Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
5,929 Commits 32 Branches 59 Tags
4f42aa5f1224cdbf4e2fff8631e055182a3f7e35
Commit Graph
462 Commits
Author SHA1 Message Date
Bobby WangandHyunsu Cho 4f42aa5f12 [pyspark] make the model saved by pyspark compatible (#8219) 
Co-authored-by: Hyunsu Cho <chohyu01@cs.washington.edu>
 2022-09-20 16:43:49 +08:00
Bobby WangandHyunsu Philip Cho 520586ffa7 [pyspark] fix empty data issue when constructing DMatrix (#8245) 
Co-authored-by: Hyunsu Philip Cho <chohyu01@cs.washington.edu>
 2022-09-20 16:43:20 +08:00
Jiaming Yuan 2e63af6117 Mitigate flaky data iter test. (#8244) 
- Reduce the number of batches.
- Verify labels.
 2022-09-14 17:54:14 +08:00
Rong OuandHyunsu Cho a2686543a9 Common interface for collective communication (#8057) 
* implement broadcast for federated communicator

* implement allreduce

* add communicator factory

* add device adapter

* add device communicator to factory

* add rabit communicator

* add rabit communicator to the factory

* add nccl device communicator

* add synchronize to device communicator

* add back print and getprocessorname

* add python wrapper and c api

* clean up types

* fix non-gpu build

* try to fix ci

* fix std::size_t

* portable string compare ignore case

* c style size_t

* fix lint errors

* cross platform setenv

* fix memory leak

* fix lint errors

* address review feedback

* add python test for rabit communicator

* fix failing gtest

* use json to configure communicators

* fix lint error

* get rid of factories

* fix cpu build

* fix include

* fix python import

* don't export collective.py yet

* skip collective communicator pytest on windows

* add review feedback

* update documentation

* remove mpi communicator type

* fix tests

* shutdown the communicator separately

Co-authored-by: Hyunsu Cho <chohyu01@cs.washington.edu>
 2022-09-12 15:21:12 -07:00
Jiaming Yuan b5eb36f1af Add max_cat_threshold to GPU and handle missing cat values. (#8212) 2022-09-07 00:57:51 +08:00
WeichenXu d03794ce7a [pyspark] Add param validation for "objective" and "eval_metric" param, and remove invalid booster params (#8173) 
Signed-off-by: Weichen Xu <weichen.xu@databricks.com>
 2022-08-24 15:29:43 +08:00
WeichenXu f4628c22a4 [pyspark] Implement SparkXGBRanker estimator (#8172) 
Signed-off-by: Weichen Xu <weichen.xu@databricks.com>
 2022-08-23 02:35:19 +08:00
WeichenXu 53d2a733b0 [pyspark] Make Xgboost estimator support using sparse matrix as optimization (#8145) 
Signed-off-by: Weichen Xu <weichen.xu@databricks.com>
 2022-08-19 01:57:28 +08:00
Jiaming Yuan 36e7c5364d [dask] Deterministic rank assignment. (#8018) 2022-08-11 19:17:58 +08:00
Jiaming Yuan 570f8ae4ba Use black on more Python files. (#8137) 2022-08-11 01:38:11 +08:00
Jiaming Yuan 446d536c23 Fix loading DMatrix binary in distributed env. (#8149) 
- Try to load DMatrix binary before trying to parse text input.
- Remove some unmaintained code.
 2022-08-10 22:53:16 +08:00
Jiaming Yuan 9ae547f994 Use config_context in sklearn interface. (#8141) 2022-08-09 14:48:54 +08:00
Bobby Wang 03cc3b359c [pyspark] support a list of feature column names (#8117) 2022-08-08 17:05:27 +08:00
Jiaming Yuan d87f69215e Quantile DMatrix for CPU. (#8130) 
- Add a new `QuantileDMatrix` that works for both CPU and GPU.
- Deprecate `DeviceQuantileDMatrix`.
 2022-08-02 15:51:23 +08:00
Jiaming Yuan 546de5efd2 [pyspark] Cleanup data processing. (#8088) 
- Use numpy stack for handling list of arrays.
- Reuse concat function from dask.
- Prepare for `QuantileDMatrix`.
- Remove unused code.
- Use iterator for prediction to avoid initializing xgboost model
 2022-07-26 15:00:52 +08:00
Bobby Wang f801d3cf15 [PySpark] change the returning model type to string from binary (#8085) 
* [PySpark] change the returning model type to string from binary

