def kflr(self):
with backpack(new_ext.KFLR()):
_, _, loss = self.problem.forward_pass()
loss.backward()
kflr = [p.kflr for p in self.problem.model.parameters()]
return kflr
def test_interface_kflr():
interface_test(new_ext.KFLR())
def test_interface_kflr_conv():
interface_test(new_ext.KFLR(), use_conv=True)
# %%
# We can now evaluate the loss and do a backward pass with Backpack
# -----------------------------------------------------------------
loss = lossfunc(model(X), y)
with backpack(
extensions.BatchGrad(),
extensions.Variance(),
extensions.SumGradSquared(),
extensions.BatchL2Grad(),
extensions.DiagGGNMC(mc_samples=1),
extensions.DiagGGNExact(),
extensions.DiagHessian(),
extensions.KFAC(mc_samples=1),
extensions.KFLR(),
extensions.KFRA(),
):
loss.backward()
# %%
# And here are the results
# -----------------------------------------------------------------
for name, param in model.named_parameters():
print(name)
print(".grad.shape: ", param.grad.shape)
print(".grad_batch.shape: ", param.grad_batch.shape)
print(".variance.shape: ", param.variance.shape)
print(".sum_grad_squared.shape: ", param.sum_grad_squared.shape)
print(".batch_l2.shape: ", param.batch_l2.shape)
"""
Compute the gradient with PyTorch and the KFLR approximation with BackPACK.
"""
from torch.nn import CrossEntropyLoss, Flatten, Linear, Sequential
from backpack import backpack, extend, extensions
from backpack.utils.examples import load_mnist_data
B = 4
X, y = load_mnist_data(B)
print("# Gradient with PyTorch, KFLR approximation with BackPACK | B =", B)
model = Sequential(Flatten(), Linear(784, 10),)
lossfunc = CrossEntropyLoss()
model = extend(model)
lossfunc = extend(lossfunc)
loss = lossfunc(model(X), y)
with backpack(extensions.KFLR()):
loss.backward()
for name, param in model.named_parameters():
print(name)
print(".grad.shape: ", param.grad.shape)
print(".kflr (shapes): ", [kflr.shape for kflr in param.kflr])
def kflr(self) -> List[List[Tensor]]: # noqa:D102
with backpack(new_ext.KFLR()):
_, _, loss = self.problem.forward_pass()
loss.backward()
return self.problem.collect_data("kflr")