Question About the ClipCocoDataset in train.py

[ycchanau]

I have a question about the above code. I observe that in the pad_tokens function, it will save the padded token. However, if we call the pad_tokens function with the same token again, the padded token saved where the padded part is set to 0, will be used to generate the mask with ge(0), as a result, the mask will always be one for all positions. Do I miss something?

[bofang98]

I found the same problem... The mask is correct for the 1st epoch only. And it will be all 1's for all positions after the 1st epoch.