Skip to content

Drop '_pg_collection' in MBridge model config when checkpointing #813

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
AAnoosheh wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Drop '_pg_collection' in MBridge model config when checkpointing #813

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
7 changes: 6 additions & 1 deletion modelopt/torch/opt/plugins/mcore_dist_checkpointing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,6 @@
DROP_SUBSTRINGS = [
"fp4",
"fp8",
"tp_",
"parallel",
"cuda_graph",
"init_",
Expand All @@ -49,6 +48,10 @@
"pipeline",
"comm",
"batch",
"pg_collection",
]
DROP_STARTSWITH = [
"tp_", # would drop 'mtp_*' otherwise
]


Expand Down Expand Up @@ -145,6 +148,8 @@ def _parse_transformer_config(transformer_config: dict) -> dict:
for k, v in transformer_config.items():
if any(substring in k for substring in DROP_SUBSTRINGS):
continue
if any(k.startswith(prefix) for prefix in DROP_STARTSWITH):
continue
if isinstance(v, (bool, int, str)):
config[k] = v
else:
Expand Down