[V1][Molmo] Fix get_multimodal_embeddings() in molmo.py (#14161)

2025-03-04 07:43:59 -08:00
parent c8525f06fc
commit b3cf368d79
22 changed files with 249 additions and 150 deletions
--- a/vllm/model_executor/models/idefics3.py
+++ b/vllm/model_executor/models/idefics3.py
@@ -617,7 +617,9 @@ class Idefics3ForConditionalGeneration(nn.Module, SupportsMultiModal,
        self.logits_processor = LogitsProcessor(config.text_config.vocab_size)
        self.sampler = get_sampler()

-    def get_multimodal_embeddings(self, **kwargs) -> Optional[NestedTensors]:
+    def get_multimodal_embeddings(
+        self, **kwargs
+    ) -> Union[list[torch.Tensor], torch.Tensor, tuple[torch.Tensor, ...]]:
        image_input = self.model._parse_and_validate_image_input(**kwargs)
        if image_input is None:
            return None