88:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.2711007595062256 + min_value: 0.0 + fl: 14 + '@model/model.8/model.8.2/Relu_output_0_89:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 15.23180866241455 + min_value: 0.0 + fl: 11 + '@model/model.8/model.8.0/Conv_output_0_90:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 15.23180866241455 + min_value: 0.0 + fl: 11 + '@model/model.7/model.7.5/Relu_output_0_91:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1436121463775635 + min_value: 0.0 + fl: 14 + '@model/model.7/model.7.3/Conv_output_0_92:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1436121463775635 + min_value: 0.0 + fl: 14 + '@model/model.7/model.7.2/Relu_output_0_93:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.2520294189453125 + min_value: 0.0 + fl: 12 + '@model/model.7/model.7.0/Conv_output_0_94:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.2520294189453125 + min_value: 0.0 + fl: 12 + '@model/model.6/model.6.5/Relu_output_0_95:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.196881890296936 + min_value: 0.0 + fl: 14 + '@model/model.6/model.6.3/Conv_output_0_96:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.196881890296936 + min_value: 0.0 + fl: 14 + '@model/model.6/model.6.2/Relu_output_0_97:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 26.618968963623047 + min_value: 0.0 + fl: 10 + '@model/model.6/model.6.0/Conv_output_0_98:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 26.618968963623047 + min_value: 0.0 + fl: 10 + '@model/model.5/model.5.5/Relu_output_0_99:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.6952513456344604 + min_value: 0.0 + fl: 14 + '@model/model.5/model.5.3/Conv_output_0_100:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.6952513456344604 + min_value: 0.0 + fl: 14 + '@model/model.5/model.5.2/Relu_output_0_101:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.9576454162597656 + min_value: 0.0 + fl: 13 + '@model/model.5/model.5.0/Conv_output_0_102:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.9576454162597656 + min_value: 0.0 + fl: 13 + '@model/model.4/model.4.5/Relu_output_0_103:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.6266145706176758 + min_value: 0.0 + fl: 14 + '@model/model.4/model.4.3/Conv_output_0_104:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.6266145706176758 + min_value: 0.0 + fl: 14 + '@model/model.4/model.4.2/Relu_output_0_105:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 27.753128051757812 + min_value: 0.0 + fl: 10 + '@model/model.4/model.4.0/Conv_output_0_106:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 27.753128051757812 + min_value: 0.0 + fl: 10 + '@model/model.3/model.3.5/Relu_output_0_107:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.9483160972595215 + min_value: 0.0 + fl: 13 + '@model/model.3/model.3.3/Conv_output_0_108:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.9483160972595215 + min_value: 0.0 + fl: 13 + '@model/model.3/model.3.2/Relu_output_0_109:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.492973327636719 + min_value: 0.0 + fl: 12 + '@model/model.3/model.3.0/Conv_output_0_110:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.492973327636719 + min_value: 0.0 + fl: 12 + '@model/model.2/model.2.5/Relu_output_0_111:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.6653316020965576 + min_value: 0.0 + fl: 13 + '@model/model.2/model.2.3/Conv_output_0_112:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.6653316020965576 + min_value: 0.0 + fl: 13 + '@model/model.2/model.2.2/Relu_output_0_113:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.853857040405273 + min_value: 0.0 + fl: 12 + '@model/model.2/model.2.0/Conv_output_0_114:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.853857040405273 + min_value: 0.0 + fl: 12 + '@model/model.1/model.1.5/Relu_output_0_115:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 7.2354888916015625 + min_value: 0.0 + fl: 12 + '@model/model.1/model.1.3/Conv_output_0_116:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 7.2354888916015625 + min_value: 0.0 + fl: 12 + '@model/model.1/model.1.2/Relu_output_0_117:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34.547969818115234 + min_value: 0.0 + fl: 9 + '@model/model.1/model.1.0/Conv_output_0_118:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34.547969818115234 + min_value: 0.0 + fl: 9 + '@model/model.0/model.0.2/Relu_output_0_119:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.835350751876831 + min_value: 0.0 + fl: 14 + '@model/model.0/model.0.0/Conv_output_0_120:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.835350751876831 + min_value: 0.0 + fl: 14 + '@input.1_121:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.49609375 + min_value: -0.5 + fl: 16 + '@attach_400/out0_3_netrans_mark_perm_122:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.068938136100769 + min_value: -0.6061154007911682 + fl: 14 + '@attach_397/out0_2_netrans_mark_perm_123:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0012764930725098 + min_value: -0.022866368293762207 + fl: 14 + '@attach_onnx//Concat_348/out0_1_netrans_mark_perm_124:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@attach_onnx//Concat_345/out0_0_netrans_mark_perm_125:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0292251110076904 + min_value: -0.47759902477264404 + fl: 14 + '@model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm_126:out0': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.49609375 + min_value: -0.5 + fl: 16 + '@400_4:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.48149600625038147 + min_value: -0.6190285682678223 + fl: 15 + '@400_4:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@397_5:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.4067230224609375 + min_value: -0.3356895446777344 + fl: 16 + '@397_5:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@onnx//Concat_348_6:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9801963567733765 + min_value: -1.1255085468292236 + fl: 14 + '@onnx//Concat_348_6:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8589934592.0 + min_value: -8589934592.0 + fl: 30 + '@onnx//Concat_345_7:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9145514965057373 + min_value: -0.691943347454071 + fl: 15 + '@onnx//Concat_345_7:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8589934592.0 + min_value: -8589934592.0 + fl: 30 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5439048409461975 + min_value: -0.5195902585983276 + fl: 15 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 68719476736.0 + min_value: -68719476736.0 + fl: 27 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3730616569519043 + min_value: -0.40091922879219055 + fl: 16 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.510176420211792 + min_value: -0.7053191065788269 + fl: 15 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3717966079711914 + min_value: -0.5089793801307678 + fl: 15 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.2983776330947876 + min_value: -0.5611649751663208 + fl: 15 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8589934592.0 + min_value: -8589934592.0 + fl: 30 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.863423764705658 + min_value: -1.2211904525756836 + fl: 14 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 68719476736.0 + min_value: -68719476736.0 + fl: 27 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.34477129578590393 + min_value: -0.6421072483062744 + fl: 15 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.321830153465271 + min_value: -0.537034809589386 + fl: 15 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8589934592.0 + min_value: -8589934592.0 + fl: 30 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7087544798851013 + min_value: -1.297533631324768 + fl: 14 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.6242436170578003 + min_value: -0.88043612241745 + fl: 15 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.29398736357688904 + min_value: -0.5265582799911499 + fl: 15 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.40234020352363586 + min_value: -0.46787166595458984 + fl: 16 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7640163898468018 + min_value: -1.2688026428222656 + fl: 14 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@cpm/conv/conv.0/Conv_output_0_38:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3061813414096832 + min_value: -0.25823432207107544 + fl: 16 + '@cpm/conv/conv.0/Conv_output_0_38:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4294967296.0 + min_value: -4294967296.0 + fl: 31 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5446580052375793 + min_value: -0.6891565322875977 + fl: 15 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.48921284079551697 + min_value: -0.9916524291038513 + fl: 15 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5472211837768555 + min_value: -0.5824781656265259 + fl: 15 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@cpm/align/align.0/Conv_output_0_49:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3047269582748413 + min_value: -0.4415889382362366 + fl: 16 + '@cpm/align/align.0/Conv_output_0_49:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8589934592.0 + min_value: -8589934592.0 + fl: 30 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.5540876853408316e+85 + min_value: -1.5541351137805832e+85 + fl: -268 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2147483648.0 + min_value: -2147483648.0 + fl: 32 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.46347150206565857 + min_value: -0.6541740298271179 + fl: 15 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5771108865737915 + min_value: -0.49492335319519043 + fl: 15 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@model/model.11/model.11.3/Conv_output_0_58:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1146647930145264 + min_value: -0.9429869055747986 + fl: 14 + '@model/model.11/model.11.3/Conv_output_0_58:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 274877906944.0 + min_value: -274877906944.0 + fl: 25 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.5540876853408316e+85 + min_value: -1.5541351137805832e+85 + fl: -268 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2147483648.0 + min_value: -2147483648.0 + fl: 32 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0274142026901245 + min_value: -0.6097877025604248 + fl: 14 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.4058336615562439 + min_value: -0.3717547357082367 + fl: 16 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@model/model.11/model.11.0/Conv_output_0_67:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8.021288871765137 + min_value: -6.133103370666504 + fl: 11 + '@model/model.11/model.11.0/Conv_output_0_67:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 274877906944.0 + min_value: -274877906944.0 + fl: 25 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.5540876853408316e+85 + min_value: -1.5541351137805832e+85 + fl: -268 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2147483648.0 + min_value: -2147483648.0 + fl: 32 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.094415545463562 + min_value: -1.0963801145553589 + fl: 14 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1496243476867676 + min_value: -1.0743796825408936 + fl: 14 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869184.0 + min_value: -17179869184.0 + fl: 29 + '@model/model.10/model.10.3/Conv_output_0_75:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.49037596583366394 + min_value: -0.5690826177597046 + fl: 15 + '@model/model.10/model.10.3/Conv_output_0_75:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.5540876853408316e+85 + min_value: -1.5541351137805832e+85 + fl: -268 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2147483648.0 + min_value: -2147483648.0 + fl: 32 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.9998779296875 + min_value: -4.0 + fl: 13 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2147483648.0 + min_value: -2147483648.0 + fl: 32 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.24999237060546875 + min_value: -0.25 + fl: 17 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2147483648.0 + min_value: -2147483648.0 + fl: 32 + '@model/model.10/model.10.0/Conv_output_0_82:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.671187877655029 + min_value: -4.681501865386963 + fl: 12 + '@model/model.10/model.10.0/Conv_output_0_82:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.9/model.9.3/Conv_output_0_84:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.44361624121665955 + min_value: -0.5646821856498718 + fl: 15 + '@model/model.9/model.9.3/Conv_output_0_84:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@model/model.9/model.9.0/Conv_output_0_86:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.577437400817871 + min_value: -5.746422290802002 + fl: 12 + '@model/model.9/model.9.0/Conv_output_0_86:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.8/model.8.3/Conv_output_0_88:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5306500792503357 + min_value: -0.5236139297485352 + fl: 15 + '@model/model.8/model.8.3/Conv_output_0_88:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.8/model.8.0/Conv_output_0_90:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.483071327209473 + min_value: -7.160202980041504 + fl: 12 + '@model/model.8/model.8.0/Conv_output_0_90:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.7/model.7.3/Conv_output_0_92:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.587432861328125 + min_value: -0.7095430493354797 + fl: 15 + '@model/model.7/model.7.3/Conv_output_0_92:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 68719476736.0 + min_value: -68719476736.0 + fl: 27 + '@model/model.7/model.7.0/Conv_output_0_94:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.470149517059326 + min_value: -6.129836082458496 + fl: 12 + '@model/model.7/model.7.0/Conv_output_0_94:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.6/model.6.3/Conv_output_0_96:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5349634289741516 + min_value: -0.6652888655662537 + fl: 15 + '@model/model.6/model.6.3/Conv_output_0_96:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 274877906944.0 + min_value: -274877906944.0 + fl: 25 + '@model/model.6/model.6.0/Conv_output_0_98:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.534395694732666 + min_value: -4.335607528686523 + fl: 12 + '@model/model.6/model.6.0/Conv_output_0_98:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.5/model.5.3/Conv_output_0_100:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.582099437713623 + min_value: -0.6120613217353821 + fl: 15 + '@model/model.5/model.5.3/Conv_output_0_100:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 34359738368.0 + min_value: -34359738368.0 + fl: 28 + '@model/model.5/model.5.0/Conv_output_0_102:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 9.850687980651855 + min_value: -9.850107192993164 + fl: 11 + '@model/model.5/model.5.0/Conv_output_0_102:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 274877906944.0 + min_value: -274877906944.0 + fl: 25 + '@model/model.4/model.4.3/Conv_output_0_104:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.8545854091644287 + min_value: -0.880989670753479 + fl: 15 + '@model/model.4/model.4.3/Conv_output_0_104:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 274877906944.0 + min_value: -274877906944.0 + fl: 25 + '@model/model.4/model.4.0/Conv_output_0_106:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.777940034866333 + min_value: -2.856468915939331 + fl: 13 + '@model/model.4/model.4.0/Conv_output_0_106:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.3/model.3.3/Conv_output_0_108:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1593574285507202 + min_value: -1.4911694526672363 + fl: 14 + '@model/model.3/model.3.3/Conv_output_0_108:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.3/model.3.0/Conv_output_0_110:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8.075460433959961 + min_value: -7.511605262756348 + fl: 11 + '@model/model.3/model.3.0/Conv_output_0_110:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 549755813888.0 + min_value: -549755813888.0 + fl: 24 + '@model/model.2/model.2.3/Conv_output_0_112:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1482876539230347 + min_value: -1.6704888343811035 + fl: 14 + '@model/model.2/model.2.3/Conv_output_0_112:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.2/model.2.0/Conv_output_0_114:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 31.821802139282227 + min_value: -32.08101272583008 + fl: 9 + '@model/model.2/model.2.0/Conv_output_0_114:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4398046511104.0 + min_value: -4398046511104.0 + fl: 21 + '@model/model.1/model.1.3/Conv_output_0_116:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.193585157394409 + min_value: -1.7845814228057861 + fl: 13 + '@model/model.1/model.1.3/Conv_output_0_116:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2199023255552.0 + min_value: -2199023255552.0 + fl: 22 + '@model/model.1/model.1.0/Conv_output_0_118:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.203607082366943 + min_value: -3.283339023590088 + fl: 12 + '@model/model.1/model.1.0/Conv_output_0_118:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 137438953472.0 + min_value: -137438953472.0 + fl: 26 + '@model/model.0/model.0.0/Conv_output_0_120:weight': + qtype: i16 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9015139937400818 + min_value: -0.923520565032959 + fl: 15 + '@model/model.0/model.0.0/Conv_output_0_120:bias': + qtype: i64 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4294967296.0 + min_value: -4294967296.0 + fl: 31 +customized_quantize_layers: {} diff --git a/resource/openpose/openpose_dynamic_fixed_point-8.quantize b/resource/openpose/openpose_dynamic_fixed_point-8.quantize new file mode 100644 index 0000000..789a6ac --- /dev/null +++ b/resource/openpose/openpose_dynamic_fixed_point-8.quantize @@ -0,0 +1,1684 @@ +# !!!This file disallow TABs!!! +version: 2 +quantize_parameters: + '@attach_onnx//Concat_345/out0_0:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@attach_onnx//Concat_348/out0_1:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@attach_397/out0_2:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.3168612718582153 + min_value: -0.28261154890060425 + fl: 6 + '@attach_400/out0_3:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9628908634185791 + min_value: -1.010817527770996 + fl: 6 + '@400_4:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9628908634185791 + min_value: -1.010817527770996 + fl: 6 + '@397_5:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.3168612718582153 + min_value: -0.28261154890060425 + fl: 6 + '@onnx//Concat_348_6:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@onnx//Concat_345_7:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.6133546829223633 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.494922161102295 + min_value: 0.0 + fl: 4 + '@initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.205663800239563 + min_value: 0.0 + fl: 6 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.7571942806243896 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.6133546829223633 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.494922161102295 + min_value: 0.0 + fl: 4 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.205663800239563 + min_value: 0.0 + fl: 6 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.7571942806243896 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.543039321899414 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.197952747344971 + min_value: 0.0 + fl: 4 + '@initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.5310773849487305 + min_value: 0.0 + fl: 4 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.5310773849487305 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.700070381164551 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.197952747344971 + min_value: 0.0 + fl: 4 + '@initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.1627414226531982 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.700070381164551 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.5712690353393555 + min_value: 0.0 + fl: 5 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.1627414226531982 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.3/Add_output_0_26:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.6256208419799805 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.5712690353393555 + min_value: 0.0 + fl: 5 + '@initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.9906160831451416 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.902176856994629 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.585700511932373 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.585700511932373 + min_value: 0.0 + fl: 5 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.9906160831451416 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.902176856994629 + min_value: 0.0 + fl: 5 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.540367603302002 + min_value: 0.0 + fl: 4 + '@cpm/conv/conv.1/Relu_output_0_35:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@refinement_stages.0/trunk/trunk.2/Add_output_0_36:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.705320358276367 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.540367603302002 + min_value: 0.0 + fl: 4 + '@cpm/conv/conv.0/Conv_output_0_38:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.1595587730407715 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.705320358276367 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.705320358276367 + min_value: 0.0 + fl: 4 + '@cpm/Add_output_0_42:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.79840087890625 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.1595587730407715 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.648444652557373 + min_value: 0.0 + fl: 4 + '@cpm/align/align.1/Relu_output_0_45:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.79840087890625 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.1/Add_output_0_47:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.917450428009033 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.648444652557373 + min_value: 0.0 + fl: 4 + '@cpm/align/align.0/Conv_output_0_49:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.79840087890625 + min_value: 0.0 + fl: 4 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.767617225646973 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.868276596069336 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.868276596069336 + min_value: 0.0 + fl: 4 + '@model/model.11/model.11.5/Relu_output_0_54:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 14.694854736328125 + min_value: 0.0 + fl: 3 + '@cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.767617225646973 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 7.235624313354492 + min_value: 0.0 + fl: 4 + '@model/model.11/model.11.3/Conv_output_0_58:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 14.694854736328125 + min_value: 0.0 + fl: 3 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.0/Add_output_0_60:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8.188901901245117 + min_value: 0.0 + fl: 3 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 7.235624313354492 + min_value: 0.0 + fl: 4 + '@model/model.11/model.11.2/Relu_output_0_62:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 25.411134719848633 + min_value: 0.0 + fl: 2 + '@cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.051402568817139 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 9.586334228515625 + min_value: 0.0 + fl: 3 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 9.586334228515625 + min_value: 0.0 + fl: 3 + '@model/model.11/model.11.0/Conv_output_0_67:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 25.411134719848633 + min_value: 0.0 + fl: 2 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.051402568817139 + min_value: 0.0 + fl: 4 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 11.0706148147583 + min_value: 0.0 + fl: 3 + '@model/model.10/model.10.5/Relu_output_0_71:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17.037574768066406 + min_value: 0.0 + fl: 2 + '@cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@Concat_output_0_73:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 11.0706148147583 + min_value: 0.0 + fl: 3 + '@model/model.10/model.10.3/Conv_output_0_75:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17.037574768066406 + min_value: 0.0 + fl: 2 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0 + min_value: 0 + fl: 300 + '@cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.32217830419540405 + min_value: -0.3667927384376526 + fl: 8 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.32217830419540405 + min_value: -0.4569574296474457 + fl: 8 + '@model/model.10/model.10.2/Relu_output_0_81:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 35.48356246948242 + min_value: 0.0 + fl: 1 + '@model/model.10/model.10.0/Conv_output_0_82:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 35.48356246948242 + min_value: 0.0 + fl: 1 + '@model/model.9/model.9.5/Relu_output_0_83:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 24.551328659057617 + min_value: 0.0 + fl: 2 + '@model/model.9/model.9.3/Conv_output_0_84:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 24.551328659057617 + min_value: 0.0 + fl: 2 + '@model/model.9/model.9.2/Relu_output_0_85:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 67.37924194335938 + min_value: 0.0 + fl: 0 + '@model/model.9/model.9.0/Conv_output_0_86:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 67.37924194335938 + min_value: 0.0 + fl: 0 + '@model/model.8/model.8.5/Relu_output_0_87:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 55.901634216308594 + min_value: 0.0 + fl: 1 + '@model/model.8/model.8.3/Conv_output_0_88:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 55.901634216308594 + min_value: 0.0 + fl: 1 + '@model/model.8/model.8.2/Relu_output_0_89:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 85.3541259765625 + min_value: 0.0 + fl: 0 + '@model/model.8/model.8.0/Conv_output_0_90:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 85.3541259765625 + min_value: 0.0 + fl: 0 + '@model/model.7/model.7.5/Relu_output_0_91:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 54.900611877441406 + min_value: 0.0 + fl: 1 + '@model/model.7/model.7.3/Conv_output_0_92:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 54.900611877441406 + min_value: 0.0 + fl: 1 + '@model/model.7/model.7.2/Relu_output_0_93:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 115.3285903930664 + min_value: 0.0 + fl: 0 + '@model/model.7/model.7.0/Conv_output_0_94:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 115.3285903930664 + min_value: 0.0 + fl: 0 + '@model/model.6/model.6.5/Relu_output_0_95:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 47.81805419921875 + min_value: 0.0 + fl: 1 + '@model/model.6/model.6.3/Conv_output_0_96:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 47.81805419921875 + min_value: 0.0 + fl: 1 + '@model/model.6/model.6.2/Relu_output_0_97:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 203.79571533203125 + min_value: 0.0 + fl: -1 + '@model/model.6/model.6.0/Conv_output_0_98:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 203.79571533203125 + min_value: 0.0 + fl: -1 + '@model/model.5/model.5.5/Relu_output_0_99:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 150.86676025390625 + min_value: 0.0 + fl: -1 + '@model/model.5/model.5.3/Conv_output_0_100:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 150.86676025390625 + min_value: 0.0 + fl: -1 + '@model/model.5/model.5.2/Relu_output_0_101:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 356.38739013671875 + min_value: 0.0 + fl: -2 + '@model/model.5/model.5.0/Conv_output_0_102:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 356.38739013671875 + min_value: 0.0 + fl: -2 + '@model/model.4/model.4.5/Relu_output_0_103:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 141.01614379882812 + min_value: 0.0 + fl: -1 + '@model/model.4/model.4.3/Conv_output_0_104:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 141.01614379882812 + min_value: 0.0 + fl: -1 + '@model/model.4/model.4.2/Relu_output_0_105:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 244.03793334960938 + min_value: 0.0 + fl: -1 + '@model/model.4/model.4.0/Conv_output_0_106:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 244.03793334960938 + min_value: 0.0 + fl: -1 + '@model/model.3/model.3.5/Relu_output_0_107:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 388.832275390625 + min_value: 0.0 + fl: -2 + '@model/model.3/model.3.3/Conv_output_0_108:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 388.832275390625 + min_value: 0.0 + fl: -2 + '@model/model.3/model.3.2/Relu_output_0_109:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 448.5357360839844 + min_value: 0.0 + fl: -2 + '@model/model.3/model.3.0/Conv_output_0_110:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 448.5357360839844 + min_value: 0.0 + fl: -2 + '@model/model.2/model.2.5/Relu_output_0_111:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 483.8823547363281 + min_value: 0.0 + fl: -2 + '@model/model.2/model.2.3/Conv_output_0_112:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 483.8823547363281 + min_value: 0.0 + fl: -2 + '@model/model.2/model.2.2/Relu_output_0_113:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 699.6900634765625 + min_value: 0.0 + fl: -3 + '@model/model.2/model.2.0/Conv_output_0_114:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 699.6900634765625 + min_value: 0.0 + fl: -3 + '@model/model.1/model.1.5/Relu_output_0_115:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1116.212646484375 + min_value: 0.0 + fl: -4 + '@model/model.1/model.1.3/Conv_output_0_116:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1116.212646484375 + min_value: 0.0 + fl: -4 + '@model/model.1/model.1.2/Relu_output_0_117:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 691.7684936523438 + min_value: 0.0 + fl: -3 + '@model/model.1/model.1.0/Conv_output_0_118:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 691.7684936523438 + min_value: 0.0 + fl: -3 + '@model/model.0/model.0.2/Relu_output_0_119:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 760.405517578125 + min_value: 0.0 + fl: -3 + '@model/model.0/model.0.0/Conv_output_0_120:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 760.405517578125 + min_value: 0.0 + fl: -3 + '@input.1_121:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 255.0 + min_value: 0.0 + fl: -1 + '@attach_400/out0_3_netrans_mark_perm_122:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9628908634185791 + min_value: -1.010817527770996 + fl: 6 + '@attach_397/out0_2_netrans_mark_perm_123:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.3168612718582153 + min_value: -0.28261154890060425 + fl: 6 + '@attach_onnx//Concat_348/out0_1_netrans_mark_perm_124:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@attach_onnx//Concat_345/out0_0_netrans_mark_perm_125:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 4.756587028503418 + min_value: -1.933040976524353 + fl: 4 + '@model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm_126:out0': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 255.0 + min_value: 0.0 + fl: -1 + '@400_4:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.48149600625038147 + min_value: -0.6190285682678223 + fl: 7 + '@400_4:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@397_5:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.4067230224609375 + min_value: -0.3356895446777344 + fl: 8 + '@397_5:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@onnx//Concat_348_6:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9801963567733765 + min_value: -1.1255085468292236 + fl: 6 + '@onnx//Concat_348_6:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@onnx//Concat_345_7:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9145514965057373 + min_value: -0.691943347454071 + fl: 7 + '@onnx//Concat_345_7:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5439048409461975 + min_value: -0.5195902585983276 + fl: 7 + '@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3730616569519043 + min_value: -0.40091922879219055 + fl: 8 + '@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.510176420211792 + min_value: -0.7053191065788269 + fl: 7 + '@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3717966079711914 + min_value: -0.5089793801307678 + fl: 7 + '@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.2983776330947876 + min_value: -0.5611649751663208 + fl: 7 + '@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.863423764705658 + min_value: -1.2211904525756836 + fl: 6 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.34477129578590393 + min_value: -0.6421072483062744 + fl: 7 + '@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.321830153465271 + min_value: -0.537034809589386 + fl: 7 + '@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7087544798851013 + min_value: -1.297533631324768 + fl: 6 + '@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.6242436170578003 + min_value: -0.88043612241745 + fl: 7 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.29398736357688904 + min_value: -0.5265582799911499 + fl: 7 + '@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.40234020352363586 + min_value: -0.46787166595458984 + fl: 8 + '@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.7640163898468018 + min_value: -1.2688026428222656 + fl: 6 + '@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@cpm/conv/conv.0/Conv_output_0_38:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3061813414096832 + min_value: -0.25823432207107544 + fl: 8 + '@cpm/conv/conv.0/Conv_output_0_38:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 524287.9997558594 + min_value: -524288.0 + fl: 12 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5446580052375793 + min_value: -0.6891565322875977 + fl: 7 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.48921284079551697 + min_value: -0.9916524291038513 + fl: 7 + '@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5472211837768555 + min_value: -0.5824781656265259 + fl: 7 + '@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@cpm/align/align.0/Conv_output_0_49:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.3047269582748413 + min_value: -0.4415889382362366 + fl: 8 + '@cpm/align/align.0/Conv_output_0_49:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.083986866408344e+85 + min_value: -3.1082702275611665e+85 + fl: -277 + '@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 255.9999998807907 + min_value: -256.0 + fl: 23 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.46347150206565857 + min_value: -0.6541740298271179 + fl: 7 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5771108865737915 + min_value: -0.49492335319519043 + fl: 7 + '@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2097151.9990234375 + min_value: -2097152.0 + fl: 10 + '@model/model.11/model.11.3/Conv_output_0_58:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1146647930145264 + min_value: -0.9429869055747986 + fl: 6 + '@model/model.11/model.11.3/Conv_output_0_58:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8388607.99609375 + min_value: -8388608.0 + fl: 8 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.083986866408344e+85 + min_value: -3.1082702275611665e+85 + fl: -277 + '@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 255.9999998807907 + min_value: -256.0 + fl: 23 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.0274142026901245 + min_value: -0.6097877025604248 + fl: 6 + '@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2097151.9990234375 + min_value: -2097152.0 + fl: 10 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.4058336615562439 + min_value: -0.3717547357082367 + fl: 8 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1048575.9995117188 + min_value: -1048576.0 + fl: 11 + '@model/model.11/model.11.0/Conv_output_0_67:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8.021288871765137 + min_value: -6.133103370666504 + fl: 3 + '@model/model.11/model.11.0/Conv_output_0_67:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 67108863.96875 + min_value: -67108864.0 + fl: 5 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.083986866408344e+85 + min_value: -3.1082702275611665e+85 + fl: -277 + '@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 255.9999998807907 + min_value: -256.0 + fl: 23 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.094415545463562 + min_value: -1.0963801145553589 + fl: 6 + '@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2097151.9990234375 + min_value: -2097152.0 + fl: 10 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1496243476867676 + min_value: -1.0743796825408936 + fl: 6 + '@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2097151.9990234375 + min_value: -2097152.0 + fl: 10 + '@model/model.10/model.10.3/Conv_output_0_75:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.49037596583366394 + min_value: -0.5690826177597046 + fl: 7 + '@model/model.10/model.10.3/Conv_output_0_75:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8388607.99609375 + min_value: -8388608.0 + fl: 8 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.083986866408344e+85 + min_value: -3.1082702275611665e+85 + fl: -277 + '@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 255.9999998807907 + min_value: -256.0 + fl: 23 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.003875732421875 + min_value: -0.00390625 + fl: 15 + '@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 255.9999998807907 + min_value: -256.0 + fl: 23 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.0470142588019371 + min_value: -0.03391284495592117 + fl: 11 + '@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 65535.99996948242 + min_value: -65536.0 + fl: 15 + '@model/model.10/model.10.0/Conv_output_0_82:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.671187877655029 + min_value: -4.681501865386963 + fl: 4 + '@model/model.10/model.10.0/Conv_output_0_82:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 33554431.984375 + min_value: -33554432.0 + fl: 6 + '@model/model.9/model.9.3/Conv_output_0_84:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.44361624121665955 + min_value: -0.5646821856498718 + fl: 7 + '@model/model.9/model.9.3/Conv_output_0_84:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 16777215.9921875 + min_value: -16777216.0 + fl: 7 + '@model/model.9/model.9.0/Conv_output_0_86:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.577437400817871 + min_value: -5.746422290802002 + fl: 4 + '@model/model.9/model.9.0/Conv_output_0_86:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 67108863.96875 + min_value: -67108864.0 + fl: 5 + '@model/model.8/model.8.3/Conv_output_0_88:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5306500792503357 + min_value: -0.5236139297485352 + fl: 7 + '@model/model.8/model.8.3/Conv_output_0_88:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 16777215.9921875 + min_value: -16777216.0 + fl: 7 + '@model/model.8/model.8.0/Conv_output_0_90:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 6.483071327209473 + min_value: -7.160202980041504 + fl: 4 + '@model/model.8/model.8.0/Conv_output_0_90:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 67108863.96875 + min_value: -67108864.0 + fl: 5 + '@model/model.7/model.7.3/Conv_output_0_92:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.587432861328125 + min_value: -0.7095430493354797 + fl: 7 + '@model/model.7/model.7.3/Conv_output_0_92:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 16777215.9921875 + min_value: -16777216.0 + fl: 7 + '@model/model.7/model.7.0/Conv_output_0_94:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.470149517059326 + min_value: -6.129836082458496 + fl: 4 + '@model/model.7/model.7.0/Conv_output_0_94:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 67108863.96875 + min_value: -67108864.0 + fl: 5 + '@model/model.6/model.6.3/Conv_output_0_96:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.5349634289741516 + min_value: -0.6652888655662537 + fl: 7 + '@model/model.6/model.6.3/Conv_output_0_96:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 33554431.984375 + min_value: -33554432.0 + fl: 6 + '@model/model.6/model.6.0/Conv_output_0_98:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 3.534395694732666 + min_value: -4.335607528686523 + fl: 4 + '@model/model.6/model.6.0/Conv_output_0_98:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 268435455.875 + min_value: -268435456.0 + fl: 3 + '@model/model.5/model.5.3/Conv_output_0_100:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.582099437713623 + min_value: -0.6120613217353821 + fl: 7 + '@model/model.5/model.5.3/Conv_output_0_100:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 67108863.96875 + min_value: -67108864.0 + fl: 5 + '@model/model.5/model.5.0/Conv_output_0_102:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 9.850687980651855 + min_value: -9.850107192993164 + fl: 3 + '@model/model.5/model.5.0/Conv_output_0_102:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 536870911.75 + min_value: -536870912.0 + fl: 2 + '@model/model.4/model.4.3/Conv_output_0_104:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.8545854091644287 + min_value: -0.880989670753479 + fl: 7 + '@model/model.4/model.4.3/Conv_output_0_104:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 33554431.984375 + min_value: -33554432.0 + fl: 6 + '@model/model.4/model.4.0/Conv_output_0_106:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.777940034866333 + min_value: -2.856468915939331 + fl: 5 + '@model/model.4/model.4.0/Conv_output_0_106:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 268435455.875 + min_value: -268435456.0 + fl: 3 + '@model/model.3/model.3.3/Conv_output_0_108:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1593574285507202 + min_value: -1.4911694526672363 + fl: 6 + '@model/model.3/model.3.3/Conv_output_0_108:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 134217727.9375 + min_value: -134217728.0 + fl: 4 + '@model/model.3/model.3.0/Conv_output_0_110:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 8.075460433959961 + min_value: -7.511605262756348 + fl: 3 + '@model/model.3/model.3.0/Conv_output_0_110:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1073741823.5 + min_value: -1073741824.0 + fl: 1 + '@model/model.2/model.2.3/Conv_output_0_112:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1.1482876539230347 + min_value: -1.6704888343811035 + fl: 6 + '@model/model.2/model.2.3/Conv_output_0_112:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 268435455.875 + min_value: -268435456.0 + fl: 3 + '@model/model.2/model.2.0/Conv_output_0_114:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 31.821802139282227 + min_value: -32.08101272583008 + fl: 1 + '@model/model.2/model.2.0/Conv_output_0_114:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 17179869176.0 + min_value: -17179869184.0 + fl: -3 + '@model/model.1/model.1.3/Conv_output_0_116:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 2.193585157394409 + min_value: -1.7845814228057861 + fl: 5 + '@model/model.1/model.1.3/Conv_output_0_116:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 536870911.75 + min_value: -536870912.0 + fl: 2 + '@model/model.1/model.1.0/Conv_output_0_118:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 5.203607082366943 + min_value: -3.283339023590088 + fl: 4 + '@model/model.1/model.1.0/Conv_output_0_118:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 1073741823.5 + min_value: -1073741824.0 + fl: 1 + '@model/model.0/model.0.0/Conv_output_0_120:weight': + qtype: i8 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 0.9015139937400818 + min_value: -0.923520565032959 + fl: 7 + '@model/model.0/model.0.0/Conv_output_0_120:bias': + qtype: i32 + quantizer: dynamic_fixed_point + rounding: rtne + max_value: 33554431.984375 + min_value: -33554432.0 + fl: 6 +customized_quantize_layers: {} diff --git a/resource/openpose/openpose_inputmeta.yml b/resource/openpose/openpose_inputmeta.yml new file mode 100644 index 0000000..21ba727 --- /dev/null +++ b/resource/openpose/openpose_inputmeta.yml @@ -0,0 +1,56 @@ +# !!!This file disallow TABs!!! +# "category" allowed values: "image, frequency, undefined" +# "database" allowed types: "TEXT, NPY, H5FS, SQLITE, LMDB, GENERATOR, ZIP" +# "tensor_name" only support in H5FS database +# "preproc_type" allowed types:"IMAGE_RGB, IMAGE_RGB888_PLANAR, IMAGE_RGB888_PLANAR_SEP, IMAGE_I420, +# IMAGE_NV12,IMAGE_NV21, IMAGE_YUV444, IMAGE_YUYV422, IMAGE_UYVY422, IMAGE_GRAY, IMAGE_BGRA, TENSOR" +input_meta: + databases: + - path: dataset.txt + type: TEXT + ports: + - lid: input.1_121 + category: image + dtype: float32 + sparse: false + tensor_name: + layout: nchw + shape: + - 1 + - 3 + - 256 + - 424 + fitting: scale + preprocess: + reverse_channel: false + mean: + - 128 + - 128 + - 128 + scale: + - 0.00390625 + - 0.00390625 + - 0.00390625 + preproc_node_params: + add_preproc_node: true + preproc_type: IMAGE_RGB + preproc_dtype_converter: + quantizer: dynamic_fixed_point + qtype: int16 + fl: 16 + preproc_image_size: + - 424 + - 256 + preproc_crop: + enable_preproc_crop: false + crop_rect: + - 0 + - 0 + - 424 + - 256 + preproc_perm: + - 0 + - 1 + - 2 + - 3 + redirect_to_output: false diff --git a/resource/openpose/openpose_postprocess_file.yml b/resource/openpose/openpose_postprocess_file.yml new file mode 100644 index 0000000..da28c8e --- /dev/null +++ b/resource/openpose/openpose_postprocess_file.yml @@ -0,0 +1,41 @@ +%YAML 1.2 +--- +# "acuity_postprocs" allowed types: "detection_validate, classification_validate, mute_built_in_actions, classification_classic, print_topn, dump_results" +postprocess: + app_postprocs: + - lid: attach_onnx//Concat_345/out0_0 + postproc_params: + add_postproc_node: true + perm: + - 0 + - 1 + - 2 + - 3 + force_float32: true + - lid: attach_onnx//Concat_348/out0_1 + postproc_params: + add_postproc_node: true + perm: + - 0 + - 1 + - 2 + - 3 + force_float32: true + - lid: attach_397/out0_2 + postproc_params: + add_postproc_node: true + perm: + - 0 + - 1 + - 2 + - 3 + force_float32: true + - lid: attach_400/out0_3 + postproc_params: + add_postproc_node: true + perm: + - 0 + - 1 + - 2 + - 3 + force_float32: true diff --git a/resource/openpose/openpose_precision_analysis.json b/resource/openpose/openpose_precision_analysis.json new file mode 100644 index 0000000..c0a982b --- /dev/null +++ b/resource/openpose/openpose_precision_analysis.json @@ -0,0 +1,3197 @@ +{ + "Layers" : { + "397_5" : { + "inputs" : [ + "@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9:out0" + ], + "name" : "397", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "40.00%", + "diff_0.01" : "9.22%", + "diff_0.025" : "5.52%", + "diff_0.05" : "3.64%", + "diff_0.1" : "2.32%", + "diff_0.25" : "1.03%", + "diff_0.5" : "0.38%", + "diff_1" : "0.00%", + "diff_max" : "1.023", + "diff_similarity" : "97.07%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 19 + } + }, + "400_4" : { + "inputs" : [ "@refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8:out0" ], + "name" : "400", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "14.67%", + "diff_0.01" : "5.17%", + "diff_0.025" : "2.57%", + "diff_0.05" : "1.39%", + "diff_0.1" : "0.83%", + "diff_0.25" : "0.39%", + "diff_0.5" : "0.16%", + "diff_1" : "0.00%", + "diff_max" : "1.069", + "diff_similarity" : "0.21%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 38 + } + }, + "Concat_output_0_73" : { + "inputs" : [ + "@cpm/conv/conv.1/Relu_output_0_35:out0", + "@onnx//Concat_345_7:out0", + "@onnx//Concat_348_6:out0" + ], + "name" : "Concat_output_0", + "op" : "concat", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "58.73%", + "diff_0.01" : "43.16%", + "diff_0.025" : "32.66%", + "diff_0.05" : "19.51%", + "diff_0.1" : "6.09%", + "diff_0.25" : "0.37%", + "diff_0.5" : "0.05%", + "diff_1" : "0.00%", + "diff_max" : "1.029", + "diff_similarity" : "85.86%", + "dim" : 3 + } + }, + "attach_397/out0_2" : { + "inputs" : [ "@attach_397/out0_2_netrans_mark_perm_123:out0" ], + "name" : "attach_397/out0", + "op" : "output", + "outputs" : [ "out0" ], + "parameters" : {} + }, + "attach_397/out0_2_netrans_mark_perm_123" : { + "inputs" : [ "@397_5:out0" ], + "name" : "attach_397/out0_2_netrans_mark_perm", + "op" : "permute", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "40.00%", + "diff_0.01" : "9.22%", + "diff_0.025" : "5.52%", + "diff_0.05" : "3.64%", + "diff_0.1" : "2.32%", + "diff_0.25" : "1.03%", + "diff_0.5" : "0.38%", + "diff_1" : "0.00%", + "diff_max" : "1.023", + "diff_similarity" : "97.07%", + "perm" : [ 0, 3, 1, 2 ] + } + }, + "attach_400/out0_3" : { + "inputs" : [ "@attach_400/out0_3_netrans_mark_perm_122:out0" ], + "name" : "attach_400/out0", + "op" : "output", + "outputs" : [ "out0" ], + "parameters" : {} + }, + "attach_400/out0_3_netrans_mark_perm_122" : { + "inputs" : [ "@400_4:out0" ], + "name" : "attach_400/out0_3_netrans_mark_perm", + "op" : "permute", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "14.67%", + "diff_0.01" : "5.17%", + "diff_0.025" : "2.57%", + "diff_0.05" : "1.39%", + "diff_0.1" : "0.83%", + "diff_0.25" : "0.39%", + "diff_0.5" : "0.16%", + "diff_1" : "0.00%", + "diff_max" : "1.069", + "diff_similarity" : "0.21%", + "perm" : [ 0, 3, 1, 2 ] + } + }, + "attach_onnx//Concat_345/out0_0" : { + "inputs" : [ "@attach_onnx//Concat_345/out0_0_netrans_mark_perm_125:out0" ], + "name" : "attach_onnx//Concat_345/out0", + "op" : "output", + "outputs" : [ "out0" ], + "parameters" : {} + }, + "attach_onnx//Concat_345/out0_0_netrans_mark_perm_125" : { + "inputs" : [ "@onnx//Concat_345_7:out0" ], + "name" : "attach_onnx//Concat_345/out0_0_netrans_mark_perm", + "op" : "permute", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "52.46%", + "diff_0.01" : "9.17%", + "diff_0.025" : "5.54%", + "diff_0.05" : "3.68%", + "diff_0.1" : "2.24%", + "diff_0.25" : "0.87%", + "diff_0.5" : "0.28%", + "diff_1" : "0.00%", + "diff_max" : "0.819", + "diff_similarity" : "97.75%", + "perm" : [ 0, 3, 1, 2 ] + } + }, + "attach_onnx//Concat_348/out0_1" : { + "inputs" : [ "@attach_onnx//Concat_348/out0_1_netrans_mark_perm_124:out0" ], + "name" : "attach_onnx//Concat_348/out0", + "op" : "output", + "outputs" : [ "out0" ], + "parameters" : {} + }, + "attach_onnx//Concat_348/out0_1_netrans_mark_perm_124" : { + "inputs" : [ "@onnx//Concat_348_6:out0" ], + "name" : "attach_onnx//Concat_348/out0_1_netrans_mark_perm", + "op" : "permute", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "27.18%", + "diff_0.01" : "3.39%", + "diff_0.025" : "1.80%", + "diff_0.05" : "1.09%", + "diff_0.1" : "0.63%", + "diff_0.25" : "0.28%", + "diff_0.5" : "0.10%", + "diff_1" : "0.00%", + "diff_max" : "1.029", + "diff_similarity" : "7.11%", + "perm" : [ 0, 3, 1, 2 ] + } + }, + "cpm/Add_output_0_42" : { + "inputs" : [ + "@cpm/align/align.1/Relu_output_0_45:out0", + "@cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46:out0" + ], + "name" : "cpm/Add_output_0", + "op" : "add", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "62.47%", + "diff_0.01" : "55.32%", + "diff_0.025" : "43.98%", + "diff_0.05" : "28.48%", + "diff_0.1" : "10.55%", + "diff_0.25" : "0.39%", + "diff_0.5" : "0.01%", + "diff_1" : "0.00%", + "diff_max" : "0.668", + "diff_similarity" : "64.51%" + } + }, + "cpm/align/align.0/Conv_output_0_49" : { + "inputs" : [ "@model/model.11/model.11.5/Relu_output_0_54:out0" ], + "name" : "cpm/align/align.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.07%", + "diff_0.01" : "90.65%", + "diff_0.025" : "77.16%", + "diff_0.05" : "57.00%", + "diff_0.1" : "28.58%", + "diff_0.25" : "2.86%", + "diff_0.5" : "0.06%", + "diff_1" : "0.00%", + "diff_max" : "1.148", + "diff_similarity" : "40.37%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/align/align.1/Relu_output_0_45" : { + "inputs" : [ "@cpm/align/align.0/Conv_output_0_49:out0" ], + "name" : "cpm/align/align.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "62.47%", + "diff_0.01" : "55.32%", + "diff_0.025" : "43.98%", + "diff_0.05" : "28.48%", + "diff_0.1" : "10.55%", + "diff_0.25" : "0.39%", + "diff_0.5" : "0.01%", + "diff_1" : "0.00%", + "diff_max" : "0.668", + "diff_similarity" : "64.51%" + } + }, + "cpm/conv/conv.0/Conv_output_0_38" : { + "inputs" : [ "@cpm/Add_output_0_42:out0" ], + "name" : "cpm/conv/conv.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "98.94%", + "diff_0.01" : "89.47%", + "diff_0.025" : "74.15%", + "diff_0.05" : "51.90%", + "diff_0.1" : "23.34%", + "diff_0.25" : "1.91%", + "diff_0.5" : "0.03%", + "diff_1" : "0.00%", + "diff_max" : "0.708", + "diff_similarity" : "50.76%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/conv/conv.1/Relu_output_0_35" : { + "inputs" : [ "@cpm/conv/conv.0/Conv_output_0_38:out0" ], + "name" : "cpm/conv/conv.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "69.02%", + "diff_0.01" : "60.00%", + "diff_0.025" : "45.81%", + "diff_0.05" : "27.31%", + "diff_0.1" : "8.28%", + "diff_0.25" : "0.33%", + "diff_0.5" : "0.01%", + "diff_1" : "0.00%", + "diff_max" : "0.695", + "diff_similarity" : "72.21%" + } + }, + "cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80" : { + "inputs" : [ "@cpm/align/align.1/Relu_output_0_45:out0" ], + "name" : "cpm/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "48.57%", + "diff_0.01" : "1.41%", + "diff_0.025" : "0.02%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.058", + "diff_similarity" : "61.68%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 128, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79" : { + "inputs" : [ "@cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:out0" ], + "name" : "cpm/trunk/trunk.0/trunk.0.1/Elu_output_0", + "op" : "elu", + "outputs" : [ "out0" ], + "parameters" : { + "alpha" : 1, + "diff_0.001" : "48.45%", + "diff_0.01" : "1.41%", + "diff_0.025" : "0.02%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.057", + "diff_similarity" : "61.67%" + } + }, + "cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78" : { + "inputs" : [ "@cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79:out0" ], + "name" : "cpm/trunk/trunk.0/trunk.0.2/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77" : { + "inputs" : [ "@cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:out0" ], + "name" : "cpm/trunk/trunk.0/trunk.0.3/Elu_output_0", + "op" : "elu", + "outputs" : [ "out0" ], + "parameters" : { + "alpha" : 1, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%" + } + }, + "cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76" : { + "inputs" : [ "@cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77:out0" ], + "name" : "cpm/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 128, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72" : { + "inputs" : [ "@cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:out0" ], + "name" : "cpm/trunk/trunk.1/trunk.1.1/Elu_output_0", + "op" : "elu", + "outputs" : [ "out0" ], + "parameters" : { + "alpha" : 1, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%" + } + }, + "cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68" : { + "inputs" : [ "@cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72:out0" ], + "name" : "cpm/trunk/trunk.1/trunk.1.2/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63" : { + "inputs" : [ "@cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:out0" ], + "name" : "cpm/trunk/trunk.1/trunk.1.3/Elu_output_0", + "op" : "elu", + "outputs" : [ "out0" ], + "parameters" : { + "alpha" : 1, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%" + } + }, + "cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59" : { + "inputs" : [ "@cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63:out0" ], + "name" : "cpm/trunk/trunk.2/trunk.2.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 128, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55" : { + "inputs" : [ "@cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:out0" ], + "name" : "cpm/trunk/trunk.2/trunk.2.1/Elu_output_0", + "op" : "elu", + "outputs" : [ "out0" ], + "parameters" : { + "alpha" : 1, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%" + } + }, + "cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50" : { + "inputs" : [ "@cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55:out0" ], + "name" : "cpm/trunk/trunk.2/trunk.2.2/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46" : { + "inputs" : [ "@cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:out0" ], + "name" : "cpm/trunk/trunk.2/trunk.2.3/Elu_output_0", + "op" : "elu", + "outputs" : [ "out0" ], + "parameters" : { + "alpha" : 1, + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "-nan%" + } + }, + "initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15" : { + "inputs" : [ "@initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0" ], + "name" : "initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "96.47%", + "diff_0.01" : "69.55%", + "diff_0.025" : "48.65%", + "diff_0.05" : "30.54%", + "diff_0.1" : "8.56%", + "diff_0.25" : "0.93%", + "diff_0.5" : "0.05%", + "diff_1" : "0.00%", + "diff_max" : "1.066", + "diff_similarity" : "67.78%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11" : { + "inputs" : [ + "@initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:out0" + ], + "name" : "initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "49.11%", + "diff_0.01" : "23.59%", + "diff_0.025" : "8.65%", + "diff_0.05" : "3.41%", + "diff_0.1" : "0.95%", + "diff_0.25" : "0.04%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.564", + "diff_similarity" : "94.45%" + } + }, + "initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14" : { + "inputs" : [ "@initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0" ], + "name" : "initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "96.66%", + "diff_0.01" : "72.19%", + "diff_0.025" : "57.04%", + "diff_0.05" : "44.24%", + "diff_0.1" : "21.93%", + "diff_0.25" : "2.42%", + "diff_0.5" : "0.21%", + "diff_1" : "0.00%", + "diff_max" : "1.227", + "diff_similarity" : "30.19%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10" : { + "inputs" : [ "@initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:out0" ], + "name" : "initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "42.46%", + "diff_0.01" : "18.95%", + "diff_0.025" : "6.61%", + "diff_0.05" : "2.44%", + "diff_0.1" : "0.63%", + "diff_0.25" : "0.02%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.462", + "diff_similarity" : "86.69%" + } + }, + "initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32" : { + "inputs" : [ "@cpm/conv/conv.1/Relu_output_0_35:out0" ], + "name" : "initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.12%", + "diff_0.01" : "91.06%", + "diff_0.025" : "76.75%", + "diff_0.05" : "57.60%", + "diff_0.1" : "31.69%", + "diff_0.25" : "4.77%", + "diff_0.5" : "0.19%", + "diff_1" : "0.00%", + "diff_max" : "0.850", + "diff_similarity" : "28.97%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28" : { + "inputs" : [ "@initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:out0" ], + "name" : "initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "42.95%", + "diff_0.01" : "35.64%", + "diff_0.025" : "25.75%", + "diff_0.05" : "14.16%", + "diff_0.1" : "4.23%", + "diff_0.25" : "0.34%", + "diff_0.5" : "0.01%", + "diff_1" : "0.00%", + "diff_max" : "0.715", + "diff_similarity" : "68.05%" + } + }, + "initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25" : { + "inputs" : [ "@initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28:out0" ], + "name" : "initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "98.92%", + "diff_0.01" : "89.41%", + "diff_0.025" : "69.90%", + "diff_0.05" : "48.29%", + "diff_0.1" : "24.15%", + "diff_0.25" : "2.22%", + "diff_0.5" : "0.07%", + "diff_1" : "0.00%", + "diff_max" : "0.780", + "diff_similarity" : "30.30%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22" : { + "inputs" : [ "@initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:out0" ], + "name" : "initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "39.40%", + "diff_0.01" : "30.84%", + "diff_0.025" : "19.64%", + "diff_0.05" : "9.26%", + "diff_0.1" : "2.87%", + "diff_0.25" : "0.32%", + "diff_0.5" : "0.01%", + "diff_1" : "0.00%", + "diff_max" : "0.772", + "diff_similarity" : "66.62%" + } + }, + "initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19" : { + "inputs" : [ "@initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22:out0" ], + "name" : "initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "98.56%", + "diff_0.01" : "85.14%", + "diff_0.025" : "61.67%", + "diff_0.05" : "33.14%", + "diff_0.1" : "14.28%", + "diff_0.25" : "2.46%", + "diff_0.5" : "0.27%", + "diff_1" : "0.00%", + "diff_max" : "1.221", + "diff_similarity" : "75.38%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18" : { + "inputs" : [ "@initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:out0" ], + "name" : "initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "58.39%", + "diff_0.01" : "46.41%", + "diff_0.025" : "29.54%", + "diff_0.05" : "14.34%", + "diff_0.1" : "5.81%", + "diff_0.25" : "1.34%", + "diff_0.5" : "0.16%", + "diff_1" : "0.00%", + "diff_max" : "1.221", + "diff_similarity" : "85.79%" + } + }, + "input.1_121" : { + "inputs" : [], + "name" : "input.1", + "op" : "input", + "outputs" : [ "out0" ], + "parameters" : { + "channels" : 1, + "is_scalar" : false, + "shape" : [ 1, 3, 256, 424 ], + "size" : "", + "type" : "float32" + } + }, + "model/model.0/model.0.0/Conv_output_0_120" : { + "inputs" : [ + "@model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm_126:out0" + ], + "name" : "model/model.0/model.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "36.46%", + "diff_0.01" : "5.74%", + "diff_0.025" : "5.59%", + "diff_0.05" : "5.54%", + "diff_0.1" : "5.44%", + "diff_0.25" : "5.14%", + "diff_0.5" : "0.56%", + "diff_1" : "0.01%", + "diff_max" : "1.203", + "diff_similarity" : "97.00%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 2, + "stride_w" : 2, + "weights" : 32 + } + }, + "model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm_126" : { + "inputs" : [ "@input.1_121:out0" ], + "name" : "model/model.0/model.0.0/Conv_output_0_120_netrans_mark_perm", + "op" : "permute", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "0.00%", + "diff_0.01" : "0.00%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.000", + "diff_similarity" : "100.00%", + "perm" : [ 0, 2, 3, 1 ] + } + }, + "model/model.0/model.0.2/Relu_output_0_119" : { + "inputs" : [ "@model/model.0/model.0.0/Conv_output_0_120:out0" ], + "name" : "model/model.0/model.0.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "27.69%", + "diff_0.01" : "0.11%", + "diff_0.025" : "0.00%", + "diff_0.05" : "0.00%", + "diff_0.1" : "0.00%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.016", + "diff_similarity" : "100.00%" + } + }, + "model/model.1/model.1.0/Conv_output_0_118" : { + "inputs" : [ "@model/model.0/model.0.2/Relu_output_0_119:out0" ], + "name" : "model/model.1/model.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "52.54%", + "diff_0.01" : "44.01%", + "diff_0.025" : "32.45%", + "diff_0.05" : "17.55%", + "diff_0.1" : "4.18%", + "diff_0.25" : "3.56%", + "diff_0.5" : "0.34%", + "diff_1" : "0.17%", + "diff_max" : "4.821", + "diff_similarity" : "99.99%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 32, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 32 + } + }, + "model/model.1/model.1.2/Relu_output_0_117" : { + "inputs" : [ "@model/model.1/model.1.0/Conv_output_0_118:out0" ], + "name" : "model/model.1/model.1.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "42.80%", + "diff_0.01" : "34.31%", + "diff_0.025" : "22.80%", + "diff_0.05" : "7.99%", + "diff_0.1" : "0.11%", + "diff_0.25" : "0.00%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.123", + "diff_similarity" : "100.00%" + } + }, + "model/model.1/model.1.3/Conv_output_0_116" : { + "inputs" : [ "@model/model.1/model.1.2/Relu_output_0_117:out0" ], + "name" : "model/model.1/model.1.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "82.54%", + "diff_0.01" : "79.98%", + "diff_0.025" : "75.45%", + "diff_0.05" : "68.74%", + "diff_0.1" : "59.16%", + "diff_0.25" : "42.98%", + "diff_0.5" : "20.17%", + "diff_1" : "0.69%", + "diff_max" : "5.555", + "diff_similarity" : "55.17%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 64 + } + }, + "model/model.1/model.1.5/Relu_output_0_115" : { + "inputs" : [ "@model/model.1/model.1.3/Conv_output_0_116:out0" ], + "name" : "model/model.1/model.1.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "73.46%", + "diff_0.01" : "71.56%", + "diff_0.025" : "67.43%", + "diff_0.05" : "61.39%", + "diff_0.1" : "52.82%", + "diff_0.25" : "37.22%", + "diff_0.5" : "16.68%", + "diff_1" : "0.04%", + "diff_max" : "1.355", + "diff_similarity" : "68.68%" + } + }, + "model/model.10/model.10.0/Conv_output_0_82" : { + "inputs" : [ "@model/model.9/model.9.5/Relu_output_0_83:out0" ], + "name" : "model/model.10/model.10.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "75.46%", + "diff_0.01" : "67.42%", + "diff_0.025" : "61.37%", + "diff_0.05" : "52.59%", + "diff_0.1" : "37.61%", + "diff_0.25" : "13.10%", + "diff_0.5" : "2.19%", + "diff_1" : "0.17%", + "diff_max" : "1.794", + "diff_similarity" : "88.15%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 512, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.10/model.10.2/Relu_output_0_81" : { + "inputs" : [ "@model/model.10/model.10.0/Conv_output_0_82:out0" ], + "name" : "model/model.10/model.10.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "54.03%", + "diff_0.01" : "47.79%", + "diff_0.025" : "43.05%", + "diff_0.05" : "36.23%", + "diff_0.1" : "25.73%", + "diff_0.25" : "8.38%", + "diff_0.5" : "1.19%", + "diff_1" : "0.03%", + "diff_max" : "1.747", + "diff_similarity" : "92.62%" + } + }, + "model/model.10/model.10.3/Conv_output_0_75" : { + "inputs" : [ "@model/model.10/model.10.2/Relu_output_0_81:out0" ], + "name" : "model/model.10/model.10.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "80.79%", + "diff_0.01" : "75.48%", + "diff_0.025" : "67.66%", + "diff_0.05" : "56.27%", + "diff_0.1" : "37.17%", + "diff_0.25" : "8.37%", + "diff_0.5" : "0.52%", + "diff_1" : "0.00%", + "diff_max" : "1.292", + "diff_similarity" : "50.49%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.10/model.10.5/Relu_output_0_71" : { + "inputs" : [ "@model/model.10/model.10.3/Conv_output_0_75:out0" ], + "name" : "model/model.10/model.10.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "47.01%", + "diff_0.01" : "43.48%", + "diff_0.025" : "37.69%", + "diff_0.05" : "29.07%", + "diff_0.1" : "16.26%", + "diff_0.25" : "2.18%", + "diff_0.5" : "0.11%", + "diff_1" : "0.00%", + "diff_max" : "1.292", + "diff_similarity" : "73.36%" + } + }, + "model/model.11/model.11.0/Conv_output_0_67" : { + "inputs" : [ "@model/model.10/model.10.5/Relu_output_0_71:out0" ], + "name" : "model/model.11/model.11.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "78.05%", + "diff_0.01" : "67.84%", + "diff_0.025" : "54.27%", + "diff_0.05" : "44.33%", + "diff_0.1" : "30.03%", + "diff_0.25" : "8.72%", + "diff_0.5" : "1.21%", + "diff_1" : "0.04%", + "diff_max" : "2.517", + "diff_similarity" : "98.09%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 512, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.11/model.11.2/Relu_output_0_62" : { + "inputs" : [ "@model/model.11/model.11.0/Conv_output_0_67:out0" ], + "name" : "model/model.11/model.11.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "62.33%", + "diff_0.01" : "53.36%", + "diff_0.025" : "41.74%", + "diff_0.05" : "33.41%", + "diff_0.1" : "22.56%", + "diff_0.25" : "6.16%", + "diff_0.5" : "0.60%", + "diff_1" : "0.01%", + "diff_max" : "2.517", + "diff_similarity" : "98.74%" + } + }, + "model/model.11/model.11.3/Conv_output_0_58" : { + "inputs" : [ "@model/model.11/model.11.2/Relu_output_0_62:out0" ], + "name" : "model/model.11/model.11.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "85.98%", + "diff_0.01" : "80.35%", + "diff_0.025" : "74.52%", + "diff_0.05" : "65.62%", + "diff_0.1" : "49.83%", + "diff_0.25" : "17.75%", + "diff_0.5" : "2.39%", + "diff_1" : "0.04%", + "diff_max" : "1.618", + "diff_similarity" : "17.04%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.11/model.11.5/Relu_output_0_54" : { + "inputs" : [ "@model/model.11/model.11.3/Conv_output_0_58:out0" ], + "name" : "model/model.11/model.11.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "45.12%", + "diff_0.01" : "42.35%", + "diff_0.025" : "37.70%", + "diff_0.05" : "30.89%", + "diff_0.1" : "19.85%", + "diff_0.25" : "4.04%", + "diff_0.5" : "0.17%", + "diff_1" : "0.00%", + "diff_max" : "1.131", + "diff_similarity" : "52.99%" + } + }, + "model/model.2/model.2.0/Conv_output_0_114" : { + "inputs" : [ "@model/model.1/model.1.5/Relu_output_0_115:out0" ], + "name" : "model/model.2/model.2.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "82.41%", + "diff_0.01" : "78.30%", + "diff_0.025" : "70.60%", + "diff_0.05" : "61.57%", + "diff_0.1" : "47.43%", + "diff_0.25" : "30.55%", + "diff_0.5" : "20.00%", + "diff_1" : "4.61%", + "diff_max" : "14.850", + "diff_similarity" : "89.06%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 64, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 2, + "stride_w" : 2, + "weights" : 64 + } + }, + "model/model.2/model.2.2/Relu_output_0_113" : { + "inputs" : [ "@model/model.2/model.2.0/Conv_output_0_114:out0" ], + "name" : "model/model.2/model.2.