本文整理汇总了Python中torch.autograd.Variable.requires_grad方法的典型用法代码示例。如果您正苦于以下问题：Python Variable.requires_grad方法的具体用法？Python Variable.requires_grad怎么用？Python Variable.requires_grad使用的例子？那么恭喜您, 这里精选的方法代码示例或许可以为您提供帮助。您也可以进一步了解该方法所在类torch.autograd.Variable的用法示例。
在下文中一共展示了Variable.requires_grad方法的3个代码示例，这些例子默认根据受欢迎程度排序。您可以为喜欢或者感觉有用的代码点赞，您的评价将有助于系统推荐出更棒的Python代码示例。
示例1: reconstruct_cells_red_first
# 需要导入模块: from torch.autograd import Variable [as 别名]
# 或者: from torch.autograd.Variable import requires_grad [as 别名]
def reconstruct_cells_red_first(imgs, netG, opt, n_bfgs_iter=100, lbfgs_lr=0.1):
assert(imgs.size(1) == 2)
n_red_noise = int(opt.nz * opt.red_portion)
n_green_noise = opt.nz - n_red_noise
red_noise = torch.FloatTensor(int(opt.batch_size), n_red_noise, 1, 1).normal_(0, 1)
green_noise = torch.FloatTensor(int(opt.batch_size), n_green_noise, 1, 1).normal_(0, 1)
if opt.cuda:
red_noise = red_noise.cuda()
green_noise = green_noise.cuda()
red_noise = Variable(red_noise)
red_noise.requires_grad = True
green_noise = Variable(green_noise)
green_noise.requires_grad = False
noise = torch.cat([red_noise, green_noise], 1)
noise_init = noise.clone()
optim_input = optim.LBFGS([red_noise], lr=lbfgs_lr)
def red_closure():
optim_input.zero_grad()
noise = torch.cat([red_noise, green_noise], 1)
gen_img = netG(noise)
l2_loss = torch.mean((imgs[:, 0, :, :] - gen_img[:, 0, :, :]) ** 2)
l2_loss.backward()
# print(l2_loss.data[0])
# sys.stdout.flush()
return l2_loss
# Do the optimization across batch
for i in tqdm(range(n_bfgs_iter)):
optim_input.step(red_closure)
# Optimization across green channel now
red_noise.requires_grad = False
green_noise.requires_grad = True
optim_input = optim.LBFGS([green_noise], lr=lbfgs_lr)
def green_closure():
optim_input.zero_grad()
noise = torch.cat([red_noise, green_noise], 1)
gen_img = netG(noise)
l2_loss = torch.mean((imgs[:, 1, :, :] - gen_img[:, 1, :, :]) ** 2)
l2_loss.backward()
# print(l2_loss.data[0])
# sys.stdout.flush()
return l2_loss
for i in tqdm(range(n_bfgs_iter)):
optim_input.step(green_closure)
# do not forget this because real variables are red_noise and green_noise
noise = torch.cat([red_noise, green_noise], 1)
return noise_init, noise示例2: reconstruct_cells
# 需要导入模块: from torch.autograd import Variable [as 别名]
# 或者: from torch.autograd.Variable import requires_grad [as 别名]
def reconstruct_cells(imgs, netG, opt, n_bfgs_iter=100, lbfgs_lr=0.1):
noise = torch.FloatTensor(int(opt.batch_size), opt.nz, 1, 1)
noise.normal_(0, 1)
if opt.cuda:
noise = noise.cuda()
noise = Variable(noise)
noise.requires_grad = True
noise_init = noise.clone()
optim_input = optim.LBFGS([noise], lr=lbfgs_lr)
def closure():
optim_input.zero_grad()
gen_img = netG(noise)
l2_loss = torch.mean((imgs - gen_img) ** 2)
l2_loss.backward()
# print(l2_loss.data[0])
# sys.stdout.flush()
return l2_loss
# Do the optimization across batch
for i in tqdm(range(n_bfgs_iter)):
optim_input.step(closure)
return noise_init, noise示例3: test_flags
# 需要导入模块: from torch.autograd import Variable [as 别名]
# 或者: from torch.autograd.Variable import requires_grad [as 别名]
def test_flags(self):
x = Variable(torch.randn(2, 2))
y = Variable(torch.randn(2, 2))
@torch.jit.compile
def fn(x, y):
return (x * x + y * y + x * y).sum()
grads = {}
for rx, ry in product((True, False), repeat=2):
x.requires_grad = rx
y.requires_grad = ry
self.assertFalse(fn.has_trace_for(x, y))
out = fn(x, y)
self.assertFalse(fn.has_trace_for(x, y))
for v, name, compute in [(x, 'x', rx), (y, 'y', ry)]:
if not compute:
continue
grad_v, = torch.autograd.grad(out, v, retain_graph=True)
expected_grad = grads.setdefault(name, grad_v)
self.assertEqual(grad_v, expected_grad)
self.assertEqual(fn.has_trace_for(x, y), rx or ry)