v25.10.00

🚨 Breaking Changes

• build with RAPIDS CUDA architectures (#295) @jameslamb
• Update rapids_config to handle user defined branch name (#272) @robertmaynard

🐛 Bug Fixes

• [BUG] Update 25.08 Dependencies to 25.10 (#329) @alexbarghi-nv
• [BUG] Fix input ID creation to use shape[-1] instead of len (#316) @alexbarghi-nv
• [BUG] Mask out unwanted vertices during negative sampling (#303) @alexbarghi-nv
• [BUG] Fix input type in Taobao example (#301) @alexbarghi-nv
• [BUG] Fix shuffle on single GPU in Taobao Example (#300) @alexbarghi-nv
• [BUG] Fix Calculation of # Sampled Nodes/Edges with Zero Input Size (#283) @alexbarghi-nv
• [CI] Disable Example Tests (#279) @alexbarghi-nv
• [BUG] Fix optional dependencies (#278) @alexbarghi-nv
• Reduce Seeds Per Call when running in CI Environment (#275) @alexbarghi-nv

🚀 New Features

• [FEA] Enable Temporal Sampling in cuGraph-PyG (#310) @alexbarghi-nv
• [FEA] Set random state using PyTorch generator (#265) @alexbarghi-nv

🛠️ Improvements

• Empty commit to trigger a build (#319) @msarahan
• fix condition to skip CUDA 13 conda-python-tests jobs (#312) @jameslamb
• Configure repo for automatic release notes generation (#308) @AyodeAwe
• Allow CMake 4 (#307) @bdice
• Add Graph Property prediction model to cugraph-pyg example (#302) @wangxiaoyunNV
• Use branch-25.10 again (#297) @jameslamb
• build with RAPIDS CUDA architectures (#295) @jameslamb
• cugraph-pyg: remove unnecessary dependency on 'numba' (#287) @jameslamb
• Build and test with CUDA 13.0.0 (#286) @jameslamb
• Update rapids-dependency-file-generator (#285) @KyleFromNVIDIA
• libwholegraph wheels: use nvidia-nccl wheels instead of vendoring libnccl.so (#284) @jameslamb
• cugraph-pyg: add missed runtime dependencies, remove runtime dependencies from 'test' extra (#277) @jameslamb
• Use build cluster in devcontainers (#274) @trxcllnt
• Use rapids_cuda_enable_fatbin_compression (#273) @robertmaynard
• Update rapids_config to handle user defined branch name (#272) @robertmaynard
• Remove cugraph Python library as a dependency (#271) @rlratzel
• Update rapids-build-backend to 0.4.0 (#269) @KyleFromNVIDIA
• cugraph_gnn RAPIDS dependency now use the new rapids-cmake branch info (#263) @robertmaynard
• Allow latest OS in devcontainers (#257) @bdice
• Update build infra to support new branching strategy (#252) @robertmaynard
• Use GCC 14 in conda builds. (#228) @vyasr

[NIGHTLY] v25.12.00

🔗 Links

• Development Branch
• Compare with main branch

🐛 Bug Fixes

• fix: fixes memory context leak (#332) @paulhendricks

📖 Documentation

• Remove Old Content and Fix Broken Links in README (#326) @alexbarghi-nv

🛠️ Improvements

• Enable sccache-dist connection pool (#333) @trxcllnt
• Update RAPIDS_BRANCH, codify changes in update-version.sh (#322) @KyleFromNVIDIA

v25.08.00

🚨 Breaking Changes

• Remove CUDA 11 from dependencies.yaml (#224) @KyleFromNVIDIA
• Remove Non-Unified API and Remaining TensorDict Code (#222) @alexbarghi-nv
• stop uploading packages to downloads.rapids.ai (#215) @jameslamb
• Remove cuGraph-DGL (#210) @alexbarghi-nv

