module `npy.onnx_numpy_wrapper`#

Short summary#

module mlprodict.npy.onnx_numpy_wrapper

Wraps numpy functions into onnx.

Classes#

class	truncated documentation
`_created_classes`	Class to store all dynamic classes created by wrappers.
`wrapper_onnxnumpy`	Intermediate wrapper to store a pointer on the compiler (type: `OnnxNumpyCompiler`).
`wrapper_onnxnumpy_np`	Intermediate wrapper to store a pointer on the compiler (type: `OnnxNumpyCompiler`) supporting multiple signatures. …

Functions#

function	truncated documentation
`onnxnumpy`	Decorator to declare a function implemented using numpy syntax but executed with ONNX operators. …
`onnxnumpy_default`	Decorator with options to declare a function implemented using numpy syntax but executed with ONNX …
`onnxnumpy_np`	Decorator to declare a function implemented using numpy syntax but executed with ONNX operators. …

Methods#

method	truncated documentation
`__call__`	Calls the compiled function with arguments args.
`__call__`	Calls the compiled function assuming the type of the first tensor in args defines the templated version of the …
`__getitem__`	Returns the instance of `wrapper_onnxnumpy` mapped to dtype.
`__getstate__`	Serializes everything but the function which generates the ONNX graph, not needed anymore.
`__getstate__`	Serializes everything but the function which generates the ONNX graph, not needed anymore.
`__init__`
`__init__`
`__init__`
`__setstate__`	Serializes everything but the function which generates the ONNX graph, not needed anymore.
`__setstate__`	Restores serialized data.
`_populate`	Creates the appropriate runtime for function fct
`_validate_onnx_data`
`append`	Adds a class into globals() to enable pickling on dynamic classes.
`to_onnx`	Returns the ONNX graph for the wrapped function. It takes additional arguments to distinguish between multiple graphs. …
`to_onnx`	Returns the ONNX graph for the wrapped function. It takes additional arguments to distinguish between multiple graphs. …

Documentation#

Wraps numpy functions into onnx.

New in version 0.6.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper._created_classes#

Bases: object

Class to store all dynamic classes created by wrappers.

source on GitHub

__init__()#

append(name, cl)#

Adds a class into globals() to enable pickling on dynamic classes.

source on GitHub

mlprodict.npy.onnx_numpy_wrapper.onnxnumpy(op_version=None, runtime=None, signature=None)#

Decorator to declare a function implemented using numpy syntax but executed with ONNX operators.

Parameters:

op_version – ONNX opset version
runtime – ‘onnxruntime’ or one implemented by OnnxInference
signature – it should be used when the function is not annoatated.

Equivalent to onnxnumpy(arg)(foo).

New in version 0.6.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_custom_fct_None_None(compiled)#

Bases: wrapper_onnxnumpy

onnx numpy abs

__fct__() → <mlprodict.npy.onnx_numpy_annotation.NDArray.ShapeType object at 0x7f3e05b0f190>#: onnx numpy abs

__name__ = 'onnxnumpy_custom_fct_None_None'#

mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_default(fct)#

Decorator with options to declare a function implemented using numpy syntax but executed with ONNX operators.

Parameters:: fct – function to wrap

New in version 0.6.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_abs_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

abs

__fct__()#: abs

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_abs_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_acos_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

acos

__fct__()#: acos

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_acos_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_acosh_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

acosh

__fct__()#: acosh

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_acosh_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_amax_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

amax

__fct__(axis=None, keepdims=0)#: amax

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_amax_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_amin_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

amin

__fct__(axis=None, keepdims=0)#: amin

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_amin_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_arange_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

arange, start, stop must be specified.

