2 Releases published by 2 people

• v3.0.1 3.0.1 Patch Release

  published May 13, 2025

• v3.0.2 3.0.2 Patch Release

  published May 25, 2025

24 Pull requests merged by 5 people

• [CI] Test with Rapids 25.06

  #11504 merged Jun 10, 2025

• Use latest Rapids

  #11323 merged Jun 9, 2025

• Fix typos in CONTRIBUTORS.md: 'a open' -> 'an open' and 'Committers comes' -> 'Committers come'

  #11503 merged Jun 9, 2025

• [EM] Log ellpack read throughput.

  #11497 merged Jun 6, 2025

• Make xgboost.testing compatible with scikit-learn 1.7

  #11502 merged Jun 6, 2025

• adds explicit <=> operators to IndexTransformIter

  #11495 merged Jun 5, 2025

• Update install doc for Conda (#11483)

  #11486 merged May 27, 2025

• Update install doc for Conda

  #11483 merged May 26, 2025

• [EM] By pass device copy if the page fits into device memory.

  #11477 merged May 23, 2025

• Bump version to 3.0.2

  #11479 merged May 22, 2025

• [backport] Fix VM fallback logic on WSL2 (#11471)

  #11480 merged May 22, 2025

• [backport] Dask 2025.4.0 scheduler info compatibility (#11462)

  #11476 merged May 22, 2025

• [EM] Implement default cache split policy.

  #11469 merged May 20, 2025

• Fix VM fallback logic on WSL2

  #11471 merged May 20, 2025

• Implement double-buffer for ellpack.

  #11465 merged May 17, 2025

• Check nvidia smi version before loading decomp API.

  #11464 merged May 16, 2025

• Dask 2025.4.0 scheduler info compatibility

  #11462 merged May 15, 2025

• Optional support for nvcomp.

  #11460 merged May 14, 2025

• Reduce flakiness of the pyspark conf test.

  #11456 merged May 14, 2025

• Add CUDA stream pool.

  #11458 merged May 14, 2025

• [EM] Initial support for splitting up the host cache.

  #11453 merged May 13, 2025

• Add wrapper for CUDA host pinned memory pool.

  #11451 merged May 13, 2025

• [backport] [CI] Rename container xgb-ci.aarch64 -> xgb-ci.manylinux_2_28_x86_64 (#11454)

  #11455 merged May 13, 2025

• [CI] Rename container xgb-ci.aarch64 -> xgb-ci.manylinux_2_28_x86_64

  #11454 merged May 12, 2025

4 Pull requests opened by 2 people

• [jvm-packages] ExtenralMemory: Overlap the caching time

  #11474 opened May 21, 2025

• Benchmark ext mem connect

  #11487 opened May 27, 2025

• [jvm-packages] Fix parameters

  #11489 opened May 27, 2025

• Documentation for Booster.java.

  #11508 opened Jun 12, 2025

14 Issues closed by 6 people

• Bug in Booster.java library when calling XGBoostJNI.checkCall(XGBoostJNI.XGBoosterLoadModelFromBuffer(ret.handle, buffer));

  #11507 closed Jun 12, 2025

• XGBoost is unusually slow

  #11505 closed Jun 10, 2025

• XGBoost PyPi package seems to lack CUDA support

  #11482 closed Jun 10, 2025

• Train when dataset doesn't fit available GPU memory?

  #7951 closed Jun 1, 2025

• Feature Request: Native MultiOutput Support on GPU (CatBoost-style)

  #11481 closed May 31, 2025

• Big difference in feature importances between xgboost version

  #11491 closed May 31, 2025

• Feature Request - Support model inference on only the used features

  #11492 closed May 31, 2025

• 3.0.2 Patch Release

  #11478 closed May 30, 2025

• When making xgboost c api inferences(predict: XGBoosterPredictFromDMatrix), is it thread-safe now?

  #11488 closed May 29, 2025

• Avoid full-dataset spill to disk when building DaskDMatrix for XGBoost on local cluster

  #11452 closed May 27, 2025

• XGBoost failing on A2 machines but the same code works in L4 GPU

  #11450 closed May 16, 2025

• XGBoostj4-spark train failure "Check failed: status == cudaDriverEntryPointSuccess: cuMemBatchDecompressAsync"

  #11463 closed May 16, 2025

• `xgboost.dask` broken for larger clusters

  #11461 closed May 15, 2025

• pip install xgboost==3.0.1 fails

  #11459 closed May 14, 2025

14 Issues opened by 11 people

• Using ndcg_exp_gain=False for non-exponential nDCG loss in XGBoost rank:ndcg

  #11506 opened Jun 10, 2025

• Importing xgboost before torch + openmp causes seg fault

  #11500 opened Jun 5, 2025

• XGBoost 3.0+ fails with a degenerate data set with identical labels

  #11499 opened Jun 4, 2025

• [jvm-packages] Upgrade to Java 17+

  #11498 opened Jun 4, 2025

• XGBoost Stage failed because barrier task ResultTask

  #11496 opened Jun 3, 2025

• 3.0.3 Patch Release

  #11494 opened May 30, 2025

• Feature proposal - groupSHAP

  #11493 opened May 30, 2025

• [CI] Drop manylinux2014 wheels in 3.1 version

  #11485 opened May 27, 2025

• AssertionError-assert s.rank not in wait_conn java.RabitTracker: Tracker Process ends with exit code 0

  #11484 opened May 26, 2025

• Mixing booster types silently fails

  #11473 opened May 21, 2025

• [CI] The operation was canceled

  #11472 opened May 19, 2025

• Support building xgboost from source for cuda code with clang++

  #11470 opened May 19, 2025

• Error building xgboost from source with Clang20 due to `FEDERATED_LEARNING=ON` incompatible forward declarations with newer versions of libc++

  #11466 opened May 15, 2025

• ExtMemQuantileDMatrix out_of_memory

  #11457 opened May 13, 2025

21 Unresolved conversations

Sometimes conversations happen on old items that aren’t yet closed. Here is a list of all the Issues and Pull Requests with unresolved conversations.

• Optimization of data initialization for large sparce datasets

  #11390 commented on Jun 2, 2025 • 4 new comments

• Add Possibility to Load XGBoost Models as PySpark Models

  #11449 commented on May 16, 2025 • 2 new comments

• [jvm-packages] Support spark connect

  #11381 commented on Jun 9, 2025 • 1 new comment

• [WIP] Avoid R API calls within OpenMP region

  #11448 commented on May 15, 2025 • 0 new comments

• [WIP] Drop the deprecated binary format.

  #11307 commented on May 29, 2025 • 0 new comments

• Optuna Pruning in 3.0.0?

  #11432 commented on Jun 4, 2025 • 0 new comments

• Make all tutorials multi-language

  #11413 commented on Jun 4, 2025 • 0 new comments

• Tracking deprecated features.

  #3986 commented on Jun 1, 2025 • 0 new comments

• Xgboost pyspark: Support external memory mode

  #8109 commented on Jun 1, 2025 • 0 new comments

• [RFC] High-level interface for external memory.

  #11000 commented on Jun 1, 2025 • 0 new comments

• xgb.train "unknown error" with External Memory and large dataset

  #11440 commented on Jun 1, 2025 • 0 new comments

• 3.0.1 Patch release.

  #11433 commented on May 31, 2025 • 0 new comments

• XGBoostError: rabit/internal/utils.h:90: Allreduce failed - Error while attempting XGboost on Dask Fargate Cluster in AWS

  #7868 commented on May 30, 2025 • 0 new comments

• Trouble Scaling XGBoost beyond in-memory training on databricks

  #10853 commented on May 28, 2025 • 0 new comments

• RAM problem using xgb.QuantileDMatrix with NA values

  #11421 commented on May 27, 2025 • 0 new comments

• Using customized subset of features when growing tree for different depths

  #11441 commented on May 27, 2025 • 0 new comments

• Why did the default base_score for reg:squaredlogerror change in xgboost 3.0.0?

  #11442 commented on May 23, 2025 • 0 new comments

• [Spark 3.5.0] Rabit Tracker Connection Failure During Distributed XGBoost Training

  #11380 commented on May 18, 2025 • 0 new comments

• CRAN package warning.

  #11431 commented on May 16, 2025 • 0 new comments

• (scikit-learn api) (feature request) XGBRegressor with early stopping, dynamic eval_set

  #7782 commented on May 16, 2025 • 0 new comments

• Question: Import xgboost as new port in vcpkg repo, build failed for arm64_osx, arm64_uwp, arm64_windows...

  #10343 commented on May 13, 2025 • 0 new comments