ci(bfcl): add --trust-remote-code to GLM-5.2 vllm_extra

key4ng · key4ng · commit 78c7a1228056 · 2026-06-23T17:25:55.000-07:00
GLM-5.2 needs --trust-remote-code to load its custom architecture, matching
the other large MoE legs (deepseek-v4, minimax-m2.7, kimi-k2.6).

Signed-off-by: key4ng &lt;rukeyang@gmail.com&gt;
diff --git a/.github/workflows/nightly-bfcl.yml b/.github/workflows/nightly-bfcl.yml
@@ -184,7 +184,7 @@ jobs:
                "vllm_tool": "glm47", "vllm_reason": "glm45",
                "smg_tool": "glm47_moe", "smg_reason": "glm45",
                # FP8 load + DeepGEMM warmup on the largest leg: generous startup ceiling.
-               "vllm_extra": "--kv-cache-dtype fp8_e4m3", "gpu_mem": "0.90", "startup_timeout": "3000",
+               "vllm_extra": "--trust-remote-code --kv-cache-dtype fp8_e4m3", "gpu_mem": "0.90", "startup_timeout": "3000",
                "model_cache": "/raid/models", "arm_mode": "sequential", "max_model_len": "32768"},
           ]
           only = os.environ.get("ONLY", "")

Original file line number	Diff line number	Diff line change
`@@ -184,7 +184,7 @@ jobs:`
`184`	`184`	`"vllm_tool": "glm47", "vllm_reason": "glm45",`
`185`	`185`	`"smg_tool": "glm47_moe", "smg_reason": "glm45",`
`186`	`186`	`# FP8 load + DeepGEMM warmup on the largest leg: generous startup ceiling.`
`187`		`- "vllm_extra": "--kv-cache-dtype fp8_e4m3", "gpu_mem": "0.90", "startup_timeout": "3000",`
	`187`	`+ "vllm_extra": "--trust-remote-code --kv-cache-dtype fp8_e4m3", "gpu_mem": "0.90", "startup_timeout": "3000",`
`188`	`188`	`"model_cache": "/raid/models", "arm_mode": "sequential", "max_model_len": "32768"},`
`189`	`189`	`]`
`190`	`190`	`only = os.environ.get("ONLY", "")`