__fct__(stop, step=1)#: arange, start, stop must be specified.

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_arange_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_argmax_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

argmax

__fct__(axis=0, keepdims=0)#: argmax

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_argmax_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_argmin_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

argmin

__fct__(axis=0, keepdims=0)#: argmin

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_argmin_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_asin_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

asin

__fct__()#: asin

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_asin_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_asinh_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

asinh

__fct__()#: asinh

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_asinh_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_atan_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

atan

__fct__()#: atan

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_atan_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_atanh_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

atanh

__fct__()#: atanh

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_atanh_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_ceil_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

ceil

__fct__()#: ceil

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_ceil_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_clip_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

clip

__fct__(a_min=None, a_max=None)#: clip

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_clip_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_compress_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

compress

__fct__(x, axis=None)#: compress

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_compress_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_concat_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

concat

__fct__(*, axis=0)#: concat

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_concat_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_cos_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

cos

__fct__()#: cos

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_cos_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_cosh_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

cosh

__fct__()#: cosh

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_cosh_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_cumsum_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

cumsum

__fct__(axis)#: cumsum

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_cumsum_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_det_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

det

__fct__()#: det

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_det_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_dot_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

dot

__fct__(b)#: dot

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_dot_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_einsum_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

einsum

__fct__(*, equation=None)#: einsum

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_einsum_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_erf_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

erf

__fct__()#: erf

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_erf_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_exp_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

exp

__fct__()#: exp

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_exp_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_expand_dims_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

expand_dims

__fct__(axis=0)#: expand_dims

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_expand_dims_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_expit_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

expit

__fct__()#: expit

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_expit_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_floor_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

floor

__fct__()#: floor

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_floor_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_hstack_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

hstack

__fct__()#: hstack

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_hstack_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_isnan_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

isnan

__fct__()#: isnan

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_isnan_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_log1p_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

log1p

__fct__()#: log1p

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_log1p_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_log_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

log

__fct__()#: log

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_log_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_logistic_regression_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

logistic_regression

__fct__(*, model=None)#: logistic_regression

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_logistic_regression_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_matmul_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

matmul

__fct__(b)#: matmul

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_matmul_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_mean_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

mean

__fct__(axis=None, keepdims=0)#: mean

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_mean_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_onnx_cube_32_None_onnxruntime(**kwargs)#

Bases: wrapper_onnxnumpy_np

__fct__()#

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_onnx_cube_32_None_onnxruntime'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_onnx_rel_diff_32_None_onnxruntime(**kwargs)#

Bases: wrapper_onnxnumpy_np

__fct__(b)#

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_onnx_rel_diff_32_None_onnxruntime'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_onnx_sq_diff_32_None_onnxruntime(**kwargs)#

Bases: wrapper_onnxnumpy_np

__fct__(b)#

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_onnx_sq_diff_32_None_onnxruntime'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_onnx_square_32_None_onnxruntime(**kwargs)#

Bases: wrapper_onnxnumpy_np

__fct__()#

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_onnx_square_32_None_onnxruntime'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_onnx_sum_32_None_onnxruntime(**kwargs)#

Bases: wrapper_onnxnumpy_np

__fct__(b)#

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_onnx_sum_32_None_onnxruntime'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_pad_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

pad

__fct__(pads, constant_value=None, mode='constant')#: pad

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_pad_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_prod_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

prod

__fct__(axis=None, keepdims=0)#: prod

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_prod_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_reciprocal_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

reciprocal

__fct__()#: reciprocal

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_reciprocal_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_relu_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

relu

__fct__()#: relu

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_relu_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_round_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

round

__fct__()#: round

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_round_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_sigmoid_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

expit

__fct__()#: expit

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_sigmoid_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_sign_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

sign

__fct__()#: sign

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_sign_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_sin_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

sin

__fct__()#: sin

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_sin_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_sinh_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

sinh

__fct__()#: sinh

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_sinh_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_sqrt_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

sqrt

__fct__()#: sqrt

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_sqrt_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_squeeze_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

squeeze

__fct__(axis=None)#: squeeze

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_squeeze_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_sum_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

sum

__fct__(axis=None, keepdims=0)#: sum

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_sum_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_tan_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

tan

__fct__()#: tan

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_tan_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_tanh_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

tanh

__fct__()#: tanh

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_tanh_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_topk_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

topk

__fct__(k, axis=-1, largest=1, sorted=1)#: topk

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_topk_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_transpose_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

transpose

__fct__(perm=(1, 0))#: transpose

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_transpose_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_unsqueeze_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

unsqueeze

__fct__(axes)#: unsqueeze

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_unsqueeze_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_vstack_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

vstack

__fct__()#: vstack

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_vstack_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_nb_where_None_None(**kwargs)#

