linux上用flask部署调用的时候，多次调用模型就卡住了

[Blue-FatMan]

Notice: In order to resolve issues more efficiently, please raise issue following the template.
（注意：为了更加高效率解决您遇到的问题，请按照模板提问，补充细节）

❓ Questions and Help

linux上多次调用模型就卡住了,卡在0%或者需要很久几分钟才能出来
模型文件是这里面下载的https://modelscope.cn/models/iic/SenseVoiceSmall/files

Before asking:

1. search the issues.
2. search the docs.

What is your question?

Code

MODEL = AutoModel(
model=model_dir,
# vad_model="fsmn-vad",
# vad_kwargs={"max_single_segment_time": 30000},
# device="cuda:0",
disable_update=True,
batch_size_s=60,
merge_vad=True, #
merge_length_s=15,
hub="hf",
)

res = MODEL.generate(
    input=audio_path,
    cache={},
    language="auto",  # "zn", "en", "yue", "ja", "ko", "nospeech"
    use_itn=True,
    batch_size_s=60,
    merge_vad=True,  #
    merge_length_s=15,
)

What have you tried?

What's your environment?

• OS (e.g., Linux): Ubuntu 18.04.1 LTS
• FunASR Version (e.g., 1.0.0): 1.2.6
• ModelScope Version (e.g.,1.11.0):
• PyTorch Version (e.g., 2.0.0): 2.6.0
• How you installed funasr (pip, source): https://pypi.tuna.tsinghua.edu.cn/packages
• Python version: 3.10
• GPU (e.g., V100M32)
• CUDA/cuDNN version (e.g., cuda11.7):
• Docker version (e.g., funasr-runtime-sdk-cpu-0.4.1)
• Any other relevant information:
  -- 系统资源： 2核8G