Skip to content

Fix moe save load #9045

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
ZHUI merged 4 commits into from
Aug 30, 2024
Merged

Fix moe save load #9045

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
cc9b7b2
sz-ernie4-1 [moe] fix save under not `use-hybrid-model` (#8782)
Meiyim Jul 24, 2024
cdd659e
fix `upload signal` under pure dp and dp-moe (#8874)
Meiyim Aug 7, 2024
4fed2f0
[fea] sharding stage1 v2 save freeze params (#8873)
bo-ke Aug 7, 2024
53a1fad
refactor model saving
Meiyim Aug 29, 2024
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
28 changes: 22 additions & 6 deletions paddlenlp/trainer/trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -354,7 +354,17 @@ def __init__(
)
self.add_callback(PrinterCallback if self.args.disable_tqdm else DEFAULT_PROGRESS_CALLBACK)

self._save_ckpt_func = dist.save_state_dict if self.args.enable_auto_parallel else paddle.save
def _save_ckpt_func(state_dict, path, signal_path=None):
if self.args.enable_auto_parallel:
dist.save_state_dict(state_dict, path)
else:
paddle.save(state_dict, path)

if signal_path is not None:
with open(signal_path, mode="w+") as f:
f.write("1")

self._save_ckpt_func = _save_ckpt_func
self._load_ckpt_func = dist.load_state_dict if self.args.enable_auto_parallel else paddle.load
if self.args.use_async_save:
self._async_optimizer_saver = AsyncSaver()
Expand Down Expand Up @@ -2297,7 +2307,9 @@ def _save_checkpoint(self, model, metrics=None):
self._save_ckpt_func(
self._filter_moe_no_sync_optimizer_params(),
os.path.join(output_dir, optimizer_name),
saved_signal_path,
)

else:
state_dict = self.optimizer.state_dict()
save_path = os.path.join(output_dir, optimizer_name)
Expand All @@ -2307,9 +2319,7 @@ def _save_checkpoint(self, model, metrics=None):
state_dict, save_path, saved_signal_path=saved_signal_path
)
else:
self._save_ckpt_func(state_dict, save_path)
with open(saved_signal_path, mode="w+") as f:
f.write("1")
self._save_ckpt_func(state_dict, save_path, saved_signal_path)
else:
if self.args.unified_checkpoint and "async_save" in self.args.unified_checkpoint_config:
global_rank = paddle.distributed.get_rank() if paddle.distributed.get_world_size() > 1 else -1
Expand All @@ -2328,10 +2338,16 @@ def _save_checkpoint(self, model, metrics=None):
else:
if self.args.data_parallel_rank > 0 and self.args.use_expert_parallel:
self._save_ckpt_func(
self._filter_moe_no_sync_optimizer_params(), os.path.join(output_dir, OPTIMIZER_NAME)
self._filter_moe_no_sync_optimizer_params(),
os.path.join(output_dir, optimizer_name),
saved_signal_path,
)
else:
self._save_ckpt_func(self.optimizer.state_dict(), os.path.join(output_dir, OPTIMIZER_NAME))
self._save_ckpt_func(
self.optimizer.state_dict(),
os.path.join(output_dir, optimizer_name),
saved_signal_path,
)

# FIXME: maybe only save one copy
paddle.save(self.lr_scheduler.state_dict(), os.path.join(output_dir, SCHEDULER_NAME))
Expand Down
3 changes: 3 additions & 0 deletions paddlenlp/trainer/utils/sharding_io.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,6 +83,9 @@ def filter_sharded_params(state_dict, optimizer, sharding_group):
for (k, v) in state_dict.items():
if v.name in filtered_parameters:
filtered_state_dict[k] = v
else:
if sharding_rank == 0:
filtered_state_dict[k] = v
return filtered_state_dict


Expand Down