[Datasets] Re-enable Parquet sampling and add progress bar #28021
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Outdated suggestions cannot be applied.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.
Signed-off-by: Cheng Su [email protected]
Why are these changes needed?
This PR is to re-enable Parquet files sampling after testing multiple time in xgboost 100GB nightly test - https://github.com/ray-project/ray/blob/master/release/air_tests/air_benchmarks/workloads/xgboost_benchmark.py .
Verified no OOM and memory usage increase during training and tuning:
Also add a progress bar
Parquet Sample Progress
for the sampling to give us more information in case sampling takes a long time (which is unexpected). Tested on sampling 10 files from 100GB dataset, and it takes 3 seconds to finish:Related issue number
Checks
git commit -s
) in this PR.scripts/format.sh
to lint the changes in this PR.