Skip to content

P
pytensor
提交 57cd51cb authored 作者: Chiheb Trabelsi's avatar Chiheb Trabelsi
浏览文件
操作

Functions parameters have been modified in test_pool.py for coherence purposes.…

Functions parameters have been modified in test_pool.py for coherence purposes. The test also contains some examples where it should print a deprecation warning. For that, you should use the following command: nosetests -s test_pool.py.
上级 c49136eb
隐藏空白字符变更
内嵌 并排
正在显示 包含 88 行增加 和 78 行删除
theano/tensor/signal/tests/test_pool.py
浏览文件 @ 57cd51cb
...@@ -30,7 +30,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -30,7 +30,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
assert Pool.out_shape((8, 6), (2, 2)) == [4, 3] assert Pool.out_shape((8, 6), (2, 2)) == [4, 3]
@staticmethod @staticmethod
def numpy_max_pool_2d(input, ds, ignore_border=False, mode='max'): def numpy_max_pool_2d(input, ws, ignore_border=False, mode='max'):
'''Helper function, implementing pool_2d in pure numpy''' '''Helper function, implementing pool_2d in pure numpy'''
if len(input.shape) < 2: if len(input.shape) < 2:
raise NotImplementedError('input should have at least 2 dim,' raise NotImplementedError('input should have at least 2 dim,'
...@@ -39,13 +39,13 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -39,13 +39,13 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
xi = 0 xi = 0
yi = 0 yi = 0
if not ignore_border: if not ignore_border:
if input.shape[-2] % ds[0]: if input.shape[-2] % ws[0]:
xi += 1 xi += 1
if input.shape[-1] % ds[1]: if input.shape[-1] % ws[1]:
yi += 1 yi += 1
out_shp = list(input.shape[:-2]) out_shp = list(input.shape[:-2])
out_shp.append(input.shape[-2] // ds[0] + xi) out_shp.append(input.shape[-2] // ws[0] + xi)
out_shp.append(input.shape[-1] // ds[1] + yi) out_shp.append(input.shape[-1] // ws[1] + yi)
output_val = numpy.zeros(out_shp) output_val = numpy.zeros(out_shp)
func = numpy.max func = numpy.max
if mode == 'sum': if mode == 'sum':
...@@ -55,29 +55,29 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -55,29 +55,29 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for k in numpy.ndindex(*input.shape[:-2]): for k in numpy.ndindex(*input.shape[:-2]):
for i in range(output_val.shape[-2]): for i in range(output_val.shape[-2]):
ii = i * ds[0] ii = i * ws[0]
for j in range(output_val.shape[-1]): for j in range(output_val.shape[-1]):
jj = j * ds[1] jj = j * ws[1]
patch = input[k][ii:ii + ds[0], jj:jj + ds[1]] patch = input[k][ii:ii + ws[0], jj:jj + ws[1]]
output_val[k][i, j] = func(patch) output_val[k][i, j] = func(patch)
return output_val return output_val
@staticmethod @staticmethod
def numpy_max_pool_nd(input, ds, ignore_border=False, mode='max'): def numpy_max_pool_nd(input, ws, ignore_border=False, mode='max'):
'''Helper function, implementing pool_nd in pure numpy''' '''Helper function, implementing pool_nd in pure numpy'''
if len(input.shape) < len(ds): if len(input.shape) < len(ws):
raise NotImplementedError('input should have at least %s dim,' raise NotImplementedError('input should have at least %s dim,'
' shape is %s' ' shape is %s'
% (str(ds), str(input.shape))) % (str(ws), str(input.shape)))
nd = len(ds) nd = len(ws)
si = [0] * nd si = [0] * nd
if not ignore_border: if not ignore_border:
for i in range(nd): for i in range(nd):
if input.shape[-nd + i] % ds[i]: if input.shape[-nd + i] % ws[i]:
si[i] += 1 si[i] += 1
out_shp = list(input.shape[:-nd]) out_shp = list(input.shape[:-nd])
