Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

add option to clip grad by value #810

Closed
wants to merge 3 commits into from
Closed

add option to clip grad by value #810

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
925dca9
add option to clip grad by value
misko Aug 14, 2024
d64974f
Merge branch 'main' into add_clip_grad_value_option
misko Aug 15, 2024
6bb0403
Merge branch 'main' into add_clip_grad_value_option
misko Aug 21, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion src/fairchem/core/models/equiformer_v2/trainers/forces_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -59,7 +59,8 @@ def multiply(obj, num):
]
self.scheduler = LRScheduler(self.optimizer, self.config["optim"])

self.clip_grad_norm = self.config["optim"].get("clip_grad_norm")
self.clip_grad_norm = self.config["optim"].get("clip_grad_norm",None)
self.clip_grad_value = self.config["optim"].get("clip_grad_value",None)
self.ema_decay = self.config["optim"].get("ema_decay")
if self.ema_decay:
self.ema = ExponentialMovingAverage(
Expand Down
20 changes: 13 additions & 7 deletions src/fairchem/core/trainers/base_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -702,6 +702,9 @@ def load_extras(self) -> None:
self.clip_grad_norm = aii(
self.config["optim"].get("clip_grad_norm", None), (int, float)
)
self.clip_grad_value = aii(
self.config["optim"].get("clip_grad_value", None), (int, float)
)
self.ema_decay = aii(self.config["optim"].get("ema_decay"), float)
if self.ema_decay:
self.ema = ExponentialMovingAverage(
Expand Down Expand Up @@ -889,15 +892,18 @@ def _backward(self, loss) -> None:
"Please check if all shared parameters are used "
"and point to PyTorch parameters."
)
if self.clip_grad_norm:
if self.clip_grad_norm or self.clip_grad_value:
if self.scaler:
self.scaler.unscale_(self.optimizer)
grad_norm = torch.nn.utils.clip_grad_norm_(
self.model.parameters(),
max_norm=self.clip_grad_norm,
)
if self.logger is not None:
self.logger.log({"grad_norm": grad_norm}, step=self.step, split="train")
if self.clip_grad_norm:
grad_norm = torch.nn.utils.clip_grad_norm_(
self.model.parameters(),
max_norm=self.clip_grad_norm,
)
if self.logger is not None:
self.logger.log({"grad_norm": grad_norm}, step=self.step, split="train")
if self.clip_grad_value:
torch.nn.utils.clip_grad_value_(self.model.parameters(), clip_value=self.clip_grad_value)
if self.scaler:
self.scaler.step(self.optimizer)
self.scaler.update()
Expand Down