Can I wrap a PyTorch model into ONNX together with tokenizers?
This is something that worked trivially in tf but in torch strings are not supported natively at all. I tried adding mapping nodes manually to the resulting onnx model but I’m getting all kinds of compilers errors.