|
| 1 | +# cuml 25.04.00 (9 Apr 2025) |
| 2 | + |
| 3 | +## 🚨 Breaking Changes |
| 4 | + |
| 5 | +- Promote experimental FIL to stable ([#6464](https://github.com/rapidsai/cuml/pull/6464)) [@wphicks](https://github.com/wphicks) |
| 6 | +- Support non-trivial `classes_` in `LogisticRegression` ([#6346](https://github.com/rapidsai/cuml/pull/6346)) [@jcrist](https://github.com/jcrist) |
| 7 | +- Use new rapids-logger library ([#6289](https://github.com/rapidsai/cuml/pull/6289)) [@vyasr](https://github.com/vyasr) |
| 8 | + |
| 9 | +## 🐛 Bug Fixes |
| 10 | + |
| 11 | +- Unblock CI for 25.04 ([#6519](https://github.com/rapidsai/cuml/pull/6519)) [@csadorf](https://github.com/csadorf) |
| 12 | +- Skip test_rf_classification_seed for cudf.pandas tests. ([#6500](https://github.com/rapidsai/cuml/pull/6500)) [@csadorf](https://github.com/csadorf) |
| 13 | +- Remove dask to sparse workarounds ([#6489](https://github.com/rapidsai/cuml/pull/6489)) [@TomAugspurger](https://github.com/TomAugspurger) |
| 14 | +- Fix accidental required sklearn import ([#6483](https://github.com/rapidsai/cuml/pull/6483)) [@jcrist](https://github.com/jcrist) |
| 15 | +- FIX Pass instances to the constructor of the meta-estimator in `SVC` ([#6471](https://github.com/rapidsai/cuml/pull/6471)) [@betatim](https://github.com/betatim) |
| 16 | +- Fix compiler dependency on ARM ([#6456](https://github.com/rapidsai/cuml/pull/6456)) [@bdice](https://github.com/bdice) |
| 17 | +- Revert "Temporarily increase `max_days_without_success` ([#6390)" (#6455](https://github.com/rapidsai/cuml/pull/6390)" (#6455)) [@divyegala](https://github.com/divyegala) |
| 18 | +- Enhance argument handling in UniversalBase to accept NoneType ([#6453](https://github.com/rapidsai/cuml/pull/6453)) [@csadorf](https://github.com/csadorf) |
| 19 | +- Fix UMAP transform ([#6449](https://github.com/rapidsai/cuml/pull/6449)) [@viclafargue](https://github.com/viclafargue) |
| 20 | +- Fallback to CPU for sparse inputs for KMeans ([#6448](https://github.com/rapidsai/cuml/pull/6448)) [@csadorf](https://github.com/csadorf) |
| 21 | +- Limited support for array-like inputs ([#6442](https://github.com/rapidsai/cuml/pull/6442)) [@csadorf](https://github.com/csadorf) |
| 22 | +- Fix `test_accuracy_score` test on `cudf.pandas` build ([#6439](https://github.com/rapidsai/cuml/pull/6439)) [@jcrist](https://github.com/jcrist) |
| 23 | +- Support positional args in `cuml.accel` estimators ([#6423](https://github.com/rapidsai/cuml/pull/6423)) [@jcrist](https://github.com/jcrist) |
| 24 | +- Fix HDBSCAN python docs related to `metric` ([#6422](https://github.com/rapidsai/cuml/pull/6422)) [@divyegala](https://github.com/divyegala) |
| 25 | +- declare runtime dependency on 'packaging', update scikit-learn and hdbscan for cuml-cpu to match cuml ([#6420](https://github.com/rapidsai/cuml/pull/6420)) [@jameslamb](https://github.com/jameslamb) |
| 26 | +- Fix forwarding of `initial_alpha` and `learning_rate` in UMAP ([#6417](https://github.com/rapidsai/cuml/pull/6417)) [@jcrist](https://github.com/jcrist) |
| 27 | +- Implement Ridge .solver_ estimated attribute ([#6415](https://github.com/rapidsai/cuml/pull/6415)) [@csadorf](https://github.com/csadorf) |
| 28 | +- Fix multi-target prediction in linear models ([#6414](https://github.com/rapidsai/cuml/pull/6414)) [@csadorf](https://github.com/csadorf) |
| 29 | +- Ensure `output_type="pandas"` returns user-facing pandas ([#6407](https://github.com/rapidsai/cuml/pull/6407)) [@jcrist](https://github.com/jcrist) |
| 30 | +- Ensure `LinearSVC` supports all input types ([#6404](https://github.com/rapidsai/cuml/pull/6404)) [@jcrist](https://github.com/jcrist) |