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "76.46%", + "diff_0.01" : "72.95%", + "diff_0.025" : "65.11%", + "diff_0.05" : "57.03%", + "diff_0.1" : "43.87%", + "diff_0.25" : "28.20%", + "diff_0.5" : "18.67%", + "diff_1" : "3.95%", + "diff_max" : "2.636", + "diff_similarity" : "90.92%" + } + }, + "model/model.2/model.2.3/Conv_output_0_112" : { + "inputs" : [ "@model/model.2/model.2.2/Relu_output_0_113:out0" ], + "name" : "model/model.2/model.2.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "90.30%", + "diff_0.01" : "87.36%", + "diff_0.025" : "81.61%", + "diff_0.05" : "73.39%", + "diff_0.1" : "55.80%", + "diff_0.25" : "19.92%", + "diff_0.5" : "5.61%", + "diff_1" : "0.32%", + "diff_max" : "8.367", + "diff_similarity" : "63.37%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "model/model.2/model.2.5/Relu_output_0_111" : { + "inputs" : [ "@model/model.2/model.2.3/Conv_output_0_112:out0" ], + "name" : "model/model.2/model.2.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "80.06%", + "diff_0.01" : "77.66%", + "diff_0.025" : "72.75%", + "diff_0.05" : "65.74%", + "diff_0.1" : "49.80%", + "diff_0.25" : "16.08%", + "diff_0.5" : "3.89%", + "diff_1" : "0.02%", + "diff_max" : "1.747", + "diff_similarity" : "73.47%" + } + }, + "model/model.3/model.3.0/Conv_output_0_110" : { + "inputs" : [ "@model/model.2/model.2.5/Relu_output_0_111:out0" ], + "name" : "model/model.3/model.3.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "90.04%", + "diff_0.01" : "83.34%", + "diff_0.025" : "76.09%", + "diff_0.05" : "65.40%", + "diff_0.1" : "49.29%", + "diff_0.25" : "23.02%", + "diff_0.5" : "7.35%", + "diff_1" : "1.04%", + "diff_max" : "3.018", + "diff_similarity" : "84.52%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 128, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "model/model.3/model.3.2/Relu_output_0_109" : { + "inputs" : [ "@model/model.3/model.3.0/Conv_output_0_110:out0" ], + "name" : "model/model.3/model.3.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "74.46%", + "diff_0.01" : "68.53%", + "diff_0.025" : "62.16%", + "diff_0.05" : "53.08%", + "diff_0.1" : "39.08%", + "diff_0.25" : "17.63%", + "diff_0.5" : "5.13%", + "diff_1" : "0.71%", + "diff_max" : "2.915", + "diff_similarity" : "89.06%" + } + }, + "model/model.3/model.3.3/Conv_output_0_108" : { + "inputs" : [ "@model/model.3/model.3.2/Relu_output_0_109:out0" ], + "name" : "model/model.3/model.3.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "89.46%", + "diff_0.01" : "85.99%", + "diff_0.025" : "80.10%", + "diff_0.05" : "70.79%", + "diff_0.1" : "53.09%", + "diff_0.25" : "20.13%", + "diff_0.5" : "7.32%", + "diff_1" : "0.82%", + "diff_max" : "2.746", + "diff_similarity" : "46.55%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "model/model.3/model.3.5/Relu_output_0_107" : { + "inputs" : [ "@model/model.3/model.3.3/Conv_output_0_108:out0" ], + "name" : "model/model.3/model.3.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "68.19%", + "diff_0.01" : "65.22%", + "diff_0.025" : "60.26%", + "diff_0.05" : "52.12%", + "diff_0.1" : "36.60%", + "diff_0.25" : "10.31%", + "diff_0.5" : "2.71%", + "diff_1" : "0.11%", + "diff_max" : "2.020", + "diff_similarity" : "68.17%" + } + }, + "model/model.4/model.4.0/Conv_output_0_106" : { + "inputs" : [ "@model/model.3/model.3.5/Relu_output_0_107:out0" ], + "name" : "model/model.4/model.4.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "89.44%", + "diff_0.01" : "85.83%", + "diff_0.025" : "80.16%", + "diff_0.05" : "71.67%", + "diff_0.1" : "59.50%", + "diff_0.25" : "29.78%", + "diff_0.5" : "8.29%", + "diff_1" : "2.00%", + "diff_max" : "2.310", + "diff_similarity" : "99.70%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 128, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 2, + "stride_w" : 2, + "weights" : 128 + } + }, + "model/model.4/model.4.2/Relu_output_0_105" : { + "inputs" : [ "@model/model.4/model.4.0/Conv_output_0_106:out0" ], + "name" : "model/model.4/model.4.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "85.04%", + "diff_0.01" : "81.66%", + "diff_0.025" : "76.40%", + "diff_0.05" : "68.43%", + "diff_0.1" : "56.84%", + "diff_0.25" : "27.52%", + "diff_0.5" : "6.83%", + "diff_1" : "1.76%", + "diff_max" : "2.023", + "diff_similarity" : "99.73%" + } + }, + "model/model.4/model.4.3/Conv_output_0_104" : { + "inputs" : [ "@model/model.4/model.4.2/Relu_output_0_105:out0" ], + "name" : "model/model.4/model.4.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "87.07%", + "diff_0.01" : "83.36%", + "diff_0.025" : "77.02%", + "diff_0.05" : "66.58%", + "diff_0.1" : "47.81%", + "diff_0.25" : "13.93%", + "diff_0.5" : "3.17%", + "diff_1" : "0.02%", + "diff_max" : "1.364", + "diff_similarity" : "73.90%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 256 + } + }, + "model/model.4/model.4.5/Relu_output_0_103" : { + "inputs" : [ "@model/model.4/model.4.3/Conv_output_0_104:out0" ], + "name" : "model/model.4/model.4.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "83.56%", + "diff_0.01" : "79.91%", + "diff_0.025" : "73.60%", + "diff_0.05" : "63.36%", + "diff_0.1" : "44.70%", + "diff_0.25" : "11.79%", + "diff_0.5" : "2.36%", + "diff_1" : "0.01%", + "diff_max" : "1.235", + "diff_similarity" : "78.35%" + } + }, + "model/model.5/model.5.0/Conv_output_0_102" : { + "inputs" : [ "@model/model.4/model.4.5/Relu_output_0_103:out0" ], + "name" : "model/model.5/model.5.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "87.04%", + "diff_0.01" : "83.24%", + "diff_0.025" : "77.58%", + "diff_0.05" : "68.92%", + "diff_0.1" : "53.74%", + "diff_0.25" : "25.03%", + "diff_0.5" : "7.71%", + "diff_1" : "1.44%", + "diff_max" : "2.477", + "diff_similarity" : "69.06%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 256, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 256 + } + }, + "model/model.5/model.5.2/Relu_output_0_101" : { + "inputs" : [ "@model/model.5/model.5.0/Conv_output_0_102:out0" ], + "name" : "model/model.5/model.5.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "69.48%", + "diff_0.01" : "66.54%", + "diff_0.025" : "61.17%", + "diff_0.05" : "53.73%", + "diff_0.1" : "41.38%", + "diff_0.25" : "18.48%", + "diff_0.5" : "5.24%", + "diff_1" : "1.07%", + "diff_max" : "2.411", + "diff_similarity" : "77.12%" + } + }, + "model/model.5/model.5.3/Conv_output_0_100" : { + "inputs" : [ "@model/model.5/model.5.2/Relu_output_0_101:out0" ], + "name" : "model/model.5/model.5.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "79.32%", + "diff_0.01" : "73.60%", + "diff_0.025" : "66.78%", + "diff_0.05" : "57.53%", + "diff_0.1" : "41.10%", + "diff_0.25" : "12.35%", + "diff_0.5" : "2.34%", + "diff_1" : "0.10%", + "diff_max" : "2.005", + "diff_similarity" : "56.36%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 256 + } + }, + "model/model.5/model.5.5/Relu_output_0_99" : { + "inputs" : [ "@model/model.5/model.5.3/Conv_output_0_100:out0" ], + "name" : "model/model.5/model.5.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "59.23%", + "diff_0.01" : "56.28%", + "diff_0.025" : "51.29%", + "diff_0.05" : "43.06%", + "diff_0.1" : "28.82%", + "diff_0.25" : "6.73%", + "diff_0.5" : "0.93%", + "diff_1" : "0.01%", + "diff_max" : "1.216", + "diff_similarity" : "72.03%" + } + }, + "model/model.6/model.6.0/Conv_output_0_98" : { + "inputs" : [ "@model/model.5/model.5.5/Relu_output_0_99:out0" ], + "name" : "model/model.6/model.6.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "76.21%", + "diff_0.01" : "71.43%", + "diff_0.025" : "65.73%", + "diff_0.05" : "57.06%", + "diff_0.1" : "44.29%", + "diff_0.25" : "22.29%", + "diff_0.5" : "7.29%", + "diff_1" : "0.80%", + "diff_max" : "2.286", + "diff_similarity" : "99.50%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 256, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 256 + } + }, + "model/model.6/model.6.2/Relu_output_0_97" : { + "inputs" : [ "@model/model.6/model.6.0/Conv_output_0_98:out0" ], + "name" : "model/model.6/model.6.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "65.94%", + "diff_0.01" : "62.68%", + "diff_0.025" : "58.03%", + "diff_0.05" : "50.71%", + "diff_0.1" : "40.65%", + "diff_0.25" : "20.42%", + "diff_0.5" : "6.29%", + "diff_1" : "0.58%", + "diff_max" : "1.845", + "diff_similarity" : "99.58%" + } + }, + "model/model.6/model.6.3/Conv_output_0_96" : { + "inputs" : [ "@model/model.6/model.6.2/Relu_output_0_97:out0" ], + "name" : "model/model.6/model.6.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "86.26%", + "diff_0.01" : "81.32%", + "diff_0.025" : "74.20%", + "diff_0.05" : "62.69%", + "diff_0.1" : "41.96%", + "diff_0.25" : "10.14%", + "diff_0.5" : "1.43%", + "diff_1" : "0.00%", + "diff_max" : "1.161", + "diff_similarity" : "69.46%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.6/model.6.5/Relu_output_0_95" : { + "inputs" : [ "@model/model.6/model.6.3/Conv_output_0_96:out0" ], + "name" : "model/model.6/model.6.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "75.74%", + "diff_0.01" : "71.93%", + "diff_0.025" : "65.51%", + "diff_0.05" : "55.04%", + "diff_0.1" : "36.44%", + "diff_0.25" : "8.18%", + "diff_0.5" : "1.14%", + "diff_1" : "0.00%", + "diff_max" : "0.901", + "diff_similarity" : "74.61%" + } + }, + "model/model.7/model.7.0/Conv_output_0_94" : { + "inputs" : [ "@model/model.6/model.6.5/Relu_output_0_95:out0" ], + "name" : "model/model.7/model.7.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "85.66%", + "diff_0.01" : "78.93%", + "diff_0.025" : "71.47%", + "diff_0.05" : "60.70%", + "diff_0.1" : "44.34%", + "diff_0.25" : "18.24%", + "diff_0.5" : "4.92%", + "diff_1" : "0.64%", + "diff_max" : "2.549", + "diff_similarity" : "78.24%", + "dilation" : [ 1, 2, 2, 1 ], + "group_number" : 512, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 2, 2, 2, 2 ], + "pad_h" : 2, + "pad_method" : "padding_const", + "pad_w" : 2, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.7/model.7.2/Relu_output_0_93" : { + "inputs" : [ "@model/model.7/model.7.0/Conv_output_0_94:out0" ], + "name" : "model/model.7/model.7.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "64.34%", + "diff_0.01" : "59.41%", + "diff_0.025" : "52.79%", + "diff_0.05" : "44.43%", + "diff_0.1" : "31.90%", + "diff_0.25" : "12.83%", + "diff_0.5" : "2.99%", + "diff_1" : "0.37%", + "diff_max" : "2.266", + "diff_similarity" : "85.62%" + } + }, + "model/model.7/model.7.3/Conv_output_0_92" : { + "inputs" : [ "@model/model.7/model.7.2/Relu_output_0_93:out0" ], + "name" : "model/model.7/model.7.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "80.74%", + "diff_0.01" : "74.44%", + "diff_0.025" : "66.26%", + "diff_0.05" : "54.05%", + "diff_0.1" : "34.37%", + "diff_0.25" : "7.32%", + "diff_0.5" : "0.73%", + "diff_1" : "0.02%", + "diff_max" : "1.432", + "diff_similarity" : "63.26%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.7/model.7.5/Relu_output_0_91" : { + "inputs" : [ "@model/model.7/model.7.3/Conv_output_0_92:out0" ], + "name" : "model/model.7/model.7.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "62.93%", + "diff_0.01" : "58.95%", + "diff_0.025" : "52.44%", + "diff_0.05" : "42.47%", + "diff_0.1" : "25.89%", + "diff_0.25" : "4.40%", + "diff_0.5" : "0.34%", + "diff_1" : "0.00%", + "diff_max" : "1.001", + "diff_similarity" : "74.58%" + } + }, + "model/model.8/model.8.0/Conv_output_0_90" : { + "inputs" : [ "@model/model.7/model.7.5/Relu_output_0_91:out0" ], + "name" : "model/model.8/model.8.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "78.77%", + "diff_0.01" : "73.28%", + "diff_0.025" : "65.84%", + "diff_0.05" : "55.94%", + "diff_0.1" : "40.34%", + "diff_0.25" : "14.60%", + "diff_0.5" : "2.69%", + "diff_1" : "0.08%", + "diff_max" : "1.891", + "diff_similarity" : "96.91%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 512, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.8/model.8.2/Relu_output_0_89" : { + "inputs" : [ "@model/model.8/model.8.0/Conv_output_0_90:out0" ], + "name" : "model/model.8/model.8.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "62.02%", + "diff_0.01" : "58.18%", + "diff_0.025" : "52.10%", + "diff_0.05" : "44.19%", + "diff_0.1" : "31.88%", + "diff_0.25" : "11.04%", + "diff_0.5" : "1.68%", + "diff_1" : "0.02%", + "diff_max" : "1.891", + "diff_similarity" : "97.73%" + } + }, + "model/model.8/model.8.3/Conv_output_0_88" : { + "inputs" : [ "@model/model.8/model.8.2/Relu_output_0_89:out0" ], + "name" : "model/model.8/model.8.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "74.15%", + "diff_0.01" : "66.88%", + "diff_0.025" : "59.24%", + "diff_0.05" : "47.09%", + "diff_0.1" : "28.56%", + "diff_0.25" : "4.97%", + "diff_0.5" : "0.28%", + "diff_1" : "0.00%", + "diff_max" : "1.380", + "diff_similarity" : "64.21%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.8/model.8.5/Relu_output_0_87" : { + "inputs" : [ "@model/model.8/model.8.3/Conv_output_0_88:out0" ], + "name" : "model/model.8/model.8.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "50.91%", + "diff_0.01" : "47.18%", + "diff_0.025" : "41.47%", + "diff_0.05" : "32.42%", + "diff_0.1" : "18.28%", + "diff_0.25" : "2.29%", + "diff_0.5" : "0.09%", + "diff_1" : "0.00%", + "diff_max" : "1.151", + "diff_similarity" : "76.28%" + } + }, + "model/model.9/model.9.0/Conv_output_0_86" : { + "inputs" : [ "@model/model.8/model.8.5/Relu_output_0_87:out0" ], + "name" : "model/model.9/model.9.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "70.35%", + "diff_0.01" : "62.08%", + "diff_0.025" : "55.61%", + "diff_0.05" : "47.43%", + "diff_0.1" : "34.87%", + "diff_0.25" : "11.95%", + "diff_0.5" : "2.09%", + "diff_1" : "0.11%", + "diff_max" : "2.654", + "diff_similarity" : "88.68%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 512, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.9/model.9.2/Relu_output_0_85" : { + "inputs" : [ "@model/model.9/model.9.0/Conv_output_0_86:out0" ], + "name" : "model/model.9/model.9.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "51.59%", + "diff_0.01" : "45.58%", + "diff_0.025" : "41.07%", + "diff_0.05" : "34.70%", + "diff_0.1" : "24.89%", + "diff_0.25" : "8.69%", + "diff_0.5" : "1.30%", + "diff_1" : "0.02%", + "diff_max" : "2.654", + "diff_similarity" : "92.19%" + } + }, + "model/model.9/model.9.3/Conv_output_0_84" : { + "inputs" : [ "@model/model.9/model.9.2/Relu_output_0_85:out0" ], + "name" : "model/model.9/model.9.3/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "78.82%", + "diff_0.01" : "71.53%", + "diff_0.025" : "63.58%", + "diff_0.05" : "51.72%", + "diff_0.1" : "32.86%", + "diff_0.25" : "6.96%", + "diff_0.5" : "0.40%", + "diff_1" : "0.00%", + "diff_max" : "1.186", + "diff_similarity" : "59.69%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 512 + } + }, + "model/model.9/model.9.5/Relu_output_0_83" : { + "inputs" : [ "@model/model.9/model.9.3/Conv_output_0_84:out0" ], + "name" : "model/model.9/model.9.5/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "54.21%", + "diff_0.01" : "50.17%", + "diff_0.025" : "43.81%", + "diff_0.05" : "34.10%", + "diff_0.1" : "19.82%", + "diff_0.25" : "3.03%", + "diff_0.5" : "0.09%", + "diff_1" : "0.00%", + "diff_max" : "1.041", + "diff_similarity" : "74.94%" + } + }, + "onnx//Concat_345_7" : { + "inputs" : [ + "@initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11:out0" + ], + "name" : "onnx//Concat_345", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "52.46%", + "diff_0.01" : "9.17%", + "diff_0.025" : "5.54%", + "diff_0.05" : "3.68%", + "diff_0.1" : "2.24%", + "diff_0.25" : "0.87%", + "diff_0.5" : "0.28%", + "diff_1" : "0.00%", + "diff_max" : "0.819", + "diff_similarity" : "97.75%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 19 + } + }, + "onnx//Concat_348_6" : { + "inputs" : [ "@initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10:out0" ], + "name" : "onnx//Concat_348", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "27.18%", + "diff_0.01" : "3.39%", + "diff_0.025" : "1.80%", + "diff_0.05" : "1.09%", + "diff_0.1" : "0.63%", + "diff_0.25" : "0.28%", + "diff_0.5" : "0.10%", + "diff_1" : "0.00%", + "diff_max" : "1.029", + "diff_similarity" : "7.11%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 38 + } + }, + "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13" : { + "inputs" : [ "@refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0" ], + "name" : "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.77%", + "diff_0.01" : "97.58%", + "diff_0.025" : "94.05%", + "diff_0.05" : "87.80%", + "diff_0.1" : "77.15%", + "diff_0.25" : "47.74%", + "diff_0.5" : "15.40%", + "diff_1" : "2.34%", + "diff_max" : "2.226", + "diff_similarity" : "35.49%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9" : { + "inputs" : [ + "@refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:out0" + ], + "name" : "refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "21.00%", + "diff_0.01" : "18.85%", + "diff_0.025" : "15.56%", + "diff_0.05" : "11.34%", + "diff_0.1" : "6.46%", + "diff_0.25" : "2.17%", + "diff_0.5" : "0.50%", + "diff_1" : "0.08%", + "diff_max" : "2.117", + "diff_similarity" : "88.00%" + } + }, + "refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12" : { + "inputs" : [ "@refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0" ], + "name" : "refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.97%", + "diff_0.01" : "99.74%", + "diff_0.025" : "99.32%", + "diff_0.05" : "98.55%", + "diff_0.1" : "96.51%", + "diff_0.25" : "81.68%", + "diff_0.5" : "40.75%", + "diff_1" : "6.65%", + "diff_max" : "2.389", + "diff_similarity" : "-4.56%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8" : { + "inputs" : [ "@refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:out0" ], + "name" : "refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "3.18%", + "diff_0.01" : "2.89%", + "diff_0.025" : "2.41%", + "diff_0.05" : "1.88%", + "diff_0.1" : "1.15%", + "diff_0.25" : "0.51%", + "diff_0.5" : "0.27%", + "diff_1" : "0.06%", + "diff_max" : "1.848", + "diff_similarity" : "1.26%" + } + }, + "refinement_stages.0/trunk/trunk.0/Add_output_0_60" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0", + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.0/Add_output_0", + "op" : "add", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "86.47%", + "diff_0.01" : "79.11%", + "diff_0.025" : "67.44%", + "diff_0.05" : "50.95%", + "diff_0.1" : "28.03%", + "diff_0.25" : "5.68%", + "diff_0.5" : "0.94%", + "diff_1" : "0.04%", + "diff_max" : "1.592", + "diff_similarity" : "80.17%" + } + }, + "refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69" : { + "inputs" : [ "@Concat_output_0_73:out0" ], + "name" : "refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "98.96%", + "diff_0.01" : "87.88%", + "diff_0.025" : "73.45%", + "diff_0.05" : "47.86%", + "diff_0.1" : "15.62%", + "diff_0.25" : "2.38%", + "diff_0.5" : "0.21%", + "diff_1" : "0.02%", + "diff_max" : "2.244", + "diff_similarity" : "76.29%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "54.86%", + "diff_0.01" : "45.99%", + "diff_0.025" : "33.15%", + "diff_0.05" : "17.68%", + "diff_0.1" : "4.56%", + "diff_0.25" : "0.13%", + "diff_0.5" : "0.00%", + "diff_1" : "0.00%", + "diff_max" : "0.685", + "diff_similarity" : "93.23%" + } + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.58%", + "diff_0.01" : "96.01%", + "diff_0.025" : "89.94%", + "diff_0.05" : "80.12%", + "diff_0.1" : "61.95%", + "diff_0.25" : "26.58%", + "diff_0.5" : "7.60%", + "diff_1" : "1.36%", + "diff_max" : "3.275", + "diff_similarity" : "28.98%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "69.37%", + "diff_0.01" : "65.73%", + "diff_0.025" : "60.25%", + "diff_0.05" : "50.60%", + "diff_0.1" : "34.83%", + "diff_0.25" : "9.51%", + "diff_0.5" : "1.51%", + "diff_1" : "0.12%", + "diff_max" : "2.119", + "diff_similarity" : "59.72%" + } + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.42%", + "diff_0.01" : "93.77%", + "diff_0.025" : "84.66%", + "diff_0.05" : "70.35%", + "diff_0.1" : "46.94%", + "diff_0.25" : "14.91%", + "diff_0.5" : "3.61%", + "diff_1" : "0.45%", + "diff_max" : "2.939", + "diff_similarity" : "27.67%", + "dilation" : [ 1, 2, 2, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 2, 2, 2, 2 ], + "pad_h" : 2, + "pad_method" : "padding_const", + "pad_w" : 2, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "70.33%", + "diff_0.01" : "65.14%", + "diff_0.025" : "56.72%", + "diff_0.05" : "43.50%", + "diff_0.1" : "24.47%", + "diff_0.25" : "4.80%", + "diff_0.5" : "0.80%", + "diff_1" : "0.03%", + "diff_max" : "1.480", + "diff_similarity" : "58.95%" + } + }, + "refinement_stages.0/trunk/trunk.1/Add_output_0_47" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0", + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.1/Add_output_0", + "op" : "add", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "88.92%", + "diff_0.01" : "84.06%", + "diff_0.025" : "76.14%", + "diff_0.05" : "63.61%", + "diff_0.1" : "42.69%", + "diff_0.25" : "11.68%", + "diff_0.5" : "1.59%", + "diff_1" : "0.16%", + "diff_max" : "1.816", + "diff_similarity" : "71.99%" + } + }, + "refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56" : { + "inputs" : [ "@refinement_stages.0/trunk/trunk.0/Add_output_0_60:out0" ], + "name" : "refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.32%", + "diff_0.01" : "93.12%", + "diff_0.025" : "82.84%", + "diff_0.05" : "66.82%", + "diff_0.1" : "41.85%", + "diff_0.25" : "10.71%", + "diff_0.5" : "1.39%", + "diff_1" : "0.04%", + "diff_max" : "1.646", + "diff_similarity" : "53.53%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "60.77%", + "diff_0.01" : "54.89%", + "diff_0.025" : "45.24%", + "diff_0.05" : "30.92%", + "diff_0.1" : "12.72%", + "diff_0.25" : "1.01%", + "diff_0.5" : "0.08%", + "diff_1" : "0.01%", + "diff_max" : "1.490", + "diff_similarity" : "83.10%" + } + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.60%", + "diff_0.01" : "95.68%", + "diff_0.025" : "89.07%", + "diff_0.05" : "78.55%", + "diff_0.1" : "59.51%", + "diff_0.25" : "24.61%", + "diff_0.5" : "7.29%", + "diff_1" : "0.99%", + "diff_max" : "2.990", + "diff_similarity" : "16.38%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "68.97%", + "diff_0.01" : "65.06%", + "diff_0.025" : "58.35%", + "diff_0.05" : "48.07%", + "diff_0.1" : "30.93%", + "diff_0.25" : "6.29%", + "diff_0.5" : "0.66%", + "diff_1" : "0.04%", + "diff_max" : "1.575", + "diff_similarity" : "53.25%" + } + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.62%", + "diff_0.01" : "96.12%", + "diff_0.025" : "90.32%", + "diff_0.05" : "80.86%", + "diff_0.1" : "63.57%", + "diff_0.25" : "28.34%", + "diff_0.5" : "7.21%", + "diff_1" : "0.56%", + "diff_max" : "2.603", + "diff_similarity" : "21.07%", + "dilation" : [ 1, 2, 2, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 2, 2, 2, 2 ], + "pad_h" : 2, + "pad_method" : "padding_const", + "pad_w" : 2, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "72.08%", + "diff_0.01" : "68.39%", + "diff_0.025" : "61.99%", + "diff_0.05" : "52.28%", + "diff_0.1" : "35.93%", + "diff_0.25" : "9.51%", + "diff_0.5" : "1.27%", + "diff_1" : "0.12%", + "diff_max" : "1.772", + "diff_similarity" : "52.71%" + } + }, + "refinement_stages.0/trunk/trunk.2/Add_output_0_36" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0", + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.2/Add_output_0", + "op" : "add", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "84.40%", + "diff_0.01" : "79.93%", + "diff_0.025" : "72.98%", + "diff_0.05" : "61.45%", + "diff_0.1" : "42.90%", + "diff_0.25" : "12.89%", + "diff_0.5" : "2.02%", + "diff_1" : "0.12%", + "diff_max" : "2.217", + "diff_similarity" : "72.72%" + } + }, + "refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43" : { + "inputs" : [ "@refinement_stages.0/trunk/trunk.1/Add_output_0_47:out0" ], + "name" : "refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.47%", + "diff_0.01" : "94.86%", + "diff_0.025" : "87.49%", + "diff_0.05" : "75.57%", + "diff_0.1" : "55.49%", + "diff_0.25" : "21.60%", + "diff_0.5" : "3.77%", + "diff_1" : "0.24%", + "diff_max" : "1.888", + "diff_similarity" : "41.47%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "50.18%", + "diff_0.01" : "45.77%", + "diff_0.025" : "38.83%", + "diff_0.05" : "28.06%", + "diff_0.1" : "13.32%", + "diff_0.25" : "1.04%", + "diff_0.5" : "0.07%", + "diff_1" : "0.01%", + "diff_max" : "1.707", + "diff_similarity" : "83.00%" + } + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.60%", + "diff_0.01" : "96.05%", + "diff_0.025" : "90.10%", + "diff_0.05" : "80.35%", + "diff_0.1" : "62.36%", + "diff_0.25" : "26.22%", + "diff_0.5" : "6.25%", + "diff_1" : "0.49%", + "diff_max" : "2.653", + "diff_similarity" : "19.26%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "61.97%", + "diff_0.01" : "58.37%", + "diff_0.025" : "52.17%", + "diff_0.05" : "42.81%", + "diff_0.1" : "26.65%", + "diff_0.25" : "4.78%", + "diff_0.5" : "0.47%", + "diff_1" : "0.04%", + "diff_max" : "1.846", + "diff_similarity" : "57.42%" + } + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.65%", + "diff_0.01" : "96.46%", + "diff_0.025" : "91.12%", + "diff_0.05" : "82.46%", + "diff_0.1" : "66.21%", + "diff_0.25" : "30.60%", + "diff_0.5" : "7.49%", + "diff_1" : "0.45%", + "diff_max" : "2.492", + "diff_similarity" : "27.77%", + "dilation" : [ 1, 2, 2, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 2, 2, 2, 2 ], + "pad_h" : 2, + "pad_method" : "padding_const", + "pad_w" : 2, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "68.90%", + "diff_0.01" : "65.66%", + "diff_0.025" : "60.09%", + "diff_0.05" : "51.68%", + "diff_0.1" : "36.46%", + "diff_0.25" : "10.63%", + "diff_0.5" : "1.58%", + "diff_1" : "0.08%", + "diff_max" : "1.744", + "diff_similarity" : "55.78%" + } + }, + "refinement_stages.0/trunk/trunk.3/Add_output_0_26" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0", + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.3/Add_output_0", + "op" : "add", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "82.72%", + "diff_0.01" : "78.43%", + "diff_0.025" : "71.34%", + "diff_0.05" : "60.26%", + "diff_0.1" : "41.53%", + "diff_0.25" : "11.68%", + "diff_0.5" : "1.55%", + "diff_1" : "0.11%", + "diff_max" : "1.910", + "diff_similarity" : "77.57%" + } + }, + "refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33" : { + "inputs" : [ "@refinement_stages.0/trunk/trunk.2/Add_output_0_36:out0" ], + "name" : "refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.59%", + "diff_0.01" : "95.67%", + "diff_0.025" : "89.27%", + "diff_0.05" : "79.21%", + "diff_0.1" : "62.45%", + "diff_0.25" : "30.66%", + "diff_0.5" : "4.93%", + "diff_1" : "0.08%", + "diff_max" : "1.601", + "diff_similarity" : "33.96%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "40.85%", + "diff_0.01" : "36.75%", + "diff_0.025" : "30.45%", + "diff_0.05" : "21.39%", + "diff_0.1" : "9.10%", + "diff_0.25" : "0.51%", + "diff_0.5" : "0.06%", + "diff_1" : "0.00%", + "diff_max" : "1.215", + "diff_similarity" : "85.62%" + } + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.59%", + "diff_0.01" : "95.97%", + "diff_0.025" : "89.76%", + "diff_0.05" : "79.78%", + "diff_0.1" : "61.70%", + "diff_0.25" : "25.74%", + "diff_0.5" : "4.70%", + "diff_1" : "0.18%", + "diff_max" : "2.018", + "diff_similarity" : "28.03%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "53.03%", + "diff_0.01" : "49.39%", + "diff_0.025" : "43.15%", + "diff_0.05" : "34.15%", + "diff_0.1" : "19.69%", + "diff_0.25" : "3.22%", + "diff_0.5" : "0.35%", + "diff_1" : "0.02%", + "diff_max" : "1.939", + "diff_similarity" : "66.09%" + } + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "98.80%", + "diff_0.01" : "94.95%", + "diff_0.025" : "88.72%", + "diff_0.05" : "78.37%", + "diff_0.1" : "59.47%", + "diff_0.25" : "23.26%", + "diff_0.5" : "4.94%", + "diff_1" : "0.33%", + "diff_max" : "2.108", + "diff_similarity" : "46.47%", + "dilation" : [ 1, 2, 2, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 2, 2, 2, 2 ], + "pad_h" : 2, + "pad_method" : "padding_const", + "pad_w" : 2, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "74.86%", + "diff_0.01" : "71.12%", + "diff_0.025" : "64.87%", + "diff_0.05" : "55.06%", + "diff_0.1" : "37.62%", + "diff_0.25" : "10.27%", + "diff_0.5" : "1.27%", + "diff_1" : "0.08%", + "diff_max" : "1.910", + "diff_similarity" : "66.99%" + } + }, + "refinement_stages.0/trunk/trunk.4/Add_output_0_16" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0", + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.4/Add_output_0", + "op" : "add", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "81.02%", + "diff_0.01" : "77.46%", + "diff_0.025" : "70.16%", + "diff_0.05" : "60.38%", + "diff_0.1" : "45.08%", + "diff_0.25" : "19.11%", + "diff_0.5" : "5.23%", + "diff_1" : "1.35%", + "diff_max" : "3.992", + "diff_similarity" : "68.04%" + } + }, + "refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23" : { + "inputs" : [ "@refinement_stages.0/trunk/trunk.3/Add_output_0_26:out0" ], + "name" : "refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.39%", + "diff_0.01" : "93.93%", + "diff_0.025" : "85.47%", + "diff_0.05" : "73.63%", + "diff_0.1" : "58.48%", + "diff_0.25" : "31.39%", + "diff_0.5" : "6.15%", + "diff_1" : "0.16%", + "diff_max" : "1.795", + "diff_similarity" : "28.44%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 1, + "ksize_w" : 1, + "pad" : [ 0, 0, 0, 0 ], + "pad_h" : 0, + "pad_method" : "padding_const", + "pad_w" : 0, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "41.23%", + "diff_0.01" : "35.67%", + "diff_0.025" : "27.31%", + "diff_0.05" : "16.15%", + "diff_0.1" : "4.89%", + "diff_0.25" : "0.29%", + "diff_0.5" : "0.05%", + "diff_1" : "0.00%", + "diff_max" : "1.106", + "diff_similarity" : "86.65%" + } + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.39%", + "diff_0.01" : "94.00%", + "diff_0.025" : "85.03%", + "diff_0.05" : "71.03%", + "diff_0.1" : "46.89%", + "diff_0.25" : "10.97%", + "diff_0.5" : "1.59%", + "diff_1" : "0.11%", + "diff_max" : "5.181", + "diff_similarity" : "64.29%", + "dilation" : [ 1, 1, 1, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 1, 1, 1, 1 ], + "pad_h" : 1, + "pad_method" : "padding_const", + "pad_w" : 1, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "60.98%", + "diff_0.01" : "55.85%", + "diff_0.025" : "47.54%", + "diff_0.05" : "36.14%", + "diff_0.1" : "20.04%", + "diff_0.25" : "3.76%", + "diff_0.5" : "0.63%", + "diff_1" : "0.04%", + "diff_max" : "2.891", + "diff_similarity" : "81.52%" + } + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0", + "op" : "convolution", + "outputs" : [ "out0" ], + "parameters" : { + "bias" : true, + "diff_0.001" : "99.67%", + "diff_0.01" : "96.66%", + "diff_0.025" : "91.70%", + "diff_0.05" : "83.71%", + "diff_0.1" : "69.12%", + "diff_0.25" : "37.56%", + "diff_0.5" : "12.83%", + "diff_1" : "2.02%", + "diff_max" : "4.061", + "diff_similarity" : "38.64%", + "dilation" : [ 1, 2, 2, 1 ], + "group_number" : 1, + "ksize_h" : 3, + "ksize_w" : 3, + "pad" : [ 2, 2, 2, 2 ], + "pad_h" : 2, + "pad_method" : "padding_const", + "pad_w" : 2, + "padding" : "VALID", + "padding_mode" : "CONSTANT", + "regularize" : false, + "stride_h" : 1, + "stride_w" : 1, + "weights" : 128 + } + }, + "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17" : { + "inputs" : [ + "@refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:out0" + ], + "name" : "refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0", + "op" : "relu", + "outputs" : [ "out0" ], + "parameters" : { + "diff_0.001" : "72.11%", + "diff_0.01" : "69.75%", + "diff_0.025" : "64.03%", + "diff_0.05" : "56.43%", + "diff_0.1" : "43.47%", + "diff_0.25" : "17.92%", + "diff_0.5" : "5.09%", + "diff_1" : "1.31%", + "diff_max" : "4.061", + "diff_similarity" : "57.17%" + } + } + }, + "MetaData" : { + "Name" : "torch_jit", + "NetransVersion" : "6", + "Org_Platform" : "onnx", + "Platform" : "tensorflow" + }, + "quantize_info" : {} +} diff --git a/resource/openpose/output_2.dat b/resource/openpose/output_2.dat new file mode 100644 index 0000000..3edf1db Binary files /dev/null and b/resource/openpose/output_2.dat differ diff --git a/resource/openpose/output_2.npy b/resource/openpose/output_2.npy new file mode 100644 index 0000000..7d3d804 Binary files /dev/null and b/resource/openpose/output_2.npy differ diff --git a/resource/openpose/output_3.dat b/resource/openpose/output_3.dat new file mode 100644 index 0000000..286d7d5 Binary files /dev/null and b/resource/openpose/output_3.dat differ diff --git a/resource/openpose/output_3.npy b/resource/openpose/output_3.npy new file mode 100644 index 0000000..c41381c Binary files /dev/null and b/resource/openpose/output_3.npy differ diff --git a/resource/openpose/result.jpg b/resource/openpose/result.jpg new file mode 100644 index 0000000..9cec841 Binary files /dev/null and b/resource/openpose/result.jpg differ diff --git a/resource/openpose/wksp/asymmetric_affine/.cproject b/resource/openpose/wksp/asymmetric_affine/.cproject new file mode 100644 index 0000000..64f6138 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/.cproject @@ -0,0 +1,966 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/resource/openpose/wksp/asymmetric_affine/.project b/resource/openpose/wksp/asymmetric_affine/.project new file mode 100644 index 0000000..657028e --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/.project @@ -0,0 +1,73 @@ + + + asymmetricaffine + + + + + + org.eclipse.cdt.managedbuilder.core.genmakebuilder + clean,full,incremental, + + + ?name? + + + + org.eclipse.cdt.make.core.append_environment + true + + + org.eclipse.cdt.make.core.buildArguments + + + + org.eclipse.cdt.make.core.buildCommand + make + + + org.eclipse.cdt.make.core.buildLocation + ${workspace_loc:/${project_name}/Debug} + + + org.eclipse.cdt.make.core.contents + org.eclipse.cdt.make.core.activeConfigSettings + + + org.eclipse.cdt.make.core.enableAutoBuild + false + + + org.eclipse.cdt.make.core.enableCleanBuild + true + + + org.eclipse.cdt.make.core.enableFullBuild + true + + + org.eclipse.cdt.make.core.stopOnError + true + + + org.eclipse.cdt.make.core.useDefaultBuildCmd + true + + + + + org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder + full,incremental, + + + + + + com.verisilicon.vdt.core.vdtnature + com.verisilicon.vdt.core.ovxnature + org.eclipse.cdt.core.cnature + org.eclipse.cdt.core.ccnature + org.eclipse.cdt.managedbuilder.core.managedBuildNature + org.eclipse.cdt.managedbuilder.core.ScannerConfigNature + + diff --git a/resource/openpose/wksp/asymmetric_affine/BUILD b/resource/openpose/wksp/asymmetric_affine/BUILD new file mode 100644 index 0000000..bbd7b4b --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/BUILD @@ -0,0 +1,28 @@ +# AUTO GENERATED FILE, BUILD AND RUN IN OVXLIB + +package(default_visibility = ["//visibility:public"]) + +filegroup( + name = "srcs", + srcs = + [ + "vnn_asymmetricaffine.c", + "vnn_asymmetricaffine.h", + "vnn_post_process.c", + "vnn_post_process.h", + "vnn_pre_process.c", + "vnn_pre_process.h", + "vnn_global.h", + "main.c", + ], +) + +cc_binary( + name = "inference", + srcs = [":srcs"] + ["//:ovxlib"], + deps = [ + "//third-party/jpeg-9b:libjpeg", + "//:ovxlib", + "@VIV_SDK//:VIV_SDK_LIB", + ], +) diff --git a/resource/openpose/wksp/asymmetric_affine/asymmetric_affine.export.data b/resource/openpose/wksp/asymmetric_affine/asymmetric_affine.export.data new file mode 100644 index 0000000..3bbd58a Binary files /dev/null and b/resource/openpose/wksp/asymmetric_affine/asymmetric_affine.export.data differ diff --git a/resource/openpose/wksp/asymmetric_affine/asymmetricaffine.2012.vcxproj b/resource/openpose/wksp/asymmetric_affine/asymmetricaffine.2012.vcxproj new file mode 100644 index 0000000..837e3b0 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/asymmetricaffine.2012.vcxproj @@ -0,0 +1,224 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Jenkins-Debug + Win32 + + + Jenkins-Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + asymmetricaffine + {816BB4FE-3813-3D87-9CF9-2B0D268ED67A} + asymmetricaffine + + + + Application + true + v110 + MultiByte + + + Application + true + v110 + MultiByte + + + Application + true + v110 + MultiByte + + + Application + true + v110 + MultiByte + + + Application + false + v110 + true + MultiByte + + + Application + false + v110 + true + MultiByte + + + + + + + + + + + + + + + + + + + + + + + + $([System.IO.File]::ReadAllText($(SolutionDir)ovxlib.path.user).Trim()) + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + + + + + + + + + + + + diff --git a/resource/openpose/wksp/asymmetric_affine/asymmetricaffine.vcxproj b/resource/openpose/wksp/asymmetric_affine/asymmetricaffine.vcxproj new file mode 100644 index 0000000..f7ac0e6 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/asymmetricaffine.vcxproj @@ -0,0 +1,224 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Jenkins-Debug + Win32 + + + Jenkins-Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + asymmetricaffine + {CFF12436-742A-4286-AA83-291F06E8D652} + asymmetricaffine + + + + Application + true + v142 + MultiByte + + + Application + true + v142 + MultiByte + + + Application + true + v142 + MultiByte + + + Application + true + v142 + MultiByte + + + Application + false + v142 + true + MultiByte + + + Application + false + v142 + true + MultiByte + + + + + + + + + + + + + + + + + + + + + + + + $([System.IO.File]::ReadAllText($(SolutionDir)ovxlib.path.user).Trim()) + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/resource/openpose/wksp/asymmetric_affine/dump_core_graph.json b/resource/openpose/wksp/asymmetric_affine/dump_core_graph.json new file mode 100644 index 0000000..b5a7f39 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/dump_core_graph.json @@ -0,0 +1,2423 @@ +{ + "MetaData": { + "Name": "core graph", + "Version": "0.0.1" + }, + "Layers": { + "node_100000": { + "inputs": [], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[424", " 256", " 3", " 1]"], + "input_lifetime": "kInput", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 3", " 32]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[32]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["2", "2"], + "dilation": ["0", "0"] + } + }, + "node_100001": { + "inputs": ["@node_100000:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100002": { + "inputs": ["@node_100001:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 32", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[32]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100003": { + "inputs": ["@node_100002:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100004": { + "inputs": ["@node_100003:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 32", " 64]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[64]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[212", " 128", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100005": { + "inputs": ["@node_100004:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[212", " 128", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[212", " 128", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100006": { + "inputs": ["@node_100005:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[212", " 128", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 64", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[64]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["2", "2"], + "dilation": ["0", "0"] + } + }, + "node_100007": { + "inputs": ["@node_100006:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100008": { + "inputs": ["@node_100007:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 64", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100009": { + "inputs": ["@node_100008:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100010": { + "inputs": ["@node_100009:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100011": { + "inputs": ["@node_100010:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100012": { + "inputs": ["@node_100011:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100013": { + "inputs": ["@node_100012:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100014": { + "inputs": ["@node_100013:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["2", "2"], + "dilation": ["0", "0"] + } + }, + "node_100015": { + "inputs": ["@node_100014:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100016": { + "inputs": ["@node_100015:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 256]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100017": { + "inputs": ["@node_100016:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100018": { + "inputs": ["@node_100017:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 256", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100019": { + "inputs": ["@node_100018:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100020": { + "inputs": ["@node_100019:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 256", " 256]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100021": { + "inputs": ["@node_100020:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100022": { + "inputs": ["@node_100021:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 256", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100023": { + "inputs": ["@node_100022:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100024": { + "inputs": ["@node_100023:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 256", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100025": { + "inputs": ["@node_100024:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100026": { + "inputs": ["@node_100025:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[5", " 5", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100027": { + "inputs": ["@node_100026:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100028": { + "inputs": ["@node_100027:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100029": { + "inputs": ["@node_100028:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100030": { + "inputs": ["@node_100029:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100031": { + "inputs": ["@node_100030:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100032": { + "inputs": ["@node_100031:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100033": { + "inputs": ["@node_100032:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100034": { + "inputs": ["@node_100033:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100035": { + "inputs": ["@node_100034:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100036": { + "inputs": ["@node_100035:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100037": { + "inputs": ["@node_100036:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100038": { + "inputs": ["@node_100037:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100039": { + "inputs": ["@node_100038:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100040": { + "inputs": ["@node_100039:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100041": { + "inputs": ["@node_100040:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100042": { + "inputs": ["@node_100041:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100043": { + "inputs": ["@node_100042:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100044": { + "inputs": ["@node_100043:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100045": { + "inputs": ["@node_100044:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100046": { + "inputs": ["@node_100045:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100047": { + "inputs": ["@node_100046:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100048": { + "inputs": ["@node_100047:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100049": { + "inputs": ["@node_100048:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100050": { + "inputs": ["@node_100049:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100051": { + "inputs": ["@node_100050:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100052": { + "inputs": ["@node_100051:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100053": { + "inputs": ["@node_100052:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100054": { + "inputs": ["@node_100053:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100055": { + "inputs": ["@node_100054:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100056": { + "inputs": ["@node_100055:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100057": { + "inputs": ["@node_100056:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100058": { + "inputs": ["@node_100057:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100059": { + "inputs": ["@node_100058:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kFloat16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100060": { + "inputs": ["@node_100047:out0", "@node_100059:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kUInt8", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kFloat16", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100061": { + "inputs": ["@node_100060:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100062": { + "inputs": ["@node_100061:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100063": { + "inputs": ["@node_100062:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100064": { + "inputs": ["@node_100063:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100065": { + "inputs": ["@node_100064:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100066": { + "inputs": ["@node_100065:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100067": { + "inputs": ["@node_100066:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100068": { + "inputs": ["@node_100067:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100069": { + "inputs": ["@node_100068:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100070": { + "inputs": ["@node_100068:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100071": { + "inputs": ["@node_100069:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100072": { + "inputs": ["@node_100070:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100073": { + "inputs": ["@node_100071:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 38]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[38]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 38", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100074": { + "inputs": ["@node_100072:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 512", " 19]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[19]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 19", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100075": { + "inputs": ["@node_100062:out0", "@node_100074:out0", "@node_100073:out0"], + "outputs": ["out0"], + "op": "Concat", + "parameters": { + "input_0_dtype": "kUInt8", + "input_0_shape": ["[53", " 32", " 128", " 1]"], + "input_0_lifetime": "kTransient", + "input_0_dma_mem_attr": "0", + "input_1_dtype": "kUInt8", + "input_1_shape": ["[53", " 32", " 19", " 1]"], + "input_1_lifetime": "kOutput", + "input_1_dma_mem_attr": "0", + "input_2_dtype": "kUInt8", + "input_2_shape": ["[53", " 32", " 38", " 1]"], + "input_2_lifetime": "kOutput", + "input_2_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 185", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "axis": 2 + } + }, + "node_100076": { + "inputs": ["@node_100075:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 185", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 185", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100077": { + "inputs": ["@node_100076:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100078": { + "inputs": ["@node_100077:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100079": { + "inputs": ["@node_100078:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100080": { + "inputs": ["@node_100079:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100081": { + "inputs": ["@node_100080:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100082": { + "inputs": ["@node_100077:out0", "@node_100081:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kUInt8", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kUInt8", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100083": { + "inputs": ["@node_100082:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100084": { + "inputs": ["@node_100083:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100085": { + "inputs": ["@node_100084:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100086": { + "inputs": ["@node_100085:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100087": { + "inputs": ["@node_100086:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100088": { + "inputs": ["@node_100087:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100089": { + "inputs": ["@node_100084:out0", "@node_100088:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kUInt8", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kUInt8", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100090": { + "inputs": ["@node_100089:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100091": { + "inputs": ["@node_100090:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100092": { + "inputs": ["@node_100091:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100093": { + "inputs": ["@node_100092:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100094": { + "inputs": ["@node_100093:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100095": { + "inputs": ["@node_100094:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100096": { + "inputs": ["@node_100091:out0", "@node_100095:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kUInt8", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kUInt8", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100097": { + "inputs": ["@node_100096:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100098": { + "inputs": ["@node_100097:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100099": { + "inputs": ["@node_100098:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100100": { + "inputs": ["@node_100099:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100101": { + "inputs": ["@node_100100:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100102": { + "inputs": ["@node_100101:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100103": { + "inputs": ["@node_100098:out0", "@node_100102:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kUInt8", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kUInt8", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100104": { + "inputs": ["@node_100103:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100105": { + "inputs": ["@node_100104:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100106": { + "inputs": ["@node_100105:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100107": { + "inputs": ["@node_100106:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100108": { + "inputs": ["@node_100107:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100109": { + "inputs": ["@node_100108:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100110": { + "inputs": ["@node_100105:out0", "@node_100109:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kUInt8", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kUInt8", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100111": { + "inputs": ["@node_100110:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100112": { + "inputs": ["@node_100110:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100113": { + "inputs": ["@node_100111:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100114": { + "inputs": ["@node_100112:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100115": { + "inputs": ["@node_100113:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 38]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[38]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 38", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100116": { + "inputs": ["@node_100114:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kUInt8", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kUInt8", + "wegith_shape": ["[1", " 1", " 128", " 19]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt32", + "bias_shape": ["[19]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kUInt8", + "output_shape": ["[53", " 32", " 19", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + } + } +} diff --git a/resource/openpose/wksp/asymmetric_affine/graph.json b/resource/openpose/wksp/asymmetric_affine/graph.json new file mode 100644 index 0000000..9b45e85 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/graph.json @@ -0,0 +1,1634 @@ +{ + "MetaData":{ + "Name": "Ovxlib_Debug_Graph", + "AcuityVersion": "UNKNOWN", + "Platform": "UNKNOWN", + "Org_Platform": "UNKNOWN" + }, + "Layers":{ + "uid_120":{ + "op": "CONV2D", + "inputs": [ "@data_input_uid_118:out0", "@data_input_uid_119:out0", "@data_input_uid_120:out0" ], + "inut_shape": [ [ 424, 256, 3, 1 ],[ 3, 3, 3, 32 ],[ 32 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_119":{ + "op": "RELU", + "inputs": [ "@uid_120:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_118":{ + "op": "CONV2D", + "inputs": [ "@uid_119:out0", "@data_input_uid_121:out0", "@data_input_uid_122:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ],[ 3, 3, 32, 1 ],[ 32 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_117":{ + "op": "RELU", + "inputs": [ "@uid_118:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_116":{ + "op": "CONV2D", + "inputs": [ "@uid_117:out0", "@data_input_uid_123:out0", "@data_input_uid_124:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ],[ 1, 1, 32, 64 ],[ 64 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 64, 1 ] ] + }, + "uid_115":{ + "op": "RELU", + "inputs": [ "@uid_116:out0" ], + "inut_shape": [ [ 212, 128, 64, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 64, 1 ] ] + }, + "uid_114":{ + "op": "CONV2D", + "inputs": [ "@uid_115:out0", "@data_input_uid_125:out0", "@data_input_uid_126:out0" ], + "inut_shape": [ [ 212, 128, 64, 1 ],[ 3, 3, 64, 1 ],[ 64 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 64, 1 ] ] + }, + "uid_113":{ + "op": "RELU", + "inputs": [ "@uid_114:out0" ], + "inut_shape": [ [ 106, 64, 64, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 64, 1 ] ] + }, + "uid_112":{ + "op": "CONV2D", + "inputs": [ "@uid_113:out0", "@data_input_uid_127:out0", "@data_input_uid_128:out0" ], + "inut_shape": [ [ 106, 64, 64, 1 ],[ 1, 1, 64, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_111":{ + "op": "RELU", + "inputs": [ "@uid_112:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_110":{ + "op": "CONV2D", + "inputs": [ "@uid_111:out0", "@data_input_uid_129:out0", "@data_input_uid_130:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_109":{ + "op": "RELU", + "inputs": [ "@uid_110:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_108":{ + "op": "CONV2D", + "inputs": [ "@uid_109:out0", "@data_input_uid_131:out0", "@data_input_uid_132:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_107":{ + "op": "RELU", + "inputs": [ "@uid_108:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_106":{ + "op": "CONV2D", + "inputs": [ "@uid_107:out0", "@data_input_uid_133:out0", "@data_input_uid_134:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_105":{ + "op": "RELU", + "inputs": [ "@uid_106:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_104":{ + "op": "CONV2D", + "inputs": [ "@uid_105:out0", "@data_input_uid_135:out0", "@data_input_uid_136:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 256 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_103":{ + "op": "RELU", + "inputs": [ "@uid_104:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_102":{ + "op": "CONV2D", + "inputs": [ "@uid_103:out0", "@data_input_uid_137:out0", "@data_input_uid_138:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 3, 3, 256, 1 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_101":{ + "op": "RELU", + "inputs": [ "@uid_102:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_100":{ + "op": "CONV2D", + "inputs": [ "@uid_101:out0", "@data_input_uid_139:out0", "@data_input_uid_140:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 1, 1, 256, 256 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_99":{ + "op": "RELU", + "inputs": [ "@uid_100:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_98":{ + "op": "CONV2D", + "inputs": [ "@uid_99:out0", "@data_input_uid_141:out0", "@data_input_uid_142:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 3, 3, 256, 1 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_97":{ + "op": "RELU", + "inputs": [ "@uid_98:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_96":{ + "op": "CONV2D", + "inputs": [ "@uid_97:out0", "@data_input_uid_143:out0", "@data_input_uid_144:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 1, 1, 256, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_95":{ + "op": "RELU", + "inputs": [ "@uid_96:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_94":{ + "op": "CONV2D", + "inputs": [ "@uid_95:out0", "@data_input_uid_145:out0", "@data_input_uid_146:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 5, 5, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_93":{ + "op": "RELU", + "inputs": [ "@uid_94:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_92":{ + "op": "CONV2D", + "inputs": [ "@uid_93:out0", "@data_input_uid_147:out0", "@data_input_uid_148:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_91":{ + "op": "RELU", + "inputs": [ "@uid_92:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_90":{ + "op": "CONV2D", + "inputs": [ "@uid_91:out0", "@data_input_uid_149:out0", "@data_input_uid_150:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_89":{ + "op": "RELU", + "inputs": [ "@uid_90:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_88":{ + "op": "CONV2D", + "inputs": [ "@uid_89:out0", "@data_input_uid_151:out0", "@data_input_uid_152:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_87":{ + "op": "RELU", + "inputs": [ "@uid_88:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_86":{ + "op": "CONV2D", + "inputs": [ "@uid_87:out0", "@data_input_uid_153:out0", "@data_input_uid_154:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_85":{ + "op": "RELU", + "inputs": [ "@uid_86:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_84":{ + "op": "CONV2D", + "inputs": [ "@uid_85:out0", "@data_input_uid_155:out0", "@data_input_uid_156:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_83":{ + "op": "RELU", + "inputs": [ "@uid_84:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_82":{ + "op": "CONV2D", + "inputs": [ "@uid_83:out0", "@data_input_uid_157:out0", "@data_input_uid_158:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_81":{ + "op": "RELU", + "inputs": [ "@uid_82:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_75":{ + "op": "CONV2D", + "inputs": [ "@uid_81:out0", "@data_input_uid_159:out0", "@data_input_uid_160:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_71":{ + "op": "RELU", + "inputs": [ "@uid_75:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_67":{ + "op": "CONV2D", + "inputs": [ "@uid_71:out0", "@data_input_uid_161:out0", "@data_input_uid_162:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_62":{ + "op": "RELU", + "inputs": [ "@uid_67:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_58":{ + "op": "CONV2D", + "inputs": [ "@uid_62:out0", "@data_input_uid_163:out0", "@data_input_uid_164:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_54":{ + "op": "RELU", + "inputs": [ "@uid_58:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_49":{ + "op": "CONV2D", + "inputs": [ "@uid_54:out0", "@data_input_uid_165:out0", "@data_input_uid_166:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_45":{ + "op": "RELU", + "inputs": [ "@uid_49:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_80":{ + "op": "CONV2D", + "inputs": [ "@uid_45:out0", "@data_input_uid_167:out0", "@data_input_uid_168:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_79":{ + "op": "ELU", + "inputs": [ "@uid_80:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_78":{ + "op": "CONV2D", + "inputs": [ "@uid_79:out0", "@data_input_uid_169:out0", "@data_input_uid_170:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_77":{ + "op": "ELU", + "inputs": [ "@uid_78:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_76":{ + "op": "CONV2D", + "inputs": [ "@uid_77:out0", "@data_input_uid_171:out0", "@data_input_uid_172:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_72":{ + "op": "ELU", + "inputs": [ "@uid_76:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_68":{ + "op": "CONV2D", + "inputs": [ "@uid_72:out0", "@data_input_uid_173:out0", "@data_input_uid_174:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_63":{ + "op": "ELU", + "inputs": [ "@uid_68:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_59":{ + "op": "CONV2D", + "inputs": [ "@uid_63:out0", "@data_input_uid_175:out0", "@data_input_uid_176:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_55":{ + "op": "ELU", + "inputs": [ "@uid_59:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_50":{ + "op": "CONV2D", + "inputs": [ "@uid_55:out0", "@data_input_uid_177:out0", "@data_input_uid_178:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_46":{ + "op": "ELU", + "inputs": [ "@uid_50:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_42":{ + "op": "ADD", + "inputs": [ "@uid_45:out0", "@uid_46:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_38":{ + "op": "CONV2D", + "inputs": [ "@uid_42:out0", "@data_input_uid_179:out0", "@data_input_uid_180:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_35":{ + "op": "RELU", + "inputs": [ "@uid_38:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_32":{ + "op": "CONV2D", + "inputs": [ "@uid_35:out0", "@data_input_uid_181:out0", "@data_input_uid_182:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_28":{ + "op": "RELU", + "inputs": [ "@uid_32:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_25":{ + "op": "CONV2D", + "inputs": [ "@uid_28:out0", "@data_input_uid_183:out0", "@data_input_uid_184:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_22":{ + "op": "RELU", + "inputs": [ "@uid_25:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_19":{ + "op": "CONV2D", + "inputs": [ "@uid_22:out0", "@data_input_uid_185:out0", "@data_input_uid_186:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_18":{ + "op": "RELU", + "inputs": [ "@uid_19:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_14":{ + "op": "CONV2D", + "inputs": [ "@uid_18:out0", "@data_input_uid_187:out0", "@data_input_uid_188:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_15":{ + "op": "CONV2D", + "inputs": [ "@uid_18:out0", "@data_input_uid_189:out0", "@data_input_uid_190:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_10":{ + "op": "RELU", + "inputs": [ "@uid_14:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_11":{ + "op": "RELU", + "inputs": [ "@uid_15:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_6":{ + "op": "CONV2D", + "inputs": [ "@uid_10:out0", "@data_input_uid_191:out0", "@data_input_uid_192:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 38 ],[ 38 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 38, 1 ] ] + }, + "uid_7":{ + "op": "CONV2D", + "inputs": [ "@uid_11:out0", "@data_input_uid_193:out0", "@data_input_uid_194:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 19 ],[ 19 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 19, 1 ] ] + }, + "uid_73":{ + "op": "CONCAT", + "inputs": [ "@uid_35:out0", "@uid_7:out0", "@uid_6:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 19, 1 ],[ 53, 32, 38, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 185, 1 ] ] + }, + "uid_69":{ + "op": "CONV2D", + "inputs": [ "@uid_73:out0", "@data_input_uid_195:out0", "@data_input_uid_196:out0" ], + "inut_shape": [ [ 53, 32, 185, 1 ],[ 1, 1, 185, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_64":{ + "op": "RELU", + "inputs": [ "@uid_69:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_74":{ + "op": "CONV2D", + "inputs": [ "@uid_64:out0", "@data_input_uid_197:out0", "@data_input_uid_198:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_70":{ + "op": "RELU", + "inputs": [ "@uid_74:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_66":{ + "op": "CONV2D", + "inputs": [ "@uid_70:out0", "@data_input_uid_199:out0", "@data_input_uid_200:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_65":{ + "op": "RELU", + "inputs": [ "@uid_66:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_60":{ + "op": "ADD", + "inputs": [ "@uid_64:out0", "@uid_65:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_56":{ + "op": "CONV2D", + "inputs": [ "@uid_60:out0", "@data_input_uid_201:out0", "@data_input_uid_202:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_51":{ + "op": "RELU", + "inputs": [ "@uid_56:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_61":{ + "op": "CONV2D", + "inputs": [ "@uid_51:out0", "@data_input_uid_203:out0", "@data_input_uid_204:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_57":{ + "op": "RELU", + "inputs": [ "@uid_61:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_53":{ + "op": "CONV2D", + "inputs": [ "@uid_57:out0", "@data_input_uid_205:out0", "@data_input_uid_206:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_52":{ + "op": "RELU", + "inputs": [ "@uid_53:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_47":{ + "op": "ADD", + "inputs": [ "@uid_51:out0", "@uid_52:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_43":{ + "op": "CONV2D", + "inputs": [ "@uid_47:out0", "@data_input_uid_207:out0", "@data_input_uid_208:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_39":{ + "op": "RELU", + "inputs": [ "@uid_43:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_48":{ + "op": "CONV2D", + "inputs": [ "@uid_39:out0", "@data_input_uid_209:out0", "@data_input_uid_210:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_44":{ + "op": "RELU", + "inputs": [ "@uid_48:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_41":{ + "op": "CONV2D", + "inputs": [ "@uid_44:out0", "@data_input_uid_211:out0", "@data_input_uid_212:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_40":{ + "op": "RELU", + "inputs": [ "@uid_41:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_36":{ + "op": "ADD", + "inputs": [ "@uid_39:out0", "@uid_40:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_33":{ + "op": "CONV2D", + "inputs": [ "@uid_36:out0", "@data_input_uid_213:out0", "@data_input_uid_214:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_29":{ + "op": "RELU", + "inputs": [ "@uid_33:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_37":{ + "op": "CONV2D", + "inputs": [ "@uid_29:out0", "@data_input_uid_215:out0", "@data_input_uid_216:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_34":{ + "op": "RELU", + "inputs": [ "@uid_37:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_31":{ + "op": "CONV2D", + "inputs": [ "@uid_34:out0", "@data_input_uid_217:out0", "@data_input_uid_218:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_30":{ + "op": "RELU", + "inputs": [ "@uid_31:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_26":{ + "op": "ADD", + "inputs": [ "@uid_29:out0", "@uid_30:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_23":{ + "op": "CONV2D", + "inputs": [ "@uid_26:out0", "@data_input_uid_219:out0", "@data_input_uid_220:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_20":{ + "op": "RELU", + "inputs": [ "@uid_23:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_27":{ + "op": "CONV2D", + "inputs": [ "@uid_20:out0", "@data_input_uid_221:out0", "@data_input_uid_222:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_24":{ + "op": "RELU", + "inputs": [ "@uid_27:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_21":{ + "op": "CONV2D", + "inputs": [ "@uid_24:out0", "@data_input_uid_223:out0", "@data_input_uid_224:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_17":{ + "op": "RELU", + "inputs": [ "@uid_21:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_16":{ + "op": "ADD", + "inputs": [ "@uid_20:out0", "@uid_17:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_12":{ + "op": "CONV2D", + "inputs": [ "@uid_16:out0", "@data_input_uid_225:out0", "@data_input_uid_226:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_13":{ + "op": "CONV2D", + "inputs": [ "@uid_16:out0", "@data_input_uid_227:out0", "@data_input_uid_228:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_8":{ + "op": "RELU", + "inputs": [ "@uid_12:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_9":{ + "op": "RELU", + "inputs": [ "@uid_13:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_4":{ + "op": "CONV2D", + "inputs": [ "@uid_8:out0", "@data_input_uid_229:out0", "@data_input_uid_230:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 38 ],[ 38 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 38, 1 ] ] + }, + "uid_5":{ + "op": "CONV2D", + "inputs": [ "@uid_9:out0", "@data_input_uid_231:out0", "@data_input_uid_232:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 19 ],[ 19 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 19, 1 ] ] + }, + "data_input_uid_118":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_119":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_120":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_121":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_122":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_123":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_124":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_125":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_126":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_127":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_128":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_129":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_130":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_131":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_132":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_133":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_134":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_135":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_136":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_137":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_138":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_139":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_140":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_141":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_142":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_143":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_144":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_145":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_146":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_147":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_148":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_149":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_150":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_151":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_152":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_153":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_154":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_155":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_156":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_157":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_158":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_159":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_160":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_161":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_162":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_163":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_164":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_165":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_166":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_167":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_168":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_169":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_170":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_171":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_172":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_173":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_174":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_175":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_176":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_177":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_178":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_179":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_180":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_181":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_182":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_183":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_184":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_185":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_186":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_187":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_188":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_189":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_190":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_191":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_192":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_193":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_194":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_195":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_196":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_197":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_198":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_199":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_200":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_201":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_202":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_203":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_204":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_205":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_206":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_207":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_208":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_209":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_210":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_211":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_212":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_213":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_214":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_215":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_216":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_217":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_218":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_219":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_220":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_221":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_222":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_223":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_224":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_225":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_226":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_227":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_228":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_229":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_230":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_231":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_232":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + } + } +} diff --git a/resource/openpose/wksp/asymmetric_affine/main.c b/resource/openpose/wksp/asymmetric_affine/main.c new file mode 100644 index 0000000..b7647ea --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/main.c @@ -0,0 +1,264 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network application project entry file +****************************************************************************/ +/*------------------------------------------- + Includes +-------------------------------------------*/ +#include +#include +#include +#ifdef __linux__ +#include +#include +#elif defined(_WIN32) +#include +#endif + +#define _BASETSD_H + +#include "vsi_nn_pub.h" + +#include "vnn_global.h" +#include "vnn_pre_process.h" +#include "vnn_post_process.h" +#include "vnn_asymmetricaffine.h" + +/*------------------------------------------- + Macros and Variables +-------------------------------------------*/ +#ifdef __linux__ +#define VSI_UINT64_SPECIFIER PRIu64 +#elif defined(_WIN32) +#define VSI_UINT64_SPECIFIER "I64u" +#endif + +/*------------------------------------------- + Functions +-------------------------------------------*/ +static void vnn_ReleaseNeuralNetwork + ( + vsi_nn_graph_t *graph + ) +{ + vnn_ReleaseAsymmetricAffine( graph, TRUE ); + if (vnn_UseImagePreprocessNode()) + { + vnn_ReleaseBufferImage(); + } +} + +static vsi_status vnn_PostProcessNeuralNetwork + ( + vsi_nn_graph_t *graph + ) +{ + return vnn_PostProcessAsymmetricAffine( graph ); +} + +#define BILLION 1000000000 +static uint64_t get_perf_count() +{ +#if defined(__linux__) || defined(__ANDROID__) || defined(__QNX__) || defined(__CYGWIN__) + struct timespec ts; + + clock_gettime(CLOCK_MONOTONIC, &ts); + + return (uint64_t)((uint64_t)ts.tv_nsec + (uint64_t)ts.tv_sec * BILLION); +#elif defined(_WIN32) || defined(UNDER_CE) + LARGE_INTEGER freq; + LARGE_INTEGER ln; + + QueryPerformanceFrequency(&freq); + QueryPerformanceCounter(&ln); + + return (uint64_t)(ln.QuadPart * BILLION / freq.QuadPart); +#endif +} + +static vsi_status vnn_VerifyGraph + ( + vsi_nn_graph_t *graph + ) +{ + vsi_status status = VSI_FAILURE; + uint64_t tmsStart, tmsEnd, msVal, usVal; + + /* Verify graph */ + printf("Verify...