Jun Chen comments

Results 17 comments of


                                            Jun Chen

This is very cool, but push to even higher gpu usage?

Another observation is, using 2 threads in the pool seems better than more

This is very cool, but push to even higher gpu usage?

Yes, this is slow?

This is very cool, but push to even higher gpu usage?

I see, thanks! Is the cpu part mostly on faster-whisper, or CTranslate2?

Request to support FlashAttention in cuda attention.cc

Some other repo claimed flash attention will be helpful to make transcribe much faster: https://github.com/Vaibhavs10/insanely-fast-whisper To my read, 2x?

Request to support FlashAttention in cuda attention.cc

thanks, looking forward to test it with faster-whisper!

Fix occasional IndexError on empty segments

hi @jongwook may I ask when this will be merged? I've recently encountered this issue, so will be good to merge it. Thanks!

Incorporating flash-attention2 [SOLVED] and subsequent testing [ONGOING]

there is some discussion here on ctranslate2: https://github.com/OpenNMT/CTranslate2/issues/1300