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
The current memory footprint is quite high - the batch size maxes out at 32.
We could significantly speed things up if we could reduce the memory footprint.
One possibility is to utilize more sparse matrices in the internal computations -- particularly for the masking and alignment matrices.
See https://stackoverflow.com/a/56887077/1167475
The text was updated successfully, but these errors were encountered:
The current memory footprint is quite high - the batch size maxes out at 32.
We could significantly speed things up if we could reduce the memory footprint.
One possibility is to utilize more sparse matrices in the internal computations -- particularly for the masking and alignment matrices.
See https://stackoverflow.com/a/56887077/1167475
The text was updated successfully, but these errors were encountered: