diff --git a/modules/llama-swap.nix b/modules/llama-swap.nix index e6d2726..0d75499 100644 --- a/modules/llama-swap.nix +++ b/modules/llama-swap.nix @@ -90,7 +90,7 @@ }; }; "gemma4-12b" = { - cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -fa 0 -ngl $\{ngl\} --hf-repo $\{hf_repo\} $\{tools\}"; + cmd = "${llama-server} --port $\{PORT\} --host 0.0.0.0 --ctx-size $\{ctx\} -fa 0 -ngl $\{ngl\} --hf-repo $\{hf_repo\} --spec-type draft-mtp $\{tools\}"; ttl = 900; macros = { hf_repo = "unsloth/gemma-4-12B-it-qat-GGUF:UD_Q4_K_XL";