|
first LoRA that actually helped
|
08-23-2026, 04:34 PM
I think the bottleneck is somewhere else. I've seen people blame HBM constraints when the real issue was much more mundane. Worth testing the simpler explanation before adopting the more interesting one.
I'm happy to test one more variant if you want a second data point.
throughput matters, but tail latency matters too
|
|
« Next Oldest | Next Newest »
|
| Messages In This Thread |
|
first LoRA that actually helped - by cacheline - 08-23-2026, 05:51 AM
Re: first LoRA that actually helped - by benchplease - 08-23-2026, 07:25 AM
Re: first LoRA that actually helped - by mangohouse - 08-23-2026, 03:45 PM
Re: first LoRA that actually helped - by northgrid33 - 08-25-2026, 01:50 PM
Re: first LoRA that actually helped - by batchsize - 08-28-2026, 12:17 AM
Re: first LoRA that actually helped - by batchsize - 08-23-2026, 04:34 PM
Re: first LoRA that actually helped - by batchpeak - 08-27-2026, 11:28 AM
Re: first LoRA that actually helped - by griffinloom66 - 08-24-2026, 01:02 AM
Re: first LoRA that actually helped - by routecraft - 08-25-2026, 10:23 PM
Re: first LoRA that actually helped - by batchsize - 08-25-2026, 11:26 PM
Re: first LoRA that actually helped - by macroforge - 08-28-2026, 01:12 AM
|
