A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 16, 07:00 UTC
[Bugfix] Allow token ID-only inputs in Qwen2-Audio (#10536)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Cyrus Leung committed
Nov 22, 2024 at 02:09 UTC
e7a8341c7c7481a0c797d50ead7a698255ac8a9f
1 file changed
+1
-1
vllm/model_executor/models/qwen2_audio.py
+1
-1
@@ -212,7 +212,7 @@ def input_processor_for_qwen2_audio(
212
213
return token_inputs(
214
prompt_token_ids=new_input_ids,
215
- prompt=inputs['prompt'],
215
+ prompt=inputs.get("prompt"),
216
multi_modal_data=multi_modal_data,
217
)
218