4 ms·
On a 1080Ti (so a 6 year old GPU), the large model runs in 1x time (so transcribing 10 minutes takes 10 minutes) and I've successfully transcribed even 1h+ file
by rolisz 4y ago
On a 1080Ti (so a 6 year old GPU), the large model runs in 1x time (so transcribing 10 minutes takes 10 minutes) and I've successfully transcribed even 1h+ files.
- kkielhofner 4y agoFWIF an optimized implementation I've been working on comes in at roughly 70x realtime (large-v2, beam size 5) on an RTX 3090.
- rolisz 4y agoNice! Are you going to release it publicly?
- kkielhofner 4y agoGreat question! We're still very early stage and stealth so it's not quite clear to us where our lines are with regards to special sauce/significant competitive advantage. As the CTO (and lead dev) I'd lean towards open sourcing it (because it's awesome and we're standing on the shoulders of open source giants already) but it may become clear it's too differentiating to open source. As I said it's just too early to tell. What I can say is if we open source it HN will be the first to hear about it!