chenyh
/
MNIST_Example
mirror of https://openi.pcl.ac.cn/OpenIOSSG/MNIST_Example.git
0
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

更新使用注释

This commit is contained in:
liuzx 2023-05-04 17:16:29 +08:00
parent 66d110fe7c
commit 4115f78ced
4 changed files with 25 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
pretrain.py
View File

@ -4,7 +4,7 @@
使用注意事项：
1、本示例只支持单数据集训练，如果是多数据集，请参考多数据集训练示例train_for_multidataset.py
2、本示例支持选择预训练模型单文件或多文件
3、本示例需要用户定义的参数有--multi_data_url,--pretrain_url,--train_url,--device_target，这4个参数在单数据集任务中必须定义
3、本示例需要用户定义的参数有--multi_data_url,--pretrain_url,--train_url，这3个参数在单数据集任务中必须定义
具体的含义如下：
--multi_data_url是启智平台上选择的数据集的obs路径
--pretrain_url是启智平台上选择的预训练模型文件的obs路径
@ -78,10 +78,13 @@ if __name__ == "__main__":
PretrainToEnv(args.pretrain_url, pretrain_dir)
device_num = int(os.getenv('RANK_SIZE'))
#使用单卡时
if device_num == 1:
DatasetToEnv(args.multi_data_url,data_dir)
context.set_context(mode=context.GRAPH_MODE,device_target=args.device_target)
ds_train = create_dataset(os.path.join(data_dir, "train"), cfg.batch_size)
#使用数据集的方式
ds_train = create_dataset(os.path.join(data_dir + "/MNISTData", "train"), cfg.batch_size)
#使用多卡时
if device_num > 1:
# set device_id and init for multi-card training
context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target, device_id=int(os.getenv('ASCEND_DEVICE_ID')))
@ -96,7 +99,7 @@ if __name__ == "__main__":
#and Wait for 0th card to finish copying data
while not os.path.exists("/cache/download_input.txt"):
time.sleep(1)
ds_train = create_dataset_parallel(os.path.join(data_dir, "train"), cfg.batch_size)
ds_train = create_dataset_parallel(os.path.join(data_dir + "/MNISTData", "train"), cfg.batch_size)
network = LeNet5(cfg.num_classes)
net_loss = nn.SoftmaxCrossEntropyWithLogits(sparse=True, reduction="mean")

9
pretrain_for_c2net.py
View File

@ -1,13 +1,13 @@
"""
######################## Attention! ########################
使用注意事项：
1、本示例需要用户定义的参数有--multi_data_url,--pretrain_url,--train_url,--device_target，这3个参数任务中必须定义
1、本示例需要用户定义的参数有--multi_data_url,--pretrain_url,--train_url,这3个参数在任务中必须定义
具体的含义如下：
--multi_data_url是启智平台上选择的数据集的obs路径
--pretrain_url是启智平台上选择的预训练模型文件的obs路径
--train_url是训练结果回传到启智平台的obs路径
2、用户需要调用OpenI.C2NETMultiDatasetToEnv等函数，来实现数据集、预训练模型文件的拷贝
3、智算网络中区别于启智的有：
3、智算网络区别于启智：
(1)智算的数据集拷贝到训练镜像后需要解压，请使用C2NETMultiDatasetToEnv函数
(2)智算任务结果不需要用户调用函数回传，会在训练结束后自动回传结果
"""
@ -72,10 +72,13 @@ if __name__ == "__main__":
PretrainToEnv(args.pretrain_url, pretrain_dir)
device_num = int(os.getenv('RANK_SIZE'))
#使用单卡时
if device_num == 1:
DatasetToEnv(args.multi_data_url,data_dir)
context.set_context(mode=context.GRAPH_MODE,device_target=args.device_target)
ds_train = create_dataset(os.path.join(data_dir, "train"), cfg.batch_size)
#使用数据集的方式
ds_train = create_dataset(os.path.join(data_dir + "/MNISTData", "train"), cfg.batch_size)
#使用多卡时
if device_num > 1:
# set device_id and init for multi-card training
context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target, device_id=int(os.getenv('ASCEND_DEVICE_ID')))

