Description
直接使用Qwen3ASRModel.from_pretrained加载并识别时为正确内容，但是通过qwen-asr-serve Qwen/Qwen3-ASR-1.7B --gpu-memory-utilization 0.8 --host 0.0.0.0 --port 8000部署的服务推理出现了重复的 “嗯”，数百个
Reproduction
直接推理：
from qwen_asr import Qwen3ASRModel
model = Qwen3ASRModel.from_pretrained(
"Qwen/Qwen3-ASR-1.7B",
dtype=torch.bfloat16,
device_map="cuda:0",
# attn_implementation="flash_attention_2",
max_inference_batch_size=32, # Batch size limit for inference. -1 means unlimited. Smaller values can help avoid OOM.
max_new_tokens=256, # Maximum number of tokens to generate. Set a larger value for long audio input.
)
results = model.transcribe(
audio=audio_url,
language=None, # set "English" to force the language
)
return results[0].text
加速服务：
qwen-asr-serve Qwen3-ASR-1.7B --gpu-memory-utilization 0.8 --host 0.0.0.0 --port 8000
client = OpenAI(base_url=“http://localhost:8000/v1”, api_key="EMPTY")
response = client.chat.completions.create(
model=ASR_MODEL,
messages=[
{
"role": "user",
"content": [{"type": "audio_url", "audio_url": {"url": audio_url}}],
}
],
)
return clean_asr_text(response.choices[0].message.content)
Logs
Environment Information
GPU：H20 96G
CUDA: 13.0
ubuntu 22.4
Known Issue
Description
直接使用Qwen3ASRModel.from_pretrained加载并识别时为正确内容，但是通过qwen-asr-serve Qwen/Qwen3-ASR-1.7B --gpu-memory-utilization 0.8 --host 0.0.0.0 --port 8000部署的服务推理出现了重复的 “嗯”，数百个
Reproduction
直接推理：
from qwen_asr import Qwen3ASRModel
model = Qwen3ASRModel.from_pretrained(
"Qwen/Qwen3-ASR-1.7B",
dtype=torch.bfloat16,
device_map="cuda:0",
# attn_implementation="flash_attention_2",
max_inference_batch_size=32, # Batch size limit for inference. -1 means unlimited. Smaller values can help avoid OOM.
max_new_tokens=256, # Maximum number of tokens to generate. Set a larger value for long audio input.
)
results = model.transcribe(
audio=audio_url,
language=None, # set "English" to force the language
)
return results[0].text
加速服务：
qwen-asr-serve Qwen3-ASR-1.7B --gpu-memory-utilization 0.8 --host 0.0.0.0 --port 8000
client = OpenAI(base_url=“http://localhost:8000/v1”, api_key="EMPTY")
response = client.chat.completions.create(
model=ASR_MODEL,
messages=[
{
"role": "user",
"content": [{"type": "audio_url", "audio_url": {"url": audio_url}}],
}
],
)
return clean_asr_text(response.choices[0].message.content)
Logs
Environment Information
GPU：H20 96G
CUDA: 13.0
ubuntu 22.4
Known Issue