[Bugfix] Standardize merging multimodal embeddings (#26771)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
2025-10-14 17:36:21 +08:00
parent 577d498212
commit d2f816d6ff
19 changed files with 57 additions and 57 deletions
--- a/vllm/model_executor/models/minicpmo.py
+++ b/vllm/model_executor/models/minicpmo.py
@@ -762,7 +762,7 @@ class MiniCPMO(MiniCPMV2_6):
        for modality in modalities:
            if modality == "audios":
                audio_input = modalities["audios"]
-                audio_features = self._process_audio_input(audio_input)
-                multimodal_embeddings += tuple(audio_features)
+                audio_embeddings = self._process_audio_input(audio_input)
+                multimodal_embeddings += tuple(audio_embeddings)

        return multimodal_embeddings