diff --git a/Services/llamacpp/llamacpp.container b/Services/llamacpp/llamacpp.container index 7502b8a..32846e7 100644 --- a/Services/llamacpp/llamacpp.container +++ b/Services/llamacpp/llamacpp.container @@ -21,6 +21,10 @@ Environment=LLAMA_ARG_NO_MMAP=true Environment=LLAMA_ARG_CTX_SIZE=131072 Environment=LLAMA_ARG_HF_REPO=unsloth/Qwen3-Coder-30B-A3B-Instruct-GGUF:Q2_K +# MTP +Environment=LLAMA_ARG_SPEC_TYPE=draft-mtp +Environment=LLAMA_ARG_SPEC_DRAFT_N_MAX=6 + # HF Environment=HF_HOME=/root/.cache/huggingface Environment=HF_TOKEN=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