Pytorch-lightning: [Bug] Progress bar displays wrong total iterations for train

[versatran01]

1018 🐛 Bug

for example, my training dataset has 448 samples, and I use a batch size of 4, then the loader should have length 448 / 4 = 112, but the progress bar is displaying N/171, which I have no idea how it computes that. The training then proceeds normally until it hits 112/171 and enters validation.

[liebkne]

Have you figure out how the total iteration is computed?

[versatran01]

I think it is the sum of train and val iteration, which is confusing.

[ThierryDeruyttere]

Yes it is very confusing. I think this should be fixed though.