Skip to content

P
pytensor
提交 104eb5fa authored 作者: Gijs van Tulder's avatar Gijs van Tulder
浏览文件
操作

Add sum mode to Downsample

上级 a0dadf5d
隐藏空白字符变更
内嵌 并排
正在显示 包含 52 行增加 和 34 行删除
theano/tensor/signal/downsample.py
浏览文件 @ 104eb5fa
......@@ -64,10 +64,10 @@ def max_pool_2d(input, ds, ignore_border=False, st=None, padding=(0, 0),
of the images, pad_h is the size of the top and bottom margins,
and pad_w is the size of the left and right margins.
:type padding: tuple of two ints
:param mode: 'max', 'average_inc_pad' or 'average_exc_pad'.
Operation executed on each window. `max` always excludes the padding
in the computation. `average` gives you the choice to include or
exclude it.
:param mode: 'max', 'sum', 'average_inc_pad' or 'average_exc_pad'.
Operation executed on each window. `max` and `sum` always exclude
the padding in the computation. `average` gives you the choice to
include or exclude it.
:type mode: string
"""
if input.ndim < 2:
......@@ -104,7 +104,7 @@ def max_pool_2d(input, ds, ignore_border=False, st=None, padding=(0, 0),
class DownsampleFactorMax(Op):
"""For N-dimensional tensors, consider that the last two
dimensions span images. This Op downsamples these images by
taking the max or average over different patch.
taking the max, sum or average over different patch.
"""
__props__ = ('ds', 'ignore_border', 'st', 'padding', 'mode')
......@@ -188,7 +188,7 @@ class DownsampleFactorMax(Op):
def __init__(self, ds, ignore_border=False, st=None, padding=(0, 0),
mode='max'):
""" Take the max or average or different input patches.
""" Take the max, sum or average or different input patches.
:param ds: downsample factor over rows and column.
ds indicates the pool region size.
......@@ -210,8 +210,8 @@ class DownsampleFactorMax(Op):
and pad_w is the size of the left and right margins.
:type padding: tuple of two ints
:param mode: 'max', 'average_inc_pad', 'average_exc_pad'.
('average_inc_pad' exclude the padding from the count,
:param mode: 'max', 'sum', 'average_inc_pad', 'average_exc_pad'.
('average_inc_pad' excludes the padding from the count,
'average_exc_pad' include it)
"""
......@@ -232,9 +232,9 @@ class DownsampleFactorMax(Op):
if self.padding[0] >= self.ds[0] or self.padding[1] >= self.ds[1]:
raise NotImplementedError(
'padding_h and padding_w must be smaller than strides')
if mode not in ['max', 'average_inc_pad', 'average_exc_pad']:
if mode not in ['max', 'average_inc_pad', 'average_exc_pad', 'sum']:
raise ValueError(
"DownsampleFactorMax mode parameter only support 'max',"
"DownsampleFactorMax mode parameter only support 'max', 'sum',"
" 'average_inc_pad' and 'average_exc_pad'. Got %s" % mode)
self.mode = mode
......@@ -277,7 +277,9 @@ class DownsampleFactorMax(Op):
else:
y = x
func = numpy.max
if self.mode != 'max':
if self.mode == 'sum':
func = numpy.sum
elif self.mode != 'max':
func = numpy.average
for n in xrange(x.shape[0]):
......@@ -317,7 +319,7 @@ class DownsampleFactorMax(Op):
return ['<algorithm>']
def c_code(self, node, name, inp, out, sub):
if self.mode not in ('max', 'average_exc_pad', 'average_inc_pad'):
