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

Jump to bottom

Support Lion with Zero Optimizer #1166

Merged
merged 5 commits into from
Mar 4, 2024
Merged

Support Lion with Zero Optimizer #1166

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
5715d7e
feat: deepspeed zero lion support
DayOfThePenguin Mar 2, 2024
383814f
feat: bump DeeperSpeed version to one that includes DeepSpeed FusedLion
DayOfThePenguin Mar 2, 2024
9b7d401
feat: bump DeeperSpeed version to include pipeline logging fix
DayOfThePenguin Mar 3, 2024
6820bfc
Merge branch 'EleutherAI:main' into zero-lion
DayOfThePenguin Mar 3, 2024
c062c20
pre-commit
Quentin-Anthony Mar 4, 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
13 changes: 11 additions & 2 deletions megatron/training.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -538,9 +538,18 @@ def get_optimizer(model, neox_args):
**neox_args.optimizer["params"],
)
elif neox_args.optimizer_type.lower() == "lion":
from .optimizers import Lion
# if we want the deepspeed zero lion...megatron lion will throw DeepSpeed Error
if neox_args.zero_optimization["stage"] != 0:
from deepspeed.ops.lion import FusedLion

optimizer = Lion(
lion_optimizer = FusedLion
# if not zero
else:
from .optimizers import Lion

lion_optimizer = Lion

optimizer = lion_optimizer(
param_groups,
weight_decay=neox_args.weight_decay,
**neox_args.optimizer["params"],
Expand Down
2 changes: 1 addition & 1 deletion requirements/requirements.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
best_download
git+https://github.com/EleutherAI/DeeperSpeed.git@b9260436e7da3e297fc6bedfd27d9e69fbba6f5c#egg=deepspeed
git+https://github.com/EleutherAI/DeeperSpeed.git@02e2ebf7dee6aaab3d89094ed470a4609763c742#egg=deepspeed
ftfy>=6.0.1
git+https://github.com/EleutherAI/lm_dataformat.git@4eec05349977071bf67fc072290b95e31c8dd836
huggingface_hub>=0.11.0
Expand Down
Loading