diff --git a/config.json b/config.json index 426b236..9d2513a 100644 --- a/config.json +++ b/config.json @@ -71,9 +71,9 @@ "ctx": "65536", "trust_remote": true, "valid_tp": [1, 2], - "max_num_seqs": "4", + "max_num_seqs": "2", "max_tokens": "8192", - "gpu_util": "0.94", + "gpu_util": "0.90", "enforce_eager": false, "tool_call_parser": "qwen3_xml", "enable_auto_tool_choice": true,