No module named UW

[jeffin07]

Hi, First of all great effort. When i tried to run the test_UWModels/test_UIEC2Net.py it shows the error No module named UW.I have already renamed the repo to UW. Can you please tell me what might be the cause?

python: 3.6 PyTorch :1.8

[BIGWangYuDong]

Hi, Firstly, please refer getting_start to install environment. Second, you can remove UW. then it can run

Best regards, Dong

[jeffin07]

@BIGWangYuDong Thank you for your reply. I did follow the instructions in getting_start to install the environment. The installation was successful but it still shows no module named UW. Did you ever encounter this problem? .Also one more thing I get this when I run test_xxx.py in test_UWModels

[BIGWangYuDong]

You can remove UW. then it can work. I have tried this.

[jeffin07]

@BIGWangYuDong Can you please explain what do you mean by removing UW?

[BIGWangYuDong]

For example from UW.core.Models.builder import NETWORK, build_backbone You can remove UW and try to run the test file.

[jeffin07]

@BIGWangYuDong Thank you. It's working perfectly. Before I close this issue can I ask you why this approach we can completely remove UW from all the files why makes a confusion? is there any specific reason. Also, I noticed a lot of operation in Cuda can we make that compatible for CPU also for testing scripts. Is it okay if I raise a PR for rectifying these issues?

[BIGWangYuDong]

I have no idea why this problem happened, I think if remove .cuda() (for example model.cuda()) then it can run in CPU. Its easy to make the code run in CPU

[jeffin07]

@BIGWangYuDong yeah, it's that simple, I just asked because it was hardcoded.If it was generalized it will be more helpful.Thank you for the response and great job :+1:

[BIGWangYuDong]

@jeffin07 I think maybe you can use gpus=0 to run the code on cpu, I did not try this, but I think it can work. I refer mmdetection to write this repo.

[BIGWangYuDong]

And if you want to debug this code, you can set works_per_gpu=0