[Core][Frontend] Support Passing Multimodal Processor Kwargs (#8657)

Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>

tests/models/decoder_only/vision_language/test_qwen.py

@@ -5,14 +5,13 @@ import pytest
 import torch
 from PIL.Image import Image
 
-from vllm.config import ModelConfig
 from vllm.inputs import InputContext, LLMInputs
 from vllm.multimodal.base import MultiModalInputs
 from vllm.multimodal.utils import cached_get_tokenizer, rescale_image_size
 
 from ....conftest import (IMAGE_ASSETS, HfRunner, ImageAsset, PromptImageInput,
                           VllmRunner, _ImageAssets)
-from ...utils import check_logprobs_close
+from ...utils import build_model_context, check_logprobs_close
 
 text_only_models = [
     "Qwen/Qwen-7B-Chat"  # Has no visual component
@@ -42,32 +41,6 @@ VIS_ENC_DIM = 4096
 IMG_SIZE = 448
 
 
-def build_model_context(model_name: str,
-                        tokenizer_name: Optional[str] = None,
-                        trust_remote_code: bool = False):
-    """Creates an InputContext for a given model.
-    
-    Args:
-        model_name: Name of the model being considered.
-        tokenizer_name: Name of the tokenizer being considered.
-        trust_remote_code: Whether or not to allow loading remote code.
-
-    Returns:
-        InputContext for the model being considered.
-    """
-    if tokenizer_name is None:
-        tokenizer_name = model_name
-    model_config = ModelConfig(
-        model_name,
-        tokenizer_name,
-        tokenizer_mode="auto",
-        trust_remote_code=trust_remote_code,
-        dtype="float32",
-        seed=0,
-    )
-    return InputContext(model_config)
-
-
 @pytest.fixture()
 def input_mapper_for_qwen():
     # Lazy import to avoid initializing CUDA during test collection