for i in range(nd): for i in range(nd):
out_shp.append(input.shape[-nd + i] // ds[i] + si[i]) out_shp.append(input.shape[-nd + i] // ws[i] + si[i])
output_val = numpy.zeros(out_shp) output_val = numpy.zeros(out_shp)
func = numpy.max func = numpy.max
if mode == 'sum': if mode == 'sum':
...@@ -87,21 +87,21 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -87,21 +87,21 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for l in numpy.ndindex(*input.shape[:-nd]): for l in numpy.ndindex(*input.shape[:-nd]):
for r in numpy.ndindex(*output_val.shape[-nd:]): for r in numpy.ndindex(*output_val.shape[-nd:]):
patch = input[l][tuple(slice(r[i] * ds[i], (r[i] + 1) * ds[i]) patch = input[l][tuple(slice(r[i] * ws[i], (r[i] + 1) * ws[i])
for i in range(nd))] for i in range(nd))]
output_val[l][r] = func(patch) output_val[l][r] = func(patch)
return output_val return output_val
@staticmethod @staticmethod
def numpy_max_pool_2d_stride_padding( def numpy_max_pool_2d_stride_padding(
x, ds, ignore_border=True, st=None, padding=(0, 0), mode='max'): x, ws, ignore_border=True, stride=None, pad=(0, 0), mode='max'):
assert ignore_border assert ignore_border
pad_h = padding[0] pad_h = pad[0]
pad_w = padding[1] pad_w = pad[1]
h = x.shape[-2] h = x.shape[-2]
w = x.shape[-1] w = x.shape[-1]
assert ds[0] > pad_h assert ws[0] > pad_h
assert ds[1] > pad_w assert ws[1] > pad_w
def pad_img(x): def pad_img(x):
y = numpy.zeros( y = numpy.zeros(
...@@ -113,13 +113,13 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -113,13 +113,13 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
return y return y
img_rows = h + 2 * pad_h img_rows = h + 2 * pad_h
img_cols = w + 2 * pad_w img_cols = w + 2 * pad_w
out_r = (img_rows - ds[0]) // st[0] + 1 out_r = (img_rows - ws[0]) // stride[0] + 1
out_c = (img_cols - ds[1]) // st[1] + 1 out_c = (img_cols - ws[1]) // stride[1] + 1
out_shp = list(x.shape[:-2]) out_shp = list(x.shape[:-2])
out_shp.append(out_r) out_shp.append(out_r)
out_shp.append(out_c) out_shp.append(out_c)
ds0, ds1 = ds ws0, ws1 = ws
st0, st1 = st st0, st1 = stride
output_val = numpy.zeros(out_shp) output_val = numpy.zeros(out_shp)
y = pad_img(x) y = pad_img(x)
func = numpy.max func = numpy.max
...@@ -131,14 +131,14 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -131,14 +131,14 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for k in numpy.ndindex(*x.shape[:-2]): for k in numpy.ndindex(*x.shape[:-2]):
for i in range(output_val.shape[-2]): for i in range(output_val.shape[-2]):
ii_st = i * st[0] ii_st = i * stride[0]
ii_end = builtins.min(ii_st + ds[0], img_rows) ii_end = builtins.min(ii_st + ws[0], img_rows)
if not inc_pad: if not inc_pad:
ii_st = builtins.max(ii_st, pad_h) ii_st = builtins.max(ii_st, pad_h)
ii_end = builtins.min(ii_end, h + pad_h) ii_end = builtins.min(ii_end, h + pad_h)
for j in range(output_val.shape[-1]): for j in range(output_val.shape[-1]):
jj_st = j * st[1] jj_st = j * stride[1]
jj_end = builtins.min(jj_st + ds[1], img_cols) jj_end = builtins.min(jj_st + ws[1], img_cols)
if not inc_pad: if not inc_pad:
jj_st = builtins.max(jj_st, pad_w) jj_st = builtins.max(jj_st, pad_w)
jj_end = builtins.min(jj_end, w + pad_w) jj_end = builtins.min(jj_end, w + pad_w)
...@@ -148,25 +148,25 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -148,25 +148,25 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
@staticmethod @staticmethod
def numpy_max_pool_nd_stride_padding( def numpy_max_pool_nd_stride_padding(
