[Meta] Official Eagle mm support, first enablement on llama4 (#20788)

Signed-off-by: morgendave <morgendave@gmail.com>
Co-authored-by: Roger Wang <hey@rogerw.me>

vllm/model_executor/models/llama_eagle.py

@@ -2,6 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 
 from collections.abc import Iterable
+from typing import Optional
 
 import torch
 import torch.nn as nn
@@ -148,7 +149,12 @@ class EagleLlamaForCausalLM(LlamaForCausalLM):
         input_ids: torch.Tensor,
         positions: torch.Tensor,
         hidden_states: torch.Tensor,
+        inputs_embeds: Optional[torch.Tensor] = None,
     ) -> tuple[torch.Tensor, torch.Tensor]:
+        if inputs_embeds is not None:
+            raise NotImplementedError(
+                f"{type(self).__name__} does not support multimodal inputs yet."
+            )
         return self.model(input_ids, positions, hidden_states)
 
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]):