Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
5,372 Commits 32 Branches 59 Tags
1d4d34563435740227c4151d10092253d9282316
Commit Graph
100 Commits
This Branch
This Branch
All Branches
Author SHA1 Message Date
Jiaming Yuan 1d4d345634 Tests for dask skl categorical data support. (#7054) 2021-06-24 16:33:57 +08:00
Jiaming Yuan da1ad798ca Convert numpy float to Python float in feat score. (#7047) 2021-06-21 20:58:43 +08:00
Jiaming Yuan bbfffb444d Fix race condition in CPU shap. (#7050) 2021-06-21 10:03:15 +08:00
Jiaming Yuan 29f8fd6fee Support categorical split in tree model dump. (#7036) 2021-06-18 16:46:20 +08:00
Jiaming Yuan 7968c0d051 Test on s390x. (#7038) 
* Fix && remove unused parameter.
 2021-06-18 14:55:08 +08:00
Jiaming Yuan 86715e4cd4 Support categorical data for dask functional interface and DQM. (#7043) 
* Support categorical data for dask functional interface and DQM.

* Implement categorical data support for GPU GK-merge.
* Add support for dask functional interface.
* Add support for DQM.

* Get newer cupy.
 2021-06-18 13:06:52 +08:00
Jiaming Yuan 7dd29ffd47 Implement feature score in GBTree. (#7041) 
* Categorical data support.
* Eliminate text parsing during feature score computation.
 2021-06-18 11:53:16 +08:00
Jiaming YuanandPhilip Hyunsu Cho dcd84b3979 [CI] Configure RAPIDS, dask, modin (#7033) 
Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>
 2021-06-18 10:27:51 +08:00
Jiaming Yuan d9799b09d0 Categorical data support for cuDF. (#7042) 
* Add support in DMatrix.
* Add support in DQM, except for iterator.
 2021-06-17 13:54:33 +08:00
Jiaming Yuan 5c2d7a18c9 Parallel model dump for trees. (#7040) 2021-06-15 14:08:26 +08:00
Jiaming Yuan 1faad825f4 Remove appveyor badge. [skip ci] (#7035) 2021-06-11 14:37:18 +08:00
Jiaming Yuan b56614e9b8 [R] Use new predict function. (#6819) 
* Call new C prediction API.
* Add `strict_shape`.
* Add `iterationrange`.
* Update document.
 2021-06-11 13:03:29 +08:00
Jiaming YuanandShvets Kirill f79cc4a7a4 Implement categorical prediction for CPU and GPU predict leaf. (#7001) 
* Categorical prediction with CPU predictor and GPU predict leaf.

* Implement categorical prediction for CPU prediction.
* Implement categorical prediction for GPU predict leaf.
* Refactor the prediction functions to have a unified get next node function.

Co-authored-by: Shvets Kirill <kirill.shvets@intel.com>
 2021-06-11 10:11:45 +08:00
Jiaming Yuan 72f9daf9b6 Fix gpu_id with custom objective. (#7015) 2021-06-09 14:51:17 +08:00
Jiaming Yuan 7beb2f7fae Hide symbols in CI build + hide symbols for C and CUDA (#6798) 
* Hide symbols in CI build.
* Hide symbols for other languages.
 2021-06-04 02:35:46 +08:00
Jiaming Yuan c4b9f4f622 Add enable_categorical to sklearn. (#7011) 2021-06-04 02:29:14 +08:00
Jiaming Yuan ee4f51a631 Support for all primitive types from array. (#7003) 
* Change C API name.
* Test for all primitive types from array.
* Add native support for CPU 128 float.
* Convert boolean and float16 in Python.

