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

Jump to bottom

move barrier to before saving checkpoint to reduce timeouts when saving #38

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

move barrier to before saving checkpoint to reduce timeouts when saving #38

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
5c4d01c
move barrier to before saving checkpoint to reduce timeouts when saving
Jun 14, 2024
64e5023
add barrier after creating state dicts
Jun 17, 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
4 changes: 2 additions & 2 deletions prismatic/training/strategies/base_strategy.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -224,8 +224,8 @@ def run_training(

# Check for Termination & Save Final Checkpoint (in case `max_steps` is not None)
if self.max_steps is not None and metrics.global_step >= self.max_steps:
self.save_checkpoint(metrics.run_dir, metrics.global_step, epoch, loss.item())
dist.barrier()
self.save_checkpoint(metrics.run_dir, metrics.global_step, epoch, loss.item())

return

Expand All @@ -235,5 +235,5 @@ def run_training(

# Save checkpoint at end each epoch (if `self.max_steps` is None)
if self.max_steps is None:
self.save_checkpoint(metrics.run_dir, metrics.global_step, epoch, loss.item())
dist.barrier()
self.save_checkpoint(metrics.run_dir, metrics.global_step, epoch, loss.item())
1 change: 1 addition & 0 deletions prismatic/training/strategies/fsdp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,6 +115,7 @@ def save_checkpoint(
if key.startswith(mprefix := f"{mkey}."):
model_state_dicts[mkey][key.removeprefix(mprefix)] = param

torch.barrier()
# Save on rank zero *only*
if overwatch.is_rank_zero():
checkpoint_dir = run_dir / "checkpoints"
Expand Down