[TRTLLM-9522][chore] implement default attach_multimodal_embeddings (#9664)

Signed-off-by: ixlmar <206748156+ixlmar@users.noreply.github.com>

tensorrt_llm/inputs/registry.py

@@ -141,6 +141,21 @@ class BaseMultimodalInputProcessor(ABC):
         self._use_fast: bool = kwargs.get('use_fast', True)
         self._multimodal_hashing_supported: Optional[bool] = None
 
+    def attach_multimodal_embeddings(
+        self,
+        inputs: TextPrompt,
+        multimodal_embedding: Dict[str, List[torch.Tensor]],
+        sampling_params: SamplingParams,
+    ) -> Tuple[List[int], Optional[ExtraProcessedInputs]]:
+        """
+        Handle externally provided multimodal input embeddings.
+
+        While inputs["multi_modal_data"] is handled by __call__, this method is intended to process
+        inputs["multi_modal_embeddings"].
+        """
+        raise NotImplementedError(
+            "Input processor does not support multimodal embedding input")
+
     @property
     @abstractmethod
     def processor(self) -> AutoProcessor:

tensorrt_llm/llmapi/llm.py

@@ -8,7 +8,7 @@ import time
 import weakref
 from collections.abc import Mapping
 from pathlib import Path
-from typing import Any, List, Literal, Optional, Sequence, Union
+from typing import Any, List, Literal, Optional, Sequence, Union, cast
 
 import transformers
 from tqdm import tqdm
@@ -17,7 +17,8 @@ from transformers import PreTrainedTokenizerBase
 from tensorrt_llm._utils import mpi_disabled
 from tensorrt_llm.inputs.data import TextPrompt
 from tensorrt_llm.inputs.multimodal import MultimodalInput, MultimodalParams
-from tensorrt_llm.inputs.registry import DefaultInputProcessor
+from tensorrt_llm.inputs.registry import (BaseMultimodalInputProcessor,
+                                          DefaultInputProcessor)
 from tensorrt_llm.llmapi import tracing
 from tensorrt_llm.metrics.enums import MetricNames
 
@@ -458,8 +459,10 @@ class BaseLLM:
                         inputs, sampling_params)
             elif 'multi_modal_embeddings' in inputs:
                 mm_embedding_info = inputs['multi_modal_embeddings']
-                prompt_token_ids, extra_processed_inputs = self.input_processor.attach_multimodal_embeddings(
+                prompt_token_ids, extra_processed_inputs = cast(
-                    inputs, mm_embedding_info, sampling_params)
+                    self.input_processor,
+                    BaseMultimodalInputProcessor).attach_multimodal_embeddings(
+                        inputs, mm_embedding_info, sampling_params)
             else:
                 with nvtx_range_debug("input_processor"):
                     prompt_token_ids, extra_processed_inputs = self.input_processor(