\n"); + tmsStart = get_perf_count(); + status = vsi_nn_VerifyGraph( graph ); + TEST_CHECK_STATUS(status, final); + tmsEnd = get_perf_count(); + msVal = (tmsEnd - tmsStart)/1000000; + usVal = (tmsEnd - tmsStart)/1000; + printf("Verify Graph: %"VSI_UINT64_SPECIFIER"ms or %"VSI_UINT64_SPECIFIER"us\n", msVal, usVal); + +final: + return status; +} + +static vsi_status vnn_ProcessGraph + ( + vsi_nn_graph_t *graph + ) +{ + vsi_status status = VSI_FAILURE; + int32_t i,loop; + char *loop_s; + uint64_t tmsStart, tmsEnd, sigStart, sigEnd; + float msVal, usVal; + + status = VSI_FAILURE; + loop = 1; /* default loop time is 1 */ + loop_s = getenv("VNN_LOOP_TIME"); + if(loop_s) + { + loop = atoi(loop_s); + } + + /* Run graph */ + tmsStart = get_perf_count(); + printf("Start run graph [%d] times...\n", loop); + for(i = 0; i < loop; i++) + { + sigStart = get_perf_count(); +#ifdef VNN_APP_ASYNC_RUN + status = vsi_nn_AsyncRunGraph( graph ); + if(status != VSI_SUCCESS) + { + printf("Async Run graph the %d time fail\n", i); + } + TEST_CHECK_STATUS( status, final ); + + //do something here... + + status = vsi_nn_AsyncRunWait( graph ); + if(status != VSI_SUCCESS) + { + printf("Wait graph the %d time fail\n", i); + } +#else + status = vsi_nn_RunGraph( graph ); + if(status != VSI_SUCCESS) + { + printf("Run graph the %d time fail\n", i); + } +#endif + TEST_CHECK_STATUS( status, final ); + + sigEnd = get_perf_count(); + msVal = (sigEnd - sigStart)/(float)1000000; + usVal = (sigEnd - sigStart)/(float)1000; + printf("Run the %u time: %.2fms or %.2fus\n", (i + 1), msVal, usVal); + } + tmsEnd = get_perf_count(); + msVal = (tmsEnd - tmsStart)/(float)1000000; + usVal = (tmsEnd - tmsStart)/(float)1000; + printf("vxProcessGraph execution time:\n"); + printf("Total %.2fms or %.2fus\n", msVal, usVal); + printf("Average %.2fms or %.2fus\n", ((float)usVal)/1000/loop, ((float)usVal)/loop); + +final: + return status; +} + +static vsi_status vnn_PreProcessNeuralNetwork + ( + vsi_nn_graph_t *graph, + int argc, + char **argv + ) +{ + /* + * argv0: execute file + * argv1: data file + * argv2~n: inputs n file + */ + const char **inputs = (const char **)argv + 2; + uint32_t input_num = argc - 2; + + return vnn_PreProcessAsymmetricAffine( graph, inputs, input_num ); +} + +static vsi_nn_graph_t *vnn_CreateNeuralNetwork + ( + const char *data_file_name + ) +{ + vsi_nn_graph_t *graph = NULL; + uint64_t tmsStart, tmsEnd, msVal, usVal; + + tmsStart = get_perf_count(); + graph = vnn_CreateAsymmetricAffine( data_file_name, NULL, + vnn_GetPreProcessMap(), vnn_GetPreProcessMapCount(), + vnn_GetPostProcessMap(), vnn_GetPostProcessMapCount() ); + TEST_CHECK_PTR(graph, final); + + tmsEnd = get_perf_count(); + msVal = (tmsEnd - tmsStart)/1000000; + usVal = (tmsEnd - tmsStart)/1000; + printf("Create Neural Network: %"VSI_UINT64_SPECIFIER"ms or %"VSI_UINT64_SPECIFIER"us\n", msVal, usVal); + +final: + return graph; +} + +/*------------------------------------------- + Main Functions +-------------------------------------------*/ +int main + ( + int argc, + char **argv + ) +{ + vsi_status status = VSI_FAILURE; + vsi_nn_graph_t *graph; + const char *data_name = NULL; + + if(argc < 3) + { + printf("Usage: %s data_file inputs...\n", argv[0]); + return -1; + } + + data_name = (const char *)argv[1]; + + /* Create the neural network */ + graph = vnn_CreateNeuralNetwork( data_name ); + TEST_CHECK_PTR( graph, final ); + + /* Verify graph */ + status = vnn_VerifyGraph( graph ); + TEST_CHECK_STATUS( status, final); + + /* Pre process the image data */ + status = vnn_PreProcessNeuralNetwork( graph, argc, argv ); + TEST_CHECK_STATUS( status, final ); + + + + /* Process graph */ + status = vnn_ProcessGraph( graph ); + TEST_CHECK_STATUS( status, final ); + + if(VNN_APP_DEBUG) + { + /* Dump all node outputs */ + vsi_nn_DumpGraphNodeOutputs(graph, "./network_dump", NULL, 0, TRUE, 0); + } + + /* Post process output data */ + status = vnn_PostProcessNeuralNetwork( graph ); + TEST_CHECK_STATUS( status, final ); + +final: + vnn_ReleaseNeuralNetwork( graph ); + fflush(stdout); + fflush(stderr); + return status; +} + diff --git a/resource/openpose/wksp/asymmetric_affine/makefile.linux b/resource/openpose/wksp/asymmetric_affine/makefile.linux new file mode 100644 index 0000000..76f98c5 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/makefile.linux @@ -0,0 +1,127 @@ +ifeq (1,$(USE_IDE_LIB)) #idelib +CC=$(CROSS_COMPILE)gcc +CXX=$(CROSS_COMPILE)g++ +DEBUG=0 +#GWG_SDK_DIR=../IDE5.4.0/cmdtools/vsimulator +INCLUDES=-I. -I$(GWG_SDK_DIR)/include/ \ + -I$(GWG_SDK_DIR)/include/CL \ + -I$(GWG_SDK_DIR)/include/VX \ + -I$(GWG_SDK_DIR)/include/ovxlib \ + -I$(GWG_SDK_DIR)/include/jpeg +CFLAGS=-Wall -std=c++0x $(INCLUDES) -D__linux__ -DLINUX +ifeq (1,$(DEBUG)) +CFLAGS+=-g +LFLAGS+=-g +else +CFLAGS+=-O3 +LFLAGS+=-O3 +endif +LIBS+= -L$(GWG_SDK_DIR)/lib \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -lArchModelSw -lNNArchPerf +LIBS+= -L$(GWG_SDK_DIR)/lib/vsim \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/lib/x64_linux \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/lib/x64_linux/vsim \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/lib/x64_linux/vsim \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/../common/lib/ \ + -lvdtproxy +File = $(GWG_SDK_DIR)/lib/libjpeg.a +File2 = $(GWG_SDK_DIR)/lib/x64_linux/libjpeg.a +File3 = $(GWG_SDK_DIR)/../common/lib/libjpeg.a +ifeq ($(File),$(wildcard $(File))) +LIBS+= $(File) +else ifeq ($(File2),$(wildcard $(File2))) +LIBS+= $(File2) +else +LIBS+= $(File3) +endif +SRCS=${wildcard *.c} +SRCS+=${wildcard *.cpp} +BIN=asymmetric_affine +OBJS=$(addsuffix .o, $(basename $(SRCS))) + +.SUFFIXES: .cpp .c + +.cpp.o: + $(CC) $(CFLAGS) -c $< + +.cpp: + $(CXX) $(CFLAGS) $< -o $@ -lm + +.c.o: + $(CC) $(CFLAGS) -c $< + +.c: + $(CC) $(CFLAGS) $< -o $@ -lm + +all: $(BIN) + +$(BIN): $(OBJS) + $(CC) $(CFLAGS) $(LFLAGS) $(EXTRALFLAGS) $(OBJS) $(LIBS) -o $@ + +clean: + rm -rf *.o + rm -rf $(BIN) + rm -rf *~ + +############################################################################## +# Netranslib. Supply necessary libraries. +else +include $(AQROOT)/makefile.linux.def +INCLUDE += -I$(GWG_SDK_INC) -I$(GWG_SDK_INC)/HAL -I$(AQROOT)/sdk/inc -I./ -I$(OVXLIB_DIR)/include/utils -I$(OVXLIB_DIR)/include/client -I$(OVXLIB_DIR)/include/ops -I$(OVXLIB_DIR)/include -I$(OVXLIB_DIR)/third-party/jpeg-9b +CFLAGS += $(INCLUDE) +ifeq ($(gcdSTATIC_LINK), 1) +LIBS += $(OVXLIB_DIR)/lib/libovxlib.a +LIBS += $(GWG_SDK_LIB)/libOpenVXU.a +LIBS += $(GWG_SDK_LIB)/libOpenVXC.a +LIBS += $(GWG_SDK_LIB)/libOpenVX.a +LIBS += $(GWG_SDK_LIB)/libCLC.a +LIBS += $(GWG_SDK_LIB)/libLLVM_viv.a +LIBS += $(GWG_SDK_LIB)/libclCompiler.a +LIBS += $(GWG_SDK_LIB)/libclPreprocessor.a +LIBS += $(GWG_SDK_LIB)/libclCommon.a +LIBS += $(GWG_SDK_LIB)/libLLVM_viv.a +LIBS += $(GWG_SDK_LIB)/libVSC.a +LIBS += $(GWG_SDK_LIB)/libhalarchuser.a +LIBS += $(GWG_SDK_LIB)/libhalosuser.a +LIBS += $(GWG_SDK_LIB)/libGAL.a +LIBS += $(GWG_SDK_LIB)/libhalarchuser.a +LIBS += $(GWG_SDK_LIB)/libGAL.a +LIBS += $(LIB_DIR)/libm.a +LIBS += $(LIB_DIR)/libpthread.a +LIBS += $(LIB_DIR)/libc.a +LIBS += $(LIB_DIR)/libdl.a +LIBS += $(LIB_DIR)/librt.a +LIBS += $(LIB_DIR)/libstdc++.a +LIBS += $(OVXLIB_DIR)/lib/libjpeg.a +else +ifeq ($(USE_VXC_BINARY)$(USE_VSC_LITE),11) +LIBS += -L$(GWG_SDK_LIB) -l OpenVX -l OpenVXU -l CLC -l VSC_Lite -lGAL +else +LIBS += -L$(GWG_SDK_LIB) -l OpenVX -l OpenVXU -l CLC -l VSC -lGAL +endif +LIBS += $(OVXLIB_DIR)/lib/libjpeg.a +LIBS += -L$(OVXLIB_DIR)/lib -l ovxlib +LIBS += -L$(LIB_DIR) -lm +endif + +############################################################################# +# Macros. +PROGRAM = 1 +TARGET_NAME = asymmetric_affine +CUR_SOURCE = ${wildcard *.c} +############################################################################# +# Objects. +OBJECTS = ${patsubst %.c, $(OBJ_DIR)/%.o, $(CUR_SOURCE)} + +# installation directory +INSTALL_DIR := ./ + +################################################################################ +# Include the common makefile. + +include $(AQROOT)/common.target +endif diff --git a/resource/openpose/wksp/asymmetric_affine/network_binary.nb b/resource/openpose/wksp/asymmetric_affine/network_binary.nb new file mode 100644 index 0000000..a6fac3e Binary files /dev/null and b/resource/openpose/wksp/asymmetric_affine/network_binary.nb differ diff --git a/resource/openpose/wksp/asymmetric_affine/vnn_asymmetricaffine.c b/resource/openpose/wksp/asymmetric_affine/vnn_asymmetricaffine.c new file mode 100644 index 0000000..e12f7fb --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/vnn_asymmetricaffine.c @@ -0,0 +1,5062 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction network definition source file +****************************************************************************/ +/*------------------------------------------- + Includes + -------------------------------------------*/ +#include +#include + +#include "vsi_nn_pub.h" + +#include "vnn_global.h" +#include "vnn_asymmetricaffine.h" + +/*------------------------------------------- + Macros + -------------------------------------------*/ + +#define NEW_VXNODE(_node, _type, _in, _out, _uid) do {\ + _node = vsi_nn_AddNode( graph, _type, _in, _out, NULL );\ + if( NULL == _node ) {\ + goto error;\ + }\ + _node->uid = (uint32_t)_uid;\ + } while(0) + +#define NEW_VIRTUAL_TENSOR(_id, _attr, _dtype) do {\ + memset( _attr.size, 0, VSI_NN_MAX_DIM_NUM * sizeof(vsi_size_t));\ + _attr.dim_num = VSI_NN_DIM_AUTO;\ + _attr.vtl = !VNN_APP_DEBUG;\ + _attr.is_const = FALSE;\ + _attr.dtype.vx_type = _dtype;\ + _id = vsi_nn_AddTensor( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +// Set const tensor dims out of this macro. +#define NEW_CONST_TENSOR(_id, _attr, _dtype, _ofst, _size) do {\ + data = load_data( fp, _ofst, _size );\ + if( NULL == data ) {\ + goto error;\ + }\ + _attr.vtl = FALSE;\ + _attr.is_const = TRUE;\ + _attr.dtype.vx_type = _dtype;\ + _id = vsi_nn_AddTensor( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, data );\ + free( data );\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +// Set generic tensor dims out of this macro. +#define NEW_NORM_TENSOR(_id, _attr, _dtype) do {\ + _attr.vtl = FALSE;\ + _attr.is_const = FALSE;\ + _attr.dtype.vx_type = _dtype;\ + if ( enable_from_handle )\ + {\ + _id = vsi_nn_AddTensorFromHandle( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + }\ + else\ + {\ + _id = vsi_nn_AddTensor( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + }\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +// Set generic tensor dims out of this macro. +#define NEW_NORM_TENSOR_FROM_HANDLE(_id, _attr, _dtype) do {\ + _attr.vtl = FALSE;\ + _attr.is_const = FALSE;\ + _attr.dtype.vx_type = _dtype;\ + _id = vsi_nn_AddTensorFromHandle( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +#define NET_NODE_NUM (117) +#define NET_NORM_TENSOR_NUM (5) +#define NET_CONST_TENSOR_NUM (114) +#define NET_VIRTUAL_TENSOR_NUM (117) +#define NET_TOTAL_TENSOR_NUM (NET_NORM_TENSOR_NUM + NET_CONST_TENSOR_NUM + NET_VIRTUAL_TENSOR_NUM) + +/*------------------------------------------- + Local Variables + -------------------------------------------*/ + +/*------------------------------------------- + Functions + -------------------------------------------*/ +static uint8_t* load_data + ( + FILE * fp, + size_t ofst, + size_t sz + ) +{ + uint8_t* data; + ssize_t ret; + size_t size; + data = NULL; + if( NULL == fp ) + { + return NULL; + } + + ret = VSI_FSEEK(fp, ofst, SEEK_SET); + if (ret != 0) + { + VSILOGE("blob seek failure."); + return NULL; + } + + data = (uint8_t*)malloc(sz); + if (data == NULL) + { + VSILOGE("buffer malloc failure."); + return NULL; + } + size = fread(data, 1, sz, fp); + if (size != sz || size == 0) + { + free(data); + data = NULL; + VSILOGE("Read file to buffer failed."); + } + return data; +} /* load_data() */ + +vsi_nn_graph_t * vnn_CreateAsymmetricAffine + ( + const char * data_file_name, + vsi_nn_context_t in_ctx, + const vsi_nn_preprocess_map_element_t * pre_process_map, + uint32_t pre_process_map_count, + const vsi_nn_postprocess_map_element_t * post_process_map, + uint32_t post_process_map_count + ) +{ + uint32_t _infinity = VSI_NN_FLOAT32_INF; + vsi_status status; + vsi_bool release_ctx; + vsi_nn_context_t ctx; + vsi_nn_graph_t * graph; + vsi_nn_node_t * node[NET_NODE_NUM]; + vsi_nn_tensor_id_t norm_tensor[NET_NORM_TENSOR_NUM]; + vsi_nn_tensor_id_t const_tensor[NET_CONST_TENSOR_NUM]; + vsi_nn_tensor_attr_t attr; + FILE * fp; + uint8_t * data; + uint32_t i = 0; + char * use_img_process_s; + char * use_from_handle = NULL; + int32_t enable_pre_post_process = 0; + int32_t enable_from_handle = 0; + vsi_bool sort = FALSE; + vsi_bool inference_with_nbg = FALSE; + char* pos = NULL; + + + + + + (void)(_infinity); + ctx = NULL; + graph = NULL; + status = VSI_FAILURE; + memset( &attr, 0, sizeof( attr ) ); + memset( &node, 0, sizeof( vsi_nn_node_t * ) * NET_NODE_NUM ); + + fp = fopen( data_file_name, "rb" ); + if( NULL == fp ) + { + VSILOGE( "Open file %s failed.", data_file_name ); + goto error; + } + + pos = strstr(data_file_name, ".nb"); + if( pos && strcmp(pos, ".nb") == 0 ) + { + inference_with_nbg = TRUE; + } + + if( NULL == in_ctx ) + { + ctx = vsi_nn_CreateContext(); + } + else + { + ctx = in_ctx; + } + + use_img_process_s = getenv( "VSI_USE_IMAGE_PROCESS" ); + if( use_img_process_s ) + { + enable_pre_post_process = atoi(use_img_process_s); + } + use_from_handle = getenv( "VSI_USE_FROM_HANDLE" ); + if ( use_from_handle ) + { + enable_from_handle = atoi(use_from_handle); + } + + graph = vsi_nn_CreateGraph( ctx, NET_TOTAL_TENSOR_NUM, NET_NODE_NUM ); + if( NULL == graph ) + { + VSILOGE( "Create graph fail." ); + goto error; + } + vsi_nn_SetGraphVersion( graph, VNN_VERSION_MAJOR, VNN_VERSION_MINOR, VNN_VERSION_PATCH ); + vsi_nn_SetGraphInputs( graph, NULL, 1 ); + vsi_nn_SetGraphOutputs( graph, NULL, 4 ); + vsi_nn_SetGraphFastMode(graph,FALSE); + +/*----------------------------------------- + Register client ops + -----------------------------------------*/ + + +/*----------------------------------------- + Node definitions + -----------------------------------------*/ + if( !inference_with_nbg ) + { + + /*----------------------------------------- + lid - model/model.0/model.0.0/Conv_output_0_120 + var - node[0] + name - model/model.0/model.0.0/Conv_output_0 + operation - convolution + input - [424, 256, 3, 1] + filter - [3, 3, 3, 32] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[0], VSI_NN_OP_CONV2D, 3, 1, 120); + node[0]->nn_param.conv2d.ksize[0] = 3; + node[0]->nn_param.conv2d.ksize[1] = 3; + node[0]->nn_param.conv2d.weights = 32; + node[0]->nn_param.conv2d.stride[0] = 2; + node[0]->nn_param.conv2d.stride[1] = 2; + node[0]->nn_param.conv2d.pad[0] = 1; + node[0]->nn_param.conv2d.pad[1] = 1; + node[0]->nn_param.conv2d.pad[2] = 1; + node[0]->nn_param.conv2d.pad[3] = 1; + node[0]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[0]->nn_param.conv2d.group = 1; + node[0]->nn_param.conv2d.dilation[0] = 1; + node[0]->nn_param.conv2d.dilation[1] = 1; + node[0]->nn_param.conv2d.multiplier = 0; + node[0]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[0]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[0]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.0/model.0.2/Relu_output_0_119 + var - node[1] + name - model/model.0/model.0.2/Relu_output_0 + operation - relu + input - [212, 128, 32, 1] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[1], VSI_NN_OP_RELU, 1, 1, 119); + + /*----------------------------------------- + lid - model/model.1/model.1.0/Conv_output_0_118 + var - node[2] + name - model/model.1/model.1.0/Conv_output_0 + operation - convolution + input - [212, 128, 32, 1] + filter - [3, 3, 32, 1] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[2], VSI_NN_OP_CONV2D, 3, 1, 118); + node[2]->nn_param.conv2d.ksize[0] = 3; + node[2]->nn_param.conv2d.ksize[1] = 3; + node[2]->nn_param.conv2d.weights = 32; + node[2]->nn_param.conv2d.stride[0] = 1; + node[2]->nn_param.conv2d.stride[1] = 1; + node[2]->nn_param.conv2d.pad[0] = 1; + node[2]->nn_param.conv2d.pad[1] = 1; + node[2]->nn_param.conv2d.pad[2] = 1; + node[2]->nn_param.conv2d.pad[3] = 1; + node[2]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[2]->nn_param.conv2d.group = 32; + node[2]->nn_param.conv2d.dilation[0] = 1; + node[2]->nn_param.conv2d.dilation[1] = 1; + node[2]->nn_param.conv2d.multiplier = 1; + node[2]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[2]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[2]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.1/model.1.2/Relu_output_0_117 + var - node[3] + name - model/model.1/model.1.2/Relu_output_0 + operation - relu + input - [212, 128, 32, 1] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[3], VSI_NN_OP_RELU, 1, 1, 117); + + /*----------------------------------------- + lid - model/model.1/model.1.3/Conv_output_0_116 + var - node[4] + name - model/model.1/model.1.3/Conv_output_0 + operation - convolution + input - [212, 128, 32, 1] + filter - [1, 1, 32, 64] + output - [212, 128, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[4], VSI_NN_OP_CONV2D, 3, 1, 116); + node[4]->nn_param.conv2d.ksize[0] = 1; + node[4]->nn_param.conv2d.ksize[1] = 1; + node[4]->nn_param.conv2d.weights = 64; + node[4]->nn_param.conv2d.stride[0] = 1; + node[4]->nn_param.conv2d.stride[1] = 1; + node[4]->nn_param.conv2d.pad[0] = 0; + node[4]->nn_param.conv2d.pad[1] = 0; + node[4]->nn_param.conv2d.pad[2] = 0; + node[4]->nn_param.conv2d.pad[3] = 0; + node[4]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[4]->nn_param.conv2d.group = 1; + node[4]->nn_param.conv2d.dilation[0] = 1; + node[4]->nn_param.conv2d.dilation[1] = 1; + node[4]->nn_param.conv2d.multiplier = 0; + node[4]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[4]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[4]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.1/model.1.5/Relu_output_0_115 + var - node[5] + name - model/model.1/model.1.5/Relu_output_0 + operation - relu + input - [212, 128, 64, 1] + output - [212, 128, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[5], VSI_NN_OP_RELU, 1, 1, 115); + + /*----------------------------------------- + lid - model/model.2/model.2.0/Conv_output_0_114 + var - node[6] + name - model/model.2/model.2.0/Conv_output_0 + operation - convolution + input - [212, 128, 64, 1] + filter - [3, 3, 64, 1] + output - [106, 64, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[6], VSI_NN_OP_CONV2D, 3, 1, 114); + node[6]->nn_param.conv2d.ksize[0] = 3; + node[6]->nn_param.conv2d.ksize[1] = 3; + node[6]->nn_param.conv2d.weights = 64; + node[6]->nn_param.conv2d.stride[0] = 2; + node[6]->nn_param.conv2d.stride[1] = 2; + node[6]->nn_param.conv2d.pad[0] = 1; + node[6]->nn_param.conv2d.pad[1] = 1; + node[6]->nn_param.conv2d.pad[2] = 1; + node[6]->nn_param.conv2d.pad[3] = 1; + node[6]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[6]->nn_param.conv2d.group = 64; + node[6]->nn_param.conv2d.dilation[0] = 1; + node[6]->nn_param.conv2d.dilation[1] = 1; + node[6]->nn_param.conv2d.multiplier = 1; + node[6]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[6]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[6]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.2/model.2.2/Relu_output_0_113 + var - node[7] + name - model/model.2/model.2.2/Relu_output_0 + operation - relu + input - [106, 64, 64, 1] + output - [106, 64, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[7], VSI_NN_OP_RELU, 1, 1, 113); + + /*----------------------------------------- + lid - model/model.2/model.2.3/Conv_output_0_112 + var - node[8] + name - model/model.2/model.2.3/Conv_output_0 + operation - convolution + input - [106, 64, 64, 1] + filter - [1, 1, 64, 128] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[8], VSI_NN_OP_CONV2D, 3, 1, 112); + node[8]->nn_param.conv2d.ksize[0] = 1; + node[8]->nn_param.conv2d.ksize[1] = 1; + node[8]->nn_param.conv2d.weights = 128; + node[8]->nn_param.conv2d.stride[0] = 1; + node[8]->nn_param.conv2d.stride[1] = 1; + node[8]->nn_param.conv2d.pad[0] = 0; + node[8]->nn_param.conv2d.pad[1] = 0; + node[8]->nn_param.conv2d.pad[2] = 0; + node[8]->nn_param.conv2d.pad[3] = 0; + node[8]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[8]->nn_param.conv2d.group = 1; + node[8]->nn_param.conv2d.dilation[0] = 1; + node[8]->nn_param.conv2d.dilation[1] = 1; + node[8]->nn_param.conv2d.multiplier = 0; + node[8]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[8]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[8]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.2/model.2.5/Relu_output_0_111 + var - node[9] + name - model/model.2/model.2.5/Relu_output_0 + operation - relu + input - [106, 64, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[9], VSI_NN_OP_RELU, 1, 1, 111); + + /*----------------------------------------- + lid - model/model.3/model.3.0/Conv_output_0_110 + var - node[10] + name - model/model.3/model.3.0/Conv_output_0 + operation - convolution + input - [106, 64, 128, 1] + filter - [3, 3, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[10], VSI_NN_OP_CONV2D, 3, 1, 110); + node[10]->nn_param.conv2d.ksize[0] = 3; + node[10]->nn_param.conv2d.ksize[1] = 3; + node[10]->nn_param.conv2d.weights = 128; + node[10]->nn_param.conv2d.stride[0] = 1; + node[10]->nn_param.conv2d.stride[1] = 1; + node[10]->nn_param.conv2d.pad[0] = 1; + node[10]->nn_param.conv2d.pad[1] = 1; + node[10]->nn_param.conv2d.pad[2] = 1; + node[10]->nn_param.conv2d.pad[3] = 1; + node[10]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[10]->nn_param.conv2d.group = 128; + node[10]->nn_param.conv2d.dilation[0] = 1; + node[10]->nn_param.conv2d.dilation[1] = 1; + node[10]->nn_param.conv2d.multiplier = 1; + node[10]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[10]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[10]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.3/model.3.2/Relu_output_0_109 + var - node[11] + name - model/model.3/model.3.2/Relu_output_0 + operation - relu + input - [106, 64, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[11], VSI_NN_OP_RELU, 1, 1, 109); + + /*----------------------------------------- + lid - model/model.3/model.3.3/Conv_output_0_108 + var - node[12] + name - model/model.3/model.3.3/Conv_output_0 + operation - convolution + input - [106, 64, 128, 1] + filter - [1, 1, 128, 128] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[12], VSI_NN_OP_CONV2D, 3, 1, 108); + node[12]->nn_param.conv2d.ksize[0] = 1; + node[12]->nn_param.conv2d.ksize[1] = 1; + node[12]->nn_param.conv2d.weights = 128; + node[12]->nn_param.conv2d.stride[0] = 1; + node[12]->nn_param.conv2d.stride[1] = 1; + node[12]->nn_param.conv2d.pad[0] = 0; + node[12]->nn_param.conv2d.pad[1] = 0; + node[12]->nn_param.conv2d.pad[2] = 0; + node[12]->nn_param.conv2d.pad[3] = 0; + node[12]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[12]->nn_param.conv2d.group = 1; + node[12]->nn_param.conv2d.dilation[0] = 1; + node[12]->nn_param.conv2d.dilation[1] = 1; + node[12]->nn_param.conv2d.multiplier = 0; + node[12]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[12]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[12]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.3/model.3.5/Relu_output_0_107 + var - node[13] + name - model/model.3/model.3.5/Relu_output_0 + operation - relu + input - [106, 64, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[13], VSI_NN_OP_RELU, 1, 1, 107); + + /*----------------------------------------- + lid - model/model.4/model.4.0/Conv_output_0_106 + var - node[14] + name - model/model.4/model.4.0/Conv_output_0 + operation - convolution + input - [106, 64, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[14], VSI_NN_OP_CONV2D, 3, 1, 106); + node[14]->nn_param.conv2d.ksize[0] = 3; + node[14]->nn_param.conv2d.ksize[1] = 3; + node[14]->nn_param.conv2d.weights = 128; + node[14]->nn_param.conv2d.stride[0] = 2; + node[14]->nn_param.conv2d.stride[1] = 2; + node[14]->nn_param.conv2d.pad[0] = 1; + node[14]->nn_param.conv2d.pad[1] = 1; + node[14]->nn_param.conv2d.pad[2] = 1; + node[14]->nn_param.conv2d.pad[3] = 1; + node[14]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[14]->nn_param.conv2d.group = 128; + node[14]->nn_param.conv2d.dilation[0] = 1; + node[14]->nn_param.conv2d.dilation[1] = 1; + node[14]->nn_param.conv2d.multiplier = 1; + node[14]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[14]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[14]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.4/model.4.2/Relu_output_0_105 + var - node[15] + name - model/model.4/model.4.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[15], VSI_NN_OP_RELU, 1, 1, 105); + + /*----------------------------------------- + lid - model/model.4/model.4.3/Conv_output_0_104 + var - node[16] + name - model/model.4/model.4.3/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 256] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[16], VSI_NN_OP_CONV2D, 3, 1, 104); + node[16]->nn_param.conv2d.ksize[0] = 1; + node[16]->nn_param.conv2d.ksize[1] = 1; + node[16]->nn_param.conv2d.weights = 256; + node[16]->nn_param.conv2d.stride[0] = 1; + node[16]->nn_param.conv2d.stride[1] = 1; + node[16]->nn_param.conv2d.pad[0] = 0; + node[16]->nn_param.conv2d.pad[1] = 0; + node[16]->nn_param.conv2d.pad[2] = 0; + node[16]->nn_param.conv2d.pad[3] = 0; + node[16]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[16]->nn_param.conv2d.group = 1; + node[16]->nn_param.conv2d.dilation[0] = 1; + node[16]->nn_param.conv2d.dilation[1] = 1; + node[16]->nn_param.conv2d.multiplier = 0; + node[16]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[16]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[16]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.4/model.4.5/Relu_output_0_103 + var - node[17] + name - model/model.4/model.4.5/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[17], VSI_NN_OP_RELU, 1, 1, 103); + + /*----------------------------------------- + lid - model/model.5/model.5.0/Conv_output_0_102 + var - node[18] + name - model/model.5/model.5.0/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [3, 3, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[18], VSI_NN_OP_CONV2D, 3, 1, 102); + node[18]->nn_param.conv2d.ksize[0] = 3; + node[18]->nn_param.conv2d.ksize[1] = 3; + node[18]->nn_param.conv2d.weights = 256; + node[18]->nn_param.conv2d.stride[0] = 1; + node[18]->nn_param.conv2d.stride[1] = 1; + node[18]->nn_param.conv2d.pad[0] = 1; + node[18]->nn_param.conv2d.pad[1] = 1; + node[18]->nn_param.conv2d.pad[2] = 1; + node[18]->nn_param.conv2d.pad[3] = 1; + node[18]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[18]->nn_param.conv2d.group = 256; + node[18]->nn_param.conv2d.dilation[0] = 1; + node[18]->nn_param.conv2d.dilation[1] = 1; + node[18]->nn_param.conv2d.multiplier = 1; + node[18]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[18]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[18]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.5/model.5.2/Relu_output_0_101 + var - node[19] + name - model/model.5/model.5.2/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[19], VSI_NN_OP_RELU, 1, 1, 101); + + /*----------------------------------------- + lid - model/model.5/model.5.3/Conv_output_0_100 + var - node[20] + name - model/model.5/model.5.3/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [1, 1, 256, 256] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[20], VSI_NN_OP_CONV2D, 3, 1, 100); + node[20]->nn_param.conv2d.ksize[0] = 1; + node[20]->nn_param.conv2d.ksize[1] = 1; + node[20]->nn_param.conv2d.weights = 256; + node[20]->nn_param.conv2d.stride[0] = 1; + node[20]->nn_param.conv2d.stride[1] = 1; + node[20]->nn_param.conv2d.pad[0] = 0; + node[20]->nn_param.conv2d.pad[1] = 0; + node[20]->nn_param.conv2d.pad[2] = 0; + node[20]->nn_param.conv2d.pad[3] = 0; + node[20]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[20]->nn_param.conv2d.group = 1; + node[20]->nn_param.conv2d.dilation[0] = 1; + node[20]->nn_param.conv2d.dilation[1] = 1; + node[20]->nn_param.conv2d.multiplier = 0; + node[20]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[20]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[20]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.5/model.5.5/Relu_output_0_99 + var - node[21] + name - model/model.5/model.5.5/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[21], VSI_NN_OP_RELU, 1, 1, 99); + + /*----------------------------------------- + lid - model/model.6/model.6.0/Conv_output_0_98 + var - node[22] + name - model/model.6/model.6.0/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [3, 3, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[22], VSI_NN_OP_CONV2D, 3, 1, 98); + node[22]->nn_param.conv2d.ksize[0] = 3; + node[22]->nn_param.conv2d.ksize[1] = 3; + node[22]->nn_param.conv2d.weights = 256; + node[22]->nn_param.conv2d.stride[0] = 1; + node[22]->nn_param.conv2d.stride[1] = 1; + node[22]->nn_param.conv2d.pad[0] = 1; + node[22]->nn_param.conv2d.pad[1] = 1; + node[22]->nn_param.conv2d.pad[2] = 1; + node[22]->nn_param.conv2d.pad[3] = 1; + node[22]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[22]->nn_param.conv2d.group = 256; + node[22]->nn_param.conv2d.dilation[0] = 1; + node[22]->nn_param.conv2d.dilation[1] = 1; + node[22]->nn_param.conv2d.multiplier = 1; + node[22]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[22]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[22]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.6/model.6.2/Relu_output_0_97 + var - node[23] + name - model/model.6/model.6.2/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[23], VSI_NN_OP_RELU, 1, 1, 97); + + /*----------------------------------------- + lid - model/model.6/model.6.3/Conv_output_0_96 + var - node[24] + name - model/model.6/model.6.3/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [1, 1, 256, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[24], VSI_NN_OP_CONV2D, 3, 1, 96); + node[24]->nn_param.conv2d.ksize[0] = 1; + node[24]->nn_param.conv2d.ksize[1] = 1; + node[24]->nn_param.conv2d.weights = 512; + node[24]->nn_param.conv2d.stride[0] = 1; + node[24]->nn_param.conv2d.stride[1] = 1; + node[24]->nn_param.conv2d.pad[0] = 0; + node[24]->nn_param.conv2d.pad[1] = 0; + node[24]->nn_param.conv2d.pad[2] = 0; + node[24]->nn_param.conv2d.pad[3] = 0; + node[24]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[24]->nn_param.conv2d.group = 1; + node[24]->nn_param.conv2d.dilation[0] = 1; + node[24]->nn_param.conv2d.dilation[1] = 1; + node[24]->nn_param.conv2d.multiplier = 0; + node[24]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[24]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[24]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.6/model.6.5/Relu_output_0_95 + var - node[25] + name - model/model.6/model.6.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[25], VSI_NN_OP_RELU, 1, 1, 95); + + /*----------------------------------------- + lid - model/model.7/model.7.0/Conv_output_0_94 + var - node[26] + name - model/model.7/model.7.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [5, 5, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[26], VSI_NN_OP_CONV2D, 3, 1, 94); + node[26]->nn_param.conv2d.ksize[0] = 5; + node[26]->nn_param.conv2d.ksize[1] = 5; + node[26]->nn_param.conv2d.weights = 512; + node[26]->nn_param.conv2d.stride[0] = 1; + node[26]->nn_param.conv2d.stride[1] = 1; + node[26]->nn_param.conv2d.pad[0] = 2; + node[26]->nn_param.conv2d.pad[1] = 2; + node[26]->nn_param.conv2d.pad[2] = 2; + node[26]->nn_param.conv2d.pad[3] = 2; + node[26]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[26]->nn_param.conv2d.group = 512; + node[26]->nn_param.conv2d.dilation[0] = 1; + node[26]->nn_param.conv2d.dilation[1] = 1; + node[26]->nn_param.conv2d.multiplier = 1; + node[26]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[26]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[26]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.7/model.7.2/Relu_output_0_93 + var - node[27] + name - model/model.7/model.7.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[27], VSI_NN_OP_RELU, 1, 1, 93); + + /*----------------------------------------- + lid - model/model.7/model.7.3/Conv_output_0_92 + var - node[28] + name - model/model.7/model.7.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[28], VSI_NN_OP_CONV2D, 3, 1, 92); + node[28]->nn_param.conv2d.ksize[0] = 1; + node[28]->nn_param.conv2d.ksize[1] = 1; + node[28]->nn_param.conv2d.weights = 512; + node[28]->nn_param.conv2d.stride[0] = 1; + node[28]->nn_param.conv2d.stride[1] = 1; + node[28]->nn_param.conv2d.pad[0] = 0; + node[28]->nn_param.conv2d.pad[1] = 0; + node[28]->nn_param.conv2d.pad[2] = 0; + node[28]->nn_param.conv2d.pad[3] = 0; + node[28]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[28]->nn_param.conv2d.group = 1; + node[28]->nn_param.conv2d.dilation[0] = 1; + node[28]->nn_param.conv2d.dilation[1] = 1; + node[28]->nn_param.conv2d.multiplier = 0; + node[28]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[28]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[28]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.7/model.7.5/Relu_output_0_91 + var - node[29] + name - model/model.7/model.7.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[29], VSI_NN_OP_RELU, 1, 1, 91); + + /*----------------------------------------- + lid - model/model.8/model.8.0/Conv_output_0_90 + var - node[30] + name - model/model.8/model.8.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[30], VSI_NN_OP_CONV2D, 3, 1, 90); + node[30]->nn_param.conv2d.ksize[0] = 3; + node[30]->nn_param.conv2d.ksize[1] = 3; + node[30]->nn_param.conv2d.weights = 512; + node[30]->nn_param.conv2d.stride[0] = 1; + node[30]->nn_param.conv2d.stride[1] = 1; + node[30]->nn_param.conv2d.pad[0] = 1; + node[30]->nn_param.conv2d.pad[1] = 1; + node[30]->nn_param.conv2d.pad[2] = 1; + node[30]->nn_param.conv2d.pad[3] = 1; + node[30]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[30]->nn_param.conv2d.group = 512; + node[30]->nn_param.conv2d.dilation[0] = 1; + node[30]->nn_param.conv2d.dilation[1] = 1; + node[30]->nn_param.conv2d.multiplier = 1; + node[30]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[30]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[30]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.8/model.8.2/Relu_output_0_89 + var - node[31] + name - model/model.8/model.8.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[31], VSI_NN_OP_RELU, 1, 1, 89); + + /*----------------------------------------- + lid - model/model.8/model.8.3/Conv_output_0_88 + var - node[32] + name - model/model.8/model.8.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[32], VSI_NN_OP_CONV2D, 3, 1, 88); + node[32]->nn_param.conv2d.ksize[0] = 1; + node[32]->nn_param.conv2d.ksize[1] = 1; + node[32]->nn_param.conv2d.weights = 512; + node[32]->nn_param.conv2d.stride[0] = 1; + node[32]->nn_param.conv2d.stride[1] = 1; + node[32]->nn_param.conv2d.pad[0] = 0; + node[32]->nn_param.conv2d.pad[1] = 0; + node[32]->nn_param.conv2d.pad[2] = 0; + node[32]->nn_param.conv2d.pad[3] = 0; + node[32]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[32]->nn_param.conv2d.group = 1; + node[32]->nn_param.conv2d.dilation[0] = 1; + node[32]->nn_param.conv2d.dilation[1] = 1; + node[32]->nn_param.conv2d.multiplier = 0; + node[32]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[32]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[32]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.8/model.8.5/Relu_output_0_87 + var - node[33] + name - model/model.8/model.8.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[33], VSI_NN_OP_RELU, 1, 1, 87); + + /*----------------------------------------- + lid - model/model.9/model.9.0/Conv_output_0_86 + var - node[34] + name - model/model.9/model.9.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[34], VSI_NN_OP_CONV2D, 3, 1, 86); + node[34]->nn_param.conv2d.ksize[0] = 3; + node[34]->nn_param.conv2d.ksize[1] = 3; + node[34]->nn_param.conv2d.weights = 512; + node[34]->nn_param.conv2d.stride[0] = 1; + node[34]->nn_param.conv2d.stride[1] = 1; + node[34]->nn_param.conv2d.pad[0] = 1; + node[34]->nn_param.conv2d.pad[1] = 1; + node[34]->nn_param.conv2d.pad[2] = 1; + node[34]->nn_param.conv2d.pad[3] = 1; + node[34]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[34]->nn_param.conv2d.group = 512; + node[34]->nn_param.conv2d.dilation[0] = 1; + node[34]->nn_param.conv2d.dilation[1] = 1; + node[34]->nn_param.conv2d.multiplier = 1; + node[34]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[34]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[34]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.9/model.9.2/Relu_output_0_85 + var - node[35] + name - model/model.9/model.9.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[35], VSI_NN_OP_RELU, 1, 1, 85); + + /*----------------------------------------- + lid - model/model.9/model.9.3/Conv_output_0_84 + var - node[36] + name - model/model.9/model.9.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[36], VSI_NN_OP_CONV2D, 3, 1, 84); + node[36]->nn_param.conv2d.ksize[0] = 1; + node[36]->nn_param.conv2d.ksize[1] = 1; + node[36]->nn_param.conv2d.weights = 512; + node[36]->nn_param.conv2d.stride[0] = 1; + node[36]->nn_param.conv2d.stride[1] = 1; + node[36]->nn_param.conv2d.pad[0] = 0; + node[36]->nn_param.conv2d.pad[1] = 0; + node[36]->nn_param.conv2d.pad[2] = 0; + node[36]->nn_param.conv2d.pad[3] = 0; + node[36]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[36]->nn_param.conv2d.group = 1; + node[36]->nn_param.conv2d.dilation[0] = 1; + node[36]->nn_param.conv2d.dilation[1] = 1; + node[36]->nn_param.conv2d.multiplier = 0; + node[36]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[36]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[36]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.9/model.9.5/Relu_output_0_83 + var - node[37] + name - model/model.9/model.9.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[37], VSI_NN_OP_RELU, 1, 1, 83); + + /*----------------------------------------- + lid - model/model.10/model.10.0/Conv_output_0_82 + var - node[38] + name - model/model.10/model.10.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[38], VSI_NN_OP_CONV2D, 3, 1, 82); + node[38]->nn_param.conv2d.ksize[0] = 3; + node[38]->nn_param.conv2d.ksize[1] = 3; + node[38]->nn_param.conv2d.weights = 512; + node[38]->nn_param.conv2d.stride[0] = 1; + node[38]->nn_param.conv2d.stride[1] = 1; + node[38]->nn_param.conv2d.pad[0] = 1; + node[38]->nn_param.conv2d.pad[1] = 1; + node[38]->nn_param.conv2d.pad[2] = 1; + node[38]->nn_param.conv2d.pad[3] = 1; + node[38]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[38]->nn_param.conv2d.group = 512; + node[38]->nn_param.conv2d.dilation[0] = 1; + node[38]->nn_param.conv2d.dilation[1] = 1; + node[38]->nn_param.conv2d.multiplier = 1; + node[38]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[38]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[38]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.10/model.10.2/Relu_output_0_81 + var - node[39] + name - model/model.10/model.10.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[39], VSI_NN_OP_RELU, 1, 1, 81); + + /*----------------------------------------- + lid - model/model.10/model.10.3/Conv_output_0_75 + var - node[40] + name - model/model.10/model.10.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[40], VSI_NN_OP_CONV2D, 3, 1, 75); + node[40]->nn_param.conv2d.ksize[0] = 1; + node[40]->nn_param.conv2d.ksize[1] = 1; + node[40]->nn_param.conv2d.weights = 512; + node[40]->nn_param.conv2d.stride[0] = 1; + node[40]->nn_param.conv2d.stride[1] = 1; + node[40]->nn_param.conv2d.pad[0] = 0; + node[40]->nn_param.conv2d.pad[1] = 0; + node[40]->nn_param.conv2d.pad[2] = 0; + node[40]->nn_param.conv2d.pad[3] = 0; + node[40]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[40]->nn_param.conv2d.group = 1; + node[40]->nn_param.conv2d.dilation[0] = 1; + node[40]->nn_param.conv2d.dilation[1] = 1; + node[40]->nn_param.conv2d.multiplier = 0; + node[40]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[40]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[40]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.10/model.10.5/Relu_output_0_71 + var - node[41] + name - model/model.10/model.10.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[41], VSI_NN_OP_RELU, 1, 1, 71); + + /*----------------------------------------- + lid - model/model.11/model.11.0/Conv_output_0_67 + var - node[42] + name - model/model.11/model.11.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[42], VSI_NN_OP_CONV2D, 3, 1, 67); + node[42]->nn_param.conv2d.ksize[0] = 3; + node[42]->nn_param.conv2d.ksize[1] = 3; + node[42]->nn_param.conv2d.weights = 512; + node[42]->nn_param.conv2d.stride[0] = 1; + node[42]->nn_param.conv2d.stride[1] = 1; + node[42]->nn_param.conv2d.pad[0] = 1; + node[42]->nn_param.conv2d.pad[1] = 1; + node[42]->nn_param.conv2d.pad[2] = 1; + node[42]->nn_param.conv2d.pad[3] = 1; + node[42]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[42]->nn_param.conv2d.group = 512; + node[42]->nn_param.conv2d.dilation[0] = 1; + node[42]->nn_param.conv2d.dilation[1] = 1; + node[42]->nn_param.conv2d.multiplier = 1; + node[42]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[42]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[42]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.11/model.11.2/Relu_output_0_62 + var - node[43] + name - model/model.11/model.11.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[43], VSI_NN_OP_RELU, 1, 1, 62); + + /*----------------------------------------- + lid - model/model.11/model.11.3/Conv_output_0_58 + var - node[44] + name - model/model.11/model.11.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[44], VSI_NN_OP_CONV2D, 3, 1, 58); + node[44]->nn_param.conv2d.ksize[0] = 1; + node[44]->nn_param.conv2d.ksize[1] = 1; + node[44]->nn_param.conv2d.weights = 512; + node[44]->nn_param.conv2d.stride[0] = 1; + node[44]->nn_param.conv2d.stride[1] = 1; + node[44]->nn_param.conv2d.pad[0] = 0; + node[44]->nn_param.conv2d.pad[1] = 0; + node[44]->nn_param.conv2d.pad[2] = 0; + node[44]->nn_param.conv2d.pad[3] = 0; + node[44]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[44]->nn_param.conv2d.group = 1; + node[44]->nn_param.conv2d.dilation[0] = 1; + node[44]->nn_param.conv2d.dilation[1] = 1; + node[44]->nn_param.conv2d.multiplier = 0; + node[44]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[44]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[44]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.11/model.11.5/Relu_output_0_54 + var - node[45] + name - model/model.11/model.11.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[45], VSI_NN_OP_RELU, 1, 1, 54); + + /*----------------------------------------- + lid - cpm/align/align.0/Conv_output_0_49 + var - node[46] + name - cpm/align/align.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[46], VSI_NN_OP_CONV2D, 3, 1, 49); + node[46]->nn_param.conv2d.ksize[0] = 1; + node[46]->nn_param.conv2d.ksize[1] = 1; + node[46]->nn_param.conv2d.weights = 128; + node[46]->nn_param.conv2d.stride[0] = 1; + node[46]->nn_param.conv2d.stride[1] = 1; + node[46]->nn_param.conv2d.pad[0] = 0; + node[46]->nn_param.conv2d.pad[1] = 0; + node[46]->nn_param.conv2d.pad[2] = 0; + node[46]->nn_param.conv2d.pad[3] = 0; + node[46]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[46]->nn_param.conv2d.group = 1; + node[46]->nn_param.conv2d.dilation[0] = 1; + node[46]->nn_param.conv2d.dilation[1] = 1; + node[46]->nn_param.conv2d.multiplier = 0; + node[46]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[46]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[46]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/align/align.1/Relu_output_0_45 + var - node[47] + name - cpm/align/align.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[47], VSI_NN_OP_RELU, 1, 1, 45); + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80 + var - node[48] + name - cpm/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[48], VSI_NN_OP_CONV2D, 3, 1, 80); + node[48]->nn_param.conv2d.ksize[0] = 3; + node[48]->nn_param.conv2d.ksize[1] = 3; + node[48]->nn_param.conv2d.weights = 128; + node[48]->nn_param.conv2d.stride[0] = 1; + node[48]->nn_param.conv2d.stride[1] = 1; + node[48]->nn_param.conv2d.pad[0] = 1; + node[48]->nn_param.conv2d.pad[1] = 1; + node[48]->nn_param.conv2d.pad[2] = 1; + node[48]->nn_param.conv2d.pad[3] = 1; + node[48]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[48]->nn_param.conv2d.group = 128; + node[48]->nn_param.conv2d.dilation[0] = 1; + node[48]->nn_param.conv2d.dilation[1] = 1; + node[48]->nn_param.conv2d.multiplier = 1; + node[48]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[48]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[48]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79 + var - node[49] + name - cpm/trunk/trunk.0/trunk.0.1/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[49], VSI_NN_OP_ELU, 1, 1, 79); + node[49]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78 + var - node[50] + name - cpm/trunk/trunk.0/trunk.0.2/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[50], VSI_NN_OP_CONV2D, 3, 1, 78); + node[50]->nn_param.conv2d.ksize[0] = 1; + node[50]->nn_param.conv2d.ksize[1] = 1; + node[50]->nn_param.conv2d.weights = 128; + node[50]->nn_param.conv2d.stride[0] = 1; + node[50]->nn_param.conv2d.stride[1] = 1; + node[50]->nn_param.conv2d.pad[0] = 0; + node[50]->nn_param.conv2d.pad[1] = 0; + node[50]->nn_param.conv2d.pad[2] = 0; + node[50]->nn_param.conv2d.pad[3] = 0; + node[50]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[50]->nn_param.conv2d.group = 1; + node[50]->nn_param.conv2d.dilation[0] = 1; + node[50]->nn_param.conv2d.dilation[1] = 1; + node[50]->nn_param.conv2d.multiplier = 0; + node[50]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[50]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[50]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77 + var - node[51] + name - cpm/trunk/trunk.0/trunk.0.3/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[51], VSI_NN_OP_ELU, 1, 1, 77); + node[51]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76 + var - node[52] + name - cpm/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[52], VSI_NN_OP_CONV2D, 3, 1, 76); + node[52]->nn_param.conv2d.ksize[0] = 3; + node[52]->nn_param.conv2d.ksize[1] = 3; + node[52]->nn_param.conv2d.weights = 128; + node[52]->nn_param.conv2d.stride[0] = 1; + node[52]->nn_param.conv2d.stride[1] = 1; + node[52]->nn_param.conv2d.pad[0] = 1; + node[52]->nn_param.conv2d.pad[1] = 1; + node[52]->nn_param.conv2d.pad[2] = 1; + node[52]->nn_param.conv2d.pad[3] = 1; + node[52]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[52]->nn_param.conv2d.group = 128; + node[52]->nn_param.conv2d.dilation[0] = 1; + node[52]->nn_param.conv2d.dilation[1] = 1; + node[52]->nn_param.conv2d.multiplier = 1; + node[52]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[52]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[52]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72 + var - node[53] + name - cpm/trunk/trunk.1/trunk.1.1/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[53], VSI_NN_OP_ELU, 1, 1, 72); + node[53]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68 + var - node[54] + name - cpm/trunk/trunk.1/trunk.1.2/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[54], VSI_NN_OP_CONV2D, 3, 1, 68); + node[54]->nn_param.conv2d.ksize[0] = 1; + node[54]->nn_param.conv2d.ksize[1] = 1; + node[54]->nn_param.conv2d.weights = 128; + node[54]->nn_param.conv2d.stride[0] = 1; + node[54]->nn_param.conv2d.stride[1] = 1; + node[54]->nn_param.conv2d.pad[0] = 0; + node[54]->nn_param.conv2d.pad[1] = 0; + node[54]->nn_param.conv2d.pad[2] = 0; + node[54]->nn_param.conv2d.pad[3] = 0; + node[54]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[54]->nn_param.conv2d.group = 1; + node[54]->nn_param.conv2d.dilation[0] = 1; + node[54]->nn_param.conv2d.dilation[1] = 1; + node[54]->nn_param.conv2d.multiplier = 0; + node[54]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[54]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[54]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63 + var - node[55] + name - cpm/trunk/trunk.1/trunk.1.3/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[55], VSI_NN_OP_ELU, 1, 1, 63); + node[55]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59 + var - node[56] + name - cpm/trunk/trunk.2/trunk.2.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[56], VSI_NN_OP_CONV2D, 3, 1, 59); + node[56]->nn_param.conv2d.ksize[0] = 3; + node[56]->nn_param.conv2d.ksize[1] = 3; + node[56]->nn_param.conv2d.weights = 128; + node[56]->nn_param.conv2d.stride[0] = 1; + node[56]->nn_param.conv2d.stride[1] = 1; + node[56]->nn_param.conv2d.pad[0] = 1; + node[56]->nn_param.conv2d.pad[1] = 1; + node[56]->nn_param.conv2d.pad[2] = 1; + node[56]->nn_param.conv2d.pad[3] = 1; + node[56]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[56]->nn_param.conv2d.group = 128; + node[56]->nn_param.conv2d.dilation[0] = 1; + node[56]->nn_param.conv2d.dilation[1] = 1; + node[56]->nn_param.conv2d.multiplier = 1; + node[56]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[56]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[56]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55 + var - node[57] + name - cpm/trunk/trunk.2/trunk.2.1/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[57], VSI_NN_OP_ELU, 1, 1, 55); + node[57]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50 + var - node[58] + name - cpm/trunk/trunk.2/trunk.2.2/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[58], VSI_NN_OP_CONV2D, 3, 1, 50); + node[58]->nn_param.conv2d.ksize[0] = 1; + node[58]->nn_param.conv2d.ksize[1] = 1; + node[58]->nn_param.conv2d.weights = 128; + node[58]->nn_param.conv2d.stride[0] = 1; + node[58]->nn_param.conv2d.stride[1] = 1; + node[58]->nn_param.conv2d.pad[0] = 0; + node[58]->nn_param.conv2d.pad[1] = 0; + node[58]->nn_param.conv2d.pad[2] = 0; + node[58]->nn_param.conv2d.pad[3] = 0; + node[58]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[58]->nn_param.conv2d.group = 1; + node[58]->nn_param.conv2d.dilation[0] = 1; + node[58]->nn_param.conv2d.dilation[1] = 1; + node[58]->nn_param.conv2d.multiplier = 0; + node[58]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[58]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[58]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46 + var - node[59] + name - cpm/trunk/trunk.2/trunk.2.3/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[59], VSI_NN_OP_ELU, 1, 1, 46); + node[59]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/Add_output_0_42 + var - node[60] + name - cpm/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[60], VSI_NN_OP_ADD, 2, 1, 42); + + /*----------------------------------------- + lid - cpm/conv/conv.0/Conv_output_0_38 + var - node[61] + name - cpm/conv/conv.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[61], VSI_NN_OP_CONV2D, 3, 1, 38); + node[61]->nn_param.conv2d.ksize[0] = 3; + node[61]->nn_param.conv2d.ksize[1] = 3; + node[61]->nn_param.conv2d.weights = 128; + node[61]->nn_param.conv2d.stride[0] = 1; + node[61]->nn_param.conv2d.stride[1] = 1; + node[61]->nn_param.conv2d.pad[0] = 1; + node[61]->nn_param.conv2d.pad[1] = 1; + node[61]->nn_param.conv2d.pad[2] = 1; + node[61]->nn_param.conv2d.pad[3] = 1; + node[61]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[61]->nn_param.conv2d.group = 1; + node[61]->nn_param.conv2d.dilation[0] = 1; + node[61]->nn_param.conv2d.dilation[1] = 1; + node[61]->nn_param.conv2d.multiplier = 0; + node[61]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[61]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[61]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/conv/conv.1/Relu_output_0_35 + var - node[62] + name - cpm/conv/conv.