diff --git a/tests/models/multimodal/processing/test_qwen3_vl.py b/tests/models/multimodal/processing/test_qwen3_vl.py index 9155fde503..a1ab94ca43 100644 --- a/tests/models/multimodal/processing/test_qwen3_vl.py +++ b/tests/models/multimodal/processing/test_qwen3_vl.py @@ -138,3 +138,49 @@ def test_processor_multi_video( assert video_phs[i].offset >= prev_end, ( f"Placeholder {i} overlaps with placeholder {i - 1}" ) + + +@pytest.mark.parametrize("model_id", [MODEL_ID]) +@pytest.mark.parametrize( + "hf_mm_kwargs", + [{"num_frames": [8, 16]}, {"fps": [2.0, 4.0]}], +) +def test_processor_multi_video_list_kwargs( + model_id: str, + hf_mm_kwargs: dict[str, Any], +) -> None: + """Regression test: a multi-video request with list-valued per-video + ``mm_processor_kwargs`` (one ``fps``/``num_frames`` per video) must not + crash. + + Before the fix, ``_call_hf_processor`` copied the whole kwargs to every + video without slicing, so ``_get_video_second_idx`` received the list + where a scalar was expected and raised ``TypeError``. + """ + ctx = build_model_context( + model_id, + limit_mm_per_prompt={"image": 0, "video": 2}, + ) + processor = MULTIMODAL_REGISTRY.create_processor(ctx.model_config) + + prompt = ( + "<|vision_start|><|video_pad|><|vision_end|>" + "<|vision_start|><|video_pad|><|vision_end|>" + ) + mm_data = { + "video": [ + _build_video_mm_data(num_frames=16)["video"][0], + _build_video_mm_data(num_frames=32)["video"][0], + ] + } + + processed = processor( + prompt, + mm_items=processor.info.parse_mm_data(mm_data), + hf_processor_mm_kwargs=hf_mm_kwargs, + ) + + video_phs = processed["mm_placeholders"].get("video", []) + assert len(video_phs) == 2, ( + f"Expected exactly 2 video placeholders, got {len(video_phs)}" + ) diff --git a/vllm/model_executor/models/qwen3_vl.py b/vllm/model_executor/models/qwen3_vl.py index 3183a23ffd..0e6ddef7f3 100644 --- a/vllm/model_executor/models/qwen3_vl.py +++ b/vllm/model_executor/models/qwen3_vl.py @@ -1271,7 +1271,7 @@ class Qwen3VLMultiModalProcessor(BaseMultiModalProcessor[Qwen3VLProcessingInfo]) vision_end_token_id = hf_config.vision_end_token_id video_token_id = hf_config.video_token_id - for item in videos: + for item_idx, item in enumerate(videos): video_array, metadata = item # NOTE: @JJJYmmm new attr metadata.frames_indices indicates @@ -1282,6 +1282,12 @@ class Qwen3VLMultiModalProcessor(BaseMultiModalProcessor[Qwen3VLProcessingInfo]) # NOTE: a copy of is created to update do_sample_frames, # otherwise mm_hash for the object will be incorrect. video_mm_kwargs = dict(**mm_kwargs) + sampled_fps = video_mm_kwargs.get("fps") + if is_list_of(sampled_fps, float): + video_mm_kwargs["fps"] = sampled_fps[item_idx] + sampled_num_frames = video_mm_kwargs.get("num_frames") + if is_list_of(sampled_num_frames, int): + video_mm_kwargs["num_frames"] = sampled_num_frames[item_idx] if "do_sample_frames" not in video_mm_kwargs: # qwen_vl_utils already has "do_sample_frames" in # mm_kwargs, don't overwrite it. @@ -1363,10 +1369,15 @@ class Qwen3VLMultiModalProcessor(BaseMultiModalProcessor[Qwen3VLProcessingInfo]) else: video_outputs = dict() + # fps/num_frames are video-only kwargs already consumed by the loop; + # exclude them so the text/image processor call below never gets a list. + non_video_mm_kwargs = { + k: v for k, v in mm_kwargs.items() if k not in ("fps", "num_frames") + } processed_outputs = super()._call_hf_processor( prompt=prompt, mm_data=mm_data, - mm_kwargs=mm_kwargs, + mm_kwargs=non_video_mm_kwargs, tok_kwargs=tok_kwargs, )