🐛 Bug Fixes

• [BUG] Correct De-Offset of Edge Label Index (#258) @alexbarghi-nv
• [BUG] Pin CPU Memory Instead of Copying to Device (#255) @alexbarghi-nv
• Fix update-version (#250) @AyodeAwe
• Use PyTorch MemPool and Disable RMM Pool Allocator to Fix Broken Tests (#237) @alexbarghi-nv
• [BUG] Specify Input Type and Assign Output to Correct Input Type (#236) @alexbarghi-nv
• Temporarily Disable Bitcoin Example Tests (#230) @alexbarghi-nv
• Branch 25.08 merge 25.06 (#226) @alexbarghi-nv

📖 Documentation

• Remove nx-cugraph reference that mentioned CUDA 11 (#243) @alexbarghi-nv
• add docs on CI workflow inputs (#235) @jameslamb

🚀 New Features

• [FEA] Add Elliptic Bitcoin fraud example (#221) @alexbarghi-nv

🛠️ Improvements

• remove cugraph-dgl CODEOWNERS rule, skip most CI on CODEOWNERS-only PRs (#264) @jameslamb
• Disable codecov comments (#256) @bdice
• Expose cugraph_pyg.tensor Subpackage (#253) @alexbarghi-nv
• fix(docker): use versioned -latest tag for all rapidsai images (#246) @gforsyth
• Propagate Changes from cuGraph Distributed Sampler (metadata addition) (#245) @alexbarghi-nv
• Migrate cugraph gnn packages to cugraph-pyg (#241) @nv-rliu
• Improving the taobao_mnmg example. (#238) @drivanov
• Use CUDA 12.9 in Conda, Devcontainers, Spark, GHA, etc. (#232) @jakirkham
• Remove nvidia and dask channels (#231) @vyasr
• Updates movielens_mnmg.py to use DDP (#229) @rlratzel
• Remove CUDA 11 from dependencies.yaml (#224) @KyleFromNVIDIA
• Remove Non-Unified API and Remaining TensorDict Code (#222) @alexbarghi-nv
• refactor(rattler): remove cuda11 options and general cleanup (#219) @gforsyth
• Remove CUDA 11 devcontainers and update CI scripts (#218) @bdice
• use branch-25.08 of shared-workflows (#216) @jameslamb
• stop uploading packages to downloads.rapids.ai (#215) @jameslamb
• Remove cuGraph-DGL (#210) @alexbarghi-nv

v25.06.00

🚨 Breaking Changes

• Remove Dask API from cuGraph-DGL (#199) @alexbarghi-nv
• [IMP] Make WholeGraph a Hard Dependency of cuGraph-PyG, Remove cuDF Spilling from Examples (#172) @alexbarghi-nv
• [IMP] Remove SG and SNMG Examples (#171) @alexbarghi-nv
• Remove Dask API from cuGraph-PyG (#166) @alexbarghi-nv
• [FEA] Support Unified WholeGraph FeatureStore and GraphStore (#156) @alexbarghi-nv

🐛 Bug Fixes

• [BUG] Fix Broken DGL Test (#223) @alexbarghi-nv
• [BUG] Fix DGL Test (#209) @alexbarghi-nv
• [BUG] Remove FeatureStore tests about to break (#207) @alexbarghi-nv
• Branch 25.06 forward resolve (fix) (#165) @alexbarghi-nv

📖 Documentation

• [DOC] Add warning of pending DGL removal (#191) @alexbarghi-nv
• [DOC] Update the README to show the current stack (#190) @alexbarghi-nv

