Others…

xbuilder.shape_helper

compatible_dimensions

experimental_experiment.xbuilder.shape_helper.compatible_dimensions(*dims: Sequence[int | str]) → bool

Evaluates the fact all the dimensions can be equal or not.

Parameters:

dims – dimensions

Returns:

compatibility

<<<

from experimental_experiment.xbuilder.shape_helper import compatible_dimensions

print(compatible_dimensions(1, 1))  # True
print(compatible_dimensions(1, 2))  # False
print(compatible_dimensions(1, "D"))  # True
print(compatible_dimensions(1, "D", "DD"))  # True

>>>

    True
    False
    True
    True

compatible_shapes

experimental_experiment.xbuilder.shape_helper.compatible_shapes(sh1: Tuple[int | torch.SymInt | str, ...], sh2: Tuple[int | torch.SymInt | str, ...]) → bool

Checks that two shapes are compatible. If both static, they must be equal. If dynamic, the variable part must be compatible meaning they could be equal.

Parameters:

• sh1 – first shape

• sh2 – seconde shape

Returns:

compatibility

<<<

from experimental_experiment.xbuilder.shape_helper import compatible_shapes

print(compatible_shapes((1, 2), (1, 2)))  # True
print(compatible_shapes((1, 2), (1, "D2")))  # True
print(compatible_shapes(("D2", 2), (1, "D2")))  # False
print(compatible_shapes(("D2", 2), (2, "D2")))  # True

>>>

    True
    True
    False
    True

xbuilder._onnx_helper

choose_consistent_domain_opset

experimental_experiment.xbuilder._onnx_helper.choose_consistent_domain_opset(domain: str, opsets: Dict[str, int] | None = None) → int

Chooses a compatible opset for a particular domain given this existing one. Only works for ai.onnx.ml, otherwise return 1.

Parameters:

• domain – new domain

• opsets – existing opsets

Returns:

version

compatible_opsets

experimental_experiment.xbuilder._onnx_helper.compatible_opsets(domain: str, op_type: str, current: int, new_version: int) → bool

Tells if two opset version for a particular operator type means the same version of it.

Parameters:

• domain – domain, only ai.onnx and ai.onnx.ml are checked.

• op_type – operator type

• current – current domain version

• new_version – new version

Returns:

result

element_wise_binary_op_types

experimental_experiment.xbuilder._onnx_helper.element_wise_binary_op_types() → Set[str]

Returns the list of element-wise operators.

<<<

import pprint
from experimental_experiment.xbuilder._onnx_helper import element_wise_binary_op_types

pprint.pprint(element_wise_binary_op_types())

>>>

    {'Sub', 'Mod', 'Add', 'Mul', 'Div', 'Xor', 'Or', 'And'}

element_wise_op_cmp_types

experimental_experiment.xbuilder._onnx_helper.element_wise_op_cmp_types() → Set[str]

Returns the list of element-wise operators doing comparisons.

<<<

import pprint
from experimental_experiment.xbuilder._onnx_helper import element_wise_op_cmp_types

pprint.pprint(element_wise_op_cmp_types())

>>>

    {'Less', 'GreaterOrEqual', 'Equal', 'LessOrEqual', 'Greater'}

unary_like_op_types

experimental_experiment.xbuilder._onnx_helper.unary_like_op_types() → Set[str]

Returns the list of unary like operators. They do not change the shape. They may change the type.

<<<

import pprint
from experimental_experiment.xbuilder._onnx_helper import unary_like_op_types

pprint.pprint(unary_like_op_types())

>>>

    {'Abs',
     'Acos',
     'Acosh',
     'Asin',
     'Asinh',
     'Atan',
     'Atanh',
     'BitShift',
     'Cast',
     'CastLike',
     'Ceil',
     'Celu',
     'Clip',
     'Cos',
     'Cosh',
     'DequantizeLinear',
     'DynamicQuantizeLinear',
     'Elu',
     'Erf',
     'Exp',
     'IsInf',
     'Log',
     'LogSoftmax',
     'Neg',
     'Not',
     'PRelu',
     'Pow',
     'QuantizeLinear',
     'Reciprocal',
     'Relu',
     'Round',
     'Selu',
     'Sigmoid',
     'Sign',
     'Sin',
     'Sinh',
     'Softmax',
     'SoftmaxCrossEntropyLoss',
     'Softplus',
     'Softsign',
     'Sqrt',
     'Tan',
     'Tanh',
     'ThresholdRelu'}