Skip to content

[Feature] Add --mg-fl-prefer argument for Megatron-LM-FL vendor selection #1183

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
lxd-cumt merged 1 commit into from
Apr 3, 2026
+5 −1
Merged

[Feature] Add --mg-fl-prefer argument for Megatron-LM-FL vendor selection #1183

Show file tree
Hide file tree
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: 3 additions & 1 deletion flagscale/train/megatron/training/arguments_fs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -765,7 +765,9 @@ def _add_regularization_args(parser):


def _add_flagos_args(parser):
group = parser.add_argument_group(title="flagscale transformer engine fl")
group = parser.add_argument_group(title="flagscale fl")
group.add_argument('--mg-fl-prefer', type=str, choices=['cuda', 'musa', 'txda'], default='',
help='Backend selection for megatron fl.')
group.add_argument('--te-fl-prefer', type=str, choices=['flagos', 'vendor', 'reference'], default='vendor',
help='Backend selection for transformer engine fl.')
group.add_argument('--te-fl-per-op', type=str, default=None,
Expand Down
2 changes: 2 additions & 0 deletions flagscale/train/megatron/training/training.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -805,6 +805,8 @@ def pretrain(

###### FlagScale Begin ######
args = get_args()
if args.mg_fl_prefer:
os.environ['MG_FL_PREFER'] = args.mg_fl_prefer
# enable flagos:triton / vendor:cuda / reference:torch backend for transformer engine fl
if args.te_fl_prefer:
os.environ['TE_FL_PREFER'] = args.te_fl_prefer
Expand Down
Loading