CE-Net
CE-Net copied to clipboard
Guzaiwang
request
Can you configure the environment for the next code? For example, torch version? Thank you
Hi Guzai Wang, Thanks for opening your code. I also need to know your environment(Cuda version, torch version, python version etc), because I got the following error when running your code under torch0.4. I am not sure whether it is caused by the environment or not.
warnings.warn("nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.") /home/wqq/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/functional.py:1961: UserWarning: Default upsampling behavior when mode=bilinear is changed to align_corners=False since 0.4.0. Please specify align_corners=True if the old behavior is desired. See the documentation of nn.Upsample for details. "See the documentation of nn.Upsample for details.".format(mode)) /home/wqq/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/nn/functional.py:1006: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead. warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.") THCudaCheck FAIL file=/opt/conda/conda-bld/pytorch_1533672544752/work/aten/src/THC/generic/THCTensorMathPairwise.cu line=21 error=59 : device-side assert triggered Traceback (most recent call last):
File "
File "/home/wqq/anaconda3/envs/pytorch/lib/python3.6/site-packages/spyder_kernels/customize/spydercustomize.py", line 786, in runfile execfile(filename, namespace)
File "/home/wqq/anaconda3/envs/pytorch/lib/python3.6/site-packages/spyder_kernels/customize/spydercustomize.py", line 110, in execfile exec(compile(f.read(), filename, 'exec'), namespace)
File "/home/wqq/qingqwan/MediSeg/CE-Net-master/main.py", line 103, in
File "/home/wqq/qingqwan/MediSeg/CE-Net-master/main.py", line 60, in CE_Net_Train train_loss, pred = solver.optimize()
File "/home/wqq/qingqwan/MediSeg/CE-Net-master/framework.py", line 66, in optimize self.optimizer.step()
File "/home/wqq/anaconda3/envs/pytorch/lib/python3.6/site-packages/torch/optim/adam.py", line 100, in step denom = exp_avg_sq.sqrt().add_(group['eps'])
RuntimeError: cuda runtime error (59) : device-side assert triggered at /opt/conda/conda-bld/pytorch_1533672544752/work/aten/src/THC/generic/THCTensorMathPairwise.cu:21
Figure it out! Torch0.4 works. The problem is caused by the loss function. It can be solved by changing: a = self.ce_loss(y_pred, y_ce_true) to: b = self.multi_class_dice_loss(y_prediction, y_true)
I succeeded in running according to your method, but it ended after running about thirty epochs. I want to ask you how the result of running is. Thank you very much.
I succeeded in running according to your method, but it ended after running about thirty epochs. I want to ask you how the result of running is. Thank you very much.
Sorry, in the original implementation, the multi-class dice loss and binary cross entropy loss were separate. When we create this repository, we integrate both of them into a loss function. And the CUDA error (or warning) occurs. Now, we have modified the loss function. It works.
I succeeded in running according to your method, but it ended after running about thirty epochs. I want to ask you how the result of running is. Thank you very much.
Sorry, in the original implementation, the multi-class dice loss and binary cross entropy loss were separate. When we create this repository, we integrate both of them into a loss function. And the CUDA error (or warning) occurs. Now, we have modified the loss function. It works.
Many thanks. It works now.
Thank you very much for your reply and code. Could you upload the test code? Thank you very much.
Hi Guzai Wang, Thanks for opening your code. i have a question as you have mention that "start the visdom file before running main.py file. i didn't understand what is "visdom" can you please tell me about visdom what is this and how to start it? i shall be very thankful to you.
Hi Guzai Wang, Thanks for opening your code. i have a question as you have mention that "start the visdom file before running main.py file. i didn't understand what is "visdom" can you please tell me about visdom what is this and how to start it? i shall be very thankful to you.
visdom is a tool that can plot errors vs epoches that helps you visually see the training phase, you actually do not need to use the visdom part.
Thanks you so much for your kind response
发自我的iPhone
------------------ Original ------------------ From: Howie <[email protected]> Date: Fri,May 29,2020 9:56 AM To: Guzaiwang/CE-Net <[email protected]> Cc: Altaf-Hussain-AI <[email protected]>, Comment <[email protected]> Subject: Re: [Guzaiwang/CE-Net] request (#3)
Hi Guzai Wang, Thanks for opening your code. i have a question as you have mention that "start the visdom file before running main.py file. i didn't understand what is "visdom" can you please tell me about visdom what is this and how to start it? i shall be very thankful to you.
visdom is a tool that can plot errors vs epoches that helps you visually see the training phase, you actually do not need to use the visdom part.
— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.
