Xop API#

API#

Automated gathering of operators#

mlprodict.npy.xop.ClassFactory (class_name, op_name, inputs, outputs, input_range, output_range, domain, attr_names, doc, deprecated, since_version, past_version)

Dynamically creates a class for a specific operator.

mlprodict.npy.xop._dynamic_class_creation (operator_names = None, cache = False, include_past = False, verbose = 0, fLOG = <built-in function print>)

Automatically generates classes for each of the operators module onnx defines and described at Operators and Operators.

mlprodict.npy.xop._GraphBuilder (self)

Graph builder. It takes a graph structure made with instances of OnnxOperatorBase. The main method is to_onnx.

  • initializer: list of initializers to add to the ONNX graph

  • node: list of nodes to add to the ONNX graph

  • input: list of inputs to add to the ONNX graph

  • output: list of inputs to add to the ONNX graph

  • opsets: opsets of the ONNX graph

  • input_names: dictionary of input names

    {name: InputDetectedVariable}

  • node_output_names: memorizes a name for a node output

    when the user did not specify any {(id(node), index): OutputDetectedVariable}

  • reserved_names: dictionary { name : (node, index) },

    name which should remain unchanged in the ONNX graph

  • names: list of uniques names

  • functions: dictionary { domain, name: function_proto }

  • function_hashes: dictionary { domain, name: hash of function_proto }

Main classes#

mlprodict.npy.xop_variable.Variable (self, name, dtype = None, shape = None, added_dtype = None, added_shape = None)

An input or output to an ONNX graph.

mlprodict.npy.xop.OnnxOperator (self, inputs, op_version = None, output_names = None, domain = None, global_context = None, kwargs)

Ancestor to every ONNX operator exposed in mlprodict.npy.xops and mlprodict.npy.xops_ml.

mlprodict.npy.xop.OnnxOperatorItem (self, onx_op, index, op_version = None)

Accessor to one of the output returned by a OnnxOperator.

mlprodict.npy.xop_convert.OnnxSubOnnx (self, model, inputs, output_names = None)

This operator is used to insert existing ONNX into the ONNX graph being built.

mlprodict.npy.xop_convert.OnnxSubEstimator (self, model, inputs, op_version = None, output_names = None, options = None, initial_types = None, kwargs)

This operator is used to call the converter of a model to insert the node coming from the conversion into a bigger ONNX graph. It supports model from scikit-learn using sklearn-onnx.

Helpers to handle API changing with opsets#

mlprodict.npy.xop_opset.OnnxReduceSumApi11 (x, axes = None, keepdims = 1, op_version = None, output_names = None)

Adds operator ReduceSum with opset>=13 following API from opset 12.

mlprodict.npy.xop_opset.OnnxSplitApi11 (x, axis = 0, split = None, op_version = None, output_names = None)

Adds operator Split with opset>=13 following API from opset 11.

mlprodict.npy.xop_opset.OnnxSqueezeApi11 (x, axes = None, op_version = None, output_names = None)

Adds operator Squeeze with opset>=13 following API from opset 11.

mlprodict.npy.xop_opset.OnnxUnsqueezeApi11 (x, axes = None, op_version = None, output_names = None)

Adds operator Unsqueeze with opset>=13 following API from opset 11.

mlprodict.npy.xop_opset.OnnxReduceL2_typed (dtype, x, axes = None, keepdims = 1, op_version = None, output_names = None)

Adds operator ReduceL2 for float or double.

mlprodict.npy.xop_opset.OnnxReshapeApi13 (x, allowzero = 0, op_version = None, output_names = None)

Adds operator Reshape with opset>=14 following API from opset 13.

Available ONNX operators#

previous

Complete Numpy API for ONNX

next

Supported ONNX operators