[train] Fix ScalingConfig(accelerator_type)
to request a small fraction of the accelerator label
#44225
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Outdated suggestions cannot be applied.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?
accelerator_type
is currently implemented as a custom resource with a quantity of 1 if an instance has an accelerator of that type. For example, both a machine with 1 A10G GPU and a machine with 4 A10G GPUs will have{"accelerator_type:A10G": 1.0}
. This label is just an indicator of whether the machine contains the accelerator, rather than a count of the number of accelerators of that type.This PR makes our accelerator type resource request match Ray Core by setting it to a fractional value (0.001). This is needed to fix autoscaling behavior to request the correct number of GPUs.
Related issue number
Checks
git commit -s
) in this PR.scripts/format.sh
to lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/
under thecorresponding
.rst
file.