input, ds, ignore_border=True, st=None, padding=None, mode='max'): input, ws, ignore_border=True, stride=None, pad=None, mode='max'):
assert ignore_border assert ignore_border
nd = len(ds) nd = len(ws)
if padding is None: if pad is None:
padding = (0,) * nd pad = (0,) * nd
if st is None: if stride is None:
st = (0,) * nd stride = (0,) * nd
assert len(padding) == len(ds) == len(st) assert len(pad) == len(ws) == len(stride)
assert all(ds[i] > padding[i] for i in range(nd)) assert all(ws[i] > pad[i] for i in range(nd))
def pad_img(x): def pad_img(x):
# initialize padded input # initialize padded input
y = numpy.zeros( y = numpy.zeros(
x.shape[0:-nd] + x.shape[0:-nd] +
tuple(x.shape[-nd + i] + padding[i] * 2 for i in range(nd)), tuple(x.shape[-nd + i] + pad[i] * 2 for i in range(nd)),
dtype=x.dtype) dtype=x.dtype)
# place the unpadded input in the center # place the unpadded input in the center
block = ((slice(None),) * (len(x.shape) - nd) + block = ((slice(None),) * (len(x.shape) - nd) +
tuple(slice(padding[i], x.shape[-nd + i] + padding[i]) tuple(slice(pad[i], x.shape[-nd + i] + pad[i])
for i in range(nd))) for i in range(nd)))
y[block] = x y[block] = x
return y return y
...@@ -174,9 +174,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -174,9 +174,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
pad_img_shp = list(input.shape[:-nd]) pad_img_shp = list(input.shape[:-nd])
out_shp = list(input.shape[:-nd]) out_shp = list(input.shape[:-nd])
for i in range(nd): for i in range(nd):
padded_size = input.shape[-nd + i] + 2 * padding[i] padded_size = input.shape[-nd + i] + 2 * pad[i]
pad_img_shp.append(padded_size) pad_img_shp.append(padded_size)
out_shp.append((padded_size - ds[i]) // st[i] + 1) out_shp.append((padded_size - ws[i]) // stride[i] + 1)
output_val = numpy.zeros(out_shp) output_val = numpy.zeros(out_shp)
padded_input = pad_img(input) padded_input = pad_img(input)
func = numpy.max func = numpy.max
...@@ -190,51 +190,51 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -190,51 +190,51 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for r in numpy.ndindex(*output_val.shape[-nd:]): for r in numpy.ndindex(*output_val.shape[-nd:]):
region = [] region = []
for i in range(nd): for i in range(nd):
r_st = r[i] * st[i] r_st = r[i] * stride[i]
r_end = builtins.min(r_st + ds[i], pad_img_shp[-nd + i]) r_end = builtins.min(r_st + ws[i], pad_img_shp[-nd + i])
if not inc_pad: if not inc_pad:
r_st = builtins.max(r_st, padding[i]) r_st = builtins.max(r_st, pad[i])
r_end = builtins.min(r_end, input.shape[-nd + i] + padding[i]) r_end = builtins.min(r_end, input.shape[-nd + i] + pad[i])
region.append(slice(r_st, r_end)) region.append(slice(r_st, r_end))
patch = padded_input[l][region] patch = padded_input[l][region]
output_val[l][r] = func(patch) output_val[l][r] = func(patch)
return output_val return output_val
@staticmethod @staticmethod
def numpy_max_pool_2d_stride(input, ds, ignore_border=False, st=None, def numpy_max_pool_2d_stride(input, ws, ignore_border=False, stride=None,
mode='max'): mode='max'):
'''Helper function, implementing pool_2d in pure numpy '''Helper function, implementing pool_2d in pure numpy
this function provides st input to indicate the stide size this function provides stride input to indicate the stride size
for the pooling regions. if not indicated, st == sd.''' for the pooling regions. if not indicated, stride == sd.'''
