[Bugfix] Allow token ID-only inputs in Qwen2-Audio (vllm-project#10536)

Signed-off-by: DarkLight1337 <[email protected]>
charlifu · Nov 21, 2024 · e7a8341 · e7a8341
1 parent c51e397
commit e7a8341
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/vllm/model_executor/models/qwen2_audio.py b/vllm/model_executor/models/qwen2_audio.py
@@ -212,7 +212,7 @@ def input_processor_for_qwen2_audio(
 
     return token_inputs(
         prompt_token_ids=new_input_ids,
-        prompt=inputs['prompt'],
+        prompt=inputs.get("prompt"),
         multi_modal_data=multi_modal_data,
     )