[4/N] Initialize MM components in context managers (M-P) (#32663)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/model_executor/models/step3_vl.py

@@ -1104,14 +1104,6 @@ class Step3VLForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP)
     ) -> torch.Tensor | IntermediateTensors:
         if intermediate_tensors is not None:
             inputs_embeds = None
-        elif inputs_embeds is None:
-            vision_embeddings = self.embed_multimodal(**kwargs)
-            inputs_embeds = self.embed_input_ids(
-                input_ids,
-                vision_embeddings,
-                is_multimodal=input_ids == self.config.image_token_id,
-            )
-            input_ids = None
 
         hidden_states = self.language_model(
             input_ids, positions, intermediate_tensors, inputs_embeds=inputs_embeds