def dconvw(img, kern, out):
desc = dnn.GpuDnnConvDesc(border_mode='valid',
subsample=(1, 1),
conv_mode='conv')(img.shape, kern.shape)
return dnn.GpuDnnConvGradW()(img,
out,
kern,
desc,
alpha=0.75,
beta=-1.0)
def test_dnn_conv_desc_merge():
img_shp = T.as_tensor_variable(numpy.asarray([2, 1, 8, 8]).astype('int64'))
kern_shp = T.as_tensor_variable(
numpy.asarray([3, 1, 2, 2]).astype('int64'))
desc1 = dnn.GpuDnnConvDesc(border_mode='valid',
subsample=(2, 2),
conv_mode='conv')(img_shp, kern_shp)
desc2 = dnn.GpuDnnConvDesc(border_mode='full',
subsample=(1, 1),
conv_mode='cross')(img_shp, kern_shp)
# CDataType is not DeepCopyable so this will crash if we don't use
# borrow=True
f = theano.function(
[], [theano.Out(desc1, borrow=True),
theano.Out(desc2, borrow=True)])
d1, d2 = f()
# This will be the case if they are merged, which would be bad.
assert d1 != d2
def test_dnn_conv_desc_merge():
if not cuda.dnn.dnn_available():
raise SkipTest(cuda.dnn.dnn_available.msg)
img_shp = T.as_tensor_variable(
numpy.asarray([2, 1, 8, 8]).astype('int64'))
kern_shp = T.as_tensor_variable(
numpy.asarray([3, 1, 2, 2]).astype('int64'))
desc1 = dnn.GpuDnnConvDesc(border_mode='valid', subsample=(2, 2),
conv_mode='conv')(img_shp, kern_shp)
desc2 = dnn.GpuDnnConvDesc(border_mode='full', subsample=(1, 1),
conv_mode='cross')(img_shp, kern_shp)
# CDataType is not DeepCopyable so this will crash if we don't use
# borrow=True
f = theano.function([], [theano.Out(desc1, borrow=True),
theano.Out(desc2, borrow=True)],
mode=mode_with_gpu)
d1, d2 = f()
# This will be the case if they are merged, which would be bad.
assert d1 != d2
desc1v2 = dnn.GpuDnnConvDesc(border_mode='valid', subsample=(2, 2),
conv_mode='conv')(img_shp, kern_shp)
f = theano.function([], [theano.Out(desc1, borrow=True),
theano.Out(desc1v2, borrow=True)],
mode=mode_with_gpu)
assert len([n for n in f.maker.fgraph.apply_nodes
if isinstance(n.op, dnn.GpuDnnConvDesc)]) == 1
# CDATA type don't equal even if they represent the same object
# So we can't use debugmode with it.
if theano.config.mode not in ["DebugMode", "DEBUG_MODE"]:
d1, d2 = f()
# They won't be equal if they aren't merged.
assert d1 == d2
def test_conv_gradw(self):
if not dnn.dnn_available():
raise SkipTest(dnn.dnn_available.msg)
img = T.ftensor4('img')
kerns = T.ftensor4('kerns')
out = T.ftensor4('out')
img_val = numpy.asarray(
numpy.random.rand(2, 5, 6, 8),
dtype='float32'
)
kern_vals = numpy.asarray(
numpy.random.rand(2, 1, 5, 6),
dtype='float32'
)
for params in product(
['valid', 'full'],
[(1, 1)], # strides besides (1, 1)
['conv', 'cross']
):
temp_img = img.dimshuffle(1, 0, 2, 3)
temp_kerns = kerns
if params[2] == 'conv':
temp_kerns = temp_kerns[:, :, ::-1, ::-1]
temp_kerns = temp_kerns.dimshuffle(1, 0, 2, 3)
shape = (
kern_vals.shape[1], img_val.shape[1],
img_val.shape[2] - kern_vals.shape[2] + 1,
img_val.shape[3] - kern_vals.shape[3] + 1
)
out_vals = numpy.zeros(shape, dtype='float32')
desc = dnn.GpuDnnConvDesc(
border_mode=params[0],
subsample=params[1],
conv_mode=params[2]
)(temp_img.shape, out.shape)
conv_grad_w = dnn.GpuDnnConvGradW()(
temp_img,
temp_kerns,
out,
desc,
)
self._compile_and_check(
[temp_img, temp_kerns, out],
[conv_grad_w],
[img_val, kern_vals, out_vals],
dnn.GpuDnnConvGradW
)
def test_conv(self):
img = T.ftensor4('img')
kerns = T.ftensor4('kerns')
img_val = numpy.asarray(numpy.random.rand(3, 4, 5, 6), dtype='float32')
kern_vals = numpy.asarray(numpy.random.rand(3, 4, 5, 6),
dtype='float32')
for params in product(['valid', 'full'], [(1, 1), (2, 2)],
['conv', 'cross']):
desc = dnn.GpuDnnConvDesc(border_mode=params[0],
subsample=params[1],
conv_mode=params[2])(img.shape,
kerns.shape)
conv = dnn.GpuDnnConv()(img_val, kern_vals, desc)
self._compile_and_check([img, kerns], [conv], [img_val, kern_vals],
dnn.GpuDnnConv)
def test_conv_gradi(self):
if not dnn.dnn_available():
raise SkipTest(dnn.dnn_available.msg)
img = T.ftensor4('img')
kerns = T.ftensor4('kerns')
out = T.ftensor4('out')
img_val = numpy.asarray(
numpy.random.rand(3, 4, 5, 6),
dtype='float32'
)
kern_vals = numpy.asarray(
numpy.random.rand(13, 14, 15, 16),
dtype='float32'
)
for params in product(
['valid'], # Should this work for 'full'?
[(1, 1)],
['conv', 'cross']
):
temp_kerns = kerns.dimshuffle(1, 0, 2, 3)
shape = (
img_val.shape[0], kern_vals.shape[1],
img_val.shape[2] + kern_vals.shape[2] - 1,
img_val.shape[3] + kern_vals.shape[3] - 1
)
out_vals = numpy.zeros(shape, dtype='float32')
desc = dnn.GpuDnnConvDesc(
border_mode=params[0],
subsample=params[1],
conv_mode=params[2]
)(out.shape, temp_kerns.shape)
conv_grad_i = dnn.GpuDnnConvGradI()(
temp_kerns,
img,
out,
desc,
)
self._compile_and_check(
[temp_kerns, img, out],
[conv_grad_i],
[kern_vals, img_val, out_vals],
dnn.GpuDnnConvGradI
)
def test_conv3d_gradi(self):
if not (cuda.dnn.dnn_available() and dnn.version() >= (2000, 2000)):
raise SkipTest('"CuDNN 3D convolution requires CuDNN v2')
ftensor5 = T.TensorType(dtype="float32", broadcastable=(False,) * 5)
img = ftensor5('img')
kerns = ftensor5('kerns')
out = ftensor5('out')
img_val = numpy.asarray(
numpy.random.rand(8, 4, 6, 7, 5),
dtype='float32'
)
kern_vals = numpy.asarray(
numpy.random.rand(9, 4, 5, 1, 2),
dtype='float32'
)
for params in product(
['valid', 'full'],
[(1, 1, 1), (2, 2, 2)],
['conv', 'cross']
):
out_vals = numpy.zeros(
dnn.GpuDnnConv3d.get_out_shape(img_val.shape, kern_vals.shape,
border_mode=params[0],
subsample=params[1]),
dtype='float32')
desc = dnn.GpuDnnConvDesc(
border_mode=params[0],
subsample=params[1],
conv_mode=params[2]
)(img.shape, kerns.shape)
conv_grad_i = dnn.GpuDnnConv3dGradI()(
kerns,
out,
img,
desc,
)
self._compile_and_check(
[kerns, out, img],
[conv_grad_i],
[kern_vals, out_vals, img_val],
dnn.GpuDnnConv3dGradI
)
def upconv(x, w, stride, x_shape=None, w_shape=None, axis_order='dnn'):
assert stride is not None
stride = tuple(stride)
conv_dim = len(stride)
border_mode = 'valid'
# if (x_shape is None) or (None in x_shape): # variable batch size or so
# x_shape = None
if conv_dim==1:
x = x.dimshuffle(0, 1, 2, 'x')
w = w.dimshuffle(0, 1, 2, 'x')
if w_shape is not None:
w_shape = list(w_shape) + [1, ]
if x_shape is not None:
x_shape = list(x_shape) + [1, ]
stride = list(stride) + [1, ]
y = conv2d_grad_wrt_inputs(x, w, x_shape, w_shape, border_mode,
subsample=stride, filter_flip=False)
y = y[:, :, :, 0]
elif conv_dim==2:
y = conv2d_grad_wrt_inputs(x, w, x_shape, w_shape, border_mode,
subsample=stride, filter_flip=False)
elif conv_dim==3:
if not dnn_avail or axis_order!='dnn':
raise ValueError("Need dnn and dnn axis order")
kerns = dnn.gpu_contiguous(w)
image = dnn.gpu_contiguous(x)
k = kerns.shape[1]
img_sh = list(image.shape)
out_sh = img_sh[:1] + [k,] + [st*sh for st, sh in zip(stride, img_sh[2:])]
out = dnn.gpu_alloc_empty(*out_sh)
desc = dnn.GpuDnnConvDesc(border_mode='valid', subsample=stride,
conv_mode='cross')(out.shape, kerns.shape)
y = dnn.GpuDnnConv3dGradI()(kerns, image, out, desc)
return y
def test_dnn_conv_merge():
"""This test that we merge correctly multiple dnn_conv.
