ling trial: layer ngram-mod under draft-mtp; add ngram args to serve script

This commit is contained in:
Frank committed 2026-08-14 01:40:46 -07:00
1 parent c275cbaf57
commit 3204b4b29e
2 files changed
+11 -1

No files matched your search

+5 -1
View File
@@ -21,8 +21,12 @@ PORT=7770
CTX_SIZE=262144
PARALLEL=2
# MTP is bundled in the Ling GGUF itself (block 42) - no separate draft file.
SPEC_TYPE=draft-mtp
# Layering ngram-mod under MTP (draft-mtp,ngram-mod) re-drafts via ngram.
SPEC_TYPE=draft-mtp,ngram-mod
SPEC_DRAFT_N_MAX=3
SPEC_NGRAM_MOD_N_MIN=48
SPEC_NGRAM_MOD_N_MAX=64
SPEC_NGRAM_MOD_N_MATCH=24
# Vulkan exposes the larger heap (UMA + GTT) vs ROCm. Use Vulkan0.
DEVICE=Vulkan0
N_GPU_LAYERS=all