def batches(self):
x = ca.empty(self.x_shape, dtype=self.x.dtype)
for start, stop in self._batch_slices():
if stop > start:
x_np = self.x[start:stop]
else:
x_np = np.concatenate((self.x[start:], self.x[:stop]))
ca.copyto(x, x_np)
yield x,
def batches(self):
x = ca.empty(self.x_shape, dtype=self.x.dtype)
for start, stop in self._batch_slices():
if stop > start:
x_np = self.x[start:stop]
else:
x_np = np.concatenate((self.x[start:], self.x[:stop]))
ca.copyto(x, x_np)
yield x,
def batches(self):
x = ca.empty(self.x_shape, dtype=dp.float_)
for start, stop in self._batch_slices():
if stop > start:
x_np = self.x[start:stop]
else:
x_np = np.concatenate((self.x[start:], self.x[:stop]))
if random.randint(0, 1) == 0:
x_np = x_np[:, :, :, ::-1]
x_np = img_transform(x_np, to_bc01=False)
x_np = np.ascontiguousarray(x_np)
ca.copyto(x, x_np)
yield {'x': x}
def batches(self):
x = ca.empty(self.x_shape, dtype=dp.float_)
for start, stop in self._batch_slices():
if stop > start:
x_np = self.x[start:stop]
else:
x_np = np.concatenate((self.x[start:], self.x[:stop]))
if random.randint(0, 1) == 0:
x_np = x_np[:, :, :, ::-1]
x_np = img_transform(x_np, to_bc01=False)
x_np = np.ascontiguousarray(x_np)
ca.copyto(x, x_np)
yield x,
def batches(self):
x1 = ca.empty(self.x_shape, dtype=self.x.dtype)
x2 = ca.empty_like(x1)
for start, stop in self._batch_slices():
if stop > start:
x1_np = self.x[start:stop]
x2_np = self.x2[start:stop]
else:
x1_np = np.concatenate((self.x[start:], self.x[:stop]))
x2_np = np.concatenate((self.x[start:], self.x[:stop]))
ca.copyto(x1, x1_np)
ca.copyto(x2, x2_np)
yield x1, x2
def batches(self):
x1 = ca.empty(self.x_shape, dtype=self.x.dtype)
x2 = ca.empty_like(x1)
for start, stop in self._batch_slices():
if stop > start:
x1_np = self.x[start:stop]
x2_np = self.x2[start:stop]
else:
x1_np = np.concatenate((self.x[start:], self.x[:stop]))
x2_np = np.concatenate((self.x[start:], self.x[:stop]))
ca.copyto(x1, x1_np)
ca.copyto(x2, x2_np)
yield x1, x2
def _update(self):
# Forward propagation
next_x = self.x.array
x_feats = [None]*len(self.layers)
x_grams = [None]*len(self.layers)
for l, layer in enumerate(self.layers):
next_x = layer.fprop(next_x)
if self.subject_weights[l] > 0:
x_feats[l] = next_x
if self.style_weights[l] > 0:
x_feats[l] = next_x
x_grams[l] = gram_matrix(next_x)
# Backward propagation
grad = ca.zeros_like(next_x)
loss = ca.zeros(1)
for l, layer in reversed(list(enumerate(self.layers))):
if self.subject_weights[l] > 0:
diff = x_feats[l] - self.subject_feats[l]
norm = ca.sum(ca.fabs(diff)) + 1e-8
weight = float(self.subject_weights[l]) / norm
grad += diff * weight
loss += 0.5*weight*ca.sum(diff**2)
if self.style_weights[l] > 0:
diff = x_grams[l] - self.style_grams[l]
n_channels = diff.shape[0]
x_feat = ca.reshape(x_feats[l], (n_channels, -1))
style_grad = ca.reshape(ca.dot(diff, x_feat), x_feats[l].shape)
norm = ca.sum(ca.fabs(style_grad))
weight = float(self.style_weights[l]) / norm
style_grad *= weight
grad += style_grad
loss += 0.25*weight*ca.sum(diff**2)
grad = layer.bprop(grad)
if self.tv_weight > 0:
x = ca.reshape(self.x.array, (3, 1) + grad.shape[2:])
tv = self.tv_conv.fprop(x, self.tv_kernel)
tv *= self.tv_weight
grad -= ca.reshape(tv, grad.shape)
ca.copyto(self.x.grad_array, grad)
return loss
def _update(self):
# Forward propagation
next_x = self.x.array
x_feats = [None] * len(self.layers)
for l, layer in enumerate(self.layers):
next_x = layer.fprop(next_x)
if self.subject_weights[l] > 0 or self.style_weights[l] > 0:
x_feats[l] = next_x
# Backward propagation
grad = ca.zeros_like(next_x)
loss = ca.zeros(1)
for l, layer in reversed(list(enumerate(self.layers))):
if self.subject_weights[l] > 0:
diff = x_feats[l] - self.subject_feats[l]
norm = ca.sum(ca.fabs(diff)) + 1e-8
weight = float(self.subject_weights[l]) / norm
grad += diff * weight
loss += 0.5 * weight * ca.sum(diff**2)
if self.style_weights[l] > 0:
diff = gram_matrix(x_feats[l]) - self.style_grams[l]
n_channels = diff.shape[0]
x_feat = ca.reshape(x_feats[l], (n_channels, -1))
style_grad = ca.reshape(ca.dot(diff, x_feat), x_feats[l].shape)
norm = ca.sum(ca.fabs(style_grad))
weight = float(self.style_weights[l]) / norm
