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

validate use_remove_padding when applying sequence parallelism #153

Merged
merged 7 commits into from
Jan 31, 2025
+12 −0
Merged

validate use_remove_padding when applying sequence parallelism #153

Changes from 2 commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
67b58f0
validate when applying sequence parallelism
chujiezheng Jan 28, 2025
1a49109
run format.sh
chujiezheng Jan 29, 2025
7c662ef
fix 'actor.use_remove_padding' to 'actor.model.use_remove_padding'
chujiezheng Jan 30, 2025
3c170b9
Merge branch 'volcengine:main' into main
chujiezheng Jan 31, 2025
440014a
fix config check
chujiezheng Jan 31, 2025
4bd7d4c
fix typo
chujiezheng Jan 31, 2025
a9f680a
fix typo
chujiezheng Jan 31, 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
9 changes: 9 additions & 0 deletions verl/utils/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -86,4 +86,13 @@ def check_mutually_exclusive(mbs, mbs_per_gpu, name: str):
assert config.critic.ppo_mini_batch_size % config.critic.ppo_micro_batch_size == 0
assert config.critic.ppo_micro_batch_size * sp_size >= n_gpus

# Check if use_remove_padding is enabled when using sequence parallelism
if config.actor_rollout_ref.actor.ulysses_sequence_parallel_size > 1:
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It seems that the correct key is config. actor_rollout_ref.model.use_remove_padding, critic.model.use_remove_padding and reward_model.model.use_remove_padding

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Fixed now!

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Unfortunately, the key is still not correct :(

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@chujiezheng The key should be actor_rollout_ref.model.use_remove_padding. The actor and ref are not necessary as they have the same model type.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@vermouth1992 @PeterSH6 Fixed now 🥹

assert config.actor_rollout_ref.actor.use_remove_padding, \
"When using sequence parallelism for actor, you must enable `use_remove_padding`."

if config.actor_rollout_ref.ref.ulysses_sequence_parallel_size > 1:
assert config.actor_rollout_ref.ref.use_remove_padding, \
"When using sequence parallelism for ref policy, you must enable `use_remove_padding`."

print("[validate_config] All configuration checks passed successfully!")
Loading