瀏覽代碼

aiden-4x-labdrafter: point at the dequantized -bf16 repo (aiden registry arch DFlash2DraftModel; MXFP8 scales unsupported)

Alexander Korolev 1 月之前
父節點
當前提交
876f54ad27
共有 1 個文件被更改,包括 1 次插入 和 1 次删除
  1. 1 1
      recipes/glm53-flash-aiden-4x-labdrafter/recipe.toml

+ 1 - 1
recipes/glm53-flash-aiden-4x-labdrafter/recipe.toml

@@ -58,7 +58,7 @@ args = [
     "--gpu-memory-utilization", "0.85",
     "--kv-cache-memory-bytes", "17179869184",
     "--speculative-config", """
-'{"method":"dflash","model":"local-inference-lab/GLM-5.3-Flash-DFlash2","num_speculative_tokens":7}'""",
+'{"method":"dflash","model":"local-inference-lab/GLM-5.3-Flash-DFlash2-bf16","num_speculative_tokens":7}'""",
     "--language-model-only",
     # image requests need the mm template shipped/in-wallpaper on aiden's rigs:
     # "--chat-template", "<path-to>/chat_template_mm.jinja",