if len(input.shape) < 2: if len(input.shape) < 2:
raise NotImplementedError('input should have at least 2 dim,' raise NotImplementedError('input should have at least 2 dim,'
' shape is %s' ' shape is %s'
% str(input.shape)) % str(input.shape))
if st is None: if stride is None:
st = ds stride = ws
img_rows = input.shape[-2] img_rows = input.shape[-2]
img_cols = input.shape[-1] img_cols = input.shape[-1]
out_r = 0 out_r = 0
out_c = 0 out_c = 0
if img_rows - ds[0] >= 0: if img_rows - ws[0] >= 0:
out_r = (img_rows - ds[0]) // st[0] + 1 out_r = (img_rows - ws[0]) // stride[0] + 1
if img_cols - ds[1] >= 0: if img_cols - ws[1] >= 0:
out_c = (img_cols - ds[1]) // st[1] + 1 out_c = (img_cols - ws[1]) // stride[1] + 1
if not ignore_border: if not ignore_border:
if out_r > 0: if out_r > 0:
if img_rows - ((out_r - 1) * st[0] + ds[0]) > 0: if img_rows - ((out_r - 1) * stride[0] + ws[0]) > 0:
rr = img_rows - out_r * st[0] rr = img_rows - out_r * stride[0]
if rr > 0: if rr > 0:
out_r += 1 out_r += 1
else: else:
if img_rows > 0: if img_rows > 0:
out_r += 1 out_r += 1
if out_c > 0: if out_c > 0:
if img_cols - ((out_c - 1) * st[1] + ds[1]) > 0: if img_cols - ((out_c - 1) * stride[1] + ws[1]) > 0:
cr = img_cols - out_c * st[1] cr = img_cols - out_c * stride[1]
if cr > 0: if cr > 0:
out_c += 1 out_c += 1
else: else:
...@@ -254,35 +254,35 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -254,35 +254,35 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
output_val = numpy.zeros(out_shp) output_val = numpy.zeros(out_shp)
for k in numpy.ndindex(*input.shape[:-2]): for k in numpy.ndindex(*input.shape[:-2]):
for i in range(output_val.shape[-2]): for i in range(output_val.shape[-2]):
ii_st = i * st[0] ii_st = i * stride[0]
ii_end = builtins.min(ii_st + ds[0], img_rows) ii_end = builtins.min(ii_st + ws[0], img_rows)
for j in range(output_val.shape[-1]): for j in range(output_val.shape[-1]):
jj_st = j * st[1] jj_st = j * stride[1]
jj_end = builtins.min(jj_st + ds[1], img_cols) jj_end = builtins.min(jj_st + ws[1], img_cols)
patch = input[k][ii_st:ii_end, jj_st:jj_end] patch = input[k][ii_st:ii_end, jj_st:jj_end]
output_val[k][i, j] = func(patch) output_val[k][i, j] = func(patch)
return output_val return output_val
@staticmethod @staticmethod
def numpy_max_pool_nd_stride(input, ds, ignore_border=False, st=None, def numpy_max_pool_nd_stride(input, ws, ignore_border=False, stride=None,
mode='max'): mode='max'):
'''Helper function, implementing pooling in pure numpy '''Helper function, implementing pooling in pure numpy
this function provides st input to indicate the stide size this function provides stride input to indicate the stide size
for the pooling regions. if not indicated, st == sd.''' for the pooling regions. if not indicated, stride == sd.'''
