Hey should 'correct' only be if the final prediction was correct? Same for accuracy?

[colinator]

Contrastive-Predictive-Coding-PyTorch/src/model/model.py

Line 100 in a9dab4e

correct1 = torch.sum(torch.eq(torch.argmax(self.softmax(total), dim=0), torch.arange(0, batch))) # correct is a tensor

Meaning, should that correct variable be += ?

[NeteeraAF]

I noticed that too.

I think you're right, and after the loop over the timesteps, the sum should be divided by batch_size*num_of_timesteps.