This can is more difficult due to GpuEmptyAlloc that aren't
merged.
"""
if not cuda.dnn.dnn_available():
raise SkipTest(cuda.dnn.dnn_available.msg)
img_shp = [2, 5, 6, 8]
kern_shp = [3, 5, 5, 6]
img = T.ftensor4('img')
kern = T.ftensor4('kern')
out = T.ftensor4('out')
desc = dnn.GpuDnnConvDesc(
border_mode='valid')(img.shape, kern.shape)
# Test forward op
o1 = dnn.dnn_conv(img, kern)
o2 = dnn.dnn_conv(img, kern)
f = theano.function([img, kern], [o1, o2], mode=mode_with_gpu)
d1, d2 = f(numpy.random.rand(*img_shp).astype('float32'),
numpy.random.rand(*kern_shp).astype('float32'))
topo = f.maker.fgraph.toposort()
assert len([n for n in topo if isinstance(n.op, dnn.GpuDnnConv)]) == 1
# Test grad w op
o1 = dnn.GpuDnnConvGradW()(img, kern, out, desc)
o2 = dnn.GpuDnnConvGradW()(img, kern, out, desc)
f = theano.function([img, kern, out], [o1, o2], mode=mode_with_gpu)
topo = f.maker.fgraph.toposort()
assert len([n for n in topo if isinstance(n.op, dnn.GpuDnnConvGradW)]) == 1
# Test grad i op
o1 = dnn.GpuDnnConvGradI()(img, kern, out, desc)
o2 = dnn.GpuDnnConvGradI()(img, kern, out, desc)
f = theano.function([img, kern, out], [o1, o2], mode=mode_with_gpu)
topo = f.maker.fgraph.toposort()
assert len([n for n in topo if isinstance(n.op, dnn.GpuDnnConvGradI)]) == 1
def test_conv(self):
if not dnn.dnn_available():
raise SkipTest(dnn.dnn_available.msg)
img = T.ftensor4('img')
kerns = T.ftensor4('kerns')
out = T.ftensor4('out')
img_val = numpy.asarray(
numpy.random.rand(7, 2, 6, 4),
dtype='float32'
)
kern_vals = numpy.asarray(
numpy.random.rand(8, 2, 4, 3),
dtype='float32'
)
for params in product(
['valid', 'full'],
[(1, 1), (2, 2)],
['conv', 'cross']
):
out_vals = numpy.zeros(
dnn.GpuDnnConv.get_out_shape(img_val.shape, kern_vals.shape,
border_mode=params[0],
subsample=params[1]),
dtype='float32')
desc = dnn.GpuDnnConvDesc(
border_mode=params[0],
subsample=params[1],
conv_mode=params[2]
)(img.shape, kerns.shape)
conv = dnn.GpuDnnConv()(img, kerns, out, desc)
self._compile_and_check(
[img, kerns, out],
[conv],
[img_val, kern_vals, out_vals],
dnn.GpuDnnConv
)
def test_conv_gradi(self):
img = T.ftensor4('img')
kerns = T.ftensor4('kerns')
img_val = numpy.asarray(numpy.random.rand(3, 4, 5, 6), dtype='float32')
kern_vals = numpy.asarray(numpy.random.rand(3, 4, 5, 6),
dtype='float32')
for params in product(
['valid'], # Should this work for 'full'?
[(1, 1)],
['conv', 'cross']):
print params
temp_kerns = kerns.dimshuffle(1, 0, 2, 3)
shape = theano.tensor.stack(img.shape[0], temp_kerns.shape[1],
img.shape[2] + temp_kerns.shape[2] - 1,
img.shape[3] + temp_kerns.shape[3] - 1)
desc = dnn.GpuDnnConvDesc(border_mode=params[0],
subsample=params[1],
conv_mode=params[2])(shape,
temp_kerns.shape)
conv_grad_i = dnn.GpuDnnConvGradI()(temp_kerns, img, desc,
shape[2], shape[3])
self._compile_and_check([temp_kerns, img], [conv_grad_i],
[kern_vals, img_val], dnn.GpuDnnConvGradI)
