module onnx_tools.compress#

Short summary#

module mlprodict.onnx_tools.compress

Functions to simplify, compress an ONNX graph.

Functions#

function

truncated documentation

_check_expression

_compress_nodes

Compresses a sequence of node to make it more readable. If possible, it creates a node Expression with a graph …

_compress_nodes_once

Compresses a sequence of node to make it more readable. If possible, it creates a node Expression with a graph …

_fuse_node

Merges two nodes having one input/output in common.

compress_proto

Compresses a ModelProto, FunctionProto, :epkg:`GraphProto`. The function detects nodes outputting …

Documentation#

Functions to simplify, compress an ONNX graph.

New in version 0.9.

source on GitHub

mlprodict.onnx_tools.compress._check_expression(expe)#
mlprodict.onnx_tools.compress._compress_nodes(nodes, verbose=0)#

Compresses a sequence of node to make it more readable. If possible, it creates a node Expression with a graph as an attribute.

Parameters:

nodes – sequence of nodes to compress

Returns:

compressed sequence of nodes

source on GitHub

mlprodict.onnx_tools.compress._compress_nodes_once(nodes, verbose=0)#

Compresses a sequence of node to make it more readable. If possible, it creates a node Expression with a graph as an attribute.

Parameters:

nodes – sequence of nodes to compress

Returns:

compressed sequence of nodes

source on GitHub

mlprodict.onnx_tools.compress._fuse_node(o, node, node_next)#

Merges two nodes having one input/output in common.

Parameters:

  • o – output name

  • node – first node (it outputs the results)

  • node_next – second node (it ingests the result)

Returns:

merged node

source on GitHub

mlprodict.onnx_tools.compress.compress_proto(proto, verbose=0)#

Compresses a ModelProto, FunctionProto, :epkg:`GraphProto`. The function detects nodes outputting results only used once. It then fuses it with the node taking it as an input.

Parameters:
Returns:

same type

New in version 0.9.

source on GitHub