Hardware limitation for RapidOCR

[Iambestfeed]

First of all, I would like to thank you for your contribution to this project.

While using RapidOCR, I wondered if I could limit the number of CPUs used during inference?

I tried looking at the issues as well as reading the Documentation but since I don't know Chinese, I really don't know if there is any way to limit the number of CPUs used? I hope if possible, can you answer me or at least give me a suggestion for a solution?

Thanks.

[SWHL]

You can change the source code as the following:

RapidOCR/python/rapidocr_onnxruntime/utils.py

Lines 31 to 34 in f686541

	sess_opt = SessionOptions()
	sess_opt.log_severity_level = 4
	sess_opt.enable_cpu_mem_arena = False
	sess_opt.graph_optimization_level = GraphOptimizationLevel.ORT_ENABLE_ALL

Add options like below:

sess_opt = SessionOptions()
sess_opt.log_severity_level = 4
sess_opt.enable_cpu_mem_arena = False
sess_opt.graph_optimization_level = GraphOptimizationLevel.ORT_ENABLE_ALL

# If you want to use 6 CPUs, you can set the intra_op_num_threads=6
sess_options.intra_op_num_threads = 6

See details for onnxruntime docs

[Iambestfeed]

Hmmm, I also found support code for onnx but it seems there is not much support code in openvino, right?

[SWHL]

I haven't noticed this before. I just searched and you can refer to this link.
I hope it can help you.

[Iambestfeed]

The good news is that I was able to configure the amount of CPU used (I haven't been able to measure the reduction rate but I see it has decreased significantly and is visible to the naked eye).
Do you think I should commit or just wait until everyone feels the need?

[SWHL]

Welcome PR, you can commit it. I won't merge here for now until everyone feels the need.

[SWHL]

I have already configure the parameter of controling the CPU, as following:

RapidOCR/python/rapidocr_onnxruntime/config.yaml

Lines 10 to 11 in a5a8eb3

	intra_op_num_threads: &intra_nums -1
	inter_op_num_threads: &inter_nums -1