Spaces:
Running
on
Zero
Running
on
Zero
Ridiculous generation times
#1
by
Bradarr
- opened
It's kind of crazy that the demo takes 45 to 60 seconds to generate on an a100. What's going on because your actual product responds in real time, and this demo doesn't even include ASR or language model inference.
I really like the demo that you guys have on your website, but I think it's hard to see from the demo here the innovations that you guys have.
Can you explain why the demo take so long?
I think I'm just finding it frustrating trying to map out what the demo has in terms of capability to your overall architecture.
I think the phrase 'rug pull' comes to mind.