mlverse/torch

nn_cross_entropy_loss (ignore_index = 0)

AGPatriota opened this issue · 0 comments

Dear fellows

I was trying to ignore the first index in the cross_entropy_loss, but the index is following the Python pattern.

It ignores the second index:
nn_cross_entropy_loss(ignore_index=1)

It ignores the first index:
nn_cross_entropy_loss(ignore_index=0)

That might happen for other functions, see my opened issue.

Best