Skip to content

[trainer] fix: reproducible problem when resume training #4156

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
wuxibin89 merged 6 commits into from
Nov 17, 2025
Merged

[trainer] fix: reproducible problem when resume training #4156

Changes from 1 commit
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
d8721e9
MOD: fix reproducible problem caused by sampler
hgtcwl Nov 17, 2025
555e36f
MOD: format code
hgtcwl Nov 17, 2025
f8a1bdc
MOD: format code
hgtcwl Nov 17, 2025
ccff31a
MOD: format code
hgtcwl Nov 17, 2025
a1658ed
MOD: format code
hgtcwl Nov 17, 2025
59dd68e
MOD: use RandomSampler from torchdata.stateful_dataloader.sampler
hgtcwl Nov 17, 2025
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
Prev Previous commit
Next Next commit
MOD: format code
  • Loading branch information
@hgtcwl
hgtcwl committed Nov 17, 2025
commit f8a1bdc78c92b91e27360333533ae0e4ebd5aa32
3 changes: 2 additions & 1 deletion verl/trainer/ppo/ray_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1317,4 +1317,5 @@ def fit(self):
# in favor of a general-purpose data buffer pool
if hasattr(self.train_dataset, "on_batch_end"):
# The dataset may be changed after each training batch
self.train_dataset.on_batch_end(batch=batch)
self.train_dataset.on_batch_end(batch=batch)