Bases: wrapper_onnxnumpy_np

where

__fct__(x, y)#: where

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__name__ = 'onnxnumpy_nb_where_None_None'#

__setstate__(state)#

Restores serialized data.

source on GitHub

mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_np(op_version=None, runtime=None, signature=None)#

Decorator to declare a function implemented using numpy syntax but executed with ONNX operators.

Parameters:

op_version – ONNX opset version
runtime – ‘onnxruntime’ or one implemented by OnnxInference
signature – it should be used when the function is not annoatated.

Equivalent to onnxnumpy(arg)(foo).

New in version 0.6.

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_onnx_exp_1_None_None(compiled)#

Bases: wrapper_onnxnumpy

__fct__() → <mlprodict.npy.onnx_numpy_annotation.NDArray.ShapeType object at 0x7f3ea80f2b20>#

__name__ = 'onnxnumpy_onnx_exp_1_None_None'#

class mlprodict.npy.onnx_numpy_wrapper.onnxnumpy_onnx_log_1_None_None(compiled)#

Bases: wrapper_onnxnumpy

__fct__() → <mlprodict.npy.onnx_numpy_annotation.NDArray.ShapeType object at 0x7f3ea80f28b0>#

__name__ = 'onnxnumpy_onnx_log_1_None_None'#

class mlprodict.npy.onnx_numpy_wrapper.wrapper_onnxnumpy(compiled)#

Bases: object

Intermediate wrapper to store a pointer on the compiler (type: OnnxNumpyCompiler).

Parameters:: compiled – instance of OnnxNumpyCompiler

New in version 0.6.

source on GitHub

__call__(*args, **kwargs)#

Calls the compiled function with arguments args.

source on GitHub

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__init__(compiled)#

__setstate__(state)#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

to_onnx(**kwargs)#

Returns the ONNX graph for the wrapped function. It takes additional arguments to distinguish between multiple graphs. This happens when a function needs to support multiple type.

Returns:: ONNX graph

source on GitHub

class mlprodict.npy.onnx_numpy_wrapper.wrapper_onnxnumpy_np(**kwargs)#

Bases: object

Intermediate wrapper to store a pointer on the compiler (type: OnnxNumpyCompiler) supporting multiple signatures.

New in version 0.6.

source on GitHub

__call__(*args, **kwargs)#

Calls the compiled function assuming the type of the first tensor in args defines the templated version of the function to convert into ONNX.

source on GitHub

__getitem__(dtype)#

Returns the instance of wrapper_onnxnumpy mapped to dtype.

Parameters:: dtype – numpy dtype corresponding to the input dtype of the function
Returns:: instance of wrapper_onnxnumpy

source on GitHub

__getstate__()#

Serializes everything but the function which generates the ONNX graph, not needed anymore.

source on GitHub

__init__(**kwargs)#

__setstate__(state)#

Restores serialized data.

source on GitHub

_populate(version)#

Creates the appropriate runtime for function fct

source on GitHub

_validate_onnx_data(X)#

to_onnx(**kwargs)#

Returns the ONNX graph for the wrapped function. It takes additional arguments to distinguish between multiple graphs. This happens when a function needs to support multiple type.

Returns:: ONNX graph

source on GitHub

module npy.onnx_numpy_wrapper#

Short summary#

Classes#

Functions#

Methods#

Documentation#

module `npy.onnx_numpy_wrapper`#