Dataloader does not function?

[luccachiang]

Hello, I am trying to run your code. However, I met some bugs. It seems that the main problem lies in get_kather2016_datasets in kather2016.py, namely the loader ia not able to load data images. The data is in t3po/data/kather2016. Did you meet this problem or would you please give some advice?

[agaldran]

Hey Lucca, did you solve this? I am sorry, I am moving from Spain to Australia this week and everything is a bit crazy. If this is still a problem, I will have a look at it next week, if you don't mind.

[luccachiang]

Hey, Agaldran. Hope everything is fine with you! With regard to this issue, I have not solved it. When I try to run train_baseline.py, after printing "==> Epoch 1/200", the model stops running. I guess the problem is in m1, m2, m3 = train_one_epoch(net, criterion, optimizer, train_loader) and specifically there is something wrong with the train_loader or maybe the hyperparameters are wrong. That's why I said that the problem lied in datasets. I do not if it is because I am using Windows 10 instead of Linux, and CPU instead of CUDA. When I debug, the last instruction PyCharm executes is n, err = decoder.decode(b"fpfp") in TiffImagePlugin.py which belongs to package PIL.

The last information it prints is the following:
Creating model: mobilenet
==> Epoch 1/200
0%| | 0/3750 [00:00<?, ?it/s]
return -1073740791 (0xC0000409)

I'd appreciate it if you can help me solve the trouble. It's too frustrating.

[luccachiang]

Oh, after I upgrade my Pillow package everything seems to work fine. Is this the correct running result?

Why does it print the index and tensor[0]? Can I turn off the output?