-
Notifications
You must be signed in to change notification settings - Fork 523
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
Experimental Flag for tune ls #1163
base: main
Are you sure you want to change the base?
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/1163
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit f2dae01 with merge base bbc48e0 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
@@ -66,7 +66,7 @@ where = [""] | |||
include = ["torchtune*", "recipes*"] | |||
|
|||
[tool.setuptools.package-data] | |||
recipes = ["configs/*.yaml", "configs/*/*.yaml"] | |||
recipes = ["configs/*.yaml", "configs/*/*.yaml", "configs/*/*/*.yaml"] |
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.
Can we just use something like configs/**.yaml
here?
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.
it should be configs/**/*.yaml
to match any number of subdirectories
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 think we already landed the existing version in another PR. Also after thinking about it more maybe it's better to be explicit here anyways
Context
What is the purpose of this PR? Is it to
Currently we have a number of recipes under dev that we would like to make accessible to users but still clearly signal to users that these are not as well tested and may have special dependencies. The current behavior is to list these all in "tune ls" which isn't ideal as it doesn't disclose the experimental nature of these recipes. One option would be to exclude them entirely from "tune ls" but then it makes it very difficult to use them outside of github.
The proposal here is to add an "experimental" flag to tune ls that displays the extra recipes. All of the recipes are always available in tune run, but if you want to view them in ls you must pass the extra flag. This is accomplished by adding an extra parameter to the recipe dataclass in _recipe_registry that allows some recipes to be marked as experimental. Then get_all_recipes can accept an argument "include_experimental". This is True for "tune run" and False by default for "tune ls".
Changelog
Test plan
Ran tests and test tune ls in terminal for expected outputs
pre-commit install
)pytest tests
pytest tests -m integration_test