@avinash_m
Thanks for pointing this out, I think this is not a log issue, What you passed in is a tensor with wrong type and wrong shape.
You program hit log_softmax, and it doesn’t accept Long, so it return a type error. If you fixed the type error, you will get shape mismatch error then. In your case, tensor type and shape are both wrong, type error returned first. (And most of the time, type check will be the first check and if wrong, you will get type error first).