style_grad *= weight
grad += style_grad
loss += 0.25 * weight * ca.sum(diff**2)
grad = layer.bprop(grad)
if self.tv_weight > 0:
x = ca.reshape(self.x.array, (3, 1) + grad.shape[2:])
tv = self.tv_conv.fprop(x, self.tv_kernel)
tv *= self.tv_weight
grad -= ca.reshape(tv, grad.shape)
ca.copyto(self.x.grad_array, grad)
return loss
def test_copyto():
a_np = np.random.random(size=(7, 11))
a_ca = ca.array(a_np)
b_np = np.zeros_like(a_np)
b_ca = np.zeros_like(a_np)
ca.copyto(b_np, a_ca)
print(np.allclose(a_np, b_np))
ca.copyto(b_ca, a_np)
print(np.allclose(np.array(a_ca), np.array(b_ca)))
ca.copyto(b_ca, a_ca)
print(np.allclose(np.array(a_ca), np.array(b_ca)))
def test_copyto():
a_np = np.random.random(size=(7, 11))
a_ca = ca.array(a_np)
b_np = np.zeros_like(a_np)
b_ca = np.zeros_like(a_np)
ca.copyto(b_np, a_ca)
print(np.allclose(a_np, b_np))
ca.copyto(b_ca, a_np)
print(np.allclose(np.array(a_ca), np.array(b_ca)))
ca.copyto(b_ca, a_ca)
print(np.allclose(np.array(a_ca), np.array(b_ca)))
def batches(self):
x1 = ca.empty(self.x_shape, dtype=self.x.dtype)
x2 = ca.empty_like(x1)
y = ca.empty(self.y_shape, dtype=self.y.dtype)
for start, stop in self._batch_slices():
if stop > start:
x1_np = self.x[start:stop]
x2_np = self.x2[start:stop]
y_np = self.y[start:stop]
else:
x1_np = np.concatenate((self.x[start:], self.x[:stop]))
x2_np = np.concatenate((self.x[start:], self.x[:stop]))
y_np = np.concatenate((self.y[start:], self.y[:stop]))
ca.copyto(x1, x1_np)
ca.copyto(x2, x2_np)
ca.copyto(y, y_np)
yield {'x1': x1, 'x2': x2, 'y': y}
def batches(self):
x1 = ca.empty(self.x_shape, dtype=self.x.dtype)
x2 = ca.empty_like(x1)
y = ca.empty(self.y_shape, dtype=self.y.dtype)
for start, stop in self._batch_slices():
if stop > start:
x1_np = self.x[start:stop]
x2_np = self.x2[start:stop]
y_np = self.y[start:stop]
else:
x1_np = np.concatenate((self.x[start:], self.x[:stop]))
x2_np = np.concatenate((self.x[start:], self.x[:stop]))
y_np = np.concatenate((self.y[start:], self.y[:stop]))
ca.copyto(x1, x1_np)
ca.copyto(x2, x2_np)
ca.copyto(y, y_np)
yield {'x1': x1, 'x2': x2, 'y': y}
def bprop(self):
for i, (start, end) in enumerate(self.slices):
ca.copyto(self.x.out_grad[start:end, :], self.outputs[i].out_grad)
def bprop(self):
for i, (start, end) in enumerate(self.slices):
ca.copyto(self.x.out_grad[start:end, :], self.outputs[i].out_grad)
def bprop(self):
for i in range(self.n_sources):
ca.copyto(self.inputs[i].out_grad, self.out_grad[i])
def fprop(self):
for i in range(self.n_sources):
ca.copyto(self.out[i], self.inputs[i].out)
def bprop(self):
for i in range(self.n_splits):
ca.copyto(self.x.out_grad[i], self.outputs[i].out_grad)
def bprop(self):
for i in range(self.n_sources):
ca.copyto(self.inputs[i].out_grad, self.out_grad[i])
def bprop(self):
ca.copyto(self.parameter.grad_array, self.out_grad)
def fprop(self):
for i in range(self.n_sources):
ca.copyto(self.array[i], self.inputs[i].array)
def fprop(self):
for i in range(self.n_sources):
ca.copyto(self.out[i], self.inputs[i].out)
def bprop(self):
ca.copyto(self.x.out_grad, self.outputs[0].out_grad)
for i in range(1, self.n_splits):
self.x.out_grad += self.outputs[i].out_grad
def bprop(self):
for i in range(self.n_splits):
ca.copyto(self.x.out_grad[i], self.outputs[i].out_grad)
def bprop(self):
ca.copyto(self.x.grad_array, self.outputs[0].grad_array)
for i in range(1, self.n_splits):
self.x.grad_array += self.outputs[i].grad_array
def bprop(self):
for i, (start, end) in enumerate(self.slices):
ca.copyto(self.x.grad_array[start:end], self.outputs[i].grad_array)
def bprop(self):
for i in range(self.n_sources):
ca.copyto(self.inputs[i].grad_array, self.grad_array[i])
def bprop(self):
ca.copyto(self.x.out_grad, self.outputs[0].out_grad)
for i in range(1, self.n_splits):
self.x.out_grad += self.outputs[i].out_grad
def bprop(self):
for i in range(self.n_splits):
ca.copyto(self.x.grad_array[i], self.outputs[i].grad_array)
def bprop(self):
ca.copyto(self.parameter.grad_array, self.out_grad)
def bprop(self):
ca.copyto(self.x.grad_array, self.outputs[0].grad_array)
for i in range(1, self.n_splits):
self.x.grad_array += self.outputs[i].grad_array
