Hi, guys… recently I am learning Pytorch C++ API and trying to reimplement the WGAN-GP in C++.
But we need to calculate the second order gradient. In the python version it’s quite simple.
def calc_gradient_penalty(netD, real_data, fake_data):
#print real_data.size()
alpha = torch.rand(BATCH_SIZE, 1)
alpha = alpha.expand(real_data.size())
alpha = alpha.cuda(gpu) if use_cuda else alpha
interpolates = alpha * real_data + ((1 - alpha) * fake_data)
if use_cuda:
interpolates = interpolates.cuda(gpu)
interpolates = autograd.Variable(interpolates, requires_grad=True)
disc_interpolates = netD(interpolates)
gradients = autograd.grad(outputs=disc_interpolates, inputs=interpolates,
grad_outputs=torch.ones(disc_interpolates.size()).cuda(gpu) if use_cuda else
torch.ones(
disc_interpolates.size()),
create_graph=True, retain_graph=True, only_inputs=True)[0]
gradient_penalty = ((gradients.norm(2, dim=1) - 1) ** 2).mean() * LAMBDA
return gradient_penalty
BUT, there is not such method torch::grad
in the C++ world.
Is there anyway can achieve the same approach?