biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Refactor] Use data parser for matching data items to multi-modal UUIDs (#32955)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-01-26 15:00:28 +08:00
committed by GitHub
parent ee484b3f4b
commit 11b556878b
14 changed files with 701 additions and 604 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
vllm/model_executor/models/qwen3_vl.py
View File

@@ -113,7 +113,11 @@ from .qwen2_5_vl import (
Qwen2_5_VLVideoInputs,
Qwen2_5_VLVideoPixelInputs,
)
from .qwen2_vl import Qwen2VLMultiModalDataParser, Qwen2VLProcessingInfo
from .qwen2_vl import (
Qwen2VLMultiModalDataParser,
Qwen2VLProcessingInfo,
_create_qwen2vl_field_factory,
)
from .qwen3 import Qwen3ForCausalLM, Qwen3Model
from .utils import (
AutoWeightsLoader,
@@ -985,28 +989,9 @@ class Qwen3VLMultiModalProcessor(BaseMultiModalProcessor[Qwen3VLProcessingInfo])
hf_inputs: BatchFeature,
hf_processor_mm_kwargs: Mapping[str, object],
) -> Mapping[str, MultiModalFieldConfig]:
image_grid_thw = hf_inputs.get("image_grid_thw", torch.empty((0, 3)))
image_grid_sizes = image_grid_thw.prod(-1)
video_grid_thw = hf_inputs.get("video_grid_thw", torch.empty((0, 3)))
video_grid_sizes = video_grid_thw.prod(-1)
return dict(
pixel_values=MultiModalFieldConfig.flat_from_sizes(
"image", image_grid_sizes
),
image_embeds=MultiModalFieldConfig.flat_from_sizes(
"image", image_grid_sizes
),
image_grid_thw=MultiModalFieldConfig.batched("image", keep_on_cpu=True),
pixel_values_videos=MultiModalFieldConfig.flat_from_sizes(
"video", video_grid_sizes
),
video_embeds=MultiModalFieldConfig.flat_from_sizes(
"video", video_grid_sizes
),
video_grid_thw=MultiModalFieldConfig.batched("video", keep_on_cpu=True),
)
return _create_qwen2vl_field_factory(
self.info.get_hf_config().vision_config.spatial_merge_size
)(hf_inputs)
def _get_prompt_updates(
self,