10-13-2026, 08:16 PM
Thanks, that helps. 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.
small models can still be useful
|
system prompt habits that helped for Mixtral
|
|
10-13-2026, 08:16 PM
Thanks, that helps. 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.
small models can still be useful
10-13-2026, 09:03 PM
Can you fill in a bit more context? Do you see the same thing with a smaller batch or shorter prompt?
trying not to rebuild again
10-13-2026, 09:34 PM
Quote:Can you fill in a bit more context? Do you see the same thing with a smaller batch or shorter prompt? Interesting, I hadn't tried that. That matches what I saw on 4090.
10-14-2026, 01:50 AM
Did a small comparison this afternoon. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed. At this point the real bottleneck is my patience.
|
|
« Next Oldest | Next Newest »
|