What is ONNX Runtime (ORT)?

[@seikatsu_ki]

It would be best if you share the code source of the demo (GitHub repo for instance). Thank you!

[@somaiahphotography2134]

Hello, I'm training an image classifier, so the input type is a numpy.ndarray of shape (784,), how do I include that array in my initial_type? Thanks :)

[@0730pleomax]

Is there a specific reason to run onnx.load(model) and onnx.checker.check_mode(onnx_model)? Since we actually run the inference by ort_sess.

[@ONNXRuntime]

No, that step is not needed. It's an optional step to check your model export.

[@gianpaj]

why would I use ONNX? why does it exist?

[@unlomtrash]

This format allows you to convert your model to any backend you want. For example you have written a model using pytorch, then exported it as onnx, then you could convert onnx to tensorflow etc