Module Reference¶

Export¶

ONNX-Chainer exports Chainer model to ONNX graph with various options.

`onnx_chainer.export`	Export function for chainer.Chain in ONNX format.
`onnx_chainer.export_testcase`	Export model and I/O tensors of the model in protobuf format.

ONNX-Chainer provides some utility functions to help exporting.

`onnx_chainer.replace_func.fake_as_funcnode`	The target function fakes FunctionNode
`onnx_chainer.replace_func.as_funcnode`	The target function fakes FunctionNode

These utilities helps converting from Chainer model to ONNX format, mainly used them internally.

onnx_chainer.context.Context Context of converter

`onnx_chainer.onnx_helper.GraphBuilder`	A helper class to build consecutive ONNX nodes.
`onnx_chainer.onnx_helper.set_func_name`	Set the name of Chainer function being converted.
`onnx_chainer.onnx_helper.get_func_name`	Return processing function name
`onnx_chainer.onnx_helper.make_node`	A thin wrapper of onnx.helper.make_node.
`onnx_chainer.onnx_helper.write_tensor_pb`
`onnx_chainer.onnx_helper.cleanse_param_name`	Converts Chainer parameter names to ONNX names.

`onnx_chainer.testing.input_generator.increasing`	Returns a monotonically increasing ndarray for test inputs.
`onnx_chainer.testing.input_generator.nonzero_increasing`	Returns a monotonically increasing ndarray for test inputs.
`onnx_chainer.testing.input_generator.positive_increasing`	Returns a monotonically increasing ndarray for test inputs.