Is the onnx's IR the same as the TVM's IR?

[calvin886]

Ask a Question

Question

Explain your question here.
I am recently trying to optimize onnx by using TVM(pytorch-> onnx -> TVM), so I am very interested in "Is the onnx's IR the same as the TVM's IR?"

Further information

• Relevant Area (e.g. model usage, backend, best practices, converters, shape_inference, version_converter, training, test, operators):

• Is this issue related to a specific model?
  Model name (e.g. mnist):
  Model opset (e.g. 7):

Notes

Any additional information, code snippets.

[jcwchen]

Hi @calvin886,
Although I don't know TVM well, I believe basically the IRs of TVM and ONNX are different... The main purpose for TVM is optimization and for ONNX is exchange.

[gramalingam]

No, they are the not the same.