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

[Model] Apply #32631 for recent models (#33785)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-02-04 20:23:01 +08:00
committed by GitHub
parent f8516a1ab9
commit e57ef99b40
4 changed files with 46 additions and 50 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
vllm/model_executor/models/qwen3_asr.py
View File

@@ -296,19 +296,21 @@ class Qwen3ASRForConditionalGeneration(
multimodal_config = vllm_config.model_config.multimodal_config
self.config = thinker_config
self.multimodal_config = multimodal_config
self.audio_tower = Qwen3OmniMoeAudioEncoder(
thinker_config.audio_config,
prefix=maybe_prefix(prefix, "audio_tower"),
)
self.quant_config = quant_config
self.language_model = Qwen3ForCausalLM(
vllm_config=vllm_config.with_hf_config(
thinker_config.text_config, architectures=["Qwen3ForCausalLM"]
),
prefix=maybe_prefix(prefix, "language_model"),
)
with self._mark_tower_model(vllm_config, "audio"):
self.audio_tower = Qwen3OmniMoeAudioEncoder(
thinker_config.audio_config,
prefix=maybe_prefix(prefix, "audio_tower"),
)
with self._mark_language_model(vllm_config):
self.language_model = Qwen3ForCausalLM(
vllm_config=vllm_config.with_hf_config(
thinker_config.text_config, architectures=["Qwen3ForCausalLM"]
),
prefix=maybe_prefix(prefix, "language_model"),
)
self.make_empty_intermediate_tensors = (
self.language_model.make_empty_intermediate_tensors
@@ -363,9 +365,6 @@ class Qwen3ASRForConditionalGeneration(
)
return audio_features.split(audio_output_lengths.tolist())
def get_language_model(self) -> torch.nn.Module:
return self.language_model
def embed_multimodal(self, **kwargs: object) -> MultiModalEmbeddings | None:
mm_input_by_modality = self._parse_and_validate_multimodal_inputs(**kwargs)
if not mm_input_by_modality: