遇到cpu环境下内存不释放 ，每次推理内存渐增长但不释放

[501956430]

app_onnx 启动占用 700m左右内存，随着运行期间推理，可以占用8g左右内存，每次api请求推理内存都会逐渐上升，推理完成内容也不见下降

[alpacaking]

You can refer to the workaround in [#61](#61 (comment)):

Set a fixed number of CPU threads when launching:

python app_onnx.py --cpu-threads 4

Meanwhile, make sure that every request from the UI/API explicitly uses the same number of threads:

cpu_threads=4