-
Notifications
You must be signed in to change notification settings - Fork 1k
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
Refactor launch
for greater extensibility
#1123
Conversation
Signed-off-by: Antoni Baum <antoni.baum@protonmail.com>
The documentation is not available anymore as the PR was closed or merged. |
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.
Thanks a lot for your PR! This looks great to me. Mind having a second look @muellerzr ?
Can you just solve the merge conflict @Yard1 ?
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.
LG2M, great work! Once the merge conflict is fixed we can go ahead and bring it in 🤗
Signed-off-by: Antoni Baum <antoni.baum@protonmail.com>
Signed-off-by: Antoni Baum <antoni.baum@protonmail.com>
@muellerzr Thanks, should be good now! |
@@ -23,8 +23,9 @@ | |||
|
|||
from ..commands.config.config_args import SageMakerConfig | |||
from ..commands.config.config_utils import DYNAMO_BACKENDS | |||
from ..utils import DynamoBackend, PrecisionType, is_torch_version, merge_dicts |
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.
Ideally it should be imported from ..utils and not using the direct path to .other
, following the same mechanism used for all other utils, isn't it?
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.
That results in a circular import, unfortunately
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.
Got it, then LGTM.
FYI @natolambert this was the PR that i had mentioned to you ages ago. It'll be useful if you guys plan to do sweeps in TRL. |
This PR refactors
commands.launch
to break up the logic into smaller functions, allowing for the environment preparation logic and actual launch logic to be decoupled. Therefore, a developer can now prepare an environment dict (and command list, where applicable) without having to launch any processes.There have been no changes to the logic - this is purely a refactor of an existing structure.
As discussed offline, this will unblock Ray Train and other distributed frameworks which launch their own jobs.