def ConvOuter(self, grad):
w = self.params['weight']
conv = self.conv_params
size = conv.size
stride = conv.stride
padding = conv.padding
num_filters = conv.num_filters
num_colors = conv.num_colors
f, numdims = w.shape
assert f == num_filters, 'f is %d but num_filters is %d' % (f, num_filters)
assert numdims == size**2 * num_colors
input_t = self.input_t
if conv.max_pool:
output_t = self.unpooled_layer
else:
output_t = self.output_t
numimages, numdims = input_t.shape
assert numdims % num_colors == 0
x = int(np.sqrt(numdims / num_colors))
assert x**2 == numdims/num_colors
n_locs = (x + 2 * padding - size) / stride + 1
self.node1.state.transpose(input_t)
cc.convOutp(input_t, output_t, grad, n_locs, size, stride, num_colors)
def ConvOuter(edge, grad):
"""Get the gradient for the weights in this edge.
Args:
grad: (output) the gradient for the weights in this edge.
"""
w = edge.params['weight']
conv = edge.conv_params
size = conv.size
stride = conv.stride
padding = conv.padding
num_filters = conv.num_filters
num_colors = conv.num_colors
f, numdims = w.shape
assert f == num_filters, 'f is %d but num_filters is %d' % (f, num_filters)
if edge.conv:
assert numdims == size**2 * num_colors
input_t = edge.input_t
if conv.max_pool:
output_t = edge.unpooled_layer
elif conv.rnorm:
output_t = edge.rnorm_temp2
else:
output_t = edge.output_t2
numdims, numimages = edge.node1.state.shape
assert numdims % num_colors == 0
x = int(math.sqrt(numdims / num_colors))
assert x**2 == numdims/num_colors
n_locs = (x + 2 * padding - size) / stride + 1
if edge.conv:
cc.convOutp(input_t, output_t, grad, n_locs, padding, size, stride, num_colors)
else:
cc.localOutp(input_t, output_t, grad, n_locs, padding, size, stride, num_colors)
