Add stream_chat and conditionally set AutoModelClass to MllamaForCond…

…itionalGeneration

llama-index-integrations/multi_modal_llms/llama-index-multi-modal-llms-huggingface/llama_index/multi_modal_llms/huggingface/base.py

@@ -19,6 +19,7 @@
     AutoConfig,
     Qwen2VLForConditionalGeneration,
     PaliGemmaForConditionalGeneration,
+    MllamaForConditionalGeneration
 )
 from qwen_vl_utils import (
     process_vision_info,
@@ -93,6 +94,8 @@ def __init__(self, **kwargs: Any) -> None:
                 AutoModelClass = Qwen2VLForConditionalGeneration
             if "PaliGemmaForConditionalGeneration" in architecture:
                 AutoModelClass = PaliGemmaForConditionalGeneration
+            if "MllamaForConditionalGeneration" in architecture:
+                AutoModelClass = MllamaForConditionalGeneration
 
             # Load the model based on the architecture
             self._model = AutoModelClass.from_pretrained(
@@ -179,6 +182,13 @@ async def astream_chat(
             "HuggingFaceMultiModal does not support async streaming chat yet."
         )
 
+    async def stream_chat(
+        self, messages: Sequence[ChatMessage], **kwargs: Any
+    ) -> ChatResponseAsyncGen:
+        raise NotImplementedError(
+            "HuggingFaceMultiModal does not support async streaming chat yet."
+        )
+
     async def astream_complete(
         self, prompt: str, image_documents: Sequence[ImageNode], **kwargs: Any
     ) -> CompletionResponseAsyncGen: