[Bugfix] Fix profiling OOM and decouple encoder multimodal profiling (#14361)

Signed-off-by: Isotr0py <2037008807@qq.com>

tests/multimodal/test_processing.py

@@ -873,7 +873,7 @@ def test_limit_mm_per_prompt_dummy(model_id, limit, num_supported, is_valid):
         exc_ctx = pytest.raises(ValueError, match="this model only supports")
 
     with exc_ctx:
-        profiler.get_dummy_data(model_config.max_model_len)
+        profiler.get_decoder_dummy_data(model_config.max_model_len)
 
 
 @pytest.mark.parametrize("model_id", ["llava-hf/llava-v1.6-mistral-7b-hf"])

vllm/inputs/registry.py

@@ -335,8 +335,10 @@ class InputRegistry:
                                                      tokenizer,
                                                      disable_cache=True)
             profiler = MultiModalProfiler(processor)
-            dummy_data = profiler.get_dummy_data(
-                seq_len, is_encoder_data=is_encoder_data)
+            dummy_data_factory = (profiler.get_encoder_dummy_data
+                                  if is_encoder_data else
+                                  profiler.get_decoder_dummy_data)
+            dummy_data = dummy_data_factory(seq_len)
         else:
             model_cls, _ = get_model_architecture(model_config)
             if is_encoder_data:

vllm/multimodal/profiling.py

@@ -3,7 +3,7 @@
 from abc import ABC, abstractmethod
 from collections.abc import Mapping
 from dataclasses import dataclass, field
-from typing import Generic, TypeVar
+from typing import Generic, TypeVar, cast
 
 import numpy as np
 import numpy.typing as npt
@@ -13,7 +13,8 @@ import vllm.envs as envs
 from vllm.inputs import DummyData
 from vllm.logger import init_logger
 
-from .inputs import MultiModalDataDict, MultiModalInputs
+from .inputs import (MultiModalDataDict, MultiModalEncDecInputs,
+                     MultiModalInputs)
 from .processing import BaseMultiModalProcessor, BaseProcessingInfo
 
 logger = init_logger(__name__)
@@ -142,14 +143,10 @@ class MultiModalProfiler(Generic[_I]):
             hf_processor_mm_kwargs=processor_inputs.hf_processor_mm_kwargs,
         )
 
-    def get_dummy_data(
+    def get_and_validate_mm_inputs(
         self,
         seq_len: int,
-        is_encoder_data: bool = False,
-    ) -> DummyData:
-        # Avoid circular import
-        from vllm.sequence import SequenceData
-
+    ) -> tuple[MultiModalInputs, Mapping[str, int]]:
         mm_counts = self.get_mm_limits()
 
         info = self.processing_info
@@ -165,11 +162,6 @@ class MultiModalProfiler(Generic[_I]):
 
         mm_inputs = self._get_dummy_mm_inputs(seq_len, mm_counts)
         placeholders_by_modality = mm_inputs["mm_placeholders"]
-        # For encoder-decoder models, use encoder prompt token ids instead of
-        # decoder prompt to construct dummy seq_data for encoder profiling.
-        prompt_token_ids = (
-            mm_inputs["prompt_token_ids"] if not is_encoder_data else
-            mm_inputs["encoder_prompt_token_ids"])  # type: ignore
 
         total_placeholders_by_modality = {
             modality: sum(item["length"] for item in placeholders)
@@ -185,12 +177,47 @@ class MultiModalProfiler(Generic[_I]):
                 f"{total_placeholders_by_modality} placeholder tokens, which "
                 f"is not the expected {expected_placeholders_by_modality} "
                 "tokens.")
+        return mm_inputs, total_placeholders_by_modality
 
+    def get_encoder_dummy_data(
+        self,
+        seq_len: int,
+    ) -> DummyData:
+        # Avoid circular import
+        from vllm.sequence import SequenceData
+
+        mm_inputs, _ = self.get_and_validate_mm_inputs(seq_len)
+        mm_inputs = cast(MultiModalEncDecInputs, mm_inputs)
+
+        # For encoder-decoder models, use encoder prompt token ids instead of
+        # decoder prompt to construct dummy seq_data for encoder profiling.
+        encoder_prompt_token_ids = mm_inputs["encoder_prompt_token_ids"]
+
+        total_len = len(encoder_prompt_token_ids)
+        num_tokens_to_pad = max(total_len, seq_len) - total_len
+        encoder_prompt_token_ids.extend([0] * num_tokens_to_pad)
+
+        return DummyData(
+            seq_data=SequenceData.from_seqs(encoder_prompt_token_ids),
+            multi_modal_data=None,
+            multi_modal_placeholders=None,
+        )
+
+    def get_decoder_dummy_data(
+        self,
+        seq_len: int,
+    ) -> DummyData:
+        # Avoid circular import
+        from vllm.sequence import SequenceData
+
+        (mm_inputs, total_placeholders_by_modality
+         ) = self.get_and_validate_mm_inputs(seq_len)
+
+        prompt_token_ids = mm_inputs["prompt_token_ids"]
         total_len = len(prompt_token_ids)
 
         # V0 does not support chunked prefill.
-        if (total_len > seq_len and not envs.VLLM_USE_V1) or is_encoder_data:
-            if total_len > seq_len and not is_encoder_data:
+        if total_len > seq_len and not envs.VLLM_USE_V1:
             logger.warning(
                 "The context length (%d) of the model is too short "
                 "to hold the multi-modal embeddings in the worst case "
@@ -202,11 +229,8 @@ class MultiModalProfiler(Generic[_I]):
                 "and/or reduce `mm_counts`.", seq_len, total_len,
                 total_placeholders_by_modality)
 
-            num_tokens_to_pad = max(total_len, seq_len) - total_len
-            prompt_token_ids.extend([0] * num_tokens_to_pad)
-
             return DummyData(
-                seq_data=SequenceData.from_seqs(prompt_token_ids),
+                seq_data=SequenceData.from_prompt_token_counts((0, seq_len)),
                 multi_modal_data=None,
                 multi_modal_placeholders=None,
             )
@@ -216,5 +240,5 @@ class MultiModalProfiler(Generic[_I]):
         return DummyData(
             seq_data=SequenceData.from_seqs(prompt_token_ids),
             multi_modal_data=mm_inputs["mm_kwargs"],
-            multi_modal_placeholders=placeholders_by_modality,
+            multi_modal_placeholders=mm_inputs["mm_placeholders"],
         )