Nice notebook and blog . Thanks for that . Any updates on implementation of avoiding loading of pytorch model alongside ONNX model to get pipeline working . Any pointers would help too.
Thanks for your message. You should consider using the huggingface/optimum package.
It allows one user to convert and use any models on the hub to ONNX. It's a lot easier!
Hi ,
Nice notebook and blog . Thanks for that . Any updates on implementation of avoiding loading of pytorch model alongside ONNX model to get pipeline working . Any pointers would help too.