word timestamp

#6
by leoh - opened

Hi team !

Super happy for this new version of distil-large. I am wondering if you did also the distillation for the specific heads of word timestamps ?

Thanks,

We've only included the segment-level timestamps in distillation. For the word-level timestamps, we use attention heads from the last half of the decoder layers, but no additional training has been done.

Your need to confirm your account before you can post a new comment.

Sign up or log in to comment