* Fix dask version for now.
 2021-06-01 08:34:48 +08:00
Jiaming Yuan 816b789bf0 Add predictor to skl constructor. (#7000) 2021-05-29 04:52:56 +08:00
Jiaming Yuan 89a49cf30e Fix dask predict on DaskDMatrix with iteration_range. (#7005) 2021-05-29 04:43:12 +08:00
Jiaming Yuan 4cf95a6041 Support numpy array interface (#6998) 2021-05-27 16:08:22 +08:00
Jiaming Yuan ab6fd304c4 [Python] Change development release postfix to dev (#6988) 2021-05-27 16:06:51 +08:00
Jiaming Yuan 29d6a5e2b8 [CI] Move appveyor tests to action (#6986) 
* Drop support for VS14, use VS15 instead.
* Drop support for mingw.
* Remove debug build.
* Split up jvm tests.
* Split up Python tests.
 2021-05-27 04:49:45 +08:00
Jiaming Yuan 86e60e3ba8 Guard against index error in prediction. (#6982) 
* Remove `best_ntree_limit` from documents.
 2021-05-25 23:24:59 +08:00
Jiaming Yuan a4bc7ecf27 Restore R cache on github action. (#6985) 2021-05-25 18:53:44 +08:00
Jiaming Yuan 6e52aefb37 Revert OMP guard. (#6987) 
The guard protects the global variable from being changed by XGBoost.  But this leads to a
bug that the `n_threads` parameter is no longer used after the first iteration.  This is
due to the fact that `omp_set_num_threads` is only called once in `Learner::Configure` at
the beginning of the training process.

The guard is still useful for `gpu_id`, since this is called all the times in our codebase
doesn't matter which iteration we are currently running.
 2021-05-25 08:56:28 +08:00
Jiaming Yuan cf06a266a8 [dask][doc] Wrap the example in main guard. (#6979) 2021-05-25 08:24:47 +08:00
Jiaming Yuan 5cb51a191e [dask][doc] Add small example for sklearn interface. (#6970) 2021-05-19 13:50:45 +08:00
Jiaming Yuan 7e846bb965 Fix prediction on df with latest dask. (#6969) 2021-05-19 12:23:03 +08:00
Jiaming Yuan 6e104f0570 Add news for 1.4.2. [skip ci] (#6963) 2021-05-17 02:50:55 +08:00
Jiaming Yuan d245bc891e Add tolerance to early stopping. (#6942) 2021-05-14 00:19:51 +08:00
Jiaming Yuan 44cc9c04ea Fix multiclass auc with empty dataset. (#6947) 2021-05-12 15:01:14 +08:00
Jiaming Yuan 05ac415780 [dask] Set dataframe index in predict. (#6944) 2021-05-12 13:24:21 +08:00
Jiaming Yuan 37ad60fe25 Enforce input data is not object. (#6927) 
* Check for object data type.
* Allow strided arrays with greater underlying buffer size.
 2021-05-02 00:09:01 +08:00
Jiaming Yuan a1d23f6613 Relax test for decision stump in distributed environment. (#6919) 2021-04-30 09:04:11 +08:00
Jiaming Yuan 45ddc39c1d Relax shotgun test. (#6918) 2021-04-30 09:03:12 +08:00
Jiaming Yuan 34df1f588b Reduce Travis environment setup time. (#6912) 
* Remove unused r from travis.
* Don't update homebrew.
* Don't install indirect/unused dependencies like libgit2, wget, openssl.
* Move graphviz installation to conda.
 2021-04-30 09:02:40 +08:00
Jiaming Yuan b31d37eac5 [CI] Fix custom metric test with empty dataset. (#6917) 2021-04-30 09:00:05 +08:00
Jiaming Yuan db6285fb55 [CI] Skip external memory gtest on osx. (#6901) 2021-04-30 08:59:33 +08:00
Jiaming Yuan ef473b1f09 Disable pylint error. (#6911) 2021-04-29 01:01:37 +08:00
Jiaming Yuan 8760ec4827 Ensure predict leaf output 1-dim vector where there's only 1 tree. (#6889) 2021-04-23 15:07:48 +08:00
Jiaming Yuan 54afa3ac7a Relax shotgun test. (#6900) 
It's non-deterministic algorithm, the test is flaky.
 2021-04-23 13:01:44 +08:00
Jiaming Yuan a2ecbdaa31 Add an API guard to prevent global variables being changed. (#6891) 2021-04-23 10:27:57 +08:00
Jiaming YuanandPhilip Hyunsu Cho 896aede340 Reorganize the installation documents. (#6877) 
* Split up installation and building from source.
* Use consistent section titles.

Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>
 2021-04-22 04:48:32 +08:00
Jiaming Yuan 74b41637de Revert "[jvm-packages] Add XGBOOST_RABIT_TRACKER_IP_FOR_TEST to set rabit tracker IP. (#6869)" (#6886) 
This reverts commit 2828da3c4c.
 2021-04-21 11:20:10 -07:00
Jiaming Yuan 233bdf105f Remove setDaemon in tracker. (#6872) 2021-04-22 01:57:13 +08:00
Jiaming Yuan 71b938f608 1.4.1 release news. (#6876) 2021-04-22 01:55:57 +08:00
Jiaming Yuan 146549260a Bump version to 1.5.0 snapshot in master. (#6875) 2021-04-22 01:53:44 +08:00
Jiaming Yuan bec2b4f094 Revert "Use CPU input for test_boost_from_prediction. (#6818)" (#6858) 
This reverts commit 74f3a2f4b5.
 2021-04-20 14:54:02 +08:00
Jiaming Yuan 556a83022d Implement unified update prediction cache for (gpu_)hist. (#6860) 
* Implement utilites for linalg.
* Unify the update prediction cache functions.
* Implement update prediction cache for multi-class gpu hist.
 2021-04-17 00:29:34 +08:00
Jiaming Yuan 1b26a2a561 Copy output data for argsort. (#6866) 
Fix GPU AUC.
 2021-04-16 21:05:01 +08:00
Jiaming Yuan a5d7094a45 Update documents. (#6856) 
* Add early stopping section to prediction doc.
* Remove best_ntree_limit.
* Better doxygen output.
 2021-04-16 12:41:03 +08:00
Jiaming Yuan bccb7e87d1 Update dmlc-core. (#6862) 
* Install pandoc, pandoc-citeproc on CI.
 2021-04-16 00:14:17 +08:00
Jiaming Yuan dee5ef2dfd Typehint for Sklearn. (#6799) 2021-04-14 06:55:21 +08:00
Jiaming Yuan 3d919db0c0 Fix pip release script. [skip ci] (#6845) 2021-04-14 06:46:02 +08:00
Jiaming Yuan b9a4f3336a 1.4 release notes. (#6843) 2021-04-13 08:38:27 +08:00
Jiaming Yuan f294c4e023 Use constexpr in dh::CopyIf. (#6828) 2021-04-08 07:37:47 +08:00
Jiaming Yuan 7bcc8b3e5c Use batched copy if. (#6826) 2021-04-06 10:34:04 +08:00
Jiaming Yuan 7e06c81894 Fix approximated predict contribution. (#6811) 2021-04-03 02:15:03 +08:00
Jiaming Yuan 0cced530ea [doc] Clarify prediction function. (#6813) 2021-04-03 02:12:04 +08:00
Jiaming Yuan b1fdb220f4 Remove deprecated n_gpus parameter. (#6821) 2021-04-02 03:02:32 +08:00
Jiaming Yuan 74f3a2f4b5 Use CPU input for test_boost_from_prediction. (#6818) 2021-04-02 00:11:35 +08:00
Jiaming YuanandPhilip Hyunsu Cho 47b62480af More general predict proba. (#6817) 
* Use `output_margin` for `softmax`.
* Add test for dask binary cls.

Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>
 2021-04-01 19:52:12 +08:00
Jiaming Yuan a5c852660b Update document for sklearn model IO. (#6809) 
* Update the use of JSON.
* Remove unnecessary type cast.
 2021-04-01 15:52:36 +08:00
Jiaming Yuan 905fdd3e08 Fix typos in AUC. (#6795) 2021-03-31 16:35:42 +08:00
Jiaming Yuan ca998df912 Clarify the behavior of use_rmm. (#6808) 
* Clarify the `use_rmm` flag in document and demo.
 2021-03-31 15:43:11 +08:00
Jiaming Yuan 3039dd194b Don't estimate sketch batch size when rmm is used. (#6807) 2021-03-31 15:29:56 +08:00
Jiaming Yuan 10ae0f9511 Fix doc for apply method. (#6796) 2021-03-31 15:28:31 +08:00
Jiaming Yuan 138fe8516a Remove unnecessary calls to iota. (#6797) 2021-03-31 15:27:23 +08:00
Jiaming Yuan 79b8b560d2 Optimize dart inplace predict perf. (#6804) 2021-03-31 15:20:54 +08:00
Jiaming Yuan a59c7323b4 Fix inplace predict missing value. (#6787) 2021-03-27 05:36:10 +08:00
Jiaming Yuan 5c87c2bba8 Update demo for prediction. (#6789) 
* Remove use of deprecated ntree_limit.
* Add sklearn demo.
 2021-03-27 03:09:25 +08:00
Jiaming Yuan a7083d3c13 Fix dart inplace prediction with GPU input. (#6777) 
* Fix dart inplace predict with data on GPU, which might trigger a fatal check
for device access right.
* Avoid copying data whenever possible.
 2021-03-25 12:00:32 +08:00
Jiaming YuanandPhilip Hyunsu Cho 1d90577800 Verify strictly positive labels for gamma regression. (#6778) 
Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>
 2021-03-25 11:46:52 +08:00
Jiaming Yuan 794fd6a46b Support v3 cuda array interface. (#6776) 2021-03-25 09:58:09 +08:00
Jiaming Yuan bcc0277338 Re-implement ROC-AUC. (#6747) 
* Re-implement ROC-AUC.

* Binary
* MultiClass
* LTR
* Add documents.

This PR resolves a few issues:
  - Define a value when the dataset is invalid, which can happen if there's an
  empty dataset, or when the dataset contains only positive or negative values.
  - Define ROC-AUC for multi-class classification.
  - Define weighted average value for distributed setting.
  - A correct implementation for learning to rank task.  Previous
  implementation is just binary classification with averaging across groups,
  which doesn't measure ordered learning to rank.
 2021-03-20 16:52:40 +08:00
Jiaming Yuan 4ee8340e79 Support column major array. (#6765) 2021-03-20 05:19:46 +08:00
Jiaming Yuan f6fe15d11f Improve parameter validation (#6769) 
* Add quotes to unused parameters.
* Check for whitespace.
 2021-03-20 01:56:55 +08:00
Jiaming Yuan 23b4165a6b Fix gamma deviance (#6761) 2021-03-20 01:56:17 +08:00
Jiaming Yuan e2d8a99413 Add document for tests directory. [skip ci] (#6760) 2021-03-18 15:15:50 +08:00
Jiaming Yuan 4f75f514ce Fix GPU RF (#6755) 
* Fix sampling.
 2021-03-17 06:23:35 +08:00
Jiaming Yuan 1a73a28511 Add device argsort. (#6749) 
This is part of https://github.com/dmlc/xgboost/pull/6747 .
 2021-03-16 16:05:22 +08:00
Jiaming Yuan 325bc93e16 [dask] Use distributed.MultiLock (#6743) 
* [dask] Use `distributed.MultiLock`

This enables training multiple models in parallel.

* Conditionally import `MultiLock`.
* Use async train directly in scikit learn interface.
* Use `worker_client` when available.
 2021-03-16 14:19:41 +08:00
Jiaming Yuan f20074e826 Check for invalid data. (#6742) 2021-03-04 14:37:20 +08:00
Jiaming Yuan a9b4a95225 Fix learning rate scheduler with cv. (#6720) 
* Expose more methods in cvpack and packed booster.
* Fix cv context in deprecated callbacks.
* Fix document.
 2021-02-28 13:57:42 +08:00
Jiaming Yuan 9da2287ab8 [breaking] Save booster feature info in JSON, remove feature name generation. (#6605) 
* Save feature info in booster in JSON model.
* [breaking] Remove automatic feature name generation in `DMatrix`.

