【论文复现】动态图模式下训练结束后计算flops错误 #38253
Comments
|
您好，我们已经收到了您的问题，会安排技术人员尽快解答您的问题，请耐心等待。请您再次检查是否提供了清晰的问题描述、复现代码、环境&版本、报错信息等。同时，您也可以通过查看官网API文档、常见问题、历史Issue、AI社区来寻求解答。祝您生活愉快～ Hi! We've received your issue and please be patient to get responded. We will arrange technicians to answer your questions as soon as possible. Please make sure that you have posted enough message to demo your request. You may also check out the API，FAQ，Github Issue and AI community to get the answer.Have a nice day! |
|
您是已经将问题解决了吗？建议到PaddleSeg的repo下提出修改意见 |
|
没有解决，并不确定这样修改会造成什么其他的影响。 |
|
您好，计算flops默认model的输出是tensor，建议您修改一下您模型的输出类型 |
|
@ghostxsl 您好，这个问题是因为maxpool2D 在指定return_mask=True时返回output和indice_mask，而不是单一tensor导致的这个问题。 |
|
您好，现在应该是只支持这些layer：https://github.com/PaddlePaddle/Paddle/blob/develop/python/paddle/hapi/dynamic_flops.py#L187-L207 |
|
您可以确认一下看看是否 model.children 的所有输出都是tensor呢？ |
|
好的
|
没有 |
您好，检查过了，都是tensor |
|
简单的复现了一下出问题的地方
|
|
您好，刚经过内部沟通，该问题为已知问题，目前正在修复中，后续进展我会在这里同步给您～ |
好的 |
1）PaddlePaddle版本：2.2.1
PR地址：https://github.com/PaddlePaddle/PaddleSeg/pull/1623
训练命令：
python PaddleSeg/train.py --config PaddleSeg/configs/enet/enet_cityscapes_1024x512_adam_0.002_80k.ymlW1218 13:14:42.893141 301 device_context.cc:465] device: 0, cuDNN Version: 7.6.
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/nn/layer/norm.py:653: UserWarning: When training, we now always track global mean and variance.
"When training, we now always track global mean and variance.")
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/math_op_patch.py:253: UserWarning: The dtype of left and right variables are not the same, left dtype is paddle.float32, but right dtype is paddle.int64, the right dtype will convert to paddle.float32
format(lhs_dtype, rhs_dtype, lhs_dtype))
2021-12-18 13:14:58 [INFO] [TRAIN] epoch: 1, iter: 10/10, loss: 2.2953, lr: 0.001259, batch_cost: 1.1241, reader_cost: 0.85701, ips: 7.1169 samples/sec | ETA 00:00:00
<class 'paddle.nn.layer.conv.Conv2D'>'s flops has been counted
Cannot find suitable count function for <class 'paddle.nn.layer.pooling.MaxPool2D'>. Treat it as zero FLOPs.
<class 'paddle.nn.layer.norm.BatchNorm2D'>'s flops has been counted
Cannot find suitable count function for <class 'paddle.nn.layer.activation.PReLU'>. Treat it as zero FLOPs.
Cannot find suitable count function for <class 'paddle.nn.layer.common.Dropout2D'>. Treat it as zero FLOPs.
<class 'paddle.nn.layer.activation.ReLU'>'s flops has been counted
<class 'paddle.nn.layer.conv.Conv2DTranspose'>'s flops has been counted
/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/tensor/creation.py:130: DeprecationWarning:
np.objectis a deprecated alias for the builtinobject. To silence this warning, useobjectby itself. Doing this will not modify any behavior and is safe.Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations
if data.dtype == np.object:
Traceback (most recent call last):
File "PaddleSeg/train.py", line 199, in
main(args)
File "PaddleSeg/train.py", line 194, in main
to_static_training=cfg.to_static_training)
File "/home/aistudio/PaddleSeg/paddleseg/core/train.py", line 311, in train
custom_ops={paddle.nn.SyncBatchNorm: op_flops_funs.count_syncbn})
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/hapi/dynamic_flops.py", line 113, in flops
print_detail=print_detail)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/hapi/dynamic_flops.py", line 254, in dynamic_flops
model(inputs)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in call
outputs = self.forward(*inputs, **kwargs)
File "/home/aistudio/PaddleSeg/paddleseg/models/enet.py", line 199, in forward
x, max_indices1_0 = self.downsample1_0(x)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 914, in call
outputs = self.forward(*inputs, **kwargs)
File "/home/aistudio/PaddleSeg/paddleseg/models/enet.py", line 504, in forward
main, max_indices = self.main_max1(x)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/fluid/dygraph/layers.py", line 917, in call
hook_result = forward_post_hook(self, inputs, outputs)
File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/hapi/dynamic_flops.py", line 184, in count_io_info
m.register_buffer('output_shape', paddle.to_tensor(y.shape))
AttributeError: 'tuple' object has no attribute 'shape'
PS:
1:self.main_max1 = paddle.nn.MaxPool2D(2, stride=2, return_mask=return_indices)
2:将File "/opt/conda/envs/python35-paddle120-env/lib/python3.7/site-packages/paddle/hapi/dynamic_flops.py", line 184, in count_io_info：
m.register_buffer('output_shape', paddle.to_tensor(y.shape))
改为
m.register_buffer('output_shape', paddle.to_tensor(y[0].shape))后，错误消除。
The text was updated successfully, but these errors were encountered: