Skip to content

added fix to ddp problems, removed rank conditions #155

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

Merged
SimarKareer merged 1 commit into from
Jan 23, 2026
Merged

added fix to ddp problems, removed rank conditions #155

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
18 changes: 11 additions & 7 deletions egomimic/pl_utils/pl_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -78,19 +78,23 @@ def on_validation_start(self):
self.model.device = self.device

if self.trainer.is_global_zero:
os.makedirs(
os.path.join(self.video_dir(), f"epoch_{self.trainer.current_epoch}"),
exist_ok=True,
)
os.makedirs(os.path.join(self.video_dir(), f"epoch_{self.trainer.current_epoch}"),exist_ok=True)


@rank_zero_only
def validation_step(self, batch, batch_idx, dataloader_idx=0):
"""
Run a validation step on the batch, and save that batch of images into the val_image_buffer. Once the buffer hits 1000 images, save that as a 30fps video using torchvision.io.write_video.
"""
print(f"[VAL_STEP] rank={self.global_rank}, batch_idx={batch_idx}",flush=True)

batch = self.model.process_batch_for_training(batch)
metrics, images_dict = self.model.forward_eval_logging(batch)

metrics = {
k: (v.to(self.device) if torch.is_tensor(v) else torch.tensor(v, device=self.device))
for k, v in metrics.items()
}

## images is now a dict
for key, images in images_dict.items():
os.makedirs(
Expand All @@ -117,10 +121,10 @@ def validation_step(self, batch, batch_idx, dataloader_idx=0):
self.val_image_buffer[key].clear()
self.val_counter[key] += 1

self.log_dict(metrics)
self.log_dict(metrics, sync_dist=True)

@rank_zero_only
def on_validation_end(self):
print(f"[ON_VALIDATION_END] rank={self.global_rank}",flush=True)
for key, buffer in self.val_image_buffer.items():
os.makedirs(
os.path.join(
Expand Down