add try catch for quant config hf download (#10340)

python/sglang/srt/configs/model_config.py

@@ -420,11 +420,20 @@ class ModelConfig:
             is_local = os.path.exists(self.model_path)
             modelopt_quant_config = {"quant_method": "modelopt"}
             if not is_local:
-                from huggingface_hub import HfApi
+                import huggingface_hub
+
+                try:
+                    from huggingface_hub import HfApi
+
+                    hf_api = HfApi()
+                    if hf_api.file_exists(self.model_path, "hf_quant_config.json"):
+                        quant_cfg = modelopt_quant_config
+                except huggingface_hub.errors.OfflineModeIsEnabled:
+                    logger.warning(
+                        "Offline mode is enabled, skipping hf_quant_config.json check"
+                    )
+                    pass
 
-                hf_api = HfApi()
-                if hf_api.file_exists(self.model_path, "hf_quant_config.json"):
-                    quant_cfg = modelopt_quant_config
             elif os.path.exists(os.path.join(self.model_path, "hf_quant_config.json")):
                 quant_config_file = os.path.join(
                     self.model_path, "hf_quant_config.json"