From da7f25e14b1a504d68b9729121ea95fda8a9c4fa Mon Sep 17 00:00:00 2001 From: Peiqi Yin <60515999+yinpeiqi@users.noreply.github.com> Date: Wed, 24 Dec 2025 18:43:00 +0800 Subject: [PATCH] [BugFix] Qwen3-Omni / Qwen2.5-Omni thinker batching with audio input (#451) Signed-off-by: yinpeiqi --- .../model_executor/models/qwen2_5_omni/qwen2_5_omni_thinker.py | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) diff --git a/vllm_omni/model_executor/models/qwen2_5_omni/qwen2_5_omni_thinker.py b/vllm_omni/model_executor/models/qwen2_5_omni/qwen2_5_omni_thinker.py index 78737ebf3..71c5e8377 100644 --- a/vllm_omni/model_executor/models/qwen2_5_omni/qwen2_5_omni_thinker.py +++ b/vllm_omni/model_executor/models/qwen2_5_omni/qwen2_5_omni_thinker.py @@ -90,7 +90,8 @@ class Qwen2_5OmniConditionalGenerationMixin(Qwen2_5OmniConditionalGenerationMixi ): feature_attention_mask = feature_attention_mask.reshape(-1, feature_attention_mask.shape[-1]) elif feature_attention_mask is not None and isinstance(feature_attention_mask, list): - feature_attention_mask = torch.cat(feature_attention_mask, dim=-1) + for i in range(len(feature_attention_mask)): + feature_attention_mask[i] = feature_attention_mask[i].reshape(-1) return Qwen2_5OmniAudioFeatureInputs( type="audio_features", input_features=input_audio_features,