if self.mode not in ('max', 'sum', 'average_exc_pad', 'average_inc_pad'):
raise theano.gof.utils.MethodNotDefined()
x, = inp
z, = out
......@@ -448,7 +450,7 @@ class DownsampleFactorMax(Op):
"""
if self.mode == 'max':
ccode += """
// use the first element as the initial value of maximum
// use the first element as the initial value of collector
collector = ((dtype_%(x)s*)(PyArray_GETPTR4(%(x)s,b,k,r_st,c_st)))[0];
// go through the pooled region in the unpadded input
for(int m=r_st; m<r_end; m++)
......@@ -461,7 +463,7 @@ class DownsampleFactorMax(Op):
}
z[0] = collector;
"""
elif self.mode == 'average_exc_pad' or self.mode == 'average_inc_pad':
elif self.mode in ('sum', 'average_exc_pad', 'average_inc_pad'):
ccode += """
// initialize the sum at zero
collector = ((dtype_%(x)s)(0));
......@@ -475,7 +477,11 @@ class DownsampleFactorMax(Op):
}
}
"""
if self.mode == 'average_inc_pad' and self.ignore_border:
if self.mode == "sum":
ccode += """
z[0] = collector;
"""
elif self.mode == 'average_inc_pad' and self.ignore_border:
ccode += """
z[0] = collector / (%(ds0)s * %(ds1)s);
"""
......@@ -493,7 +499,7 @@ class DownsampleFactorMax(Op):
return ccode % locals()
def c_code_cache_version(self):
return (0, 6, 8, 1)
return (0, 6, 8, 3)
class DownsampleFactorMaxGrad(Op):
__props__ = ('ds', 'ignore_border', 'st', 'padding', 'mode')
......@@ -505,9 +511,9 @@ class DownsampleFactorMaxGrad(Op):
st = ds
self.st = tuple(st)
self.padding = tuple(padding)
if mode not in ['max', 'average_inc_pad', 'average_exc_pad']:
if mode not in ['max', 'sum', 'average_inc_pad', 'average_exc_pad']:
raise ValueError(
"DownsampleFactorMax mode parameter only support 'max',"
"DownsampleFactorMax mode parameter only support 'max', 'sum',"
" 'average_inc_pad' and 'average_exc_pad'. Got %s" % mode)
self.mode = mode
......@@ -524,7 +530,7 @@ class DownsampleFactorMaxGrad(Op):
return Apply(self, [x, maxout, gz], [x.type()])
def perform(self, node, inp, out):
if self.mode != 'max' and self.padding != (0, 0):
if self.mode not in ('max', 'sum') and self.padding != (0, 0):
raise NotImplementedError()
x, maxout, gz = inp
gx_stg, = out
......@@ -539,6 +545,7 @@ class DownsampleFactorMaxGrad(Op):
img_rows = x.shape[-2] + 2 * pad_h
img_cols = x.shape[-1] + 2 * pad_w
inc_pad = self.mode == 'average_inc_pad'
sum_mode = self.mode == 'sum'
# pad the image
if self.padding != (0, 0):
......@@ -566,20 +573,23 @@ class DownsampleFactorMaxGrad(Op):
for n in xrange(x.shape[0]):
for k in xrange(x.shape[1]):
for r in xrange(pr):
if inc_pad:
if sum_mode or inc_pad:
row_st = r * st0
else:
row_st = __builtin__.max(r * st0, self.padding[0])
row_end = __builtin__.min(row_st + ds0, img_rows)
for c in xrange(pc):
if inc_pad:
if sum_mode or inc_pad:
col_st = c * st1
else:
col_st = __builtin__.max(c * st1,
self.padding[1])
col_end = __builtin__.min(col_st + ds1, img_cols)
val = gz[n, k, r, c] / ((row_end - row_st) *
(col_end - col_st))
if sum_mode:
val = gz[n, k, r, c]
else:
val = gz[n, k, r, c] / ((row_end - row_st) *
(col_end - col_st))
gx[n, k, row_st:row_end, col_st:col_end] += val
# unpad the image
gx = gx[:, :, pad_h:(img_rows-pad_h), pad_w:(img_cols-pad_w)]
......
