Commit Graph
6942 Commits
Author SHA1 Message Date
Jiaming Yuan 8fe1a2213c Cleanup code for distributed training. (#9805)
* Cleanup code for distributed training.

- Merge `GetNcclResult` into nccl stub.
- Split up utilities from the main dask module.
- Let Channel return `Result` to accommodate nccl channel.
- Remove old `use_label_encoder` parameter.
2023-11-25 09:10:56 +08:00
Jiaming Yuan e9260de3f3 [breaking] Remove dense libsvm parser plugin. (#9799) 2023-11-23 00:12:39 +08:00
Jiaming Yuan 1877cb8e83 Change default metric for gamma regression to deviance. (#9757)
* Change default metric for gamma regression to deviance.

- Cleanup the gamma implementation.
- Use deviance instead since the objective is derived from deviance.
2023-11-22 21:17:48 +08:00
Jiaming Yuan 0715ab3c10 Use dlopen to load NCCL. (#9796)
This PR adds optional support for loading nccl with `dlopen` as an alternative of compile time linking. This is to address the size bloat issue with the PyPI binary release.
- Add CMake option to load `nccl` at runtime.
- Add an NCCL stub.

After this, `nccl` will be fetched from PyPI when using pip to install XGBoost, either by a user or by `pyproject.toml`. Others who want to link the nccl at compile time can continue to do so without any change.

At the moment, this is Linux only since we only support MNMG on Linux.
2023-11-22 19:27:31 +08:00
Jiaming Yuan fedd9674c8 Implement column sampler in CUDA. (#9785)
- CUDA implementation.
- Extract the broadcasting logic, we will need the context parameter after revamping the collective implementation.
- Some changes to the event loop for fixing a deadlock in CI.
- Move argsort into algorithms.cuh, add support for cuda stream.
2023-11-17 04:29:08 +08:00
Bobby Wang 178cfe70a8 [pyspark][doc] Test and doc for stage-level scheduling. (#9786) 2023-11-16 18:15:59 +08:00
Jiaming Yuan ada377c57e [coll] Reduce the scope of lock in the event loop. (#9784) 2023-11-15 14:16:19 +08:00
Bobby Wang 36a552ac98 [jvm-packages] support stage-level scheduling (#9775) 2023-11-14 08:59:45 +08:00
Ken Geis 162da7b52b fix typo in Parameters doc (#9781) 2023-11-13 03:09:06 +08:00
Jiaming Yuan 6fd4a30667 [coll] Increase timeout for allgather test. (#9777) 2023-11-09 05:26:40 +08:00
Jiaming Yuan 44099f585d [coll] Add C API for the tracker. (#9773) 2023-11-08 18:17:14 +08:00
Jiaming Yuan 06bdc15e9b [coll] Pass context to various functions. (#9772)
* [coll] Pass context to various functions.

In the future, the `Context` object would be required for collective operations, this PR
passes the context object to some required functions to prepare for swapping out the
implementation.
2023-11-08 09:54:05 +08:00
Jiaming Yuan 6c0a190f6d [coll] Add comm group. (#9759)
- Implement `CommGroup` for double dispatching.
- Small cleanup to tracker for handling abort.
2023-11-07 11:12:31 +08:00
Jiaming Yuan c3a0622b49 Fix using categorical data with the score function of ranker. (#9753) 2023-11-07 07:29:11 +08:00
Jiaming Yuan 82828621d0 [doc] Add doc for linters and simplify c++ lint script. (#9750) 2023-11-07 05:03:30 +08:00
Jiaming Yuan 98238d63fa [dask] Change document to avoid using default import. (#9742)
This aligns dask with pyspark, users need to explicitly call:

```
from xgboost.dask import DaskXGBClassifier
from xgboost import dask as dxgb
```

In future releases, we might stop using the default import and remove the lazy loader.
2023-11-07 02:44:39 +08:00
Bobby WangandJiaming Yuan 093b675838 [Doc] update the tutorial of xgboost4j-spark-gpu (#9752)
---------

Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
2023-11-03 18:19:28 +08:00
david-cortes be20df8c23 [Python] Accept numpy generators as random_state (#9743)
* accept numpy generators for random_state

* make linter happy

* fix tests
2023-11-01 16:20:44 -07:00
Jiaming Yuan 4da4e092b5 [coll] Improvements and fixes for tracker and allreduce. (#9745)
- Allow the tracker to wait.
- Fix allreduce type cast
- Return args from the federated tracker.
2023-11-02 04:06:46 +08:00
Philip Hyunsu Cho 0ff8572737 [CI] Build libxgboost4j.dylib with CMAKE_OSX_DEPLOYMENT_TARGET (#9749) 2023-11-01 11:20:28 -07:00
Philip Hyunsu Cho 1b9ed4a4a1 [CI] Improve CI for Mac M1 (#9748)
* [CI] Improve CI for Mac M1

