chore(halo): set spec-draft-p-min for coder model

Add a 0.74 confidence threshold so speculative drafting stops early once the draft model's predicted token probability drops below it, favoring shorter, higher-acceptance draft sequences.
2026-05-21 23:15:09 +02:00 · 2026-05-21 23:15:09 +02:00 · ccd8750899
commit ccd8750899
parent 3a070413e4
1 changed files with 1 additions and 0 deletions
--- a/systems/x86_64-linux/halo/models.ini
+++ b/systems/x86_64-linux/halo/models.ini
@ -18,6 +18,7 @@ c                   = 131072
 hf                  = unsloth/Qwen3.6-27B-MTP-GGUF:UD-Q8_K_XL
 spec-type           = ngram-simple,draft-mtp
 spec-draft-n-max    = 5
+spec-draft-p-min    = 0.74
 threads-batch       = 16
 temp                = 0.6
 top-p               = 0.95