fix

vllm/config.py

@@ -169,7 +169,7 @@ def __init__(
                                     code_revision, rope_scaling, rope_theta)
         self.hf_text_config = get_hf_text_config(self.hf_config)
         self.hf_image_processor_config = get_hf_image_processor_config(
-            self.model, trust_remote_code, revision)
+            self.model, revision)
         self.dtype = _get_and_verify_dtype(self.hf_text_config, dtype)
 
         # Choose a default enforce_eager value if the user did not specify

vllm/inputs/registry.py

@@ -57,12 +57,7 @@ def get_hf_config(self, hf_config_type: Type[C] = PretrainedConfig) -> C:
 
     def get_hf_image_processor_config(self) -> Dict:
         """
-        Get the HuggingFace configuration
-        (:class:`transformers.PretrainedConfig`) of the model,
-        additionally checking its type.
-
-        Raises:
-            TypeError: If the model is not of the specified type.
+        Get the HuggingFace image processor configuration of the model.
         """
 
         return self.model_config.hf_image_processor_config

vllm/model_executor/models/phi3v.py

@@ -15,7 +15,7 @@
 # limitations under the License.
 import re
 from functools import lru_cache
-from typing import (Iterable, List, Literal, Mapping, Optional, Tuple,
+from typing import (Dict, Iterable, List, Literal, Mapping, Optional, Tuple,
                     TypedDict, Union)
 
 import numpy as np
@@ -324,7 +324,7 @@ def _calc_hd_transform_size(*, width: int, height: int, hd_num: int = 16):
 
 # Based on https://huggingface.co/microsoft/Phi-3-vision-128k-instruct/blob/main/image_processing_phi3_v.py#L181
 def get_phi3v_image_feature_size(
-    hf_config: PretrainedConfig,
+    hf_config: Dict,
     *,
     input_height: int,
     input_width: int,

vllm/transformers_utils/config.py

@@ -101,29 +101,10 @@ def get_config(
 
 def get_hf_image_processor_config(
     model: Union[str, Path],
-    trust_remote_code: bool,
     revision: Optional[str] = None,
     **kwargs,
 ) -> Dict:
-    try:
-        config = get_image_processor_config(
-            model,
-            trust_remote_code=trust_remote_code,
-            revision=revision,
-            **kwargs)
-    except ValueError as e:
-        if (not trust_remote_code and
-                "requires you to execute the configuration file" in str(e)):
-            err_msg = (
-                "Failed to load the model config. If the model is a custom "
-                "model not yet available in the HuggingFace transformers "
-                "library, consider setting `trust_remote_code=True` in LLM "
-                "or using the `--trust-remote-code` flag in the CLI.")
-            raise RuntimeError(err_msg) from e
-        else:
-            raise e
-
-    return config
+    return get_image_processor_config(model, revision=revision, **kwargs)
 
 
 def get_hf_text_config(config: PretrainedConfig):