Thanks very much for this share.
I noticed that the original paper said they do not use batch normalizaiton in the ResNet101 encoder, however, bn is still applied in this implementation. Will this be problematic? How will this change affects the representations extracted by the model?

Thank you for releasing the source code, but I wonder have you got the same result of its paper?

I've rewrite your code and struggling for the result for a couple of days.

Hi,
I wonder if there is any hint on the final classification accuracy on the cpc representations. (One linear classifier built on top of the well-extracted cpc-features). I ran the code but only got around 25% accuracy on the FASHION_MNIST data after 30000 iters training in the final evaluation, which is weird.

There might be something wrong with my experiment, is here any thing that I should pay more attention run the code? I just directly run the model, train the cpc feature extractor and then set "mode" to be "validation" and train the final classifier.

While applying ResNet are you passing labels?
If yes then for every patch the label has been included? As there are of 49 patches per image.
So I would like to know patch wise labeling(i.e 49 times same label is repeated) is there or image wise?

Thank you for your source code! But I have question in model.py[#37],I have not seen the @ on this condition!Could you give me some clues?