| 31 | +- Make UMAP callback pickleable ([#6402](https://github.com/rapidsai/cuml/pull/6402)) [@jcrist](https://github.com/jcrist) |
| 32 | +- Correctly align trees in experimental FIL ([#6397](https://github.com/rapidsai/cuml/pull/6397)) [@wphicks](https://github.com/wphicks) |
| 33 | +- A few log level handling cleanups ([#6393](https://github.com/rapidsai/cuml/pull/6393)) [@jcrist](https://github.com/jcrist) |
| 34 | +- Fix cupy error in `KernelDensity` with `epanechnikov` ([#6388](https://github.com/rapidsai/cuml/pull/6388)) [@jcrist](https://github.com/jcrist) |
| 35 | +- Check if both ground truth and result are NaN in MSEObjectiveTest for RF ([#6387](https://github.com/rapidsai/cuml/pull/6387)) [@wphicks](https://github.com/wphicks) |
| 36 | +- Support non-native endian inputs to `LabelEncoder` ([#6384](https://github.com/rapidsai/cuml/pull/6384)) [@jcrist](https://github.com/jcrist) |
| 37 | +- Fix hypothesis test in `test_kernel_ridge.py` ([#6382](https://github.com/rapidsai/cuml/pull/6382)) [@jcrist](https://github.com/jcrist) |
| 38 | +- Remove debug logging from nearest neighbors test ([#6376](https://github.com/rapidsai/cuml/pull/6376)) [@csadorf](https://github.com/csadorf) |
| 39 | +- Ensure FIL CPU can be run without an available GPU ([#6373](https://github.com/rapidsai/cuml/pull/6373)) [@wphicks](https://github.com/wphicks) |
| 40 | +- Correctly pass through unhashable objects during hyperparam lookup ([#6369](https://github.com/rapidsai/cuml/pull/6369)) [@wphicks](https://github.com/wphicks) |
| 41 | +- Correct translation of RandomForest criterion hyperparameter ([#6363](https://github.com/rapidsai/cuml/pull/6363)) [@wphicks](https://github.com/wphicks) |
| 42 | +- UMAP fallback improvements ([#6358](https://github.com/rapidsai/cuml/pull/6358)) [@viclafargue](https://github.com/viclafargue) |
| 43 | +- A few GPU<->CPU interop fixes ([#6355](https://github.com/rapidsai/cuml/pull/6355)) [@jcrist](https://github.com/jcrist) |
| 44 | +- Fix output type of `KernelRidge.predict` ([#6354](https://github.com/rapidsai/cuml/pull/6354)) [@jcrist](https://github.com/jcrist) |
| 45 | +- Skip test_extract_partitions_shape test. ([#6338](https://github.com/rapidsai/cuml/pull/6338)) [@csadorf](https://github.com/csadorf) |
| 46 | +- Require sphinx<8.2.0 ([#6336](https://github.com/rapidsai/cuml/pull/6336)) [@csadorf](https://github.com/csadorf) |
| 47 | +- Fix call to NearestNeighbors.kneighbors() without args. ([#6333](https://github.com/rapidsai/cuml/pull/6333)) [@csadorf](https://github.com/csadorf) |
| 48 | +- Commenting out debug print statements in UMAP ([#6332](https://github.com/rapidsai/cuml/pull/6332)) [@viclafargue](https://github.com/viclafargue) |
| 49 | +- Correct UMAP dispatch trigger condition ([#6330](https://github.com/rapidsai/cuml/pull/6330)) [@viclafargue](https://github.com/viclafargue) |
| 50 | +- Fix issue caused by None arg provided by sklearn pipeline ([#6326](https://github.com/rapidsai/cuml/pull/6326)) [@viclafargue](https://github.com/viclafargue) |
| 51 | +- Improvements for interoperability of Random Forest ([#6320](https://github.com/rapidsai/cuml/pull/6320)) [@dantegd](https://github.com/dantegd) |
| 52 | +- Improvements for PCA CPU interoperability ([#6319](https://github.com/rapidsai/cuml/pull/6319)) [@dantegd](https://github.com/dantegd) |
| 53 | +- Plumb `metric` and `metric_kwds` through to UMAP with `nn_descent` ([#6304](https://github.com/rapidsai/cuml/pull/6304)) [@jcrist](https://github.com/jcrist) |