🚀 New Features

• Branch 25.06 forward resolve (#163) @alexbarghi-nv

🛠️ Improvements

• use 'rapids-init-pip' in wheel CI, other CI changes (#212) @jameslamb
• Finish CUDA 12.9 migration and use branch-25.06 workflows (#208) @bdice
• Update to clang 20 (#205) @bdice
• Quote head_rev in conda recipes (#204) @bdice
• use GitHub Actions artifacts (#203) @jameslamb
• Enable device code compression (#202) @robertmaynard
• chore: skip DGL tests on Python 3.13 (#201) @gforsyth
• Build and test with CUDA 12.9.0 (#200) @bdice
• Remove Dask API from cuGraph-DGL (#199) @alexbarghi-nv
• Add support for Python 3.13 (#197) @gforsyth
• fix(libwholegraph): pass branch, sha, and date (#195) @gforsyth
• fix(libwholegraph): set build_type to "nightly" and also publish wheel (#193) @gforsyth
• Reenable example tests (#192) @alexbarghi-nv
• added downloads from github (#189) @VenkateshJaya
• feat: add libwholegraph wheel (#182) @gforsyth
• feat(rattler): port conda build recipe to rattler build (#178) @gforsyth
• Vendor RAPIDS.cmake (#177) @bdice
• Add ARM conda environments (#176) @bdice
• restrict dgl dependency to x86 (#175) @trxcllnt
• [IMP] Make WholeGraph a Hard Dependency of cuGraph-PyG, Remove cuDF Spilling from Examples (#172) @alexbarghi-nv
• [IMP] Remove SG and SNMG Examples (#171) @alexbarghi-nv
• cugraph-pyg: remove Dask dependencies and related test code (#168) @jameslamb
• Remove Dask API from cuGraph-PyG (#166) @alexbarghi-nv
• Moving wheel builds to specified location and uploading build artifacts to Github (#162) @VenkateshJaya
• [FEA] Support Unified WholeGraph FeatureStore and GraphStore (#156) @alexbarghi-nv

v25.04.02

🚨 Breaking Changes

• Modify CMakeLists to install major version (#181) @alexbarghi-nv
• Update PyTorch to use CUDA 12.6 (#155) @alexbarghi-nv
• Deprecate Unbuffered Sampling in cuGraph-PyG (#151) @alexbarghi-nv
• Update DGL Packages to PyTorch 2.4 (#148) @alexbarghi-nv
• Update CuPy Version to 13.2 (#132) @alexbarghi-nv
• [FEA] Support Link Prediction and Negative Sampling in cuGraph-DGL (#50) @alexbarghi-nv

🐛 Bug Fixes

• Modify CMakeLists to install major version (#181) @alexbarghi-nv
• [HOTFIX] Add deprecation warnings for DGL classes (#170) @alexbarghi-nv
• [FIX] Add mkl version, limit tensordict to 0.6.2 (#161) @alexbarghi-nv
• Fix update version (#157) @raydouglass
• Drop PyTorch Constraint for OGB (#153) @alexbarghi-nv
• Update DGL Packages to PyTorch 2.4 (#148) @alexbarghi-nv
• [BUG] Fix bug with drop_last when mod is 0 (#136) @alexbarghi-nv
• Update CuPy Version to 13.2 (#132) @alexbarghi-nv

🚀 New Features

• [FEA] New MovieLens Example, Add Timing to Taobao Example (#141) @alexbarghi-nv
• [FEA] Support Link Prediction and Negative Sampling in cuGraph-DGL (#50) @alexbarghi-nv

🛠️ Improvements

• Deprecate TensorDictFeatureStore in cuGraph-PyG (#160) @alexbarghi-nv
• Update PyTorch to use CUDA 12.6 (#155) @alexbarghi-nv
• Deprecate Unbuffered Sampling in cuGraph-PyG (#151) @alexbarghi-nv
• Use conda-build instead of conda-mambabuild (#150) @bdice
• add cugraph-notebook-codeowners to CODEOWNERS (#149) @AyodeAwe
• Require CMake 3.30.4 (#147) @robertmaynard
• Create Conda CI test env in one step (#144) @KyleFromNVIDIA
• Add build_type input to test.yaml (#143) @gforsyth
• Use shared-workflows branch-25.04 (#138) @bdice
• disallow fallback to Make in Python builds (#123) @jameslamb
• Forward-merge branch-25.02 to branch-25.04 (#122) @bdice
• Migrate to NVKS for amd64 CI runners (#119) @bdice

v25.04.01

🚨 Breaking Changes

• Update PyTorch to use CUDA 12.6 (#155) @alexbarghi-nv
• Deprecate Unbuffered Sampling in cuGraph-PyG (#151) @alexbarghi-nv
• Update DGL Packages to PyTorch 2.4 (#148) @alexbarghi-nv
• Update CuPy Version to 13.2 (#132) @alexbarghi-nv
• [FEA] Support Link Prediction and Negative Sampling in cuGraph-DGL (#50) @alexbarghi-nv

