Skip to content

Configure FSDP to keep module params #12074

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
maanug-nv merged 1 commit into from
Mar 6, 2025
Merged

Configure FSDP to keep module params #12074

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
4 changes: 4 additions & 0 deletions nemo/collections/nlp/models/language_modeling/megatron_base_model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1324,6 +1324,10 @@ def find_frozen_submodules(model):
for submodule in frozen_submodule_names:
logging.debug(f"Ignoring state {submodule} in FSDP.")
self.trainer.strategy.kwargs['ignored_states'] = frozen_submodules

# Transformer Engine expects that module parameters are available, so FSDP should not replace them
self.trainer.strategy.kwargs['use_orig_params'] = True

# FSDP requires uniform status of require_grads
# Diffusion models like SD has frozen parts and needs to be added to 'ignored_states'
# from sharding for FSDP to work
Expand Down
Loading