| 54 | +- Use `L2Expanded`/`L2SqrtExpanded` instead of the `*Unexpanded` variants ([#6303](https://github.com/rapidsai/cuml/pull/6303)) [@jcrist](https://github.com/jcrist) |
| 55 | +- Fix UMAP issues with large inputs ([#6245](https://github.com/rapidsai/cuml/pull/6245)) [@viclafargue](https://github.com/viclafargue) |
| 56 | +- Fix `LogisticRegression.decision_function` output shape ([#6235](https://github.com/rapidsai/cuml/pull/6235)) [@jcrist](https://github.com/jcrist) |
| 57 | +- Compute `noise_variance_` in PCA implementation ([#6234](https://github.com/rapidsai/cuml/pull/6234)) [@jcrist](https://github.com/jcrist) |
| 58 | +- Multiple CPU interop fixes for serialization and cloning ([#6223](https://github.com/rapidsai/cuml/pull/6223)) [@dantegd](https://github.com/dantegd) |
| 59 | + |
| 60 | +## 📖 Documentation |
| 61 | + |
| 62 | +- Clarify support in docs ([#6452](https://github.com/rapidsai/cuml/pull/6452)) [@bdice](https://github.com/bdice) |
| 63 | +- Provide docs for cuml.accel command line feature ([#6322](https://github.com/rapidsai/cuml/pull/6322)) [@wphicks](https://github.com/wphicks) |
| 64 | + |
| 65 | +## 🚀 New Features |
| 66 | + |
| 67 | +- Promote experimental FIL to stable ([#6464](https://github.com/rapidsai/cuml/pull/6464)) [@wphicks](https://github.com/wphicks) |
| 68 | +- UMAP 32bits dispatch mechanism ([#6314](https://github.com/rapidsai/cuml/pull/6314)) [@viclafargue](https://github.com/viclafargue) |
| 69 | +- Add aarch64 conda environments ([#6226](https://github.com/rapidsai/cuml/pull/6226)) [@jcrist](https://github.com/jcrist) |
| 70 | +- SVM cpu/gpu interop ([#6089](https://github.com/rapidsai/cuml/pull/6089)) [@viclafargue](https://github.com/viclafargue) |
| 71 | + |
| 72 | +## 🛠️ Improvements |
| 73 | + |
| 74 | +- Don't use private APIs in FIL notebook ([#6488](https://github.com/rapidsai/cuml/pull/6488)) [@jcrist](https://github.com/jcrist) |
| 75 | +- Only use released versions of dask ([#6485](https://github.com/rapidsai/cuml/pull/6485)) [@jcrist](https://github.com/jcrist) |
| 76 | +- Move SVM working set log to debug ([#6477](https://github.com/rapidsai/cuml/pull/6477)) [@jcrist](https://github.com/jcrist) |
| 77 | +- C++ `SG_RF_TEST` OOMs in 11.4 nightly ([#6474](https://github.com/rapidsai/cuml/pull/6474)) [@divyegala](https://github.com/divyegala) |
| 78 | +- Revert "ci: pre-filter 11.4 jobs before they are enabled in shared workflows ([#6447)" (#6470](https://github.com/rapidsai/cuml/pull/6447)" (#6470)) [@divyegala](https://github.com/divyegala) |
| 79 | +- Support `__len__` in ensemble estimators ([#6468](https://github.com/rapidsai/cuml/pull/6468)) [@jcrist](https://github.com/jcrist) |
| 80 | +- Add deprecation warning to `cuml-cpu` builds ([#6466](https://github.com/rapidsai/cuml/pull/6466)) [@jcrist](https://github.com/jcrist) |
| 81 | +- Fix template argument of `DeviceSegmentedReduce` kernels ([#6459](https://github.com/rapidsai/cuml/pull/6459)) [@miscco](https://github.com/miscco) |
| 82 | +- Add scikit-learn acceleration tests to nightly CI tests ([#6457](https://github.com/rapidsai/cuml/pull/6457)) [@csadorf](https://github.com/csadorf) |
| 83 | +- Fix benchmark builds in devcontainers ([#6450](https://github.com/rapidsai/cuml/pull/6450)) [@bdice](https://github.com/bdice) |
| 84 | +- ci: pre-filter 11.4 jobs before they are enabled in shared workflows ([#6447](https://github.com/rapidsai/cuml/pull/6447)) [@gforsyth](https://github.com/gforsyth) |
| 85 | +- Reenable more UMAP tests on ARM ([#6446](https://github.com/rapidsai/cuml/pull/6446)) [@jcrist](https://github.com/jcrist) |