This PR is to enable reliable feature validation in Python package.
 2021-02-25 18:54:16 +08:00
Jiaming Yuan c375173dca Support pylint 2.7.0 (#6726) 2021-02-25 12:49:58 +08:00
Jiaming Yuan 872e559b91 Use inplace predict for sklearn. (#6718) 
* Use inplace predict for sklearn when possible.
 2021-02-22 12:27:04 +08:00
Jiaming Yuan bdedaab8d1 Fix pylint. (#6714) 2021-02-19 11:53:27 +08:00
Jiaming Yuan e8c5c53e2f Use Predictor for dart. (#6693) 
* Use normal predictor for dart booster.
* Implement `inplace_predict` for dart.
* Enable `dart` for dask interface now that it's thread-safe.
* categorical data should be working out of box for dart now.

The implementation is not very efficient as it has to pull back the data and
apply weight for each tree, but still a significant improvement over previous
implementation as now we no longer binary search for each sample.

* Fix output prediction shape on dataframe.
 2021-02-09 23:30:19 +08:00
Jiaming Yuan dbf7e9d3cb Remove R cache in github action. (#6695) 
The cache stores outdated packages with wrong linkage.  Right now there's no way to clear the cache.
 2021-02-09 18:53:20 +08:00
Jiaming Yuan 1335db6113 [dask] Improve documents. (#6687) 
* Add tag for versions.
* use autoclass in sphinx build.
Made some class methods to be private to avoid exporting documents.
 2021-02-09 09:20:58 +08:00
Jiaming Yuan 5d48d40d9a Fix DMatrix slice with feature types. (#6689) 2021-02-09 08:13:51 +08:00
Jiaming Yuan 218a5fb6dd Simplify Span checks. (#6685) 
* Stop printing out message.
* Remove R specialization.

The printed message is not really useful anyway, without a reproducible example
there's no way to fix it.  But if there's a reproducible example, we can always
obtain these information by a debugger.  Removing the `printf` function avoids
creating the context in kernel.
 2021-02-09 08:12:58 +08:00
Jiaming Yuan 4656b09d5d [breaking] Add prediction fucntion for DMatrix and use inplace predict for dask. (#6668) 
* Add a new API function for predicting on `DMatrix`.  This function aligns
with rest of the `XGBoosterPredictFrom*` functions on semantic of function
arguments.
* Purge `ntree_limit` from libxgboost, use iteration instead.
* [dask] Use `inplace_predict` by default for dask sklearn models.
* [dask] Run prediction shape inference on worker instead of client.

The breaking change is in the Python sklearn `apply` function, I made it to be
consistent with other prediction functions where `best_iteration` is used by
default.
 2021-02-08 18:26:32 +08:00
Jiaming Yuan dbb5208a0a Use __array_interface__ for creating DMatrix from CSR. (#6675) 
* Use __array_interface__ for creating DMatrix from CSR.
* Add configuration.
 2021-02-05 21:09:47 +08:00
Jiaming Yuan 1e949110da Use generic dispatching routine for array interface. (#6672) 2021-02-05 09:23:38 +08:00
Jiaming Yuan a4101de678 Fix divide by 0 in feature importance when no split is found. (#6676) 2021-02-05 03:39:30 +08:00
Jiaming Yuan 72892cc80d [dask] Disable gblinear and dart. (#6665) 2021-02-04 09:13:09 +08:00
Jiaming Yuan 9d62b14591 Fix document. [skip ci] (#6669) 2021-02-02 20:43:31 +08:00
Jiaming Yuan 411592a347 Enhance inplace prediction. (#6653) 
* Accept array interface for csr and array.
* Accept an optional proxy dmatrix for metainfo.

This constructs an explicit `_ProxyDMatrix` type in Python.

* Remove unused doc.
* Add strict output.
 2021-02-02 11:41:46 +08:00