# -*- encoding: utf-8 -*-
# pylint: disable=E0203,E1101,C0111
"""
Runtime operator.
:githublink:`%|py|7`
"""
import numpy
from numpy.random import RandomState
from onnx.defs import onnx_opset_version
from ._op import OpRun
[docs]def _dropout(X, drop_probability=0.5, seed=0,
training_mode=False, return_mask=False):
if drop_probability == 0 or not training_mode:
if return_mask:
return X, numpy.ones(X.shape, dtype=bool)
return (X, )
rnd = RandomState(seed)
mask = rnd.uniform(0, 1.0, X.shape) >= drop_probability
scale = (1. / (1. - drop_probability))
if return_mask:
return mask * X * scale, mask.astype(bool)
return (mask * X * scale, )
[docs]class DropoutBase(OpRun):
[docs] def __init__(self, onnx_node, desc=None, expected_attributes=None, **options):
OpRun.__init__(self, onnx_node, desc=desc,
expected_attributes=expected_attributes,
**options)
self.nb_outputs = len(onnx_node.output)
[docs] def _private_run(self, X, seed=None, ratio=0.5, training_mode=False): # pylint: disable=W0221
return _dropout(X, ratio, seed=seed, return_mask=self.nb_outputs == 2,
training_mode=training_mode)
[docs] def _infer_shapes(self, *inputs): # pylint: disable=W0221
X = inputs[0]
if self.nb_outputs == 1:
return (X.copy(), )
if self.nb_outputs == 2:
return (X.copy(), X.copy())
raise RuntimeError( # pragma: no cover
"Unexpected numbers of output {} > 2.".format(self.nb_outputs))
[docs]class Dropout_7(DropoutBase):
atts = {'ratio': 0.5}
[docs] def __init__(self, onnx_node, desc=None, **options):
DropoutBase.__init__(self, onnx_node, desc=desc,
expected_attributes=Dropout_7.atts,
**options)
[docs] def _run(self, X): # pylint: disable=W0221
return self._private_run(X, self.ratio)
[docs]class Dropout_12(DropoutBase):
atts = {'seed': 0}
[docs] def __init__(self, onnx_node, desc=None, **options):
DropoutBase.__init__(self, onnx_node, desc=desc,
expected_attributes=Dropout_12.atts,
**options)
[docs] def _run(self, *inputs): # pylint: disable=W0221
X = inputs[0]
ratio = 0.5 if len(inputs) <= 1 else inputs[1]
training_mode = False if len(inputs) <= 2 else inputs[2]
return self._private_run(X, seed=self.seed, ratio=ratio,
training_mode=training_mode)
if onnx_opset_version() >= 12:
Dropout = Dropout_12
else:
Dropout = Dropout_7 # pragma: no cover
