self.nb_batch_samples_world = input.size(0)
self.nb_batch_samples_quizzes = 0
+ # Shuffle
+ input = input[torch.randperm(input.size(0))]
+
if desc is None:
desc = f"epoch-{split}"
for batch in tqdm.tqdm(
device=self.device,
)
+ # Should not be necessary though, the autoregression is done
+ # in eval mode
+ sum_logits = sum_logits.detach()
+
average_logits = sum_logits / quizzes.numel()
# It's a bit brutal to do it twice, we should probably have a