XGBoost pyspark can be can be accelerated by RAPIDS Accelerator seamlessly by
changing the returning model type from binary to string.
 2022-07-19 18:39:20 +08:00
Jiaming Yuan 2365f82750 [dask] Mitigate non-deterministic test. (#8077) 2022-07-19 16:55:59 +08:00
Jiaming Yuan 647d3844dd Make test for categorical data deterministic. (#8080) 2022-07-15 14:48:39 +08:00
Jiaming Yuan dae7a41baa Update Python requirement to >=3.8. (#8071) 
Additional changes:
- Use mamba for CPU test on Jenkins.
- Cleanup CPU test dependencies.
- Restore some of the modin tests
 2022-07-14 18:01:47 +08:00
WeichenXuHyunsu ChoJiaming Yuan
 		176fec8789 PySpark XGBoost integration (#8020) 
Co-authored-by: Hyunsu Cho <chohyu01@cs.washington.edu>
Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
 2022-07-13 13:11:18 +08:00
Jiaming Yuan 8959622836 [dask] Use an invalid port for test. (#8064) 2022-07-13 11:59:02 +08:00
Jiaming Yuan 4a87ea49b8 Reduce regularization for CPU gblinear. (#8013) 2022-06-21 01:05:27 +08:00
Jiaming Yuan b90c6d25e8 Implement max_cat_threshold for CPU. (#7957) 2022-06-04 11:02:46 +08:00
Jiaming Yuan 13b15e07e8 Handle formatted JSON input. (#7953) 2022-06-01 16:20:58 +08:00
Jiaming Yuan bde4f25794 Handle missing categorical value in CPU evaluator. (#7948) 2022-05-27 14:15:47 +08:00
Jiaming Yuan 18cbebaeb9 Unify the cat split storage for CPU. (#7937) 
* Unify the cat split storage for CPU.

* Cleanup.

* Workaround.
 2022-05-26 04:14:40 -07:00
Jiaming Yuan 606be9e663 Handle missing values in one hot splits. (#7934) 2022-05-24 20:48:41 +08:00
Jiaming Yuan 474366c020 Add convergence test for sparse datasets. (#7922) 2022-05-23 18:07:26 +08:00
Jiaming Yuan f93a727869 Address remaining mypy errors in python package. (#7914) 2022-05-18 22:46:15 +08:00
Rong Ou 77d4a53c32 use RabitContext intead of init/finalize (#7911) 2022-05-17 12:15:41 +08:00
Jiaming Yuan db80671d6b Fix monotone constraint with tuple input. (#7891) 2022-05-13 04:00:03 +08:00
Jiaming Yuan 46e0bce212 Use maximum category in sketch. (#7853) 2022-05-05 19:56:49 +08:00
Jiaming Yuan fdf533f2b9 [POC] Experimental support for l1 error. (#7812) 
Support adaptive tree, a feature supported by both sklearn and lightgbm.  The tree leaf is recomputed based on residue of labels and predictions after construction.

For l1 error, the optimal value is the median (50 percentile).

This is marked as experimental support for the following reasons:
- The value is not well defined for distributed training, where we might have empty leaves for local workers. Right now I just use the original leaf value for computing the average with other workers, which might cause significant errors.
- Some follow-ups are required, for exact, pruner, and optimization for quantile function. Also, we need to calculate the initial estimation.
 2022-04-26 21:41:55 +08:00
Jiaming Yuan 332380479b Avoid warning in np primitive type tests. (#7833) 2022-04-23 02:07:01 +08:00
Jiaming Yuan c70fa502a5 Expose feature_types to sklearn interface. (#7821) 2022-04-21 20:23:35 +08:00
Jiaming Yuan 52d4eda786 Deprecate use_label_encoder in XGBClassifier. (#7822) 
* Deprecate `use_label_encoder` in XGBClassifier.

* We have removed the encoder, now prepare to remove the indicator.
 2022-04-21 13:14:02 +08:00
Jiaming Yuan 9150fdbd4d Support pandas nullable types. (#7760) 2022-03-30 08:51:52 +08:00
Jiaming Yuan a50b84244e Cleanup configuration for constraints. (#7758) 2022-03-29 04:22:46 +08:00
Jiaming Yuan 3c9b04460a Move num_parallel_tree to model parameter. (#7751) 
The size of forest should be a property of model itself instead of a training
hyper-parameter.
 2022-03-29 02:32:42 +08:00
Jiaming Yuan 8b3ecfca25 Mitigate flaky tests. (#7749) 
* Skip non-increasing test with external memory when subsample is used.
* Increase bin numbers for boost from prediction test. This mitigates the effect of
  non-deterministic partitioning.
 2022-03-28 21:20:50 +08:00
Jiaming Yuan 4d81c741e9 External memory support for hist (#7531) 
* Generate column matrix from gHistIndex.
* Avoid synchronization with the sparse page once the cache is written.
* Cleanups: Remove member variables/functions, change the update routine to look like approx and gpu_hist.
* Remove pruner.
 2022-03-22 00:13:20 +08:00
Xiaochang WuJiaming YuanZhang Zhang
 		613ec36c5a Support building SimpleDMatrix from Arrow data format (#7512) 
* Integrate with Arrow C data API.
* Support Arrow dataset.
* Support Arrow table.

Co-authored-by: Xiaochang Wu <xiaochang.wu@intel.com>
Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
Co-authored-by: Zhang Zhang <zhang.zhang@intel.com>
 2022-03-15 13:25:19 +08:00
Jiaming Yuan a62a3d991d [dask] prediction with categorical data. (#7708) 2022-03-10 00:21:48 +08:00
Cheng Li a92e0f6240 multi groups in the constraints (#7711) 2022-03-01 18:10:15 +08:00
Jiaming Yuan 83a66b4994 Support categorical data for hist. (#7695) 
* Extract partitioner from hist.
* Implement categorical data support by passing the gradient index directly into the partitioner.
* Organize/update document.
* Remove code for negative hessian.
 2022-02-25 03:47:14 +08:00
Jiaming Yuan f08c5dcb06 Cleanup some pylint errors. (#7667) 
* Cleanup some pylint errors.

* Cleanup pylint errors in rabit modules.
* Make data iter an abstract class and cleanup private access.
* Cleanup no-self-use for booster.
 2022-02-19 18:53:12 +08:00
Jiaming Yuan 7366d3b20c Ensure models with categorical splits don't use old binary format. (#7666) 2022-02-19 08:05:28 +08:00
Jiaming Yuan 0d0abe1845 Support optimal partitioning for GPU hist. (#7652) 
* Implement `MaxCategory` in quantile.
* Implement partition-based split for GPU evaluation.  Currently, it's based on the existing evaluation function.
* Extract an evaluator from GPU Hist to store the needed states.
* Added some CUDA stream/event utilities.
* Update document with references.
* Fixed a bug in approx evaluator where the number of data points is less than the number of categories.
 2022-02-15 03:03:12 +08:00
Jiaming Yuan 5cd1f71b51 [dask] Improve configuration for port. (#7645) 
- Try port 0 to let the OS return the available port.
- Add port configuration.
 2022-02-14 21:34:34 +08:00
Jiaming Yuan 3e693e4f97 [dask] Fix nthread config with dask sklearn wrapper. (#7633) 2022-02-08 06:38:32 +08:00