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[62], VSI_NN_OP_RELU, 1, 1, 35); + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32 + var - node[63] + name - initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[63], VSI_NN_OP_CONV2D, 3, 1, 32); + node[63]->nn_param.conv2d.ksize[0] = 3; + node[63]->nn_param.conv2d.ksize[1] = 3; + node[63]->nn_param.conv2d.weights = 128; + node[63]->nn_param.conv2d.stride[0] = 1; + node[63]->nn_param.conv2d.stride[1] = 1; + node[63]->nn_param.conv2d.pad[0] = 1; + node[63]->nn_param.conv2d.pad[1] = 1; + node[63]->nn_param.conv2d.pad[2] = 1; + node[63]->nn_param.conv2d.pad[3] = 1; + node[63]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[63]->nn_param.conv2d.group = 1; + node[63]->nn_param.conv2d.dilation[0] = 1; + node[63]->nn_param.conv2d.dilation[1] = 1; + node[63]->nn_param.conv2d.multiplier = 0; + node[63]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[63]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[63]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28 + var - node[64] + name - initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[64], VSI_NN_OP_RELU, 1, 1, 28); + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25 + var - node[65] + name - initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[65], VSI_NN_OP_CONV2D, 3, 1, 25); + node[65]->nn_param.conv2d.ksize[0] = 3; + node[65]->nn_param.conv2d.ksize[1] = 3; + node[65]->nn_param.conv2d.weights = 128; + node[65]->nn_param.conv2d.stride[0] = 1; + node[65]->nn_param.conv2d.stride[1] = 1; + node[65]->nn_param.conv2d.pad[0] = 1; + node[65]->nn_param.conv2d.pad[1] = 1; + node[65]->nn_param.conv2d.pad[2] = 1; + node[65]->nn_param.conv2d.pad[3] = 1; + node[65]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[65]->nn_param.conv2d.group = 1; + node[65]->nn_param.conv2d.dilation[0] = 1; + node[65]->nn_param.conv2d.dilation[1] = 1; + node[65]->nn_param.conv2d.multiplier = 0; + node[65]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[65]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[65]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22 + var - node[66] + name - initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[66], VSI_NN_OP_RELU, 1, 1, 22); + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19 + var - node[67] + name - initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[67], VSI_NN_OP_CONV2D, 3, 1, 19); + node[67]->nn_param.conv2d.ksize[0] = 3; + node[67]->nn_param.conv2d.ksize[1] = 3; + node[67]->nn_param.conv2d.weights = 128; + node[67]->nn_param.conv2d.stride[0] = 1; + node[67]->nn_param.conv2d.stride[1] = 1; + node[67]->nn_param.conv2d.pad[0] = 1; + node[67]->nn_param.conv2d.pad[1] = 1; + node[67]->nn_param.conv2d.pad[2] = 1; + node[67]->nn_param.conv2d.pad[3] = 1; + node[67]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[67]->nn_param.conv2d.group = 1; + node[67]->nn_param.conv2d.dilation[0] = 1; + node[67]->nn_param.conv2d.dilation[1] = 1; + node[67]->nn_param.conv2d.multiplier = 0; + node[67]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[67]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[67]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18 + var - node[68] + name - initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[68], VSI_NN_OP_RELU, 1, 1, 18); + + /*----------------------------------------- + lid - initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14 + var - node[69] + name - initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[69], VSI_NN_OP_CONV2D, 3, 1, 14); + node[69]->nn_param.conv2d.ksize[0] = 1; + node[69]->nn_param.conv2d.ksize[1] = 1; + node[69]->nn_param.conv2d.weights = 512; + node[69]->nn_param.conv2d.stride[0] = 1; + node[69]->nn_param.conv2d.stride[1] = 1; + node[69]->nn_param.conv2d.pad[0] = 0; + node[69]->nn_param.conv2d.pad[1] = 0; + node[69]->nn_param.conv2d.pad[2] = 0; + node[69]->nn_param.conv2d.pad[3] = 0; + node[69]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[69]->nn_param.conv2d.group = 1; + node[69]->nn_param.conv2d.dilation[0] = 1; + node[69]->nn_param.conv2d.dilation[1] = 1; + node[69]->nn_param.conv2d.multiplier = 0; + node[69]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[69]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[69]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15 + var - node[70] + name - initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[70], VSI_NN_OP_CONV2D, 3, 1, 15); + node[70]->nn_param.conv2d.ksize[0] = 1; + node[70]->nn_param.conv2d.ksize[1] = 1; + node[70]->nn_param.conv2d.weights = 512; + node[70]->nn_param.conv2d.stride[0] = 1; + node[70]->nn_param.conv2d.stride[1] = 1; + node[70]->nn_param.conv2d.pad[0] = 0; + node[70]->nn_param.conv2d.pad[1] = 0; + node[70]->nn_param.conv2d.pad[2] = 0; + node[70]->nn_param.conv2d.pad[3] = 0; + node[70]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[70]->nn_param.conv2d.group = 1; + node[70]->nn_param.conv2d.dilation[0] = 1; + node[70]->nn_param.conv2d.dilation[1] = 1; + node[70]->nn_param.conv2d.multiplier = 0; + node[70]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[70]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[70]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10 + var - node[71] + name - initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[71], VSI_NN_OP_RELU, 1, 1, 10); + + /*----------------------------------------- + lid - initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11 + var - node[72] + name - initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[72], VSI_NN_OP_RELU, 1, 1, 11); + + /*----------------------------------------- + lid - onnx//Concat_348_6 + var - node[73] + name - onnx//Concat_348 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 38] + output - [53, 32, 38, 1] + -----------------------------------------*/ + NEW_VXNODE(node[73], VSI_NN_OP_CONV2D, 3, 1, 6); + node[73]->nn_param.conv2d.ksize[0] = 1; + node[73]->nn_param.conv2d.ksize[1] = 1; + node[73]->nn_param.conv2d.weights = 38; + node[73]->nn_param.conv2d.stride[0] = 1; + node[73]->nn_param.conv2d.stride[1] = 1; + node[73]->nn_param.conv2d.pad[0] = 0; + node[73]->nn_param.conv2d.pad[1] = 0; + node[73]->nn_param.conv2d.pad[2] = 0; + node[73]->nn_param.conv2d.pad[3] = 0; + node[73]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[73]->nn_param.conv2d.group = 1; + node[73]->nn_param.conv2d.dilation[0] = 1; + node[73]->nn_param.conv2d.dilation[1] = 1; + node[73]->nn_param.conv2d.multiplier = 0; + node[73]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[73]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[73]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - onnx//Concat_345_7 + var - node[74] + name - onnx//Concat_345 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 19] + output - [53, 32, 19, 1] + -----------------------------------------*/ + NEW_VXNODE(node[74], VSI_NN_OP_CONV2D, 3, 1, 7); + node[74]->nn_param.conv2d.ksize[0] = 1; + node[74]->nn_param.conv2d.ksize[1] = 1; + node[74]->nn_param.conv2d.weights = 19; + node[74]->nn_param.conv2d.stride[0] = 1; + node[74]->nn_param.conv2d.stride[1] = 1; + node[74]->nn_param.conv2d.pad[0] = 0; + node[74]->nn_param.conv2d.pad[1] = 0; + node[74]->nn_param.conv2d.pad[2] = 0; + node[74]->nn_param.conv2d.pad[3] = 0; + node[74]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[74]->nn_param.conv2d.group = 1; + node[74]->nn_param.conv2d.dilation[0] = 1; + node[74]->nn_param.conv2d.dilation[1] = 1; + node[74]->nn_param.conv2d.multiplier = 0; + node[74]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[74]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[74]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - Concat_output_0_73 + var - node[75] + name - Concat_output_0 + operation - concat + input - [53, 32, 128, 1] + [53, 32, 19, 1] + [53, 32, 38, 1] + output - [53, 32, 185, 1] + -----------------------------------------*/ + NEW_VXNODE(node[75], VSI_NN_OP_CONCAT, 3, 1, 73); + node[75]->nn_param.concat.axis = 2; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69 + var - node[76] + name - refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 185, 1] + filter - [1, 1, 185, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[76], VSI_NN_OP_CONV2D, 3, 1, 69); + node[76]->nn_param.conv2d.ksize[0] = 1; + node[76]->nn_param.conv2d.ksize[1] = 1; + node[76]->nn_param.conv2d.weights = 128; + node[76]->nn_param.conv2d.stride[0] = 1; + node[76]->nn_param.conv2d.stride[1] = 1; + node[76]->nn_param.conv2d.pad[0] = 0; + node[76]->nn_param.conv2d.pad[1] = 0; + node[76]->nn_param.conv2d.pad[2] = 0; + node[76]->nn_param.conv2d.pad[3] = 0; + node[76]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[76]->nn_param.conv2d.group = 1; + node[76]->nn_param.conv2d.dilation[0] = 1; + node[76]->nn_param.conv2d.dilation[1] = 1; + node[76]->nn_param.conv2d.multiplier = 0; + node[76]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[76]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[76]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64 + var - node[77] + name - refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[77], VSI_NN_OP_RELU, 1, 1, 64); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74 + var - node[78] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[78], VSI_NN_OP_CONV2D, 3, 1, 74); + node[78]->nn_param.conv2d.ksize[0] = 3; + node[78]->nn_param.conv2d.ksize[1] = 3; + node[78]->nn_param.conv2d.weights = 128; + node[78]->nn_param.conv2d.stride[0] = 1; + node[78]->nn_param.conv2d.stride[1] = 1; + node[78]->nn_param.conv2d.pad[0] = 1; + node[78]->nn_param.conv2d.pad[1] = 1; + node[78]->nn_param.conv2d.pad[2] = 1; + node[78]->nn_param.conv2d.pad[3] = 1; + node[78]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[78]->nn_param.conv2d.group = 1; + node[78]->nn_param.conv2d.dilation[0] = 1; + node[78]->nn_param.conv2d.dilation[1] = 1; + node[78]->nn_param.conv2d.multiplier = 0; + node[78]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[78]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[78]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70 + var - node[79] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[79], VSI_NN_OP_RELU, 1, 1, 70); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66 + var - node[80] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[80], VSI_NN_OP_CONV2D, 3, 1, 66); + node[80]->nn_param.conv2d.ksize[0] = 5; + node[80]->nn_param.conv2d.ksize[1] = 5; + node[80]->nn_param.conv2d.weights = 128; + node[80]->nn_param.conv2d.stride[0] = 1; + node[80]->nn_param.conv2d.stride[1] = 1; + node[80]->nn_param.conv2d.pad[0] = 2; + node[80]->nn_param.conv2d.pad[1] = 2; + node[80]->nn_param.conv2d.pad[2] = 2; + node[80]->nn_param.conv2d.pad[3] = 2; + node[80]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[80]->nn_param.conv2d.group = 1; + node[80]->nn_param.conv2d.dilation[0] = 1; + node[80]->nn_param.conv2d.dilation[1] = 1; + node[80]->nn_param.conv2d.multiplier = 0; + node[80]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[80]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[80]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65 + var - node[81] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[81], VSI_NN_OP_RELU, 1, 1, 65); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/Add_output_0_60 + var - node[82] + name - refinement_stages.0/trunk/trunk.0/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[82], VSI_NN_OP_ADD, 2, 1, 60); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56 + var - node[83] + name - refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[83], VSI_NN_OP_CONV2D, 3, 1, 56); + node[83]->nn_param.conv2d.ksize[0] = 1; + node[83]->nn_param.conv2d.ksize[1] = 1; + node[83]->nn_param.conv2d.weights = 128; + node[83]->nn_param.conv2d.stride[0] = 1; + node[83]->nn_param.conv2d.stride[1] = 1; + node[83]->nn_param.conv2d.pad[0] = 0; + node[83]->nn_param.conv2d.pad[1] = 0; + node[83]->nn_param.conv2d.pad[2] = 0; + node[83]->nn_param.conv2d.pad[3] = 0; + node[83]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[83]->nn_param.conv2d.group = 1; + node[83]->nn_param.conv2d.dilation[0] = 1; + node[83]->nn_param.conv2d.dilation[1] = 1; + node[83]->nn_param.conv2d.multiplier = 0; + node[83]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[83]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[83]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51 + var - node[84] + name - refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[84], VSI_NN_OP_RELU, 1, 1, 51); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61 + var - node[85] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[85], VSI_NN_OP_CONV2D, 3, 1, 61); + node[85]->nn_param.conv2d.ksize[0] = 3; + node[85]->nn_param.conv2d.ksize[1] = 3; + node[85]->nn_param.conv2d.weights = 128; + node[85]->nn_param.conv2d.stride[0] = 1; + node[85]->nn_param.conv2d.stride[1] = 1; + node[85]->nn_param.conv2d.pad[0] = 1; + node[85]->nn_param.conv2d.pad[1] = 1; + node[85]->nn_param.conv2d.pad[2] = 1; + node[85]->nn_param.conv2d.pad[3] = 1; + node[85]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[85]->nn_param.conv2d.group = 1; + node[85]->nn_param.conv2d.dilation[0] = 1; + node[85]->nn_param.conv2d.dilation[1] = 1; + node[85]->nn_param.conv2d.multiplier = 0; + node[85]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[85]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[85]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57 + var - node[86] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[86], VSI_NN_OP_RELU, 1, 1, 57); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53 + var - node[87] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[87], VSI_NN_OP_CONV2D, 3, 1, 53); + node[87]->nn_param.conv2d.ksize[0] = 5; + node[87]->nn_param.conv2d.ksize[1] = 5; + node[87]->nn_param.conv2d.weights = 128; + node[87]->nn_param.conv2d.stride[0] = 1; + node[87]->nn_param.conv2d.stride[1] = 1; + node[87]->nn_param.conv2d.pad[0] = 2; + node[87]->nn_param.conv2d.pad[1] = 2; + node[87]->nn_param.conv2d.pad[2] = 2; + node[87]->nn_param.conv2d.pad[3] = 2; + node[87]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[87]->nn_param.conv2d.group = 1; + node[87]->nn_param.conv2d.dilation[0] = 1; + node[87]->nn_param.conv2d.dilation[1] = 1; + node[87]->nn_param.conv2d.multiplier = 0; + node[87]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[87]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[87]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52 + var - node[88] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[88], VSI_NN_OP_RELU, 1, 1, 52); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/Add_output_0_47 + var - node[89] + name - refinement_stages.0/trunk/trunk.1/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[89], VSI_NN_OP_ADD, 2, 1, 47); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43 + var - node[90] + name - refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[90], VSI_NN_OP_CONV2D, 3, 1, 43); + node[90]->nn_param.conv2d.ksize[0] = 1; + node[90]->nn_param.conv2d.ksize[1] = 1; + node[90]->nn_param.conv2d.weights = 128; + node[90]->nn_param.conv2d.stride[0] = 1; + node[90]->nn_param.conv2d.stride[1] = 1; + node[90]->nn_param.conv2d.pad[0] = 0; + node[90]->nn_param.conv2d.pad[1] = 0; + node[90]->nn_param.conv2d.pad[2] = 0; + node[90]->nn_param.conv2d.pad[3] = 0; + node[90]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[90]->nn_param.conv2d.group = 1; + node[90]->nn_param.conv2d.dilation[0] = 1; + node[90]->nn_param.conv2d.dilation[1] = 1; + node[90]->nn_param.conv2d.multiplier = 0; + node[90]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[90]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[90]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39 + var - node[91] + name - refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[91], VSI_NN_OP_RELU, 1, 1, 39); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48 + var - node[92] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[92], VSI_NN_OP_CONV2D, 3, 1, 48); + node[92]->nn_param.conv2d.ksize[0] = 3; + node[92]->nn_param.conv2d.ksize[1] = 3; + node[92]->nn_param.conv2d.weights = 128; + node[92]->nn_param.conv2d.stride[0] = 1; + node[92]->nn_param.conv2d.stride[1] = 1; + node[92]->nn_param.conv2d.pad[0] = 1; + node[92]->nn_param.conv2d.pad[1] = 1; + node[92]->nn_param.conv2d.pad[2] = 1; + node[92]->nn_param.conv2d.pad[3] = 1; + node[92]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[92]->nn_param.conv2d.group = 1; + node[92]->nn_param.conv2d.dilation[0] = 1; + node[92]->nn_param.conv2d.dilation[1] = 1; + node[92]->nn_param.conv2d.multiplier = 0; + node[92]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[92]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[92]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44 + var - node[93] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[93], VSI_NN_OP_RELU, 1, 1, 44); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41 + var - node[94] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[94], VSI_NN_OP_CONV2D, 3, 1, 41); + node[94]->nn_param.conv2d.ksize[0] = 5; + node[94]->nn_param.conv2d.ksize[1] = 5; + node[94]->nn_param.conv2d.weights = 128; + node[94]->nn_param.conv2d.stride[0] = 1; + node[94]->nn_param.conv2d.stride[1] = 1; + node[94]->nn_param.conv2d.pad[0] = 2; + node[94]->nn_param.conv2d.pad[1] = 2; + node[94]->nn_param.conv2d.pad[2] = 2; + node[94]->nn_param.conv2d.pad[3] = 2; + node[94]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[94]->nn_param.conv2d.group = 1; + node[94]->nn_param.conv2d.dilation[0] = 1; + node[94]->nn_param.conv2d.dilation[1] = 1; + node[94]->nn_param.conv2d.multiplier = 0; + node[94]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[94]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[94]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40 + var - node[95] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[95], VSI_NN_OP_RELU, 1, 1, 40); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/Add_output_0_36 + var - node[96] + name - refinement_stages.0/trunk/trunk.2/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[96], VSI_NN_OP_ADD, 2, 1, 36); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33 + var - node[97] + name - refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[97], VSI_NN_OP_CONV2D, 3, 1, 33); + node[97]->nn_param.conv2d.ksize[0] = 1; + node[97]->nn_param.conv2d.ksize[1] = 1; + node[97]->nn_param.conv2d.weights = 128; + node[97]->nn_param.conv2d.stride[0] = 1; + node[97]->nn_param.conv2d.stride[1] = 1; + node[97]->nn_param.conv2d.pad[0] = 0; + node[97]->nn_param.conv2d.pad[1] = 0; + node[97]->nn_param.conv2d.pad[2] = 0; + node[97]->nn_param.conv2d.pad[3] = 0; + node[97]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[97]->nn_param.conv2d.group = 1; + node[97]->nn_param.conv2d.dilation[0] = 1; + node[97]->nn_param.conv2d.dilation[1] = 1; + node[97]->nn_param.conv2d.multiplier = 0; + node[97]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[97]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[97]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29 + var - node[98] + name - refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[98], VSI_NN_OP_RELU, 1, 1, 29); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37 + var - node[99] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[99], VSI_NN_OP_CONV2D, 3, 1, 37); + node[99]->nn_param.conv2d.ksize[0] = 3; + node[99]->nn_param.conv2d.ksize[1] = 3; + node[99]->nn_param.conv2d.weights = 128; + node[99]->nn_param.conv2d.stride[0] = 1; + node[99]->nn_param.conv2d.stride[1] = 1; + node[99]->nn_param.conv2d.pad[0] = 1; + node[99]->nn_param.conv2d.pad[1] = 1; + node[99]->nn_param.conv2d.pad[2] = 1; + node[99]->nn_param.conv2d.pad[3] = 1; + node[99]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[99]->nn_param.conv2d.group = 1; + node[99]->nn_param.conv2d.dilation[0] = 1; + node[99]->nn_param.conv2d.dilation[1] = 1; + node[99]->nn_param.conv2d.multiplier = 0; + node[99]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[99]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[99]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34 + var - node[100] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[100], VSI_NN_OP_RELU, 1, 1, 34); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31 + var - node[101] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[101], VSI_NN_OP_CONV2D, 3, 1, 31); + node[101]->nn_param.conv2d.ksize[0] = 5; + node[101]->nn_param.conv2d.ksize[1] = 5; + node[101]->nn_param.conv2d.weights = 128; + node[101]->nn_param.conv2d.stride[0] = 1; + node[101]->nn_param.conv2d.stride[1] = 1; + node[101]->nn_param.conv2d.pad[0] = 2; + node[101]->nn_param.conv2d.pad[1] = 2; + node[101]->nn_param.conv2d.pad[2] = 2; + node[101]->nn_param.conv2d.pad[3] = 2; + node[101]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[101]->nn_param.conv2d.group = 1; + node[101]->nn_param.conv2d.dilation[0] = 1; + node[101]->nn_param.conv2d.dilation[1] = 1; + node[101]->nn_param.conv2d.multiplier = 0; + node[101]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[101]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[101]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30 + var - node[102] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[102], VSI_NN_OP_RELU, 1, 1, 30); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/Add_output_0_26 + var - node[103] + name - refinement_stages.0/trunk/trunk.3/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[103], VSI_NN_OP_ADD, 2, 1, 26); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23 + var - node[104] + name - refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[104], VSI_NN_OP_CONV2D, 3, 1, 23); + node[104]->nn_param.conv2d.ksize[0] = 1; + node[104]->nn_param.conv2d.ksize[1] = 1; + node[104]->nn_param.conv2d.weights = 128; + node[104]->nn_param.conv2d.stride[0] = 1; + node[104]->nn_param.conv2d.stride[1] = 1; + node[104]->nn_param.conv2d.pad[0] = 0; + node[104]->nn_param.conv2d.pad[1] = 0; + node[104]->nn_param.conv2d.pad[2] = 0; + node[104]->nn_param.conv2d.pad[3] = 0; + node[104]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[104]->nn_param.conv2d.group = 1; + node[104]->nn_param.conv2d.dilation[0] = 1; + node[104]->nn_param.conv2d.dilation[1] = 1; + node[104]->nn_param.conv2d.multiplier = 0; + node[104]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[104]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[104]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20 + var - node[105] + name - refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[105], VSI_NN_OP_RELU, 1, 1, 20); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27 + var - node[106] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[106], VSI_NN_OP_CONV2D, 3, 1, 27); + node[106]->nn_param.conv2d.ksize[0] = 3; + node[106]->nn_param.conv2d.ksize[1] = 3; + node[106]->nn_param.conv2d.weights = 128; + node[106]->nn_param.conv2d.stride[0] = 1; + node[106]->nn_param.conv2d.stride[1] = 1; + node[106]->nn_param.conv2d.pad[0] = 1; + node[106]->nn_param.conv2d.pad[1] = 1; + node[106]->nn_param.conv2d.pad[2] = 1; + node[106]->nn_param.conv2d.pad[3] = 1; + node[106]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[106]->nn_param.conv2d.group = 1; + node[106]->nn_param.conv2d.dilation[0] = 1; + node[106]->nn_param.conv2d.dilation[1] = 1; + node[106]->nn_param.conv2d.multiplier = 0; + node[106]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[106]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[106]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24 + var - node[107] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[107], VSI_NN_OP_RELU, 1, 1, 24); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21 + var - node[108] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[108], VSI_NN_OP_CONV2D, 3, 1, 21); + node[108]->nn_param.conv2d.ksize[0] = 5; + node[108]->nn_param.conv2d.ksize[1] = 5; + node[108]->nn_param.conv2d.weights = 128; + node[108]->nn_param.conv2d.stride[0] = 1; + node[108]->nn_param.conv2d.stride[1] = 1; + node[108]->nn_param.conv2d.pad[0] = 2; + node[108]->nn_param.conv2d.pad[1] = 2; + node[108]->nn_param.conv2d.pad[2] = 2; + node[108]->nn_param.conv2d.pad[3] = 2; + node[108]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[108]->nn_param.conv2d.group = 1; + node[108]->nn_param.conv2d.dilation[0] = 1; + node[108]->nn_param.conv2d.dilation[1] = 1; + node[108]->nn_param.conv2d.multiplier = 0; + node[108]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[108]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[108]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17 + var - node[109] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[109], VSI_NN_OP_RELU, 1, 1, 17); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/Add_output_0_16 + var - node[110] + name - refinement_stages.0/trunk/trunk.4/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[110], VSI_NN_OP_ADD, 2, 1, 16); + + /*----------------------------------------- + lid - refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12 + var - node[111] + name - refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[111], VSI_NN_OP_CONV2D, 3, 1, 12); + node[111]->nn_param.conv2d.ksize[0] = 1; + node[111]->nn_param.conv2d.ksize[1] = 1; + node[111]->nn_param.conv2d.weights = 128; + node[111]->nn_param.conv2d.stride[0] = 1; + node[111]->nn_param.conv2d.stride[1] = 1; + node[111]->nn_param.conv2d.pad[0] = 0; + node[111]->nn_param.conv2d.pad[1] = 0; + node[111]->nn_param.conv2d.pad[2] = 0; + node[111]->nn_param.conv2d.pad[3] = 0; + node[111]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[111]->nn_param.conv2d.group = 1; + node[111]->nn_param.conv2d.dilation[0] = 1; + node[111]->nn_param.conv2d.dilation[1] = 1; + node[111]->nn_param.conv2d.multiplier = 0; + node[111]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[111]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[111]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13 + var - node[112] + name - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[112], VSI_NN_OP_CONV2D, 3, 1, 13); + node[112]->nn_param.conv2d.ksize[0] = 1; + node[112]->nn_param.conv2d.ksize[1] = 1; + node[112]->nn_param.conv2d.weights = 128; + node[112]->nn_param.conv2d.stride[0] = 1; + node[112]->nn_param.conv2d.stride[1] = 1; + node[112]->nn_param.conv2d.pad[0] = 0; + node[112]->nn_param.conv2d.pad[1] = 0; + node[112]->nn_param.conv2d.pad[2] = 0; + node[112]->nn_param.conv2d.pad[3] = 0; + node[112]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[112]->nn_param.conv2d.group = 1; + node[112]->nn_param.conv2d.dilation[0] = 1; + node[112]->nn_param.conv2d.dilation[1] = 1; + node[112]->nn_param.conv2d.multiplier = 0; + node[112]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[112]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[112]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8 + var - node[113] + name - refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[113], VSI_NN_OP_RELU, 1, 1, 8); + + /*----------------------------------------- + lid - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9 + var - node[114] + name - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[114], VSI_NN_OP_RELU, 1, 1, 9); + + /*----------------------------------------- + lid - 400_4 + var - node[115] + name - 400 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 38] + output - [53, 32, 38, 1] + -----------------------------------------*/ + NEW_VXNODE(node[115], VSI_NN_OP_CONV2D, 3, 1, 4); + node[115]->nn_param.conv2d.ksize[0] = 1; + node[115]->nn_param.conv2d.ksize[1] = 1; + node[115]->nn_param.conv2d.weights = 38; + node[115]->nn_param.conv2d.stride[0] = 1; + node[115]->nn_param.conv2d.stride[1] = 1; + node[115]->nn_param.conv2d.pad[0] = 0; + node[115]->nn_param.conv2d.pad[1] = 0; + node[115]->nn_param.conv2d.pad[2] = 0; + node[115]->nn_param.conv2d.pad[3] = 0; + node[115]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[115]->nn_param.conv2d.group = 1; + node[115]->nn_param.conv2d.dilation[0] = 1; + node[115]->nn_param.conv2d.dilation[1] = 1; + node[115]->nn_param.conv2d.multiplier = 0; + node[115]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[115]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[115]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - 397_5 + var - node[116] + name - 397 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 19] + output - [53, 32, 19, 1] + -----------------------------------------*/ + NEW_VXNODE(node[116], VSI_NN_OP_CONV2D, 3, 1, 5); + node[116]->nn_param.conv2d.ksize[0] = 1; + node[116]->nn_param.conv2d.ksize[1] = 1; + node[116]->nn_param.conv2d.weights = 19; + node[116]->nn_param.conv2d.stride[0] = 1; + node[116]->nn_param.conv2d.stride[1] = 1; + node[116]->nn_param.conv2d.pad[0] = 0; + node[116]->nn_param.conv2d.pad[1] = 0; + node[116]->nn_param.conv2d.pad[2] = 0; + node[116]->nn_param.conv2d.pad[3] = 0; + node[116]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[116]->nn_param.conv2d.group = 1; + node[116]->nn_param.conv2d.dilation[0] = 1; + node[116]->nn_param.conv2d.dilation[1] = 1; + node[116]->nn_param.conv2d.multiplier = 0; + node[116]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[116]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[116]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + } + else + { + NEW_VXNODE(node[0], VSI_NN_OP_NBG, 1, 4, 0); + node[0]->nn_param.nbg.type = VSI_NN_NBG_FILE; + node[0]->nn_param.nbg.url = data_file_name; + + } + +/*----------------------------------------- + Tensor initialize + -----------------------------------------*/ + attr.dtype.fmt = VSI_NN_DIM_FMT_NCHW; + /* @attach_onnx//Concat_345/out0_0:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 19; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.005909114144742489; + attr.dtype.zero_point = 81; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_NORM_TENSOR(norm_tensor[0], attr, VSI_NN_TYPE_UINT8); + + /* @attach_onnx//Concat_348/out0_1:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 38; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.005909114144742489; + attr.dtype.zero_point = 81; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_NORM_TENSOR(norm_tensor[1], attr, VSI_NN_TYPE_UINT8); + + /* @attach_397/out0_2:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 19; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.004016246646642685; + attr.dtype.zero_point = 6; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_NORM_TENSOR(norm_tensor[2], attr, VSI_NN_TYPE_UINT8); + + /* @attach_400/out0_3:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 38; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.006568837445229292; + attr.dtype.zero_point = 92; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_NORM_TENSOR(norm_tensor[3], attr, VSI_NN_TYPE_UINT8); + + /* @input.1_121:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 424; + attr.size[1] = 256; + attr.size[2] = 3; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.00390625; + attr.dtype.zero_point = 128; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_NORM_TENSOR(norm_tensor[4], attr, VSI_NN_TYPE_UINT8); + + + + if( !inference_with_nbg ) + { + /* @model/model.0/model.0.0/Conv_output_0_120:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 3; + attr.size[3] = 32; + attr.dim_num = 4; + attr.dtype.scale = 0.007156998384743929; + attr.dtype.zero_point = 129; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[0], attr, VSI_NN_TYPE_UINT8, 821092, 864); + + /* @model/model.0/model.0.0/Conv_output_0_120:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 32; + attr.dim_num = 1; + attr.dtype.scale = 2.7957024940405972e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[1], attr, VSI_NN_TYPE_INT32, 820964, 128); + + /* @model/model.1/model.1.0/Conv_output_0_118:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 32; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.033282142132520676; + attr.dtype.zero_point = 99; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[2], attr, VSI_NN_TYPE_UINT8, 822084, 288); + + /* @model/model.1/model.1.0/Conv_output_0_118:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 32; + attr.dim_num = 1; + attr.dtype.scale = 0.00023954668722581118; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[3], attr, VSI_NN_TYPE_INT32, 821956, 128); + + /* @model/model.1/model.1.3/Conv_output_0_116:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 32; + attr.size[3] = 64; + attr.dim_num = 4; + attr.dtype.scale = 0.015600653365254402; + attr.dtype.zero_point = 114; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[4], attr, VSI_NN_TYPE_UINT8, 822628, 2048); + + /* @model/model.1/model.1.3/Conv_output_0_116:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 64; + attr.dim_num = 1; + attr.dtype.scale = 0.002113611437380314; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[5], attr, VSI_NN_TYPE_INT32, 822372, 256); + + /* @model/model.2/model.2.0/Conv_output_0_114:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 64; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.25059929490089417; + attr.dtype.zero_point = 128; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[6], attr, VSI_NN_TYPE_UINT8, 1366628, 576); + + /* @model/model.2/model.2.0/Conv_output_0_114:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 64; + attr.dim_num = 1; + attr.dtype.scale = 0.005779115948826075; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[7], attr, VSI_NN_TYPE_INT32, 1366372, 256); + + /* @model/model.2/model.2.3/Conv_output_0_112:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 64; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0110540259629488; + attr.dtype.zero_point = 151; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[8], attr, VSI_NN_TYPE_UINT8, 1367716, 8192); + + /* @model/model.2/model.2.3/Conv_output_0_112:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 0.0002537595573812723; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[9], attr, VSI_NN_TYPE_INT32, 1367204, 512); + + /* @model/model.3/model.3.0/Conv_output_0_110:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.061125747859478; + attr.dtype.zero_point = 123; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[10], attr, VSI_NN_TYPE_UINT8, 1376420, 1152); + + /* @model/model.3/model.3.0/Conv_output_0_110:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 0.0004793335683643818; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[11], attr, VSI_NN_TYPE_INT32, 1375908, 512); + + /* @model/model.3/model.3.3/Conv_output_0_108:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.010394223965704441; + attr.dtype.zero_point = 143; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[12], attr, VSI_NN_TYPE_UINT8, 1378084, 16384); + + /* @model/model.3/model.3.3/Conv_output_0_108:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 0.0001374005078105256; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[13], attr, VSI_NN_TYPE_INT32, 1377572, 512); + + /* @model/model.4/model.4.0/Conv_output_0_106:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.02209572121500969; + attr.dtype.zero_point = 129; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[14], attr, VSI_NN_TYPE_UINT8, 1394980, 1152); + + /* @model/model.4/model.4.0/Conv_output_0_106:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 0.00020763278007507324; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[15], attr, VSI_NN_TYPE_INT32, 1394468, 512); + + /* @model/model.4/model.4.3/Conv_output_0_104:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 256; + attr.dim_num = 4; + attr.dtype.scale = 0.006806176621466875; + attr.dtype.zero_point = 129; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[16], attr, VSI_NN_TYPE_UINT8, 1397156, 32768); + + /* @model/model.4/model.4.3/Conv_output_0_104:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.scale = 0.0007407556404359639; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[17], attr, VSI_NN_TYPE_INT32, 1396132, 1024); + + /* @model/model.5/model.5.0/Conv_output_0_102:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 256; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.07725801318883896; + attr.dtype.zero_point = 127; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[18], attr, VSI_NN_TYPE_UINT8, 1430948, 2304); + + /* @model/model.5/model.5.0/Conv_output_0_102:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.scale = 0.00043133748113177717; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[19], attr, VSI_NN_TYPE_INT32, 1429924, 1024); + + /* @model/model.5/model.5.3/Conv_output_0_100:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 256; + attr.size[3] = 256; + attr.dim_num = 4; + attr.dtype.scale = 0.00468298327177763; + attr.dtype.zero_point = 131; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[20], attr, VSI_NN_TYPE_UINT8, 1434276, 65536); + + /* @model/model.5/model.5.3/Conv_output_0_100:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.scale = 5.431609315564856e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[21], attr, VSI_NN_TYPE_INT32, 1433252, 1024); + + /* @model/model.6/model.6.0/Conv_output_0_98:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 256; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.030862757936120033; + attr.dtype.zero_point = 140; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[22], attr, VSI_NN_TYPE_UINT8, 1500836, 2304); + + /* @model/model.6/model.6.0/Conv_output_0_98:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.scale = 0.00020517698430921882; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[23], attr, VSI_NN_TYPE_INT32, 1499812, 1024); + + /* @model/model.6/model.6.3/Conv_output_0_96:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 256; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.004706871695816517; + attr.dtype.zero_point = 141; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[24], attr, VSI_NN_TYPE_UINT8, 1505188, 131072); + + /* @model/model.6/model.6.3/Conv_output_0_96:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.0004913414595648646; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[25], attr, VSI_NN_TYPE_INT32, 1503140, 2048); + + /* @model/model.7/model.7.0/Conv_output_0_94:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.04549013823270798; + attr.dtype.zero_point = 135; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[26], attr, VSI_NN_TYPE_UINT8, 1638308, 12800); + + /* @model/model.7/model.7.0/Conv_output_0_94:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.0001868777471827343; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[27], attr, VSI_NN_TYPE_INT32, 1636260, 2048); + + /* @model/model.7/model.7.3/Conv_output_0_92:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.005086179822683334; + attr.dtype.zero_point = 140; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[28], attr, VSI_NN_TYPE_UINT8, 1653156, 262144); + + /* @model/model.7/model.7.3/Conv_output_0_92:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.00010475594172021374; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[29], attr, VSI_NN_TYPE_INT32, 1651108, 2048); + + /* @model/model.8/model.8.0/Conv_output_0_90:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.05350303649902344; + attr.dtype.zero_point = 134; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[30], attr, VSI_NN_TYPE_UINT8, 1917348, 4608); + + /* @model/model.8/model.8.0/Conv_output_0_90:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.00021001300774514675; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[31], attr, VSI_NN_TYPE_INT32, 1915300, 2048); + + /* @model/model.8/model.8.3/Conv_output_0_88:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.00413436908274889; + attr.dtype.zero_point = 127; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[32], attr, VSI_NN_TYPE_UINT8, 1924004, 262144); + + /* @model/model.8/model.8.3/Conv_output_0_88:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.0002469565370120108; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[33], attr, VSI_NN_TYPE_INT32, 1921956, 2048); + + /* @model/model.9/model.9.0/Conv_output_0_86:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.04440729320049286; + attr.dtype.zero_point = 129; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[34], attr, VSI_NN_TYPE_UINT8, 2188196, 4608); + + /* @model/model.9/model.9.0/Conv_output_0_86:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.00019374178373254836; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[35], attr, VSI_NN_TYPE_INT32, 2186148, 2048); + + /* @model/model.9/model.9.3/Conv_output_0_84:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.003954111132770777; + attr.dtype.zero_point = 143; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[36], attr, VSI_NN_TYPE_UINT8, 2194852, 262144); + + /* @model/model.9/model.9.3/Conv_output_0_84:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 3.859875141642988e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[37], attr, VSI_NN_TYPE_INT32, 2192804, 2048); + + /* @model/model.10/model.10.0/Conv_output_0_82:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.04452035203576088; + attr.dtype.zero_point = 105; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[38], attr, VSI_NN_TYPE_UINT8, 826724, 4608); + + /* @model/model.10/model.10.0/Conv_output_0_82:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.00019735554815270007; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[39], attr, VSI_NN_TYPE_INT32, 824676, 2048); + + /* @model/model.10/model.10.3/Conv_output_0_75:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.004154739435762167; + attr.dtype.zero_point = 137; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[40], attr, VSI_NN_TYPE_UINT8, 833380, 262144); + + /* @model/model.10/model.10.3/Conv_output_0_75:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 3.935426866519265e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[41], attr, VSI_NN_TYPE_INT32, 831332, 2048); + + /* @model/model.11/model.11.0/Conv_output_0_67:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.05550742149353027; + attr.dtype.zero_point = 110; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[42], attr, VSI_NN_TYPE_UINT8, 1097572, 4608); + + /* @model/model.11/model.11.0/Conv_output_0_67:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.00022851143148727715; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[43], attr, VSI_NN_TYPE_INT32, 1095524, 2048); + + /* @model/model.11/model.11.3/Conv_output_0_58:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.008069222792983055; + attr.dtype.zero_point = 117; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[44], attr, VSI_NN_TYPE_UINT8, 1104228, 262144); + + /* @model/model.11/model.11.3/Conv_output_0_58:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 0.00030899656121619046; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[45], attr, VSI_NN_TYPE_INT32, 1102180, 2048); + + /* @cpm/align/align.0/Conv_output_0_49:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0029267289210110903; + attr.dtype.zero_point = 151; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[46], attr, VSI_NN_TYPE_UINT8, 8036, 65536); + + /* @cpm/align/align.0/Conv_output_0_49:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 1.2175480151199736e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[47], attr, VSI_NN_TYPE_INT32, 7524, 512); + + /* @cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.00031736117671243846; + attr.dtype.zero_point = 107; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[48], attr, VSI_NN_TYPE_UINT8, 222052, 1152); + + /* @cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 8.34300351471029e-07; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[49], attr, VSI_NN_TYPE_INT32, 221540, 512); + + /* @cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:weight + @cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:weight + @cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[50], attr, VSI_NN_TYPE_UINT8, 223204, 16384); + + /* @cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 0.000355197349563241; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[51], attr, VSI_NN_TYPE_INT32, 221540, 512); + + /* @cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.00028086593374609947; + attr.dtype.zero_point = 113; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[52], attr, VSI_NN_TYPE_UINT8, 239588, 1152); + + /* @cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 0.00028086593374609947; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[53], attr, VSI_NN_TYPE_INT32, 221540, 512); + + /* @cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:bias + @cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[54], attr, VSI_NN_TYPE_INT32, 221540, 512); + + /* @cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.scale = 0.00025757146067917347; + attr.dtype.zero_point = 126; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[55], attr, VSI_NN_TYPE_UINT8, 240740, 1152); + + /* @cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 0.00025757146067917347; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[56], attr, VSI_NN_TYPE_INT32, 221540, 512); + + /* @cpm/conv/conv.0/Conv_output_0_38:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.002213394967839122; + attr.dtype.zero_point = 117; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[57], attr, VSI_NN_TYPE_UINT8, 74084, 147456); + + /* @cpm/conv/conv.0/Conv_output_0_38:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 5.818721092509804e-06; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[58], attr, VSI_NN_TYPE_INT32, 73572, 512); + + /* @initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0032178261317312717; + attr.dtype.zero_point = 164; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[59], attr, VSI_NN_TYPE_UINT8, 377572, 147456); + + /* @initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 1.298771530855447e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[60], attr, VSI_NN_TYPE_INT32, 377060, 512); + + /* @initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.003368097823113203; + attr.dtype.zero_point = 159; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[61], attr, VSI_NN_TYPE_UINT8, 525540, 147456); + + /* @initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 8.261245966423303e-06; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[62], attr, VSI_NN_TYPE_INT32, 525028, 512); + + /* @initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0033707553520798683; + attr.dtype.zero_point = 166; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[63], attr, VSI_NN_TYPE_UINT8, 673508, 147456); + + /* @initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 1.0206378647126257e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[64], attr, VSI_NN_TYPE_INT32, 672996, 512); + + /* @initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.004766649566590786; + attr.dtype.zero_point = 148; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[65], attr, VSI_NN_TYPE_UINT8, 311524, 65536); + + /* @initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 2.4971695893327706e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[66], attr, VSI_NN_TYPE_INT32, 309476, 2048); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.scale = 0.0034540235064923763; + attr.dtype.zero_point = 147; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[67], attr, VSI_NN_TYPE_UINT8, 243940, 65536); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.scale = 1.8095062841894105e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[68], attr, VSI_NN_TYPE_INT32, 241892, 2048); + + /* @onnx//Concat_348_6:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 38; + attr.dim_num = 4; + attr.dtype.scale = 0.008257665671408176; + attr.dtype.zero_point = 136; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[69], attr, VSI_NN_TYPE_UINT8, 2466952, 19456); + + /* @onnx//Concat_348_6:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 38; + attr.dim_num = 1; + attr.dtype.scale = 1.401876397721935e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[70], attr, VSI_NN_TYPE_INT32, 2466800, 152); + + /* @onnx//Concat_345_7:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 19; + attr.dim_num = 4; + attr.dtype.scale = 0.006299979984760284; + attr.dtype.zero_point = 110; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[71], attr, VSI_NN_TYPE_UINT8, 2457072, 9728); + + /* @onnx//Concat_345_7:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 19; + attr.dim_num = 1; + attr.dtype.scale = 1.2204181984998286e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[72], attr, VSI_NN_TYPE_INT32, 2456996, 76); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 185; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.008591355755925179; + attr.dtype.zero_point = 128; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[73], attr, VSI_NN_TYPE_UINT8, 2520712, 23680); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 5.076730303699151e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[74], attr, VSI_NN_TYPE_INT32, 2520200, 512); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.00872158445417881; + attr.dtype.zero_point = 123; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[75], attr, VSI_NN_TYPE_UINT8, 2544904, 147456); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 2.469811079208739e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[76], attr, VSI_NN_TYPE_INT32, 2544392, 512); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0030493661761283875; + attr.dtype.zero_point = 122; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[77], attr, VSI_NN_TYPE_UINT8, 2692872, 409600); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 2.846899769792799e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[78], attr, VSI_NN_TYPE_INT32, 2692360, 512); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.00420405576005578; + attr.dtype.zero_point = 118; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[79], attr, VSI_NN_TYPE_UINT8, 3102984, 16384); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 3.088738230871968e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[80], attr, VSI_NN_TYPE_INT32, 3102472, 512); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0064203995279967785; + attr.dtype.zero_point = 95; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[81], attr, VSI_NN_TYPE_UINT8, 3119880, 147456); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 4.450359847396612e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[82], attr, VSI_NN_TYPE_INT32, 3119368, 512); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.00438292371109128; + attr.dtype.zero_point = 149; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[83], attr, VSI_NN_TYPE_UINT8, 3267848, 409600); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 3.485388879198581e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[84], attr, VSI_NN_TYPE_INT32, 3267336, 512); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.005807314533740282; + attr.dtype.zero_point = 171; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[85], attr, VSI_NN_TYPE_UINT8, 3677960, 16384); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 5.061101182946004e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[86], attr, VSI_NN_TYPE_INT32, 3677448, 512); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.004430193454027176; + attr.dtype.zero_point = 131; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[87], attr, VSI_NN_TYPE_UINT8, 3694856, 147456); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 2.9648092095158063e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[88], attr, VSI_NN_TYPE_INT32, 3694344, 512); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0048384880647063255; + attr.dtype.zero_point = 142; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[89], attr, VSI_NN_TYPE_UINT8, 3842824, 409600); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 3.502295658108778e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[90], attr, VSI_NN_TYPE_INT32, 3842312, 512); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.003412595484405756; + attr.dtype.zero_point = 137; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[91], attr, VSI_NN_TYPE_UINT8, 4252936, 16384); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 3.3315638575004414e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[92], attr, VSI_NN_TYPE_INT32, 4252424, 512); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.007971839047968388; + attr.dtype.zero_point = 159; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[93], attr, VSI_NN_TYPE_UINT8, 4269832, 147456); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 3.797268800553866e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[94], attr, VSI_NN_TYPE_INT32, 4269320, 512); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.005900704767554998; + attr.dtype.zero_point = 149; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[95], attr, VSI_NN_TYPE_UINT8, 4417800, 409600); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 4.4868818804388866e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[96], attr, VSI_NN_TYPE_INT32, 4417288, 512); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0038701118901371956; + attr.dtype.zero_point = 166; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[97], attr, VSI_NN_TYPE_UINT8, 4827912, 16384); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 2.9839047783752903e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[98], attr, VSI_NN_TYPE_INT32, 4827400, 512); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.00786779634654522; + attr.dtype.zero_point = 165; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[99], attr, VSI_NN_TYPE_UINT8, 4844808, 147456); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 3.4125168895116076e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[100], attr, VSI_NN_TYPE_INT32, 4844296, 512); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0081749577075243; + attr.dtype.zero_point = 149; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[101], attr, VSI_NN_TYPE_UINT8, 4992776, 409600); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 6.3744155340828e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[102], attr, VSI_NN_TYPE_INT32, 4992264, 512); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.004170569125562906; + attr.dtype.zero_point = 125; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[103], attr, VSI_NN_TYPE_UINT8, 2503816, 16384); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 8.69764553499408e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[104], attr, VSI_NN_TYPE_INT32, 2503304, 512); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.scale = 0.0030352191533893347; + attr.dtype.zero_point = 132; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[105], attr, VSI_NN_TYPE_UINT8, 2486920, 16384); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.scale = 6.329893949441612e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[106], attr, VSI_NN_TYPE_INT32, 2486408, 512); + + /* @400_4:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 38; + attr.dim_num = 4; + attr.dtype.scale = 0.004315782338380814; + attr.dtype.zero_point = 143; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[107], attr, VSI_NN_TYPE_UINT8, 2660, 4864); + + /* @400_4:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 38; + attr.dim_num = 1; + attr.dtype.scale = 3.128240496153012e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[108], attr, VSI_NN_TYPE_INT32, 2508, 152); + + /* @397_5:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 19; + attr.dim_num = 4; + attr.dtype.scale = 0.002911421936005354; + attr.dtype.zero_point = 115; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[109], attr, VSI_NN_TYPE_UINT8, 76, 2432); + + /* @397_5:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 19; + attr.dim_num = 1; + attr.dtype.scale = 2.4169421521946788e-05; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_CONST_TENSOR(const_tensor[110], attr, VSI_NN_TYPE_INT32, 0, 76); + + + + /* @model/model.0/model.0.0/Conv_output_0_120:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007197454106062651; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[0]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.0/model.0.2/Relu_output_0_119:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007197454106062651; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[1]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.1/model.1.0/Conv_output_0_118:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.13548223674297333; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[2]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.1/model.1.2/Relu_output_0_117:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.13548223674297333; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[3]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.1/model.1.3/Conv_output_0_116:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.023061182349920273; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[4]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.1/model.1.5/Relu_output_0_115:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.023061182349920273; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[5]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.2/model.2.0/Conv_output_0_114:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.022956302389502525; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[6]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.2/model.2.2/Relu_output_0_113:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.022956302389502525; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[7]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.2/model.2.3/Conv_output_0_112:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007841762155294418; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[8]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.2/model.2.5/Relu_output_0_111:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007841762155294418; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[9]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.3/model.3.0/Conv_output_0_110:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.013218929059803486; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[10]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.3/model.3.2/Relu_output_0_109:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.013218929059803486; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[11]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.3/model.3.3/Conv_output_0_108:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.009396967478096485; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[12]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.3/model.3.5/Relu_output_0_107:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.009396967478096485; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[13]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.4/model.4.0/Conv_output_0_106:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.10883579403162003; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[14]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.4/model.4.2/Relu_output_0_105:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.10883579403162003; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[15]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.4/model.4.3/Conv_output_0_104:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.005583077669143677; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[16]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.4/model.4.5/Relu_output_0_103:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.005583077669143677; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[17]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.5/model.5.0/Conv_output_0_102:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.011598609387874603; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[18]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.5/model.5.2/Relu_output_0_101:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.011598609387874603; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[19]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.5/model.5.3/Conv_output_0_100:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.006648044567555189; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[20]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.5/model.5.5/Relu_output_0_99:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.006648044567555189; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[21]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.6/model.6.0/Conv_output_0_98:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.10438811033964157; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[22]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.6/model.6.2/Relu_output_0_97:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.10438811033964157; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[23]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.6/model.6.3/Conv_output_0_96:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004108093678951263; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[24]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.6/model.6.5/Relu_output_0_95:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004108093678951263; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[25]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.7/model.7.0/Conv_output_0_94:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.02059619314968586; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[26]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.7/model.7.2/Relu_output_0_93:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.02059619314968586; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[27]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.7/model.7.3/Conv_output_0_92:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.003925254102796316; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[28]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.7/model.7.5/Relu_output_0_91:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.003925254102796316; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[29]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.8/model.8.0/Conv_output_0_90:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0597325824201107; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[30]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.8/model.8.2/Relu_output_0_89:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0597325824201107; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[31]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.8/model.8.3/Conv_output_0_88:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004362836945801973; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[32]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.8/model.8.5/Relu_output_0_87:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004362836945801973; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[33]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.9/model.9.0/Conv_output_0_86:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.00976167619228363; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[34]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.9/model.9.2/Relu_output_0_85:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.00976167619228363; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[35]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.9/model.9.3/Conv_output_0_84:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0044329287484288216; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[36]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.9/model.9.5/Relu_output_0_83:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0044329287484288216; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[37]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.10/model.10.0/Conv_output_0_82:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.009472139179706573; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[38]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.10/model.10.2/Relu_output_0_81:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.009472139179706573; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[39]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.10/model.10.3/Conv_output_0_75:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004116772674024105; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[40]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.10/model.10.5/Relu_output_0_71:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004116772674024105; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[41]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.11/model.11.0/Conv_output_0_67:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.038293223828077316; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[42]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.11/model.11.2/Relu_output_0_62:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.038293223828077316; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[43]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.11/model.11.3/Conv_output_0_58:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0041600982658565044; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[44]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @model/model.11/model.11.5/Relu_output_0_54:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0041600982658565044; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[45]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/align/align.0/Conv_output_0_49:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0026288670487701893; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[46]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/align/align.1/Relu_output_0_45:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0026288670487701893; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[47]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.00036175409331917763; + attr.dtype.zero_point = 161; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[48]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.000355197349563241; + attr.dtype.zero_point = 160; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[49]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[50]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[51]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[52]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[53]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[54]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[55]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[56]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[57]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 1.0; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[58]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_NONE; + NEW_VIRTUAL_TENSOR(node[59]->output.tensors[0], attr, VSI_NN_TYPE_FLOAT16); + + /* @cpm/Add_output_0_42:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0026288670487701893; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[60]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/conv/conv.0/Conv_output_0_38:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004036176949739456; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[61]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @cpm/conv/conv.1/Relu_output_0_35:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004036176949739456; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[62]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0024527928326278925; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[63]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0024527928326278925; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[64]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.003027920378372073; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[65]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.003027920378372073; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[66]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.005238836165517569; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[67]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.005238836165517569; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[68]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.001697666710242629; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[69]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0019371778471395373; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[70]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.001697666710242629; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[71]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0019371778471395373; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[72]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @Concat_output_0_73:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.005909114144742489; + attr.dtype.zero_point = 81; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[75]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0028318376280367374; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[76]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0028318376280367374; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[77]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.009336037561297417; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[78]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.009336037561297417; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[79]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.00706364493817091; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[80]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.00706364493817091; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[81]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.0/Add_output_0_60:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007347043603658676; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[82]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0069315931759774685; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[83]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.0069315931759774685; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[84]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007952200248837471; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[85]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007952200248837471; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[86]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.008267718367278576; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[87]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.008267718367278576; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[88]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.1/Add_output_0_47:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.008715045638382435; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[89]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.006692279130220413; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[90]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.006692279130220413; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[91]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007238409481942654; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[92]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007238409481942654; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[93]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007330504711717367; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[94]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007330504711717367; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[95]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.2/Add_output_0_36:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.009762551635503769; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[96]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004763353615999222; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[97]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004763353615999222; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[98]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.00760397594422102; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[99]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.00760397594422102; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[100]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007583816070109606; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[101]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007583816070109606; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[102]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.3/Add_output_0_26:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007710125297307968; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[103]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004337322432547808; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[104]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.004337322432547808; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[105]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007797490805387497; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[106]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007797490805387497; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[107]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.020372943952679634; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[108]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.020372943952679634; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[109]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.02085481770336628; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[110]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007248373702168465; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[111]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.008301586844027042; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[112]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.007248373702168465; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[113]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.scale = 0.008301586844027042; + attr.dtype.zero_point = 0; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_AFFINE_ASYMMETRIC; + NEW_VIRTUAL_TENSOR(node[114]->output.tensors[0], attr, VSI_NN_TYPE_UINT8); + + + +/*----------------------------------------- + Connection initialize + -----------------------------------------*/ + node[0]->input.tensors[0] = norm_tensor[4]; + node[73]->output.tensors[0] = norm_tensor[1]; + node[74]->output.tensors[0] = norm_tensor[0]; + node[115]->output.tensors[0] = norm_tensor[3]; + node[116]->output.tensors[0] = norm_tensor[2]; + + /* model/model.0/model.0.0/Conv_output_0_120 */ + node[0]->input.tensors[1] = const_tensor[0]; /* data_weight */ + node[0]->input.tensors[2] = const_tensor[1]; /* data_bias */ + + /* model/model.0/model.0.2/Relu_output_0_119 */ + node[1]->input.tensors[0] = node[0]->output.tensors[0]; + + /* model/model.1/model.1.0/Conv_output_0_118 */ + node[2]->input.tensors[0] = node[1]->output.tensors[0]; + node[2]->input.tensors[1] = const_tensor[2]; /* data_weight */ + node[2]->input.tensors[2] = const_tensor[3]; /* data_bias */ + + /* model/model.1/model.1.2/Relu_output_0_117 */ + node[3]->input.tensors[0] = node[2]->output.tensors[0]; + + /* model/model.1/model.1.3/Conv_output_0_116 */ + node[4]->input.tensors[0] = node[3]->output.tensors[0]; + node[4]->input.tensors[1] = const_tensor[4]; /* data_weight */ + node[4]->input.tensors[2] = const_tensor[5]; /* data_bias */ + + /* model/model.1/model.1.5/Relu_output_0_115 */ + node[5]->input.tensors[0] = node[4]->output.tensors[0]; + + /* model/model.2/model.2.0/Conv_output_0_114 */ + node[6]->input.tensors[0] = node[5]->output.tensors[0]; + node[6]->input.tensors[1] = const_tensor[6]; /* data_weight */ + node[6]->input.tensors[2] = const_tensor[7]; /* data_bias */ + + /* model/model.2/model.2.2/Relu_output_0_113 */ + node[7]->input.tensors[0] = node[6]->output.tensors[0]; + + /* model/model.2/model.2.3/Conv_output_0_112 */ + node[8]->input.tensors[0] = node[7]->output.tensors[0]; + node[8]->input.tensors[1] = const_tensor[8]; /* data_weight */ + node[8]->input.tensors[2] = const_tensor[9]; /* data_bias */ + + /* model/model.2/model.2.5/Relu_output_0_111 */ + node[9]->input.tensors[0] = node[8]->output.tensors[0]; + + /* model/model.3/model.3.0/Conv_output_0_110 */ + node[10]->input.tensors[0] = node[9]->output.tensors[0]; + node[10]->input.tensors[1] = const_tensor[10]; /* data_weight */ + node[10]->input.tensors[2] = const_tensor[11]; /* data_bias */ + + /* model/model.3/model.3.2/Relu_output_0_109 */ + node[11]->input.tensors[0] = node[10]->output.tensors[0]; + + /* model/model.3/model.3.3/Conv_output_0_108 */ + node[12]->input.tensors[0] = node[11]->output.tensors[0]; + node[12]->input.tensors[1] = const_tensor[12]; /* data_weight */ + node[12]->input.tensors[2] = const_tensor[13]; /* data_bias */ + + /* model/model.3/model.3.5/Relu_output_0_107 */ + node[13]->input.tensors[0] = node[12]->output.tensors[0]; + + /* model/model.4/model.4.0/Conv_output_0_106 */ + node[14]->input.tensors[0] = node[13]->output.tensors[0]; + node[14]->input.tensors[1] = const_tensor[14]; /* data_weight */ + node[14]->input.tensors[2] = const_tensor[15]; /* data_bias */ + + /* model/model.4/model.4.2/Relu_output_0_105 */ + node[15]->input.tensors[0] = node[14]->output.tensors[0]; + + /* model/model.4/model.4.3/Conv_output_0_104 */ + node[16]->input.tensors[0] = node[15]->output.tensors[0]; + node[16]->input.tensors[1] = const_tensor[16]; /* data_weight */ + node[16]->input.tensors[2] = const_tensor[17]; /* data_bias */ + + /* model/model.4/model.4.5/Relu_output_0_103 */ + node[17]->input.tensors[0] = node[16]->output.tensors[0]; + + /* model/model.5/model.5.0/Conv_output_0_102 */ + node[18]->input.tensors[0] = node[17]->output.tensors[0]; + node[18]->input.tensors[1] = const_tensor[18]; /* data_weight */ + node[18]->input.tensors[2] = const_tensor[19]; /* data_bias */ + + /* model/model.5/model.5.2/Relu_output_0_101 */ + node[19]->input.tensors[0] = node[18]->output.tensors[0]; + + /* model/model.5/model.5.3/Conv_output_0_100 */ + node[20]->input.tensors[0] = node[19]->output.tensors[0]; + node[20]->input.tensors[1] = const_tensor[20]; /* data_weight */ + node[20]->input.tensors[2] = const_tensor[21]; /* data_bias */ + + /* model/model.5/model.5.5/Relu_output_0_99 */ + node[21]->input.tensors[0] = node[20]->output.tensors[0]; + + /* model/model.6/model.6.0/Conv_output_0_98 */ + node[22]->input.tensors[0] = node[21]->output.tensors[0]; + node[22]->input.tensors[1] = const_tensor[22]; /* data_weight */ + node[22]->input.tensors[2] = const_tensor[23]; /* data_bias */ + + /* model/model.6/model.6.2/Relu_output_0_97 */ + node[23]->input.tensors[0] = node[22]->output.tensors[0]; + + /* model/model.6/model.6.3/Conv_output_0_96 */ + node[24]->input.tensors[0] = node[23]->output.tensors[0]; + node[24]->input.tensors[1] = const_tensor[24]; /* data_weight */ + node[24]->input.tensors[2] = const_tensor[25]; /* data_bias */ + + /* model/model.6/model.6.5/Relu_output_0_95 */ + node[25]->input.tensors[0] = node[24]->output.tensors[0]; + + /* model/model.7/model.7.0/Conv_output_0_94 */ + node[26]->input.tensors[0] = node[25]->output.tensors[0]; + node[26]->input.tensors[1] = const_tensor[26]; /* data_weight */ + node[26]->input.tensors[2] = const_tensor[27]; /* data_bias */ + + /* model/model.7/model.7.2/Relu_output_0_93 */ + node[27]->input.tensors[0] = node[26]->output.tensors[0]; + + /* model/model.7/model.7.3/Conv_output_0_92 */ + node[28]->input.tensors[0] = node[27]->output.tensors[0]; + node[28]->input.tensors[1] = const_tensor[28]; /* data_weight */ + node[28]->input.tensors[2] = const_tensor[29]; /* data_bias */ + + /* model/model.7/model.7.5/Relu_output_0_91 */ + node[29]->input.tensors[0] = node[28]->output.tensors[0]; + + /* model/model.8/model.8.0/Conv_output_0_90 */ + node[30]->input.tensors[0] = node[29]->output.tensors[0]; + node[30]->input.tensors[1] = const_tensor[30]; /* data_weight */ + node[30]->input.tensors[2] = const_tensor[31]; /* data_bias */ + + /* model/model.8/model.8.2/Relu_output_0_89 */ + node[31]->input.tensors[0] = node[30]->output.tensors[0]; + + /* model/model.8/model.8.3/Conv_output_0_88 */ + node[32]->input.tensors[0] = node[31]->output.tensors[0]; + node[32]->input.tensors[1] = const_tensor[32]; /* data_weight */ + node[32]->input.tensors[2] = const_tensor[33]; /* data_bias */ + + /* model/model.8/model.8.5/Relu_output_0_87 */ + node[33]->input.tensors[0] = node[32]->output.tensors[0]; + + /* model/model.9/model.9.0/Conv_output_0_86 */ + node[34]->input.tensors[0] = node[33]->output.tensors[0]; + node[34]->input.tensors[1] = const_tensor[34]; /* data_weight */ + node[34]->input.tensors[2] = const_tensor[35]; /* data_bias */ + + /* model/model.9/model.9.2/Relu_output_0_85 */ + node[35]->input.tensors[0] = node[34]->output.tensors[0]; + + /* model/model.9/model.9.3/Conv_output_0_84 */ + node[36]->input.tensors[0] = node[35]->output.tensors[0]; + node[36]->input.tensors[1] = const_tensor[36]; /* data_weight */ + node[36]->input.tensors[2] = const_tensor[37]; /* data_bias */ + + /* model/model.9/model.9.5/Relu_output_0_83 */ + node[37]->input.tensors[0] = node[36]->output.tensors[0]; + + /* model/model.10/model.10.0/Conv_output_0_82 */ + node[38]->input.tensors[0] = node[37]->output.tensors[0]; + node[38]->input.tensors[1] = const_tensor[38]; /* data_weight */ + node[38]->input.tensors[2] = const_tensor[39]; /* data_bias */ + + /* model/model.10/model.10.2/Relu_output_0_81 */ + node[39]->input.tensors[0] = node[38]->output.tensors[0]; + + /* model/model.10/model.10.3/Conv_output_0_75 */ + node[40]->input.tensors[0] = node[39]->output.tensors[0]; + node[40]->input.tensors[1] = const_tensor[40]; /* data_weight */ + node[40]->input.tensors[2] = const_tensor[41]; /* data_bias */ + + /* model/model.10/model.10.5/Relu_output_0_71 */ + node[41]->input.tensors[0] = node[40]->output.tensors[0]; + + /* model/model.11/model.11.0/Conv_output_0_67 */ + node[42]->input.tensors[0] = node[41]->output.tensors[0]; + node[42]->input.tensors[1] = const_tensor[42]; /* data_weight */ + node[42]->input.tensors[2] = const_tensor[43]; /* data_bias */ + + /* model/model.11/model.11.2/Relu_output_0_62 */ + node[43]->input.tensors[0] = node[42]->output.tensors[0]; + + /* model/model.11/model.11.3/Conv_output_0_58 */ + node[44]->input.tensors[0] = node[43]->output.tensors[0]; + node[44]->input.tensors[1] = const_tensor[44]; /* data_weight */ + node[44]->input.tensors[2] = const_tensor[45]; /* data_bias */ + + /* model/model.11/model.11.5/Relu_output_0_54 */ + node[45]->input.tensors[0] = node[44]->output.tensors[0]; + + /* cpm/align/align.0/Conv_output_0_49 */ + node[46]->input.tensors[0] = node[45]->output.tensors[0]; + node[46]->input.tensors[1] = const_tensor[46]; /* data_weight */ + node[46]->input.tensors[2] = const_tensor[47]; /* data_bias */ + + /* cpm/align/align.1/Relu_output_0_45 */ + node[47]->input.tensors[0] = node[46]->output.tensors[0]; + + /* cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80 */ + node[48]->input.tensors[0] = node[47]->output.tensors[0]; + node[48]->input.tensors[1] = const_tensor[48]; /* data_weight */ + node[48]->input.tensors[2] = const_tensor[49]; /* data_bias */ + + /* cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79 */ + node[49]->input.tensors[0] = node[48]->output.tensors[0]; + + /* cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78 */ + node[50]->input.tensors[0] = node[49]->output.tensors[0]; + node[50]->input.tensors[1] = const_tensor[50]; /* data_weight */ + node[50]->input.tensors[2] = const_tensor[51]; /* data_bias */ + + /* cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77 */ + node[51]->input.tensors[0] = node[50]->output.tensors[0]; + + /* cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76 */ + node[52]->input.tensors[0] = node[51]->output.tensors[0]; + node[52]->input.tensors[1] = const_tensor[52]; /* data_weight */ + node[52]->input.tensors[2] = const_tensor[53]; /* data_bias */ + + /* cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72 */ + node[53]->input.tensors[0] = node[52]->output.tensors[0]; + + /* cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68 */ + node[54]->input.tensors[0] = node[53]->output.tensors[0]; + node[54]->input.tensors[1] = const_tensor[50]; /* data_weight */ + node[54]->input.tensors[2] = const_tensor[54]; /* data_bias */ + + /* cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63 */ + node[55]->input.tensors[0] = node[54]->output.tensors[0]; + + /* cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59 */ + node[56]->input.tensors[0] = node[55]->output.tensors[0]; + node[56]->input.tensors[1] = const_tensor[55]; /* data_weight */ + node[56]->input.tensors[2] = const_tensor[56]; /* data_bias */ + + /* cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55 */ + node[57]->input.tensors[0] = node[56]->output.tensors[0]; + + /* cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50 */ + node[58]->input.tensors[0] = node[57]->output.tensors[0]; + node[58]->input.tensors[1] = const_tensor[50]; /* data_weight */ + node[58]->input.tensors[2] = const_tensor[54]; /* data_bias */ + + /* cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46 */ + node[59]->input.tensors[0] = node[58]->output.tensors[0]; + + /* cpm/Add_output_0_42 */ + node[60]->input.tensors[0] = node[47]->output.tensors[0]; + node[60]->input.tensors[1] = node[59]->output.tensors[0]; + + /* cpm/conv/conv.0/Conv_output_0_38 */ + node[61]->input.tensors[0] = node[60]->output.tensors[0]; + node[61]->input.tensors[1] = const_tensor[57]; /* data_weight */ + node[61]->input.tensors[2] = const_tensor[58]; /* data_bias */ + + /* cpm/conv/conv.1/Relu_output_0_35 */ + node[62]->input.tensors[0] = node[61]->output.tensors[0]; + + /* initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32 */ + node[63]->input.tensors[0] = node[62]->output.tensors[0]; + node[63]->input.tensors[1] = const_tensor[59]; /* data_weight */ + node[63]->input.tensors[2] = const_tensor[60]; /* data_bias */ + + /* initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28 */ + node[64]->input.tensors[0] = node[63]->output.tensors[0]; + + /* initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25 */ + node[65]->input.tensors[0] = node[64]->output.tensors[0]; + node[65]->input.tensors[1] = const_tensor[61]; /* data_weight */ + node[65]->input.tensors[2] = const_tensor[62]; /* data_bias */ + + /* initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22 */ + node[66]->input.tensors[0] = node[65]->output.tensors[0]; + + /* initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19 */ + node[67]->input.tensors[0] = node[66]->output.tensors[0]; + node[67]->input.tensors[1] = const_tensor[63]; /* data_weight */ + node[67]->input.tensors[2] = const_tensor[64]; /* data_bias */ + + /* initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18 */ + node[68]->input.tensors[0] = node[67]->output.tensors[0]; + + /* initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14 */ + node[69]->input.tensors[0] = node[68]->output.tensors[0]; + node[69]->input.tensors[1] = const_tensor[65]; /* data_weight */ + node[69]->input.tensors[2] = const_tensor[66]; /* data_bias */ + + /* initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15 */ + node[70]->input.tensors[0] = node[68]->output.tensors[0]; + node[70]->input.tensors[1] = const_tensor[67]; /* data_weight */ + node[70]->input.tensors[2] = const_tensor[68]; /* data_bias */ + + /* initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10 */ + node[71]->input.tensors[0] = node[69]->output.tensors[0]; + + /* initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11 */ + node[72]->input.tensors[0] = node[70]->output.tensors[0]; + + /* onnx//Concat_348_6 */ + node[73]->input.tensors[0] = node[71]->output.tensors[0]; + node[73]->input.tensors[1] = const_tensor[69]; /* data_weight */ + node[73]->input.tensors[2] = const_tensor[70]; /* data_bias */ + + /* onnx//Concat_345_7 */ + node[74]->input.tensors[0] = node[72]->output.tensors[0]; + node[74]->input.tensors[1] = const_tensor[71]; /* data_weight */ + node[74]->input.tensors[2] = const_tensor[72]; /* data_bias */ + + /* Concat_output_0_73 */ + node[75]->input.tensors[0] = node[62]->output.tensors[0]; + node[75]->input.tensors[1] = node[74]->output.tensors[0]; + node[75]->input.tensors[2] = node[73]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69 */ + node[76]->input.tensors[0] = node[75]->output.tensors[0]; + node[76]->input.tensors[1] = const_tensor[73]; /* data_weight */ + node[76]->input.tensors[2] = const_tensor[74]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64 */ + node[77]->input.tensors[0] = node[76]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74 */ + node[78]->input.tensors[0] = node[77]->output.tensors[0]; + node[78]->input.tensors[1] = const_tensor[75]; /* data_weight */ + node[78]->input.tensors[2] = const_tensor[76]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70 */ + node[79]->input.tensors[0] = node[78]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66 */ + node[80]->input.tensors[0] = node[79]->output.tensors[0]; + node[80]->input.tensors[1] = const_tensor[77]; /* data_weight */ + node[80]->input.tensors[2] = const_tensor[78]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65 */ + node[81]->input.tensors[0] = node[80]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/Add_output_0_60 */ + node[82]->input.tensors[0] = node[77]->output.tensors[0]; + node[82]->input.tensors[1] = node[81]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56 */ + node[83]->input.tensors[0] = node[82]->output.tensors[0]; + node[83]->input.tensors[1] = const_tensor[79]; /* data_weight */ + node[83]->input.tensors[2] = const_tensor[80]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51 */ + node[84]->input.tensors[0] = node[83]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61 */ + node[85]->input.tensors[0] = node[84]->output.tensors[0]; + node[85]->input.tensors[1] = const_tensor[81]; /* data_weight */ + node[85]->input.tensors[2] = const_tensor[82]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57 */ + node[86]->input.tensors[0] = node[85]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53 */ + node[87]->input.tensors[0] = node[86]->output.tensors[0]; + node[87]->input.tensors[1] = const_tensor[83]; /* data_weight */ + node[87]->input.tensors[2] = const_tensor[84]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52 */ + node[88]->input.tensors[0] = node[87]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/Add_output_0_47 */ + node[89]->input.tensors[0] = node[84]->output.tensors[0]; + node[89]->input.tensors[1] = node[88]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43 */ + node[90]->input.tensors[0] = node[89]->output.tensors[0]; + node[90]->input.tensors[1] = const_tensor[85]; /* data_weight */ + node[90]->input.tensors[2] = const_tensor[86]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39 */ + node[91]->input.tensors[0] = node[90]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48 */ + node[92]->input.tensors[0] = node[91]->output.tensors[0]; + node[92]->input.tensors[1] = const_tensor[87]; /* data_weight */ + node[92]->input.tensors[2] = const_tensor[88]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44 */ + node[93]->input.tensors[0] = node[92]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41 */ + node[94]->input.tensors[0] = node[93]->output.tensors[0]; + node[94]->input.tensors[1] = const_tensor[89]; /* data_weight */ + node[94]->input.tensors[2] = const_tensor[90]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40 */ + node[95]->input.tensors[0] = node[94]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/Add_output_0_36 */ + node[96]->input.tensors[0] = node[91]->output.tensors[0]; + node[96]->input.tensors[1] = node[95]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33 */ + node[97]->input.tensors[0] = node[96]->output.tensors[0]; + node[97]->input.tensors[1] = const_tensor[91]; /* data_weight */ + node[97]->input.tensors[2] = const_tensor[92]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29 */ + node[98]->input.tensors[0] = node[97]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37 */ + node[99]->input.tensors[0] = node[98]->output.tensors[0]; + node[99]->input.tensors[1] = const_tensor[93]; /* data_weight */ + node[99]->input.tensors[2] = const_tensor[94]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34 */ + node[100]->input.tensors[0] = node[99]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31 */ + node[101]->input.tensors[0] = node[100]->output.tensors[0]; + node[101]->input.tensors[1] = const_tensor[95]; /* data_weight */ + node[101]->input.tensors[2] = const_tensor[96]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30 */ + node[102]->input.tensors[0] = node[101]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/Add_output_0_26 */ + node[103]->input.tensors[0] = node[98]->output.tensors[0]; + node[103]->input.tensors[1] = node[102]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23 */ + node[104]->input.tensors[0] = node[103]->output.tensors[0]; + node[104]->input.tensors[1] = const_tensor[97]; /* data_weight */ + node[104]->input.tensors[2] = const_tensor[98]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20 */ + node[105]->input.tensors[0] = node[104]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27 */ + node[106]->input.tensors[0] = node[105]->output.tensors[0]; + node[106]->input.tensors[1] = const_tensor[99]; /* data_weight */ + node[106]->input.tensors[2] = const_tensor[100]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24 */ + node[107]->input.tensors[0] = node[106]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21 */ + node[108]->input.tensors[0] = node[107]->output.tensors[0]; + node[108]->input.tensors[1] = const_tensor[101]; /* data_weight */ + node[108]->input.tensors[2] = const_tensor[102]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17 */ + node[109]->input.tensors[0] = node[108]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/Add_output_0_16 */ + node[110]->input.tensors[0] = node[105]->output.tensors[0]; + node[110]->input.tensors[1] = node[109]->output.tensors[0]; + + /* refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12 */ + node[111]->input.tensors[0] = node[110]->output.tensors[0]; + node[111]->input.tensors[1] = const_tensor[103]; /* data_weight */ + node[111]->input.tensors[2] = const_tensor[104]; /* data_bias */ + + /* refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13 */ + node[112]->input.tensors[0] = node[110]->output.tensors[0]; + node[112]->input.tensors[1] = const_tensor[105]; /* data_weight */ + node[112]->input.tensors[2] = const_tensor[106]; /* data_bias */ + + /* refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8 */ + node[113]->input.tensors[0] = node[111]->output.tensors[0]; + + /* refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9 */ + node[114]->input.tensors[0] = node[112]->output.tensors[0]; + + /* 400_4 */ + node[115]->input.tensors[0] = node[113]->output.tensors[0]; + node[115]->input.tensors[1] = const_tensor[107]; /* data_weight */ + node[115]->input.tensors[2] = const_tensor[108]; /* data_bias */ + + /* 397_5 */ + node[116]->input.tensors[0] = node[114]->output.tensors[0]; + node[116]->input.tensors[1] = const_tensor[109]; /* data_weight */ + node[116]->input.tensors[2] = const_tensor[110]; /* data_bias */ + + + } + else + { + node[0]->output.tensors[0] = norm_tensor[0]; + node[0]->output.tensors[1] = norm_tensor[1]; + node[0]->output.tensors[2] = norm_tensor[2]; + node[0]->output.tensors[3] = norm_tensor[3]; + node[0]->input.tensors[0] = norm_tensor[4]; + + } + graph->output.tensors[0] = norm_tensor[0]; + graph->output.tensors[1] = norm_tensor[1]; + graph->output.tensors[2] = norm_tensor[2]; + graph->output.tensors[3] = norm_tensor[3]; + graph->input.tensors[0] = norm_tensor[4]; + + + if( enable_pre_post_process ) + { + sort = TRUE; + if( pre_process_map_count > 0 ) + { + for( i = 0; i < pre_process_map_count; i++ ) + { + status = vsi_nn_AddGraphPreProcess(graph, pre_process_map[i].graph_input_idx, + pre_process_map[i].preprocesses, + pre_process_map[i].preprocess_count); + TEST_CHECK_STATUS( status, error ); + } + } + + if( post_process_map_count > 0 ) + { + for( i = 0; i < post_process_map_count; i++ ) + { + status = vsi_nn_AddGraphPostProcess(graph, post_process_map[i].graph_output_idx, + post_process_map[i].postprocesses, + post_process_map[i].postprocess_count); + TEST_CHECK_STATUS( status, error ); + } + } + } + + status = vsi_nn_SetupGraph( graph, sort ); + TEST_CHECK_STATUS( status, error ); + vsi_nn_DumpGraphToJson( graph ); + + if( VSI_FAILURE == status ) + { + goto error; + } + + fclose( fp ); + + return graph; + +error: + if( NULL != fp ) + { + fclose( fp ); + } + + release_ctx = ( NULL == in_ctx ); + vsi_nn_DumpGraphToJson( graph ); + vnn_ReleaseAsymmetricAffine( graph, release_ctx ); + + return NULL; +} /* vsi_nn_CreateAsymmetricAffine() */ + +void vnn_ReleaseAsymmetricAffine + ( + vsi_nn_graph_t * graph, + vsi_bool release_ctx + ) +{ + vsi_nn_context_t ctx; + if( NULL != graph ) + { + ctx = graph->ctx; + vsi_nn_ReleaseGraph( &graph ); + + /*----------------------------------------- + Unregister client ops + -----------------------------------------*/ + + + if( release_ctx ) + { + vsi_nn_ReleaseContext( &ctx ); + } + } +} /* vsi_nn_ReleaseAsymmetricAffine() */ + diff --git a/resource/openpose/wksp/asymmetric_affine/vnn_asymmetricaffine.h b/resource/openpose/wksp/asymmetric_affine/vnn_asymmetricaffine.h new file mode 100644 index 0000000..daae091 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/vnn_asymmetricaffine.h @@ -0,0 +1,39 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction network definition header file +****************************************************************************/ + +#ifndef _VNN_ASYMMETRICAFFINE_H +#define _VNN_ASYMMETRICAFFINE_H + +#include "vsi_nn_pub.h" + +#define VNN_APP_DEBUG (FALSE) +#define VNN_VERSION_MAJOR 1 +#define VNN_VERSION_MINOR 1 +#define VNN_VERSION_PATCH 53 +#define VNN_RUNTIME_VERSION \ + (VNN_VERSION_MAJOR * 10000 + VNN_VERSION_MINOR * 100 + VNN_VERSION_PATCH) + +_version_assert(VNN_RUNTIME_VERSION <= VSI_NN_VERSION, + CASE_VERSION_is_higher_than_OVXLIB_VERSION) + +void vnn_ReleaseAsymmetricAffine + ( + vsi_nn_graph_t * graph, + vsi_bool release_ctx + ); + +vsi_nn_graph_t * vnn_CreateAsymmetricAffine + ( + const char * data_file_name, + vsi_nn_context_t in_ctx, + const vsi_nn_preprocess_map_element_t * pre_process_map, + uint32_t pre_process_map_count, + const vsi_nn_postprocess_map_element_t * post_process_map, + uint32_t post_process_map_count + ); + +#endif diff --git a/resource/openpose/wksp/asymmetric_affine/vnn_global.h b/resource/openpose/wksp/asymmetric_affine/vnn_global.h new file mode 100644 index 0000000..e8b3704 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/vnn_global.h @@ -0,0 +1,38 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network global header file +****************************************************************************/ +#ifndef _VNN_GLOBAL_H_ +#define _VNN_GLOBAL_H_ + +typedef struct { + uint32_t graph_input_idx; + vsi_nn_preprocess_base_t *preprocesses; + uint32_t preprocess_count; +} vsi_nn_preprocess_map_element_t; + + +typedef struct { + uint32_t graph_output_idx; + vsi_nn_postprocess_base_t *postprocesses; + uint32_t postprocess_count; +} vsi_nn_postprocess_map_element_t; + +#ifndef VSI_SIZE_T +typedef uint32_t vsi_size_t; +typedef int32_t vsi_ssize_t; +#endif + +#ifdef _WIN32 +#define VSI_FSEEK _fseeki64 +#else +#define VSI_FSEEK fseek +#endif + +/* + * This file will be deprecated in the future + */ + +#endif \ No newline at end of file diff --git a/resource/openpose/wksp/asymmetric_affine/vnn_post_process.c b/resource/openpose/wksp/asymmetric_affine/vnn_post_process.c new file mode 100644 index 0000000..685da64 --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/vnn_post_process.c @@ -0,0 +1,173 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction post-process source file +****************************************************************************/ +/*------------------------------------------- + Includes +-------------------------------------------*/ +#include +#include +#include + +#include "vsi_nn_pub.h" + +#include "vnn_global.h" +#include "vnn_post_process.h" + +#define _BASETSD_H + +/*------------------------------------------- + Variable definitions +-------------------------------------------*/ + +/*{graph_output_idx, postprocess}*/ +const static vsi_nn_postprocess_map_element_t* postprocess_map = NULL; + + +/*------------------------------------------- + Functions +-------------------------------------------*/ +static void save_output_data(vsi_nn_graph_t *graph) +{ + uint32_t i; +#define _DUMP_FILE_LENGTH 1028 +#define _DUMP_SHAPE_LENGTH 128 + char filename[_DUMP_FILE_LENGTH] = {0}, shape[_DUMP_SHAPE_LENGTH] = {0}; + vsi_nn_tensor_t *tensor; + + for(i = 0; i < graph->output.num; i++) + { + tensor = vsi_nn_GetTensor(graph, graph->output.tensors[i]); + vsi_nn_ShapeToString( tensor->attr.size, tensor->attr.dim_num, + shape, _DUMP_SHAPE_LENGTH, FALSE ); + snprintf(filename, _DUMP_FILE_LENGTH, "output%u_%s.dat", i, shape); + vsi_nn_SaveTensorToBinary(graph, tensor, filename); + + } +} + +static vsi_bool get_top + ( + float *pfProb, + float *pfMaxProb, + vsi_size_t *pMaxClass, + vsi_size_t outputCount, + vsi_size_t topNum + ) +{ + vsi_size_t i, j, k; + + #define MAX_TOP_NUM 20 + if (topNum > MAX_TOP_NUM) return FALSE; + + memset(pfMaxProb, 0xfe, sizeof(float) * topNum); + memset(pMaxClass, 0xff, sizeof(vsi_size_t) * topNum); + + for (j = 0; j < topNum; j++) + { + for (i=0; i *(pfMaxProb+j)) + { + *(pfMaxProb+j) = pfProb[i]; + *(pMaxClass+j) = i; + } + } + } + + return TRUE; +} + +static vsi_status show_top5 + ( + vsi_nn_graph_t *graph, + vsi_nn_tensor_t *tensor + ) +{ + vsi_status status = VSI_FAILURE; + vsi_size_t i,sz,stride; + float *buffer = NULL; + uint8_t *tensor_data = NULL; + vsi_size_t MaxClass[5]; + float fMaxProb[5]; + vsi_size_t topk = 5; + + sz = 1; + for(i = 0; i < tensor->attr.dim_num; i++) + { + sz *= tensor->attr.size[i]; + } + + if(topk > sz) + topk = sz; + + stride = (vsi_size_t)vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + tensor_data = (uint8_t *)vsi_nn_ConvertTensorToData(graph, tensor); + buffer = (float *)malloc(sizeof(float) * sz); + + for(i = 0; i < sz; i++) + { + status = vsi_nn_DtypeToFloat32(&tensor_data[stride * i], &buffer[i], &tensor->attr.dtype); + } + + if (!get_top(buffer, fMaxProb, MaxClass, sz, topk)) + { + printf("Fail to show result.\n"); + goto final; + } + + printf(" --- Top%d ---\n", topk); + for(i = 0; i< topk; i++) + { + printf("%3d: %8.6f\n", MaxClass[i], fMaxProb[i]); + } + status = VSI_SUCCESS; + +final: + if(tensor_data)vsi_nn_Free(tensor_data); + if(buffer)free(buffer); + return status; +} + +vsi_status vnn_PostProcessAsymmetricAffine(vsi_nn_graph_t *graph) +{ + vsi_status status = VSI_FAILURE; + + /* Show the top5 result */ + status = show_top5(graph, vsi_nn_GetTensor(graph, graph->output.tensors[0])); + TEST_CHECK_STATUS(status, final); + + /* Save all output tensor data to txt file */ + save_output_data(graph); + +final: + return VSI_SUCCESS; +} + +const vsi_nn_postprocess_map_element_t * vnn_GetPostProcessMap() +{ + return postprocess_map; +} + +uint32_t vnn_GetPostProcessMapCount() +{ + if (postprocess_map == NULL) + return 0; + else + return sizeof(postprocess_map) / sizeof(vsi_nn_postprocess_map_element_t); +} diff --git a/resource/openpose/wksp/asymmetric_affine/vnn_post_process.h b/resource/openpose/wksp/asymmetric_affine/vnn_post_process.h new file mode 100644 index 0000000..36351dc --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/vnn_post_process.h @@ -0,0 +1,16 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction post-process header file +****************************************************************************/ +#ifndef _VNN_POST_PROCESS_H_ +#define _VNN_POST_PROCESS_H_ + +vsi_status vnn_PostProcessAsymmetricAffine(vsi_nn_graph_t *graph); + +const vsi_nn_postprocess_map_element_t * vnn_GetPostProcessMap(); + +uint32_t vnn_GetPostProcessMapCount(); + +#endif diff --git a/resource/openpose/wksp/asymmetric_affine/vnn_pre_process.c b/resource/openpose/wksp/asymmetric_affine/vnn_pre_process.c new file mode 100644 index 0000000..73d81ec --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/vnn_pre_process.c @@ -0,0 +1,900 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction pre-process source file +****************************************************************************/ +/*------------------------------------------- + Includes +-------------------------------------------*/ +#include +#include +#include + +#include "jpeglib.h" +#include "vsi_nn_pub.h" +#include "vnn_global.h" +#include "vnn_pre_process.h" + +#define _BASETSD_H + +/*------------------------------------------- + Variable definitions +-------------------------------------------*/ + +/*{graph_input_idx, preprocess}*/ +const static vsi_nn_preprocess_map_element_t* preprocess_map = NULL; + +/*------------------------------------------- + Functions +-------------------------------------------*/ +#define INPUT_META_NUM 1 +static vnn_input_meta_t input_meta_tab[INPUT_META_NUM]; +static void _load_input_meta() +{ + uint32_t i; + for (i = 0; i < INPUT_META_NUM; i++) + { + memset(&input_meta_tab[i].image.preprocess, + VNN_PREPRO_NONE, sizeof(int32_t) * VNN_PREPRO_NUM); + } + /* lid: input.1_121 */ + input_meta_tab[0].image.preprocess[0] = VNN_PREPRO_REORDER; + input_meta_tab[0].image.preprocess[1] = VNN_PREPRO_MEAN; + input_meta_tab[0].image.preprocess[2] = VNN_PREPRO_SCALE; + input_meta_tab[0].image.reorder[0] = 2; + input_meta_tab[0].image.reorder[1] = 1; + input_meta_tab[0].image.reorder[2] = 0; + input_meta_tab[0].image.mean[0] = 128; + input_meta_tab[0].image.mean[1] = 128; + input_meta_tab[0].image.mean[2] = 128; + input_meta_tab[0].image.scale[0] = 0.00390625; + input_meta_tab[0].image.scale[1] = 0.00390625; + input_meta_tab[0].image.scale[2] = 0.00390625; + + +} + +static vsi_enum _get_file_type(const char *file_name) +{ + vsi_enum type = 0; + const char *ptr; + char sep = '.'; + uint32_t pos,n; + char buff[32] = {0}; + + ptr = strrchr(file_name, sep); + pos = ptr - file_name; + n = strlen(file_name) - (pos + 1); + strncpy(buff, file_name+(pos+1), n); + + if(strcmp(buff, "jpg") == 0 + || strcmp(buff, "jpeg") == 0 + || strcmp(buff, "JPG") == 0 + || strcmp(buff, "JPEG") == 0 ) + { + type = NN_FILE_JPG; + } + else if(strcmp(buff, "tensor") == 0 + || strcmp(buff, "txt") == 0) + { + char *qnt_suffix = ".qnt.tensor"; + ptr = strstr(file_name, qnt_suffix); + if(ptr && strlen(qnt_suffix)) + { + type = NN_FILE_QTENSOR; + } + else + { + type = NN_FILE_TENSOR; + } + } + else if(strcmp(buff, "qtensor") == 0) + { + type = NN_FILE_QTENSOR; + } + else if(strcmp(buff, "bin") == 0 + || strcmp(buff, "dat") == 0) + { + type = NN_FILE_BINARY; + } + else + { + type = NN_FILE_NONE; + } + + return type; +} + +static vsi_status _jpeg_to_bmp + ( + FILE * inputFile, + unsigned char* bmpData, + vsi_size_t bmpWidth, + vsi_size_t bmpHeight, + vsi_size_t channel + ) +{ + struct jpeg_decompress_struct cinfo; + struct jpeg_error_mgr jerr; + JSAMPARRAY buffer; + unsigned char *point = NULL; + unsigned long width, height; + unsigned short depth = 0; + + cinfo.err = jpeg_std_error(&jerr); + jpeg_create_decompress(&cinfo); + jpeg_stdio_src(&cinfo,inputFile); + jpeg_read_header(&cinfo,TRUE); + + cinfo.dct_method = JDCT_IFAST; + + if (bmpData == NULL) + { + return VSI_FAILURE; + } + else + { + jpeg_start_decompress(&cinfo); + + width = cinfo.output_width; + height = cinfo.output_height; + depth = cinfo.output_components; + if(width * height * depth != bmpWidth * bmpHeight * channel) + { + printf("wrong jpg file , the jpg file size should be %u %u %u\n", + bmpWidth, bmpHeight, channel); + return VSI_FAILURE; + } + + buffer = (*cinfo.mem->alloc_sarray) + ((j_common_ptr)&cinfo, JPOOL_IMAGE, width*depth, 1); + + point = bmpData; + + while (cinfo.output_scanline < height) + { + jpeg_read_scanlines(&cinfo, buffer, 1); + memcpy(point, *buffer, width * depth); + point += width * depth; + } + + jpeg_finish_decompress(&cinfo); + } + + jpeg_destroy_decompress(&cinfo); + + return VSI_SUCCESS; +} + +static uint8_t *_float32_to_dtype + ( + float *fdata, + vsi_nn_tensor_t *tensor + ) +{ + vsi_status status; + uint8_t *data; + vsi_size_t sz,i,stride; + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + data = (uint8_t *)malloc(stride * sz * sizeof(uint8_t)); + TEST_CHECK_PTR(data, final); + memset(data, 0, stride * sz * sizeof(uint8_t)); + + for(i = 0; i < sz; i++) + { + status = vsi_nn_Float32ToDtype(fdata[i], &data[stride * i], &tensor->attr.dtype); + if(status != VSI_SUCCESS) + { + if(data)free(data); + return NULL; + } + } + +final: + return data; +} + +static float *_imageData_to_float32 + ( + uint8_t *bmpData, + vsi_nn_tensor_t *tensor + ) +{ + float *fdata; + vsi_size_t sz,i; + + fdata = NULL; + sz = vsi_nn_GetElementNum(tensor); + fdata = (float *)malloc(sz * sizeof(float)); + TEST_CHECK_PTR(fdata, final); + + for(i = 0; i < sz; i++) + { + fdata[i] = (float)bmpData[i]; + } + +final: + return fdata; +} + +/* + jpg file --> BMP data(dataformat: RGBRGBRGB...) +*/ +static uint8_t *_decode_jpeg + ( + const char *name, + vsi_nn_tensor_t *tensor + ) +{ + FILE *bmpFile; + uint8_t *bmpData; + vsi_size_t sz,w,h,c; + vsi_status status; + + bmpFile = NULL; + bmpData = NULL; + w = tensor->attr.size[0]; + h = tensor->attr.size[1]; + c = tensor->attr.size[2]; + sz = vsi_nn_GetElementNum(tensor); + + bmpFile = fopen( name, "rb" ); + TEST_CHECK_PTR(bmpFile, final); + + bmpData = (uint8_t *)malloc(sz * sizeof(uint8_t)); + TEST_CHECK_PTR(bmpData, final); + memset(bmpData, 0, sz * sizeof(uint8_t)); + + status = _jpeg_to_bmp( bmpFile, bmpData, w, h, c); + if(status == VSI_FAILURE) + { + free(bmpData); + fclose(bmpFile); + return NULL; + } + +final: + if(bmpFile)fclose(bmpFile); + return bmpData; +} + +static void _data_scale + ( + float *fdata, + vnn_input_meta_t *meta, + vsi_nn_tensor_t *tensor + ) +{ + vsi_size_t s0,s1,s2; + vsi_size_t i,j,offset; + float val,scale; + + s0 = tensor->attr.size[0]; + s1 = tensor->attr.size[1]; + s2 = tensor->attr.size[2]; + for(i = 0; i < s2; i++) + { + offset = s0 * s1 * i; + scale = meta->image.scale[i]; + for(j = 0; j < s0 * s1; j++) + { + val = fdata[offset + j] * scale; + fdata[offset + j ] = val; + } + } + +} + +static void _data_mean + ( + float *fdata, + vnn_input_meta_t *meta, + vsi_nn_tensor_t *tensor + ) +{ + vsi_size_t s0,s1,s2; + vsi_size_t i,j,offset; + float val,mean; + + s0 = tensor->attr.size[0]; + s1 = tensor->attr.size[1]; + s2 = tensor->attr.size[2]; + + for(i = 0; i < s2; i++) + { + offset = s0 * s1 * i; + mean = meta->image.mean[i]; + for(j = 0; j < s0 * s1; j++) + { + val = fdata[offset + j] - mean; + fdata[offset + j ] = val; + } + } + +} + +/* + caffe: transpose + reorder + tf: reorder +*/ +static void _data_transform + ( + float *fdata, + vnn_input_meta_t *meta, + vsi_nn_tensor_t *tensor + ) +{ + vsi_size_t s0,s1,s2; + vsi_size_t i,j,offset,sz,order; + float * data; + uint32_t * reorder; + + data = NULL; + reorder = meta->image.reorder; + s0 = tensor->attr.size[0]; + s1 = tensor->attr.size[1]; + s2 = tensor->attr.size[2]; + sz = vsi_nn_GetElementNum(tensor); + data = (float *)malloc(sz * sizeof(float)); + TEST_CHECK_PTR(data, final); + memset(data, 0, sizeof(float) * sz); + + for(i = 0; i < s2; i++) + { + if(s2 > 1 && reorder[i] <= s2) + { + order = reorder[i]; + } + else + { + order = i; + } + + offset = s0 * s1 * i; + for(j = 0; j < s0 * s1; j++) + { + data[j + offset] = fdata[j * s2 + order]; + } + } + + + memcpy(fdata, data, sz * sizeof(float)); +final: + if(data)free(data); +} + +static uint8_t *_get_binary_data + ( + vsi_nn_tensor_t *tensor, + const char *name + ) +{ + uint8_t *tensorData; + vsi_size_t sz,stride,ret,total_sz; + FILE *tensorFile; + + tensorData = NULL; + tensorFile = fopen(name, "rb"); + TEST_CHECK_PTR(tensorFile, error); + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + total_sz = sz * stride; + tensorData = (uint8_t *)malloc(total_sz * sizeof(uint8_t)); + TEST_CHECK_PTR(tensorData, error); + + memset(tensorData, 0, total_sz * sizeof(uint8_t)); + ret = fread(tensorData, 1, total_sz, tensorFile); + if(ret != total_sz) + { + printf("Read %s fail\n", name); + printf("read data %u != tensor sz %u\n", ret, total_sz); + if(tensorData)free(tensorData); + goto error; + } + + if(tensorFile)fclose(tensorFile); + return tensorData; +error: + if(tensorFile)fclose(tensorFile); + return NULL; +} + +static uint8_t *_get_qtensor_data + ( + vsi_nn_tensor_t *tensor, + const char *name + ) +{ + vsi_size_t i = 0; + float fval = 0.0; + uint8_t *tensorData; + vsi_size_t sz = 1,stride = 1; + FILE *tensorFile; + uint16_t uint16_temp_value = 0; + int16_t int16_temp_value = 0; + + tensorData = NULL; + tensorFile = fopen(name, "rb"); + TEST_CHECK_PTR(tensorFile, error); + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + tensorData = (uint8_t *)malloc(sz * stride * sizeof(uint8_t)); + TEST_CHECK_PTR(tensorData, error); + memset(tensorData, 0, sz * stride * sizeof(uint8_t)); + + for(i = 0; i < sz; i++) + { + if(fscanf( tensorFile, "%f ", &fval ) != 1) + { + printf("Read tensor file fail.\n"); + printf("Please check file lines or if the file contains illegal characters\n"); + goto error; + } + if(1 == stride) + { + if(VSI_NN_TYPE_INT8 == tensor->attr.dtype.vx_type) + tensorData[i * stride] = (int8_t)fval; + else + tensorData[i * stride] = (uint8_t)fval; + } + else if(2 == stride) + { + if(VSI_NN_TYPE_INT16 == tensor->attr.dtype.vx_type) + { + int16_temp_value = (int16_t)fval; + memcpy(tensorData + i * stride, &int16_temp_value, stride * sizeof(uint8_t)); + } + else + { + uint16_temp_value = (uint16_t)fval; + memcpy(tensorData + i * stride, &uint16_temp_value, stride * sizeof(uint8_t)); + } + } + else + { + printf("Do not support quant data with length of %u.\n", stride); + goto error; + } + } + + if(tensorFile)fclose(tensorFile); + return tensorData; +error: + if(tensorFile)fclose(tensorFile); + return NULL; +} + +static uint8_t *_get_tensor_data + ( + vsi_nn_tensor_t *tensor, + const char *name + ) +{ + vsi_status status = VSI_FAILURE; + vsi_size_t i = 0; + float fval = 0.0; + uint8_t *tensorData; + vsi_size_t sz = 1; + vsi_size_t stride = 1; + FILE *tensorFile; + + tensorData = NULL; + tensorFile = fopen(name, "rb"); + TEST_CHECK_PTR(tensorFile, error); + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride ==0) + { + stride = 1; + } + tensorData = (uint8_t *)malloc(stride * sz * sizeof(uint8_t)); + TEST_CHECK_PTR(tensorData, error); + memset(tensorData, 0, stride * sz * sizeof(uint8_t)); + + for(i = 0; i < sz; i++) + { + if(fscanf( tensorFile, "%f ", &fval ) != 1) + { + printf("Read tensor file fail.\n"); + printf("Please check file lines or if the file contains illegal characters\n"); + goto error; + } + status = vsi_nn_Float32ToDtype(fval, &tensorData[stride * i], &tensor->attr.dtype); + TEST_CHECK_STATUS(status, error); + } + + if(tensorFile)fclose(tensorFile); + return tensorData; +error: + if(tensorFile)fclose(tensorFile); + return NULL; +} + +static uint8_t *_get_jpeg_data + ( + vsi_nn_tensor_t *tensor, + vnn_input_meta_t *meta, + const char *filename + ) +{ + uint32_t i; + uint8_t *bmpData,*data; + float *fdata; + vsi_bool use_image_process = vnn_UseImagePreprocessNode(); + + bmpData = NULL; + fdata = NULL; + data = NULL; + + bmpData = _decode_jpeg(filename, tensor); + TEST_CHECK_PTR(bmpData, final); + + if(use_image_process) + { + data = bmpData; + goto final; + } + + fdata = _imageData_to_float32(bmpData, tensor); + TEST_CHECK_PTR(fdata, final); + + for(i = 0; i < _cnt_of_array(meta->image.preprocess); i++) + { + switch (meta->image.preprocess[i]) + { + case VNN_PREPRO_NONE: + break; + case VNN_PREPRO_REORDER: + _data_transform(fdata, meta, tensor); + break; + case VNN_PREPRO_MEAN: + _data_mean(fdata, meta, tensor); + break; + case VNN_PREPRO_SCALE: + _data_scale(fdata, meta, tensor); + break; + default: + break; + } + } + + data = _float32_to_dtype(fdata, tensor); + TEST_CHECK_PTR(data, final); +final: + if(fdata) + { + free(fdata); + fdata = NULL; + } + if(use_image_process) + { + ; + } + else + { + if(bmpData) + { + free(bmpData); + bmpData = NULL; + } + } + + return data; +} + +#define IMAGE_ADDR_ALIGN_START_SIZE 64 +#define IMAGE_ADDR_ALIGN_BLOCK_SIZE 64 + +static uint8_t *buffer_img = NULL; +static uint8_t *buffer_img_align_addr = NULL; + +static void _get_image_handle_buffer + ( + vsi_size_t width, + vsi_size_t height, + vsi_size_t channels, + vsi_size_t align_start_size, + vsi_size_t align_block_size + ) +{ + vsi_size_t sz; + uint64_t temp; + + sz = width * height * channels + align_start_size + align_block_size; + buffer_img = (uint8_t *)malloc( sz * sizeof( uint8_t ) ); + memset(buffer_img, 0, sizeof( uint8_t ) * sz); + + temp = (uint64_t)(buffer_img) % align_start_size; + if (temp == 0) + { + buffer_img_align_addr = buffer_img; + } + else + { + buffer_img_align_addr = buffer_img + align_start_size - temp; + } +} + +static vsi_status _handle_multiple_inputs + ( + vsi_nn_graph_t *graph, + uint32_t idx, + const char *input_file + ) +{ + vsi_status status; + vsi_nn_tensor_t *tensor; + uint8_t *data; + vnn_input_meta_t meta; + vsi_enum fileType; + char dumpInput[128]; + char *p1 = NULL; + + status = VSI_FAILURE; + data = NULL; + tensor = NULL; + memset(&meta, 0, sizeof(vnn_input_meta_t)); + tensor = vsi_nn_GetTensor( graph, graph->input.tensors[idx] ); + meta = input_meta_tab[idx]; + fileType = _get_file_type(input_file); + switch(fileType) + { + case NN_FILE_JPG: + data = _get_jpeg_data(tensor, &meta, input_file); + TEST_CHECK_PTR(data, final); + break; + case NN_FILE_TENSOR: + data = _get_tensor_data(tensor, input_file); + TEST_CHECK_PTR(data, final); + break; + case NN_FILE_QTENSOR: + data = _get_qtensor_data(tensor, input_file); + TEST_CHECK_PTR(data, final); + break; + case NN_FILE_BINARY: + data = _get_binary_data(tensor, input_file); + TEST_CHECK_PTR(data, final); + break; + default: + printf("error input file type\n"); + break; + } + + /* Copy the Pre-processed data to input tensor */ + status = vsi_nn_CopyDataToTensor(graph, tensor, data); + TEST_CHECK_STATUS(status, final); + + /* Save the image data to file */ + p1 = getenv( "VSI_SAVE_FILE_TYPE"); + + snprintf(dumpInput, sizeof(dumpInput), "input_%d.dat", idx); + vsi_nn_SaveTensorToBinary(graph, tensor, dumpInput); + + + status = VSI_SUCCESS; +final: + if(data)free(data); + return status; +} + +void vnn_ReleaseBufferImage() +{ + if (buffer_img) free(buffer_img); + buffer_img = NULL; +} + +vsi_bool vnn_UseImagePreprocessNode() +{ + int32_t use_img_process; + char *use_img_process_s; + use_img_process = 0; /* default is 0 */ + use_img_process_s = getenv("VSI_USE_IMAGE_PROCESS"); + if(use_img_process_s) + { + use_img_process = atoi(use_img_process_s); + } + if (use_img_process) + { + return TRUE; + } + return FALSE; +} + +vsi_status vnn_PreProcessAsymmetricAffine + ( + vsi_nn_graph_t *graph, + const char **inputs, + uint32_t input_num + ) +{ + uint32_t i; + vsi_status status; + status = VSI_FAILURE; + _load_input_meta(); + if(input_num != graph->input.num) + { + printf("Graph need %u inputs, but enter %u inputs!!!\n", + graph->input.num, input_num); + return status; + } + for(i = 0; i < input_num; i++) + { + status = _handle_multiple_inputs(graph, i, inputs[i]); + TEST_CHECK_STATUS(status, final); + } + + status = VSI_SUCCESS; +final: + return status; +} + +vsi_size_t vnn_LoadFP32DataFromTextFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ) +{ + float fval = 0.0; + vsi_size_t i = 0; + uint8_t * buffer = NULL; + vsi_size_t item_ount = 0; + vsi_size_t read_size = 0; + vsi_size_t stride = sizeof(fval); + FILE *fp = NULL; + + if(!fname || !buffer_ptr || !buffer_sz) + { + return read_size; + } + + fp = fopen(fname, "rb"); + if(fp) + { + while(!feof(fp) && fscanf( fp, "%f ", &fval ) == 1) + { + item_ount++; + } + + if(item_ount > 0) + { + read_size = item_ount * stride; + buffer = (uint8_t *)malloc(read_size); + if(buffer) + { + int fail_to_read = FALSE; + + VSI_FSEEK(fp, 0, SEEK_SET); + for(i = 0; i < item_ount && !fail_to_read; i++) + { + if(fscanf( fp, "%f ", (float *)&buffer[stride * i] ) != 1) + { + printf("Read tensor file fail.\n"); + printf("Please check file lines or if the file contains illegal characters\n"); + free(buffer); + fail_to_read = TRUE; + read_size = 0; + break; + } + } + + if(!fail_to_read) + { + *buffer_ptr = buffer; + *buffer_sz = read_size; + } + } + else + { + read_size = 0; + printf("Allocate memory fail!\n"); + } + } + else + { + printf("No available data found!\n"); + } + fclose(fp); + } + else + { + printf("Fail to open %s\n", fname); + } + + if(!read_size) + { + printf("Load data from %s fail!\n", fname); + } + + return read_size; +} + +vsi_size_t vnn_LoadRawDataFromBinaryFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ) +{ + FILE * fp = NULL; + vsi_size_t fsize = 0; + vsi_size_t read_size = 0; + uint8_t* buffer = NULL; + + if(!fname || !buffer_ptr || !buffer_sz) + { + return fsize; + } + + fp = fopen(fname, "rb"); + if(fp) + { + fsize = VSI_FSEEK(fp, 0, SEEK_END); + fsize = ftell(fp); + + buffer = (uint8_t *)malloc(fsize); + if(buffer) + { + VSI_FSEEK(fp, 0, SEEK_SET); + read_size = fread(buffer, 1, fsize, fp); + if(read_size == fsize) + { + *buffer_ptr = buffer; + *buffer_sz = read_size; + } + else + { + fsize = 0; + free(buffer); + buffer = NULL; + } + } + else + { + fsize = 0; + printf("Allocate memory fail!\n"); + } + + if(fp) + { + fclose(fp); + } + } + + if(!fsize) + { + printf("Load data from %s fail!\n", fname); + } + return fsize; +} + +const vsi_nn_preprocess_map_element_t * vnn_GetPreProcessMap() +{ + return preprocess_map; +} + +uint32_t vnn_GetPreProcessMapCount() +{ + if (preprocess_map == NULL) + return 0; + else + return sizeof(preprocess_map) / sizeof(vsi_nn_preprocess_map_element_t); +} diff --git a/resource/openpose/wksp/asymmetric_affine/vnn_pre_process.h b/resource/openpose/wksp/asymmetric_affine/vnn_pre_process.h new file mode 100644 index 0000000..384067d --- /dev/null +++ b/resource/openpose/wksp/asymmetric_affine/vnn_pre_process.h @@ -0,0 +1,72 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction pre-process header file +****************************************************************************/ +#ifndef _VNN_PRE_PROCESS_H_ +#define _VNN_PRE_PROCESS_H_ + +typedef enum _vnn_file_type +{ + NN_FILE_NONE, + NN_FILE_TENSOR, + NN_FILE_QTENSOR, + NN_FILE_JPG, + NN_FILE_BINARY +} vnn_file_type_e; + +typedef enum _vnn_pre_order +{ + VNN_PREPRO_NONE = -1, + VNN_PREPRO_REORDER, + VNN_PREPRO_MEAN, + VNN_PREPRO_SCALE, + VNN_PREPRO_NUM +} vnn_pre_order_e; + +typedef struct _vnn_input_meta +{ + union + { + struct + { + int32_t preprocess[VNN_PREPRO_NUM]; + uint32_t reorder[4]; + float mean[4]; + float scale[4]; + int32_t channel_count; + } image; + }; +} vnn_input_meta_t; + +vsi_status vnn_PreProcessAsymmetricAffine + ( + vsi_nn_graph_t *graph, + const char **inputs, + uint32_t input_num + ); + +vsi_bool vnn_UseImagePreprocessNode(); + +void vnn_ReleaseBufferImage(); + +vsi_size_t vnn_LoadFP32DataFromTextFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ); + +vsi_size_t vnn_LoadRawDataFromBinaryFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ); + +const vsi_nn_preprocess_map_element_t * vnn_GetPreProcessMap(); + +uint32_t vnn_GetPreProcessMapCount(); + +#endif diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/.cproject b/resource/openpose/wksp/dynamic_fixed_point-16/.cproject new file mode 100644 index 0000000..64f6138 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/.cproject @@ -0,0 +1,966 @@ + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/.project b/resource/openpose/wksp/dynamic_fixed_point-16/.project new file mode 100644 index 0000000..ff2eb12 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/.project @@ -0,0 +1,73 @@ + + + dynamicfixedpoint16 + + + + + + org.eclipse.cdt.managedbuilder.core.genmakebuilder + clean,full,incremental, + + + ?name? + + + + org.eclipse.cdt.make.core.append_environment + true + + + org.eclipse.cdt.make.core.buildArguments + + + + org.eclipse.cdt.make.core.buildCommand + make + + + org.eclipse.cdt.make.core.buildLocation + ${workspace_loc:/${project_name}/Debug} + + + org.eclipse.cdt.make.core.contents + org.eclipse.cdt.make.core.activeConfigSettings + + + org.eclipse.cdt.make.core.enableAutoBuild + false + + + org.eclipse.cdt.make.core.enableCleanBuild + true + + + org.eclipse.cdt.make.core.enableFullBuild + true + + + org.eclipse.cdt.make.core.stopOnError + true + + + org.eclipse.cdt.make.core.useDefaultBuildCmd + true + + + + + org.eclipse.cdt.managedbuilder.core.ScannerConfigBuilder + full,incremental, + + + + + + com.verisilicon.vdt.core.vdtnature + com.verisilicon.vdt.core.ovxnature + org.eclipse.cdt.core.cnature + org.eclipse.cdt.core.ccnature + org.eclipse.cdt.managedbuilder.core.managedBuildNature + org.eclipse.cdt.managedbuilder.core.ScannerConfigNature + + diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/BUILD b/resource/openpose/wksp/dynamic_fixed_point-16/BUILD new file mode 100644 index 0000000..7b1cceb --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/BUILD @@ -0,0 +1,28 @@ +# AUTO GENERATED FILE, BUILD AND RUN IN OVXLIB + +package(default_visibility = ["//visibility:public"]) + +filegroup( + name = "srcs", + srcs = + [ + "vnn_dynamicfixedpoint16.c", + "vnn_dynamicfixedpoint16.h", + "vnn_post_process.c", + "vnn_post_process.h", + "vnn_pre_process.c", + "vnn_pre_process.h", + "vnn_global.h", + "main.c", + ], +) + +cc_binary( + name = "inference", + srcs = [":srcs"] + ["//:ovxlib"], + deps = [ + "//third-party/jpeg-9b:libjpeg", + "//:ovxlib", + "@VIV_SDK//:VIV_SDK_LIB", + ], +) diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/dump_core_graph.json b/resource/openpose/wksp/dynamic_fixed_point-16/dump_core_graph.json new file mode 100644 index 0000000..8991bd3 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/dump_core_graph.json @@ -0,0 +1,2483 @@ +{ + "MetaData": { + "Name": "core graph", + "Version": "0.0.1" + }, + "Layers": { + "node_100000": { + "inputs": [], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[424", " 256", " 3", " 1]"], + "input_lifetime": "kInput", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 3", " 32]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[32]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["2", "2"], + "dilation": ["0", "0"] + } + }, + "node_100001": { + "inputs": ["@node_100000:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100002": { + "inputs": ["@node_100001:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 32", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[32]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100003": { + "inputs": ["@node_100002:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[212", " 128", " 32", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100004": { + "inputs": ["@node_100003:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[212", " 128", " 32", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 32", " 64]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[64]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[212", " 128", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100005": { + "inputs": ["@node_100004:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[212", " 128", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[212", " 128", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100006": { + "inputs": ["@node_100005:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[212", " 128", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 64", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[64]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["2", "2"], + "dilation": ["0", "0"] + } + }, + "node_100007": { + "inputs": ["@node_100006:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 64", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100008": { + "inputs": ["@node_100007:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 64", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 64", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100009": { + "inputs": ["@node_100008:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100010": { + "inputs": ["@node_100009:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100011": { + "inputs": ["@node_100010:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100012": { + "inputs": ["@node_100011:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100013": { + "inputs": ["@node_100012:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[106", " 64", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100014": { + "inputs": ["@node_100013:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[106", " 64", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["2", "2"], + "dilation": ["0", "0"] + } + }, + "node_100015": { + "inputs": ["@node_100014:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100016": { + "inputs": ["@node_100015:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 256]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100017": { + "inputs": ["@node_100016:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100018": { + "inputs": ["@node_100017:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 256", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100019": { + "inputs": ["@node_100018:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100020": { + "inputs": ["@node_100019:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 256", " 256]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100021": { + "inputs": ["@node_100020:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100022": { + "inputs": ["@node_100021:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 256", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[256]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100023": { + "inputs": ["@node_100022:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 256", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100024": { + "inputs": ["@node_100023:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 256", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 256", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100025": { + "inputs": ["@node_100024:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100026": { + "inputs": ["@node_100025:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[5", " 5", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100027": { + "inputs": ["@node_100026:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100028": { + "inputs": ["@node_100027:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100029": { + "inputs": ["@node_100028:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100030": { + "inputs": ["@node_100029:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100031": { + "inputs": ["@node_100030:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100032": { + "inputs": ["@node_100031:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100033": { + "inputs": ["@node_100032:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100034": { + "inputs": ["@node_100033:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100035": { + "inputs": ["@node_100034:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100036": { + "inputs": ["@node_100035:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100037": { + "inputs": ["@node_100036:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100038": { + "inputs": ["@node_100037:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100039": { + "inputs": ["@node_100038:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100040": { + "inputs": ["@node_100039:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100041": { + "inputs": ["@node_100040:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100042": { + "inputs": ["@node_100041:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 512", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100043": { + "inputs": ["@node_100042:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100044": { + "inputs": ["@node_100043:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100045": { + "inputs": ["@node_100044:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100046": { + "inputs": ["@node_100045:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100047": { + "inputs": ["@node_100046:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100048": { + "inputs": ["@node_100047:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100049": { + "inputs": ["@node_100048:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100050": { + "inputs": ["@node_100049:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100051": { + "inputs": ["@node_100050:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100052": { + "inputs": ["@node_100051:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100053": { + "inputs": ["@node_100052:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100054": { + "inputs": ["@node_100053:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100055": { + "inputs": ["@node_100054:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100056": { + "inputs": ["@node_100055:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 1]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100057": { + "inputs": ["@node_100056:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100058": { + "inputs": ["@node_100057:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100059": { + "inputs": ["@node_100058:out0"], + "outputs": ["out0"], + "op": "lut", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kFloat16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100060": { + "inputs": ["@node_100047:out0", "@node_100059:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kInt16", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kFloat16", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100061": { + "inputs": ["@node_100060:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100062": { + "inputs": ["@node_100061:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100063": { + "inputs": ["@node_100062:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100064": { + "inputs": ["@node_100063:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100065": { + "inputs": ["@node_100064:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100066": { + "inputs": ["@node_100065:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100067": { + "inputs": ["@node_100066:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100068": { + "inputs": ["@node_100067:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100069": { + "inputs": ["@node_100068:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100070": { + "inputs": ["@node_100068:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 512]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[512]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100071": { + "inputs": ["@node_100069:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100072": { + "inputs": ["@node_100070:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 512", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100073": { + "inputs": ["@node_100071:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 19]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[19]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 19", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100074": { + "inputs": ["@node_100072:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 512", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 512", " 38]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[38]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 38", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100075": { + "inputs": ["@node_100074:out0"], + "outputs": ["out0"], + "op": "TensorCopy", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 38", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kFloat32", + "output_shape": ["[53", " 32", " 38", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0" + } + }, + "node_100076": { + "inputs": ["@node_100062:out0", "@node_100073:out0", "@node_100074:out0"], + "outputs": ["out0"], + "op": "Concat", + "parameters": { + "input_0_dtype": "kInt16", + "input_0_shape": ["[53", " 32", " 128", " 1]"], + "input_0_lifetime": "kTransient", + "input_0_dma_mem_attr": "0", + "input_1_dtype": "kInt16", + "input_1_shape": ["[53", " 32", " 19", " 1]"], + "input_1_lifetime": "kTransient", + "input_1_dma_mem_attr": "0", + "input_2_dtype": "kInt16", + "input_2_shape": ["[53", " 32", " 38", " 1]"], + "input_2_lifetime": "kTransient", + "input_2_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 185", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "axis": 2 + } + }, + "node_100077": { + "inputs": ["@node_100076:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 185", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 185", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100078": { + "inputs": ["@node_100077:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100079": { + "inputs": ["@node_100078:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100080": { + "inputs": ["@node_100079:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100081": { + "inputs": ["@node_100080:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100082": { + "inputs": ["@node_100081:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100083": { + "inputs": ["@node_100078:out0", "@node_100082:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kInt16", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kInt16", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100084": { + "inputs": ["@node_100083:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100085": { + "inputs": ["@node_100084:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100086": { + "inputs": ["@node_100085:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100087": { + "inputs": ["@node_100086:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100088": { + "inputs": ["@node_100087:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100089": { + "inputs": ["@node_100088:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100090": { + "inputs": ["@node_100085:out0", "@node_100089:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kInt16", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kInt16", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100091": { + "inputs": ["@node_100090:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100092": { + "inputs": ["@node_100091:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100093": { + "inputs": ["@node_100092:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100094": { + "inputs": ["@node_100093:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100095": { + "inputs": ["@node_100094:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100096": { + "inputs": ["@node_100095:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100097": { + "inputs": ["@node_100092:out0", "@node_100096:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kInt16", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kInt16", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100098": { + "inputs": ["@node_100097:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100099": { + "inputs": ["@node_100098:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100100": { + "inputs": ["@node_100099:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100101": { + "inputs": ["@node_100100:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100102": { + "inputs": ["@node_100101:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100103": { + "inputs": ["@node_100102:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100104": { + "inputs": ["@node_100099:out0", "@node_100103:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kInt16", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kInt16", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100105": { + "inputs": ["@node_100104:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100106": { + "inputs": ["@node_100105:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100107": { + "inputs": ["@node_100106:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[3", " 3", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["1", "1", "1", "1"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100108": { + "inputs": ["@node_100107:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100109": { + "inputs": ["@node_100108:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[5", " 5", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["2", "2", "2", "2"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100110": { + "inputs": ["@node_100109:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100111": { + "inputs": ["@node_100106:out0", "@node_100110:out0"], + "outputs": ["out0"], + "op": "TensorAdd", + "parameters": { + "input0_dtype": "kInt16", + "input0_shape": ["[53", " 32", " 128", " 1]"], + "input0_lifetime": "kTransient", + "input0_dma_mem_attr": "0", + "input1_dtype": "kInt16", + "input1_shape": ["[53", " 32", " 128", " 1]"], + "input1_lifetime": "kTransient", + "input1_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100112": { + "inputs": ["@node_100111:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100113": { + "inputs": ["@node_100111:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 128]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[128]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100114": { + "inputs": ["@node_100112:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100115": { + "inputs": ["@node_100113:out0"], + "outputs": ["out0"], + "op": "relu", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 128", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0" + } + }, + "node_100116": { + "inputs": ["@node_100114:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 38]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[38]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 38", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100117": { + "inputs": ["@node_100115:out0"], + "outputs": ["out0"], + "op": "Conv2d", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 128", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "wegith_dtype": "kInt16", + "wegith_shape": ["[1", " 1", " 128", " 19]"], + "wegith_lifetime": "kConstant", + "wegith_dma_mem_attr": "0", + "bias_dtype": "kInt64", + "bias_shape": ["[19]"], + "bias_lifetime": "kConstant", + "bias_dma_mem_attr": "0", + "output_dtype": "kInt16", + "output_shape": ["[53", " 32", " 19", " 1]"], + "output_lifetime": "kTransient", + "output_dma_mem_attr": "0", + "pad": ["0", "0", "0", "0"], + "stride": ["1", "1"], + "dilation": ["0", "0"] + } + }, + "node_100118": { + "inputs": ["@node_100116:out0"], + "outputs": ["out0"], + "op": "TensorCopy", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 38", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kFloat32", + "output_shape": ["[53", " 32", " 38", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0" + } + }, + "node_100119": { + "inputs": ["@node_100073:out0"], + "outputs": ["out0"], + "op": "TensorCopy", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 19", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kFloat32", + "output_shape": ["[53", " 32", " 19", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0" + } + }, + "node_100120": { + "inputs": ["@node_100117:out0"], + "outputs": ["out0"], + "op": "TensorCopy", + "parameters": { + "input_dtype": "kInt16", + "input_shape": ["[53", " 32", " 19", " 1]"], + "input_lifetime": "kTransient", + "input_dma_mem_attr": "0", + "output_dtype": "kFloat32", + "output_shape": ["[53", " 32", " 19", " 1]"], + "output_lifetime": "kOutput", + "output_dma_mem_attr": "0" + } + } + } +} diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/dynamic_fixed_point-16.export.data b/resource/openpose/wksp/dynamic_fixed_point-16/dynamic_fixed_point-16.export.data new file mode 100644 index 0000000..bb5e513 Binary files /dev/null and b/resource/openpose/wksp/dynamic_fixed_point-16/dynamic_fixed_point-16.export.data differ diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/dynamicfixedpoint16.2012.vcxproj b/resource/openpose/wksp/dynamic_fixed_point-16/dynamicfixedpoint16.2012.vcxproj new file mode 100644 index 0000000..e6e8683 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/dynamicfixedpoint16.2012.vcxproj @@ -0,0 +1,224 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Jenkins-Debug + Win32 + + + Jenkins-Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + dynamicfixedpoint16 + {816BB4FE-3813-3D87-9CF9-2B0D268ED67A} + dynamicfixedpoint16 + + + + Application + true + v110 + MultiByte + + + Application + true + v110 + MultiByte + + + Application + true + v110 + MultiByte + + + Application + true + v110 + MultiByte + + + Application + false + v110 + true + MultiByte + + + Application + false + v110 + true + MultiByte + + + + + + + + + + + + + + + + + + + + + + + + $([System.IO.File]::ReadAllText($(SolutionDir)ovxlib.path.user).Trim()) + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + + + + + + + + + + + + diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/dynamicfixedpoint16.vcxproj b/resource/openpose/wksp/dynamic_fixed_point-16/dynamicfixedpoint16.vcxproj new file mode 100644 index 0000000..06f9cff --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/dynamicfixedpoint16.vcxproj @@ -0,0 +1,224 @@ + + + + + Debug + Win32 + + + Debug + x64 + + + Jenkins-Debug + Win32 + + + Jenkins-Debug + x64 + + + Release + Win32 + + + Release + x64 + + + + dynamicfixedpoint16 + {CFF12436-742A-4286-AA83-291F06E8D652} + dynamicfixedpoint16 + + + + Application + true + v142 + MultiByte + + + Application + true + v142 + MultiByte + + + Application + true + v142 + MultiByte + + + Application + true + v142 + MultiByte + + + Application + false + v142 + true + MultiByte + + + Application + false + v142 + true + MultiByte + + + + + + + + + + + + + + + + + + + + + + + + $([System.IO.File]::ReadAllText($(SolutionDir)ovxlib.path.user).Trim()) + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(SolutionDir)$(Configuration)\objs\$(ProjectName)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + .exe + $(GWG_SDK_DIR)\bin\ + $(Platform)\$(Configuration)\ + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + Disabled + $(VIV_SDK_PATH)\include;$(SolutionDir)\include;$(SolutionDir)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + + + true + $(VIV_SDK_PATH)\lib\win32;$(SolutionDir)$(Configuration) + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + Level3 + MaxSpeed + $(AQROOT)\sdk\inc;$(OVXLIB_PATH)\include;$(OVXLIB_PATH)\third-party\jpeg-9b + WIN32;_MBCS;%(PreprocessorDefinitions) + true + true + + + true + $(GWG_SDK_DIR)\lib;$(GWG_SDK_DIR)\bin;$(SolutionDir)$(Configuration); + %(AdditionalDependencies)libCLC.lib;libVSC.lib;libOpenVX.lib;libopenvxu.lib;libovxlib.lib;jpeg.lib; + true + true + + + + + + + + + + + + + + + + \ No newline at end of file diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/graph.json b/resource/openpose/wksp/dynamic_fixed_point-16/graph.json new file mode 100644 index 0000000..7d610a0 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/graph.json @@ -0,0 +1,1669 @@ +{ + "MetaData":{ + "Name": "Ovxlib_Debug_Graph", + "AcuityVersion": "UNKNOWN", + "Platform": "UNKNOWN", + "Org_Platform": "UNKNOWN" + }, + "Layers":{ + "uid_120":{ + "op": "CONV2D", + "inputs": [ "@uid_10000:out0", "@data_input_uid_123:out0", "@data_input_uid_124:out0" ], + "inut_shape": [ [ 424, 256, 3, 1 ],[ 3, 3, 3, 32 ],[ 32 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_119":{ + "op": "RELU", + "inputs": [ "@uid_120:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_118":{ + "op": "CONV2D", + "inputs": [ "@uid_119:out0", "@data_input_uid_125:out0", "@data_input_uid_126:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ],[ 3, 3, 32, 1 ],[ 32 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_117":{ + "op": "RELU", + "inputs": [ "@uid_118:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 32, 1 ] ] + }, + "uid_116":{ + "op": "CONV2D", + "inputs": [ "@uid_117:out0", "@data_input_uid_127:out0", "@data_input_uid_128:out0" ], + "inut_shape": [ [ 212, 128, 32, 1 ],[ 1, 1, 32, 64 ],[ 64 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 64, 1 ] ] + }, + "uid_115":{ + "op": "RELU", + "inputs": [ "@uid_116:out0" ], + "inut_shape": [ [ 212, 128, 64, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 212, 128, 64, 1 ] ] + }, + "uid_114":{ + "op": "CONV2D", + "inputs": [ "@uid_115:out0", "@data_input_uid_129:out0", "@data_input_uid_130:out0" ], + "inut_shape": [ [ 212, 128, 64, 1 ],[ 3, 3, 64, 1 ],[ 64 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 64, 1 ] ] + }, + "uid_113":{ + "op": "RELU", + "inputs": [ "@uid_114:out0" ], + "inut_shape": [ [ 106, 64, 64, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 64, 1 ] ] + }, + "uid_112":{ + "op": "CONV2D", + "inputs": [ "@uid_113:out0", "@data_input_uid_131:out0", "@data_input_uid_132:out0" ], + "inut_shape": [ [ 106, 64, 64, 1 ],[ 1, 1, 64, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_111":{ + "op": "RELU", + "inputs": [ "@uid_112:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_110":{ + "op": "CONV2D", + "inputs": [ "@uid_111:out0", "@data_input_uid_133:out0", "@data_input_uid_134:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_109":{ + "op": "RELU", + "inputs": [ "@uid_110:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_108":{ + "op": "CONV2D", + "inputs": [ "@uid_109:out0", "@data_input_uid_135:out0", "@data_input_uid_136:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_107":{ + "op": "RELU", + "inputs": [ "@uid_108:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 106, 64, 128, 1 ] ] + }, + "uid_106":{ + "op": "CONV2D", + "inputs": [ "@uid_107:out0", "@data_input_uid_137:out0", "@data_input_uid_138:out0" ], + "inut_shape": [ [ 106, 64, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_105":{ + "op": "RELU", + "inputs": [ "@uid_106:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_104":{ + "op": "CONV2D", + "inputs": [ "@uid_105:out0", "@data_input_uid_139:out0", "@data_input_uid_140:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 256 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_103":{ + "op": "RELU", + "inputs": [ "@uid_104:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_102":{ + "op": "CONV2D", + "inputs": [ "@uid_103:out0", "@data_input_uid_141:out0", "@data_input_uid_142:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 3, 3, 256, 1 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_101":{ + "op": "RELU", + "inputs": [ "@uid_102:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_100":{ + "op": "CONV2D", + "inputs": [ "@uid_101:out0", "@data_input_uid_143:out0", "@data_input_uid_144:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 1, 1, 256, 256 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_99":{ + "op": "RELU", + "inputs": [ "@uid_100:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_98":{ + "op": "CONV2D", + "inputs": [ "@uid_99:out0", "@data_input_uid_145:out0", "@data_input_uid_146:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 3, 3, 256, 1 ],[ 256 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_97":{ + "op": "RELU", + "inputs": [ "@uid_98:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 256, 1 ] ] + }, + "uid_96":{ + "op": "CONV2D", + "inputs": [ "@uid_97:out0", "@data_input_uid_147:out0", "@data_input_uid_148:out0" ], + "inut_shape": [ [ 53, 32, 256, 1 ],[ 1, 1, 256, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_95":{ + "op": "RELU", + "inputs": [ "@uid_96:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_94":{ + "op": "CONV2D", + "inputs": [ "@uid_95:out0", "@data_input_uid_149:out0", "@data_input_uid_150:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 5, 5, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_93":{ + "op": "RELU", + "inputs": [ "@uid_94:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_92":{ + "op": "CONV2D", + "inputs": [ "@uid_93:out0", "@data_input_uid_151:out0", "@data_input_uid_152:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_91":{ + "op": "RELU", + "inputs": [ "@uid_92:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_90":{ + "op": "CONV2D", + "inputs": [ "@uid_91:out0", "@data_input_uid_153:out0", "@data_input_uid_154:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_89":{ + "op": "RELU", + "inputs": [ "@uid_90:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_88":{ + "op": "CONV2D", + "inputs": [ "@uid_89:out0", "@data_input_uid_155:out0", "@data_input_uid_156:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_87":{ + "op": "RELU", + "inputs": [ "@uid_88:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_86":{ + "op": "CONV2D", + "inputs": [ "@uid_87:out0", "@data_input_uid_157:out0", "@data_input_uid_158:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_85":{ + "op": "RELU", + "inputs": [ "@uid_86:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_84":{ + "op": "CONV2D", + "inputs": [ "@uid_85:out0", "@data_input_uid_159:out0", "@data_input_uid_160:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_83":{ + "op": "RELU", + "inputs": [ "@uid_84:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_82":{ + "op": "CONV2D", + "inputs": [ "@uid_83:out0", "@data_input_uid_161:out0", "@data_input_uid_162:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_81":{ + "op": "RELU", + "inputs": [ "@uid_82:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_75":{ + "op": "CONV2D", + "inputs": [ "@uid_81:out0", "@data_input_uid_163:out0", "@data_input_uid_164:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_71":{ + "op": "RELU", + "inputs": [ "@uid_75:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_67":{ + "op": "CONV2D", + "inputs": [ "@uid_71:out0", "@data_input_uid_165:out0", "@data_input_uid_166:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 3, 3, 512, 1 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_62":{ + "op": "RELU", + "inputs": [ "@uid_67:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_58":{ + "op": "CONV2D", + "inputs": [ "@uid_62:out0", "@data_input_uid_167:out0", "@data_input_uid_168:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_54":{ + "op": "RELU", + "inputs": [ "@uid_58:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_49":{ + "op": "CONV2D", + "inputs": [ "@uid_54:out0", "@data_input_uid_169:out0", "@data_input_uid_170:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_45":{ + "op": "RELU", + "inputs": [ "@uid_49:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_80":{ + "op": "CONV2D", + "inputs": [ "@uid_45:out0", "@data_input_uid_171:out0", "@data_input_uid_172:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_79":{ + "op": "ELU", + "inputs": [ "@uid_80:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_78":{ + "op": "CONV2D", + "inputs": [ "@uid_79:out0", "@data_input_uid_173:out0", "@data_input_uid_174:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_77":{ + "op": "ELU", + "inputs": [ "@uid_78:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_76":{ + "op": "CONV2D", + "inputs": [ "@uid_77:out0", "@data_input_uid_175:out0", "@data_input_uid_176:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_72":{ + "op": "ELU", + "inputs": [ "@uid_76:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_68":{ + "op": "CONV2D", + "inputs": [ "@uid_72:out0", "@data_input_uid_177:out0", "@data_input_uid_178:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_63":{ + "op": "ELU", + "inputs": [ "@uid_68:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_59":{ + "op": "CONV2D", + "inputs": [ "@uid_63:out0", "@data_input_uid_179:out0", "@data_input_uid_180:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 1 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_55":{ + "op": "ELU", + "inputs": [ "@uid_59:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_50":{ + "op": "CONV2D", + "inputs": [ "@uid_55:out0", "@data_input_uid_181:out0", "@data_input_uid_182:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_46":{ + "op": "ELU", + "inputs": [ "@uid_50:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_42":{ + "op": "ADD", + "inputs": [ "@uid_45:out0", "@uid_46:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_38":{ + "op": "CONV2D", + "inputs": [ "@uid_42:out0", "@data_input_uid_183:out0", "@data_input_uid_184:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_35":{ + "op": "RELU", + "inputs": [ "@uid_38:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_32":{ + "op": "CONV2D", + "inputs": [ "@uid_35:out0", "@data_input_uid_185:out0", "@data_input_uid_186:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_28":{ + "op": "RELU", + "inputs": [ "@uid_32:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_25":{ + "op": "CONV2D", + "inputs": [ "@uid_28:out0", "@data_input_uid_187:out0", "@data_input_uid_188:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_22":{ + "op": "RELU", + "inputs": [ "@uid_25:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_19":{ + "op": "CONV2D", + "inputs": [ "@uid_22:out0", "@data_input_uid_189:out0", "@data_input_uid_190:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_18":{ + "op": "RELU", + "inputs": [ "@uid_19:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_14":{ + "op": "CONV2D", + "inputs": [ "@uid_18:out0", "@data_input_uid_191:out0", "@data_input_uid_192:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_15":{ + "op": "CONV2D", + "inputs": [ "@uid_18:out0", "@data_input_uid_193:out0", "@data_input_uid_194:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 512 ],[ 512 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_10":{ + "op": "RELU", + "inputs": [ "@uid_14:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_11":{ + "op": "RELU", + "inputs": [ "@uid_15:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 512, 1 ] ] + }, + "uid_6":{ + "op": "CONV2D", + "inputs": [ "@uid_10:out0", "@data_input_uid_195:out0", "@data_input_uid_196:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 38 ],[ 38 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 38, 1 ] ] + }, + "uid_7":{ + "op": "CONV2D", + "inputs": [ "@uid_11:out0", "@data_input_uid_197:out0", "@data_input_uid_198:out0" ], + "inut_shape": [ [ 53, 32, 512, 1 ],[ 1, 1, 512, 19 ],[ 19 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 19, 1 ] ] + }, + "uid_73":{ + "op": "CONCAT", + "inputs": [ "@uid_35:out0", "@uid_7:out0", "@uid_6:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 19, 1 ],[ 53, 32, 38, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 185, 1 ] ] + }, + "uid_69":{ + "op": "CONV2D", + "inputs": [ "@uid_73:out0", "@data_input_uid_199:out0", "@data_input_uid_200:out0" ], + "inut_shape": [ [ 53, 32, 185, 1 ],[ 1, 1, 185, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_64":{ + "op": "RELU", + "inputs": [ "@uid_69:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_74":{ + "op": "CONV2D", + "inputs": [ "@uid_64:out0", "@data_input_uid_201:out0", "@data_input_uid_202:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_70":{ + "op": "RELU", + "inputs": [ "@uid_74:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_66":{ + "op": "CONV2D", + "inputs": [ "@uid_70:out0", "@data_input_uid_203:out0", "@data_input_uid_204:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_65":{ + "op": "RELU", + "inputs": [ "@uid_66:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_60":{ + "op": "ADD", + "inputs": [ "@uid_64:out0", "@uid_65:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_56":{ + "op": "CONV2D", + "inputs": [ "@uid_60:out0", "@data_input_uid_205:out0", "@data_input_uid_206:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_51":{ + "op": "RELU", + "inputs": [ "@uid_56:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_61":{ + "op": "CONV2D", + "inputs": [ "@uid_51:out0", "@data_input_uid_207:out0", "@data_input_uid_208:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_57":{ + "op": "RELU", + "inputs": [ "@uid_61:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_53":{ + "op": "CONV2D", + "inputs": [ "@uid_57:out0", "@data_input_uid_209:out0", "@data_input_uid_210:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_52":{ + "op": "RELU", + "inputs": [ "@uid_53:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_47":{ + "op": "ADD", + "inputs": [ "@uid_51:out0", "@uid_52:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_43":{ + "op": "CONV2D", + "inputs": [ "@uid_47:out0", "@data_input_uid_211:out0", "@data_input_uid_212:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_39":{ + "op": "RELU", + "inputs": [ "@uid_43:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_48":{ + "op": "CONV2D", + "inputs": [ "@uid_39:out0", "@data_input_uid_213:out0", "@data_input_uid_214:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_44":{ + "op": "RELU", + "inputs": [ "@uid_48:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_41":{ + "op": "CONV2D", + "inputs": [ "@uid_44:out0", "@data_input_uid_215:out0", "@data_input_uid_216:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_40":{ + "op": "RELU", + "inputs": [ "@uid_41:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_36":{ + "op": "ADD", + "inputs": [ "@uid_39:out0", "@uid_40:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_33":{ + "op": "CONV2D", + "inputs": [ "@uid_36:out0", "@data_input_uid_217:out0", "@data_input_uid_218:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_29":{ + "op": "RELU", + "inputs": [ "@uid_33:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_37":{ + "op": "CONV2D", + "inputs": [ "@uid_29:out0", "@data_input_uid_219:out0", "@data_input_uid_220:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_34":{ + "op": "RELU", + "inputs": [ "@uid_37:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_31":{ + "op": "CONV2D", + "inputs": [ "@uid_34:out0", "@data_input_uid_221:out0", "@data_input_uid_222:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_30":{ + "op": "RELU", + "inputs": [ "@uid_31:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_26":{ + "op": "ADD", + "inputs": [ "@uid_29:out0", "@uid_30:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_23":{ + "op": "CONV2D", + "inputs": [ "@uid_26:out0", "@data_input_uid_223:out0", "@data_input_uid_224:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_20":{ + "op": "RELU", + "inputs": [ "@uid_23:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_27":{ + "op": "CONV2D", + "inputs": [ "@uid_20:out0", "@data_input_uid_225:out0", "@data_input_uid_226:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 3, 3, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_24":{ + "op": "RELU", + "inputs": [ "@uid_27:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_21":{ + "op": "CONV2D", + "inputs": [ "@uid_24:out0", "@data_input_uid_227:out0", "@data_input_uid_228:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 5, 5, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_17":{ + "op": "RELU", + "inputs": [ "@uid_21:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_16":{ + "op": "ADD", + "inputs": [ "@uid_20:out0", "@uid_17:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_12":{ + "op": "CONV2D", + "inputs": [ "@uid_16:out0", "@data_input_uid_229:out0", "@data_input_uid_230:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_13":{ + "op": "CONV2D", + "inputs": [ "@uid_16:out0", "@data_input_uid_231:out0", "@data_input_uid_232:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 128 ],[ 128 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_8":{ + "op": "RELU", + "inputs": [ "@uid_12:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_9":{ + "op": "RELU", + "inputs": [ "@uid_13:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 128, 1 ] ] + }, + "uid_4":{ + "op": "CONV2D", + "inputs": [ "@uid_8:out0", "@data_input_uid_233:out0", "@data_input_uid_234:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 38 ],[ 38 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 38, 1 ] ] + }, + "uid_5":{ + "op": "CONV2D", + "inputs": [ "@uid_9:out0", "@data_input_uid_235:out0", "@data_input_uid_236:out0" ], + "inut_shape": [ [ 53, 32, 128, 1 ],[ 1, 1, 128, 19 ],[ 19 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 19, 1 ] ] + }, + "uid_10000":{ + "op": "PRE_PROCESS", + "inputs": [ "@data_input_uid_237:out0" ], + "inut_shape": [ [ 1272, 256, 1, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 424, 256, 3, 1 ] ] + }, + "uid_20000":{ + "op": "POST_PROCESS", + "inputs": [ "@uid_7:out0" ], + "inut_shape": [ [ 53, 32, 19, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 19, 1 ] ] + }, + "uid_20001":{ + "op": "POST_PROCESS", + "inputs": [ "@uid_6:out0" ], + "inut_shape": [ [ 53, 32, 38, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 38, 1 ] ] + }, + "uid_20002":{ + "op": "POST_PROCESS", + "inputs": [ "@uid_5:out0" ], + "inut_shape": [ [ 53, 32, 19, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 19, 1 ] ] + }, + "uid_20003":{ + "op": "POST_PROCESS", + "inputs": [ "@uid_4:out0" ], + "inut_shape": [ [ 53, 32, 38, 1 ] ], + "outputs": [ "out0" ], + "output_shape": [ [ 53, 32, 38, 1 ] ] + }, + "data_input_uid_123":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_124":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_125":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_126":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_127":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_128":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_129":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_130":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_131":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_132":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_133":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_134":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_135":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_136":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_137":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_138":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_139":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_140":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_141":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_142":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_143":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_144":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_145":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_146":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_147":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_148":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_149":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_150":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_151":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_152":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_153":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_154":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_155":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_156":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_157":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_158":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_159":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_160":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_161":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_162":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_163":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_164":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_165":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_166":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_167":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_168":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_169":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_170":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_171":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_172":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_173":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_174":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_175":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_176":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_177":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_178":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_179":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_180":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_181":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_182":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_183":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_184":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_185":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_186":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_187":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_188":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_189":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_190":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_191":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_192":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_193":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_194":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_195":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_196":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_197":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_198":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_199":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_200":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_201":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_202":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_203":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_204":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_205":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_206":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_207":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_208":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_209":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_210":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_211":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_212":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_213":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_214":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_215":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_216":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_217":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_218":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_219":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_220":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_221":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_222":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_223":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_224":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_225":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_226":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_227":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_228":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_229":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_230":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_231":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_232":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_233":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_234":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_235":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_236":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + }, + "data_input_uid_237":{ + "op": "DATA_INPUT", + "inputs": [ ], + "inut_shape": [ [ ] ], + "outputs": [ "out0" ], + "output_shape": [ [ ] ] + } + } +} diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/main.c b/resource/openpose/wksp/dynamic_fixed_point-16/main.c new file mode 100644 index 0000000..59c45ab --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/main.c @@ -0,0 +1,264 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network application project entry file +****************************************************************************/ +/*------------------------------------------- + Includes +-------------------------------------------*/ +#include +#include +#include +#ifdef __linux__ +#include +#include +#elif defined(_WIN32) +#include +#endif + +#define _BASETSD_H + +#include "vsi_nn_pub.h" + +#include "vnn_global.h" +#include "vnn_pre_process.h" +#include "vnn_post_process.h" +#include "vnn_dynamicfixedpoint16.h" + +/*------------------------------------------- + Macros and Variables +-------------------------------------------*/ +#ifdef __linux__ +#define VSI_UINT64_SPECIFIER PRIu64 +#elif defined(_WIN32) +#define VSI_UINT64_SPECIFIER "I64u" +#endif + +/*------------------------------------------- + Functions +-------------------------------------------*/ +static void vnn_ReleaseNeuralNetwork + ( + vsi_nn_graph_t *graph + ) +{ + vnn_ReleaseDynamicFixedPoint16( graph, TRUE ); + if (vnn_UseImagePreprocessNode()) + { + vnn_ReleaseBufferImage(); + } +} + +static vsi_status vnn_PostProcessNeuralNetwork + ( + vsi_nn_graph_t *graph + ) +{ + return vnn_PostProcessDynamicFixedPoint16( graph ); +} + +#define BILLION 1000000000 +static uint64_t get_perf_count() +{ +#if defined(__linux__) || defined(__ANDROID__) || defined(__QNX__) || defined(__CYGWIN__) + struct timespec ts; + + clock_gettime(CLOCK_MONOTONIC, &ts); + + return (uint64_t)((uint64_t)ts.tv_nsec + (uint64_t)ts.tv_sec * BILLION); +#elif defined(_WIN32) || defined(UNDER_CE) + LARGE_INTEGER freq; + LARGE_INTEGER ln; + + QueryPerformanceFrequency(&freq); + QueryPerformanceCounter(&ln); + + return (uint64_t)(ln.QuadPart * BILLION / freq.QuadPart); +#endif +} + +static vsi_status vnn_VerifyGraph + ( + vsi_nn_graph_t *graph + ) +{ + vsi_status status = VSI_FAILURE; + uint64_t tmsStart, tmsEnd, msVal, usVal; + + /* Verify graph */ + printf("Verify...\n"); + tmsStart = get_perf_count(); + status = vsi_nn_VerifyGraph( graph ); + TEST_CHECK_STATUS(status, final); + tmsEnd = get_perf_count(); + msVal = (tmsEnd - tmsStart)/1000000; + usVal = (tmsEnd - tmsStart)/1000; + printf("Verify Graph: %"VSI_UINT64_SPECIFIER"ms or %"VSI_UINT64_SPECIFIER"us\n", msVal, usVal); + +final: + return status; +} + +static vsi_status vnn_ProcessGraph + ( + vsi_nn_graph_t *graph + ) +{ + vsi_status status = VSI_FAILURE; + int32_t i,loop; + char *loop_s; + uint64_t tmsStart, tmsEnd, sigStart, sigEnd; + float msVal, usVal; + + status = VSI_FAILURE; + loop = 1; /* default loop time is 1 */ + loop_s = getenv("VNN_LOOP_TIME"); + if(loop_s) + { + loop = atoi(loop_s); + } + + /* Run graph */ + tmsStart = get_perf_count(); + printf("Start run graph [%d] times...\n", loop); + for(i = 0; i < loop; i++) + { + sigStart = get_perf_count(); +#ifdef VNN_APP_ASYNC_RUN + status = vsi_nn_AsyncRunGraph( graph ); + if(status != VSI_SUCCESS) + { + printf("Async Run graph the %d time fail\n", i); + } + TEST_CHECK_STATUS( status, final ); + + //do something here... + + status = vsi_nn_AsyncRunWait( graph ); + if(status != VSI_SUCCESS) + { + printf("Wait graph the %d time fail\n", i); + } +#else + status = vsi_nn_RunGraph( graph ); + if(status != VSI_SUCCESS) + { + printf("Run graph the %d time fail\n", i); + } +#endif + TEST_CHECK_STATUS( status, final ); + + sigEnd = get_perf_count(); + msVal = (sigEnd - sigStart)/(float)1000000; + usVal = (sigEnd - sigStart)/(float)1000; + printf("Run the %u time: %.2fms or %.2fus\n", (i + 1), msVal, usVal); + } + tmsEnd = get_perf_count(); + msVal = (tmsEnd - tmsStart)/(float)1000000; + usVal = (tmsEnd - tmsStart)/(float)1000; + printf("vxProcessGraph execution time:\n"); + printf("Total %.2fms or %.2fus\n", msVal, usVal); + printf("Average %.2fms or %.2fus\n", ((float)usVal)/1000/loop, ((float)usVal)/loop); + +final: + return status; +} + +static vsi_status vnn_PreProcessNeuralNetwork + ( + vsi_nn_graph_t *graph, + int argc, + char **argv + ) +{ + /* + * argv0: execute file + * argv1: data file + * argv2~n: inputs n file + */ + const char **inputs = (const char **)argv + 2; + uint32_t input_num = argc - 2; + + return vnn_PreProcessDynamicFixedPoint16( graph, inputs, input_num ); +} + +static vsi_nn_graph_t *vnn_CreateNeuralNetwork + ( + const char *data_file_name + ) +{ + vsi_nn_graph_t *graph = NULL; + uint64_t tmsStart, tmsEnd, msVal, usVal; + + tmsStart = get_perf_count(); + graph = vnn_CreateDynamicFixedPoint16( data_file_name, NULL, + vnn_GetPreProcessMap(), vnn_GetPreProcessMapCount(), + vnn_GetPostProcessMap(), vnn_GetPostProcessMapCount() ); + TEST_CHECK_PTR(graph, final); + + tmsEnd = get_perf_count(); + msVal = (tmsEnd - tmsStart)/1000000; + usVal = (tmsEnd - tmsStart)/1000; + printf("Create Neural Network: %"VSI_UINT64_SPECIFIER"ms or %"VSI_UINT64_SPECIFIER"us\n", msVal, usVal); + +final: + return graph; +} + +/*------------------------------------------- + Main Functions +-------------------------------------------*/ +int main + ( + int argc, + char **argv + ) +{ + vsi_status status = VSI_FAILURE; + vsi_nn_graph_t *graph; + const char *data_name = NULL; + + if(argc < 3) + { + printf("Usage: %s data_file inputs...\n", argv[0]); + return -1; + } + + data_name = (const char *)argv[1]; + + /* Create the neural network */ + graph = vnn_CreateNeuralNetwork( data_name ); + TEST_CHECK_PTR( graph, final ); + + /* Verify graph */ + status = vnn_VerifyGraph( graph ); + TEST_CHECK_STATUS( status, final); + + /* Pre process the image data */ + status = vnn_PreProcessNeuralNetwork( graph, argc, argv ); + TEST_CHECK_STATUS( status, final ); + + + + /* Process graph */ + status = vnn_ProcessGraph( graph ); + TEST_CHECK_STATUS( status, final ); + + if(VNN_APP_DEBUG) + { + /* Dump all node outputs */ + vsi_nn_DumpGraphNodeOutputs(graph, "./network_dump", NULL, 0, TRUE, 0); + } + + /* Post process output data */ + status = vnn_PostProcessNeuralNetwork( graph ); + TEST_CHECK_STATUS( status, final ); + +final: + vnn_ReleaseNeuralNetwork( graph ); + fflush(stdout); + fflush(stderr); + return status; +} + diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/makefile.linux b/resource/openpose/wksp/dynamic_fixed_point-16/makefile.linux new file mode 100644 index 0000000..70cec05 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/makefile.linux @@ -0,0 +1,127 @@ +ifeq (1,$(USE_IDE_LIB)) #idelib +CC=$(CROSS_COMPILE)gcc +CXX=$(CROSS_COMPILE)g++ +DEBUG=0 +#GWG_SDK_DIR=../IDE5.4.0/cmdtools/vsimulator +INCLUDES=-I. -I$(GWG_SDK_DIR)/include/ \ + -I$(GWG_SDK_DIR)/include/CL \ + -I$(GWG_SDK_DIR)/include/VX \ + -I$(GWG_SDK_DIR)/include/ovxlib \ + -I$(GWG_SDK_DIR)/include/jpeg +CFLAGS=-Wall -std=c++0x $(INCLUDES) -D__linux__ -DLINUX +ifeq (1,$(DEBUG)) +CFLAGS+=-g +LFLAGS+=-g +else +CFLAGS+=-O3 +LFLAGS+=-O3 +endif +LIBS+= -L$(GWG_SDK_DIR)/lib \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy -lArchModelSw -lNNArchPerf +LIBS+= -L$(GWG_SDK_DIR)/lib/vsim \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/lib/x64_linux \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/lib/x64_linux/vsim \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/lib/x64_linux/vsim \ + -lOpenVX -lOpenVXU -lCLC -lVSC -lGAL -lovxlib -lEmulator -lvdtproxy +LIBS+= -L$(GWG_SDK_DIR)/../common/lib/ \ + -lvdtproxy +File = $(GWG_SDK_DIR)/lib/libjpeg.a +File2 = $(GWG_SDK_DIR)/lib/x64_linux/libjpeg.a +File3 = $(GWG_SDK_DIR)/../common/lib/libjpeg.a +ifeq ($(File),$(wildcard $(File))) +LIBS+= $(File) +else ifeq ($(File2),$(wildcard $(File2))) +LIBS+= $(File2) +else +LIBS+= $(File3) +endif +SRCS=${wildcard *.c} +SRCS+=${wildcard *.cpp} +BIN=dynamic_fixed_point-16 +OBJS=$(addsuffix .o, $(basename $(SRCS))) + +.SUFFIXES: .cpp .c + +.cpp.o: + $(CC) $(CFLAGS) -c $< + +.cpp: + $(CXX) $(CFLAGS) $< -o $@ -lm + +.c.o: + $(CC) $(CFLAGS) -c $< + +.c: + $(CC) $(CFLAGS) $< -o $@ -lm + +all: $(BIN) + +$(BIN): $(OBJS) + $(CC) $(CFLAGS) $(LFLAGS) $(EXTRALFLAGS) $(OBJS) $(LIBS) -o $@ + +clean: + rm -rf *.o + rm -rf $(BIN) + rm -rf *~ + +############################################################################## +# Netranslib. Supply necessary libraries. +else +include $(AQROOT)/makefile.linux.def +INCLUDE += -I$(GWG_SDK_INC) -I$(GWG_SDK_INC)/HAL -I$(AQROOT)/sdk/inc -I./ -I$(OVXLIB_DIR)/include/utils -I$(OVXLIB_DIR)/include/client -I$(OVXLIB_DIR)/include/ops -I$(OVXLIB_DIR)/include -I$(OVXLIB_DIR)/third-party/jpeg-9b +CFLAGS += $(INCLUDE) +ifeq ($(gcdSTATIC_LINK), 1) +LIBS += $(OVXLIB_DIR)/lib/libovxlib.a +LIBS += $(GWG_SDK_LIB)/libOpenVXU.a +LIBS += $(GWG_SDK_LIB)/libOpenVXC.a +LIBS += $(GWG_SDK_LIB)/libOpenVX.a +LIBS += $(GWG_SDK_LIB)/libCLC.a +LIBS += $(GWG_SDK_LIB)/libLLVM_viv.a +LIBS += $(GWG_SDK_LIB)/libclCompiler.a +LIBS += $(GWG_SDK_LIB)/libclPreprocessor.a +LIBS += $(GWG_SDK_LIB)/libclCommon.a +LIBS += $(GWG_SDK_LIB)/libLLVM_viv.a +LIBS += $(GWG_SDK_LIB)/libVSC.a +LIBS += $(GWG_SDK_LIB)/libhalarchuser.a +LIBS += $(GWG_SDK_LIB)/libhalosuser.a +LIBS += $(GWG_SDK_LIB)/libGAL.a +LIBS += $(GWG_SDK_LIB)/libhalarchuser.a +LIBS += $(GWG_SDK_LIB)/libGAL.a +LIBS += $(LIB_DIR)/libm.a +LIBS += $(LIB_DIR)/libpthread.a +LIBS += $(LIB_DIR)/libc.a +LIBS += $(LIB_DIR)/libdl.a +LIBS += $(LIB_DIR)/librt.a +LIBS += $(LIB_DIR)/libstdc++.a +LIBS += $(OVXLIB_DIR)/lib/libjpeg.a +else +ifeq ($(USE_VXC_BINARY)$(USE_VSC_LITE),11) +LIBS += -L$(GWG_SDK_LIB) -l OpenVX -l OpenVXU -l CLC -l VSC_Lite -lGAL +else +LIBS += -L$(GWG_SDK_LIB) -l OpenVX -l OpenVXU -l CLC -l VSC -lGAL +endif +LIBS += $(OVXLIB_DIR)/lib/libjpeg.a +LIBS += -L$(OVXLIB_DIR)/lib -l ovxlib +LIBS += -L$(LIB_DIR) -lm +endif + +############################################################################# +# Macros. +PROGRAM = 1 +TARGET_NAME = dynamic_fixed_point-16 +CUR_SOURCE = ${wildcard *.c} +############################################################################# +# Objects. +OBJECTS = ${patsubst %.c, $(OBJ_DIR)/%.o, $(CUR_SOURCE)} + +# installation directory +INSTALL_DIR := ./ + +################################################################################ +# Include the common makefile. + +include $(AQROOT)/common.target +endif diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/network_binary.nb b/resource/openpose/wksp/dynamic_fixed_point-16/network_binary.nb new file mode 100644 index 0000000..40f4a24 Binary files /dev/null and b/resource/openpose/wksp/dynamic_fixed_point-16/network_binary.nb differ diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/vnn_dynamicfixedpoint16.c b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_dynamicfixedpoint16.c new file mode 100644 index 0000000..b4ebd99 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_dynamicfixedpoint16.c @@ -0,0 +1,4806 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction network definition source file +****************************************************************************/ +/*------------------------------------------- + Includes + -------------------------------------------*/ +#include +#include + +#include "vsi_nn_pub.h" + +#include "vnn_global.h" +#include "vnn_dynamicfixedpoint16.h" + +/*------------------------------------------- + Macros + -------------------------------------------*/ + +#define NEW_VXNODE(_node, _type, _in, _out, _uid) do {\ + _node = vsi_nn_AddNode( graph, _type, _in, _out, NULL );\ + if( NULL == _node ) {\ + goto error;\ + }\ + _node->uid = (uint32_t)_uid;\ + } while(0) + +#define NEW_VIRTUAL_TENSOR(_id, _attr, _dtype) do {\ + memset( _attr.size, 0, VSI_NN_MAX_DIM_NUM * sizeof(vsi_size_t));\ + _attr.dim_num = VSI_NN_DIM_AUTO;\ + _attr.vtl = !VNN_APP_DEBUG;\ + _attr.is_const = FALSE;\ + _attr.dtype.vx_type = _dtype;\ + _id = vsi_nn_AddTensor( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +// Set const tensor dims out of this macro. +#define NEW_CONST_TENSOR(_id, _attr, _dtype, _ofst, _size) do {\ + data = load_data( fp, _ofst, _size );\ + if( NULL == data ) {\ + goto error;\ + }\ + _attr.vtl = FALSE;\ + _attr.is_const = TRUE;\ + _attr.dtype.vx_type = _dtype;\ + _id = vsi_nn_AddTensor( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, data );\ + free( data );\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +// Set generic tensor dims out of this macro. +#define NEW_NORM_TENSOR(_id, _attr, _dtype) do {\ + _attr.vtl = FALSE;\ + _attr.is_const = FALSE;\ + _attr.dtype.vx_type = _dtype;\ + if ( enable_from_handle )\ + {\ + _id = vsi_nn_AddTensorFromHandle( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + }\ + else\ + {\ + _id = vsi_nn_AddTensor( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + }\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +// Set generic tensor dims out of this macro. +#define NEW_NORM_TENSOR_FROM_HANDLE(_id, _attr, _dtype) do {\ + _attr.vtl = FALSE;\ + _attr.is_const = FALSE;\ + _attr.dtype.vx_type = _dtype;\ + _id = vsi_nn_AddTensorFromHandle( graph, VSI_NN_TENSOR_ID_AUTO,\ + & _attr, NULL );\ + if( VSI_NN_TENSOR_ID_NA == _id ) {\ + goto error;\ + }\ + } while(0) + +#define NET_NODE_NUM (117) +#define NET_NORM_TENSOR_NUM (5) +#define NET_CONST_TENSOR_NUM (114) +#define NET_VIRTUAL_TENSOR_NUM (117) +#define NET_TOTAL_TENSOR_NUM (NET_NORM_TENSOR_NUM + NET_CONST_TENSOR_NUM + NET_VIRTUAL_TENSOR_NUM) + +/*------------------------------------------- + Local Variables + -------------------------------------------*/ + +/*------------------------------------------- + Functions + -------------------------------------------*/ +static uint8_t* load_data + ( + FILE * fp, + size_t ofst, + size_t sz + ) +{ + uint8_t* data; + ssize_t ret; + size_t size; + data = NULL; + if( NULL == fp ) + { + return NULL; + } + + ret = VSI_FSEEK(fp, ofst, SEEK_SET); + if (ret != 0) + { + VSILOGE("blob seek failure."); + return NULL; + } + + data = (uint8_t*)malloc(sz); + if (data == NULL) + { + VSILOGE("buffer malloc failure."); + return NULL; + } + size = fread(data, 1, sz, fp); + if (size != sz || size == 0) + { + free(data); + data = NULL; + VSILOGE("Read file to buffer failed."); + } + return data; +} /* load_data() */ + +vsi_nn_graph_t * vnn_CreateDynamicFixedPoint16 + ( + const char * data_file_name, + vsi_nn_context_t in_ctx, + const vsi_nn_preprocess_map_element_t * pre_process_map, + uint32_t pre_process_map_count, + const vsi_nn_postprocess_map_element_t * post_process_map, + uint32_t post_process_map_count + ) +{ + uint32_t _infinity = VSI_NN_FLOAT32_INF; + vsi_status status; + vsi_bool release_ctx; + vsi_nn_context_t ctx; + vsi_nn_graph_t * graph; + vsi_nn_node_t * node[NET_NODE_NUM]; + vsi_nn_tensor_id_t norm_tensor[NET_NORM_TENSOR_NUM]; + vsi_nn_tensor_id_t const_tensor[NET_CONST_TENSOR_NUM]; + vsi_nn_tensor_attr_t attr; + FILE * fp; + uint8_t * data; + uint32_t i = 0; + char * use_img_process_s; + char * use_from_handle = NULL; + int32_t enable_pre_post_process = 0; + int32_t enable_from_handle = 0; + vsi_bool sort = FALSE; + vsi_bool inference_with_nbg = FALSE; + char* pos = NULL; + + + + + + (void)(_infinity); + ctx = NULL; + graph = NULL; + status = VSI_FAILURE; + memset( &attr, 0, sizeof( attr ) ); + memset( &node, 0, sizeof( vsi_nn_node_t * ) * NET_NODE_NUM ); + + fp = fopen( data_file_name, "rb" ); + if( NULL == fp ) + { + VSILOGE( "Open file %s failed.", data_file_name ); + goto error; + } + + pos = strstr(data_file_name, ".nb"); + if( pos && strcmp(pos, ".nb") == 0 ) + { + inference_with_nbg = TRUE; + } + + if( NULL == in_ctx ) + { + ctx = vsi_nn_CreateContext(); + } + else + { + ctx = in_ctx; + } + + use_img_process_s = getenv( "VSI_USE_IMAGE_PROCESS" ); + if( use_img_process_s ) + { + enable_pre_post_process = atoi(use_img_process_s); + } + use_from_handle = getenv( "VSI_USE_FROM_HANDLE" ); + if ( use_from_handle ) + { + enable_from_handle = atoi(use_from_handle); + } + + graph = vsi_nn_CreateGraph( ctx, NET_TOTAL_TENSOR_NUM, NET_NODE_NUM ); + if( NULL == graph ) + { + VSILOGE( "Create graph fail." ); + goto error; + } + vsi_nn_SetGraphVersion( graph, VNN_VERSION_MAJOR, VNN_VERSION_MINOR, VNN_VERSION_PATCH ); + vsi_nn_SetGraphInputs( graph, NULL, 1 ); + vsi_nn_SetGraphOutputs( graph, NULL, 4 ); + vsi_nn_SetGraphFastMode(graph,FALSE); + +/*----------------------------------------- + Register client ops + -----------------------------------------*/ + + +/*----------------------------------------- + Node definitions + -----------------------------------------*/ + if( !inference_with_nbg ) + { + + /*----------------------------------------- + lid - model/model.0/model.0.0/Conv_output_0_120 + var - node[0] + name - model/model.0/model.0.0/Conv_output_0 + operation - convolution + input - [424, 256, 3, 1] + filter - [3, 3, 3, 32] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[0], VSI_NN_OP_CONV2D, 3, 1, 120); + node[0]->nn_param.conv2d.ksize[0] = 3; + node[0]->nn_param.conv2d.ksize[1] = 3; + node[0]->nn_param.conv2d.weights = 32; + node[0]->nn_param.conv2d.stride[0] = 2; + node[0]->nn_param.conv2d.stride[1] = 2; + node[0]->nn_param.conv2d.pad[0] = 1; + node[0]->nn_param.conv2d.pad[1] = 1; + node[0]->nn_param.conv2d.pad[2] = 1; + node[0]->nn_param.conv2d.pad[3] = 1; + node[0]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[0]->nn_param.conv2d.group = 1; + node[0]->nn_param.conv2d.dilation[0] = 1; + node[0]->nn_param.conv2d.dilation[1] = 1; + node[0]->nn_param.conv2d.multiplier = 0; + node[0]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[0]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[0]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.0/model.0.2/Relu_output_0_119 + var - node[1] + name - model/model.0/model.0.2/Relu_output_0 + operation - relu + input - [212, 128, 32, 1] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[1], VSI_NN_OP_RELU, 1, 1, 119); + + /*----------------------------------------- + lid - model/model.1/model.1.0/Conv_output_0_118 + var - node[2] + name - model/model.1/model.1.0/Conv_output_0 + operation - convolution + input - [212, 128, 32, 1] + filter - [3, 3, 32, 1] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[2], VSI_NN_OP_CONV2D, 3, 1, 118); + node[2]->nn_param.conv2d.ksize[0] = 3; + node[2]->nn_param.conv2d.ksize[1] = 3; + node[2]->nn_param.conv2d.weights = 32; + node[2]->nn_param.conv2d.stride[0] = 1; + node[2]->nn_param.conv2d.stride[1] = 1; + node[2]->nn_param.conv2d.pad[0] = 1; + node[2]->nn_param.conv2d.pad[1] = 1; + node[2]->nn_param.conv2d.pad[2] = 1; + node[2]->nn_param.conv2d.pad[3] = 1; + node[2]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[2]->nn_param.conv2d.group = 32; + node[2]->nn_param.conv2d.dilation[0] = 1; + node[2]->nn_param.conv2d.dilation[1] = 1; + node[2]->nn_param.conv2d.multiplier = 1; + node[2]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[2]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[2]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.1/model.1.2/Relu_output_0_117 + var - node[3] + name - model/model.1/model.1.2/Relu_output_0 + operation - relu + input - [212, 128, 32, 1] + output - [212, 128, 32, 1] + -----------------------------------------*/ + NEW_VXNODE(node[3], VSI_NN_OP_RELU, 1, 1, 117); + + /*----------------------------------------- + lid - model/model.1/model.1.3/Conv_output_0_116 + var - node[4] + name - model/model.1/model.1.3/Conv_output_0 + operation - convolution + input - [212, 128, 32, 1] + filter - [1, 1, 32, 64] + output - [212, 128, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[4], VSI_NN_OP_CONV2D, 3, 1, 116); + node[4]->nn_param.conv2d.ksize[0] = 1; + node[4]->nn_param.conv2d.ksize[1] = 1; + node[4]->nn_param.conv2d.weights = 64; + node[4]->nn_param.conv2d.stride[0] = 1; + node[4]->nn_param.conv2d.stride[1] = 1; + node[4]->nn_param.conv2d.pad[0] = 0; + node[4]->nn_param.conv2d.pad[1] = 0; + node[4]->nn_param.conv2d.pad[2] = 0; + node[4]->nn_param.conv2d.pad[3] = 0; + node[4]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[4]->nn_param.conv2d.group = 1; + node[4]->nn_param.conv2d.dilation[0] = 1; + node[4]->nn_param.conv2d.dilation[1] = 1; + node[4]->nn_param.conv2d.multiplier = 0; + node[4]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[4]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[4]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.1/model.1.5/Relu_output_0_115 + var - node[5] + name - model/model.1/model.1.5/Relu_output_0 + operation - relu + input - [212, 128, 64, 1] + output - [212, 128, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[5], VSI_NN_OP_RELU, 1, 1, 115); + + /*----------------------------------------- + lid - model/model.2/model.2.0/Conv_output_0_114 + var - node[6] + name - model/model.2/model.2.0/Conv_output_0 + operation - convolution + input - [212, 128, 64, 1] + filter - [3, 3, 64, 1] + output - [106, 64, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[6], VSI_NN_OP_CONV2D, 3, 1, 114); + node[6]->nn_param.conv2d.ksize[0] = 3; + node[6]->nn_param.conv2d.ksize[1] = 3; + node[6]->nn_param.conv2d.weights = 64; + node[6]->nn_param.conv2d.stride[0] = 2; + node[6]->nn_param.conv2d.stride[1] = 2; + node[6]->nn_param.conv2d.pad[0] = 1; + node[6]->nn_param.conv2d.pad[1] = 1; + node[6]->nn_param.conv2d.pad[2] = 1; + node[6]->nn_param.conv2d.pad[3] = 1; + node[6]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[6]->nn_param.conv2d.group = 64; + node[6]->nn_param.conv2d.dilation[0] = 1; + node[6]->nn_param.conv2d.dilation[1] = 1; + node[6]->nn_param.conv2d.multiplier = 1; + node[6]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[6]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[6]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.2/model.2.2/Relu_output_0_113 + var - node[7] + name - model/model.2/model.2.2/Relu_output_0 + operation - relu + input - [106, 64, 64, 1] + output - [106, 64, 64, 1] + -----------------------------------------*/ + NEW_VXNODE(node[7], VSI_NN_OP_RELU, 1, 1, 113); + + /*----------------------------------------- + lid - model/model.2/model.2.3/Conv_output_0_112 + var - node[8] + name - model/model.2/model.2.3/Conv_output_0 + operation - convolution + input - [106, 64, 64, 1] + filter - [1, 1, 64, 128] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[8], VSI_NN_OP_CONV2D, 3, 1, 112); + node[8]->nn_param.conv2d.ksize[0] = 1; + node[8]->nn_param.conv2d.ksize[1] = 1; + node[8]->nn_param.conv2d.weights = 128; + node[8]->nn_param.conv2d.stride[0] = 1; + node[8]->nn_param.conv2d.stride[1] = 1; + node[8]->nn_param.conv2d.pad[0] = 0; + node[8]->nn_param.conv2d.pad[1] = 0; + node[8]->nn_param.conv2d.pad[2] = 0; + node[8]->nn_param.conv2d.pad[3] = 0; + node[8]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[8]->nn_param.conv2d.group = 1; + node[8]->nn_param.conv2d.dilation[0] = 1; + node[8]->nn_param.conv2d.dilation[1] = 1; + node[8]->nn_param.conv2d.multiplier = 0; + node[8]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[8]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[8]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.2/model.2.5/Relu_output_0_111 + var - node[9] + name - model/model.2/model.2.5/Relu_output_0 + operation - relu + input - [106, 64, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[9], VSI_NN_OP_RELU, 1, 1, 111); + + /*----------------------------------------- + lid - model/model.3/model.3.0/Conv_output_0_110 + var - node[10] + name - model/model.3/model.3.0/Conv_output_0 + operation - convolution + input - [106, 64, 128, 1] + filter - [3, 3, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[10], VSI_NN_OP_CONV2D, 3, 1, 110); + node[10]->nn_param.conv2d.ksize[0] = 3; + node[10]->nn_param.conv2d.ksize[1] = 3; + node[10]->nn_param.conv2d.weights = 128; + node[10]->nn_param.conv2d.stride[0] = 1; + node[10]->nn_param.conv2d.stride[1] = 1; + node[10]->nn_param.conv2d.pad[0] = 1; + node[10]->nn_param.conv2d.pad[1] = 1; + node[10]->nn_param.conv2d.pad[2] = 1; + node[10]->nn_param.conv2d.pad[3] = 1; + node[10]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[10]->nn_param.conv2d.group = 128; + node[10]->nn_param.conv2d.dilation[0] = 1; + node[10]->nn_param.conv2d.dilation[1] = 1; + node[10]->nn_param.conv2d.multiplier = 1; + node[10]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[10]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[10]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.3/model.3.2/Relu_output_0_109 + var - node[11] + name - model/model.3/model.3.2/Relu_output_0 + operation - relu + input - [106, 64, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[11], VSI_NN_OP_RELU, 1, 1, 109); + + /*----------------------------------------- + lid - model/model.3/model.3.3/Conv_output_0_108 + var - node[12] + name - model/model.3/model.3.3/Conv_output_0 + operation - convolution + input - [106, 64, 128, 1] + filter - [1, 1, 128, 128] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[12], VSI_NN_OP_CONV2D, 3, 1, 108); + node[12]->nn_param.conv2d.ksize[0] = 1; + node[12]->nn_param.conv2d.ksize[1] = 1; + node[12]->nn_param.conv2d.weights = 128; + node[12]->nn_param.conv2d.stride[0] = 1; + node[12]->nn_param.conv2d.stride[1] = 1; + node[12]->nn_param.conv2d.pad[0] = 0; + node[12]->nn_param.conv2d.pad[1] = 0; + node[12]->nn_param.conv2d.pad[2] = 0; + node[12]->nn_param.conv2d.pad[3] = 0; + node[12]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[12]->nn_param.conv2d.group = 1; + node[12]->nn_param.conv2d.dilation[0] = 1; + node[12]->nn_param.conv2d.dilation[1] = 1; + node[12]->nn_param.conv2d.multiplier = 0; + node[12]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[12]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[12]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.3/model.3.5/Relu_output_0_107 + var - node[13] + name - model/model.3/model.3.5/Relu_output_0 + operation - relu + input - [106, 64, 128, 1] + output - [106, 64, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[13], VSI_NN_OP_RELU, 1, 1, 107); + + /*----------------------------------------- + lid - model/model.4/model.4.0/Conv_output_0_106 + var - node[14] + name - model/model.4/model.4.0/Conv_output_0 + operation - convolution + input - [106, 64, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[14], VSI_NN_OP_CONV2D, 3, 1, 106); + node[14]->nn_param.conv2d.ksize[0] = 3; + node[14]->nn_param.conv2d.ksize[1] = 3; + node[14]->nn_param.conv2d.weights = 128; + node[14]->nn_param.conv2d.stride[0] = 2; + node[14]->nn_param.conv2d.stride[1] = 2; + node[14]->nn_param.conv2d.pad[0] = 1; + node[14]->nn_param.conv2d.pad[1] = 1; + node[14]->nn_param.conv2d.pad[2] = 1; + node[14]->nn_param.conv2d.pad[3] = 1; + node[14]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[14]->nn_param.conv2d.group = 128; + node[14]->nn_param.conv2d.dilation[0] = 1; + node[14]->nn_param.conv2d.dilation[1] = 1; + node[14]->nn_param.conv2d.multiplier = 1; + node[14]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[14]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[14]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.4/model.4.2/Relu_output_0_105 + var - node[15] + name - model/model.4/model.4.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[15], VSI_NN_OP_RELU, 1, 1, 105); + + /*----------------------------------------- + lid - model/model.4/model.4.3/Conv_output_0_104 + var - node[16] + name - model/model.4/model.4.3/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 256] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[16], VSI_NN_OP_CONV2D, 3, 1, 104); + node[16]->nn_param.conv2d.ksize[0] = 1; + node[16]->nn_param.conv2d.ksize[1] = 1; + node[16]->nn_param.conv2d.weights = 256; + node[16]->nn_param.conv2d.stride[0] = 1; + node[16]->nn_param.conv2d.stride[1] = 1; + node[16]->nn_param.conv2d.pad[0] = 0; + node[16]->nn_param.conv2d.pad[1] = 0; + node[16]->nn_param.conv2d.pad[2] = 0; + node[16]->nn_param.conv2d.pad[3] = 0; + node[16]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[16]->nn_param.conv2d.group = 1; + node[16]->nn_param.conv2d.dilation[0] = 1; + node[16]->nn_param.conv2d.dilation[1] = 1; + node[16]->nn_param.conv2d.multiplier = 0; + node[16]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[16]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[16]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.4/model.4.5/Relu_output_0_103 + var - node[17] + name - model/model.4/model.4.5/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[17], VSI_NN_OP_RELU, 1, 1, 103); + + /*----------------------------------------- + lid - model/model.5/model.5.0/Conv_output_0_102 + var - node[18] + name - model/model.5/model.5.0/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [3, 3, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[18], VSI_NN_OP_CONV2D, 3, 1, 102); + node[18]->nn_param.conv2d.ksize[0] = 3; + node[18]->nn_param.conv2d.ksize[1] = 3; + node[18]->nn_param.conv2d.weights = 256; + node[18]->nn_param.conv2d.stride[0] = 1; + node[18]->nn_param.conv2d.stride[1] = 1; + node[18]->nn_param.conv2d.pad[0] = 1; + node[18]->nn_param.conv2d.pad[1] = 1; + node[18]->nn_param.conv2d.pad[2] = 1; + node[18]->nn_param.conv2d.pad[3] = 1; + node[18]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[18]->nn_param.conv2d.group = 256; + node[18]->nn_param.conv2d.dilation[0] = 1; + node[18]->nn_param.conv2d.dilation[1] = 1; + node[18]->nn_param.conv2d.multiplier = 1; + node[18]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[18]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[18]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.5/model.5.2/Relu_output_0_101 + var - node[19] + name - model/model.5/model.5.2/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[19], VSI_NN_OP_RELU, 1, 1, 101); + + /*----------------------------------------- + lid - model/model.5/model.5.3/Conv_output_0_100 + var - node[20] + name - model/model.5/model.5.3/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [1, 1, 256, 256] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[20], VSI_NN_OP_CONV2D, 3, 1, 100); + node[20]->nn_param.conv2d.ksize[0] = 1; + node[20]->nn_param.conv2d.ksize[1] = 1; + node[20]->nn_param.conv2d.weights = 256; + node[20]->nn_param.conv2d.stride[0] = 1; + node[20]->nn_param.conv2d.stride[1] = 1; + node[20]->nn_param.conv2d.pad[0] = 0; + node[20]->nn_param.conv2d.pad[1] = 0; + node[20]->nn_param.conv2d.pad[2] = 0; + node[20]->nn_param.conv2d.pad[3] = 0; + node[20]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[20]->nn_param.conv2d.group = 1; + node[20]->nn_param.conv2d.dilation[0] = 1; + node[20]->nn_param.conv2d.dilation[1] = 1; + node[20]->nn_param.conv2d.multiplier = 0; + node[20]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[20]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[20]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.5/model.5.5/Relu_output_0_99 + var - node[21] + name - model/model.5/model.5.5/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[21], VSI_NN_OP_RELU, 1, 1, 99); + + /*----------------------------------------- + lid - model/model.6/model.6.0/Conv_output_0_98 + var - node[22] + name - model/model.6/model.6.0/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [3, 3, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[22], VSI_NN_OP_CONV2D, 3, 1, 98); + node[22]->nn_param.conv2d.ksize[0] = 3; + node[22]->nn_param.conv2d.ksize[1] = 3; + node[22]->nn_param.conv2d.weights = 256; + node[22]->nn_param.conv2d.stride[0] = 1; + node[22]->nn_param.conv2d.stride[1] = 1; + node[22]->nn_param.conv2d.pad[0] = 1; + node[22]->nn_param.conv2d.pad[1] = 1; + node[22]->nn_param.conv2d.pad[2] = 1; + node[22]->nn_param.conv2d.pad[3] = 1; + node[22]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[22]->nn_param.conv2d.group = 256; + node[22]->nn_param.conv2d.dilation[0] = 1; + node[22]->nn_param.conv2d.dilation[1] = 1; + node[22]->nn_param.conv2d.multiplier = 1; + node[22]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[22]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[22]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.6/model.6.2/Relu_output_0_97 + var - node[23] + name - model/model.6/model.6.2/Relu_output_0 + operation - relu + input - [53, 32, 256, 1] + output - [53, 32, 256, 1] + -----------------------------------------*/ + NEW_VXNODE(node[23], VSI_NN_OP_RELU, 1, 1, 97); + + /*----------------------------------------- + lid - model/model.6/model.6.3/Conv_output_0_96 + var - node[24] + name - model/model.6/model.6.3/Conv_output_0 + operation - convolution + input - [53, 32, 256, 1] + filter - [1, 1, 256, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[24], VSI_NN_OP_CONV2D, 3, 1, 96); + node[24]->nn_param.conv2d.ksize[0] = 1; + node[24]->nn_param.conv2d.ksize[1] = 1; + node[24]->nn_param.conv2d.weights = 512; + node[24]->nn_param.conv2d.stride[0] = 1; + node[24]->nn_param.conv2d.stride[1] = 1; + node[24]->nn_param.conv2d.pad[0] = 0; + node[24]->nn_param.conv2d.pad[1] = 0; + node[24]->nn_param.conv2d.pad[2] = 0; + node[24]->nn_param.conv2d.pad[3] = 0; + node[24]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[24]->nn_param.conv2d.group = 1; + node[24]->nn_param.conv2d.dilation[0] = 1; + node[24]->nn_param.conv2d.dilation[1] = 1; + node[24]->nn_param.conv2d.multiplier = 0; + node[24]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[24]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[24]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.6/model.6.5/Relu_output_0_95 + var - node[25] + name - model/model.6/model.6.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[25], VSI_NN_OP_RELU, 1, 1, 95); + + /*----------------------------------------- + lid - model/model.7/model.7.0/Conv_output_0_94 + var - node[26] + name - model/model.7/model.7.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [5, 5, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[26], VSI_NN_OP_CONV2D, 3, 1, 94); + node[26]->nn_param.conv2d.ksize[0] = 5; + node[26]->nn_param.conv2d.ksize[1] = 5; + node[26]->nn_param.conv2d.weights = 512; + node[26]->nn_param.conv2d.stride[0] = 1; + node[26]->nn_param.conv2d.stride[1] = 1; + node[26]->nn_param.conv2d.pad[0] = 2; + node[26]->nn_param.conv2d.pad[1] = 2; + node[26]->nn_param.conv2d.pad[2] = 2; + node[26]->nn_param.conv2d.pad[3] = 2; + node[26]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[26]->nn_param.conv2d.group = 512; + node[26]->nn_param.conv2d.dilation[0] = 1; + node[26]->nn_param.conv2d.dilation[1] = 1; + node[26]->nn_param.conv2d.multiplier = 1; + node[26]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[26]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[26]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.7/model.7.2/Relu_output_0_93 + var - node[27] + name - model/model.7/model.7.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[27], VSI_NN_OP_RELU, 1, 1, 93); + + /*----------------------------------------- + lid - model/model.7/model.7.3/Conv_output_0_92 + var - node[28] + name - model/model.7/model.7.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[28], VSI_NN_OP_CONV2D, 3, 1, 92); + node[28]->nn_param.conv2d.ksize[0] = 1; + node[28]->nn_param.conv2d.ksize[1] = 1; + node[28]->nn_param.conv2d.weights = 512; + node[28]->nn_param.conv2d.stride[0] = 1; + node[28]->nn_param.conv2d.stride[1] = 1; + node[28]->nn_param.conv2d.pad[0] = 0; + node[28]->nn_param.conv2d.pad[1] = 0; + node[28]->nn_param.conv2d.pad[2] = 0; + node[28]->nn_param.conv2d.pad[3] = 0; + node[28]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[28]->nn_param.conv2d.group = 1; + node[28]->nn_param.conv2d.dilation[0] = 1; + node[28]->nn_param.conv2d.dilation[1] = 1; + node[28]->nn_param.conv2d.multiplier = 0; + node[28]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[28]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[28]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.7/model.7.5/Relu_output_0_91 + var - node[29] + name - model/model.7/model.7.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[29], VSI_NN_OP_RELU, 1, 1, 91); + + /*----------------------------------------- + lid - model/model.8/model.8.0/Conv_output_0_90 + var - node[30] + name - model/model.8/model.8.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[30], VSI_NN_OP_CONV2D, 3, 1, 90); + node[30]->nn_param.conv2d.ksize[0] = 3; + node[30]->nn_param.conv2d.ksize[1] = 3; + node[30]->nn_param.conv2d.weights = 512; + node[30]->nn_param.conv2d.stride[0] = 1; + node[30]->nn_param.conv2d.stride[1] = 1; + node[30]->nn_param.conv2d.pad[0] = 1; + node[30]->nn_param.conv2d.pad[1] = 1; + node[30]->nn_param.conv2d.pad[2] = 1; + node[30]->nn_param.conv2d.pad[3] = 1; + node[30]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[30]->nn_param.conv2d.group = 512; + node[30]->nn_param.conv2d.dilation[0] = 1; + node[30]->nn_param.conv2d.dilation[1] = 1; + node[30]->nn_param.conv2d.multiplier = 1; + node[30]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[30]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[30]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.8/model.8.2/Relu_output_0_89 + var - node[31] + name - model/model.8/model.8.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[31], VSI_NN_OP_RELU, 1, 1, 89); + + /*----------------------------------------- + lid - model/model.8/model.8.3/Conv_output_0_88 + var - node[32] + name - model/model.8/model.8.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[32], VSI_NN_OP_CONV2D, 3, 1, 88); + node[32]->nn_param.conv2d.ksize[0] = 1; + node[32]->nn_param.conv2d.ksize[1] = 1; + node[32]->nn_param.conv2d.weights = 512; + node[32]->nn_param.conv2d.stride[0] = 1; + node[32]->nn_param.conv2d.stride[1] = 1; + node[32]->nn_param.conv2d.pad[0] = 0; + node[32]->nn_param.conv2d.pad[1] = 0; + node[32]->nn_param.conv2d.pad[2] = 0; + node[32]->nn_param.conv2d.pad[3] = 0; + node[32]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[32]->nn_param.conv2d.group = 1; + node[32]->nn_param.conv2d.dilation[0] = 1; + node[32]->nn_param.conv2d.dilation[1] = 1; + node[32]->nn_param.conv2d.multiplier = 0; + node[32]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[32]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[32]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.8/model.8.5/Relu_output_0_87 + var - node[33] + name - model/model.8/model.8.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[33], VSI_NN_OP_RELU, 1, 1, 87); + + /*----------------------------------------- + lid - model/model.9/model.9.0/Conv_output_0_86 + var - node[34] + name - model/model.9/model.9.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[34], VSI_NN_OP_CONV2D, 3, 1, 86); + node[34]->nn_param.conv2d.ksize[0] = 3; + node[34]->nn_param.conv2d.ksize[1] = 3; + node[34]->nn_param.conv2d.weights = 512; + node[34]->nn_param.conv2d.stride[0] = 1; + node[34]->nn_param.conv2d.stride[1] = 1; + node[34]->nn_param.conv2d.pad[0] = 1; + node[34]->nn_param.conv2d.pad[1] = 1; + node[34]->nn_param.conv2d.pad[2] = 1; + node[34]->nn_param.conv2d.pad[3] = 1; + node[34]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[34]->nn_param.conv2d.group = 512; + node[34]->nn_param.conv2d.dilation[0] = 1; + node[34]->nn_param.conv2d.dilation[1] = 1; + node[34]->nn_param.conv2d.multiplier = 1; + node[34]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[34]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[34]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.9/model.9.2/Relu_output_0_85 + var - node[35] + name - model/model.9/model.9.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[35], VSI_NN_OP_RELU, 1, 1, 85); + + /*----------------------------------------- + lid - model/model.9/model.9.3/Conv_output_0_84 + var - node[36] + name - model/model.9/model.9.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[36], VSI_NN_OP_CONV2D, 3, 1, 84); + node[36]->nn_param.conv2d.ksize[0] = 1; + node[36]->nn_param.conv2d.ksize[1] = 1; + node[36]->nn_param.conv2d.weights = 512; + node[36]->nn_param.conv2d.stride[0] = 1; + node[36]->nn_param.conv2d.stride[1] = 1; + node[36]->nn_param.conv2d.pad[0] = 0; + node[36]->nn_param.conv2d.pad[1] = 0; + node[36]->nn_param.conv2d.pad[2] = 0; + node[36]->nn_param.conv2d.pad[3] = 0; + node[36]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[36]->nn_param.conv2d.group = 1; + node[36]->nn_param.conv2d.dilation[0] = 1; + node[36]->nn_param.conv2d.dilation[1] = 1; + node[36]->nn_param.conv2d.multiplier = 0; + node[36]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[36]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[36]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.9/model.9.5/Relu_output_0_83 + var - node[37] + name - model/model.9/model.9.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[37], VSI_NN_OP_RELU, 1, 1, 83); + + /*----------------------------------------- + lid - model/model.10/model.10.0/Conv_output_0_82 + var - node[38] + name - model/model.10/model.10.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[38], VSI_NN_OP_CONV2D, 3, 1, 82); + node[38]->nn_param.conv2d.ksize[0] = 3; + node[38]->nn_param.conv2d.ksize[1] = 3; + node[38]->nn_param.conv2d.weights = 512; + node[38]->nn_param.conv2d.stride[0] = 1; + node[38]->nn_param.conv2d.stride[1] = 1; + node[38]->nn_param.conv2d.pad[0] = 1; + node[38]->nn_param.conv2d.pad[1] = 1; + node[38]->nn_param.conv2d.pad[2] = 1; + node[38]->nn_param.conv2d.pad[3] = 1; + node[38]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[38]->nn_param.conv2d.group = 512; + node[38]->nn_param.conv2d.dilation[0] = 1; + node[38]->nn_param.conv2d.dilation[1] = 1; + node[38]->nn_param.conv2d.multiplier = 1; + node[38]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[38]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[38]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.10/model.10.2/Relu_output_0_81 + var - node[39] + name - model/model.10/model.10.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[39], VSI_NN_OP_RELU, 1, 1, 81); + + /*----------------------------------------- + lid - model/model.10/model.10.3/Conv_output_0_75 + var - node[40] + name - model/model.10/model.10.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[40], VSI_NN_OP_CONV2D, 3, 1, 75); + node[40]->nn_param.conv2d.ksize[0] = 1; + node[40]->nn_param.conv2d.ksize[1] = 1; + node[40]->nn_param.conv2d.weights = 512; + node[40]->nn_param.conv2d.stride[0] = 1; + node[40]->nn_param.conv2d.stride[1] = 1; + node[40]->nn_param.conv2d.pad[0] = 0; + node[40]->nn_param.conv2d.pad[1] = 0; + node[40]->nn_param.conv2d.pad[2] = 0; + node[40]->nn_param.conv2d.pad[3] = 0; + node[40]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[40]->nn_param.conv2d.group = 1; + node[40]->nn_param.conv2d.dilation[0] = 1; + node[40]->nn_param.conv2d.dilation[1] = 1; + node[40]->nn_param.conv2d.multiplier = 0; + node[40]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[40]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[40]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.10/model.10.5/Relu_output_0_71 + var - node[41] + name - model/model.10/model.10.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[41], VSI_NN_OP_RELU, 1, 1, 71); + + /*----------------------------------------- + lid - model/model.11/model.11.0/Conv_output_0_67 + var - node[42] + name - model/model.11/model.11.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [3, 3, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[42], VSI_NN_OP_CONV2D, 3, 1, 67); + node[42]->nn_param.conv2d.ksize[0] = 3; + node[42]->nn_param.conv2d.ksize[1] = 3; + node[42]->nn_param.conv2d.weights = 512; + node[42]->nn_param.conv2d.stride[0] = 1; + node[42]->nn_param.conv2d.stride[1] = 1; + node[42]->nn_param.conv2d.pad[0] = 1; + node[42]->nn_param.conv2d.pad[1] = 1; + node[42]->nn_param.conv2d.pad[2] = 1; + node[42]->nn_param.conv2d.pad[3] = 1; + node[42]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[42]->nn_param.conv2d.group = 512; + node[42]->nn_param.conv2d.dilation[0] = 1; + node[42]->nn_param.conv2d.dilation[1] = 1; + node[42]->nn_param.conv2d.multiplier = 1; + node[42]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[42]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[42]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.11/model.11.2/Relu_output_0_62 + var - node[43] + name - model/model.11/model.11.2/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[43], VSI_NN_OP_RELU, 1, 1, 62); + + /*----------------------------------------- + lid - model/model.11/model.11.3/Conv_output_0_58 + var - node[44] + name - model/model.11/model.11.3/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[44], VSI_NN_OP_CONV2D, 3, 1, 58); + node[44]->nn_param.conv2d.ksize[0] = 1; + node[44]->nn_param.conv2d.ksize[1] = 1; + node[44]->nn_param.conv2d.weights = 512; + node[44]->nn_param.conv2d.stride[0] = 1; + node[44]->nn_param.conv2d.stride[1] = 1; + node[44]->nn_param.conv2d.pad[0] = 0; + node[44]->nn_param.conv2d.pad[1] = 0; + node[44]->nn_param.conv2d.pad[2] = 0; + node[44]->nn_param.conv2d.pad[3] = 0; + node[44]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[44]->nn_param.conv2d.group = 1; + node[44]->nn_param.conv2d.dilation[0] = 1; + node[44]->nn_param.conv2d.dilation[1] = 1; + node[44]->nn_param.conv2d.multiplier = 0; + node[44]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[44]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[44]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - model/model.11/model.11.5/Relu_output_0_54 + var - node[45] + name - model/model.11/model.11.5/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[45], VSI_NN_OP_RELU, 1, 1, 54); + + /*----------------------------------------- + lid - cpm/align/align.0/Conv_output_0_49 + var - node[46] + name - cpm/align/align.0/Conv_output_0 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[46], VSI_NN_OP_CONV2D, 3, 1, 49); + node[46]->nn_param.conv2d.ksize[0] = 1; + node[46]->nn_param.conv2d.ksize[1] = 1; + node[46]->nn_param.conv2d.weights = 128; + node[46]->nn_param.conv2d.stride[0] = 1; + node[46]->nn_param.conv2d.stride[1] = 1; + node[46]->nn_param.conv2d.pad[0] = 0; + node[46]->nn_param.conv2d.pad[1] = 0; + node[46]->nn_param.conv2d.pad[2] = 0; + node[46]->nn_param.conv2d.pad[3] = 0; + node[46]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[46]->nn_param.conv2d.group = 1; + node[46]->nn_param.conv2d.dilation[0] = 1; + node[46]->nn_param.conv2d.dilation[1] = 1; + node[46]->nn_param.conv2d.multiplier = 0; + node[46]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[46]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[46]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/align/align.1/Relu_output_0_45 + var - node[47] + name - cpm/align/align.