| 86 | +- Clean up after removing KMeans deprecation warning ([#6445](https://github.com/rapidsai/cuml/pull/6445)) [@betatim](https://github.com/betatim) |
| 87 | +- Port all conda recipes to `rattler-build` ([#6440](https://github.com/rapidsai/cuml/pull/6440)) [@gforsyth](https://github.com/gforsyth) |
| 88 | +- Remove the deprecation warning and switch to new default ([#6433](https://github.com/rapidsai/cuml/pull/6433)) [@betatim](https://github.com/betatim) |
| 89 | +- Remove old deprecations ([#6427](https://github.com/rapidsai/cuml/pull/6427)) [@jcrist](https://github.com/jcrist) |
| 90 | +- Fix oob_score=True handling with accelerator ([#6425](https://github.com/rapidsai/cuml/pull/6425)) [@betatim](https://github.com/betatim) |
| 91 | +- remove unused imports ([#6424](https://github.com/rapidsai/cuml/pull/6424)) [@jameslamb](https://github.com/jameslamb) |
| 92 | +- Disable codecov comments ([#6419](https://github.com/rapidsai/cuml/pull/6419)) [@jcrist](https://github.com/jcrist) |
| 93 | +- Require hypothesis tests have explicit examples ([#6409](https://github.com/rapidsai/cuml/pull/6409)) [@aamijar](https://github.com/aamijar) |
| 94 | +- Improve `accuracy_score` compatibility with `sklearn`. ([#6406](https://github.com/rapidsai/cuml/pull/6406)) [@jcrist](https://github.com/jcrist) |
| 95 | +- Drop `cub{Min, Max}` ([#6405](https://github.com/rapidsai/cuml/pull/6405)) [@miscco](https://github.com/miscco) |
| 96 | +- Move `accel` implementation fully to `cuml.accel` ([#6401](https://github.com/rapidsai/cuml/pull/6401)) [@jcrist](https://github.com/jcrist) |
| 97 | +- cuml-cpu: fix import issues, enable conda import tests ([#6400](https://github.com/rapidsai/cuml/pull/6400)) [@jameslamb](https://github.com/jameslamb) |
| 98 | +- Drop `numba_utils` ([#6396](https://github.com/rapidsai/cuml/pull/6396)) [@jakirkham](https://github.com/jakirkham) |
| 99 | +- Simplify `cpu_to_gpu` and `gpu_to_cpu` ([#6395](https://github.com/rapidsai/cuml/pull/6395)) [@jcrist](https://github.com/jcrist) |
| 100 | +- Adds get_level() ([#6392](https://github.com/rapidsai/cuml/pull/6392)) [@Intron7](https://github.com/Intron7) |
| 101 | +- Temporarily increase `max_days_without_success` ([#6390](https://github.com/rapidsai/cuml/pull/6390)) [@jcrist](https://github.com/jcrist) |
| 102 | +- Drop deprecated CCCL functors ([#6389](https://github.com/rapidsai/cuml/pull/6389)) [@miscco](https://github.com/miscco) |
| 103 | +- Solve forward merger conflicts of #6313 ([#6385](https://github.com/rapidsai/cuml/pull/6385)) [@dantegd](https://github.com/dantegd) |
| 104 | +- CI: Run extended accel mode tests ([#6377](https://github.com/rapidsai/cuml/pull/6377)) [@csadorf](https://github.com/csadorf) |
| 105 | +- Use conda-build instead of conda-mambabuild ([#6368](https://github.com/rapidsai/cuml/pull/6368)) [@bdice](https://github.com/bdice) |
| 106 | +- Don't log coverage output to terminal in CI ([#6367](https://github.com/rapidsai/cuml/pull/6367)) [@jcrist](https://github.com/jcrist) |
| 107 | +- Remove `KernelRidge` from `cuml.accel` ([#6361](https://github.com/rapidsai/cuml/pull/6361)) [@jcrist](https://github.com/jcrist) |
| 108 | +- Enable UVM in cuml.accel ([#6359](https://github.com/rapidsai/cuml/pull/6359)) [@wphicks](https://github.com/wphicks) |
| 109 | +- Drop deprecated `CUB` iterators in favor of `Thrust` iterators ([#6356](https://github.com/rapidsai/cuml/pull/6356)) [@miscco](https://github.com/miscco) |
| 110 | +- Adjust accel verbosity ([#6353](https://github.com/rapidsai/cuml/pull/6353)) [@csadorf](https://github.com/csadorf) |
| 111 | +- Avoid importing `umap` with `import cuml` ([#6352](https://github.com/rapidsai/cuml/pull/6352)) [@jcrist](https://github.com/jcrist) |