7
train.py
View File

@ -64,10 +64,13 @@ if __name__ == "__main__":
DatasetToEnv(args.multi_data_url, data_dir)
device_num = int(os.getenv('RANK_SIZE'))
#使用单卡时
if device_num == 1:
DatasetToEnv(args.multi_data_url,data_dir)
context.set_context(mode=context.GRAPH_MODE,device_target=args.device_target)
ds_train = create_dataset(os.path.join(data_dir, "train"), cfg.batch_size)
#使用数据集的方式
ds_train = create_dataset(os.path.join(data_dir + "/MNISTData", "train"), cfg.batch_size)
#使用多卡时
if device_num > 1:
# set device_id and init for multi-card training
context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target, device_id=int(os.getenv('ASCEND_DEVICE_ID')))
@ -81,7 +84,7 @@ if __name__ == "__main__":
#If the cache file does not exist, it means that the copy data has not been completed,
#and Wait for 0th card to finish copying data
while not os.path.exists("/cache/download_input.txt"):
time.sleep(1)
time.sleep(1)
ds_train = create_dataset_parallel(os.path.join(data_dir + "/MNISTData", "train"), cfg.batch_size)
network = LeNet5(cfg.num_classes)

13
train_for_c2net.py
View File

@ -1,12 +1,12 @@
"""
######################## Attention! ########################
使用注意事项：
1、本示例需要用户定义的参数有--multi_data_url,--train_url,--device_target，这3个参数任务中必须定义
1、本示例需要用户定义的参数有--multi_data_url,--train_url,这2个参数任务中必须定义
具体的含义如下：
--multi_data_url是启智平台上选择的数据集的obs路径
--train_url是训练结果回传到启智平台的obs路径
2、用户需要调用OpenI.C2NETMultiDatasetToEnv等函数，来实现数据集、预训练模型文件的拷贝
3、智算网络中区别于启智的有：
3、智算网络区别于启智：
(1)智算的数据集拷贝到训练镜像后需要解压，请使用C2NETMultiDatasetToEnv函数
(2)智算任务结果不需要用户调用函数回传，会在训练结束后自动回传
"""
@ -30,7 +30,7 @@ from OpenI import C2NETMultiDatasetToEnv as DatasetToEnv
parser = argparse.ArgumentParser(description='MindSpore Lenet Example')
parser.add_argument('--multi_data_url',
help='使用单数据集训练时，需要定义的参数',
help='使用数据集，需要定义的参数',
default= '[{}]')
parser.add_argument('--train_url',
@ -61,10 +61,13 @@ if __name__ == "__main__":
###Initialize and copy data to training image
DatasetToEnv(args.multi_data_url, data_dir)
device_num = int(os.getenv('RANK_SIZE'))
#使用单卡时
if device_num == 1:
DatasetToEnv(args.multi_data_url,data_dir)
context.set_context(mode=context.GRAPH_MODE,device_target=args.device_target)
ds_train = create_dataset(os.path.join(data_dir, "train"), cfg.batch_size)
#使用数据集的方式
ds_train = create_dataset(os.path.join(data_dir + "/MNISTData", "train"), cfg.batch_size)
#使用多卡时
if device_num > 1:
# set device_id and init for multi-card training
context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target, device_id=int(os.getenv('ASCEND_DEVICE_ID')))
@ -78,7 +81,7 @@ if __name__ == "__main__":
#If the cache file does not exist, it means that the copy data has not been completed,
#and Wait for 0th card to finish copying data
while not os.path.exists("/cache/download_input.txt"):
time.sleep(1)
time.sleep(1)
ds_train = create_dataset_parallel(os.path.join(data_dir + "/MNISTData", "train"), cfg.batch_size)
network = LeNet5(cfg.num_classes)