[None][chore] Small cleanups to MultimodalModelMixin

Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

Author: 2ez4bz

tensorrt_llm/_torch/models/modeling_mistral.py

@@ -23,7 +23,7 @@
 from tensorrt_llm._torch.models.modeling_mistral_large3 import (
     Mistral3Gate, MistralLarge3ForCausalLM)
 from tensorrt_llm._torch.models.modeling_multimodal_mixin import (
-    MultimodalEncoderOutput, MultimodalModelMixin, PreparedLlmInputs)
+    MultimodalModelMixin, PreparedLlmInputs)
 from tensorrt_llm._torch.models.modeling_multimodal_utils import (
     _MULTIMODAL_ENV_NAME, _is_mm_disagg)
 from tensorrt_llm._torch.models.modeling_utils import (DecoderModel,
@@ -706,9 +706,9 @@ def encode_multimodal_inputs(
         self,
         multimodal_params: Sequence[MultimodalParams],
         **encoder_kwargs: Any,
-    ) -> MultimodalEncoderOutput:
+    ) -> torch.Tensor:
         mm_embeds = self._vision_forward(list(multimodal_params))
-        return MultimodalEncoderOutput(embeddings=mm_embeds[0])
+        return mm_embeds[0]
 
     def get_language_model_forward_kwargs(
         self,

tensorrt_llm/_torch/models/modeling_multimodal_mixin.py

@@ -29,27 +29,6 @@
 )
 
 
-@dataclass(frozen=True)
-class MultimodalEncoderOutput:
-    """Output produced by a model-owned multimodal encoder hook.
-
-    Contract:
-    - `embeddings` contains all multimodal embedding rows for the supplied
-      `multimodal_params`.
-    - Rows are concatenated in the same order as `multimodal_params`.
-    - Per-request row counts match `total_embeds_in_request` from runtime
-      metadata when that metadata is available.
-    - Special multimodal tokens occupy token positions but do not have rows in
-      this tensor.
-
-    The single-tensor shape is required for chunked-prefill embedding reuse,
-    which lets later chunks skip the encoder. See
-    `modeling_multimodal_utils.py` for the caching machinery.
-    """
-
-    embeddings: torch.Tensor
-
-
 @dataclass(frozen=True)
 class PreparedLlmInputs:
     """Prepared inputs returned by `MultimodalModelMixin`."""
@@ -71,8 +50,13 @@ def encode_multimodal_inputs(
         self,
         multimodal_params: Sequence[MultimodalParams],
         **encoder_kwargs: Any,
-    ) -> MultimodalEncoderOutput:
-        """Run model-specific multimodal encoder work."""
+    ) -> torch.Tensor:
+        """Run model-specific multimodal encoder work.
+
+        Returns the single primary multimodal embedding tensor for the supplied params. Rows are
+        expected to be concatenated in request order, and special multimodal tokens occupy token
+        positions but do not have rows here.
+        """
         raise NotImplementedError
 
     @property
@@ -122,16 +106,16 @@ def after_full_multimodal_embeddings(
         *,
         input_ids: torch.Tensor,
         multimodal_params: Sequence[MultimodalParams],
-        encoder_output: MultimodalEncoderOutput,
+        embeddings: torch.Tensor,
         **forward_kwargs: Any,
-    ) -> tuple[torch.Tensor, MultimodalEncoderOutput]:
+    ) -> tuple[torch.Tensor, torch.Tensor]:
         """Optional hook before active chunk rows are selected.
 
         Runs after cache lookup or encoder execution has produced full
         per-request multimodal embeddings, but before the mixin selects rows
         active in the current forward chunk.
         """
-        return input_ids, encoder_output
+        return input_ids, embeddings
 
     def after_active_multimodal_embeddings(
         self,
@@ -179,24 +163,19 @@ def prepare_multimodal_inputs(
             multimodal_params=context_params,
             **forward_kwargs,
         )
-        full_output = self._get_or_encode_multimodal_embeddings(
+        full_embeddings = self._get_or_encode_multimodal_embeddings(
             context_params,
             **encoder_kwargs,
         )
 
-        input_ids, full_output = self.after_full_multimodal_embeddings(
+        input_ids, full_embeddings = self.after_full_multimodal_embeddings(
             input_ids=input_ids,
             multimodal_params=context_params,
-            encoder_output=full_output,
+            embeddings=full_embeddings,
             **forward_kwargs,
         )
 
-        active_embeddings = self._find_active_multimodal_embeddings(
-            [full_output.embeddings],
-            input_ids=input_ids,
-            positions=positions,
-            multimodal_params=context_params,
-        )
+        active_embeddings = find_input_mm_embeds([full_embeddings], list(context_params))
         active_embeddings, extra_embeds = self.after_active_multimodal_embeddings(
             active_embeddings=active_embeddings,
             multimodal_params=context_params,
@@ -228,50 +207,21 @@ def _get_or_encode_multimodal_embeddings(
         self,
         multimodal_params: Sequence[MultimodalParams],
         **encoder_kwargs: Any,
-    ) -> MultimodalEncoderOutput:
+    ) -> torch.Tensor:
         """Return cached multimodal embeddings or run the encoder for misses.
 
