The easy caseΒΆ
The easy case is when the machine learned model can be converter into ONNX with a converting library without writing any specific code. That means that a converter exists for the model or each piece of the model, the converter produces an ONNX graph where every node is part of the existing ONNX specifications, and the runtime used to compute the predictions implements every node used in the ONNX graph.
- Train and deploy a scikit-learn pipeline
- Benchmark ONNX conversion
- What is the opset number?
- One model, many possible conversions with options
- Choose appropriate output of a classifier
- Black list operators when converting
- Issues when switching to float
- Intermediate results and investigation
- Store arrays in one onnx graph
- Dataframe as an input
- Modify the ONNX graph