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

add expandable segments var #775

Merged
merged 6 commits into from
Jul 22, 2024
Merged

add expandable segments var #775

Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
5743a59
adding new notebook for using fairchem models with NEBs without CatTS…
brookwander Jul 16, 2024
863a01d
add expandable segments var
rayg1234 Jul 18, 2024
90f673e
Merge branch 'main' into rgao_expandable_segments
mshuaibii Jul 19, 2024
dfc59e1
Merge branch 'main' into rgao_expandable_segments
rayg1234 Jul 19, 2024
51ba122
add note
rayg1234 Jul 19, 2024
b043c83
Merge branch 'main' into rgao_expandable_segments
rayg1234 Jul 19, 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: 13 additions & 0 deletions src/fairchem/core/common/utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,12 @@
from torch.nn.modules.module import _IncompatibleKeys


DEFAULT_ENV_VARS = {
# Expandable segments is a new cuda feature that helps with memory fragmentation during frequent allocations (ie: in the case of variable batch sizes).
# see https://pytorch.org/docs/stable/notes/cuda.html.
"PYTORCH_CUDA_ALLOC_CONF" : "expandable_segments:True",
}

# copied from https://stackoverflow.com/questions/33490870/parsing-yaml-in-python-detect-duplicated-keys
# prevents loading YAMLS where keys have been overwritten
class UniqueKeyLoader(yaml.SafeLoader):
Expand Down Expand Up @@ -953,6 +959,12 @@ def check_traj_files(batch, traj_dir) -> bool:
return all(fl.exists() for fl in traj_files)


def setup_env_vars() -> None:
for k, v in DEFAULT_ENV_VARS.items():
os.environ[k] = v
logging.info(f"Setting env {k}={v}")


@contextmanager
def new_trainer_context(*, config: dict[str, Any], distributed: bool = False):
from fairchem.core.common import distutils, gp_utils
Expand All @@ -969,6 +981,7 @@ class _TrainingContext:
trainer: BaseTrainer

setup_logging()
setup_env_vars()
original_config = config
config = copy.deepcopy(original_config)

Expand Down