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[47], VSI_NN_OP_RELU, 1, 1, 45); + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80 + var - node[48] + name - cpm/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[48], VSI_NN_OP_CONV2D, 3, 1, 80); + node[48]->nn_param.conv2d.ksize[0] = 3; + node[48]->nn_param.conv2d.ksize[1] = 3; + node[48]->nn_param.conv2d.weights = 128; + node[48]->nn_param.conv2d.stride[0] = 1; + node[48]->nn_param.conv2d.stride[1] = 1; + node[48]->nn_param.conv2d.pad[0] = 1; + node[48]->nn_param.conv2d.pad[1] = 1; + node[48]->nn_param.conv2d.pad[2] = 1; + node[48]->nn_param.conv2d.pad[3] = 1; + node[48]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[48]->nn_param.conv2d.group = 128; + node[48]->nn_param.conv2d.dilation[0] = 1; + node[48]->nn_param.conv2d.dilation[1] = 1; + node[48]->nn_param.conv2d.multiplier = 1; + node[48]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[48]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[48]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79 + var - node[49] + name - cpm/trunk/trunk.0/trunk.0.1/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[49], VSI_NN_OP_ELU, 1, 1, 79); + node[49]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78 + var - node[50] + name - cpm/trunk/trunk.0/trunk.0.2/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[50], VSI_NN_OP_CONV2D, 3, 1, 78); + node[50]->nn_param.conv2d.ksize[0] = 1; + node[50]->nn_param.conv2d.ksize[1] = 1; + node[50]->nn_param.conv2d.weights = 128; + node[50]->nn_param.conv2d.stride[0] = 1; + node[50]->nn_param.conv2d.stride[1] = 1; + node[50]->nn_param.conv2d.pad[0] = 0; + node[50]->nn_param.conv2d.pad[1] = 0; + node[50]->nn_param.conv2d.pad[2] = 0; + node[50]->nn_param.conv2d.pad[3] = 0; + node[50]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[50]->nn_param.conv2d.group = 1; + node[50]->nn_param.conv2d.dilation[0] = 1; + node[50]->nn_param.conv2d.dilation[1] = 1; + node[50]->nn_param.conv2d.multiplier = 0; + node[50]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[50]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[50]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77 + var - node[51] + name - cpm/trunk/trunk.0/trunk.0.3/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[51], VSI_NN_OP_ELU, 1, 1, 77); + node[51]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76 + var - node[52] + name - cpm/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[52], VSI_NN_OP_CONV2D, 3, 1, 76); + node[52]->nn_param.conv2d.ksize[0] = 3; + node[52]->nn_param.conv2d.ksize[1] = 3; + node[52]->nn_param.conv2d.weights = 128; + node[52]->nn_param.conv2d.stride[0] = 1; + node[52]->nn_param.conv2d.stride[1] = 1; + node[52]->nn_param.conv2d.pad[0] = 1; + node[52]->nn_param.conv2d.pad[1] = 1; + node[52]->nn_param.conv2d.pad[2] = 1; + node[52]->nn_param.conv2d.pad[3] = 1; + node[52]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[52]->nn_param.conv2d.group = 128; + node[52]->nn_param.conv2d.dilation[0] = 1; + node[52]->nn_param.conv2d.dilation[1] = 1; + node[52]->nn_param.conv2d.multiplier = 1; + node[52]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[52]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[52]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72 + var - node[53] + name - cpm/trunk/trunk.1/trunk.1.1/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[53], VSI_NN_OP_ELU, 1, 1, 72); + node[53]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68 + var - node[54] + name - cpm/trunk/trunk.1/trunk.1.2/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[54], VSI_NN_OP_CONV2D, 3, 1, 68); + node[54]->nn_param.conv2d.ksize[0] = 1; + node[54]->nn_param.conv2d.ksize[1] = 1; + node[54]->nn_param.conv2d.weights = 128; + node[54]->nn_param.conv2d.stride[0] = 1; + node[54]->nn_param.conv2d.stride[1] = 1; + node[54]->nn_param.conv2d.pad[0] = 0; + node[54]->nn_param.conv2d.pad[1] = 0; + node[54]->nn_param.conv2d.pad[2] = 0; + node[54]->nn_param.conv2d.pad[3] = 0; + node[54]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[54]->nn_param.conv2d.group = 1; + node[54]->nn_param.conv2d.dilation[0] = 1; + node[54]->nn_param.conv2d.dilation[1] = 1; + node[54]->nn_param.conv2d.multiplier = 0; + node[54]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[54]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[54]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63 + var - node[55] + name - cpm/trunk/trunk.1/trunk.1.3/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[55], VSI_NN_OP_ELU, 1, 1, 63); + node[55]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59 + var - node[56] + name - cpm/trunk/trunk.2/trunk.2.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[56], VSI_NN_OP_CONV2D, 3, 1, 59); + node[56]->nn_param.conv2d.ksize[0] = 3; + node[56]->nn_param.conv2d.ksize[1] = 3; + node[56]->nn_param.conv2d.weights = 128; + node[56]->nn_param.conv2d.stride[0] = 1; + node[56]->nn_param.conv2d.stride[1] = 1; + node[56]->nn_param.conv2d.pad[0] = 1; + node[56]->nn_param.conv2d.pad[1] = 1; + node[56]->nn_param.conv2d.pad[2] = 1; + node[56]->nn_param.conv2d.pad[3] = 1; + node[56]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[56]->nn_param.conv2d.group = 128; + node[56]->nn_param.conv2d.dilation[0] = 1; + node[56]->nn_param.conv2d.dilation[1] = 1; + node[56]->nn_param.conv2d.multiplier = 1; + node[56]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[56]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[56]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55 + var - node[57] + name - cpm/trunk/trunk.2/trunk.2.1/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[57], VSI_NN_OP_ELU, 1, 1, 55); + node[57]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50 + var - node[58] + name - cpm/trunk/trunk.2/trunk.2.2/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[58], VSI_NN_OP_CONV2D, 3, 1, 50); + node[58]->nn_param.conv2d.ksize[0] = 1; + node[58]->nn_param.conv2d.ksize[1] = 1; + node[58]->nn_param.conv2d.weights = 128; + node[58]->nn_param.conv2d.stride[0] = 1; + node[58]->nn_param.conv2d.stride[1] = 1; + node[58]->nn_param.conv2d.pad[0] = 0; + node[58]->nn_param.conv2d.pad[1] = 0; + node[58]->nn_param.conv2d.pad[2] = 0; + node[58]->nn_param.conv2d.pad[3] = 0; + node[58]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[58]->nn_param.conv2d.group = 1; + node[58]->nn_param.conv2d.dilation[0] = 1; + node[58]->nn_param.conv2d.dilation[1] = 1; + node[58]->nn_param.conv2d.multiplier = 0; + node[58]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[58]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[58]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46 + var - node[59] + name - cpm/trunk/trunk.2/trunk.2.3/Elu_output_0 + operation - elu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[59], VSI_NN_OP_ELU, 1, 1, 46); + node[59]->nn_param.elu.alpha = 1.0; + + /*----------------------------------------- + lid - cpm/Add_output_0_42 + var - node[60] + name - cpm/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[60], VSI_NN_OP_ADD, 2, 1, 42); + + /*----------------------------------------- + lid - cpm/conv/conv.0/Conv_output_0_38 + var - node[61] + name - cpm/conv/conv.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[61], VSI_NN_OP_CONV2D, 3, 1, 38); + node[61]->nn_param.conv2d.ksize[0] = 3; + node[61]->nn_param.conv2d.ksize[1] = 3; + node[61]->nn_param.conv2d.weights = 128; + node[61]->nn_param.conv2d.stride[0] = 1; + node[61]->nn_param.conv2d.stride[1] = 1; + node[61]->nn_param.conv2d.pad[0] = 1; + node[61]->nn_param.conv2d.pad[1] = 1; + node[61]->nn_param.conv2d.pad[2] = 1; + node[61]->nn_param.conv2d.pad[3] = 1; + node[61]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[61]->nn_param.conv2d.group = 1; + node[61]->nn_param.conv2d.dilation[0] = 1; + node[61]->nn_param.conv2d.dilation[1] = 1; + node[61]->nn_param.conv2d.multiplier = 0; + node[61]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[61]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[61]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - cpm/conv/conv.1/Relu_output_0_35 + var - node[62] + name - cpm/conv/conv.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[62], VSI_NN_OP_RELU, 1, 1, 35); + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32 + var - node[63] + name - initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[63], VSI_NN_OP_CONV2D, 3, 1, 32); + node[63]->nn_param.conv2d.ksize[0] = 3; + node[63]->nn_param.conv2d.ksize[1] = 3; + node[63]->nn_param.conv2d.weights = 128; + node[63]->nn_param.conv2d.stride[0] = 1; + node[63]->nn_param.conv2d.stride[1] = 1; + node[63]->nn_param.conv2d.pad[0] = 1; + node[63]->nn_param.conv2d.pad[1] = 1; + node[63]->nn_param.conv2d.pad[2] = 1; + node[63]->nn_param.conv2d.pad[3] = 1; + node[63]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[63]->nn_param.conv2d.group = 1; + node[63]->nn_param.conv2d.dilation[0] = 1; + node[63]->nn_param.conv2d.dilation[1] = 1; + node[63]->nn_param.conv2d.multiplier = 0; + node[63]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[63]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[63]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28 + var - node[64] + name - initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[64], VSI_NN_OP_RELU, 1, 1, 28); + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25 + var - node[65] + name - initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[65], VSI_NN_OP_CONV2D, 3, 1, 25); + node[65]->nn_param.conv2d.ksize[0] = 3; + node[65]->nn_param.conv2d.ksize[1] = 3; + node[65]->nn_param.conv2d.weights = 128; + node[65]->nn_param.conv2d.stride[0] = 1; + node[65]->nn_param.conv2d.stride[1] = 1; + node[65]->nn_param.conv2d.pad[0] = 1; + node[65]->nn_param.conv2d.pad[1] = 1; + node[65]->nn_param.conv2d.pad[2] = 1; + node[65]->nn_param.conv2d.pad[3] = 1; + node[65]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[65]->nn_param.conv2d.group = 1; + node[65]->nn_param.conv2d.dilation[0] = 1; + node[65]->nn_param.conv2d.dilation[1] = 1; + node[65]->nn_param.conv2d.multiplier = 0; + node[65]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[65]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[65]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22 + var - node[66] + name - initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[66], VSI_NN_OP_RELU, 1, 1, 22); + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19 + var - node[67] + name - initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[67], VSI_NN_OP_CONV2D, 3, 1, 19); + node[67]->nn_param.conv2d.ksize[0] = 3; + node[67]->nn_param.conv2d.ksize[1] = 3; + node[67]->nn_param.conv2d.weights = 128; + node[67]->nn_param.conv2d.stride[0] = 1; + node[67]->nn_param.conv2d.stride[1] = 1; + node[67]->nn_param.conv2d.pad[0] = 1; + node[67]->nn_param.conv2d.pad[1] = 1; + node[67]->nn_param.conv2d.pad[2] = 1; + node[67]->nn_param.conv2d.pad[3] = 1; + node[67]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[67]->nn_param.conv2d.group = 1; + node[67]->nn_param.conv2d.dilation[0] = 1; + node[67]->nn_param.conv2d.dilation[1] = 1; + node[67]->nn_param.conv2d.multiplier = 0; + node[67]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[67]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[67]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18 + var - node[68] + name - initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[68], VSI_NN_OP_RELU, 1, 1, 18); + + /*----------------------------------------- + lid - initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14 + var - node[69] + name - initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[69], VSI_NN_OP_CONV2D, 3, 1, 14); + node[69]->nn_param.conv2d.ksize[0] = 1; + node[69]->nn_param.conv2d.ksize[1] = 1; + node[69]->nn_param.conv2d.weights = 512; + node[69]->nn_param.conv2d.stride[0] = 1; + node[69]->nn_param.conv2d.stride[1] = 1; + node[69]->nn_param.conv2d.pad[0] = 0; + node[69]->nn_param.conv2d.pad[1] = 0; + node[69]->nn_param.conv2d.pad[2] = 0; + node[69]->nn_param.conv2d.pad[3] = 0; + node[69]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[69]->nn_param.conv2d.group = 1; + node[69]->nn_param.conv2d.dilation[0] = 1; + node[69]->nn_param.conv2d.dilation[1] = 1; + node[69]->nn_param.conv2d.multiplier = 0; + node[69]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[69]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[69]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15 + var - node[70] + name - initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 512] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[70], VSI_NN_OP_CONV2D, 3, 1, 15); + node[70]->nn_param.conv2d.ksize[0] = 1; + node[70]->nn_param.conv2d.ksize[1] = 1; + node[70]->nn_param.conv2d.weights = 512; + node[70]->nn_param.conv2d.stride[0] = 1; + node[70]->nn_param.conv2d.stride[1] = 1; + node[70]->nn_param.conv2d.pad[0] = 0; + node[70]->nn_param.conv2d.pad[1] = 0; + node[70]->nn_param.conv2d.pad[2] = 0; + node[70]->nn_param.conv2d.pad[3] = 0; + node[70]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[70]->nn_param.conv2d.group = 1; + node[70]->nn_param.conv2d.dilation[0] = 1; + node[70]->nn_param.conv2d.dilation[1] = 1; + node[70]->nn_param.conv2d.multiplier = 0; + node[70]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[70]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[70]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10 + var - node[71] + name - initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[71], VSI_NN_OP_RELU, 1, 1, 10); + + /*----------------------------------------- + lid - initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11 + var - node[72] + name - initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0 + operation - relu + input - [53, 32, 512, 1] + output - [53, 32, 512, 1] + -----------------------------------------*/ + NEW_VXNODE(node[72], VSI_NN_OP_RELU, 1, 1, 11); + + /*----------------------------------------- + lid - onnx//Concat_348_6 + var - node[73] + name - onnx//Concat_348 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 38] + output - [53, 32, 38, 1] + -----------------------------------------*/ + NEW_VXNODE(node[73], VSI_NN_OP_CONV2D, 3, 1, 6); + node[73]->nn_param.conv2d.ksize[0] = 1; + node[73]->nn_param.conv2d.ksize[1] = 1; + node[73]->nn_param.conv2d.weights = 38; + node[73]->nn_param.conv2d.stride[0] = 1; + node[73]->nn_param.conv2d.stride[1] = 1; + node[73]->nn_param.conv2d.pad[0] = 0; + node[73]->nn_param.conv2d.pad[1] = 0; + node[73]->nn_param.conv2d.pad[2] = 0; + node[73]->nn_param.conv2d.pad[3] = 0; + node[73]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[73]->nn_param.conv2d.group = 1; + node[73]->nn_param.conv2d.dilation[0] = 1; + node[73]->nn_param.conv2d.dilation[1] = 1; + node[73]->nn_param.conv2d.multiplier = 0; + node[73]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[73]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[73]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - onnx//Concat_345_7 + var - node[74] + name - onnx//Concat_345 + operation - convolution + input - [53, 32, 512, 1] + filter - [1, 1, 512, 19] + output - [53, 32, 19, 1] + -----------------------------------------*/ + NEW_VXNODE(node[74], VSI_NN_OP_CONV2D, 3, 1, 7); + node[74]->nn_param.conv2d.ksize[0] = 1; + node[74]->nn_param.conv2d.ksize[1] = 1; + node[74]->nn_param.conv2d.weights = 19; + node[74]->nn_param.conv2d.stride[0] = 1; + node[74]->nn_param.conv2d.stride[1] = 1; + node[74]->nn_param.conv2d.pad[0] = 0; + node[74]->nn_param.conv2d.pad[1] = 0; + node[74]->nn_param.conv2d.pad[2] = 0; + node[74]->nn_param.conv2d.pad[3] = 0; + node[74]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[74]->nn_param.conv2d.group = 1; + node[74]->nn_param.conv2d.dilation[0] = 1; + node[74]->nn_param.conv2d.dilation[1] = 1; + node[74]->nn_param.conv2d.multiplier = 0; + node[74]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[74]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[74]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - Concat_output_0_73 + var - node[75] + name - Concat_output_0 + operation - concat + input - [53, 32, 128, 1] + [53, 32, 19, 1] + [53, 32, 38, 1] + output - [53, 32, 185, 1] + -----------------------------------------*/ + NEW_VXNODE(node[75], VSI_NN_OP_CONCAT, 3, 1, 73); + node[75]->nn_param.concat.axis = 2; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69 + var - node[76] + name - refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 185, 1] + filter - [1, 1, 185, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[76], VSI_NN_OP_CONV2D, 3, 1, 69); + node[76]->nn_param.conv2d.ksize[0] = 1; + node[76]->nn_param.conv2d.ksize[1] = 1; + node[76]->nn_param.conv2d.weights = 128; + node[76]->nn_param.conv2d.stride[0] = 1; + node[76]->nn_param.conv2d.stride[1] = 1; + node[76]->nn_param.conv2d.pad[0] = 0; + node[76]->nn_param.conv2d.pad[1] = 0; + node[76]->nn_param.conv2d.pad[2] = 0; + node[76]->nn_param.conv2d.pad[3] = 0; + node[76]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[76]->nn_param.conv2d.group = 1; + node[76]->nn_param.conv2d.dilation[0] = 1; + node[76]->nn_param.conv2d.dilation[1] = 1; + node[76]->nn_param.conv2d.multiplier = 0; + node[76]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[76]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[76]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64 + var - node[77] + name - refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[77], VSI_NN_OP_RELU, 1, 1, 64); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74 + var - node[78] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[78], VSI_NN_OP_CONV2D, 3, 1, 74); + node[78]->nn_param.conv2d.ksize[0] = 3; + node[78]->nn_param.conv2d.ksize[1] = 3; + node[78]->nn_param.conv2d.weights = 128; + node[78]->nn_param.conv2d.stride[0] = 1; + node[78]->nn_param.conv2d.stride[1] = 1; + node[78]->nn_param.conv2d.pad[0] = 1; + node[78]->nn_param.conv2d.pad[1] = 1; + node[78]->nn_param.conv2d.pad[2] = 1; + node[78]->nn_param.conv2d.pad[3] = 1; + node[78]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[78]->nn_param.conv2d.group = 1; + node[78]->nn_param.conv2d.dilation[0] = 1; + node[78]->nn_param.conv2d.dilation[1] = 1; + node[78]->nn_param.conv2d.multiplier = 0; + node[78]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[78]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[78]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70 + var - node[79] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[79], VSI_NN_OP_RELU, 1, 1, 70); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66 + var - node[80] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[80], VSI_NN_OP_CONV2D, 3, 1, 66); + node[80]->nn_param.conv2d.ksize[0] = 5; + node[80]->nn_param.conv2d.ksize[1] = 5; + node[80]->nn_param.conv2d.weights = 128; + node[80]->nn_param.conv2d.stride[0] = 1; + node[80]->nn_param.conv2d.stride[1] = 1; + node[80]->nn_param.conv2d.pad[0] = 2; + node[80]->nn_param.conv2d.pad[1] = 2; + node[80]->nn_param.conv2d.pad[2] = 2; + node[80]->nn_param.conv2d.pad[3] = 2; + node[80]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[80]->nn_param.conv2d.group = 1; + node[80]->nn_param.conv2d.dilation[0] = 1; + node[80]->nn_param.conv2d.dilation[1] = 1; + node[80]->nn_param.conv2d.multiplier = 0; + node[80]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[80]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[80]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65 + var - node[81] + name - refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[81], VSI_NN_OP_RELU, 1, 1, 65); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.0/Add_output_0_60 + var - node[82] + name - refinement_stages.0/trunk/trunk.0/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[82], VSI_NN_OP_ADD, 2, 1, 60); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56 + var - node[83] + name - refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[83], VSI_NN_OP_CONV2D, 3, 1, 56); + node[83]->nn_param.conv2d.ksize[0] = 1; + node[83]->nn_param.conv2d.ksize[1] = 1; + node[83]->nn_param.conv2d.weights = 128; + node[83]->nn_param.conv2d.stride[0] = 1; + node[83]->nn_param.conv2d.stride[1] = 1; + node[83]->nn_param.conv2d.pad[0] = 0; + node[83]->nn_param.conv2d.pad[1] = 0; + node[83]->nn_param.conv2d.pad[2] = 0; + node[83]->nn_param.conv2d.pad[3] = 0; + node[83]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[83]->nn_param.conv2d.group = 1; + node[83]->nn_param.conv2d.dilation[0] = 1; + node[83]->nn_param.conv2d.dilation[1] = 1; + node[83]->nn_param.conv2d.multiplier = 0; + node[83]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[83]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[83]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51 + var - node[84] + name - refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[84], VSI_NN_OP_RELU, 1, 1, 51); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61 + var - node[85] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[85], VSI_NN_OP_CONV2D, 3, 1, 61); + node[85]->nn_param.conv2d.ksize[0] = 3; + node[85]->nn_param.conv2d.ksize[1] = 3; + node[85]->nn_param.conv2d.weights = 128; + node[85]->nn_param.conv2d.stride[0] = 1; + node[85]->nn_param.conv2d.stride[1] = 1; + node[85]->nn_param.conv2d.pad[0] = 1; + node[85]->nn_param.conv2d.pad[1] = 1; + node[85]->nn_param.conv2d.pad[2] = 1; + node[85]->nn_param.conv2d.pad[3] = 1; + node[85]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[85]->nn_param.conv2d.group = 1; + node[85]->nn_param.conv2d.dilation[0] = 1; + node[85]->nn_param.conv2d.dilation[1] = 1; + node[85]->nn_param.conv2d.multiplier = 0; + node[85]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[85]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[85]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57 + var - node[86] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[86], VSI_NN_OP_RELU, 1, 1, 57); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53 + var - node[87] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[87], VSI_NN_OP_CONV2D, 3, 1, 53); + node[87]->nn_param.conv2d.ksize[0] = 5; + node[87]->nn_param.conv2d.ksize[1] = 5; + node[87]->nn_param.conv2d.weights = 128; + node[87]->nn_param.conv2d.stride[0] = 1; + node[87]->nn_param.conv2d.stride[1] = 1; + node[87]->nn_param.conv2d.pad[0] = 2; + node[87]->nn_param.conv2d.pad[1] = 2; + node[87]->nn_param.conv2d.pad[2] = 2; + node[87]->nn_param.conv2d.pad[3] = 2; + node[87]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[87]->nn_param.conv2d.group = 1; + node[87]->nn_param.conv2d.dilation[0] = 1; + node[87]->nn_param.conv2d.dilation[1] = 1; + node[87]->nn_param.conv2d.multiplier = 0; + node[87]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[87]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[87]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52 + var - node[88] + name - refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[88], VSI_NN_OP_RELU, 1, 1, 52); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.1/Add_output_0_47 + var - node[89] + name - refinement_stages.0/trunk/trunk.1/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[89], VSI_NN_OP_ADD, 2, 1, 47); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43 + var - node[90] + name - refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[90], VSI_NN_OP_CONV2D, 3, 1, 43); + node[90]->nn_param.conv2d.ksize[0] = 1; + node[90]->nn_param.conv2d.ksize[1] = 1; + node[90]->nn_param.conv2d.weights = 128; + node[90]->nn_param.conv2d.stride[0] = 1; + node[90]->nn_param.conv2d.stride[1] = 1; + node[90]->nn_param.conv2d.pad[0] = 0; + node[90]->nn_param.conv2d.pad[1] = 0; + node[90]->nn_param.conv2d.pad[2] = 0; + node[90]->nn_param.conv2d.pad[3] = 0; + node[90]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[90]->nn_param.conv2d.group = 1; + node[90]->nn_param.conv2d.dilation[0] = 1; + node[90]->nn_param.conv2d.dilation[1] = 1; + node[90]->nn_param.conv2d.multiplier = 0; + node[90]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[90]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[90]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39 + var - node[91] + name - refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[91], VSI_NN_OP_RELU, 1, 1, 39); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48 + var - node[92] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[92], VSI_NN_OP_CONV2D, 3, 1, 48); + node[92]->nn_param.conv2d.ksize[0] = 3; + node[92]->nn_param.conv2d.ksize[1] = 3; + node[92]->nn_param.conv2d.weights = 128; + node[92]->nn_param.conv2d.stride[0] = 1; + node[92]->nn_param.conv2d.stride[1] = 1; + node[92]->nn_param.conv2d.pad[0] = 1; + node[92]->nn_param.conv2d.pad[1] = 1; + node[92]->nn_param.conv2d.pad[2] = 1; + node[92]->nn_param.conv2d.pad[3] = 1; + node[92]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[92]->nn_param.conv2d.group = 1; + node[92]->nn_param.conv2d.dilation[0] = 1; + node[92]->nn_param.conv2d.dilation[1] = 1; + node[92]->nn_param.conv2d.multiplier = 0; + node[92]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[92]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[92]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44 + var - node[93] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[93], VSI_NN_OP_RELU, 1, 1, 44); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41 + var - node[94] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[94], VSI_NN_OP_CONV2D, 3, 1, 41); + node[94]->nn_param.conv2d.ksize[0] = 5; + node[94]->nn_param.conv2d.ksize[1] = 5; + node[94]->nn_param.conv2d.weights = 128; + node[94]->nn_param.conv2d.stride[0] = 1; + node[94]->nn_param.conv2d.stride[1] = 1; + node[94]->nn_param.conv2d.pad[0] = 2; + node[94]->nn_param.conv2d.pad[1] = 2; + node[94]->nn_param.conv2d.pad[2] = 2; + node[94]->nn_param.conv2d.pad[3] = 2; + node[94]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[94]->nn_param.conv2d.group = 1; + node[94]->nn_param.conv2d.dilation[0] = 1; + node[94]->nn_param.conv2d.dilation[1] = 1; + node[94]->nn_param.conv2d.multiplier = 0; + node[94]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[94]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[94]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40 + var - node[95] + name - refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[95], VSI_NN_OP_RELU, 1, 1, 40); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.2/Add_output_0_36 + var - node[96] + name - refinement_stages.0/trunk/trunk.2/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[96], VSI_NN_OP_ADD, 2, 1, 36); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33 + var - node[97] + name - refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[97], VSI_NN_OP_CONV2D, 3, 1, 33); + node[97]->nn_param.conv2d.ksize[0] = 1; + node[97]->nn_param.conv2d.ksize[1] = 1; + node[97]->nn_param.conv2d.weights = 128; + node[97]->nn_param.conv2d.stride[0] = 1; + node[97]->nn_param.conv2d.stride[1] = 1; + node[97]->nn_param.conv2d.pad[0] = 0; + node[97]->nn_param.conv2d.pad[1] = 0; + node[97]->nn_param.conv2d.pad[2] = 0; + node[97]->nn_param.conv2d.pad[3] = 0; + node[97]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[97]->nn_param.conv2d.group = 1; + node[97]->nn_param.conv2d.dilation[0] = 1; + node[97]->nn_param.conv2d.dilation[1] = 1; + node[97]->nn_param.conv2d.multiplier = 0; + node[97]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[97]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[97]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29 + var - node[98] + name - refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[98], VSI_NN_OP_RELU, 1, 1, 29); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37 + var - node[99] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[99], VSI_NN_OP_CONV2D, 3, 1, 37); + node[99]->nn_param.conv2d.ksize[0] = 3; + node[99]->nn_param.conv2d.ksize[1] = 3; + node[99]->nn_param.conv2d.weights = 128; + node[99]->nn_param.conv2d.stride[0] = 1; + node[99]->nn_param.conv2d.stride[1] = 1; + node[99]->nn_param.conv2d.pad[0] = 1; + node[99]->nn_param.conv2d.pad[1] = 1; + node[99]->nn_param.conv2d.pad[2] = 1; + node[99]->nn_param.conv2d.pad[3] = 1; + node[99]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[99]->nn_param.conv2d.group = 1; + node[99]->nn_param.conv2d.dilation[0] = 1; + node[99]->nn_param.conv2d.dilation[1] = 1; + node[99]->nn_param.conv2d.multiplier = 0; + node[99]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[99]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[99]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34 + var - node[100] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[100], VSI_NN_OP_RELU, 1, 1, 34); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31 + var - node[101] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[101], VSI_NN_OP_CONV2D, 3, 1, 31); + node[101]->nn_param.conv2d.ksize[0] = 5; + node[101]->nn_param.conv2d.ksize[1] = 5; + node[101]->nn_param.conv2d.weights = 128; + node[101]->nn_param.conv2d.stride[0] = 1; + node[101]->nn_param.conv2d.stride[1] = 1; + node[101]->nn_param.conv2d.pad[0] = 2; + node[101]->nn_param.conv2d.pad[1] = 2; + node[101]->nn_param.conv2d.pad[2] = 2; + node[101]->nn_param.conv2d.pad[3] = 2; + node[101]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[101]->nn_param.conv2d.group = 1; + node[101]->nn_param.conv2d.dilation[0] = 1; + node[101]->nn_param.conv2d.dilation[1] = 1; + node[101]->nn_param.conv2d.multiplier = 0; + node[101]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[101]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[101]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30 + var - node[102] + name - refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[102], VSI_NN_OP_RELU, 1, 1, 30); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.3/Add_output_0_26 + var - node[103] + name - refinement_stages.0/trunk/trunk.3/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[103], VSI_NN_OP_ADD, 2, 1, 26); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23 + var - node[104] + name - refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[104], VSI_NN_OP_CONV2D, 3, 1, 23); + node[104]->nn_param.conv2d.ksize[0] = 1; + node[104]->nn_param.conv2d.ksize[1] = 1; + node[104]->nn_param.conv2d.weights = 128; + node[104]->nn_param.conv2d.stride[0] = 1; + node[104]->nn_param.conv2d.stride[1] = 1; + node[104]->nn_param.conv2d.pad[0] = 0; + node[104]->nn_param.conv2d.pad[1] = 0; + node[104]->nn_param.conv2d.pad[2] = 0; + node[104]->nn_param.conv2d.pad[3] = 0; + node[104]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[104]->nn_param.conv2d.group = 1; + node[104]->nn_param.conv2d.dilation[0] = 1; + node[104]->nn_param.conv2d.dilation[1] = 1; + node[104]->nn_param.conv2d.multiplier = 0; + node[104]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[104]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[104]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20 + var - node[105] + name - refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[105], VSI_NN_OP_RELU, 1, 1, 20); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27 + var - node[106] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [3, 3, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[106], VSI_NN_OP_CONV2D, 3, 1, 27); + node[106]->nn_param.conv2d.ksize[0] = 3; + node[106]->nn_param.conv2d.ksize[1] = 3; + node[106]->nn_param.conv2d.weights = 128; + node[106]->nn_param.conv2d.stride[0] = 1; + node[106]->nn_param.conv2d.stride[1] = 1; + node[106]->nn_param.conv2d.pad[0] = 1; + node[106]->nn_param.conv2d.pad[1] = 1; + node[106]->nn_param.conv2d.pad[2] = 1; + node[106]->nn_param.conv2d.pad[3] = 1; + node[106]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[106]->nn_param.conv2d.group = 1; + node[106]->nn_param.conv2d.dilation[0] = 1; + node[106]->nn_param.conv2d.dilation[1] = 1; + node[106]->nn_param.conv2d.multiplier = 0; + node[106]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[106]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[106]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24 + var - node[107] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[107], VSI_NN_OP_RELU, 1, 1, 24); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21 + var - node[108] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [5, 5, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[108], VSI_NN_OP_CONV2D, 3, 1, 21); + node[108]->nn_param.conv2d.ksize[0] = 5; + node[108]->nn_param.conv2d.ksize[1] = 5; + node[108]->nn_param.conv2d.weights = 128; + node[108]->nn_param.conv2d.stride[0] = 1; + node[108]->nn_param.conv2d.stride[1] = 1; + node[108]->nn_param.conv2d.pad[0] = 2; + node[108]->nn_param.conv2d.pad[1] = 2; + node[108]->nn_param.conv2d.pad[2] = 2; + node[108]->nn_param.conv2d.pad[3] = 2; + node[108]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[108]->nn_param.conv2d.group = 1; + node[108]->nn_param.conv2d.dilation[0] = 1; + node[108]->nn_param.conv2d.dilation[1] = 1; + node[108]->nn_param.conv2d.multiplier = 0; + node[108]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[108]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[108]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17 + var - node[109] + name - refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[109], VSI_NN_OP_RELU, 1, 1, 17); + + /*----------------------------------------- + lid - refinement_stages.0/trunk/trunk.4/Add_output_0_16 + var - node[110] + name - refinement_stages.0/trunk/trunk.4/Add_output_0 + operation - add + input - [53, 32, 128, 1] + [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[110], VSI_NN_OP_ADD, 2, 1, 16); + + /*----------------------------------------- + lid - refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12 + var - node[111] + name - refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[111], VSI_NN_OP_CONV2D, 3, 1, 12); + node[111]->nn_param.conv2d.ksize[0] = 1; + node[111]->nn_param.conv2d.ksize[1] = 1; + node[111]->nn_param.conv2d.weights = 128; + node[111]->nn_param.conv2d.stride[0] = 1; + node[111]->nn_param.conv2d.stride[1] = 1; + node[111]->nn_param.conv2d.pad[0] = 0; + node[111]->nn_param.conv2d.pad[1] = 0; + node[111]->nn_param.conv2d.pad[2] = 0; + node[111]->nn_param.conv2d.pad[3] = 0; + node[111]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[111]->nn_param.conv2d.group = 1; + node[111]->nn_param.conv2d.dilation[0] = 1; + node[111]->nn_param.conv2d.dilation[1] = 1; + node[111]->nn_param.conv2d.multiplier = 0; + node[111]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[111]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[111]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13 + var - node[112] + name - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 128] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[112], VSI_NN_OP_CONV2D, 3, 1, 13); + node[112]->nn_param.conv2d.ksize[0] = 1; + node[112]->nn_param.conv2d.ksize[1] = 1; + node[112]->nn_param.conv2d.weights = 128; + node[112]->nn_param.conv2d.stride[0] = 1; + node[112]->nn_param.conv2d.stride[1] = 1; + node[112]->nn_param.conv2d.pad[0] = 0; + node[112]->nn_param.conv2d.pad[1] = 0; + node[112]->nn_param.conv2d.pad[2] = 0; + node[112]->nn_param.conv2d.pad[3] = 0; + node[112]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[112]->nn_param.conv2d.group = 1; + node[112]->nn_param.conv2d.dilation[0] = 1; + node[112]->nn_param.conv2d.dilation[1] = 1; + node[112]->nn_param.conv2d.multiplier = 0; + node[112]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[112]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[112]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8 + var - node[113] + name - refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[113], VSI_NN_OP_RELU, 1, 1, 8); + + /*----------------------------------------- + lid - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9 + var - node[114] + name - refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0 + operation - relu + input - [53, 32, 128, 1] + output - [53, 32, 128, 1] + -----------------------------------------*/ + NEW_VXNODE(node[114], VSI_NN_OP_RELU, 1, 1, 9); + + /*----------------------------------------- + lid - 400_4 + var - node[115] + name - 400 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 38] + output - [53, 32, 38, 1] + -----------------------------------------*/ + NEW_VXNODE(node[115], VSI_NN_OP_CONV2D, 3, 1, 4); + node[115]->nn_param.conv2d.ksize[0] = 1; + node[115]->nn_param.conv2d.ksize[1] = 1; + node[115]->nn_param.conv2d.weights = 38; + node[115]->nn_param.conv2d.stride[0] = 1; + node[115]->nn_param.conv2d.stride[1] = 1; + node[115]->nn_param.conv2d.pad[0] = 0; + node[115]->nn_param.conv2d.pad[1] = 0; + node[115]->nn_param.conv2d.pad[2] = 0; + node[115]->nn_param.conv2d.pad[3] = 0; + node[115]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[115]->nn_param.conv2d.group = 1; + node[115]->nn_param.conv2d.dilation[0] = 1; + node[115]->nn_param.conv2d.dilation[1] = 1; + node[115]->nn_param.conv2d.multiplier = 0; + node[115]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[115]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[115]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + /*----------------------------------------- + lid - 397_5 + var - node[116] + name - 397 + operation - convolution + input - [53, 32, 128, 1] + filter - [1, 1, 128, 19] + output - [53, 32, 19, 1] + -----------------------------------------*/ + NEW_VXNODE(node[116], VSI_NN_OP_CONV2D, 3, 1, 5); + node[116]->nn_param.conv2d.ksize[0] = 1; + node[116]->nn_param.conv2d.ksize[1] = 1; + node[116]->nn_param.conv2d.weights = 19; + node[116]->nn_param.conv2d.stride[0] = 1; + node[116]->nn_param.conv2d.stride[1] = 1; + node[116]->nn_param.conv2d.pad[0] = 0; + node[116]->nn_param.conv2d.pad[1] = 0; + node[116]->nn_param.conv2d.pad[2] = 0; + node[116]->nn_param.conv2d.pad[3] = 0; + node[116]->nn_param.conv2d.pad_mode = VSI_NN_PAD_MODE_CONSTANT; + node[116]->nn_param.conv2d.group = 1; + node[116]->nn_param.conv2d.dilation[0] = 1; + node[116]->nn_param.conv2d.dilation[1] = 1; + node[116]->nn_param.conv2d.multiplier = 0; + node[116]->vx_param.overflow_policy = VX_CONVERT_POLICY_SATURATE; + node[116]->vx_param.rounding_policy = VX_ROUND_POLICY_TO_NEAREST_EVEN; + node[116]->vx_param.down_scale_size_rounding = VX_CONVOLUTIONAL_NETWORK_DS_SIZE_ROUNDING_FLOOR; + + } + else + { + NEW_VXNODE(node[0], VSI_NN_OP_NBG, 1, 4, 0); + node[0]->nn_param.nbg.type = VSI_NN_NBG_FILE; + node[0]->nn_param.nbg.url = data_file_name; + + } + +/*----------------------------------------- + Tensor initialize + -----------------------------------------*/ + attr.dtype.fmt = VSI_NN_DIM_FMT_NCHW; + /* @attach_onnx//Concat_345/out0_0:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 19; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_NORM_TENSOR(norm_tensor[0], attr, VSI_NN_TYPE_INT16); + + /* @attach_onnx//Concat_348/out0_1:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 38; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_NORM_TENSOR(norm_tensor[1], attr, VSI_NN_TYPE_INT16); + + /* @attach_397/out0_2:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 19; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_NORM_TENSOR(norm_tensor[2], attr, VSI_NN_TYPE_INT16); + + /* @attach_400/out0_3:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 53; + attr.size[1] = 32; + attr.size[2] = 38; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_NORM_TENSOR(norm_tensor[3], attr, VSI_NN_TYPE_INT16); + + /* @input.1_121:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 424; + attr.size[1] = 256; + attr.size[2] = 3; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_NORM_TENSOR(norm_tensor[4], attr, VSI_NN_TYPE_INT16); + + + + if( !inference_with_nbg ) + { + /* @model/model.0/model.0.0/Conv_output_0_120:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 3; + attr.size[3] = 32; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[0], attr, VSI_NN_TYPE_INT16, 1639880, 1728); + + /* @model/model.0/model.0.0/Conv_output_0_120:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 32; + attr.dim_num = 1; + attr.dtype.fl = 31; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[1], attr, VSI_NN_TYPE_INT64, 1639624, 256); + + /* @model/model.1/model.1.0/Conv_output_0_118:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 32; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[2], attr, VSI_NN_TYPE_INT16, 1641864, 576); + + /* @model/model.1/model.1.0/Conv_output_0_118:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 32; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[3], attr, VSI_NN_TYPE_INT64, 1641608, 256); + + /* @model/model.1/model.1.3/Conv_output_0_116:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 32; + attr.size[3] = 64; + attr.dim_num = 4; + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[4], attr, VSI_NN_TYPE_INT16, 1642952, 4096); + + /* @model/model.1/model.1.3/Conv_output_0_116:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 64; + attr.dim_num = 1; + attr.dtype.fl = 22; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[5], attr, VSI_NN_TYPE_INT64, 1642440, 512); + + /* @model/model.2/model.2.0/Conv_output_0_114:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 64; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 9; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[6], attr, VSI_NN_TYPE_INT16, 2730952, 1152); + + /* @model/model.2/model.2.0/Conv_output_0_114:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 64; + attr.dim_num = 1; + attr.dtype.fl = 21; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[7], attr, VSI_NN_TYPE_INT64, 2730440, 512); + + /* @model/model.2/model.2.3/Conv_output_0_112:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 64; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[8], attr, VSI_NN_TYPE_INT16, 2733128, 16384); + + /* @model/model.2/model.2.3/Conv_output_0_112:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[9], attr, VSI_NN_TYPE_INT64, 2732104, 1024); + + /* @model/model.3/model.3.0/Conv_output_0_110:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 11; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[10], attr, VSI_NN_TYPE_INT16, 2750536, 2304); + + /* @model/model.3/model.3.0/Conv_output_0_110:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 24; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[11], attr, VSI_NN_TYPE_INT64, 2749512, 1024); + + /* @model/model.3/model.3.3/Conv_output_0_108:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[12], attr, VSI_NN_TYPE_INT16, 2753864, 32768); + + /* @model/model.3/model.3.3/Conv_output_0_108:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[13], attr, VSI_NN_TYPE_INT64, 2752840, 1024); + + /* @model/model.4/model.4.0/Conv_output_0_106:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[14], attr, VSI_NN_TYPE_INT16, 2787656, 2304); + + /* @model/model.4/model.4.0/Conv_output_0_106:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[15], attr, VSI_NN_TYPE_INT64, 2786632, 1024); + + /* @model/model.4/model.4.3/Conv_output_0_104:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 256; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[16], attr, VSI_NN_TYPE_INT16, 2792008, 65536); + + /* @model/model.4/model.4.3/Conv_output_0_104:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.fl = 25; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[17], attr, VSI_NN_TYPE_INT64, 2789960, 2048); + + /* @model/model.5/model.5.0/Conv_output_0_102:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 256; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 11; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[18], attr, VSI_NN_TYPE_INT16, 2859592, 4608); + + /* @model/model.5/model.5.0/Conv_output_0_102:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.fl = 25; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[19], attr, VSI_NN_TYPE_INT64, 2857544, 2048); + + /* @model/model.5/model.5.3/Conv_output_0_100:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 256; + attr.size[3] = 256; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[20], attr, VSI_NN_TYPE_INT16, 2866248, 131072); + + /* @model/model.5/model.5.3/Conv_output_0_100:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[21], attr, VSI_NN_TYPE_INT64, 2864200, 2048); + + /* @model/model.6/model.6.0/Conv_output_0_98:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 256; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[22], attr, VSI_NN_TYPE_INT16, 2999368, 4608); + + /* @model/model.6/model.6.0/Conv_output_0_98:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 256; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[23], attr, VSI_NN_TYPE_INT64, 2997320, 2048); + + /* @model/model.6/model.6.3/Conv_output_0_96:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 256; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[24], attr, VSI_NN_TYPE_INT16, 3008072, 262144); + + /* @model/model.6/model.6.3/Conv_output_0_96:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 25; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[25], attr, VSI_NN_TYPE_INT64, 3003976, 4096); + + /* @model/model.7/model.7.0/Conv_output_0_94:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[26], attr, VSI_NN_TYPE_INT16, 3274312, 25600); + + /* @model/model.7/model.7.0/Conv_output_0_94:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[27], attr, VSI_NN_TYPE_INT64, 3270216, 4096); + + /* @model/model.7/model.7.3/Conv_output_0_92:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[28], attr, VSI_NN_TYPE_INT16, 3304008, 524288); + + /* @model/model.7/model.7.3/Conv_output_0_92:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 27; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[29], attr, VSI_NN_TYPE_INT64, 3299912, 4096); + + /* @model/model.8/model.8.0/Conv_output_0_90:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[30], attr, VSI_NN_TYPE_INT16, 3832392, 9216); + + /* @model/model.8/model.8.0/Conv_output_0_90:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[31], attr, VSI_NN_TYPE_INT64, 3828296, 4096); + + /* @model/model.8/model.8.3/Conv_output_0_88:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[32], attr, VSI_NN_TYPE_INT16, 3845704, 524288); + + /* @model/model.8/model.8.3/Conv_output_0_88:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[33], attr, VSI_NN_TYPE_INT64, 3841608, 4096); + + /* @model/model.9/model.9.0/Conv_output_0_86:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[34], attr, VSI_NN_TYPE_INT16, 4374088, 9216); + + /* @model/model.9/model.9.0/Conv_output_0_86:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[35], attr, VSI_NN_TYPE_INT64, 4369992, 4096); + + /* @model/model.9/model.9.3/Conv_output_0_84:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[36], attr, VSI_NN_TYPE_INT16, 4387400, 524288); + + /* @model/model.9/model.9.3/Conv_output_0_84:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[37], attr, VSI_NN_TYPE_INT64, 4383304, 4096); + + /* @model/model.10/model.10.0/Conv_output_0_82:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[38], attr, VSI_NN_TYPE_INT16, 1651144, 9216); + + /* @model/model.10/model.10.0/Conv_output_0_82:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 26; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[39], attr, VSI_NN_TYPE_INT64, 1647048, 4096); + + /* @model/model.10/model.10.3/Conv_output_0_75:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[40], attr, VSI_NN_TYPE_INT16, 1664456, 524288); + + /* @model/model.10/model.10.3/Conv_output_0_75:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[41], attr, VSI_NN_TYPE_INT64, 1660360, 4096); + + /* @model/model.11/model.11.0/Conv_output_0_67:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 512; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 11; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[42], attr, VSI_NN_TYPE_INT16, 2192840, 9216); + + /* @model/model.11/model.11.0/Conv_output_0_67:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 25; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[43], attr, VSI_NN_TYPE_INT64, 2188744, 4096); + + /* @model/model.11/model.11.3/Conv_output_0_58:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[44], attr, VSI_NN_TYPE_INT16, 2206152, 524288); + + /* @model/model.11/model.11.3/Conv_output_0_58:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 25; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[45], attr, VSI_NN_TYPE_INT64, 2202056, 4096); + + /* @cpm/align/align.0/Conv_output_0_49:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[46], attr, VSI_NN_TYPE_INT16, 16072, 131072); + + /* @cpm/align/align.0/Conv_output_0_49:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 30; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[47], attr, VSI_NN_TYPE_INT64, 15048, 1024); + + /* @cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = 17; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[48], attr, VSI_NN_TYPE_INT16, 444104, 2304); + + /* @cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:bias + @cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:bias + @cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:bias + @cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:bias + @cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:bias + @cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 32; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[49], attr, VSI_NN_TYPE_INT64, 443080, 1024); + + /* @cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[50], attr, VSI_NN_TYPE_INT16, 446408, 32768); + + /* @cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:weight + @cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 1; + attr.dim_num = 4; + attr.dtype.fl = -268; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[51], attr, VSI_NN_TYPE_INT16, 479176, 2304); + + /* @cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:weight + @cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = -268; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[52], attr, VSI_NN_TYPE_INT16, 446408, 32768); + + /* @cpm/conv/conv.0/Conv_output_0_38:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[53], attr, VSI_NN_TYPE_INT16, 148168, 294912); + + /* @cpm/conv/conv.0/Conv_output_0_38:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 31; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[54], attr, VSI_NN_TYPE_INT64, 147144, 1024); + + /* @initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[55], attr, VSI_NN_TYPE_INT16, 752840, 294912); + + /* @initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[56], attr, VSI_NN_TYPE_INT64, 751816, 1024); + + /* @initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[57], attr, VSI_NN_TYPE_INT16, 1048776, 294912); + + /* @initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 30; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[58], attr, VSI_NN_TYPE_INT64, 1047752, 1024); + + /* @initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[59], attr, VSI_NN_TYPE_INT16, 1344712, 294912); + + /* @initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 30; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[60], attr, VSI_NN_TYPE_INT64, 1343688, 1024); + + /* @initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[61], attr, VSI_NN_TYPE_INT16, 620744, 131072); + + /* @initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[62], attr, VSI_NN_TYPE_INT64, 616648, 4096); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 512; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[63], attr, VSI_NN_TYPE_INT16, 485576, 131072); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 512; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[64], attr, VSI_NN_TYPE_INT64, 481480, 4096); + + /* @onnx//Concat_348_6:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 38; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[65], attr, VSI_NN_TYPE_INT16, 4931600, 38912); + + /* @onnx//Concat_348_6:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 38; + attr.dim_num = 1; + attr.dtype.fl = 30; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[66], attr, VSI_NN_TYPE_INT64, 4931296, 304); + + /* @onnx//Concat_345_7:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 512; + attr.size[3] = 19; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[67], attr, VSI_NN_TYPE_INT16, 4911840, 19456); + + /* @onnx//Concat_345_7:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 19; + attr.dim_num = 1; + attr.dtype.fl = 30; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[68], attr, VSI_NN_TYPE_INT64, 4911688, 152); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 185; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[69], attr, VSI_NN_TYPE_INT16, 5039120, 47360); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[70], attr, VSI_NN_TYPE_INT64, 5038096, 1024); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[71], attr, VSI_NN_TYPE_INT16, 5087504, 294912); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[72], attr, VSI_NN_TYPE_INT64, 5086480, 1024); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[73], attr, VSI_NN_TYPE_INT16, 5383440, 819200); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[74], attr, VSI_NN_TYPE_INT64, 5382416, 1024); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[75], attr, VSI_NN_TYPE_INT16, 6203664, 32768); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[76], attr, VSI_NN_TYPE_INT64, 6202640, 1024); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[77], attr, VSI_NN_TYPE_INT16, 6237456, 294912); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[78], attr, VSI_NN_TYPE_INT64, 6236432, 1024); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[79], attr, VSI_NN_TYPE_INT16, 6533392, 819200); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[80], attr, VSI_NN_TYPE_INT64, 6532368, 1024); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[81], attr, VSI_NN_TYPE_INT16, 7353616, 32768); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[82], attr, VSI_NN_TYPE_INT64, 7352592, 1024); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[83], attr, VSI_NN_TYPE_INT16, 7387408, 294912); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[84], attr, VSI_NN_TYPE_INT64, 7386384, 1024); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[85], attr, VSI_NN_TYPE_INT16, 7683344, 819200); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[86], attr, VSI_NN_TYPE_INT64, 7682320, 1024); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[87], attr, VSI_NN_TYPE_INT16, 8503568, 32768); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[88], attr, VSI_NN_TYPE_INT64, 8502544, 1024); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[89], attr, VSI_NN_TYPE_INT16, 8537360, 294912); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[90], attr, VSI_NN_TYPE_INT64, 8536336, 1024); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[91], attr, VSI_NN_TYPE_INT16, 8833296, 819200); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[92], attr, VSI_NN_TYPE_INT64, 8832272, 1024); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[93], attr, VSI_NN_TYPE_INT16, 9653520, 32768); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[94], attr, VSI_NN_TYPE_INT64, 9652496, 1024); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 3; + attr.size[1] = 3; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[95], attr, VSI_NN_TYPE_INT16, 9687312, 294912); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[96], attr, VSI_NN_TYPE_INT64, 9686288, 1024); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 5; + attr.size[1] = 5; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[97], attr, VSI_NN_TYPE_INT16, 9983248, 819200); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 27; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[98], attr, VSI_NN_TYPE_INT64, 9982224, 1024); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[99], attr, VSI_NN_TYPE_INT16, 5005328, 32768); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 27; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[100], attr, VSI_NN_TYPE_INT64, 5004304, 1024); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 128; + attr.dim_num = 4; + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[101], attr, VSI_NN_TYPE_INT16, 4971536, 32768); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 128; + attr.dim_num = 1; + attr.dtype.fl = 28; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[102], attr, VSI_NN_TYPE_INT64, 4970512, 1024); + + /* @400_4:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 38; + attr.dim_num = 4; + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[103], attr, VSI_NN_TYPE_INT16, 5320, 9728); + + /* @400_4:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 38; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[104], attr, VSI_NN_TYPE_INT64, 5016, 304); + + /* @397_5:weight */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 1; + attr.size[1] = 1; + attr.size[2] = 128; + attr.size[3] = 19; + attr.dim_num = 4; + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[105], attr, VSI_NN_TYPE_INT16, 152, 4864); + + /* @397_5:bias */ + memset( &attr, 0, sizeof( attr ) ); + attr.size[0] = 19; + attr.dim_num = 1; + attr.dtype.fl = 29; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_CONST_TENSOR(const_tensor[106], attr, VSI_NN_TYPE_INT64, 0, 152); + + + + /* @model/model.0/model.0.0/Conv_output_0_120:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[0]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.0/model.0.2/Relu_output_0_119:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[1]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.1/model.1.0/Conv_output_0_118:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 9; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[2]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.1/model.1.2/Relu_output_0_117:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 9; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[3]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.1/model.1.3/Conv_output_0_116:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[4]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.1/model.1.5/Relu_output_0_115:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[5]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.2/model.2.0/Conv_output_0_114:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[6]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.2/model.2.2/Relu_output_0_113:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[7]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.2/model.2.3/Conv_output_0_112:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[8]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.2/model.2.5/Relu_output_0_111:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[9]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.3/model.3.0/Conv_output_0_110:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[10]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.3/model.3.2/Relu_output_0_109:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[11]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.3/model.3.3/Conv_output_0_108:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[12]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.3/model.3.5/Relu_output_0_107:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[13]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.4/model.4.0/Conv_output_0_106:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 10; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[14]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.4/model.4.2/Relu_output_0_105:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 10; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[15]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.4/model.4.3/Conv_output_0_104:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[16]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.4/model.4.5/Relu_output_0_103:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[17]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.5/model.5.0/Conv_output_0_102:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[18]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.5/model.5.2/Relu_output_0_101:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[19]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.5/model.5.3/Conv_output_0_100:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[20]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.5/model.5.5/Relu_output_0_99:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[21]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.6/model.6.0/Conv_output_0_98:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 10; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[22]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.6/model.6.2/Relu_output_0_97:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 10; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[23]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.6/model.6.3/Conv_output_0_96:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[24]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.6/model.6.5/Relu_output_0_95:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[25]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.7/model.7.0/Conv_output_0_94:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[26]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.7/model.7.2/Relu_output_0_93:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[27]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.7/model.7.3/Conv_output_0_92:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[28]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.7/model.7.5/Relu_output_0_91:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[29]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.8/model.8.0/Conv_output_0_90:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 11; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[30]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.8/model.8.2/Relu_output_0_89:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 11; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[31]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.8/model.8.3/Conv_output_0_88:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[32]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.8/model.8.5/Relu_output_0_87:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[33]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.9/model.9.0/Conv_output_0_86:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[34]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.9/model.9.2/Relu_output_0_85:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[35]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.9/model.9.3/Conv_output_0_84:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[36]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.9/model.9.5/Relu_output_0_83:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[37]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.10/model.10.0/Conv_output_0_82:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[38]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.10/model.10.2/Relu_output_0_81:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[39]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.10/model.10.3/Conv_output_0_75:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[40]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.10/model.10.5/Relu_output_0_71:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[41]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.11/model.11.0/Conv_output_0_67:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 11; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[42]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.11/model.11.2/Relu_output_0_62:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 11; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[43]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.11/model.11.3/Conv_output_0_58:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[44]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @model/model.11/model.11.5/Relu_output_0_54:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[45]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/align/align.0/Conv_output_0_49:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[46]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/align/align.1/Relu_output_0_45:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[47]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 19; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[48]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 19; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[49]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[50]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[51]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[52]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[53]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[54]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[55]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[56]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[57]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 300; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[58]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_NONE; + NEW_VIRTUAL_TENSOR(node[59]->output.tensors[0], attr, VSI_NN_TYPE_FLOAT16); + + /* @cpm/Add_output_0_42:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[60]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/conv/conv.0/Conv_output_0_38:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[61]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @cpm/conv/conv.1/Relu_output_0_35:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[62]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[63]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[64]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[65]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[66]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[67]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[68]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[69]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[70]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 16; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[71]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[72]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @Concat_output_0_73:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[75]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[76]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 15; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[77]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[78]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[79]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[80]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[81]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.0/Add_output_0_60:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[82]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[83]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[84]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[85]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[86]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[87]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[88]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.1/Add_output_0_47:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[89]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[90]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[91]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[92]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[93]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[94]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[95]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.2/Add_output_0_36:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[96]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[97]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[98]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[99]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[100]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[101]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[102]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.3/Add_output_0_26:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[103]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[104]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[105]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[106]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[107]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[108]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[109]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/trunk/trunk.4/Add_output_0_16:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 12; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[110]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[111]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[112]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 14; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[113]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + /* @refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9:out0 */ + memset( &attr, 0, sizeof( attr ) ); + attr.dtype.fl = 13; + attr.dtype.qnt_type = VSI_NN_QNT_TYPE_DFP; + NEW_VIRTUAL_TENSOR(node[114]->output.tensors[0], attr, VSI_NN_TYPE_INT16); + + + +/*----------------------------------------- + Connection initialize + -----------------------------------------*/ + node[0]->input.tensors[0] = norm_tensor[4]; + node[73]->output.tensors[0] = norm_tensor[1]; + node[74]->output.tensors[0] = norm_tensor[0]; + node[115]->output.tensors[0] = norm_tensor[3]; + node[116]->output.tensors[0] = norm_tensor[2]; + + /* model/model.0/model.0.0/Conv_output_0_120 */ + node[0]->input.tensors[1] = const_tensor[0]; /* data_weight */ + node[0]->input.tensors[2] = const_tensor[1]; /* data_bias */ + + /* model/model.0/model.0.2/Relu_output_0_119 */ + node[1]->input.tensors[0] = node[0]->output.tensors[0]; + + /* model/model.1/model.1.0/Conv_output_0_118 */ + node[2]->input.tensors[0] = node[1]->output.tensors[0]; + node[2]->input.tensors[1] = const_tensor[2]; /* data_weight */ + node[2]->input.tensors[2] = const_tensor[3]; /* data_bias */ + + /* model/model.1/model.1.2/Relu_output_0_117 */ + node[3]->input.tensors[0] = node[2]->output.tensors[0]; + + /* model/model.1/model.1.3/Conv_output_0_116 */ + node[4]->input.tensors[0] = node[3]->output.tensors[0]; + node[4]->input.tensors[1] = const_tensor[4]; /* data_weight */ + node[4]->input.tensors[2] = const_tensor[5]; /* data_bias */ + + /* model/model.1/model.1.5/Relu_output_0_115 */ + node[5]->input.tensors[0] = node[4]->output.tensors[0]; + + /* model/model.2/model.2.0/Conv_output_0_114 */ + node[6]->input.tensors[0] = node[5]->output.tensors[0]; + node[6]->input.tensors[1] = const_tensor[6]; /* data_weight */ + node[6]->input.tensors[2] = const_tensor[7]; /* data_bias */ + + /* model/model.2/model.2.2/Relu_output_0_113 */ + node[7]->input.tensors[0] = node[6]->output.tensors[0]; + + /* model/model.2/model.2.3/Conv_output_0_112 */ + node[8]->input.tensors[0] = node[7]->output.tensors[0]; + node[8]->input.tensors[1] = const_tensor[8]; /* data_weight */ + node[8]->input.tensors[2] = const_tensor[9]; /* data_bias */ + + /* model/model.2/model.2.5/Relu_output_0_111 */ + node[9]->input.tensors[0] = node[8]->output.tensors[0]; + + /* model/model.3/model.3.0/Conv_output_0_110 */ + node[10]->input.tensors[0] = node[9]->output.tensors[0]; + node[10]->input.tensors[1] = const_tensor[10]; /* data_weight */ + node[10]->input.tensors[2] = const_tensor[11]; /* data_bias */ + + /* model/model.3/model.3.2/Relu_output_0_109 */ + node[11]->input.tensors[0] = node[10]->output.tensors[0]; + + /* model/model.3/model.3.3/Conv_output_0_108 */ + node[12]->input.tensors[0] = node[11]->output.tensors[0]; + node[12]->input.tensors[1] = const_tensor[12]; /* data_weight */ + node[12]->input.tensors[2] = const_tensor[13]; /* data_bias */ + + /* model/model.3/model.3.5/Relu_output_0_107 */ + node[13]->input.tensors[0] = node[12]->output.tensors[0]; + + /* model/model.4/model.4.0/Conv_output_0_106 */ + node[14]->input.tensors[0] = node[13]->output.tensors[0]; + node[14]->input.tensors[1] = const_tensor[14]; /* data_weight */ + node[14]->input.tensors[2] = const_tensor[15]; /* data_bias */ + + /* model/model.4/model.4.2/Relu_output_0_105 */ + node[15]->input.tensors[0] = node[14]->output.tensors[0]; + + /* model/model.4/model.4.3/Conv_output_0_104 */ + node[16]->input.tensors[0] = node[15]->output.tensors[0]; + node[16]->input.tensors[1] = const_tensor[16]; /* data_weight */ + node[16]->input.tensors[2] = const_tensor[17]; /* data_bias */ + + /* model/model.4/model.4.5/Relu_output_0_103 */ + node[17]->input.tensors[0] = node[16]->output.tensors[0]; + + /* model/model.5/model.5.0/Conv_output_0_102 */ + node[18]->input.tensors[0] = node[17]->output.tensors[0]; + node[18]->input.tensors[1] = const_tensor[18]; /* data_weight */ + node[18]->input.tensors[2] = const_tensor[19]; /* data_bias */ + + /* model/model.5/model.5.2/Relu_output_0_101 */ + node[19]->input.tensors[0] = node[18]->output.tensors[0]; + + /* model/model.5/model.5.3/Conv_output_0_100 */ + node[20]->input.tensors[0] = node[19]->output.tensors[0]; + node[20]->input.tensors[1] = const_tensor[20]; /* data_weight */ + node[20]->input.tensors[2] = const_tensor[21]; /* data_bias */ + + /* model/model.5/model.5.5/Relu_output_0_99 */ + node[21]->input.tensors[0] = node[20]->output.tensors[0]; + + /* model/model.6/model.6.0/Conv_output_0_98 */ + node[22]->input.tensors[0] = node[21]->output.tensors[0]; + node[22]->input.tensors[1] = const_tensor[22]; /* data_weight */ + node[22]->input.tensors[2] = const_tensor[23]; /* data_bias */ + + /* model/model.6/model.6.2/Relu_output_0_97 */ + node[23]->input.tensors[0] = node[22]->output.tensors[0]; + + /* model/model.6/model.6.3/Conv_output_0_96 */ + node[24]->input.tensors[0] = node[23]->output.tensors[0]; + node[24]->input.tensors[1] = const_tensor[24]; /* data_weight */ + node[24]->input.tensors[2] = const_tensor[25]; /* data_bias */ + + /* model/model.6/model.6.5/Relu_output_0_95 */ + node[25]->input.tensors[0] = node[24]->output.tensors[0]; + + /* model/model.7/model.7.0/Conv_output_0_94 */ + node[26]->input.tensors[0] = node[25]->output.tensors[0]; + node[26]->input.tensors[1] = const_tensor[26]; /* data_weight */ + node[26]->input.tensors[2] = const_tensor[27]; /* data_bias */ + + /* model/model.7/model.7.2/Relu_output_0_93 */ + node[27]->input.tensors[0] = node[26]->output.tensors[0]; + + /* model/model.7/model.7.3/Conv_output_0_92 */ + node[28]->input.tensors[0] = node[27]->output.tensors[0]; + node[28]->input.tensors[1] = const_tensor[28]; /* data_weight */ + node[28]->input.tensors[2] = const_tensor[29]; /* data_bias */ + + /* model/model.7/model.7.5/Relu_output_0_91 */ + node[29]->input.tensors[0] = node[28]->output.tensors[0]; + + /* model/model.8/model.8.0/Conv_output_0_90 */ + node[30]->input.tensors[0] = node[29]->output.tensors[0]; + node[30]->input.tensors[1] = const_tensor[30]; /* data_weight */ + node[30]->input.tensors[2] = const_tensor[31]; /* data_bias */ + + /* model/model.8/model.8.2/Relu_output_0_89 */ + node[31]->input.tensors[0] = node[30]->output.tensors[0]; + + /* model/model.8/model.8.3/Conv_output_0_88 */ + node[32]->input.tensors[0] = node[31]->output.tensors[0]; + node[32]->input.tensors[1] = const_tensor[32]; /* data_weight */ + node[32]->input.tensors[2] = const_tensor[33]; /* data_bias */ + + /* model/model.8/model.8.5/Relu_output_0_87 */ + node[33]->input.tensors[0] = node[32]->output.tensors[0]; + + /* model/model.9/model.9.0/Conv_output_0_86 */ + node[34]->input.tensors[0] = node[33]->output.tensors[0]; + node[34]->input.tensors[1] = const_tensor[34]; /* data_weight */ + node[34]->input.tensors[2] = const_tensor[35]; /* data_bias */ + + /* model/model.9/model.9.2/Relu_output_0_85 */ + node[35]->input.tensors[0] = node[34]->output.tensors[0]; + + /* model/model.9/model.9.3/Conv_output_0_84 */ + node[36]->input.tensors[0] = node[35]->output.tensors[0]; + node[36]->input.tensors[1] = const_tensor[36]; /* data_weight */ + node[36]->input.tensors[2] = const_tensor[37]; /* data_bias */ + + /* model/model.9/model.9.5/Relu_output_0_83 */ + node[37]->input.tensors[0] = node[36]->output.tensors[0]; + + /* model/model.10/model.10.0/Conv_output_0_82 */ + node[38]->input.tensors[0] = node[37]->output.tensors[0]; + node[38]->input.tensors[1] = const_tensor[38]; /* data_weight */ + node[38]->input.tensors[2] = const_tensor[39]; /* data_bias */ + + /* model/model.10/model.10.2/Relu_output_0_81 */ + node[39]->input.tensors[0] = node[38]->output.tensors[0]; + + /* model/model.10/model.10.3/Conv_output_0_75 */ + node[40]->input.tensors[0] = node[39]->output.tensors[0]; + node[40]->input.tensors[1] = const_tensor[40]; /* data_weight */ + node[40]->input.tensors[2] = const_tensor[41]; /* data_bias */ + + /* model/model.10/model.10.5/Relu_output_0_71 */ + node[41]->input.tensors[0] = node[40]->output.tensors[0]; + + /* model/model.11/model.11.0/Conv_output_0_67 */ + node[42]->input.tensors[0] = node[41]->output.tensors[0]; + node[42]->input.tensors[1] = const_tensor[42]; /* data_weight */ + node[42]->input.tensors[2] = const_tensor[43]; /* data_bias */ + + /* model/model.11/model.11.2/Relu_output_0_62 */ + node[43]->input.tensors[0] = node[42]->output.tensors[0]; + + /* model/model.11/model.11.3/Conv_output_0_58 */ + node[44]->input.tensors[0] = node[43]->output.tensors[0]; + node[44]->input.tensors[1] = const_tensor[44]; /* data_weight */ + node[44]->input.tensors[2] = const_tensor[45]; /* data_bias */ + + /* model/model.11/model.11.5/Relu_output_0_54 */ + node[45]->input.tensors[0] = node[44]->output.tensors[0]; + + /* cpm/align/align.0/Conv_output_0_49 */ + node[46]->input.tensors[0] = node[45]->output.tensors[0]; + node[46]->input.tensors[1] = const_tensor[46]; /* data_weight */ + node[46]->input.tensors[2] = const_tensor[47]; /* data_bias */ + + /* cpm/align/align.1/Relu_output_0_45 */ + node[47]->input.tensors[0] = node[46]->output.tensors[0]; + + /* cpm/trunk/trunk.0/trunk.0.0/Conv_output_0_80 */ + node[48]->input.tensors[0] = node[47]->output.tensors[0]; + node[48]->input.tensors[1] = const_tensor[48]; /* data_weight */ + node[48]->input.tensors[2] = const_tensor[49]; /* data_bias */ + + /* cpm/trunk/trunk.0/trunk.0.1/Elu_output_0_79 */ + node[49]->input.tensors[0] = node[48]->output.tensors[0]; + + /* cpm/trunk/trunk.0/trunk.0.2/Conv_output_0_78 */ + node[50]->input.tensors[0] = node[49]->output.tensors[0]; + node[50]->input.tensors[1] = const_tensor[50]; /* data_weight */ + node[50]->input.tensors[2] = const_tensor[49]; /* data_bias */ + + /* cpm/trunk/trunk.0/trunk.0.3/Elu_output_0_77 */ + node[51]->input.tensors[0] = node[50]->output.tensors[0]; + + /* cpm/trunk/trunk.1/trunk.1.0/Conv_output_0_76 */ + node[52]->input.tensors[0] = node[51]->output.tensors[0]; + node[52]->input.tensors[1] = const_tensor[51]; /* data_weight */ + node[52]->input.tensors[2] = const_tensor[49]; /* data_bias */ + + /* cpm/trunk/trunk.1/trunk.1.1/Elu_output_0_72 */ + node[53]->input.tensors[0] = node[52]->output.tensors[0]; + + /* cpm/trunk/trunk.1/trunk.1.2/Conv_output_0_68 */ + node[54]->input.tensors[0] = node[53]->output.tensors[0]; + node[54]->input.tensors[1] = const_tensor[52]; /* data_weight */ + node[54]->input.tensors[2] = const_tensor[49]; /* data_bias */ + + /* cpm/trunk/trunk.1/trunk.1.3/Elu_output_0_63 */ + node[55]->input.tensors[0] = node[54]->output.tensors[0]; + + /* cpm/trunk/trunk.2/trunk.2.0/Conv_output_0_59 */ + node[56]->input.tensors[0] = node[55]->output.tensors[0]; + node[56]->input.tensors[1] = const_tensor[51]; /* data_weight */ + node[56]->input.tensors[2] = const_tensor[49]; /* data_bias */ + + /* cpm/trunk/trunk.2/trunk.2.1/Elu_output_0_55 */ + node[57]->input.tensors[0] = node[56]->output.tensors[0]; + + /* cpm/trunk/trunk.2/trunk.2.2/Conv_output_0_50 */ + node[58]->input.tensors[0] = node[57]->output.tensors[0]; + node[58]->input.tensors[1] = const_tensor[52]; /* data_weight */ + node[58]->input.tensors[2] = const_tensor[49]; /* data_bias */ + + /* cpm/trunk/trunk.2/trunk.2.3/Elu_output_0_46 */ + node[59]->input.tensors[0] = node[58]->output.tensors[0]; + + /* cpm/Add_output_0_42 */ + node[60]->input.tensors[0] = node[47]->output.tensors[0]; + node[60]->input.tensors[1] = node[59]->output.tensors[0]; + + /* cpm/conv/conv.0/Conv_output_0_38 */ + node[61]->input.tensors[0] = node[60]->output.tensors[0]; + node[61]->input.tensors[1] = const_tensor[53]; /* data_weight */ + node[61]->input.tensors[2] = const_tensor[54]; /* data_bias */ + + /* cpm/conv/conv.1/Relu_output_0_35 */ + node[62]->input.tensors[0] = node[61]->output.tensors[0]; + + /* initial_stage/trunk/trunk.0/trunk.0.0/Conv_output_0_32 */ + node[63]->input.tensors[0] = node[62]->output.tensors[0]; + node[63]->input.tensors[1] = const_tensor[55]; /* data_weight */ + node[63]->input.tensors[2] = const_tensor[56]; /* data_bias */ + + /* initial_stage/trunk/trunk.0/trunk.0.1/Relu_output_0_28 */ + node[64]->input.tensors[0] = node[63]->output.tensors[0]; + + /* initial_stage/trunk/trunk.1/trunk.1.0/Conv_output_0_25 */ + node[65]->input.tensors[0] = node[64]->output.tensors[0]; + node[65]->input.tensors[1] = const_tensor[57]; /* data_weight */ + node[65]->input.tensors[2] = const_tensor[58]; /* data_bias */ + + /* initial_stage/trunk/trunk.1/trunk.1.1/Relu_output_0_22 */ + node[66]->input.tensors[0] = node[65]->output.tensors[0]; + + /* initial_stage/trunk/trunk.2/trunk.2.0/Conv_output_0_19 */ + node[67]->input.tensors[0] = node[66]->output.tensors[0]; + node[67]->input.tensors[1] = const_tensor[59]; /* data_weight */ + node[67]->input.tensors[2] = const_tensor[60]; /* data_bias */ + + /* initial_stage/trunk/trunk.2/trunk.2.1/Relu_output_0_18 */ + node[68]->input.tensors[0] = node[67]->output.tensors[0]; + + /* initial_stage/pafs/pafs.0/pafs.0.0/Conv_output_0_14 */ + node[69]->input.tensors[0] = node[68]->output.tensors[0]; + node[69]->input.tensors[1] = const_tensor[61]; /* data_weight */ + node[69]->input.tensors[2] = const_tensor[62]; /* data_bias */ + + /* initial_stage/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_15 */ + node[70]->input.tensors[0] = node[68]->output.tensors[0]; + node[70]->input.tensors[1] = const_tensor[63]; /* data_weight */ + node[70]->input.tensors[2] = const_tensor[64]; /* data_bias */ + + /* initial_stage/pafs/pafs.0/pafs.0.1/Relu_output_0_10 */ + node[71]->input.tensors[0] = node[69]->output.tensors[0]; + + /* initial_stage/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_11 */ + node[72]->input.tensors[0] = node[70]->output.tensors[0]; + + /* onnx//Concat_348_6 */ + node[73]->input.tensors[0] = node[71]->output.tensors[0]; + node[73]->input.tensors[1] = const_tensor[65]; /* data_weight */ + node[73]->input.tensors[2] = const_tensor[66]; /* data_bias */ + + /* onnx//Concat_345_7 */ + node[74]->input.tensors[0] = node[72]->output.tensors[0]; + node[74]->input.tensors[1] = const_tensor[67]; /* data_weight */ + node[74]->input.tensors[2] = const_tensor[68]; /* data_bias */ + + /* Concat_output_0_73 */ + node[75]->input.tensors[0] = node[62]->output.tensors[0]; + node[75]->input.tensors[1] = node[74]->output.tensors[0]; + node[75]->input.tensors[2] = node[73]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/initial/initial.0/Conv_output_0_69 */ + node[76]->input.tensors[0] = node[75]->output.tensors[0]; + node[76]->input.tensors[1] = const_tensor[69]; /* data_weight */ + node[76]->input.tensors[2] = const_tensor[70]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.0/initial/initial.1/Relu_output_0_64 */ + node[77]->input.tensors[0] = node[76]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.0/Conv_output_0_74 */ + node[78]->input.tensors[0] = node[77]->output.tensors[0]; + node[78]->input.tensors[1] = const_tensor[71]; /* data_weight */ + node[78]->input.tensors[2] = const_tensor[72]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.0/trunk.0.2/Relu_output_0_70 */ + node[79]->input.tensors[0] = node[78]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.0/Conv_output_0_66 */ + node[80]->input.tensors[0] = node[79]->output.tensors[0]; + node[80]->input.tensors[1] = const_tensor[73]; /* data_weight */ + node[80]->input.tensors[2] = const_tensor[74]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.0/trunk/trunk.1/trunk.1.2/Relu_output_0_65 */ + node[81]->input.tensors[0] = node[80]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.0/Add_output_0_60 */ + node[82]->input.tensors[0] = node[77]->output.tensors[0]; + node[82]->input.tensors[1] = node[81]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/initial/initial.0/Conv_output_0_56 */ + node[83]->input.tensors[0] = node[82]->output.tensors[0]; + node[83]->input.tensors[1] = const_tensor[75]; /* data_weight */ + node[83]->input.tensors[2] = const_tensor[76]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.1/initial/initial.1/Relu_output_0_51 */ + node[84]->input.tensors[0] = node[83]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.0/Conv_output_0_61 */ + node[85]->input.tensors[0] = node[84]->output.tensors[0]; + node[85]->input.tensors[1] = const_tensor[77]; /* data_weight */ + node[85]->input.tensors[2] = const_tensor[78]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.0/trunk.0.2/Relu_output_0_57 */ + node[86]->input.tensors[0] = node[85]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.0/Conv_output_0_53 */ + node[87]->input.tensors[0] = node[86]->output.tensors[0]; + node[87]->input.tensors[1] = const_tensor[79]; /* data_weight */ + node[87]->input.tensors[2] = const_tensor[80]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.1/trunk/trunk.1/trunk.1.2/Relu_output_0_52 */ + node[88]->input.tensors[0] = node[87]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.1/Add_output_0_47 */ + node[89]->input.tensors[0] = node[84]->output.tensors[0]; + node[89]->input.tensors[1] = node[88]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/initial/initial.0/Conv_output_0_43 */ + node[90]->input.tensors[0] = node[89]->output.tensors[0]; + node[90]->input.tensors[1] = const_tensor[81]; /* data_weight */ + node[90]->input.tensors[2] = const_tensor[82]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.2/initial/initial.1/Relu_output_0_39 */ + node[91]->input.tensors[0] = node[90]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.0/Conv_output_0_48 */ + node[92]->input.tensors[0] = node[91]->output.tensors[0]; + node[92]->input.tensors[1] = const_tensor[83]; /* data_weight */ + node[92]->input.tensors[2] = const_tensor[84]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.0/trunk.0.2/Relu_output_0_44 */ + node[93]->input.tensors[0] = node[92]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.0/Conv_output_0_41 */ + node[94]->input.tensors[0] = node[93]->output.tensors[0]; + node[94]->input.tensors[1] = const_tensor[85]; /* data_weight */ + node[94]->input.tensors[2] = const_tensor[86]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.2/trunk/trunk.1/trunk.1.2/Relu_output_0_40 */ + node[95]->input.tensors[0] = node[94]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.2/Add_output_0_36 */ + node[96]->input.tensors[0] = node[91]->output.tensors[0]; + node[96]->input.tensors[1] = node[95]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/initial/initial.0/Conv_output_0_33 */ + node[97]->input.tensors[0] = node[96]->output.tensors[0]; + node[97]->input.tensors[1] = const_tensor[87]; /* data_weight */ + node[97]->input.tensors[2] = const_tensor[88]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.3/initial/initial.1/Relu_output_0_29 */ + node[98]->input.tensors[0] = node[97]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.0/Conv_output_0_37 */ + node[99]->input.tensors[0] = node[98]->output.tensors[0]; + node[99]->input.tensors[1] = const_tensor[89]; /* data_weight */ + node[99]->input.tensors[2] = const_tensor[90]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.0/trunk.0.2/Relu_output_0_34 */ + node[100]->input.tensors[0] = node[99]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.0/Conv_output_0_31 */ + node[101]->input.tensors[0] = node[100]->output.tensors[0]; + node[101]->input.tensors[1] = const_tensor[91]; /* data_weight */ + node[101]->input.tensors[2] = const_tensor[92]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.3/trunk/trunk.1/trunk.1.2/Relu_output_0_30 */ + node[102]->input.tensors[0] = node[101]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.3/Add_output_0_26 */ + node[103]->input.tensors[0] = node[98]->output.tensors[0]; + node[103]->input.tensors[1] = node[102]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/initial/initial.0/Conv_output_0_23 */ + node[104]->input.tensors[0] = node[103]->output.tensors[0]; + node[104]->input.tensors[1] = const_tensor[93]; /* data_weight */ + node[104]->input.tensors[2] = const_tensor[94]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.4/initial/initial.1/Relu_output_0_20 */ + node[105]->input.tensors[0] = node[104]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.0/Conv_output_0_27 */ + node[106]->input.tensors[0] = node[105]->output.tensors[0]; + node[106]->input.tensors[1] = const_tensor[95]; /* data_weight */ + node[106]->input.tensors[2] = const_tensor[96]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.0/trunk.0.2/Relu_output_0_24 */ + node[107]->input.tensors[0] = node[106]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.0/Conv_output_0_21 */ + node[108]->input.tensors[0] = node[107]->output.tensors[0]; + node[108]->input.tensors[1] = const_tensor[97]; /* data_weight */ + node[108]->input.tensors[2] = const_tensor[98]; /* data_bias */ + + /* refinement_stages.0/trunk/trunk.4/trunk/trunk.1/trunk.1.2/Relu_output_0_17 */ + node[109]->input.tensors[0] = node[108]->output.tensors[0]; + + /* refinement_stages.0/trunk/trunk.4/Add_output_0_16 */ + node[110]->input.tensors[0] = node[105]->output.tensors[0]; + node[110]->input.tensors[1] = node[109]->output.tensors[0]; + + /* refinement_stages.0/pafs/pafs.0/pafs.0.0/Conv_output_0_12 */ + node[111]->input.tensors[0] = node[110]->output.tensors[0]; + node[111]->input.tensors[1] = const_tensor[99]; /* data_weight */ + node[111]->input.tensors[2] = const_tensor[100]; /* data_bias */ + + /* refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.0/Conv_output_0_13 */ + node[112]->input.tensors[0] = node[110]->output.tensors[0]; + node[112]->input.tensors[1] = const_tensor[101]; /* data_weight */ + node[112]->input.tensors[2] = const_tensor[102]; /* data_bias */ + + /* refinement_stages.0/pafs/pafs.0/pafs.0.1/Relu_output_0_8 */ + node[113]->input.tensors[0] = node[111]->output.tensors[0]; + + /* refinement_stages.0/heatmaps/heatmaps.0/heatmaps.0.1/Relu_output_0_9 */ + node[114]->input.tensors[0] = node[112]->output.tensors[0]; + + /* 400_4 */ + node[115]->input.tensors[0] = node[113]->output.tensors[0]; + node[115]->input.tensors[1] = const_tensor[103]; /* data_weight */ + node[115]->input.tensors[2] = const_tensor[104]; /* data_bias */ + + /* 397_5 */ + node[116]->input.tensors[0] = node[114]->output.tensors[0]; + node[116]->input.tensors[1] = const_tensor[105]; /* data_weight */ + node[116]->input.tensors[2] = const_tensor[106]; /* data_bias */ + + + } + else + { + node[0]->output.tensors[0] = norm_tensor[0]; + node[0]->output.tensors[1] = norm_tensor[1]; + node[0]->output.tensors[2] = norm_tensor[2]; + node[0]->output.tensors[3] = norm_tensor[3]; + node[0]->input.tensors[0] = norm_tensor[4]; + + } + graph->output.tensors[0] = norm_tensor[0]; + graph->output.tensors[1] = norm_tensor[1]; + graph->output.tensors[2] = norm_tensor[2]; + graph->output.tensors[3] = norm_tensor[3]; + graph->input.tensors[0] = norm_tensor[4]; + + + if( enable_pre_post_process ) + { + sort = TRUE; + if( pre_process_map_count > 0 ) + { + for( i = 0; i < pre_process_map_count; i++ ) + { + status = vsi_nn_AddGraphPreProcess(graph, pre_process_map[i].graph_input_idx, + pre_process_map[i].preprocesses, + pre_process_map[i].preprocess_count); + TEST_CHECK_STATUS( status, error ); + } + } + + if( post_process_map_count > 0 ) + { + for( i = 0; i < post_process_map_count; i++ ) + { + status = vsi_nn_AddGraphPostProcess(graph, post_process_map[i].graph_output_idx, + post_process_map[i].postprocesses, + post_process_map[i].postprocess_count); + TEST_CHECK_STATUS( status, error ); + } + } + } + + status = vsi_nn_SetupGraph( graph, sort ); + TEST_CHECK_STATUS( status, error ); + vsi_nn_DumpGraphToJson( graph ); + + if( VSI_FAILURE == status ) + { + goto error; + } + + fclose( fp ); + + return graph; + +error: + if( NULL != fp ) + { + fclose( fp ); + } + + release_ctx = ( NULL == in_ctx ); + vsi_nn_DumpGraphToJson( graph ); + vnn_ReleaseDynamicFixedPoint16( graph, release_ctx ); + + return NULL; +} /* vsi_nn_CreateDynamicFixedPoint16() */ + +void vnn_ReleaseDynamicFixedPoint16 + ( + vsi_nn_graph_t * graph, + vsi_bool release_ctx + ) +{ + vsi_nn_context_t ctx; + if( NULL != graph ) + { + ctx = graph->ctx; + vsi_nn_ReleaseGraph( &graph ); + + /*----------------------------------------- + Unregister client ops + -----------------------------------------*/ + + + if( release_ctx ) + { + vsi_nn_ReleaseContext( &ctx ); + } + } +} /* vsi_nn_ReleaseDynamicFixedPoint16() */ + diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/vnn_dynamicfixedpoint16.h b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_dynamicfixedpoint16.h new file mode 100644 index 0000000..5ecec4a --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_dynamicfixedpoint16.h @@ -0,0 +1,39 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction network definition header file +****************************************************************************/ + +#ifndef _VNN_DYNAMICFIXEDPOINT16_H +#define _VNN_DYNAMICFIXEDPOINT16_H + +#include "vsi_nn_pub.h" + +#define VNN_APP_DEBUG (FALSE) +#define VNN_VERSION_MAJOR 1 +#define VNN_VERSION_MINOR 1 +#define VNN_VERSION_PATCH 53 +#define VNN_RUNTIME_VERSION \ + (VNN_VERSION_MAJOR * 10000 + VNN_VERSION_MINOR * 100 + VNN_VERSION_PATCH) + +_version_assert(VNN_RUNTIME_VERSION <= VSI_NN_VERSION, + CASE_VERSION_is_higher_than_OVXLIB_VERSION) + +void vnn_ReleaseDynamicFixedPoint16 + ( + vsi_nn_graph_t * graph, + vsi_bool release_ctx + ); + +vsi_nn_graph_t * vnn_CreateDynamicFixedPoint16 + ( + const char * data_file_name, + vsi_nn_context_t in_ctx, + const vsi_nn_preprocess_map_element_t * pre_process_map, + uint32_t pre_process_map_count, + const vsi_nn_postprocess_map_element_t * post_process_map, + uint32_t post_process_map_count + ); + +#endif diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/vnn_global.h b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_global.h new file mode 100644 index 0000000..e8b3704 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_global.h @@ -0,0 +1,38 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network global header file +****************************************************************************/ +#ifndef _VNN_GLOBAL_H_ +#define _VNN_GLOBAL_H_ + +typedef struct { + uint32_t graph_input_idx; + vsi_nn_preprocess_base_t *preprocesses; + uint32_t preprocess_count; +} vsi_nn_preprocess_map_element_t; + + +typedef struct { + uint32_t graph_output_idx; + vsi_nn_postprocess_base_t *postprocesses; + uint32_t postprocess_count; +} vsi_nn_postprocess_map_element_t; + +#ifndef VSI_SIZE_T +typedef uint32_t vsi_size_t; +typedef int32_t vsi_ssize_t; +#endif + +#ifdef _WIN32 +#define VSI_FSEEK _fseeki64 +#else +#define VSI_FSEEK fseek +#endif + +/* + * This file will be deprecated in the future + */ + +#endif \ No newline at end of file diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/vnn_post_process.c b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_post_process.c new file mode 100644 index 0000000..e75de56 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_post_process.c @@ -0,0 +1,215 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction post-process source file +****************************************************************************/ +/*------------------------------------------- + Includes +-------------------------------------------*/ +#include +#include +#include + +#include "vsi_nn_pub.h" + +#include "vnn_global.h" +#include "vnn_post_process.h" + +#define _BASETSD_H + +/*------------------------------------------- + Variable definitions +-------------------------------------------*/ +/*post process for lid: attach_onnx//Concat_345/out0_0*/ +int32_t perm_0[] = {0, 1, 2, 3}; +vsi_nn_postprocess_permute_t permute_for_norm_tensor_0 = {perm_0, 4}; +vsi_nn_postprocess_dtype_convert_t dtype_convert_for_norm_tensor_0 = {{VSI_NN_DIM_FMT_NCHW, VSI_NN_TYPE_FLOAT32, {VSI_NN_QNT_TYPE_NONE}}}; +vsi_nn_postprocess_base_t post_process_for_norm_tensor_0[] = + { + {VSI_NN_POSTPROCESS_PERMUTE, &permute_for_norm_tensor_0}, + {VSI_NN_POSTPROCESS_DTYPE_CONVERT, &dtype_convert_for_norm_tensor_0}, + }; +/*post process for lid: attach_onnx//Concat_348/out0_1*/ +int32_t perm_1[] = {0, 1, 2, 3}; +vsi_nn_postprocess_permute_t permute_for_norm_tensor_1 = {perm_1, 4}; +vsi_nn_postprocess_dtype_convert_t dtype_convert_for_norm_tensor_1 = {{VSI_NN_DIM_FMT_NCHW, VSI_NN_TYPE_FLOAT32, {VSI_NN_QNT_TYPE_NONE}}}; +vsi_nn_postprocess_base_t post_process_for_norm_tensor_1[] = + { + {VSI_NN_POSTPROCESS_PERMUTE, &permute_for_norm_tensor_1}, + {VSI_NN_POSTPROCESS_DTYPE_CONVERT, &dtype_convert_for_norm_tensor_1}, + }; +/*post process for lid: attach_397/out0_2*/ +int32_t perm_2[] = {0, 1, 2, 3}; +vsi_nn_postprocess_permute_t permute_for_norm_tensor_2 = {perm_2, 4}; +vsi_nn_postprocess_dtype_convert_t dtype_convert_for_norm_tensor_2 = {{VSI_NN_DIM_FMT_NCHW, VSI_NN_TYPE_FLOAT32, {VSI_NN_QNT_TYPE_NONE}}}; +vsi_nn_postprocess_base_t post_process_for_norm_tensor_2[] = + { + {VSI_NN_POSTPROCESS_PERMUTE, &permute_for_norm_tensor_2}, + {VSI_NN_POSTPROCESS_DTYPE_CONVERT, &dtype_convert_for_norm_tensor_2}, + }; +/*post process for lid: attach_400/out0_3*/ +int32_t perm_3[] = {0, 1, 2, 3}; +vsi_nn_postprocess_permute_t permute_for_norm_tensor_3 = {perm_3, 4}; +vsi_nn_postprocess_dtype_convert_t dtype_convert_for_norm_tensor_3 = {{VSI_NN_DIM_FMT_NCHW, VSI_NN_TYPE_FLOAT32, {VSI_NN_QNT_TYPE_NONE}}}; +vsi_nn_postprocess_base_t post_process_for_norm_tensor_3[] = + { + {VSI_NN_POSTPROCESS_PERMUTE, &permute_for_norm_tensor_3}, + {VSI_NN_POSTPROCESS_DTYPE_CONVERT, &dtype_convert_for_norm_tensor_3}, + }; + +/*{graph_output_idx, postprocess}*/ +const static vsi_nn_postprocess_map_element_t postprocess_map[] = +{ +{0, post_process_for_norm_tensor_0, sizeof(post_process_for_norm_tensor_0) / sizeof(vsi_nn_postprocess_base_t)}, +{1, post_process_for_norm_tensor_1, sizeof(post_process_for_norm_tensor_1) / sizeof(vsi_nn_postprocess_base_t)}, +{2, post_process_for_norm_tensor_2, sizeof(post_process_for_norm_tensor_2) / sizeof(vsi_nn_postprocess_base_t)}, +{3, post_process_for_norm_tensor_3, sizeof(post_process_for_norm_tensor_3) / sizeof(vsi_nn_postprocess_base_t)}, +}; + + +/*------------------------------------------- + Functions +-------------------------------------------*/ +static void save_output_data(vsi_nn_graph_t *graph) +{ + uint32_t i; +#define _DUMP_FILE_LENGTH 1028 +#define _DUMP_SHAPE_LENGTH 128 + char filename[_DUMP_FILE_LENGTH] = {0}, shape[_DUMP_SHAPE_LENGTH] = {0}; + vsi_nn_tensor_t *tensor; + + for(i = 0; i < graph->output.num; i++) + { + tensor = vsi_nn_GetTensor(graph, graph->output.tensors[i]); + vsi_nn_ShapeToString( tensor->attr.size, tensor->attr.dim_num, + shape, _DUMP_SHAPE_LENGTH, FALSE ); + snprintf(filename, _DUMP_FILE_LENGTH, "output%u_%s.dat", i, shape); + vsi_nn_SaveTensorToBinary(graph, tensor, filename); + + } +} + +static vsi_bool get_top + ( + float *pfProb, + float *pfMaxProb, + vsi_size_t *pMaxClass, + vsi_size_t outputCount, + vsi_size_t topNum + ) +{ + vsi_size_t i, j, k; + + #define MAX_TOP_NUM 20 + if (topNum > MAX_TOP_NUM) return FALSE; + + memset(pfMaxProb, 0xfe, sizeof(float) * topNum); + memset(pMaxClass, 0xff, sizeof(vsi_size_t) * topNum); + + for (j = 0; j < topNum; j++) + { + for (i=0; i *(pfMaxProb+j)) + { + *(pfMaxProb+j) = pfProb[i]; + *(pMaxClass+j) = i; + } + } + } + + return TRUE; +} + +static vsi_status show_top5 + ( + vsi_nn_graph_t *graph, + vsi_nn_tensor_t *tensor + ) +{ + vsi_status status = VSI_FAILURE; + vsi_size_t i,sz,stride; + float *buffer = NULL; + uint8_t *tensor_data = NULL; + vsi_size_t MaxClass[5]; + float fMaxProb[5]; + vsi_size_t topk = 5; + + sz = 1; + for(i = 0; i < tensor->attr.dim_num; i++) + { + sz *= tensor->attr.size[i]; + } + + if(topk > sz) + topk = sz; + + stride = (vsi_size_t)vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + tensor_data = (uint8_t *)vsi_nn_ConvertTensorToData(graph, tensor); + buffer = (float *)malloc(sizeof(float) * sz); + + for(i = 0; i < sz; i++) + { + status = vsi_nn_DtypeToFloat32(&tensor_data[stride * i], &buffer[i], &tensor->attr.dtype); + } + + if (!get_top(buffer, fMaxProb, MaxClass, sz, topk)) + { + printf("Fail to show result.\n"); + goto final; + } + + printf(" --- Top%d ---\n", topk); + for(i = 0; i< topk; i++) + { + printf("%3d: %8.6f\n", MaxClass[i], fMaxProb[i]); + } + status = VSI_SUCCESS; + +final: + if(tensor_data)vsi_nn_Free(tensor_data); + if(buffer)free(buffer); + return status; +} + +vsi_status vnn_PostProcessDynamicFixedPoint16(vsi_nn_graph_t *graph) +{ + vsi_status status = VSI_FAILURE; + + /* Show the top5 result */ + status = show_top5(graph, vsi_nn_GetTensor(graph, graph->output.tensors[0])); + TEST_CHECK_STATUS(status, final); + + /* Save all output tensor data to txt file */ + save_output_data(graph); + +final: + return VSI_SUCCESS; +} + +const vsi_nn_postprocess_map_element_t * vnn_GetPostProcessMap() +{ + return postprocess_map; +} + +uint32_t vnn_GetPostProcessMapCount() +{ + if (postprocess_map == NULL) + return 0; + else + return sizeof(postprocess_map) / sizeof(vsi_nn_postprocess_map_element_t); +} diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/vnn_post_process.h b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_post_process.h new file mode 100644 index 0000000..1aa1229 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_post_process.h @@ -0,0 +1,16 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction post-process header file +****************************************************************************/ +#ifndef _VNN_POST_PROCESS_H_ +#define _VNN_POST_PROCESS_H_ + +vsi_status vnn_PostProcessDynamicFixedPoint16(vsi_nn_graph_t *graph); + +const vsi_nn_postprocess_map_element_t * vnn_GetPostProcessMap(); + +uint32_t vnn_GetPostProcessMapCount(); + +#endif diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/vnn_pre_process.c b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_pre_process.c new file mode 100644 index 0000000..89fab19 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_pre_process.c @@ -0,0 +1,938 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction pre-process source file +****************************************************************************/ +/*------------------------------------------- + Includes +-------------------------------------------*/ +#include +#include +#include + +#include "jpeglib.h" +#include "vsi_nn_pub.h" +#include "vnn_global.h" +#include "vnn_pre_process.h" + +#define _BASETSD_H + +/*------------------------------------------- + Variable definitions +-------------------------------------------*/ +/*pre process for lid: input.1_121*/ +vsi_nn_preprocess_source_layout_e source_layout_for_norm_tensor_4 = VSI_NN_SOURCE_LAYOUT_NCHW; +vsi_nn_preprocess_source_format_e source_format_for_norm_tensor_4 = VSI_NN_SOURCE_FORMAT_IMAGE_RGB; +vsi_nn_preprocess_image_size_t size_for_norm_tensor_4 = {424, 256, 3}; + +vsi_nn_preprocess_image_resize_t resize_for_norm_tensor_4 = {424, 256, 3}; +int8_t reverse_channel_for_norm_tensor_4 = 1; +float mean_and_scale_4[] = {128, 128, 128}; +vsi_nn_preprocess_mean_and_scale_t mean_and_scale_for_norm_tensor_4 = {mean_and_scale_4, 3, 0.00390625}; +int32_t perm_4[] = {0, 1, 2, 3}; +vsi_nn_preprocess_permute_t permute_for_norm_tensor_4 = {perm_4, 4}; +vsi_nn_preprocess_dtype_convert_t dtype_converter_for_norm_tensor4={.dtype.fmt=VSI_NN_DIM_FMT_NCHW, .dtype.vx_type=VSI_NN_TYPE_INT16, .dtype.qnt_type=VSI_NN_QNT_TYPE_DFP, .dtype.fl=16}; +vsi_nn_preprocess_base_t pre_process_for_norm_tensor_4[] = + { + {VSI_NN_PREPROCESS_SOURCE_LAYOUT, &source_layout_for_norm_tensor_4}, + {VSI_NN_PREPROCESS_SET_SOURCE_FORMAT, &source_format_for_norm_tensor_4}, + + {VSI_NN_PREPROCESS_IMAGE_SIZE, &size_for_norm_tensor_4}, + {VSI_NN_PREPROCESS_IMAGE_RESIZE_BILINEAR, &resize_for_norm_tensor_4}, + {VSI_NN_PREPROCESS_REVERSE_CHANNEL, &reverse_channel_for_norm_tensor_4}, + {VSI_NN_PREPROCESS_MEAN_AND_SCALE, &mean_and_scale_for_norm_tensor_4}, + {VSI_NN_PREPROCESS_PERMUTE, &permute_for_norm_tensor_4}, + {VSI_NN_PREPROCESS_DTYPE_CONVERT, &dtype_converter_for_norm_tensor4}, + }; + +/*{graph_input_idx, preprocess}*/ +const static vsi_nn_preprocess_map_element_t preprocess_map[] = +{ +{0, pre_process_for_norm_tensor_4, sizeof(pre_process_for_norm_tensor_4) / sizeof(vsi_nn_preprocess_base_t)}, +}; + +/*------------------------------------------- + Functions +-------------------------------------------*/ +#define INPUT_META_NUM 1 +static vnn_input_meta_t input_meta_tab[INPUT_META_NUM]; +static void _load_input_meta() +{ + uint32_t i; + for (i = 0; i < INPUT_META_NUM; i++) + { + memset(&input_meta_tab[i].image.preprocess, + VNN_PREPRO_NONE, sizeof(int32_t) * VNN_PREPRO_NUM); + } + if (vnn_UseImagePreprocessNode()) + { + /* lid: input.1_121 */ + input_meta_tab[0].image.preprocess[0] = VNN_PREPRO_NONE; + input_meta_tab[0].image.preprocess[1] = VNN_PREPRO_NONE; + input_meta_tab[0].image.preprocess[2] = VNN_PREPRO_NONE; + + } + else + { + /* lid: input.1_121 */ + input_meta_tab[0].image.preprocess[0] = VNN_PREPRO_REORDER; + input_meta_tab[0].image.preprocess[1] = VNN_PREPRO_MEAN; + input_meta_tab[0].image.preprocess[2] = VNN_PREPRO_SCALE; + input_meta_tab[0].image.reorder[0] = 2; + input_meta_tab[0].image.reorder[1] = 1; + input_meta_tab[0].image.reorder[2] = 0; + input_meta_tab[0].image.mean[0] = 128; + input_meta_tab[0].image.mean[1] = 128; + input_meta_tab[0].image.mean[2] = 128; + input_meta_tab[0].image.scale[0] = 0.00390625; + input_meta_tab[0].image.scale[1] = 0.00390625; + input_meta_tab[0].image.scale[2] = 0.00390625; + + } + +} + +static vsi_enum _get_file_type(const char *file_name) +{ + vsi_enum type = 0; + const char *ptr; + char sep = '.'; + uint32_t pos,n; + char buff[32] = {0}; + + ptr = strrchr(file_name, sep); + pos = ptr - file_name; + n = strlen(file_name) - (pos + 1); + strncpy(buff, file_name+(pos+1), n); + + if(strcmp(buff, "jpg") == 0 + || strcmp(buff, "jpeg") == 0 + || strcmp(buff, "JPG") == 0 + || strcmp(buff, "JPEG") == 0 ) + { + type = NN_FILE_JPG; + } + else if(strcmp(buff, "tensor") == 0 + || strcmp(buff, "txt") == 0) + { + char *qnt_suffix = ".qnt.tensor"; + ptr = strstr(file_name, qnt_suffix); + if(ptr && strlen(qnt_suffix)) + { + type = NN_FILE_QTENSOR; + } + else + { + type = NN_FILE_TENSOR; + } + } + else if(strcmp(buff, "qtensor") == 0) + { + type = NN_FILE_QTENSOR; + } + else if(strcmp(buff, "bin") == 0 + || strcmp(buff, "dat") == 0) + { + type = NN_FILE_BINARY; + } + else + { + type = NN_FILE_NONE; + } + + return type; +} + +static vsi_status _jpeg_to_bmp + ( + FILE * inputFile, + unsigned char* bmpData, + vsi_size_t bmpWidth, + vsi_size_t bmpHeight, + vsi_size_t channel + ) +{ + struct jpeg_decompress_struct cinfo; + struct jpeg_error_mgr jerr; + JSAMPARRAY buffer; + unsigned char *point = NULL; + unsigned long width, height; + unsigned short depth = 0; + + cinfo.err = jpeg_std_error(&jerr); + jpeg_create_decompress(&cinfo); + jpeg_stdio_src(&cinfo,inputFile); + jpeg_read_header(&cinfo,TRUE); + + cinfo.dct_method = JDCT_IFAST; + + if (bmpData == NULL) + { + return VSI_FAILURE; + } + else + { + jpeg_start_decompress(&cinfo); + + width = cinfo.output_width; + height = cinfo.output_height; + depth = cinfo.output_components; + if(width * height * depth != bmpWidth * bmpHeight * channel) + { + printf("wrong jpg file , the jpg file size should be %u %u %u\n", + bmpWidth, bmpHeight, channel); + return VSI_FAILURE; + } + + buffer = (*cinfo.mem->alloc_sarray) + ((j_common_ptr)&cinfo, JPOOL_IMAGE, width*depth, 1); + + point = bmpData; + + while (cinfo.output_scanline < height) + { + jpeg_read_scanlines(&cinfo, buffer, 1); + memcpy(point, *buffer, width * depth); + point += width * depth; + } + + jpeg_finish_decompress(&cinfo); + } + + jpeg_destroy_decompress(&cinfo); + + return VSI_SUCCESS; +} + +static uint8_t *_float32_to_dtype + ( + float *fdata, + vsi_nn_tensor_t *tensor + ) +{ + vsi_status status; + uint8_t *data; + vsi_size_t sz,i,stride; + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + data = (uint8_t *)malloc(stride * sz * sizeof(uint8_t)); + TEST_CHECK_PTR(data, final); + memset(data, 0, stride * sz * sizeof(uint8_t)); + + for(i = 0; i < sz; i++) + { + status = vsi_nn_Float32ToDtype(fdata[i], &data[stride * i], &tensor->attr.dtype); + if(status != VSI_SUCCESS) + { + if(data)free(data); + return NULL; + } + } + +final: + return data; +} + +static float *_imageData_to_float32 + ( + uint8_t *bmpData, + vsi_nn_tensor_t *tensor + ) +{ + float *fdata; + vsi_size_t sz,i; + + fdata = NULL; + sz = vsi_nn_GetElementNum(tensor); + fdata = (float *)malloc(sz * sizeof(float)); + TEST_CHECK_PTR(fdata, final); + + for(i = 0; i < sz; i++) + { + fdata[i] = (float)bmpData[i]; + } + +final: + return fdata; +} + +/* + jpg file --> BMP data(dataformat: RGBRGBRGB...) +*/ +static uint8_t *_decode_jpeg + ( + const char *name, + vsi_nn_tensor_t *tensor + ) +{ + FILE *bmpFile; + uint8_t *bmpData; + vsi_size_t sz,w,h,c; + vsi_status status; + + bmpFile = NULL; + bmpData = NULL; + w = tensor->attr.size[0]; + h = tensor->attr.size[1]; + c = tensor->attr.size[2]; + sz = vsi_nn_GetElementNum(tensor); + + bmpFile = fopen( name, "rb" ); + TEST_CHECK_PTR(bmpFile, final); + + bmpData = (uint8_t *)malloc(sz * sizeof(uint8_t)); + TEST_CHECK_PTR(bmpData, final); + memset(bmpData, 0, sz * sizeof(uint8_t)); + + status = _jpeg_to_bmp( bmpFile, bmpData, w, h, c); + if(status == VSI_FAILURE) + { + free(bmpData); + fclose(bmpFile); + return NULL; + } + +final: + if(bmpFile)fclose(bmpFile); + return bmpData; +} + +static void _data_scale + ( + float *fdata, + vnn_input_meta_t *meta, + vsi_nn_tensor_t *tensor + ) +{ + vsi_size_t s0,s1,s2; + vsi_size_t i,j,offset; + float val,scale; + + s0 = tensor->attr.size[0]; + s1 = tensor->attr.size[1]; + s2 = tensor->attr.size[2]; + for(i = 0; i < s2; i++) + { + offset = s0 * s1 * i; + scale = meta->image.scale[i]; + for(j = 0; j < s0 * s1; j++) + { + val = fdata[offset + j] * scale; + fdata[offset + j ] = val; + } + } + +} + +static void _data_mean + ( + float *fdata, + vnn_input_meta_t *meta, + vsi_nn_tensor_t *tensor + ) +{ + vsi_size_t s0,s1,s2; + vsi_size_t i,j,offset; + float val,mean; + + s0 = tensor->attr.size[0]; + s1 = tensor->attr.size[1]; + s2 = tensor->attr.size[2]; + + for(i = 0; i < s2; i++) + { + offset = s0 * s1 * i; + mean = meta->image.mean[i]; + for(j = 0; j < s0 * s1; j++) + { + val = fdata[offset + j] - mean; + fdata[offset + j ] = val; + } + } + +} + +/* + caffe: transpose + reorder + tf: reorder +*/ +static void _data_transform + ( + float *fdata, + vnn_input_meta_t *meta, + vsi_nn_tensor_t *tensor + ) +{ + vsi_size_t s0,s1,s2; + vsi_size_t i,j,offset,sz,order; + float * data; + uint32_t * reorder; + + data = NULL; + reorder = meta->image.reorder; + s0 = tensor->attr.size[0]; + s1 = tensor->attr.size[1]; + s2 = tensor->attr.size[2]; + sz = vsi_nn_GetElementNum(tensor); + data = (float *)malloc(sz * sizeof(float)); + TEST_CHECK_PTR(data, final); + memset(data, 0, sizeof(float) * sz); + + for(i = 0; i < s2; i++) + { + if(s2 > 1 && reorder[i] <= s2) + { + order = reorder[i]; + } + else + { + order = i; + } + + offset = s0 * s1 * i; + for(j = 0; j < s0 * s1; j++) + { + data[j + offset] = fdata[j * s2 + order]; + } + } + + + memcpy(fdata, data, sz * sizeof(float)); +final: + if(data)free(data); +} + +static uint8_t *_get_binary_data + ( + vsi_nn_tensor_t *tensor, + const char *name + ) +{ + uint8_t *tensorData; + vsi_size_t sz,stride,ret,total_sz; + FILE *tensorFile; + + tensorData = NULL; + tensorFile = fopen(name, "rb"); + TEST_CHECK_PTR(tensorFile, error); + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + total_sz = sz * stride; + tensorData = (uint8_t *)malloc(total_sz * sizeof(uint8_t)); + TEST_CHECK_PTR(tensorData, error); + + memset(tensorData, 0, total_sz * sizeof(uint8_t)); + ret = fread(tensorData, 1, total_sz, tensorFile); + if(ret != total_sz) + { + printf("Read %s fail\n", name); + printf("read data %u != tensor sz %u\n", ret, total_sz); + if(tensorData)free(tensorData); + goto error; + } + + if(tensorFile)fclose(tensorFile); + return tensorData; +error: + if(tensorFile)fclose(tensorFile); + return NULL; +} + +static uint8_t *_get_qtensor_data + ( + vsi_nn_tensor_t *tensor, + const char *name + ) +{ + vsi_size_t i = 0; + float fval = 0.0; + uint8_t *tensorData; + vsi_size_t sz = 1,stride = 1; + FILE *tensorFile; + uint16_t uint16_temp_value = 0; + int16_t int16_temp_value = 0; + + tensorData = NULL; + tensorFile = fopen(name, "rb"); + TEST_CHECK_PTR(tensorFile, error); + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride == 0) + { + stride = 1; + } + tensorData = (uint8_t *)malloc(sz * stride * sizeof(uint8_t)); + TEST_CHECK_PTR(tensorData, error); + memset(tensorData, 0, sz * stride * sizeof(uint8_t)); + + for(i = 0; i < sz; i++) + { + if(fscanf( tensorFile, "%f ", &fval ) != 1) + { + printf("Read tensor file fail.\n"); + printf("Please check file lines or if the file contains illegal characters\n"); + goto error; + } + if(1 == stride) + { + if(VSI_NN_TYPE_INT8 == tensor->attr.dtype.vx_type) + tensorData[i * stride] = (int8_t)fval; + else + tensorData[i * stride] = (uint8_t)fval; + } + else if(2 == stride) + { + if(VSI_NN_TYPE_INT16 == tensor->attr.dtype.vx_type) + { + int16_temp_value = (int16_t)fval; + memcpy(tensorData + i * stride, &int16_temp_value, stride * sizeof(uint8_t)); + } + else + { + uint16_temp_value = (uint16_t)fval; + memcpy(tensorData + i * stride, &uint16_temp_value, stride * sizeof(uint8_t)); + } + } + else + { + printf("Do not support quant data with length of %u.\n", stride); + goto error; + } + } + + if(tensorFile)fclose(tensorFile); + return tensorData; +error: + if(tensorFile)fclose(tensorFile); + return NULL; +} + +static uint8_t *_get_tensor_data + ( + vsi_nn_tensor_t *tensor, + const char *name + ) +{ + vsi_status status = VSI_FAILURE; + vsi_size_t i = 0; + float fval = 0.0; + uint8_t *tensorData; + vsi_size_t sz = 1; + vsi_size_t stride = 1; + FILE *tensorFile; + + tensorData = NULL; + tensorFile = fopen(name, "rb"); + TEST_CHECK_PTR(tensorFile, error); + + sz = vsi_nn_GetElementNum(tensor); + stride = vsi_nn_TypeGetBytes(tensor->attr.dtype.vx_type); + if(stride ==0) + { + stride = 1; + } + tensorData = (uint8_t *)malloc(stride * sz * sizeof(uint8_t)); + TEST_CHECK_PTR(tensorData, error); + memset(tensorData, 0, stride * sz * sizeof(uint8_t)); + + for(i = 0; i < sz; i++) + { + if(fscanf( tensorFile, "%f ", &fval ) != 1) + { + printf("Read tensor file fail.\n"); + printf("Please check file lines or if the file contains illegal characters\n"); + goto error; + } + status = vsi_nn_Float32ToDtype(fval, &tensorData[stride * i], &tensor->attr.dtype); + TEST_CHECK_STATUS(status, error); + } + + if(tensorFile)fclose(tensorFile); + return tensorData; +error: + if(tensorFile)fclose(tensorFile); + return NULL; +} + +static uint8_t *_get_jpeg_data + ( + vsi_nn_tensor_t *tensor, + vnn_input_meta_t *meta, + const char *filename + ) +{ + uint32_t i; + uint8_t *bmpData,*data; + float *fdata; + vsi_bool use_image_process = vnn_UseImagePreprocessNode(); + + bmpData = NULL; + fdata = NULL; + data = NULL; + + bmpData = _decode_jpeg(filename, tensor); + TEST_CHECK_PTR(bmpData, final); + + if(use_image_process) + { + data = bmpData; + goto final; + } + + fdata = _imageData_to_float32(bmpData, tensor); + TEST_CHECK_PTR(fdata, final); + + for(i = 0; i < _cnt_of_array(meta->image.preprocess); i++) + { + switch (meta->image.preprocess[i]) + { + case VNN_PREPRO_NONE: + break; + case VNN_PREPRO_REORDER: + _data_transform(fdata, meta, tensor); + break; + case VNN_PREPRO_MEAN: + _data_mean(fdata, meta, tensor); + break; + case VNN_PREPRO_SCALE: + _data_scale(fdata, meta, tensor); + break; + default: + break; + } + } + + data = _float32_to_dtype(fdata, tensor); + TEST_CHECK_PTR(data, final); +final: + if(fdata) + { + free(fdata); + fdata = NULL; + } + if(use_image_process) + { + ; + } + else + { + if(bmpData) + { + free(bmpData); + bmpData = NULL; + } + } + + return data; +} + +#define IMAGE_ADDR_ALIGN_START_SIZE 64 +#define IMAGE_ADDR_ALIGN_BLOCK_SIZE 64 + +static uint8_t *buffer_img = NULL; +static uint8_t *buffer_img_align_addr = NULL; + +static void _get_image_handle_buffer + ( + vsi_size_t width, + vsi_size_t height, + vsi_size_t channels, + vsi_size_t align_start_size, + vsi_size_t align_block_size + ) +{ + vsi_size_t sz; + uint64_t temp; + + sz = width * height * channels + align_start_size + align_block_size; + buffer_img = (uint8_t *)malloc( sz * sizeof( uint8_t ) ); + memset(buffer_img, 0, sizeof( uint8_t ) * sz); + + temp = (uint64_t)(buffer_img) % align_start_size; + if (temp == 0) + { + buffer_img_align_addr = buffer_img; + } + else + { + buffer_img_align_addr = buffer_img + align_start_size - temp; + } +} + +static vsi_status _handle_multiple_inputs + ( + vsi_nn_graph_t *graph, + uint32_t idx, + const char *input_file + ) +{ + vsi_status status; + vsi_nn_tensor_t *tensor; + uint8_t *data; + vnn_input_meta_t meta; + vsi_enum fileType; + char dumpInput[128]; + char *p1 = NULL; + + status = VSI_FAILURE; + data = NULL; + tensor = NULL; + memset(&meta, 0, sizeof(vnn_input_meta_t)); + tensor = vsi_nn_GetTensor( graph, graph->input.tensors[idx] ); + meta = input_meta_tab[idx]; + fileType = _get_file_type(input_file); + switch(fileType) + { + case NN_FILE_JPG: + data = _get_jpeg_data(tensor, &meta, input_file); + TEST_CHECK_PTR(data, final); + break; + case NN_FILE_TENSOR: + data = _get_tensor_data(tensor, input_file); + TEST_CHECK_PTR(data, final); + break; + case NN_FILE_QTENSOR: + data = _get_qtensor_data(tensor, input_file); + TEST_CHECK_PTR(data, final); + break; + case NN_FILE_BINARY: + data = _get_binary_data(tensor, input_file); + TEST_CHECK_PTR(data, final); + break; + default: + printf("error input file type\n"); + break; + } + + /* Copy the Pre-processed data to input tensor */ + status = vsi_nn_CopyDataToTensor(graph, tensor, data); + TEST_CHECK_STATUS(status, final); + + /* Save the image data to file */ + p1 = getenv( "VSI_SAVE_FILE_TYPE"); + + snprintf(dumpInput, sizeof(dumpInput), "input_%d.dat", idx); + vsi_nn_SaveTensorToBinary(graph, tensor, dumpInput); + + + status = VSI_SUCCESS; +final: + if(data)free(data); + return status; +} + +void vnn_ReleaseBufferImage() +{ + if (buffer_img) free(buffer_img); + buffer_img = NULL; +} + +vsi_bool vnn_UseImagePreprocessNode() +{ + int32_t use_img_process; + char *use_img_process_s; + use_img_process = 0; /* default is 0 */ + use_img_process_s = getenv("VSI_USE_IMAGE_PROCESS"); + if(use_img_process_s) + { + use_img_process = atoi(use_img_process_s); + } + if (use_img_process) + { + return TRUE; + } + return FALSE; +} + +vsi_status vnn_PreProcessDynamicFixedPoint16 + ( + vsi_nn_graph_t *graph, + const char **inputs, + uint32_t input_num + ) +{ + uint32_t i; + vsi_status status; + status = VSI_FAILURE; + _load_input_meta(); + if(input_num != graph->input.num) + { + printf("Graph need %u inputs, but enter %u inputs!!!\n", + graph->input.num, input_num); + return status; + } + for(i = 0; i < input_num; i++) + { + status = _handle_multiple_inputs(graph, i, inputs[i]); + TEST_CHECK_STATUS(status, final); + } + + status = VSI_SUCCESS; +final: + return status; +} + +vsi_size_t vnn_LoadFP32DataFromTextFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ) +{ + float fval = 0.0; + vsi_size_t i = 0; + uint8_t * buffer = NULL; + vsi_size_t item_ount = 0; + vsi_size_t read_size = 0; + vsi_size_t stride = sizeof(fval); + FILE *fp = NULL; + + if(!fname || !buffer_ptr || !buffer_sz) + { + return read_size; + } + + fp = fopen(fname, "rb"); + if(fp) + { + while(!feof(fp) && fscanf( fp, "%f ", &fval ) == 1) + { + item_ount++; + } + + if(item_ount > 0) + { + read_size = item_ount * stride; + buffer = (uint8_t *)malloc(read_size); + if(buffer) + { + int fail_to_read = FALSE; + + VSI_FSEEK(fp, 0, SEEK_SET); + for(i = 0; i < item_ount && !fail_to_read; i++) + { + if(fscanf( fp, "%f ", (float *)&buffer[stride * i] ) != 1) + { + printf("Read tensor file fail.\n"); + printf("Please check file lines or if the file contains illegal characters\n"); + free(buffer); + fail_to_read = TRUE; + read_size = 0; + break; + } + } + + if(!fail_to_read) + { + *buffer_ptr = buffer; + *buffer_sz = read_size; + } + } + else + { + read_size = 0; + printf("Allocate memory fail!\n"); + } + } + else + { + printf("No available data found!\n"); + } + fclose(fp); + } + else + { + printf("Fail to open %s\n", fname); + } + + if(!read_size) + { + printf("Load data from %s fail!\n", fname); + } + + return read_size; +} + +vsi_size_t vnn_LoadRawDataFromBinaryFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ) +{ + FILE * fp = NULL; + vsi_size_t fsize = 0; + vsi_size_t read_size = 0; + uint8_t* buffer = NULL; + + if(!fname || !buffer_ptr || !buffer_sz) + { + return fsize; + } + + fp = fopen(fname, "rb"); + if(fp) + { + fsize = VSI_FSEEK(fp, 0, SEEK_END); + fsize = ftell(fp); + + buffer = (uint8_t *)malloc(fsize); + if(buffer) + { + VSI_FSEEK(fp, 0, SEEK_SET); + read_size = fread(buffer, 1, fsize, fp); + if(read_size == fsize) + { + *buffer_ptr = buffer; + *buffer_sz = read_size; + } + else + { + fsize = 0; + free(buffer); + buffer = NULL; + } + } + else + { + fsize = 0; + printf("Allocate memory fail!\n"); + } + + if(fp) + { + fclose(fp); + } + } + + if(!fsize) + { + printf("Load data from %s fail!\n", fname); + } + return fsize; +} + +const vsi_nn_preprocess_map_element_t * vnn_GetPreProcessMap() +{ + return preprocess_map; +} + +uint32_t vnn_GetPreProcessMapCount() +{ + if (preprocess_map == NULL) + return 0; + else + return sizeof(preprocess_map) / sizeof(vsi_nn_preprocess_map_element_t); +} diff --git a/resource/openpose/wksp/dynamic_fixed_point-16/vnn_pre_process.h b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_pre_process.h new file mode 100644 index 0000000..e7de270 --- /dev/null +++ b/resource/openpose/wksp/dynamic_fixed_point-16/vnn_pre_process.h @@ -0,0 +1,72 @@ +/**************************************************************************** +* Generated by NETRANS 6.27.0 +* Match ovxlib 1.1.53 +* +* Neural Network appliction pre-process header file +****************************************************************************/ +#ifndef _VNN_PRE_PROCESS_H_ +#define _VNN_PRE_PROCESS_H_ + +typedef enum _vnn_file_type +{ + NN_FILE_NONE, + NN_FILE_TENSOR, + NN_FILE_QTENSOR, + NN_FILE_JPG, + NN_FILE_BINARY +} vnn_file_type_e; + +typedef enum _vnn_pre_order +{ + VNN_PREPRO_NONE = -1, + VNN_PREPRO_REORDER, + VNN_PREPRO_MEAN, + VNN_PREPRO_SCALE, + VNN_PREPRO_NUM +} vnn_pre_order_e; + +typedef struct _vnn_input_meta +{ + union + { + struct + { + int32_t preprocess[VNN_PREPRO_NUM]; + uint32_t reorder[4]; + float mean[4]; + float scale[4]; + int32_t channel_count; + } image; + }; +} vnn_input_meta_t; + +vsi_status vnn_PreProcessDynamicFixedPoint16 + ( + vsi_nn_graph_t *graph, + const char **inputs, + uint32_t input_num + ); + +vsi_bool vnn_UseImagePreprocessNode(); + +void vnn_ReleaseBufferImage(); + +vsi_size_t vnn_LoadFP32DataFromTextFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ); + +vsi_size_t vnn_LoadRawDataFromBinaryFile + ( + const char * fname, + uint8_t ** buffer_ptr, + vsi_size_t * buffer_sz + ); + +const vsi_nn_preprocess_map_element_t * vnn_GetPreProcessMap(); + +uint32_t vnn_GetPreProcessMapCount(); + +#endif diff --git a/src/openpose.py b/src/openpose.py new file mode 100644 index 0000000..46022dc --- /dev/null +++ b/src/openpose.py @@ -0,0 +1,475 @@ +import cv2 +import numpy as np +import math +from operator import itemgetter +from utils.client import * +from utils.common import load_image_cv, init_service_rpc, fiducial_e + + +BODY_PARTS_KPT_IDS = [[1, 2], [1, 5], [2, 3], [3, 4], [5, 6], [6, 7], [1, 8], [8, 9], [9, 10], [1, 11], + [11, 12], [12, 13], [1, 0], [0, 14], [14, 16], [0, 15], [15, 17], [2, 16], [5, 17]] +BODY_PARTS_PAF_IDS = ([12, 13], [20, 21], [14, 15], [16, 17], [22, 23], [24, 25], [0, 1], [2, 3], [4, 5], + [6, 7], [8, 9], [10, 11], [28, 29], [30, 31], [34, 35], [32, 33], [36, 37], [18, 19], [26, 27]) + +def get_alpha(rate=30, cutoff=1): + tau = 1 / (2 * math.pi * cutoff) + te = 1 / rate + return 1 / (1 + tau / te) + + +class LowPassFilter: + """低通滤波器 + """ + def __init__(self): + self.x_previous = None + + def __call__(self, x, alpha=0.5): + if self.x_previous is None: + self.x_previous = x + return x + x_filtered = alpha * x + (1 - alpha) * self.x_previous + self.x_previous = x_filtered + return x_filtered + +class OneEuroFilter: + """ 由两路低通滤波组成： + filter_dx：对输入导数的低通滤波 + filter_x：对原始信号本身的低通滤波 + """ + def __init__(self, freq=15, mincutoff=1, beta=0.05, dcutoff=1): + self.freq = freq + self.mincutoff = mincutoff + self.beta = beta + self.dcutoff = dcutoff + self.filter_x = LowPassFilter() + self.filter_dx = LowPassFilter() + self.x_previous = None + self.dx = None + + def __call__(self, x): + if self.dx is None: + self.dx = 0 + else: + self.dx = (x - self.x_previous) * self.freq + dx_smoothed = self.filter_dx(self.dx, get_alpha(self.freq, self.dcutoff)) + cutoff = self.mincutoff + self.beta * abs(dx_smoothed) + x_filtered = self.filter_x(x, get_alpha(self.freq, cutoff)) + self.x_previous = x + return x_filtered + + +class Pose: + """解析关键节的位置并画图 + """ + num_kpts = 18 + kpt_names = ['nose', 'neck', + 'r_sho', 'r_elb', 'r_wri', 'l_sho', 'l_elb', 'l_wri', + 'r_hip', 'r_knee', 'r_ank', 'l_hip', 'l_knee', 'l_ank', + 'r_eye', 'l_eye', + 'r_ear', 'l_ear'] + sigmas = np.array([.26, .79, .79, .72, .62, .79, .72, .62, 1.07, .87, .89, 1.07, .87, .89, .25, .25, .35, .35], + dtype=np.float32) / 10.0 + vars = (sigmas * 2) ** 2 + last_id = -1 + color = [0, 224, 255] + + def __init__(self, keypoints, confidence): + super().__init__() + self.keypoints = keypoints + self.confidence = confidence + self.bbox = Pose.get_bbox(self.keypoints) + self.id = None + self.filters = [[OneEuroFilter(), OneEuroFilter()] for _ in range(Pose.num_kpts)] + + @staticmethod + def get_bbox(keypoints): + """从一组关键点(keypoints)中筛选出有效的关键点（即坐标不是 -1）的子集， + 并计算这些有效关键点的最小外接矩形（bounding box） + ---------- + keypoints : 关键点 + + Returns + ------- + 最小外接矩形 + """ + found_keypoints = np.zeros((np.count_nonzero(keypoints[:, 0] != -1), 2), dtype=np.int32) + found_kpt_id = 0 + for kpt_id in range(Pose.num_kpts): + if keypoints[kpt_id, 0] == -1: + continue + found_keypoints[found_kpt_id] = keypoints[kpt_id] + found_kpt_id += 1 + bbox = cv2.boundingRect(found_keypoints) + return bbox + + + def draw(self, img): + """将关键点画到原图上 + + Parameters + ---------- + img : 原图 + """ + assert self.keypoints.shape == (Pose.num_kpts, 2) + + for part_id in range(len(BODY_PARTS_PAF_IDS) - 2): + kpt_a_id = BODY_PARTS_KPT_IDS[part_id][0] + global_kpt_a_id = self.keypoints[kpt_a_id, 0] + if global_kpt_a_id != -1: + x_a, y_a = self.keypoints[kpt_a_id] + cv2.circle(img, (int(x_a), int(y_a)), 3, Pose.color, -1) + kpt_b_id = BODY_PARTS_KPT_IDS[part_id][1] + global_kpt_b_id = self.keypoints[kpt_b_id, 0] + if global_kpt_b_id != -1: + x_b, y_b = self.keypoints[kpt_b_id] + cv2.circle(img, (int(x_b), int(y_b)), 3, Pose.color, -1) + if global_kpt_a_id != -1 and global_kpt_b_id != -1: + cv2.line(img, (int(x_a), int(y_a)), (int(x_b), int(y_b)), Pose.color, 1) + + + + +def extract_keypoints(heatmap, all_keypoints, total_keypoint_num): + """从一个热力图（heatmap）中检测并提取关键点（keypoints） + + Parameters + ---------- + heatmap : 二维的热力图，通常表示某个关键点的检测概率或置信度分布 + all_keypoints : 所有检测出的关键点及其信息 + total_keypoint_num : 已存在的总关键点数量 + + Returns + ------- + 本次检测到的有效关键点数量 + """ + heatmap[heatmap < 0.1] = 0 + heatmap_with_borders = np.pad(heatmap, [(2, 2), (2, 2)], mode='constant') + heatmap_center = heatmap_with_borders[1:heatmap_with_borders.shape[0]-1, 1:heatmap_with_borders.shape[1]-1] + heatmap_left = heatmap_with_borders[1:heatmap_with_borders.shape[0]-1, 2:heatmap_with_borders.shape[1]] + heatmap_right = heatmap_with_borders[1:heatmap_with_borders.shape[0]-1, 0:heatmap_with_borders.shape[1]-2] + heatmap_up = heatmap_with_borders[2:heatmap_with_borders.shape[0], 1:heatmap_with_borders.shape[1]-1] + heatmap_down = heatmap_with_borders[0:heatmap_with_borders.shape[0]-2, 1:heatmap_with_borders.shape[1]-1] + + heatmap_peaks = (heatmap_center > heatmap_left) &\ + (heatmap_center > heatmap_right) &\ + (heatmap_center > heatmap_up) &\ + (heatmap_center > heatmap_down) + heatmap_peaks = heatmap_peaks[1:heatmap_center.shape[0]-1, 1:heatmap_center.shape[1]-1] + keypoints = list(zip(np.nonzero(heatmap_peaks)[1], np.nonzero(heatmap_peaks)[0])) # (w, h) + keypoints = sorted(keypoints, key=itemgetter(0)) + + suppressed = np.zeros(len(keypoints), np.uint8) + keypoints_with_score_and_id = [] + keypoint_num = 0 + for i in range(len(keypoints)): + if suppressed[i]: + continue + for j in range(i+1, len(keypoints)): + if math.sqrt((keypoints[i][0] - keypoints[j][0]) ** 2 + + (keypoints[i][1] - keypoints[j][1]) ** 2) < 6: + suppressed[j] = 1 + keypoint_with_score_and_id = (keypoints[i][0], keypoints[i][1], heatmap[keypoints[i][1], keypoints[i][0]], + total_keypoint_num + keypoint_num) + keypoints_with_score_and_id.append(keypoint_with_score_and_id) + keypoint_num += 1 + all_keypoints.append(keypoints_with_score_and_id) + return keypoint_num + + +def connections_nms(a_idx, b_idx, affinity_scores): + """对一组成对点对及其对应的亲和度分数（affinity scores）进行排序和去重选择， + 得到一个“非极大抑制（NMS）”风格的点对子集 + + Parameters + ---------- + a_idx : 第一个关键点集合的索引数组 + b_idx : 第二个关键点集合的索引数组 + affinity_scores : 每对点的亲和度分数，一维数组，越大表示越可信的连接 + + Returns + ------- + 一个不重叠的、置信度较高的点对集合 + """ + order = affinity_scores.argsort()[::-1] + affinity_scores = affinity_scores[order] + a_idx = a_idx[order] + b_idx = b_idx[order] + idx = [] + has_kpt_a = set() + has_kpt_b = set() + for t, (i, j) in enumerate(zip(a_idx, b_idx)): + if i not in has_kpt_a and j not in has_kpt_b: + idx.append(t) + has_kpt_a.add(i) + has_kpt_b.add(j) + idx = np.asarray(idx, dtype=np.int32) + return a_idx[idx], b_idx[idx], affinity_scores[idx] + + +def group_keypoints(all_keypoints_by_type, pafs, pose_entry_size=20, min_paf_score=0.05): + """一组人体关键点及其关节连线的检测结果中，基于对齐的对齐关节（骨架）关系和部分字段的评分， + 组装成一个或多个姿态入口（pose_entries），并返回更新后的所有关键点集合（all_keypoints） + + Parameters + ---------- + all_keypoints_by_type : 一组按类型划分的关键点集合 + pafs : 部分关节对场，用于计算候选骨架的亲和度评分 + pose_entry_size : 每个姿态条目的固定长度，默认值为 20 + min_paf_score : 匹配分数阈值 + + Returns + ------- + 过滤后的姿态条目集合;更新后的关键点集合 + """ + pose_entries = [] + all_keypoints = np.array([item for sublist in all_keypoints_by_type for item in sublist]) + points_per_limb = 10 + grid = np.arange(points_per_limb, dtype=np.float32).reshape(1, -1, 1) + all_keypoints_by_type = [np.array(keypoints, np.float32) for keypoints in all_keypoints_by_type] + for part_id in range(len(BODY_PARTS_PAF_IDS)): + part_pafs = pafs[:, :, BODY_PARTS_PAF_IDS[part_id]] + kpts_a = all_keypoints_by_type[BODY_PARTS_KPT_IDS[part_id][0]] + kpts_b = all_keypoints_by_type[BODY_PARTS_KPT_IDS[part_id][1]] + n = len(kpts_a) + m = len(kpts_b) + if n == 0 or m == 0: + continue + + # Get vectors between all pairs of keypoints, i.e. candidate limb vectors. + a = kpts_a[:, :2] + a = np.broadcast_to(a[None], (m, n, 2)) + b = kpts_b[:, :2] + vec_raw = (b[:, None, :] - a).reshape(-1, 1, 2) + + # Sample points along every candidate limb vector. + steps = (1 / (points_per_limb - 1) * vec_raw) + points = steps * grid + a.reshape(-1, 1, 2) + points = points.round().astype(dtype=np.int32) + x = points[..., 0].ravel() + y = points[..., 1].ravel() + + # Compute affinity score between candidate limb vectors and part affinity field. + field = part_pafs[y, x].reshape(-1, points_per_limb, 2) + vec_norm = np.linalg.norm(vec_raw, ord=2, axis=-1, keepdims=True) + vec = vec_raw / (vec_norm + 1e-6) + affinity_scores = (field * vec).sum(-1).reshape(-1, points_per_limb) + valid_affinity_scores = affinity_scores > min_paf_score + valid_num = valid_affinity_scores.sum(1) + affinity_scores = (affinity_scores * valid_affinity_scores).sum(1) / (valid_num + 1e-6) + success_ratio = valid_num / points_per_limb + + # Get a list of limbs according to the obtained affinity score. + valid_limbs = np.where(np.logical_and(affinity_scores > 0, success_ratio > 0.8))[0] + if len(valid_limbs) == 0: + continue + b_idx, a_idx = np.divmod(valid_limbs, n) + affinity_scores = affinity_scores[valid_limbs] + + # Suppress incompatible connections. + a_idx, b_idx, affinity_scores = connections_nms(a_idx, b_idx, affinity_scores) + connections = list(zip(kpts_a[a_idx, 3].astype(np.int32), + kpts_b[b_idx, 3].astype(np.int32), + affinity_scores)) + if len(connections) == 0: + continue + + if part_id == 0: + pose_entries = [np.ones(pose_entry_size) * -1 for _ in range(len(connections))] + for i in range(len(connections)): + pose_entries[i][BODY_PARTS_KPT_IDS[0][0]] = connections[i][0] + pose_entries[i][BODY_PARTS_KPT_IDS[0][1]] = connections[i][1] + pose_entries[i][-1] = 2 + pose_entries[i][-2] = np.sum(all_keypoints[connections[i][0:2], 2]) + connections[i][2] + elif part_id == 17 or part_id == 18: + kpt_a_id = BODY_PARTS_KPT_IDS[part_id][0] + kpt_b_id = BODY_PARTS_KPT_IDS[part_id][1] + for i in range(len(connections)): + for j in range(len(pose_entries)): + if pose_entries[j][kpt_a_id] == connections[i][0] and pose_entries[j][kpt_b_id] == -1: + pose_entries[j][kpt_b_id] = connections[i][1] + elif pose_entries[j][kpt_b_id] == connections[i][1] and pose_entries[j][kpt_a_id] == -1: + pose_entries[j][kpt_a_id] = connections[i][0] + continue + else: + kpt_a_id = BODY_PARTS_KPT_IDS[part_id][0] + kpt_b_id = BODY_PARTS_KPT_IDS[part_id][1] + for i in range(len(connections)): + num = 0 + for j in range(len(pose_entries)): + if pose_entries[j][kpt_a_id] == connections[i][0]: + pose_entries[j][kpt_b_id] = connections[i][1] + num += 1 + pose_entries[j][-1] += 1 + pose_entries[j][-2] += all_keypoints[connections[i][1], 2] + connections[i][2] + if num == 0: + pose_entry = np.ones(pose_entry_size) * -1 + pose_entry[kpt_a_id] = connections[i][0] + pose_entry[kpt_b_id] = connections[i][1] + pose_entry[-1] = 2 + pose_entry[-2] = np.sum(all_keypoints[connections[i][0:2], 2]) + connections[i][2] + pose_entries.append(pose_entry) + + filtered_entries = [] + for i in range(len(pose_entries)): + if pose_entries[i][-1] < 3 or (pose_entries[i][-2] / pose_entries[i][-1] < 0.2): + continue + filtered_entries.append(pose_entries[i]) + pose_entries = np.asarray(filtered_entries) + return pose_entries, all_keypoints + + + +def calculate_poses(heatmaps, pafs): + """通过热力图（heatmaps）和部分关节对场（PAFs）估计并组装出一个或多个姿态（poses） + + Parameters + ---------- + heatmaps : 原始的热力图张量 + pafs : 部分关节对场（PAFs），用于评估骨架连接的置信度 + + Returns + ------- + 包含若干 Pose 对象的列表 current_poses + """ + stride = 8 + upsample_ratio = 4 + scale = 1 + num_keypoints = Pose.num_kpts + heatmaps = cv2.resize(heatmaps, (0, 0), fx=upsample_ratio, fy=upsample_ratio, interpolation=cv2.INTER_CUBIC) + pafs = cv2.resize(pafs, (0, 0), fx=upsample_ratio, fy=upsample_ratio, interpolation=cv2.INTER_CUBIC) + total_keypoints_num = 0 + all_keypoints_by_type = [] + for kpt_idx in range(num_keypoints): # 19th for bg + total_keypoints_num += extract_keypoints(heatmaps[:, :, kpt_idx], all_keypoints_by_type, total_keypoints_num) + + pose_entries, all_keypoints = group_keypoints(all_keypoints_by_type, pafs) + for kpt_id in range(all_keypoints.shape[0]): + all_keypoints[kpt_id, 0] = (all_keypoints[kpt_id, 0] * stride / upsample_ratio) / scale + all_keypoints[kpt_id, 1] = (all_keypoints[kpt_id, 1] * stride / upsample_ratio) / scale + current_poses = [] + for n in range(len(pose_entries)): + if len(pose_entries[n]) == 0: + continue + pose_keypoints = np.ones((num_keypoints, 2), dtype=np.int32) * -1 + for kpt_id in range(num_keypoints): + if pose_entries[n][kpt_id] != -1.0: # keypoint was found + pose_keypoints[kpt_id, 0] = int(all_keypoints[int(pose_entries[n][kpt_id]), 0]) + pose_keypoints[kpt_id, 1] = int(all_keypoints[int(pose_entries[n][kpt_id]), 1]) + pose = Pose(pose_keypoints, pose_entries[n][18]) + current_poses.append(pose) + return current_poses + + +def binary2np(rsp): + """ + 二进制数据转为指定size的np + + Parameters + ---------- + rsp : 板卡推理结果 + + Returns + ------- + 返回np + """ + ret_bytes = [] + for ret in rsp['tensor']: + ret_bytes.append(base64.b64decode(ret)) + + stage2_heatmaps = np.frombuffer(ret_bytes[2], dtype=np.float32).flatten() + stage2_pafs = np.frombuffer(ret_bytes[3], dtype=np.float32).flatten() + stage2_heatmaps = np.transpose(stage2_heatmaps.reshape(19, 32, 53), (1, 2, 0)) + stage2_pafs = np.transpose(stage2_pafs.reshape(38, 32, 53), (1, 2, 0)) + return stage2_heatmaps, stage2_pafs + + +def draw_content(im_src, current_poses): + """图像上标注检测结果 + + Args: + im_src : 输入图片 + current_poses : 关键点信息 + + Returns: + 结果图 + """ + draw_img = im_src.copy() + if len(current_poses) > 0: + orig_img = im_src.copy() + for pose in current_poses: + pose.draw(draw_img) + + img_tmp = cv2.addWeighted(orig_img, 0.6, draw_img, 0.4, 0) + for pose in current_poses: + cv2.rectangle(img_tmp, (pose.bbox[0], pose.bbox[1]), + (pose.bbox[0] + pose.bbox[2], pose.bbox[1] + pose.bbox[3]), (0, 255, 0)) + return img_tmp + else: + return draw_img + + +def post_process(rsp, im_src): + """后处理 + + Args: + rsp : 板卡推理结果 + im_src : 原图 + + Returns: + 结果图 + """ + stage2_heatmaps, stage2_pafs = binary2np(rsp) + current_poses = calculate_poses(stage2_heatmaps, stage2_pafs) + result_image = draw_content(im_src, current_poses) + return result_image + +def compare_result(rsp): + """ + 对比PC端网络输出和板卡的结果的fe + Parameters + ---------- + rsp : 板卡推理结果 + """ + ret_bytes = [] + for ret in rsp['tensor']: + ret_bytes.append(base64.b64decode(ret)) + infer_result_0 = np.frombuffer(ret_bytes[2], dtype=np.float32).flatten() + infer_result_1 = np.frombuffer(ret_bytes[3], dtype=np.float32).flatten() + + output_2 = np.load('../resource/openpose/output_2.npy') + output_3 = np.load('../resource/openpose/output_3.npy') + fe_result_0 = fiducial_e(infer_result_0, output_2) + fe_result_1 = fiducial_e(infer_result_1, output_3) + + print('infer_time', rsp['infer_time']) + print(' 输出1的板卡 和pc fe_mean fe_max', fe_result_0.mean(), fe_result_0.max()) + print(' 输出2的板卡 和pc fe_mean fe_max', fe_result_1.mean(), fe_result_1.max()) + +def data_preprocessing(img_path): + """数据前处理后转成base64格式 + + Parameters + ---------- + img_path : 图片路径 + Returns + ------- + 原图以及base64格式数据 + """ + img_src = load_image_cv(img_path) + resize_imag = cv2.resize(img_src, (424, 256)) + img_src_b64 = base64.b64encode(np.ascontiguousarray(resize_imag)).decode('utf-8') + return resize_imag, img_src_b64 + +if __name__ == "__main__": + method_model = 'openpose' + src_img, img_b64 = data_preprocessing('../resource/openpose/1.jpg') + + init_service_rpc('../resource/openpose/wksp/dynamic_fixed_point-16/network_binary.nb', method_model) + ret_infer = rpc_call(method_model, img_b64) + + result_image = post_process(ret_infer, src_img) + cv2.imwrite('../resource/openpose/result.jpg', result_image) + compare_result(ret_infer) + + + diff --git a/src/utils/__pycache__/client.cpython-39.pyc b/src/utils/__pycache__/client.cpython-39.pyc deleted file mode 100644 index 240cb81..0000000 Binary files a/src/utils/__pycache__/client.cpython-39.pyc and /dev/null differ diff --git a/src/utils/__pycache__/common.cpython-39.pyc b/src/utils/__pycache__/common.cpython-39.pyc deleted file mode 100644 index be9b219..0000000 Binary files a/src/utils/__pycache__/common.cpython-39.pyc and /dev/null differ diff --git a/src/yolov5s_seg.py b/src/yolov5s_seg.py index 892ce15..354b73e 100644 --- a/src/yolov5s_seg.py +++ b/src/yolov5s_seg.py @@ -86,25 +86,6 @@ names = {0: 'person', 1: 'bicycle', 2: 'car', 3: 'motorcycle', 4: 'airplane', 5: -def infer(img_data, web_service): - """下位机推理 - - Args: - img_data (np.array): 推理数据 - web_service: 实例化service - - Returns: - 推理结果信息 - """ - rsp = web_service.infer(img_data) - infer_result0 = np.frombuffer(rsp['output_0.dat'], dtype=np.float32).flatten() - infer_result1 = np.frombuffer(rsp['output_1.dat'], dtype=np.float32).flatten() - np_reshaped = infer_result0.reshape(1, 25200, 117) - pred = torch.from_numpy(np_reshaped) - np_reshaped1 = infer_result1.reshape(1, 32, 160, 160) - proto = torch.from_numpy(np_reshaped1) - return pred, proto - def binary2tensor(rsp): """ 二进制数据转为指定size的tensor