memory leak in funasr-runtime-sdk-cpu-0.4.4

[FD-Liekkas]

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）

🐛 Bug

1. Use docker Version: registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.4
2. Run /workspace/FunASR/runtime/run_server.sh as referred to https://github.com/alibaba-damo-academy/FunASR/blob/main/runtime/docs/SDK_advanced_guide_offline_zh.md#python-client.
3. Execute python3.8 funasr_wss_client.py --host 127.0.0.1 --port 10095 --ssl 1 --mode offline --audio_in "xxx.mp4". Here, "xxx.mp4" is a long video, more than 30 minutes in duration. After step 3, the funasr-wss-server process's memory usage increases with each audio recognition. This increase seems to be related to the size of the audio.

To Reproduce

Steps to reproduce the behavior (always include the command you ran):

1. Run cmd in my macOS terminal

   sudo docker pull registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.4
   mkdir -p ./funasr-runtime-resources/models
   sudo docker run -p 10095:10095 -it --privileged=true -v $PWD/funasr-runtime-resources/models:/workspace/models registry.cn-hangzhou.aliyuncs.com/funasr_repo/funasr:funasr-runtime-sdk-cpu-0.4.4
   

2. Enter contanier from terminal

   docker exec -it contanierId sh
   

3. Lanuch server

   cd /workspace/FunASR/runtime
   nohup bash run_server.sh > log.txt 2>&1 &
   

4. Query by python client. "xxx.mp4" is a long video, more than 30 minutes in duration

   cd /workspace/FunASR/runtime/python/websocket/
   python3.8 funasr_wss_client.py --host 127.0.0.1 --port 10095 --ssl 1 --mode offline --audio_in "xxx.mp4"
   

The funasr-wss-server process's memory usage increases with each audio recognition. Using the "xxx.wav" audio file for over 30 minutes will result in a similar reault. You can see memory usage with top command.

Code sample

No

Expected behavior

Solving the memory leak issue in the FunASR Offline File Transcription Software Package.

Environment

• OS (e.g., Linux): MacOS
• FunASR Version (e.g., 1.0.0): 1.0.13, also have tried 1.0.25
• ModelScope Version (e.g., 1.11.0): 1.12.0
• PyTorch Version (e.g., 2.0.0): 2.2.1
• How you installed funasr (pip, source): installed in official image
• Python version: 3.8.18
• GPU (e.g., V100M32): use cpu
• CUDA/cuDNN version (e.g., cuda11.7): use cpu
• Docker version (e.g., funasr-runtime-sdk-cpu-0.4.1): funasr-runtime-sdk-cpu-0.4.4
• Any other relevant information: No

Additional context

No

[hoavt-54]

We're having this on GPU as well. Memory keeps increasing after each inference.