Skip to content

Releases: NVIDIA/spark-rapids-tools

v25.02.3

09 May 00:57
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Packaging tools jar with the python package (#1634)
  • Integrate Qualx train_and_evaluate into spark_rapids CLI with docs (#1660)
  • Adding 'Not Recommended Reason' to qualification summary (#1649)
  • Use config file and add docs for Qualx hash_util (#1657)
  • Add more qualx unit tests (#1654)
  • Qualx pipeline API (#1645)

Core

  • Fix Bootstrap to avoid carrying forward CPU run memory configs when insufficient for GPU runs (#1663)
  • AutoTuner: Fix incorrect memoryOverheadFactor recommendation for YARN and k8s and add default master in unit tests (#1659)
  • Improve unit handling for memory-related configs in AutoTuner (#1652)
  • Improve memory allocations in ProfileResults classes (#1642)
  • Fix filtered eventlogs in Profiling tool (#1639)

v25.02.2

15 Apr 22:31
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Disable assertions in core-tools (#1635)

Core

  • Fix handling of driver logs in core-tools (#1636)
  • Disable assertions in core-tools (#1635)
  • Remove the calculation of job frequency from core-tools (#1630)
  • Add java-cmd to the system-properties CSV file (#1632)
  • Add attemptId to app summaries (#1627)

v25.02.1

10 Apr 20:32
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Remove Compare/Combined modes from Profiling tool (#1619)
  • Disable dropping sqlIDs with failures during qualx prediction (#1615)
  • Qualx model updates from weekly KPI run 2025-03-30 (#1604)
  • Fix unused global warnings from new flake8 7.2.0 (#1609)
  • Reduce qualx logging noise (#1603)

Core

  • Sync supported ops with RAPIDS plugin as of 04-01-2025 (#1618)
  • Remove Compare/Combined modes from Profiling tool (#1619)
  • Adding sql_plan_info.json file generation for qualX consumption (#1612)
  • Improve Profiling AutoTuner to provide cluster-aware configuration recommendations (#1607)

v25.02.0

26 Mar 18:38
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Revert "Update version by jenkins-spark-rapids-tools-auto-release-111" (#1600)
  • Qualx unit tests (#1599)
  • Qualx model updates from weekly KPI run 2025-03-09 (#1584)
  • Add new qualx features for custom model training (#1573)
  • Qualx model updates from weekly KPI run 2025-02-23 (#1559)

Core

  • Revert "[FEA] Add filtered diagnostic output for GPU slowness in Profiler tool (#1548)" (#1602)
  • Revert "Update version by jenkins-spark-rapids-tools-auto-release-111" (#1600)
  • [FEA] Add filtered diagnostic output for GPU slowness in Profiler tool (#1548)
  • AutoTuner recommends increasing shuffle partitions when shuffle stages have OOM failures on YARN (#1593)
  • Add insertIntoHiveTable to the WriteOps report (#1591)
  • Fix AutoTuner unit test with dynamic plugin JAR URL value (#1592)
  • Adjust maxPartitionBytes if the table scan stage had OOM task failures (#1578)
  • Adds aggregation across metrics for failed/succeeded and non completed stages (#1558)
  • Support Bin and Slice expressions in Qual tool (#1581)
  • Add kryo related settings to the AutoTuner's Bootstrap conf (#1574)
  • Add sqlID column to failed_jobs.csv (#1567)
  • Pretty print FileFormat in the write_operations file (#1562)

Miscellaneous

  • Change labeler workflow to pull_request_target (#1596)
  • Add Github workflow to add label to PR automatically (#1585)

v24.12.4

21 Feb 20:11
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Add comment for each expected_raw_feature indicating CSV source (#1547)

Core

  • Calculate task metric aggregates on-the-fly to reduce memory usage (#1543)
  • Generate a detailed report for the write ops (#1544)

Miscellaneous

  • Disable dataproc enhanced optimizer configs (#1554)

v24.12.3

13 Feb 22:00
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Add support for configurable qualx label column (#1528)
  • Merge Distributed Qualification Tools CLI (#1516)

Core

  • AutoTuner/Bootstrapper should recommend Dataproc Spark performance enhancements (#1539)
  • Disable Per-SQL summary text output (#1530)
  • Use a stub to store Spark StageInfo (#1525)
  • Recommend G6 instead of G5 on EMR (#1523)
  • Generate a separate file to list bootstrap properties (#1517)

Miscellaneous

  • Disable diagnostics pytests (#1532)

v24.12.2

31 Jan 00:02
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Revert "Follow Up: Make '--platform' argument mandatory in CLI (#1473)" (#1498)
  • Fix Hadoop JAR Download Timeouts in Behave Tests (#1503)

Core

  • AutoTuner: Set recommendation for spark.task.resource.gpu.amount to a very low value (#1514)
  • [FEA] Add IO diagnostic output for GPU slowness in Profiler tool (#1451)
  • [BUG] Qual tool should convert time units at stage/job/sql level (#1511)
  • Fix string comparison for memory overhead in pinned pool size recommendation in AutoTuner (#1508)
  • Update core tools rules to allow cross-build between 2.12 and 2.13 (#1510)
  • Sync plugin support as of 2024-12-31 (#1478)
  • Add stringType and binaryType to the list of dataType map (#1506)
  • [BUG] Remove duplicated executor CPU time and runtime metric from SQLTaskAggMetricsProfileResult (#1504)
  • Improve AutoTuner cluster configuration recommendations for GPU runs (#1501)

Miscellaneous

  • Use common add-to-project action [skip ci] (#1505)

v24.12.1

14 Jan 03:32
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Add compute_precision_recall utility function (#1500)
  • Fix additional FutureWarning issues (#1499)
  • Qualx model updates from weekly KPI run 2025-01-10 (#1495)
  • Fix future warnings for pandas>=2.2 (#1494)
  • Pin scikit-learn dependency for shap (#1491)
  • Make spill heuristic 1 TB by default (#1488)
  • Support Python 3.9-3.12 (#1486)
  • Update models for latest code/datasets (#1485)

Core

  • Improve scalastyle rules to detect spaces (#1493)
  • Improve shuffle manager recommendation in AutoTuner with version validation (#1483)
  • Support group-limit optimization for ROW_NUMBER in Qualification (#1487)
  • Bump minimum Spark version to 3.2.0 and improve AutoTuner unit tests for multiple Spark versions (#1482)
  • Fix inconsistent shuffle write time sum results in Profiler output (#1450)
  • Refine Qualification AutoTuner recommendations for shuffle partitions for CPU event logs (#1479)
  • Split AutoTuner for Profiling and Qualification and Override BATCH_SIZE_BYTES (#1471)

v24.12.0

20 Dec 20:23
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Make '--platform' argument mandatory in qualification and profiling CLI to prevent incorrect behavior (#1463)

Core

  • Skip processing apps with invalid platform and spark runtime configurations (#1421)
  • Improve implementation of finding median in StatisticsMetrics (#1474)
  • Optimize implementation of getAggregateRawMetrics in core-tools (#1468)
  • Adding Spark 3.5.2 support in auto tuner for EMR (#1466)
  • Mark RunningWindowFunction as supported in Qual tool (#1465)
  • Deduplicate calls to aggregateSparkMetricsBySql (#1464)

Miscellaneous

  • Follow Up: Make '--platform' argument mandatory in CLI (#1473)

v24.10.3

13 Dec 01:57
Compare
Choose a tag to compare

Packages

Changes

User Tools

  • Fix dataframe handling of column-types (#1458)