bjutsecurity22
/
mindspore2022
forked from huawei/mindspore2022
0
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

fix parallel timeout

This commit is contained in:
yao_yf 2021-04-02 20:19:26 +08:00
parent 5298b1d225
commit a83fb3316b
32 changed files with 45 additions and 45 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
tests/ut/python/parallel/test_auto_parallel_double_subgraphs.py
View File

@ -105,7 +105,7 @@ class TrainStepWarp(nn.Cell):
def test_double_subgraphs():
_set_multi_subgraphs()
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(device_num=8, global_rank=0)
context.set_auto_parallel_context(parallel_mode="auto_parallel")
net = TrainStepWarp(NetWithLoss(Net()))
@ -156,7 +156,7 @@ class DatasetLenet():
return self
def test_double_subgraphs_train():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(device_num=1, global_rank=0)
context.set_auto_parallel_context(parallel_mode="auto_parallel")
net = TrainStepWarp(NetWithLoss(Net()))

2
tests/ut/python/parallel/test_auto_parallel_for_loop.py
View File

@ -118,7 +118,7 @@ _w1 = Tensor(np.ones([512, 128]), dtype=ms.float32)
def test_auto_parallel():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=16, global_rank=0)
net = Full(_w1, 3)
net.set_auto_parallel()

2
tests/ut/python/parallel/test_auto_parallel_for_loop_multi_subgraph.py
View File

@ -121,7 +121,7 @@ class TrainStepWarp(nn.Cell):
def test_double_subgraphs():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=8, global_rank=0)
net = TrainStepWarp(NetWithLoss(Net()))
_set_multi_subgraphs()

2
tests/ut/python/parallel/test_auto_parallel_for_loop_reshape.py
View File

@ -125,7 +125,7 @@ _w1 = Tensor(np.ones([512, 128, 1]), dtype=ms.float32)
def test_auto_parallel():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=16, global_rank=0)
net = Full(_w1, 3)
net.set_auto_parallel()

2
tests/ut/python/parallel/test_auto_parallel_for_loop_simplify.py
View File

@ -83,7 +83,7 @@ _w1 = Tensor(np.ones([512, 128]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
learning_rate = 0.1
momentum = 0.9
epoch_size = 2

2
tests/ut/python/parallel/test_auto_parallel_two_bn.py
View File

@ -68,7 +68,7 @@ def test_two_bn():
out = self.block2(out)
return out
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(device_num=8, global_rank=0)
context.set_auto_parallel_context(parallel_mode="auto_parallel")
net = NetWithLoss(Net())

4
tests/ut/python/parallel/test_broadcast_to.py
View File

@ -64,7 +64,7 @@ _x2 = Tensor(np.ones([64, 64]), dtype=ms.float32)
def compile_net(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()
@ -74,7 +74,7 @@ def compile_net(net):
def compile_net2(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_concat.py
View File

@ -80,7 +80,7 @@ w3 = Tensor(np.ones([64, 64, 32]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_embeddinglookup.py
View File

@ -95,7 +95,7 @@ def test_embeddinglookup_reducescatter_false_grad():
def test_embeddinglookup_reducescatter_true_grad():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
shape = [8, 8]
offset = 8
net = GradWrap(NetWithLoss(Net(shape, offset)))

4
tests/ut/python/parallel/test_eval.py
View File

@ -52,7 +52,7 @@ _b = Tensor(np.ones([64, 64]), dtype=ms.float32)
def test_train_and_eval():
context.set_context(save_graphs=True, mode=0)
context.set_context(save_graphs=False, mode=0)
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=16)
strategy1 = ((4, 4), (4, 4))
strategy2 = ((4, 4),)
@ -69,7 +69,7 @@ def test_train_and_eval():
context.reset_auto_parallel_context()
def test_train_and_eval_auto():
context.set_context(save_graphs=True, mode=0)
context.set_context(save_graphs=False, mode=0)
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=16)
strategy1 = ((4, 4), (4, 4))
strategy2 = ((4, 4),)