nd = len(ds) nd = len(ws)
if st is None: if stride is None:
st = ds stride = ws
assert len(st) == len(ds) assert len(stride) == len(ws)
out_shp = list(input.shape[:-nd]) out_shp = list(input.shape[:-nd])
for i in range(nd): for i in range(nd):
out = 0 out = 0
if input.shape[-nd + i] - ds[i] >= 0: if input.shape[-nd + i] - ws[i] >= 0:
out = (input.shape[-nd + i] - ds[i]) // st[i] + 1 out = (input.shape[-nd + i] - ws[i]) // stride[i] + 1
if not ignore_border: if not ignore_border:
if out > 0: if out > 0:
if input.shape[-nd + i] - ((out - 1) * st[i] + ds[i]) > 0: if input.shape[-nd + i] - ((out - 1) * stride[i] + ws[i]) > 0:
if input.shape[-nd + i] - out * st[i] > 0: if input.shape[-nd + i] - out * stride[i] > 0:
out += 1 out += 1
else: else:
if input.shape[-nd + i] > 0: if input.shape[-nd + i] > 0:
...@@ -300,8 +300,8 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -300,8 +300,8 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
for r in numpy.ndindex(*output_val.shape[-nd:]): for r in numpy.ndindex(*output_val.shape[-nd:]):
region = [] region = []
for i in range(nd): for i in range(nd):
r_st = r[i] * st[i] r_st = r[i] * stride[i]
r_end = builtins.min(r_st + ds[i], input.shape[-nd + i]) r_end = builtins.min(r_st + ws[i], input.shape[-nd + i])
region.append(slice(r_st, r_end)) region.append(slice(r_st, r_end))
patch = input[l][region] patch = input[l][region]
output_val[l][r] = func(patch) output_val[l][r] = func(patch)
...@@ -697,6 +697,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -697,6 +697,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
grad_shape = Pool.out_shape( grad_shape = Pool.out_shape(
imval.shape, maxpoolshp, ndim=len(maxpoolshp), imval.shape, maxpoolshp, ndim=len(maxpoolshp),
ignore_border=ignore_border, st=stride) ignore_border=ignore_border, st=stride)
grad_shape = Pool.out_shape(
imval.shape, maxpoolshp, ndim=len(maxpoolshp),
ignore_border=ignore_border, stride=stride)
# skip the grad verification when the output is empty # skip the grad verification when the output is empty
if numpy.prod(grad_shape) != 0: if numpy.prod(grad_shape) != 0:
...@@ -728,7 +731,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -728,7 +731,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
grad_shape = Pool.out_shape( grad_shape = Pool.out_shape(
imval.shape, avgpoolshp, imval.shape, avgpoolshp,
ndim=len(avgpoolshp), ndim=len(avgpoolshp),
ignore_border=ignore_border, st=stride) ignore_border=ignore_border, stride=stride)
# skip the grad verification when the output is empty # skip the grad verification when the output is empty
if numpy.prod(grad_shape) != 0: if numpy.prod(grad_shape) != 0:
...@@ -765,9 +768,15 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -765,9 +768,15 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
grad_shape = Pool.out_shape(imval.shape, grad_shape = Pool.out_shape(imval.shape,
maxpoolshp, maxpoolshp,
ndim=len(maxpoolshp), ndim=len(maxpoolshp),
st=stridesize, stride=stridesize,
ignore_border=True, ignore_border=True,
padding=paddingsize) padding=paddingsize)
grad_shape = Pool.out_shape(imval.shape,
maxpoolshp,
ndim=len(maxpoolshp),
stride=stridesize,
ignore_border=True,
pad=paddingsize)
grad_val = rng.rand(*grad_shape) * 10.0 grad_val = rng.rand(*grad_shape) * 10.0
def mp(input, grad): def mp(input, grad):
...@@ -804,9 +813,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -804,9 +813,9 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
grad_shape = Pool.out_shape(imval.shape, grad_shape = Pool.out_shape(imval.shape,
avgpoolshp, avgpoolshp,
ndim=len(avgpoolshp), ndim=len(avgpoolshp),
st=stridesize, stride=stridesize,
ignore_border=True, ignore_border=True,
padding=paddingsize) pad=paddingsize)
grad_val = rng.rand(*grad_shape) * 10.0 grad_val = rng.rand(*grad_shape) * 10.0
def mp(input, grad): def mp(input, grad):
...@@ -823,6 +832,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester): ...@@ -823,6 +832,7 @@ class TestDownsampleFactorMax(utt.InferShapeTester):
z = tensor.dot(x_vec.dimshuffle(0, 'x'), z = tensor.dot(x_vec.dimshuffle(0, 'x'),
x_vec.dimshuffle('x', 0)) x_vec.dimshuffle('x', 0))
y = pool_2d(input=z, ds=(2, 2), ignore_border=True) y = pool_2d(input=z, ds=(2, 2), ignore_border=True)
y = pool_2d(input=z, ws=(2, 2), ignore_border=True)
C = tensor.exp(tensor.sum(y)) C = tensor.exp(tensor.sum(y))
grad_hess = tensor.hessian(cost=C, wrt=x_vec) grad_hess = tensor.hessian(cost=C, wrt=x_vec)
......
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