Harald Hoyer 5ee2f65337 chore(halo): tune llama models.ini and drop 35B-A3B model ... Serve only Qwen3.6-27B; remove the unused 35B-A3B preset. Tuning: - Move model-specific keys (spec-type, sampling temp/top-p/top-k/min-p) out of the [*] defaults into [Qwen3.6-27B] so they no longer leak onto other models; draft-mtp in particular only works on MTP-weighted models. - Drop the duplicate parallel key from [*]. - Bump ubatch-size 256 -> 512 for faster iGPU prefill on Strix Halo. - Add threads-batch = 16 to use all cores for prefill while keeping generation at threads = 8 under full GPU offload.		2026-05-20 14:23:42 +02:00
..
aarch64-darwin	feat(rialo): add pi	2026-05-19 14:27:50 +02:00
aarch64-linux	nix fmt	2026-02-24 13:25:42 +01:00
x86_64-darwin/mpro	nix fmt	2024-11-19 10:31:29 +01:00
x86_64-linux	chore(halo): tune llama models.ini and drop 35B-A3B model	2026-05-20 14:23:42 +02:00
nixbuild.nix	chore: nix fmt	2026-05-03 14:57:49 +02:00