Last active
December 27, 2023 06:12
-
-
Save seanbenhur/55ba0f278beb6da3d5ffd7b0d38573f8 to your computer and use it in GitHub Desktop.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import torch | |
class SAM(torch.optim.Optimizer): | |
def __init__(self, params, base_optimizer, rho=0.05, **kwargs): | |
assert rho >= 0.0, f"Invalid rho, should be non-negative: {rho}" | |
defaults = dict(rho=rho, **kwargs) | |
super(SAM, self).__init__(params, defaults) | |
self.base_optimizer = base_optimizer(self.param_groups, **kwargs) | |
self.param_groups = self.base_optimizer.param_groups | |
@torch.no_grad() | |
def first_step(self, zero_grad=False): | |
grad_norm = self._grad_norm() | |
for group in self.param_groups: | |
scale = group["rho"] / (grad_norm + 1e-12) | |
for p in group["params"]: | |
if p.grad is None: continue | |
e_w = p.grad * scale.to(p) | |
p.add_(e_w) # climb to the local maximum "w + e(w)" | |
self.state[p]["e_w"] = e_w | |
if zero_grad: self.zero_grad() | |
@torch.no_grad() | |
def second_step(self, zero_grad=False): | |
for group in self.param_groups: | |
for p in group["params"]: | |
if p.grad is None: continue | |
p.sub_(self.state[p]["e_w"]) # get back to "w" from "w + e(w)" | |
self.base_optimizer.step() # do the actual "sharpness-aware" update | |
if zero_grad: self.zero_grad() | |
def _grad_norm(self): | |
shared_device = self.param_groups[0]["params"][0].device # put everything on the same device, in case of model parallelism | |
norm = torch.norm( | |
torch.stack([ | |
p.grad.norm(p=2).to(shared_device) | |
for group in self.param_groups for p in group["params"] | |
if p.grad is not None | |
]), | |
p=2 | |
) | |
return norm |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment