v24.10.00

🚨 Breaking Changes

• Replace cuDF (de)serializer with cuDF spill-aware (de)serializer (#1369) @pentschev

📖 Documentation

• Fix typo in spilling documentation (#1384) @rjzamora
• Add notes on cudf spilling to docs (#1383) @rjzamora

🚀 New Features

• [Benchmark] Add parquet read benchmark (#1371) @rjzamora
• Replace cuDF (de)serializer with cuDF spill-aware (de)serializer (#1369) @pentschev

🛠️ Improvements

• Update update-version.sh to use packaging lib (#1387) @AyodeAwe
• Use CI workflow branch 'branch-24.10' again (#1386) @jameslamb
• Update to flake8 7.1.1. (#1385) @bdice
• enable Python 3.12 tests on PRs (#1382) @jameslamb
• Add support for Python 3.12 (#1380) @jameslamb
• Update rapidsai/pre-commit-hooks (#1379) @KyleFromNVIDIA
• Drop Python 3.9 support (#1377) @jameslamb
• Remove NumPy <2 pin (#1375) @seberg
• Update pre-commit hooks (#1373) @KyleFromNVIDIA
• Merge branch-24.08 into branch-24.10 (#1368) @jameslamb

[NIGHTLY] v24.12.00

🔗 Links

• Development Branch
• Compare with main branch

🚨 Breaking Changes

• Add warmup runs and profile all iterations to benchmarks (#1402) @pentschev

🐛 Bug Fixes

• Ignore legacy Dask dataframe warnings (#1397) @pentschev
• Reenable UCXX in CI (#1396) @pentschev

🚀 New Features

• Enable Pytorch to share same memory pool as RMM via cli (#1392) @VibhuJawa

🛠️ Improvements

• enforce wheel size limits, README formatting in CI (#1404) @jameslamb
• Add warmup runs and profile all iterations to benchmarks (#1402) @pentschev
• remove unnecessary cmake and sccache configuration (#1400) @jameslamb
• make conda installs in CI stricter (#1395) @jameslamb
• Limit output of pytest durations (#1393) @pentschev
• Switch pytest traceback to native (#1389) @galipremsagar

v24.08.00

🐛 Bug Fixes

• Fix partitioning in explicit-comms shuffle (#1356) @rjzamora
• Update cuDF's assert_eq import (#1353) @pentschev

🚀 New Features

• Add arguments to enable cuDF spilling and set statistics (#1362) @pentschev
• Allow disabling RMM in benchmarks (#1352) @pentschev

🛠️ Improvements

• consolidate cuda_suffixed=false blocks in dependencies.yaml, fix update-version.sh (#1367) @jameslamb
• split up CUDA-suffixed dependencies in dependencies.yaml (#1364) @jameslamb
• Use verify-alpha-spec hook (#1360) @KyleFromNVIDIA
• Use workflow branch 24.08 again (#1359) @KyleFromNVIDIA
• Build and test with CUDA 12.5.1 (#1357) @KyleFromNVIDIA
• Drop setup.py (#1354) @jakirkham
• remove .gitattributes (#1350) @jameslamb
• make conda recipe data-loading stricter (#1349) @jameslamb
• Adopt CI/packaging codeowners (#1347) @bdice
• Remove text builds of documentation (#1346) @vyasr
• use rapids-build-backend (#1343) @jameslamb

v24.06.00

🐛 Bug Fixes

• Fix license name (#1337) @raydouglass
• Skip TCP-only DGX tests with UCX 1.16 (#1331) @pentschev
• Update explicit-comms for dask-expr support (#1323) @rjzamora
• Skip explicit-comms tests when dask-expr is active (#1322) @rjzamora
• Relax type-check in test_proxy.py (#1321) @rjzamora

📖 Documentation

• Fix broken links in docs (#1329) @pentschev

🛠️ Improvements

• remove 'tomli' dependency (#1338) @jameslamb
• Trap CI test errors with their original exit codes (#1330) @pentschev
• Prevent path conflict in builds (#1325) @AyodeAwe

v24.04.00

🐛 Bug Fixes

• handle more RAPIDS version formats in update-version.sh (#1307) @jameslamb

🚀 New Features

• Allow using pandas 2 (#1308) @vyasr
• Support CUDA 12.2 (#1302) @jameslamb

🛠️ Improvements

• Use conda env create --yes instead of --force (#1326) @bdice
• Add upper bound to prevent usage of NumPy 2 (#1320) @bdice
• Generalize GHA selectors for pure Python testing (#1318) @jakirkham
• Requre NumPy 1.23+ (#1316) @jakirkham
• Add support for Python 3.11 (#1315) @jameslamb
• target branch-24.04 for GitHub Actions workflows (#1314) @jameslamb
• Filter dd deprecation (#1312) @rjzamora
• Update ops-bot.yaml (#1310) @AyodeAwe

v24.02.00

🚨 Breaking Changes

• Publish nightly wheels to NVIDIA index instead of PyPI (#1294) @pentschev

🐛 Bug Fixes

• Fix get_device_memory_ids (#1305) @wence-
• Prevent double UCX initialization in test_dgx (#1301) @pentschev
• Update to Dask's shuffle_method kwarg (#1300) @pentschev
• Add timeout to test_dask_use_explicit_comms (#1298) @pentschev
• Publish nightly wheels to NVIDIA index instead of PyPI (#1294) @pentschev
• Make versions PEP440 compliant (#1279) @vyasr
• Generate pyproject.toml with dfg (#1276) @vyasr
• Fix rapids dask dependency version (#1275) @vyasr

