torch-model-compression icon indicating copy to clipboard operation
torch-model-compression copied to clipboard

Published 20 hours ago • verified publisher icon THU-MIG

examples中prune.py运行报错

Open chihoz opened this issue 2 years ago • 3 comments

运行到第六个epoch时报错如下： Traceback (most recent call last): File "D:\Pycharm\PyCharm 2021.2.3\plugins\python\helpers\pydev\pydevd.py", line 1483, in _exec pydev_imports.execfile(file, globals, locals) # execute the script File "D:\Pycharm\PyCharm 2021.2.3\plugins\python\helpers\pydev_pydev_imps_pydev_execfile.py", line 18, in execfile exec(compile(contents+"\n", file, 'exec'), glob, loc) File "E:/python project/torch-model-compression-main/torch-model-compression-main/examples/torchslim/pytorch-cifar/prune.py", line 94, in solver.run() File "D:\Anoconda\lib\site-packages\torchpruner-0.1.0-py3.8.egg\torchslim\slim_solver.py", line 354, in run File "D:\Anoconda\lib\site-packages\torchpruner-0.1.0-py3.8.egg\torchslim\slim_solver.py", line 224, in run_hook File "D:\Anoconda\lib\site-packages\torchpruner-0.1.0-py3.8.egg\torchslim\pruning\resrep.py", line 297, in after_iteration_hook File "D:\Anoconda\lib\site-packages\torchpruner-0.1.0-py3.8.egg\torchpruner\graph.py", line 622, in build_graph File "D:\Anoconda\lib\site-packages\torchpruner-0.1.0-py3.8.egg\torchpruner\operator\operator.py", line 186, in fill_shape """ RuntimeError: Fail to predict the shape on operator name: 'self.bn1.BatchNormalization', type: 'BatchNormalization' 是什么原因呢？

chihoz avatar Feb 15 '22 03:02 chihoz

我也遇到了这个问题，请问你解决了吗？

stivensss avatar Mar 09 '22 03:03 stivensss

我是在window上运行报Fail to predict the shape on operator name: 'self.bn1.BatchNormalization', type: 'BatchNormalization'这个错误，在linux上运行正常。估计是安装包环境问题

stivensss avatar Mar 09 '22 03:03 stivensss

我是在window上运行报Fail to predict the shape on operator name: 'self.bn1.BatchNormalization', type: 'BatchNormalization'这个错误，在linux上运行正常。估计是安装包环境问题

我在linux上同样也遇到了这个问题...

Juelianqvq avatar Apr 27 '22 04:04 Juelianqvq

我也遇到了这个问题，请问你解决了吗？

请问下您解决了吗

EricBlaine007 avatar Apr 18 '23 09:04 EricBlaine007

我是在window上运行报Fail to predict the shape on operator name: 'self.bn1.BatchNormalization', type: 'BatchNormalization'这个错误，在linux上运行正常。估计是安装包环境问题

我在linux上同样也遇到了这个问题...

请问下您解决了吗这个问题，我在Linux系统上也遇到了相同的问题。

EricBlaine007 avatar Apr 18 '23 09:04 EricBlaine007

你们 pytorch 版本太新了，我在 torch==1.9.0+cu111 上测试正常，只是 torch==1.13.1+cu117 就不行了

测试发现是 pytorch 改了 BatchNormalization 的功能，某些时候会增加输出个数，从 1 个增加到 5 个。

Commit 3410752 强行忽略了后几个，应该能跑了。

gdh1995 avatar Apr 19 '23 16:04 gdh1995

你们 pytorch 版本太新了，我在 torch==1.9.0+cu111 上测试正常，只是 torch==1.13.1+cu117 就不行了

测试发现是 pytorch 改了 BatchNormalization 的功能，某些时候会增加输出个数，从 1 个增加到 5 个。

Commit 3410752 强行忽略了后几个，应该能跑了。

好的，谢谢前辈，已经解决了，看到您都是凌晨解答，注意休息！

EricBlaine007 avatar Apr 20 '23 03:04 EricBlaine007