Skip to content

fix: fsdp sharded state dict wont work for save_only_model knob #36627

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
SunMarc merged 4 commits into from
Mar 13, 2025
Merged

fix: fsdp sharded state dict wont work for save_only_model knob #36627

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
5fead11
fix: fsdp sharded state dict wont work for save_only_model knob
kmehant Mar 10, 2025
d98a593
Merge branch 'main' into bug-fsdp-save
SunMarc Mar 13, 2025
be0c246
Merge branch 'main' into bug-fsdp-save
SunMarc Mar 13, 2025
a0325c0
Merge branch 'main' into bug-fsdp-save
SunMarc Mar 13, 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
6 changes: 6 additions & 0 deletions src/transformers/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5175,6 +5175,12 @@ def create_accelerator_and_postprocess(self):
raise ValueError(
"`auto_find_batch_size` isn't supported yet with DeepSpeed Zero-3. Please consider using Zero-2, Zero-1, or FSDP"
)
if (
self.args.save_only_model
and self.is_fsdp_enabled
and "SHARDED_STATE_DICT" in str(self.accelerator.state.fsdp_plugin.state_dict_type)
):
raise ValueError("save_only_model option is not compatible with FSDP state dict type 'SHARDED_STATE_DICT'")

def propagate_args_to_deepspeed(self, auto_find_batch_size=False):
"""
Expand Down