onnx_extended.tools.onnx_nodes

convert_onnx_model

onnx_extended.tools.onnx_nodes.convert_onnx_model(onnx_model: ModelProto | GraphProto | NodeProto | FunctionProto, opsets: Dict[str, int], recursive: bool = True, use_as_tensor_attributes: bool = True, verbose: int = 0, _from_opset: Dict[str, int] | None = None, debug_info: List[str] | None = None) ModelProto | GraphProto | NodeProto | FunctionProto[source]

Upgrades a model to the latest opsets.

Parameters:
  • onnx_model – proto

  • opsets – list of opsets to update

  • recursive – looks into subgraphs

  • use_as_tensor_attributes – use attributes siffixed with as_tensor for trees

  • verbose – verbosity

  • _from_opset – tells which opset a node belongs too, only used when onnx_model is a NodeProto

  • debug_info – unused

Returns:

new proto

enumerate_onnx_node_types

onnx_extended.tools.onnx_nodes.enumerate_onnx_node_types(model: str | ModelProto | GraphProto, level: int = 0, shapes: Dict[str, TypeProto] | None = None, external: bool = True) Iterable[Dict[str, str | float]][source]

Looks into types for every node in a model.

Parameters:
  • model – a string or a proto

  • level – level (recursivity level)

  • shapes – known shapes, returned by :func:onnx.shape_inference.infer_shapes`

  • external – loads the external data if the model is loaded

Returns:

a list of dictionary which can be turned into a dataframe.

get_hidden_inputs

onnx_extended.tools.onnx_nodes.get_hidden_inputs(nodes: Iterable[NodeProto]) Set[str][source]

Returns the list of hidden inputs used by subgraphs.

Parameters:

nodes – list of nodes

Returns:

list of names

multiply_tree

onnx_extended.tools.onnx_nodes.multiply_tree(node: NodeProto, n: int, random: bool = True) NodeProto[source]

Multiplies the number of trees in TreeEnsemble operator. It replicates the existing trees but permutes features ids and node values if random is True.

Parameters:
  • node – tree ensemble operator

  • n – number of times the existing trees must be multiplied

  • random – permutation or thresholds

Returns:

the new trees

onnx_merge_models

onnx_extended.tools.onnx_nodes.onnx_merge_models(m1: ModelProto, m2: ModelProto, io_map: List[Tuple[str, str]], verbose: int = 0) ModelProto[source]

Merges two models. The functions also checks that the model have the same defined opsets (except for function). If not, the most recent opset is selected.

Parameters:
  • m1 – first model

  • m2 – second model

  • io_map – mapping between outputs of the first model and and the input of the second one

  • verbose – display some information if one of the model was updated

Returns:

new model

onnx_remove_node_unused

onnx_extended.tools.onnx_nodes.onnx_remove_node_unused(onnx_model, recursive=True, debug_info=None, **options)[source]

Removes unused nodes of the graph. An unused node is not involved in the output computation.

Parameters:
  • onnx_model – onnx model

  • recursive – looks into subgraphs

  • debug_info – debug information (private)

  • options – unused

Returns:

new onnx _model

select_model_inputs_outputs

onnx_extended.tools.onnx_nodes.select_model_inputs_outputs(model: ModelProto, outputs: List[str] | None = None, inputs: List[str] | None = None, infer_shapes: bool = True, overwrite: Dict[str, Any] | None = None, remove_unused: bool = True, verbose: int = 0)[source]

Takes a model and changes its outputs.

Parameters:
  • modelONNX model

  • inputs – new inputs, same ones if None

  • outputs – new outputs, same ones if None

  • infer_shapes – infer inputs and outputs shapes

  • overwrite – overwrite type and shapes for inputs or outputs, overwrite is a dictionary {‘name’: (numpy dtype, shape)}

  • remove_unused – remove unused nodes from the graph

  • verbose – display information while converting

Returns:

modified model

The function removes unneeded nodes.

The following example shows how to change the inputs of model to bypass the first nodes. Shape inferences fails to determine the new inputs type. They need to be overwritten. verbose=1 shows the number of deleted nodes.

import onnx
from onnx_extended.tools.onnx_nodes import select_model_inputs_outputs

onx = onnx.load(path)
onx2 = select_model_inputs_outputs(
    onx, inputs=["a", "b"],
    infer_shapes=True, verbose=1,
    overwrite={'a': (numpy.int32, None), 'b': (numpy.int64, None)})
onnx.save(onx2, path2)