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

[VLM] Remove BaseProcessingInfo.get_mm_max_tokens_per_item (#16408)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-04-11 00:06:58 +08:00
committed by GitHub
parent 7678fcd5b6
commit 83b824c8b4
39 changed files with 104 additions and 677 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
vllm/model_executor/models/whisper.py
View File

@@ -538,16 +538,9 @@ class WhisperProcessingInfo(BaseProcessingInfo):
assert isinstance(feature_extractor, WhisperFeatureExtractor)
return feature_extractor
def get_max_audio_tokens(self) -> int:
def get_num_audio_tokens(self) -> int:
return self.get_hf_config().max_source_positions
def get_mm_max_tokens_per_item(
self,
seq_len: int,
mm_counts: Mapping[str, int],
) -> Mapping[str, int]:
return {"audio": self.get_max_audio_tokens()}
class WhisperDummyInputsBuilder(BaseDummyInputsBuilder[WhisperProcessingInfo]):
@@ -630,7 +623,7 @@ class WhisperMultiModalProcessor(
hf_processor_mm_kwargs: Mapping[str, object],
out_mm_kwargs: MultiModalKwargs,
) -> Sequence[PromptUpdate]:
num_tokens = self.info.get_max_audio_tokens()
num_tokens = self.info.get_num_audio_tokens()
return [
PromptReplacement(
modality="audio",