npx.npx_functions

onnx_array_api.npx.npx_functions.abs(*inputs, **kwargs)

See numpy.absolute().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.absolute(*inputs, **kwargs)

See numpy.absolute().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.arccos(*inputs, **kwargs)

See numpy.arccos().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.arccosh(*inputs, **kwargs)

See numpy.arccosh().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.amax(*inputs, **kwargs)

See numpy.amax().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: OptParType[int],
    keepdims: OptParType[int],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.amin(*inputs, **kwargs)

See numpy.amin().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: OptParType[int],
    keepdims: OptParType[int],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.arange(*inputs, **kwargs)

See numpy.arange().

Signature:

(
    start_or_stop: TensorType[{'int16', 'int32', 'float64', 'int64', 'float32'}, (1,), 'I'],
    stop_or_step: OptTensorType[{'int16', 'int32', 'float64', 'int64', 'float32'}, (1,), 'I'],
    step: OptTensorType[{'int16', 'int32', 'float64', 'int64', 'float32'}, (1,), 'I'],
    dtype: OptParType[DType],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.argmax(*inputs, **kwargs)

See numpy.amax().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: OptParType[int],
    keepdims: OptParType[int],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.argmin(*inputs, **kwargs)

See numpy.argmin().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: OptParType[int],
    keepdims: OptParType[int],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.arcsin(*inputs, **kwargs)

See numpy.arcsin().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.arcsinh(*inputs, **kwargs)

See numpy.arcsinh().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.arctan(*inputs, **kwargs)

See numpy.arctan().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.arctanh(*inputs, **kwargs)

See numpy.arctanh().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.cdist(*inputs, **kwargs)

See scipy.special.distance.cdist().

Signature:

(
    xa: TensorType[numerics, 'T'],
    xb: TensorType[numerics, 'T'],
    metric: OptParType[str],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.ceil(*inputs, **kwargs)

See numpy.ceil().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.clip(*inputs, **kwargs)

See numpy.clip().

Signature:

(
    x: TensorType[numerics, 'T'],
    a_min: TensorType[numerics, 'T'],
    a_max: TensorType[numerics, 'T'],
):

onnx_array_api.npx.npx_functions.compress(*inputs, **kwargs)

See numpy.compress(). np.compress(condition, x) or npnx.compress(x, condition).

Signature:

(
    condition: TensorType['bool_', 'B'],
    x: TensorType[numerics, 'T'],
    axis: OptParType[int],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.compute(*inputs, **kwargs)

Executes an onnx proto.

param x:

inputs

param proto:

proto to execute

param name:

model name

return:

outputs

Signature:

(
    x: SequenceType["TensorType[numerics, 'T']"],
    proto: ParType[typing.Union[onnx.onnx_ml_pb2.FunctionProto, onnx.onnx_ml_pb2.ModelProto, onnx.onnx_ml_pb2.NodeProto]],
    name: ParType[str],
) -> TupleType[TensorType[numerics, 'T']]:

onnx_array_api.npx.npx_functions.concat(*inputs, **kwargs)

Operator concat, handle numpy.vstack() and numpy.hstack().

Signature:

(
    x: SequenceType["TensorType[numerics, 'T']"],
    axis: ParType[int],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.cos(*inputs, **kwargs)

See numpy.cos().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.cosh(*inputs, **kwargs)

See numpy.cosh().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.cumsum(*inputs, **kwargs)

See numpy.cumsum().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: OptTensorType['int64', 'I'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.det(*inputs, **kwargs)

See numpy.linalg:det().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.dot(*inputs, **kwargs)

See numpy.dot() dot is equivalent to npx.matmul == np.matmul != np.dot with arrays with more than 3D dimensions.

Signature:

(
    a: TensorType[numerics, 'T'],
    b: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.einsum(*inputs, **kwargs)

See numpy.einsum().

Signature:

(
    x: SequenceType["TensorType[numerics, 'T']"],
    equation: ParType[str],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.erf(*inputs, **kwargs)

See scipy.special.erf().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.exp(*inputs, **kwargs)

See numpy.exp().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.expand_dims(*inputs, **kwargs)

See numpy.expand_dims().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: TensorType['int64', 'I'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.expit(*inputs, **kwargs)

See scipy.special.expit().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.floor(*inputs, **kwargs)

See numpy.floor().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.hstack(*inputs, **kwargs)

See numpy.hstack().

Signature:

(
    x: SequenceType["TensorType[numerics, 'T']"],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.copy(*inputs, **kwargs)

Makes a copy.

Signature:

(
    x: TensorType[allowed, 'T'],
) -> TensorType[allowed, 'T']:

onnx_array_api.npx.npx_functions.identity(*inputs, **kwargs)

Makes a copy.

Signature:

(
    n: ParType[int],
    dtype: OptParType[DType],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.isnan(*inputs, **kwargs)

See numpy.isnan().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType['bool_', 'T1']:

onnx_array_api.npx.npx_functions.log(*inputs, **kwargs)

See numpy.log().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.log1p(*inputs, **kwargs)

See numpy.log1p().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.matmul(*inputs, **kwargs)

See numpy.matmul().

Signature:

(
    a: TensorType[numerics, 'T'],
    b: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.pad(*inputs, **kwargs)

It does not implement numpy.pad() but the ONNX version.

Signature:

(
    x: TensorType[numerics, 'T'],
    pads: TensorType['int64', 'I'],
    constant_value: OptTensorType[numerics, 'T'],
    axes: OptTensorType['int64', 'I'],
    mode: ParType[str],
):

onnx_array_api.npx.npx_functions.reciprocal(*inputs, **kwargs)

See numpy.reciprocal().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.relu(*inputs, **kwargs)

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.round(*inputs, **kwargs)

See numpy.round().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.sigmoid(*inputs, **kwargs)

See scipy.special.expit().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.sign(*inputs, **kwargs)

See numpy.sign().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.sin(*inputs, **kwargs)

See numpy.sin().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.sinh(*inputs, **kwargs)

See numpy.sinh().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.squeeze(*inputs, **kwargs)

See numpy.squeeze().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: OptTensorType['int64', 'I'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.tan(*inputs, **kwargs)

See numpy.tan().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.tanh(*inputs, **kwargs)

See numpy.tanh().

Signature:

(
    x: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.topk(*inputs, **kwargs)

See numpy.argsort().

Signature:

(
    x: TensorType[numerics, 'T'],
    k: TensorType['int64', (1,), 'I'],
    axis: OptParType[int],
    largest: OptParType[int],
    sorted: OptParType[int],
) -> TupleType[TensorType[numerics, 'T'], TensorType['int64', 'I']]:

onnx_array_api.npx.npx_functions.transpose(*inputs, **kwargs)

See numpy.transpose().

Signature:

(
    x: TensorType[numerics, 'T'],
    perm: ParType[typing.Tuple[int, ...]],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.unsqueeze(*inputs, **kwargs)

See numpy.expand_dims().

Signature:

(
    x: TensorType[numerics, 'T'],
    axis: TensorType['int64', 'I'],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.vstack(*inputs, **kwargs)

See numpy.vstack().

Signature:

(
    x: SequenceType["TensorType[numerics, 'T']"],
) -> TensorType[numerics, 'T']:

onnx_array_api.npx.npx_functions.where(*inputs, **kwargs)

See numpy.where().

Signature:

(
    cond: TensorType['bool_', 'B'],
    x: TensorType[numerics, 'T'],
    y: TensorType[numerics, 'T'],
) -> TensorType[numerics, 'T']: