Skip to content

[Bugfix] Fix qwen3-omni audio truncation issue #26815

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
DarkLight1337 merged 4 commits into from
Oct 15, 2025
Merged

[Bugfix] Fix qwen3-omni audio truncation issue #26815

Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
70532f6
fix qwen3-omni audio truncation issue
Isotr0py Oct 14, 2025
9281723
enforce truncation=False
Isotr0py Oct 14, 2025
bd61ae5
ooops
Isotr0py Oct 14, 2025
7fd991a
add comment and correct truncation case
Isotr0py Oct 14, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions vllm/model_executor/models/qwen3_omni_moe_thinker.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -711,11 +711,12 @@ def pad_to_hop_length(x: np.ndarray, hop_length: int) -> np.ndarray:
return x

# NOTE: WhisperFeatureExtractor cannot handle empty list of audios
feature_extractor = self.info.get_feature_extractor()

@DarkLight1337 DarkLight1337 Oct 14, 2025

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you add a comment so we can revert once it's fixed on transformers side?

hop_length = feature_extractor.hop_length
if audios:
# NOTE: Qwen3-Omni processor accept "audio"
# To make sure the cache works with padding=True, we pre-padded
# the audio to multiple of hop_length.
hop_length = self.info.get_feature_extractor().hop_length
mm_data["audio"] = [
pad_to_hop_length(audio, hop_length)
if isinstance(audio, np.ndarray)
Expand All @@ -725,6 +726,7 @@ def pad_to_hop_length(x: np.ndarray, hop_length: int) -> np.ndarray:
mm_kwargs = dict(
**mm_kwargs,
)
mm_kwargs["truncation"] = False

hf_inputs = super()._call_hf_processor(
prompt=prompt,
Expand All @@ -738,7 +740,6 @@ def pad_to_hop_length(x: np.ndarray, hop_length: int) -> np.ndarray:
and "feature_attention_mask" in hf_inputs
and (audios := mm_data.get("audio", []))
):
hop_length = self.info.get_feature_extractor().hop_length
audio_num_frames = []
for _, audio in enumerate(audios):
audio_length = len(audio[0]) if isinstance(audio, tuple) else len(audio)
Expand Down