-
Notifications
You must be signed in to change notification settings - Fork 415
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.
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
Deepspeed Integration #109
Conversation
35bbee2
to
7556fcd
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM, maybe wait on one more lgtm for the training loop changes
ead595e
to
5c14354
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'll take another look once the TODOs are adressed, but looks awesome!
@pytest.mark.gpu | ||
@pytest.mark.parametrize("world_size", [pytest.param(1), pytest.param(2, marks=pytest.mark.world_size(2))]) | ||
def test_deepspeed(world_size: int, mosaic_trainer_hparams: TrainerHparams, tmpdir: pathlib.Path) -> None: | ||
"""Pretty much just copied from ./test_ddp""" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
#110 changes the DDP test fyi
Support for using DeepSpeed instead of Pytorch DDP in the trainer. Somewhat WIP, but good enough to merge for now. The major TODOs are:
I'm waiting on a node to run a quick regression test.