🛠️ Improvements

• Remove usages of rapids-env-update (#1304) @KyleFromNVIDIA
• refactor CUDA versions in dependencies.yaml (#1303) @jameslamb
• Start generating conda test environments (#1291) @charlesbluca
• Branch 24.02 merge branch 23.12 (#1286) @vyasr

v23.12.00

🐛 Bug Fixes

• Update actions/labeler to v4 (#1292) @raydouglass
• Increase Nanny close timeout for test_spilling_local_cuda_cluster (#1289) @pentschev
• Fix path (#1277) @vyasr
• Add missing alpha spec (#1273) @vyasr
• Set minimum click to 8.1 (#1272) @jacobtomlinson
• Reenable tests that were segfaulting (#1266) @pentschev
• Increase close timeout of Nanny in LocalCUDACluster (#1260) @pentschev
• Small reorganization and fixes for test_spill (#1255) @pentschev
• Update plugins to inherit from WorkerPlugin (#1230) @jrbourbeau

🚀 New Features

• Add support for UCXX (#1268) @pentschev

🛠️ Improvements

• Fix license (#1285) @vyasr
• Build concurrency for nightly and merge triggers (#1282) @bdice
• Use new rapids-dask-dependency metapackage for managing dask versions (#1270) @galipremsagar
• Remove ucp.reset() requirement from test_dgx (#1269) @pentschev
• Generate proper, consistent nightly versions for pip and conda packages (#1267) @galipremsagar
• Unpin dask and distributed for 23.12 development (#1264) @galipremsagar
• Move some dask_cuda.utils pieces to their own modules (#1263) @pentschev
• Update shared-action-workflows references (#1261) @AyodeAwe
• Use branch-23.12 workflows. (#1259) @bdice
• dask-cuda: Build CUDA 12.0 ARM conda packages. (#1238) @bdice

v23.10.00

🐛 Bug Fixes

• Monkeypatch protocol.loads ala dask/distributed#8216 (#1247) @wence-
• Explicit-comms: preserve partition IDs (#1240) @madsbk
• Increase test timeouts further to reduce CI failures (#1234) @pentschev
• Use conda mambabuild not mamba mambabuild (#1231) @bdice
• Increate timeouts of tests that frequently timeout in CI (#1228) @pentschev
• Adapt to non-string task keys in distributed (#1225) @wence-
• Update test_worker_timeout (#1223) @pentschev
• Avoid importing loads_function from distributed (#1220) @rjzamora

🚀 New Features

• Enable maximum pool size for RMM async allocator (#1221) @pentschev

🛠️ Improvements

• Pin dask and distributed for 23.10 release (#1251) @galipremsagar
• Update test_spill.py to avoid FutureWarnings (#1243) @pentschev
• Remove obsolete pytest filterwarnings (#1241) @pentschev
• Update image names (#1233) @AyodeAwe
• Use copy-pr-bot (#1227) @ajschmidt8
• Unpin dask and distributed for 23.10 development (#1222) @galipremsagar

v23.08.00

🐛 Bug Fixes

• Ensure plugin config can be passed from worker to client (#1212) @wence-
• Adjust to new get_default_shuffle_method name (#1200) @pentschev
• Increase minimum timeout to wait for workers in CI (#1192) @pentschev

📖 Documentation

• Remove RTD configuration and references to RTD page (#1211) @charlesbluca
• Clarify memory_limit docs (#1207) @pentschev

🚀 New Features

• Remove versioneer (#1204) @pentschev
• Remove code for Distributed<2023.5.1 compatibility (#1191) @pentschev
• Specify disk spill compression based on Dask config (#1190) @pentschev

🛠️ Improvements

• Pin dask and distributed for 23.08 release (#1214) @galipremsagar
• Revert CUDA 12.0 CI workflows to branch-23.08. (#1210) @bdice
• Use minimal Numba dependencies for CUDA 12 (#1209) @jakirkham
• Aggregate reads & writes in disk_io (#1205) @jakirkham
• CUDA 12 Support (#1201) @quasiben
• Remove explicit UCX config from tests (#1199) @pentschev
• use rapids-upload-docs script (#1194) @AyodeAwe
• Unpin dask and distributed for development (#1189) @galipremsagar
• Remove documentation build scripts for Jenkins (#1187) @ajschmidt8
• Use KvikIO in Dask-CUDA (#925) @jakirkham

v23.06.00

🚨 Breaking Changes

• Update minimum Python version to Python 3.9 (#1164) @shwina

🐛 Bug Fixes

• Increase pytest CI timeout (#1196) @pentschev
• Increase minimum timeout to wait for workers in CI (#1193) @pentschev
• Disable np.bool deprecation warning (#1182) @pentschev
• Always upload on branch/nightly builds (#1177) @raydouglass
• Workaround for DeviceHostFile tests with CuPy>=12.0.0 (#1175) @pentschev
• Temporarily relax Python constraint (#1166) @vyasr

📖 Documentation

• [doc] Add document about main guard. (#1157) @trivialfis

🚀 New Features

• Require Numba 0.57.0+ (#1185) @jakirkham
• Revert "Temporarily relax Python constraint" (#1171) @vyasr
• Update to zict 3.0 (#1160) @pentschev

🛠️ Improvements

• Add __main__ entrypoint to dask-cuda-worker CLI (#1181) @hmacdope
• run docs nightly too (#1176) @AyodeAwe
• Fix GHAs Workflows (#1172) @ajschmidt8
• Remove matrix_filter from workflows (#1168) @charlesbluca
• Revert to branch-23.06 for shared-action-workflows (#1167) @shwina
• Update minimum Python version to Python 3.9 (#1164) @shwina
• Remove usage of rapids-get-rapids-version-from-git (#1163) @jjacobelli
• Use ARC V2 self-hosted runners for GPU jobs (#1159) @jjacobelli