Exporter

Vortex includes onnx exporter to ease exporting model and embedding model properties.

class vortex.development.exporter.onnx.ONNXExporter(dataset=None, batch_size=1, shape_inference=True, opset_version: int = 11)

Vortex onnx example, handle onnx export for vortex.development.networks.models.ModelBase

__call__(model, filename, shape_inference=None, **kwargs)

Export given model and save to filename

Parameters

  • model – model to be exported

  • filename (str) – desired output filename

  • shape_inference (bool, optional) – enable or disable shape inference. Defaults to None.

export(model, filename, **kwargs)

First part of exporter calls, will do the following: set model on eval mode, calls model’s on_export_start, read input_names and output_names from model, export the model,

Parameters

  • model – model to be exported

  • filename (str) – [description]

embed_properties(model, filename, shape_inference=None)

Second part of export calls, embed required model’s properties to exported model

Parameters

  • model – model in question

  • filename (str) – exported model filename

  • shape_inference (bool, optional) – enable or disable shape inference. Defaults to None.

finalize_export(model, filename)

Final part of export calls, call model.on_export_end

Parameters

  • model – model in question

  • filename (str) – exported model filename