A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 9, 00:00 UTC
[Bugfix] fix qwen3-asr response error (#33644)
Signed-off-by: jesse <szxfml@gmail.com> Signed-off-by: Cyrus Leung <cyrus.tl.leung@gmail.com> Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Song Zhixin committed
Feb 3, 2026 at 19:33 UTC
ceab70c89d2b1f5eeaeb4582eb927b16dacb7671
1 file changed
+7
-6
vllm/model_executor/models/qwen3_asr.py
+7
-6
@@ -125,6 +125,13 @@ class Qwen3ASRProcessingInfo(BaseProcessingInfo):
125
def get_supported_mm_limits(self) -> Mapping[str, int | None]:
126
return {"audio": None}
127
128
+ def get_data_parser(self) -> MultiModalDataParser:
129
+ feature_extractor = self.get_feature_extractor()
130
+ return Qwen3ASRMultiModalDataParser(
131
+ target_sr=feature_extractor.sampling_rate,
132
+ expected_hidden_size=self._get_expected_hidden_size(),
133
+ )
134
+
135
136
class Qwen3ASRDummyInputsBuilder(BaseDummyInputsBuilder[Qwen3ASRProcessingInfo]):
137
def get_dummy_text(self, mm_counts: Mapping[str, int]) -> str:
@@ -194,12 +201,6 @@ class Qwen3ASRMultiModalDataParser(MultiModalDataParser):
201
class Qwen3ASRMultiModalProcessor(
202
Qwen3OmniMoeThinkerMultiModalProcessor,
203
):
197
- def _get_data_parser(self) -> MultiModalDataParser:
198
- feature_extractor = self.info.get_feature_extractor()
199
- return Qwen3ASRMultiModalDataParser(
200
- target_sr=feature_extractor.sampling_rate,
201
- )
202
-
204
def _get_mm_fields_config(
205
self,
206
hf_inputs: BatchFeature,