theano/tensor/signal/tests/test_downsample.py
浏览文件 @ 104eb5fa
......@@ -33,7 +33,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
out_shp.append(input.shape[-1] / ds[1] + yi)
output_val = numpy.zeros(out_shp)
func = numpy.max
if mode != 'max':
if mode == 'sum':
func = numpy.sum
elif mode != 'max':
func = numpy.average
for k in numpy.ndindex(*input.shape[:-2]):
......@@ -76,7 +78,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
tt = []
y = pad_img(x)
func = numpy.max
if mode != 'max':
if mode == 'sum':
func = numpy.sum
elif mode != 'max':
func = numpy.average
inc_pad = mode == 'average_inc_pad'
......@@ -145,7 +149,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
out_shp.append(out_c)
func = numpy.max
if mode != 'max':
if mode == 'sum':
func = numpy.sum
elif mode != 'max':
func = numpy.average
output_val = numpy.zeros(out_shp)
......@@ -169,6 +175,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for maxpoolshp, ignore_border, mode in product(maxpoolshps,
[True, False],
['max',
'sum',
'average_inc_pad',
'average_exc_pad']):
# print 'maxpoolshp =', maxpoolshp
......@@ -198,23 +205,23 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
stridesizes = ((1, 1), (3, 3), (5, 7))
# generate random images
imval = rng.rand(4, 10, 16, 16)
# The same for each mode
outputshps = ((4, 10, 16, 16), (4, 10, 6, 6), (4, 10, 4, 3),
(4, 10, 16, 16), (4, 10, 6, 6), (4, 10, 4, 3),
(4, 10, 14, 14), (4, 10, 5, 5), (4, 10, 3, 2),
(4, 10, 14, 14), (4, 10, 6, 6), (4, 10, 4, 3),
(4, 10, 12, 14), (4, 10, 4, 5), (4, 10, 3, 2),
(4, 10, 12, 14), (4, 10, 5, 6), (4, 10, 4, 3))
# The same for each mode
outputshps = outputshps + outputshps + outputshps
images = tensor.dtensor4()
indx = 0
for mode, maxpoolshp, ignore_border in product(['max',
'sum',
'average_inc_pad',
'average_exc_pad'],
maxpoolshps,
[True, False]):
for stride in stridesizes:
outputshp = outputshps[indx]
outputshp = outputshps[indx % len(outputshps)]
indx += 1
# DownsampleFactorMax op
numpy_output_val = \
......@@ -251,7 +258,8 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
stride = stridesizes[indx]
maxpoolshp = maxpoolshps[indx]
for ignore_border, mode in product([True, False],
['max', 'average_inc_pad',
['max', 'sum',
'average_inc_pad',
'average_exc_pad']):
indx_out = indx * 2
if not ignore_border:
......@@ -283,7 +291,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
c = 2 # channel size
images = tensor.dtensor4()
for indx, mode in product(numpy.arange(len(maxpoolsizes)),
['max', 'average_inc_pad',
['max', 'sum', 'average_inc_pad',
'average_exc_pad']):
imgsize = imgsizes[indx]
imval = rng.rand(m, c, imgsize[0], imgsize[1]) - 0.5
......@@ -486,7 +494,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for maxpoolshp, ignore_border, mode in product(maxpoolshps,
[True, False],
['max',
['max', 'sum',
'average_inc_pad',
'average_exc_pad']):
# print 'maxpoolshp =', maxpoolshp
......@@ -537,7 +545,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for maxpoolshp, ignore_border, mode in product(maxpoolshps,
[True, False],
['max',
['max', 'sum',
'average_inc_pad',
'average_exc_pad']):
# print 'maxpoolshp =', maxpoolshp
......@@ -574,7 +582,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for maxpoolshp, ignore_border, mode in product(maxpoolshps,
[True, False],
['max',
['max', 'sum',
'average_inc_pad',
'average_exc_pad']):
# print 'maxpoolshp =', maxpoolshp
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论