Fixup no_trainer examples scripts and add more tests #16765
Merged
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
muellerzr
added
Examples
muellerzr
changed the title
sgugger
approved these changes
Apr 13, 2022
|
The documentation is not available anymore as the PR was closed or merged. |
elusenji
pushed a commit
to elusenji/transformers
that referenced
this pull request
Jun 12, 2022
* Change tracking to store_true * Remove step param and use it in the log dictionary directly * use vars(args) when passing args to init_trackers * Include tracking tests since tensorboard is already a dep
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Outdated suggestions cannot be applied.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.
Fixup
no_trainerExamples and Bolster their testsWhat does this add?
This changes the logging behavior inside the
no_trainerscripts, slightly changes how the initial configuration is stored, and adds tests for the tracking API.Who is it for?
Users of
transformerswho want to try outAcceleratequicklyWhy is this needed?
I was made aware that the scripts were laggy when it came to how logs were sent to weights and biases when using the
no_trainerscripts, and this was due to the step being passed in as a parameter, causing a lag in when it gets uploaded.To follow akin to the original Accelerate scripts, these are now passed in as a
"step"parameter to the overall dictionary logged viaaccelerate.log()TensorBoardalso does not like whenEnum's are logged, so there is a manual adjustment rightr before saving the hyperparemeters to get the enum value from the LR Scheduler type.Finally, as
TensorBoardis a test requirement, I added in tests for tracking inside the no_trainer tests, asTensorBoardis also how we test that behavior in the CI in Accelerate proper.