From 7ab9fe6e59490d1908ed8f627d23d14eb7b4186b Mon Sep 17 00:00:00 2001 From: SZCJW <792430652@qq.com> Date: Thu, 2 Apr 2026 23:10:43 +0800 Subject: [PATCH] x --- config.json | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/config.json b/config.json index cd3f9cb..4ca2f46 100644 --- a/config.json +++ b/config.json @@ -20,7 +20,7 @@ "revision": "", "models": { "default": "Qwen3.5-35B-A3B-GPTQ-Int4", - "selected": "Qwen3-Next-80B-A3B-Instruct-AWQ-4bit", + "selected": "Qwen3.5-35B-A3B-FP8", "profiles": { "GLM-4.7-Flash-AWQ": { "local_path": "GLM-4.7-Flash-AWQ", @@ -73,13 +73,13 @@ }, "Qwen3.5-35B-A3B-FP8": { "local_path": "Qwen3.5-35B-A3B-FP8", - "dtype": "bfloat16", + "dtype": "float16", "ctx": "65536", "trust_remote": true, "valid_tp": [1, 2], "max_num_seqs": "8", - "max_tokens": "32768", - "gpu_util": "0.98", + "max_tokens": "4096", + "gpu_util": "0.92", "tool_call_parser": "qwen3_coder", "enable_auto_tool_choice": true, "served_model_name": "Qwen3.5-35B-A3B-FP8",