Bug - need to modify code to run on onnx cuda

[montvid]

llm output:

Fix for future RapidVideOCR GPU runs — the correct param key is EngineConfig.onnxruntime.use_cuda (not Det.use_cuda), since engine_cfg
is set from EngineConfig after params are merged.

[SWHL]

Could you please elaborate further?

[montvid]

my RapidVideOCR local debian 12 setup with onnx ran with cpu but not cuda. Fixed with a llm model but don't have the diff. Just a heads up that onnx did not work for me.

I can close it as the chinese ocr from video frames with RapidVideOCR does not satisfy me one of the reasons frame detection is quite bad in my case.