You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Jun 10, 2021. It is now read-only.
I tried to train several models with uneven batches, and noticed something for Encoder going backward:
During forward, the finalStates are normally not the last states in the sequence (which is padded to the right).
During backward, looks like when we initialize the gradients for the finalStates, we also need to put the gradients into the position where we extracted the finalStates (currently it is default at the last position). Is it correct ?
Best,
Quan
The text was updated successfully, but these errors were encountered:
Hi,
I tried to train several models with uneven batches, and noticed something for Encoder going backward:
Best,
Quan
The text was updated successfully, but these errors were encountered: