x

2026-04-11 01:03:03 +08:00
parent db43278cb7
commit e46558bc28
2 changed files with 103 additions and 2 deletions
@@ -27,7 +27,7 @@
        "ctx": "32768",
        "trust_remote": true,
        "valid_tp": [1, 2],
-        "max_num_seqs": "8",
+        "max_num_seqs": "16",
        "max_tokens": "32768",
        "gpu_util": "0.98",
        "tool_call_parser": "qwen3_coder",
@@ -102,7 +102,9 @@
        "valid_tp": [2],
        "max_num_seqs": "12",
        "max_tokens": "65536",
-        "gpu_util": "0.94",
+        "max_num_batched_tokens": 32768,
+        "kv_cache_dtype": "fp8",
+        "gpu_util": "0.92",
        "enforce_eager": false,
        "enable_prefix_caching": true,
        "tool_call_parser": "qwen3_coder",