/
maximum.py
75 lines (58 loc) · 2.09 KB
/
maximum.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
import numpy
import chainer
from chainer.backends import cuda
from chainer import function_node
from chainer import utils
from chainer.utils import type_check
class Maximum(function_node.FunctionNode):
"""Element-wise maximum of input variables."""
def check_type_forward(self, in_types):
type_check.expect(
in_types.size() == 2,
in_types[0].dtype.kind == 'f',
in_types[0].dtype == in_types[1].dtype,
in_types[0].shape == in_types[1].shape
)
def forward_cpu(self, inputs):
self.retain_inputs((0, 1))
x1, x2 = inputs
y = numpy.maximum(x1, x2)
return utils.force_array(y),
def forward_gpu(self, inputs):
self.retain_inputs((0, 1))
x1, x2 = inputs
return cuda.cupy.maximum(x1, x2),
def backward(self, indexes, grad_outputs):
x1, x2 = self.get_retained_inputs()
return MaximumGrad(
utils.force_array(x1.data >= x2.data)).apply((grad_outputs[0],))
class MaximumGrad(function_node.FunctionNode):
def __init__(self, cond):
self.cond = cond
def forward_cpu(self, inputs):
gy, = inputs
gx1 = numpy.where(self.cond, gy, gy.dtype.type(0))
gx2 = numpy.where(self.cond, gy.dtype.type(0), gy)
return utils.force_array(gx1), utils.force_array(gx2)
def forward_gpu(self, inputs):
gy, = inputs
gx1, gx2 = cuda.elementwise(
'S cond, T gy', 'T gx1, T gx2',
'''
gx1 = cond ? gy : (T)0.0;
gx2 = cond ? (T)0.0 : gy;
''',
'maximum_bwd1')(self.cond, gy)
return gx1, gx2
def backward(self, indexes, grad_outputs):
return chainer.functions.where(
self.cond, grad_outputs[0], grad_outputs[1]),
def maximum(x1, x2):
"""Element-wise maximum of input variables.
Args:
x1 (~chainer.Variable): Input variables to be compared.
x2 (~chainer.Variable): Input variables to be compared.
Returns:
~chainer.Variable: Output variable.
"""
return Maximum().apply((x1, x2))[0]