-        Delegates cache lookup and gather behavior to
-        `get_multimodal_embeddings`, then validates the single primary tensor
-        contract for both encoded and cached-only paths.
+        Delegates cache lookup and gather behavior to `get_multimodal_embeddings`, then validates
+        the single tensor contract for both encoded and cached-only paths.
         """
-
-        def encoder_forward_fn(params: list[MultimodalParams], **kwargs: Any) -> list[torch.Tensor]:
-            encoder_output = self.encode_multimodal_inputs(params, **kwargs)
-            if not isinstance(encoder_output, MultimodalEncoderOutput):
-                raise TypeError("encode_multimodal_inputs must return MultimodalEncoderOutput.")
-            if not isinstance(encoder_output.embeddings, torch.Tensor):
-                raise TypeError("MultimodalEncoderOutput.embeddings must be a torch.Tensor.")
-            return [encoder_output.embeddings]
-
         embeddings = get_multimodal_embeddings(
-            encoder_forward_fn=encoder_forward_fn,
+            encoder_forward_fn=self.encode_multimodal_inputs,
             multimodal_params=list(multimodal_params),
             encoder_kwargs=encoder_kwargs,
         )
-        primary = self._require_primary_embedding(embeddings)
-        # Validate post-gather so cached-only paths (KV reuse, all-cached chunked
-        # prefill) are also checked, not just paths that ran the encoder.
-        self._validate_primary_embedding_rows(primary, multimodal_params)
-        return MultimodalEncoderOutput(embeddings=primary)
-
-    def _find_active_multimodal_embeddings(
-        self,
-        multimodal_embeddings: list[torch.Tensor],
-        *,
-        input_ids: torch.Tensor,
-        positions: Optional[torch.Tensor],
-        multimodal_params: Sequence[MultimodalParams],
-    ) -> list[torch.Tensor]:
-        """Named internal stage for selecting active chunk multimodal rows.
-
-        This initial template stage currently delegates to
-        `find_input_mm_embeds`. Model-specific behavior around slicing should
-        use `after_full_multimodal_embeddings` or
-        `after_active_multimodal_embeddings` so the common mixin sequence stays
-        centralized.
-        """
-        return find_input_mm_embeds(multimodal_embeddings, list(multimodal_params))
+        # Validate post-gather so cached-only paths (KV reuse, all-cached chunked prefill) are also
+        # checked, not just paths that ran the encoder.
+        self._validate_embeddings(embeddings, multimodal_params)
+        return embeddings[0]
 
     def _fuse_multimodal_embeddings(
         self,
@@ -313,38 +263,35 @@ def _fuse_multimodal_embeddings(
         return fused_input_ids, inputs_embeds, ()
 
     @staticmethod
-    def _require_primary_embedding(embeddings: list[torch.Tensor]) -> torch.Tensor:
-        if len(embeddings) != 1:
-            raise ValueError(
-                "MultimodalModelMixin requires a single primary embedding tensor, "
-                f"got {len(embeddings)} tensors."
-            )
-        return embeddings[0]
-
-    @staticmethod
-    def _validate_primary_embedding_rows(
-        primary: torch.Tensor,
+    def _validate_embeddings(
+        embeddings: list[torch.Tensor],
         multimodal_params: Sequence[MultimodalParams],
     ) -> None:
-        """Validate gathered primary embedding row count against runtime metadata.
+        """Validate gathered embeddings embedding row count against runtime metadata.
 
         Skipped if any param lacks `multimodal_runtime.total_embeds_in_request`, since the contract
         cannot be evaluated without complete metadata.
         """
+        if len(embeddings) != 1:
+            raise ValueError(
+                f"MultimodalModelMixin requires a single embedding tensor, got {len(embeddings)} "
+                "tensors."
+            )
+
+        embeddings_tensor = embeddings[0]
         expected_rows = 0
         for param in multimodal_params:
             runtime = param.multimodal_runtime
             if runtime is None or runtime.total_embeds_in_request is None:
                 logger.debug(
-                    "Skipping multimodal embedding row-count validation: "
-                    "runtime metadata missing or incomplete for at least one param."
+                    "Skipping multimodal embedding row-count validation: runtime metadata missing "
+                    "or incomplete for at least one param."
                 )
                 return
             expected_rows += runtime.total_embeds_in_request
 
-        actual_rows = primary.shape[0]
+        actual_rows = embeddings_tensor.shape[0]
         if actual_rows != expected_rows:
             raise ValueError(
-                "Multimodal embedding row count mismatch: "
-                f"expected {expected_rows}, got {actual_rows}."
+                f"Multimodal embedding row count mismatch: expected {expected_rows}, got {actual_rows}."
             )

tensorrt_llm/_torch/models/modeling_multimodal_utils.py

@@ -196,10 +196,26 @@ def _cache_multimodal_embeddings(
     )
 
