[Multi Modal] Configurable MM Profiling (#25631)

Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>

vllm/model_executor/models/deepseek_vl2.py

@@ -14,6 +14,7 @@ from einops import rearrange, repeat
 from transformers import BatchFeature
 
 from vllm.config import VllmConfig
+from vllm.config.multimodal import BaseDummyOptions
 from vllm.distributed import get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.model_loader.utils import set_default_torch_dtype
@@ -191,16 +192,20 @@ class DeepseekVL2DummyInputsBuilder(
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
+        mm_options: Optional[Mapping[str, BaseDummyOptions]] = None,
     ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         max_image_size = self.info.get_image_size_with_most_features()
 
+        image_overrides = mm_options.get("image") if mm_options else None
+
         return {
             "image":
             self._get_dummy_images(width=max_image_size.width,
                                    height=max_image_size.height,
-                                   num_images=num_images)
+                                   num_images=num_images,
+                                   overrides=image_overrides)
         }