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

[Multi Modal] Configurable MM Profiling (#25631)

Browse Source 
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Wenlong Wang
2025-10-03 03:59:10 -07:00
committed by GitHub
parent 2ed3f20dba
commit 79aa244678
60 changed files with 654 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

15
vllm/model_executor/models/qwen2_5_omni_thinker.py
View File

@@ -39,6 +39,7 @@ from transformers.models.qwen2_5_omni.processing_qwen2_5_omni import (
from transformers.models.whisper import WhisperFeatureExtractor
from vllm.config import VllmConfig
from vllm.config.multimodal import BaseDummyOptions
from vllm.logger import init_logger
from vllm.model_executor.layers.rotary_embedding import MRotaryEmbedding
from vllm.model_executor.models.module_mapping import MultiModelKeys
@@ -212,6 +213,7 @@ class Qwen2_5OmniThinkerDummyInputsBuilder(
self,
seq_len: int,
mm_counts: Mapping[str, int],
mm_options: Optional[Mapping[str, BaseDummyOptions]] = None,
) -> MultiModalDataDict:
num_audios = mm_counts.get("audio", 0)
num_images = mm_counts.get("image", 0)
@@ -228,19 +230,26 @@ class Qwen2_5OmniThinkerDummyInputsBuilder(
target_num_frames = \
self.info.get_num_frames_with_most_features(seq_len, mm_counts)
image_overrides = mm_options.get("image") if mm_options else None
video_overrides = mm_options.get("video") if mm_options else None
audio_overrides = mm_options.get("audio") if mm_options else None
mm_data = {
"audio":
self._get_dummy_audios(length=target_audio_length,
num_audios=num_audios),
num_audios=num_audios,
overrides=audio_overrides),
"image":
self._get_dummy_images(width=target_width,
height=target_height,
num_images=num_images),
num_images=num_images,
overrides=image_overrides),
"video":
self._get_dummy_videos(width=target_width,
height=target_height,
num_frames=target_num_frames,
num_videos=num_videos),
num_videos=num_videos,
overrides=video_overrides),
}
return mm_data