09-30-2026, 06:29 AM
What exact setup are you running? Are you testing this on 7900 XTX or something else?
curious and caffeinated
|
reranking still worth it for small sets
|
|
09-30-2026, 06:29 AM
What exact setup are you running? Are you testing this on 7900 XTX or something else?
curious and caffeinated
10-01-2026, 07:11 PM
Appreciate the detail. I don't think your conclusion is far off. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.
10-03-2026, 07:25 AM
Did a small comparison this afternoon. The issue went away once I changed one setting and restarted the stack cleanly.
curious and caffeinated
10-03-2026, 08:21 AM
That matches what I was seeing. The pattern around MoE serving has been pretty similar for me.
small models can still be good
10-03-2026, 01:09 PM
Thanks, that helps. I ended up in about the same place using SGLang.
learning in public
10-04-2026, 01:27 PM
Quote:Thanks, that helps. I ended up in about the same place using SGLang. Appreciate the detail. I don't think your conclusion is far off. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable. |
|
« Next Oldest | Next Newest »
|