2
tests/ut/python/parallel/test_loss_and_o2_level.py
View File

@ -63,7 +63,7 @@ _w1 = Tensor(np.ones([512, 128]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
learning_rate = 0.1
momentum = 0.9
epoch_size = 2

2
tests/ut/python/parallel/test_loss_scale.py
View File

@ -194,7 +194,7 @@ def test_loss_scale():
def test_loss_scale2():
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
context.set_auto_parallel_context(parallel_mode=ParallelMode.SEMI_AUTO_PARALLEL, device_num=8)
predict = Tensor(np.ones([64, 64]), dtype=ms.float32)
label = Tensor(np.ones([64,]), dtype=ms.int32)

6
tests/ut/python/parallel/test_manual_embedding_lookup.py
View File

@ -66,7 +66,7 @@ _b = Tensor(np.ones([8, 8, 8]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
optimizer = LazyAdam(net.trainable_params(), learning_rate=0.1)
optimizer.sparse_opt.add_prim_attr("primitive_target", "CPU")
train_net = TrainOneStepCell(net, optimizer)
@ -113,7 +113,7 @@ def test_normal_split_with_offset():
def test_auto_parallel_error():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=2, global_rank=0)
net = Net()
with pytest.raises(RuntimeError):
@ -121,7 +121,7 @@ def test_auto_parallel_error():
def test_auto_parallel():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=2, global_rank=0)
net = Net(split_string="fake")
compile_net(net)

4
tests/ut/python/parallel/test_manual_gatherv2.py
View File

@ -60,7 +60,7 @@ _b = Tensor(np.ones([64, 8]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()
@ -106,7 +106,7 @@ def test_normal_split_with_offset():
def test_auto_parallel_error():
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(parallel_mode="auto_parallel", device_num=2, global_rank=0)
net = Net()
with pytest.raises(RuntimeError):

2
tests/ut/python/parallel/test_model_with_loss.py
View File

@ -63,7 +63,7 @@ _w1 = Tensor(np.ones([512, 128]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
learning_rate = 0.1
momentum = 0.9
epoch_size = 2

2
tests/ut/python/parallel/test_model_without_loss.py
View File

@ -103,7 +103,7 @@ w3 = Tensor(np.ones([64, 64, 32]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
learning_rate = 0.1
momentum = 0.9
epoch_size = 2

2
tests/ut/python/parallel/test_o2_level.py
View File

@ -90,7 +90,7 @@ _w2 = Tensor(np.ones([128, 64, 1]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
learning_rate = 0.1
momentum = 0.9
epoch_size = 2

8
tests/ut/python/parallel/test_pack.py
View File

@ -136,7 +136,7 @@ _x_c = Tensor(np.ones([8, 8, 8]), dtype=ms.float32)
def compile_net(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()
@ -146,7 +146,7 @@ def compile_net(net):
def compile_net1(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()
@ -156,7 +156,7 @@ def compile_net1(net):
def compile_net2(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()
@ -166,7 +166,7 @@ def compile_net2(net):
def compile_net_con(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_parameter_merge.py
View File

@ -79,7 +79,7 @@ def clean_all_ckpt_files(folder_path):
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
learning_rate = 0.1
momentum = 0.9
epoch_size = 2

2
tests/ut/python/parallel/test_range.py
View File

@ -74,7 +74,7 @@ _w1 = Tensor(np.ones([64, 8]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
learning_rate = 0.1
momentum = 0.9
epoch_size = 2

2
tests/ut/python/parallel/test_reluv2.py
View File

@ -40,7 +40,7 @@ _x = Tensor(np.ones([32, 16, 48, 64]), dtype=ms.float32)
def compile_net(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

4
tests/ut/python/parallel/test_repeated_calc.py
View File

@ -73,7 +73,7 @@ def test_tensoradd_reshape_matmul():
strategy2 = ((8, 1), (1, 8))
net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
x = Tensor(np.ones([32, 8, 16]), dtype=ms.float32)
y = Tensor(np.ones([32, 8, 16]), dtype=ms.float32)
@ -99,7 +99,7 @@ def test_two_matmul():
strategy2 = ((8, 1), (1, 1))
net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel")
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
x = Tensor(np.ones([128, 32]), dtype=ms.float32)
y = Tensor(np.ones([32, 64]), dtype=ms.float32)

