Add error handler

Update README.md

Update multiprocess_inference.py

Update performance metrics

Enable RKNN2 internal multicore tensor parallel function, speed up inference latency ~1.5s.

Update dependency version in document

Update model path

Update rkllm_binding.py

They finally fixed this stupid bug...

Upload 3 files

update model path

Update README.md

Update README.md

Upload 23 files

initial commit