🐛 Bug Fixes

• [HOTFIX] Add deprecation warnings for DGL classes (#170) @alexbarghi-nv
• [FIX] Add mkl version, limit tensordict to 0.6.2 (#161) @alexbarghi-nv
• Fix update version (#157) @raydouglass
• Drop PyTorch Constraint for OGB (#153) @alexbarghi-nv
• Update DGL Packages to PyTorch 2.4 (#148) @alexbarghi-nv
• [BUG] Fix bug with drop_last when mod is 0 (#136) @alexbarghi-nv
• Update CuPy Version to 13.2 (#132) @alexbarghi-nv

🚀 New Features

• [FEA] New MovieLens Example, Add Timing to Taobao Example (#141) @alexbarghi-nv
• [FEA] Support Link Prediction and Negative Sampling in cuGraph-DGL (#50) @alexbarghi-nv

🛠️ Improvements

• Deprecate TensorDictFeatureStore in cuGraph-PyG (#160) @alexbarghi-nv
• Update PyTorch to use CUDA 12.6 (#155) @alexbarghi-nv
• Deprecate Unbuffered Sampling in cuGraph-PyG (#151) @alexbarghi-nv
• Use conda-build instead of conda-mambabuild (#150) @bdice
• add cugraph-notebook-codeowners to CODEOWNERS (#149) @AyodeAwe
• Require CMake 3.30.4 (#147) @robertmaynard
• Create Conda CI test env in one step (#144) @KyleFromNVIDIA
• Add build_type input to test.yaml (#143) @gforsyth
• Use shared-workflows branch-25.04 (#138) @bdice
• disallow fallback to Make in Python builds (#123) @jameslamb
• Forward-merge branch-25.02 to branch-25.04 (#122) @bdice
• Migrate to NVKS for amd64 CI runners (#119) @bdice

v25.04.00

🚨 Breaking Changes

• Update PyTorch to use CUDA 12.6 (#155) @alexbarghi-nv
• Deprecate Unbuffered Sampling in cuGraph-PyG (#151) @alexbarghi-nv
• Update DGL Packages to PyTorch 2.4 (#148) @alexbarghi-nv
• Update CuPy Version to 13.2 (#132) @alexbarghi-nv
• [FEA] Support Link Prediction and Negative Sampling in cuGraph-DGL (#50) @alexbarghi-nv

🐛 Bug Fixes

• [FIX] Add mkl version, limit tensordict to 0.6.2 (#161) @alexbarghi-nv
• Fix update version (#157) @raydouglass
• Drop PyTorch Constraint for OGB (#153) @alexbarghi-nv
• Update DGL Packages to PyTorch 2.4 (#148) @alexbarghi-nv
• [BUG] Fix bug with drop_last when mod is 0 (#136) @alexbarghi-nv
• Update CuPy Version to 13.2 (#132) @alexbarghi-nv

🚀 New Features

• [FEA] New MovieLens Example, Add Timing to Taobao Example (#141) @alexbarghi-nv
• [FEA] Support Link Prediction and Negative Sampling in cuGraph-DGL (#50) @alexbarghi-nv

🛠️ Improvements

• Deprecate TensorDictFeatureStore in cuGraph-PyG (#160) @alexbarghi-nv
• Update PyTorch to use CUDA 12.6 (#155) @alexbarghi-nv
• Deprecate Unbuffered Sampling in cuGraph-PyG (#151) @alexbarghi-nv
• Use conda-build instead of conda-mambabuild (#150) @bdice
• add cugraph-notebook-codeowners to CODEOWNERS (#149) @AyodeAwe
• Require CMake 3.30.4 (#147) @robertmaynard
• Create Conda CI test env in one step (#144) @KyleFromNVIDIA
• Add build_type input to test.yaml (#143) @gforsyth
• Use shared-workflows branch-25.04 (#138) @bdice
• disallow fallback to Make in Python builds (#123) @jameslamb
• Forward-merge branch-25.02 to branch-25.04 (#122) @bdice
• Migrate to NVKS for amd64 CI runners (#119) @bdice

v25.02.00

🚨 Breaking Changes

• update pip devcontainers to UCX 1.18 (#112) @jameslamb
• remove dependency on cugraph-ops (#99) @tingyu66
• [FEA] Support Heterogeneous Sampling in cuGraph-PyG (#82) @alexbarghi-nv