2
tests/ut/python/parallel/test_reshape_optimized.py
View File

@ -40,7 +40,7 @@ _b = Tensor(np.ones([128, 64, 32]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_reshape_skip_redistribution.py
View File

@ -43,7 +43,7 @@ _x = Tensor(np.ones([64, 64]), dtype=ms.float32)
_b = Tensor(np.ones([128, 64, 32]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

4
tests/ut/python/parallel/test_split.py
View File

@ -75,7 +75,7 @@ _x1 = Tensor(np.ones([48, 64, 32]), dtype=ms.float32)
_w2 = Tensor(np.ones([48, 64, 32]), dtype=ms.float32)
def compile_net(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()
@ -85,7 +85,7 @@ def compile_net(net):
def compile_net1(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_stridedslice.py
View File

@ -67,7 +67,7 @@ _b = Tensor(np.ones([128, 64, 32]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_tile.py
View File

@ -60,7 +60,7 @@ _b = Tensor(np.ones([128, 64, 32]), dtype=ms.float32)
def compile_net(net):
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_train_and_eval.py
View File

@ -52,7 +52,7 @@ _b = Tensor(np.ones([64, 64]), dtype=ms.float32)
def test_train_and_eval():
context.set_context(save_graphs=True, mode=0)
context.set_context(save_graphs=False, mode=0)
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=16)
strategy1 = ((4, 4), (4, 4))
strategy2 = ((4, 4),)

4
tests/ut/python/parallel/test_two_matmul.py
View File

@ -141,7 +141,7 @@ def test_matmul_forward_reduce_scatter():
return out
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=8, global_rank=0)
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
strategy1 = ((2, 2), (2, 2))
strategy2 = ((4, 2), (4, 2))
net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))
@ -166,7 +166,7 @@ def test_matmul_forward_reduce_scatter_transpose():
return out
context.set_auto_parallel_context(parallel_mode="semi_auto_parallel", device_num=16, global_rank=0)
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
strategy1 = ((2, 4), (2, 4))
strategy2 = ((8, 2), (8, 2))
net = GradWrap(NetWithLoss(Net(strategy1, strategy2)))

2
tests/ut/python/parallel/test_uniform_candidate_sampler.py
View File

@ -73,7 +73,7 @@ _x = Tensor(np.ones([48, 16]), dtype=ms.int32)
def compile_net(net):
context.set_context(mode=context.GRAPH_MODE, save_graphs=True)
context.set_context(mode=context.GRAPH_MODE, save_graphs=False)
optimizer = Momentum(net.trainable_params(), learning_rate=0.1, momentum=0.9)
train_net = TrainOneStepCell(net, optimizer)
train_net.set_auto_parallel()

2
tests/ut/python/parallel/test_virtual_dataset_3_input.py
View File

@ -93,7 +93,7 @@ def test_virtualdataset_cell_3_inputs():
out = self.matmul2(out, b)
return out
context.set_context(save_graphs=True)
context.set_context(save_graphs=False)
context.set_auto_parallel_context(parallel_mode="auto_parallel")
context.set_auto_parallel_context(device_num=8, global_rank=0)
net = GradWrap(VirtualDatasetCellTriple(NetWithLoss(Net(None, None, None))))

4
tests/ut/python/runtest.sh
View File

@ -41,7 +41,7 @@ if [ $# -eq 1 ] && ([ "$1" == "stage1" ] || [ "$1" == "stage2" ] || [ "$1" ==
elif [ $1 == "stage2" ]; then
echo "run python parallel"
pytest -v $CURRPATH/parallel
pytest -s $CURRPATH/parallel/*.py
RET=$?
if [ ${RET} -ne 0 ]; then
@ -87,7 +87,7 @@ else
exit ${RET}
fi
pytest -v $CURRPATH/parallel
pytest -v $CURRPATH/parallel/*.py
RET=$?
if [ ${RET} -ne 0 ]; then
exit ${RET}