Skip to content

Save only model sharded sd #40208

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

Open
winglian wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Save only model sharded sd #40208

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
3f92d8c
allow sharded_state_dict when save_only_model
winglian Aug 15, 2025
bda74b7
make sure to save the sharded_state_dict too
winglian Aug 15, 2025
629573e
chore: lint
winglian Aug 15, 2025
1b45bf7
Merge branch 'main' into save_only_model-sharded_Sd
S1ro1 Aug 19, 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
14 changes: 8 additions & 6 deletions src/transformers/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3315,6 +3315,7 @@ def _save_checkpoint(self, model, trial):

run_dir = self._get_output_dir(trial=trial)
output_dir = os.path.join(run_dir, checkpoint_folder)
os.makedirs(output_dir, exist_ok=True)
self.save_model(output_dir, _internal_call=True)

if self.args.save_strategy in [SaveStrategy.STEPS, SaveStrategy.EPOCH] and self.state.best_global_step:
Expand All @@ -3330,6 +3331,13 @@ def _save_checkpoint(self, model, trial):
self._save_scaler(output_dir)
# Save RNG state
self._save_rng_state(output_dir)
elif self.is_fsdp_enabled and (
"SHARDED_STATE_DICT" in str(self.accelerator.state.fsdp_plugin.state_dict_type)
):
# self.save_model above only handles FULL_STATE_DICT
save_fsdp_model(
self.accelerator.state.fsdp_plugin, self.accelerator, self.model, output_dir, **_get_fsdp_ckpt_kwargs()
)

# Save the Trainer state
if self.args.should_save:
Expand Down Expand Up @@ -5384,12 +5392,6 @@ 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