[Bugfix] Fix image input for Pixtral-HF (#11741)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
2025-01-08 10:17:16 +08:00
parent 5950f555a1
commit 91445c7bc8
4 changed files with 52 additions and 6 deletions
--- a/vllm/model_executor/models/llava.py
+++ b/vllm/model_executor/models/llava.py
@@ -546,6 +546,12 @@ class LlavaForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP):
                raise ValueError("Incorrect type of pixel values. "
                                 f"Got type: {type(pixel_values)}")

+            if self.config.vision_config.model_type == "pixtral":
+                return LlavaImagePixelInputs(
+                    type="pixel_values",
+                    data=flatten_bn(pixel_values),
+                )
+
            return LlavaImagePixelInputs(
                type="pixel_values",
                data=self._validate_pixel_values(