xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2025-12-17 02:45:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Misc] Rename Audios -> Audio in Qwen2audio Processing (#17507)

Browse Source 
Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
This commit is contained in:
Alex Brooks 2025-04-30 20:51:36 -06:00 committed by GitHub
parent 200bbf92e8
commit 02bd654846
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 8 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
vllm/model_executor/models/qwen2_audio.py
View File

@ -150,8 +150,15 @@ class Qwen2AudioMultiModalProcessor(
mm_data: Mapping[str, object], mm_data: Mapping[str, object],
mm_kwargs: Mapping[str, Any], mm_kwargs: Mapping[str, Any],
) -> BatchFeature: ) -> BatchFeature:
# NOTE - we rename audios -> audio in mm data because transformers has
# deprecated audios for the qwen2audio processor and will remove
# support for it in transformers 4.54.
audios = mm_data.pop("audios", [])
if audios:
mm_data["audio"] = audios
# Text-only input not supported in composite processor # Text-only input not supported in composite processor
if not mm_data.get("audios", []): if not mm_data.get("audio", []):
prompt_ids = self.info.get_tokenizer().encode(prompt) prompt_ids = self.info.get_tokenizer().encode(prompt)
prompt_ids = self._apply_hf_processor_tokens_only(prompt_ids) prompt_ids = self._apply_hf_processor_tokens_only(prompt_ids)
return BatchFeature(dict(input_ids=[prompt_ids]), tensor_type="pt") return BatchFeature(dict(input_ids=[prompt_ids]), tensor_type="pt")