* Add -v flag

* Disable OpenMP in libxgboost4j.dylib

* Target MacOS 10.15+ to use C++17
2023-11-01 10:03:56 -07:00
david-cortes d3f0646779 [R] Avoid modifying importance dt in-place, fix aggregation (#9740) 2023-11-01 05:10:59 +08:00
Jiaming Yuan bc995a4865 [coll] Add federated coll. (#9738)
- Define a new data type, the proto file is copied for now.
- Merge client and communicator into `FederatedColl`.
- Define CUDA variant.
- Migrate tests for CPU, add tests for CUDA.
2023-11-01 04:06:46 +08:00
Philip Hyunsu Cho 6b98305db4 [CI] Enable gmock in gtest (#9737) 2023-10-31 20:09:35 +08:00
Jiaming Yuan 80390e6cb6 [coll] Federated comm. (#9732) 2023-10-31 02:39:55 +08:00
Bobby WangandJiaming Yuan fa65cf6646 [doc] How to configure regarding to stage-level (#9727)
---------

Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
2023-10-31 01:28:34 +08:00
omahs 2cfc90e8db Fix typos (#9731) 2023-10-30 16:52:12 +08:00
Jiaming Yuan 6755179e77 [coll] Add nccl. (#9726) 2023-10-28 16:33:58 +08:00
James Lamb 0c621094b3 [CI] enforce cmakelint checks (#9728) 2023-10-28 05:38:04 +08:00
Dmitry RazdoburdinandDmitry Razdoburdin <> 9c22df9342 Fix mingw hanging on regex in context (#9729)
---------

Co-authored-by: Dmitry Razdoburdin <>
2023-10-27 20:01:35 +08:00
Bobby Wang 1323531323 [pyspark] unify the way for determining whether runs on the GPU. (#9724) 2023-10-27 11:21:30 +08:00
Dmitry RazdoburdinandDmitry Razdoburdin <> f41a08fda8 Add 'sycl' devices to the context (#9691)
Co-authored-by: Dmitry Razdoburdin <>
2023-10-26 22:17:56 +08:00
Philip Hyunsu Cho d4d7097acc Update JVM script (#9714) 2023-10-24 17:48:38 -07:00
Philip Hyunsu Cho 01d59ded00 Fix libpath logic for Windows (#9712)
* Fix libpath logic for Windows (#9687)

* Use sys.base_prefix instead of sys.prefix (#9711)

* Use sys.base_prefix instead of sys.prefix

* Update libpath.py too
2023-10-24 17:25:28 -07:00
Jiaming Yuan 7a02facc9d Serialize expand entry for allgather. (#9702) 2023-10-24 14:33:28 +08:00
Hyunsu Cho ee8b29c843 [CI] Hotfix for JVM test on GH Action 2023-10-23 20:02:33 -07:00
Philip Hyunsu Cho 87621322ed [CI] Build libxgboost4j.dylib for Intel Mac (#9704)
* [CI] Build libxgboost4j.dylib for Intel Mac

* Use correct runner name

* Fix shell command

* Add back branch condition
2023-10-23 19:15:24 -07:00
Jiaming Yuan 3ca06ac51e [doc] Mention data consistency for categorical features. (#9678) 2023-10-24 10:11:33 +08:00
Philip Hyunsu Cho 5e6cb63a56 [CI] Set up CI for Mac M1 (#9699) 2023-10-22 23:33:19 -07:00
Philip Hyunsu Cho 791de7789b [jvm-packages] Remove hard dependency on libjvm (#9698) 2023-10-21 23:14:38 -07:00
Jiaming Yuan b771f58453 [coll] Define interface for bridging. (#9695)
* Define the basic interface that will shared by nccl, federated and native.
2023-10-20 16:20:48 +08:00
Rong Ou 6fbe6248f4 More in-memory input support for column split (#9685) 2023-10-20 16:02:36 +08:00
Chuck Atkins 83cdf14b2c CMake LTO and CUDA arch (#9677) 2023-10-20 13:01:37 +08:00
Philip Hyunsu Cho 3b86260b50 Fix build for AppleClang 11 (#9684) (#9693) 2023-10-18 12:27:21 -07:00
Jiaming Yuan 5d1bcde719 [coll] allgatherv. (#9688) 2023-10-19 03:13:50 +08:00
Dmitry Razdoburdin ea9f09716b Reorder if-else statements to allow using of cpu branches for sycl-devices (#9682) 2023-10-18 10:55:33 +08:00
Jiaming Yuan 4c0e4422d0 [coll] allgather. (#9681) 2023-10-18 10:22:18 +08:00
Jiaming Yuan 48ac9b6cbe [coll] Allreduce. (#9679) 2023-10-17 13:57:14 +08:00
Rong OuandJiaming Yuan da6803b75b Support column-wise data split with in-memory inputs (#9628)
---------

Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com>
2023-10-17 12:16:39 +08:00
Bobby Wang 4d1607eefd [pyspark] Support stage-level scheduling for training (#9519) 2023-10-17 10:35:39 +08:00