 
+def _normalize_encoder_embeddings(
+    encoder_embeddings: Union[torch.Tensor, List[torch.Tensor]],
+) -> List[torch.Tensor]:
+    if isinstance(encoder_embeddings, torch.Tensor):
+        return [encoder_embeddings]
+
+    if (not isinstance(encoder_embeddings, list) or not all(
+            isinstance(embedding, torch.Tensor)
+            for embedding in encoder_embeddings)):
+        raise TypeError(
+            "encoder_forward_fn must return a torch.Tensor or a list of torch.Tensor."
+        )
+
+    return encoder_embeddings
+
+
 def get_multimodal_embeddings(
     encoder_forward_fn: Callable[
-        [List[MultimodalParams]],
-        List[torch.Tensor],
+        ...,
+        torch.Tensor | List[torch.Tensor],
     ],
     multimodal_params: List[MultimodalParams],
     encoder_kwargs: Optional[Dict[str, Any]] = None,
@@ -215,7 +231,8 @@ def get_multimodal_embeddings(
 
     Args:
         encoder_forward_fn: Callable that performs encoder forward pass.
-                           Should accept List[MultimodalParams] and return List[torch.Tensor].
+                           Should accept List[MultimodalParams] and return either
+                           a single torch.Tensor or List[torch.Tensor].
         multimodal_params: All multimodal parameters in the batch.
         encoder_kwargs: Optional kwargs to pass to encoder_forward_fn.
     Returns:
@@ -233,6 +250,7 @@ def get_multimodal_embeddings(
         kwargs = encoder_kwargs or {}
         encoder_embeddings = encoder_forward_fn(uncached_multimodal_params,
                                                 **kwargs)
+        encoder_embeddings = _normalize_encoder_embeddings(encoder_embeddings)
 
         # TODO: support multiple multimodal modalities per request
         if len(encoder_embeddings) > 1:

tests/unittest/_torch/multimodal/test_multimodal_mixin.py

@@ -48,10 +48,28 @@ def encode_multimodal_inputs(self, multimodal_params, **encoder_kwargs):
         raise AssertionError("Tests use cached multimodal embeddings and should not encode.")
 
 
+class TensorEncoderMultimodalModel(DummyMultimodalModel):
+    def __init__(
+        self,
+        embedding: Embedding,
+        mm_token_ids: torch.Tensor,
+        mm_embeds: torch.Tensor,
+    ):
+        super().__init__(embedding, mm_token_ids)
+        self.mm_embeds = mm_embeds
+
+    def encode_multimodal_inputs(self, multimodal_params, **encoder_kwargs) -> torch.Tensor:
+        return self.mm_embeds
+
+
 def make_cached_multimodal_param(mm_embeds: torch.Tensor) -> MultimodalParams:
     return MultimodalParams(multimodal_data={"multimodal_embedding": mm_embeds})
 
 
+def make_raw_multimodal_param() -> MultimodalParams:
+    return MultimodalParams(multimodal_data={"image": {"pixel_values": torch.empty(1)}})
+
+
 @pytest.mark.parametrize("device", ["cpu"] + (["cuda"] if torch.cuda.is_available() else []))
 def test_prepare_multimodal_inputs_forwards_precomputed_indices(device):
     hidden = 8
@@ -84,3 +102,37 @@ def test_prepare_multimodal_inputs_forwards_precomputed_indices(device):
         mm_emb.to(dtype=out.inputs_embeds.dtype, device=out.inputs_embeds.device),
     )
     torch.testing.assert_close(out.inputs_embeds[text_idx], emb(input_ids[text_idx]))
+
+
+@pytest.mark.parametrize("device", ["cpu"] + (["cuda"] if torch.cuda.is_available() else []))
+def test_prepare_multimodal_inputs_accepts_tensor_encoder_output(device):
+    hidden = 8
+    mm_token_id = 7
+    emb = make_embedding(num_embeddings=40, hidden_size=hidden, device=device)
+
+    input_ids = torch.tensor([0, mm_token_id, 1], dtype=torch.long, device=device)
+    text_idx = torch.tensor([0, 2], dtype=torch.long, device=device)
+    mm_idx = torch.tensor([1], dtype=torch.long, device=device)
+    mm_emb = torch.randn(mm_idx.shape[0], hidden, device=device)
+    model = TensorEncoderMultimodalModel(
+        emb,
+        torch.tensor([mm_token_id], dtype=torch.long, device=device),
+        mm_emb,
+    )
+
+    out = model.prepare_multimodal_inputs(
+        input_ids=input_ids,
+        positions=None,
+        multimodal_params=[make_raw_multimodal_param()],
+        num_context_requests=1,
+        text_token_indices=text_idx,
+        mm_token_indices=mm_idx,
+    )
+
+    assert out.input_ids is None
+    assert out.inputs_embeds is not None
+    torch.testing.assert_close(
+        out.inputs_embeds[mm_idx],
+        mm_emb.to(dtype=out.inputs_embeds.dtype, device=out.inputs_embeds.device),
+    )
+    torch.testing.assert_close(out.inputs_embeds[text_idx], emb(input_ids[text_idx]))