wehub-resource-sync
5cbd3f29e3
CodeQL / Analyze (actions) (push) Waiting to run
CodeQL / Analyze (cpp) (push) Waiting to run
CodeQL / Analyze (python) (push) Waiting to run
Copilot Setup Steps / copilot-setup-steps (push) Waiting to run
Generate and publish ONNX docs / build (push) Waiting to run
Generate and publish ONNX docs / deploy (push) Blocked by required conditions
Scorecard supply-chain security / Scorecard analysis (push) Waiting to run
Fuzz / Run fuzz harnesses (${{ github.event_name == 'schedule' && 'nightly' || 'smoke' }}) (push) Has been cancelled
Create Releases / call-mac (push) Has been cancelled
Create Releases / call-linux (push) Has been cancelled
Create Releases / call-sdist (push) Has been cancelled
Create Releases / call-win (push) Has been cancelled
Create Releases / call-pyodide (push) Has been cancelled
Windows_No_Exception_CI / build (x64, 3.10) (push) Has been cancelled
Check URLs / build (push) Has been cancelled
Create Releases / Attest CI build artifacts (push) Has been cancelled
Create Releases / Check for Publish release build to pypi (push) Has been cancelled
Create Releases / Check for Publish preview build to test.pypi-weekly (push) Has been cancelled
Create Releases / Publish preview build to test.pypi-weekly (push) Has been cancelled
Create Releases / Check for Publish release build to test.pypi (rc-candidates) (push) Has been cancelled
Create Releases / Publish release build to test.pypi (push) Has been cancelled
Create Releases / Check for Publish preview build to pypi-weekly (push) Has been cancelled
Create Releases / Publish preview build to pypi-weekly (push) Has been cancelled
Create Releases / Publish release build to pypi (push) Has been cancelled
Create Releases / test source distribution (push) Has been cancelled
clang-tidy / clang-tidy (push) Has been cancelled
Lint / Validate SBOM (push) Has been cancelled
Lint / Enforce style (push) Has been cancelled
CI / Test windows-2022, 3.14, External, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test windows-latest, 3.10, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test windows-latest, 3.14, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test windows-latest, 3.14t, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test ubuntu-24.04, 3.14, Internal, debug=1, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test ubuntu-24.04, 3.14, External, debug=0, unity_build=1, onnx_ml=1, autogen=1 (push) Has been cancelled
CI / Test ubuntu-24.04, 3.14, External, debug=0, unity_build=0, onnx_ml=0, autogen=0 (push) Has been cancelled
CI / Test macos-latest, 3.10, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test macos-latest, 3.14, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test macos-latest, 3.14t, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test ubuntu-24.04, 3.14, External, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test ubuntu-24.04, 3.10, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test ubuntu-24.04, 3.14, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
CI / Test ubuntu-24.04, 3.14t, Internal, debug=0, unity_build=0, onnx_ml=1, autogen=0 (push) Has been cancelled
Pixi CI / Install and lint (ubuntu-24.04-arm) (push) Has been cancelled
Pixi CI / Install and lint (windows-2022) (push) Has been cancelled
Pixi CI / Xcode generator build (push) Has been cancelled
Pixi CI / Install and test (macos-latest, default) (push) Has been cancelled
Pixi CI / Install and test (ubuntu-24.04-arm, default) (push) Has been cancelled
Pixi CI / Install and test (ubuntu-latest, default) (push) Has been cancelled
Pixi CI / Install and test (windows-2022, default) (push) Has been cancelled
Pixi CI / Install and test (macos-latest, oldies) (push) Has been cancelled
Pixi CI / Install and test (ubuntu-24.04-arm, oldies) (push) Has been cancelled
Pixi CI / Install and test (ubuntu-latest, oldies) (push) Has been cancelled
Pixi CI / Install and test (windows-2022, oldies) (push) Has been cancelled
3.1 KiB
3.1 KiB
Broadcasting in ONNX
In ONNX, element-wise operators can take inputs with different shape, as long as the input tensors are broadcastable to the same shape. ONNX supports two types of broadcasting: multidirectional broadcasting and unidirectional broadcasting. We will introduce these two types of broadcasting respectively in the following sections.
Multidirectional Broadcasting
In ONNX, a set of tensors are multidirectional broadcastable to the same shape if one of the following is true:
- The tensors all have exactly the same shape.
- The tensors all have the same number of dimensions and the length of each dimensions is either a common length or 1.
- The tensors that have too few dimensions can have their shapes prepended with a dimension of length 1 to satisfy property 2.
For example, the following tensor shapes are supported by multidirectional broadcasting:
- shape(A) = (2, 3, 4, 5), shape(B) = (,), i.e. B is a scalar ==> shape(result) = (2, 3, 4, 5)
- shape(A) = (2, 3, 4, 5), shape(B) = (5,), ==> shape(result) = (2, 3, 4, 5)
- shape(A) = (4, 5), shape(B) = (2, 3, 4, 5), ==> shape(result) = (2, 3, 4, 5)
- shape(A) = (1, 4, 5), shape(B) = (2, 3, 1, 1), ==> shape(result) = (2, 3, 4, 5)
- shape(A) = (3, 4, 5), shape(B) = (2, 1, 1, 1), ==> shape(result) = (2, 3, 4, 5)
Multidirectional broadcasting is the same as Numpy's broadcasting.
Multidirectional broadcasting is supported by the following operators in ONNX:
Unidirectional Broadcasting
In ONNX, tensor B is unidirectional broadcastable to tensor A if one of the following is true:
- Tensor A and B both have exactly the same shape.
- Tensor A and B all have the same number of dimensions and the length of each dimensions is either a common length or B's length is 1.
- Tensor B has too few dimensions, and B can have its shapes prepended with a dimension of length 1 to satisfy property 2.
When unidirectional broadcasting happens, the output's shape is the same as the shape of A (i.e., the larger shape of two input tensors).
In the following examples, tensor B is unidirectional broadcastable to tensor A:
- shape(A) = (2, 3, 4, 5), shape(B) = (,), i.e. B is a scalar ==> shape(result) = (2, 3, 4, 5)
- shape(A) = (2, 3, 4, 5), shape(B) = (5,), ==> shape(result) = (2, 3, 4, 5)
- shape(A) = (2, 3, 4, 5), shape(B) = (2, 1, 1, 5), ==> shape(result) = (2, 3, 4, 5)
- shape(A) = (2, 3, 4, 5), shape(B) = (1, 3, 1, 5), ==> shape(result) = (2, 3, 4, 5)
Unidirectional broadcasting is supported by the following operators in ONNX: