10-09-2026, 07:15 PM
Good point. I ended up in about the same place using vLLM. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
measure twice, benchmark three times
|
first LoRA that actually helped on 7900 XTX
|
|
« Next Oldest | Next Newest »
|
| Messages In This Thread |
|
first LoRA that actually helped on 7900 XTX - by batchsize - 10-07-2026, 05:44 PM
Re: first LoRA that actually helped on 7900 XTX - by papermixer4 - 10-07-2026, 06:20 PM
Re: first LoRA that actually helped on 7900 XTX - by nanobrook - 10-07-2026, 06:33 PM
Re: first LoRA that actually helped on 7900 XTX - by batchpeak - 10-09-2026, 06:55 AM
Re: first LoRA that actually helped on 7900 XTX - by cacheline - 10-09-2026, 07:15 PM
Re: first LoRA that actually helped on 7900 XTX - by quiethouse - 10-10-2026, 03:49 AM
Re: first LoRA that actually helped on 7900 XTX - by thetashed24 - 10-13-2026, 04:30 AM
Re: first LoRA that actually helped on 7900 XTX - by cacheline - 10-13-2026, 04:58 AM
Re: first LoRA that actually helped on 7900 XTX - by cachelab - 10-16-2026, 05:18 AM
|
Users browsing this thread: 1 Guest(s)