nemoramo/acoustic_model

ACE loss

Opened this issue · 0 comments

viig99 commented

Hi I am using ACE Loss in my asr model too
https://github.com/viig99/SABER/blob/master/utils/aggloss.py#L24 this worked for ocr, but doesn't seem to be working for speech, i can see in your codebase that you have commented it out and instead using https://github.com/ramosmy/acoustic_model/blob/master/ace.py#L45, can you please let me know why this is the case ?