Help Needed: Configuring ONNX in PaddleOCR (onnx_sess_options = None)

[AnshuTrivedi]

I'm trying to run ONNX inference with PaddleOCR and I could use some help understanding how to properly set and use ONNX-related configurations.

Here is what I currently have in my config:
use_onnx = True
onnx_sess_options = None
onnx_providers = ['CPUExecutionProvider']

I have two main questions:

What are the available options for onnx_sess_options?
I'd like to understand what session options I can configure (e.g. thread settings, optimization levels, memory options), and what each of them does.

How do I properly pass onnx_sess_options if I want to customize performance settings?
For example, if I want to increase performance on CPU, what values should I pass in onnx_sess_options, and in what format?

Any examples of how to structure the config or set these options in PaddleOCR's Python scripts would be really appreciated!

Thanks in advance 🙏

[cuicheng01]

Hello, it is recommended to use the high-performance inference in PaddleOCR 3.0, which supports ONNX as a backend. Here is the documentation.