+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Optimization & Performance (https://www.datasculpting.ai/inference/forumdisplay.php?fid=10)
+--- Forum: Benchmarks (https://www.datasculpting.ai/inference/forumdisplay.php?fid=43)
+--- Thread: domain tune without wrecking base behavior with PyTorch (/showthread.php?tid=484)
domain tune without wrecking base behavior with PyTorch - benchplease - 09-25-2026
Spent some time on this today and wanted to compare notes.
Two paths both seem viable, but they break down differently once continuous batching becomes important.
Trying to avoid rebuilding the whole thing twice.
Re: domain tune without wrecking base behavior with PyTorch - fusedhall56 - 09-26-2026
Tried to reproduce it just now. My result was mixed: first run looked bad, second run looked normal. I expected the model choice to matter more than it did.
Re: domain tune without wrecking base behavior with PyTorch - kernelghost - 09-26-2026
Need a little more detail here. What model, what runtime, and roughly what context size?
Re: domain tune without wrecking base behavior with PyTorch - deltastation - 09-26-2026
I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with Phi and compare from there.
Re: domain tune without wrecking base behavior with PyTorch - cacheline - 09-26-2026
that fixed it for me too.
Re: domain tune without wrecking base behavior with PyTorch - mlxgarden - 09-29-2026
Good point. The pattern around quant drift has been pretty similar for me.
Re: domain tune without wrecking base behavior with PyTorch - kernelghost - 09-29-2026
Good point. I ended up in about the same place using llama.cpp. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.