Hendrik/xgboost
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
5,393 Commits 32 Branches 59 Tags
Commit Graph

1114 Commits

Author SHA1 Message Date
Jiaming Yuan
77f6cf2d13
Support hessian in host sketch container. (#7081) 
Prepare for migrating approx onto hist's codebase.
 2021-07-08 16:33:58 +08:00
Jiaming Yuan
84d359efb8
Support host data in proxy DMatrix. (#7087) 2021-07-08 11:35:48 +08:00
Jiaming Yuan
5d7cdf2e36
[Breaking] Rename Quantile DMatrix C API. (#7082) 
The role of ProxyDMatrix is going beyond what it was designed.  Now it's used by both
QuantileDeviceDMatrix and inplace prediction.  After the refactoring of sparse DMatrix it
will also be used for external memory.  Renaming the C API to extract it from
QuantileDeviceDMatrix.
 2021-07-08 11:34:14 +08:00
Jiaming Yuan
c766f143ab
Refactor external memory formats. (#7089) 
* Save base_rowid.
* Return write size.
* Remove unused function.
 2021-07-08 04:04:51 +08:00
Jiaming Yuan
689eb8f620
Check external memory support for exact tree method. (#7088) 2021-07-08 02:12:57 +08:00
Jiaming Yuan
615ab2b03e
Extract evaluate splits from CPU hist. (#7079) 
Other than modularizing the split evaluation function, this PR also removes some more functions including `InitNewNodes` and `BuildNodeStats` among some other unused variables.  Also, scattered code like setting leaf weights is grouped into the split evaluator and `NodeEntry` is simplified and made private.  Another subtle difference with the original implementation is that the modified code doesn't call `tree[nidx].Parent()` to traversal upward.
 2021-07-07 15:16:25 +08:00
Jiaming Yuan
116d711815
Make SimpleDMatrix ctor reusable. (#7075) 2021-07-06 13:38:24 +08:00
Jiaming Yuan
d7e1fa7664
Fix feature names and types in output model slice. (#7078) 2021-07-06 11:47:49 +08:00
Jiaming Yuan
1cd20efe68
Move GHistIndex into DMatrix. (#7064) 2021-07-01 00:44:49 +08:00
Jiaming Yuan
1c8fdf2218
Remove use of device_idx in dh::LaunchN. (#7063) 
It's an unused parameter, removing it can make the CI log more readable.
 2021-06-29 11:37:26 +08:00
Jiaming Yuan
8fa32fdda2
Implement categorical data support for SHAP. (#7053) 
* Add CPU implementation.
* Update GPUTreeSHAP.
* Add GPU implementation by defining custom split condition.
 2021-06-25 19:02:46 +08:00
Jiaming Yuan
663136aa08
Implement feature score for linear model. (#7048) 
* Add feature score support for linear model.
* Port R interface to the new implementation.
* Add linear model support in Python.

Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu>
 2021-06-25 14:34:02 +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 Yuan
5c2d7a18c9
Parallel model dump for trees. (#7040) 2021-06-15 14:08:26 +08:00
ShvetsKS
2567404ab6
Simplify sparse and dense CPU hist kernels (#7029) 
* Simplify sparse and dense kernels
* Extract row partitioner.

Co-authored-by: Kirill Shvets <kirill.shvets@intel.com>
 2021-06-11 18:26:30 +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 Yuan
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
TP Boudreau
bd2ca543c4
Fix BinarySearchBin() argument types (#7026) 2021-06-08 19:05:46 +08:00
ShvetsKS
5cdaac00c1
Remove feature grouping (#7018) 
Co-authored-by: Kirill Shvets <kirill.shvets@intel.com>
 2021-06-03 04:35:26 +08:00
ShvetsKS
57c732655e
Merge lossgude and depthwise strategies for CPU hist (#7007) 
* fix java/scala test: max depth is also valid parameter for lossguide

Co-authored-by: Kirill Shvets <kirill.shvets@intel.com>
 2021-06-03 01:49:43 +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
ShvetsKS
55b823b27d
Reduce 'InitSampling' complexity and set gradients to zero (#6922) 
Co-authored-by: Kirill Shvets <kirill.shvets@intel.com>
 2021-05-29 04:52:23 +08:00
Jiaming Yuan
4cf95a6041
Support numpy array interface (#6998) 2021-05-27 16:08:22 +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
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
Livius
a4886c404a
Fix compilation error on x86 (#6964) 
Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
 2021-05-14 13:31:49 +08:00
Jiaming Yuan
44cc9c04ea
Fix multiclass auc with empty dataset. (#6947) 2021-05-12 15:01:14 +08:00
Andrew Ziem
3e7e426b36
Fix spelling in documents (#6948) 
* Update roxygen2 doc.

Co-authored-by: fis <jm.yuan@outlook.com>
 2021-05-11 20:44:36 +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
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
a2ecbdaa31
Add an API guard to prevent global variables being changed. (#6891) 2021-04-23 10:27:57 +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
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
b1fdb220f4
Remove deprecated n_gpus parameter. (#6821) 2021-04-02 03:02:32 +08:00
Jiaming Yuan
905fdd3e08
Fix typos in AUC. (#6795) 2021-03-31 16:35:42 +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
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
ShvetsKS
8825670c9c
Memory consumption fix for row-major adapters (#6779) 
Co-authored-by: Kirill Shvets <kirill.shvets@intel.com>
Co-authored-by: fis <jm.yuan@outlook.com>
 2021-03-26 08:44:30 +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