[Tune] Make `Tuner.restore(trainable=...) a required argument #34982
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
…nable/param_space Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
…ore/required_trainable
Signed-off-by: Justin Yu <[email protected]>
…nable = None Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
…ore/required_trainable
krfricke
reviewed
May 3, 2023
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
…ore/required_trainable
Signed-off-by: Justin Yu <[email protected]>
Signed-off-by: Justin Yu <[email protected]>
…ore/required_trainable
justinvyu
added
the
tests-ok
|
Tests look ok - apart from the GPU early kickoff weirdness. |
architkulkarni
pushed a commit
to architkulkarni/ray
that referenced
this pull request
May 16, 2023
…oject#34982) **Hard deprecation**: This PR makes `Tuner.restore(..., trainable)` a required argument. This removes the special case of `tune.with_parameters` throwing a very complicated error message if you try to restore without re-specifying the trainable. Users will have access to the `trainable` on restore anyways, and the restore API now mirrors the regular `Tuner` constructor, where `trainable` is required and `param_space` is optional. This PR also simplifies validation logic on restore -- and also adding loose validation that the `param_space` has at least the same keys as before (preventing users from trying to add new hyperparameters after restoration). Signed-off-by: Justin Yu <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Why are these changes needed?
Hard deprecation: This PR makes
Tuner.restore(..., trainable)a required argument. This removes the special case oftune.with_parametersthrowing a very complicated error message if you try to restore without re-specifying the trainable. Users will have access to thetrainableon restore anyways, and the restore API now mirrors the regularTunerconstructor, wheretrainableis required andparam_spaceis optional.This PR also simplifies validation logic on restore -- and also adding loose validation that the
param_spacehas at least the same keys as before (preventing users from trying to add new hyperparameters after restoration).Context
See here for context: #32912, #32912 (comment)
Related issue number
Closes #33546
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/under thecorresponding
.rstfile.