| 112 | +- Make proxy module patching skip patching if patched library is not installed ([#6348](https://github.com/rapidsai/cuml/pull/6348)) [@dantegd](https://github.com/dantegd) |
| 113 | +- Add new `cuml.accel` module ([#6347](https://github.com/rapidsai/cuml/pull/6347)) [@jcrist](https://github.com/jcrist) |
| 114 | +- Support non-trivial `classes_` in `LogisticRegression` ([#6346](https://github.com/rapidsai/cuml/pull/6346)) [@jcrist](https://github.com/jcrist) |
| 115 | +- Make `r2_score` compatible with Scikit-Learn ([#6337](https://github.com/rapidsai/cuml/pull/6337)) [@jcrist](https://github.com/jcrist) |
| 116 | +- Remove left-over code related to strict mode. ([#6335](https://github.com/rapidsai/cuml/pull/6335)) [@csadorf](https://github.com/csadorf) |
| 117 | +- Make sure specifying default value or nothing behaves the same in `KMeans` ([#6331](https://github.com/rapidsai/cuml/pull/6331)) [@betatim](https://github.com/betatim) |
| 118 | +- Make KernelRidge inherit from UniversalBase ([#6327](https://github.com/rapidsai/cuml/pull/6327)) [@betatim](https://github.com/betatim) |
| 119 | +- Reduce peak memory in `UMAP.fit`/`UMAP.fit_transform` ([#6323](https://github.com/rapidsai/cuml/pull/6323)) [@jcrist](https://github.com/jcrist) |
| 120 | +- Consolidate more Conda solves in CI ([#6321](https://github.com/rapidsai/cuml/pull/6321)) [@KyleFromNVIDIA](https://github.com/KyleFromNVIDIA) |
| 121 | +- Require CMake 3.30.4 ([#6318](https://github.com/rapidsai/cuml/pull/6318)) [@robertmaynard](https://github.com/robertmaynard) |
| 122 | +- Copy attributes from the original class to the proxy ([#6306](https://github.com/rapidsai/cuml/pull/6306)) [@betatim](https://github.com/betatim) |
| 123 | +- Use shared-workflows branch-25.04 ([#6298](https://github.com/rapidsai/cuml/pull/6298)) [@bdice](https://github.com/bdice) |
| 124 | +- Add build_type input field for `test.yaml` ([#6296](https://github.com/rapidsai/cuml/pull/6296)) [@gforsyth](https://github.com/gforsyth) |
| 125 | +- Do not use `LEGACY_PTX_ARCH` argument for `BlockReduce` ([#6290](https://github.com/rapidsai/cuml/pull/6290)) [@miscco](https://github.com/miscco) |
| 126 | +- Use new rapids-logger library ([#6289](https://github.com/rapidsai/cuml/pull/6289)) [@vyasr](https://github.com/vyasr) |
| 127 | +- disallow fallback to Make in wheel builds ([#6286](https://github.com/rapidsai/cuml/pull/6286)) [@jameslamb](https://github.com/jameslamb) |
| 128 | +- Forward-merge branch-25.02 to branch-25.04 ([#6283](https://github.com/rapidsai/cuml/pull/6283)) [@bdice](https://github.com/bdice) |
| 129 | +- Migrate to NVKS for amd64 CI runners ([#6280](https://github.com/rapidsai/cuml/pull/6280)) [@bdice](https://github.com/bdice) |
| 130 | +- Add `verify-codeowners` hook ([#6275](https://github.com/rapidsai/cuml/pull/6275)) [@KyleFromNVIDIA](https://github.com/KyleFromNVIDIA) |
| 131 | +- Speed up FIL and TSNE pytests in CI ([#6270](https://github.com/rapidsai/cuml/pull/6270)) [@dantegd](https://github.com/dantegd) |
| 132 | +- Bracnh 25.04 merge branch 25.02 ([#6268](https://github.com/rapidsai/cuml/pull/6268)) [@vyasr](https://github.com/vyasr) |
| 133 | +- Add shellcheck to pre-commit and fix warnings ([#6246](https://github.com/rapidsai/cuml/pull/6246)) [@gforsyth](https://github.com/gforsyth) |
| 134 | +- Prepare for n_init=auto in KMeans ([#6142](https://github.com/rapidsai/cuml/pull/6142)) [@betatim](https://github.com/betatim) |
| 135 | + |
1 | 136 | # cuml 25.02.00 (13 Feb 2025)
|
2 | 137 |
|
3 | 138 | ## 🚨 Breaking Changes
|
|
0 commit comments