[Model] Define merge_by_field_config MM interface (U-Z) (vllm-project#26261)

Signed-off-by: Ayush Satyam <ayushsatyam146@gmail.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk>

Author: 2 people

vllm/model_executor/models/ultravox.py

@@ -69,18 +69,16 @@ class UltravoxAudioFeatureInputs(TensorSchema):
     type: Literal["audio_features"]
     data: Annotated[
         Union[torch.Tensor, list[torch.Tensor], list[list[torch.Tensor]]],
-        TensorShape("b", "n", "nmb", "t", dynamic_dims={"n"}),
+        TensorShape("bn", "nmb", "t"),
     ]
-    lens: Annotated[
-        Union[torch.Tensor, list[torch.Tensor]],
-        TensorShape("b", "n", dynamic_dims={"n"}),
-    ]
-    """Length of the audio frames. Used for attention mask in WhisperEncoder."""
-    token_len: Annotated[
-        Union[torch.Tensor, list[torch.Tensor]],
-        TensorShape("b", "n", dynamic_dims={"n"}),
-    ]
-    """Length of the audio tokens. Used for flattening the audio features."""
+    lens: Annotated[torch.Tensor, TensorShape("bn")]
+    """
+    Length of the audio frames per chunk. Used for attention mask in WhisperEncoder.
+    """
+    token_len: Annotated[torch.Tensor, TensorShape("bn")]
+    """Length of the audio tokens per chunk. Used for flattening the audio features."""
+    num_chunks: Annotated[torch.Tensor, TensorShape("n")]
+    """Number of chunks per audio. Used for flattening the audio features."""
 
 
 class UltravoxAudioEmbeddingInputs(TensorSchema):
@@ -421,6 +419,8 @@ def forward(
     dummy_inputs=UltravoxDummyInputsBuilder,
 )
 class UltravoxModel(nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA):
+    merge_by_field_config = True
+
     packed_modules_mapping = {
         "qkv_proj": ["q_proj", "k_proj", "v_proj"],
         "gate_up_proj": ["gate_proj", "up_proj"],
@@ -519,6 +519,7 @@ def _parse_and_validate_audio_input(
         audio_embeds = kwargs.pop("audio_embeds", None)
         audio_lens = kwargs.pop("audio_lens", None)
         audio_token_len = kwargs.pop("audio_token_len", None)
+        audio_num_chunks = kwargs.pop("audio_num_chunks", None)
 
         if audio_features is None and audio_embeds is None:
             return None
@@ -529,6 +530,7 @@ def _parse_and_validate_audio_input(
                 data=audio_features,
                 lens=audio_lens,
                 token_len=audio_token_len,
+                num_chunks=audio_num_chunks,
             )
 
         if audio_embeds is not None:
@@ -547,9 +549,8 @@ def _process_audio_input(
         # [[B1, 80, M1], [B2, 80, M2]] -> [B1+B2, 80, max(M1, M2)]
         audio_features = pad_and_concat_to_dim3(audio_input["data"])
 
-        # [B1, B2] -> [B1+B2]
-        audio_lens = flatten_bn(audio_input["lens"], concat=True)
-        audio_token_len = flatten_bn(audio_input["token_len"], concat=True)
+        audio_lens = audio_input["lens"]
+        audio_token_len = audio_input["token_len"]
 
         embeddings = self._audio_features_to_embeddings(audio_features, audio_lens)
 
@@ -568,7 +569,8 @@ def _process_audio_input(
 
         # Return one tensor per input audio
         embed_lens = [
-            token_len_item.sum().item() for token_len_item in audio_input["token_len"]
+            chunk_lens.sum().item()
+            for chunk_lens in audio_token_len.split(audio_input["num_chunks"].tolist())
         ]
         return flattened_embeddings.split(embed_lens)
 
@@ -663,6 +665,7 @@ def pad_and_concat_to_dim3(
         if features.ndim > 3:
             # Flatten [B, N, 80, M] -> [B * N, 80, M]
             features = flatten_bn(features)
+
         return features
 
     features = [pad_and_concat_to_dim3(f) for f in features]

vllm/model_executor/models/voxtral.py

@@ -61,7 +61,7 @@
 )
 
 from .interfaces import SupportsLoRA, SupportsMultiModal, SupportsTranscription
-from .utils import flatten_bn, init_vllm_registered_model, maybe_prefix
+from .utils import init_vllm_registered_model, maybe_prefix
 
 logger = init_logger(__name__)
 
@@ -337,6 +337,8 @@ def _get_data_parser(self) -> MultiModalDataParser:
 class VoxtralForConditionalGeneration(
     nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA, SupportsTranscription
 ):
+    merge_by_field_config = True
+
     supported_languages = ISO639_1_SUPPORTED_LANGS
 
     packed_modules_mapping = {
@@ -445,7 +447,6 @@ def _parse_and_validate_audio_arrays(
                 f"Incorrect type of audio_arrays. Got type: {type(audio_arrays)}"
             )
 
-        audio_arrays = flatten_bn(audio_arrays)
         if isinstance(audio_arrays, torch.Tensor):
             audio_arrays = list(audio_arrays.unbind(0))
         return audio_arrays

vllm/model_executor/models/whisper.py

@@ -36,7 +36,7 @@
 from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
 from vllm.model_executor.model_loader.utils import set_default_torch_dtype
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
-from vllm.multimodal import MULTIMODAL_REGISTRY, NestedTensors
+from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.multimodal.inputs import (
     MultiModalDataDict,
     MultiModalFieldConfig,
@@ -51,6 +51,7 @@
 )
 from vllm.multimodal.profiling import BaseDummyInputsBuilder
 from vllm.transformers_utils.processor import cached_get_processor
+from vllm.utils.jsontree import json_map_leaves
 from vllm.utils.tensor_schema import TensorSchema, TensorShape
 
 from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsTranscription
@@ -135,7 +136,10 @@ class WhisperAudioInputs(TensorSchema):
         - t: Time frames (M)
     """
 
-    input_features: Annotated[Optional[NestedTensors], TensorShape("b", "nmb", "t")]
+    input_features: Annotated[
+        Optional[list[torch.Tensor]],
+        TensorShape("b", "nmb", "t"),
+    ]
 
 
 class WhisperEncoderAttention(MultiHeadAttention):
@@ -781,6 +785,7 @@ def _get_prompt_updates(
 class WhisperForConditionalGeneration(
     nn.Module, SupportsTranscription, SupportsMultiModal
 ):
+    merge_by_field_config = True
     packed_modules_mapping = {
         "self_attn.qkv_proj": [
             "self_attn.q_proj",
@@ -936,12 +941,7 @@ def _parse_and_validate_audio_input(self, **kwargs: object) -> WhisperAudioInput
         input_features = kwargs.pop("input_features", None)
 
         if input_features is not None:
-            if not isinstance(input_features, (torch.Tensor, list)):
-                raise ValueError(
-                    "Incorrect type of audio features. "
-                    f"Got type: {type(input_features)}"
-                )
-            input_features = torch.cat([feat.to(self.dtype) for feat in input_features])
+            input_features = json_map_leaves(lambda x: x.to(self.dtype), input_features)
 
         return WhisperAudioInputs(input_features=input_features)
 

vllm/multimodal/inputs.py

@@ -677,6 +677,9 @@ def __init__(self, field: BaseMultiModalField, modality: str) -> None:
         self.field = field
         self.modality = modality
 
+    def __repr__(self) -> str:
+        return f"MultiModalFieldConfig(field={self.field}, modality={self.modality})"
+
     def build_elems(
         self,
         key: str,