🐛 Bug Fixes

• Remove Build Directory (#107) @alexbarghi-nv
• Fix inputs for the workflow (#106) @vyasr
• Remove invalid conditional (#105) @vyasr

📖 Documentation

• build libwholegraph docs in CI (#96) @jameslamb

🚀 New Features

• Fix of create_node_classification_datasets (#128) @linhu-nv
• [FEA] Knowledge Graph/Graph Database Renumbering (#101) @alexbarghi-nv
• [FEA] Support Heterogeneous Sampling in cuGraph-PyG (#82) @alexbarghi-nv

🛠️ Improvements

• Use rapids-pip-retry in CI jobs that might need retries (#133) @gforsyth
• Revert CUDA 12.8 shared workflow branch changes (#121) @vyasr
• Deprecate the Dask API in cuGraph-PyG (#118) @alexbarghi-nv
• Build and test with CUDA 12.8.0 (#115) @bdice
• Support PyG 2.6 in cuGraph-PyG (#114) @alexbarghi-nv
• update pip devcontainers to UCX 1.18 (#112) @jameslamb
• enforce cmake-format and cmake-lint, other small packaging changes (#111) @jameslamb
• wholegraph: remove debugging details from CMake (#109) @jameslamb
• Use GCC 13 in CUDA 12 conda builds. (#108) @bdice
• Heterogeneous Link Prediction Example for cuGraph-PyG (#104) @alexbarghi-nv
• remove flake8, clang tools from wholegraph CMake (#103) @jameslamb
• simplify wholegraph CMake, other small building and testing changes (#102) @jameslamb
• Check if nightlies have succeeded recently enough (#100) @vyasr
• remove dependency on cugraph-ops (#99) @tingyu66
• remove unused dependencies.yaml entries, other small cleanup (#98) @jameslamb
• fix nightly docs workflow dependencies (#97) @jameslamb
• merge branch-24.12 into branch-25.02 (#95) @jameslamb
• Update version references in workflow (#93) @AyodeAwe
• Require approval to run CI on draft PRs (#80) @bdice
• add devcontainers (#79) @jameslamb
• skip conda-python-tests on arm64 (#70) @jameslamb
• Add breaking change workflow trigger (#64) @AyodeAwe

v24.12.00

🚨 Breaking Changes

• Disable RockyLinux Tests for DGL, Disable CUDA All 11.4 Tests (#89) @alexbarghi-nv

🐛 Bug Fixes

• Disable RockyLinux Tests for DGL, Disable CUDA All 11.4 Tests (#89) @alexbarghi-nv
• [Bugfix] Add stream synchronization before the scatter operation (#73) @chang-l
• [Bugfix] Dynamic load NVML symbols for better compatibility (#72) @chang-l

🚀 New Features

• [Feature] Add gather/scatter support 1D tensor (#74) @chang-l
• update wholegraph (#65) @zhuofan1123

🛠️ Improvements

• skip CUDA 11.4 conda-python-tests jobs (#91) @jameslamb
• Relax PyTorch upper bound (allowing 2.4) (#75) @jakirkham
• remove versioning workaround for nightlies (#69) @jameslamb
• fix devcontainer builds (#68) @jameslamb
• enforce wheel size limits, README formatting in CI (#67) @jameslamb
• start publishing packages (#66) @jameslamb
• add notebook tests, build.sh args (#62) @jameslamb
• add nightly builds/tests, remove some cugraph references (#61) @jameslamb
• remove docs support in build.sh, remove test_wheel.sh, other small cleanup (#60) @jameslamb
• add PR CI for cugraph-pyg and cugraph-dgl (#59) @jameslamb
• add full CI for wholegraph (#58) @jameslamb
• add alpha specs, pre-commit hook to enforce them (#57) @jameslamb
• introduce minimal CI for PRs (#56) @jameslamb
• Add Copy PR Bot (#55) @alexbarghi-nv