GitLab

* new cuda framework

* add histogram construction kernel

* before removing multi-gpu

* new cuda framework

* tree learner cuda kernels

* single tree framework ready

* single tree training framework

* remove comments

* boosting with cuda

* optimize for best split find

* data split

* move boosting into cuda

* parallel synchronize best split point

* merge split data kernels

* before code refactor

* use tasks instead of features as units for split finding

* refactor cuda best split finder

* fix configuration error with small leaves in data split

* skip histogram construction of too small leaf

* skip split finding of invalid leaves

stop when no leaf to split

* support row wise with CUDA

* copy data for split by column

* copy data from host to CPU by column for data partition

* add synchronize best splits for one leaf from multiple blocks

* partition dense row data

* fix sync best split from task blocks

* add support for sparse row wise for CUDA

* remove useless code

* add l2 regression objective

* sparse multi value bin enabled for CUDA

* fix cuda ranking objective

* support for number of items <= 2048 per query

* speedup histogram construction by interleaving global memory access

* split optimization

* add cuda tree predictor

* remove comma

* refactor objective and score updater

* before use struct

* use structure for split information

* use structure for leaf splits

* return CUDASplitInfo directly after finding best split

* split with CUDATree directly

* use cuda row data in cuda histogram constructor

* clean src/treelearner/cuda

* gather shared cuda device functions

* put shared CUDA functions into header file

* change smaller leaf from <= back to < for consistent result with CPU

* add tree predictor

* remove useless cuda_tree_predictor

* predict on CUDA with pipeline

* add global sort algorithms

* add global argsort for queries with many items in ranking tasks

* remove limitation of maximum number of items per query in ranking

* add cuda metrics

* fix CUDA AUC

* remove debug code

* add regression metrics

* remove useless file

* don't use mask in shuffle reduce

* add more regression objectives

* fix cuda mape loss

add cuda xentropy loss

* use template for different versions of BitonicArgSortDevice

* add multiclass metrics

* add ndcg metric

* fix cross entropy objectives and metrics

* fix cross entropy and ndcg metrics

* add support for customized objective in CUDA

* complete multiclass ova for CUDA

* separate cuda tree learner

* use shuffle based prefix sum

* clean up cuda_algorithms.hpp

* add copy subset on CUDA

* add bagging for CUDA

* clean up code

* copy gradients from host to device

* support bagging without using subset

* add support of bagging with subset for CUDAColumnData

* add support of bagging with subset for dense CUDARowData

* refactor copy sparse subrow

* use copy subset for column subset

* add reset train data and reset config for CUDA tree learner

add deconstructors for cuda tree learner

* add USE_CUDA ifdef to cuda tree learner files

* check that dataset doesn't contain CUDA tree learner

* remove printf debug information

* use full new cuda tree learner only when using single GPU

* disable all CUDA code when using CPU version

* recover main.cpp

* add cpp files for multi value bins

* update LightGBM.vcxproj

* update LightGBM.vcxproj

fix lint errors

* fix lint errors

* fix lint errors

* update Makevars

fix lint errors

* fix the case with 0 feature and 0 bin

fix split finding for invalid leaves

create cuda column data when loaded from bin file

* fix lint errors

hide GetRowWiseData when cuda is not used

* recover default device type to cpu

* fix na_as_missing case

fix cuda feature meta information

* fix UpdateDataIndexToLeafIndexKernel

* create CUDA trees when needed in CUDADataPartition::UpdateTrainScore

* add refit by tree for cuda tree learner

* fix test_refit in test_engine.py

* create set of large bin partitions in CUDARowData

* add histogram construction for columns with a large number of bins

* add find best split for categorical features on CUDA

* add bitvectors for categorical split

* cuda data partition split for categorical features

* fix split tree with categorical feature

* fix categorical feature splits

* refactor cuda_data_partition.cu with multi-level templates

* refactor CUDABestSplitFinder by grouping task information into struct

* pre-allocate space for vector split_find_tasks_ in CUDABestSplitFinder

* fix misuse of reference

* remove useless changes

* add support for path smoothing

* virtual destructor for LightGBM::Tree

* fix overlapped cat threshold in best split infos

* reset histogram pointers in data partition and spllit finder in ResetConfig

* comment useless parameter

* fix reverse case when na is missing and default bin is zero

* fix mfb_is_na and mfb_is_zero and is_single_feature_column

* remove debug log

* fix cat_l2 when one-hot

fix gradient copy when data subset is used

* switch shared histogram size according to CUDA version

* gpu_use_dp=true when cuda test

* revert modification in config.h

* fix setting of gpu_use_dp=true in .ci/test.sh

* fix linter errors

* fix linter error

remove useless change

* recover main.cpp

* separate cuda_exp and cuda

* fix ci bash scripts

add description for cuda_exp

* add USE_CUDA_EXP flag

* switch off USE_CUDA_EXP

* revert changes in python-packages

* more careful separation for USE_CUDA_EXP

* fix CUDARowData::DivideCUDAFeatureGroups

fix set fields for cuda metadata

* revert config.h

* fix test settings for cuda experimental version

* skip some tests due to unsupported features or differences in implementation details for CUDA Experimental version

* fix lint issue by adding a blank line

* fix lint errors by resorting imports

* fix lint errors by resorting imports

* fix lint errors by resorting imports

* merge cuda.yml and cuda_exp.yml

* update python version in cuda.yml

* remove cuda_exp.yml

* remove unrelated changes

* fix compilation warnings

fix cuda exp ci task name

* recover task

* use multi-level template in histogram construction

check split only in debug mode

* ignore NVCC related lines in parameter_generator.py

* update job name for CUDA tests

* apply review suggestions

* Update .github/workflows/cuda.yml
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update .github/workflows/cuda.yml
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* update header

* remove useless TODOs

* remove [TODO(shiyu1994): constrain the split with min_data_in_group] and record in #5062

* #include <LightGBM/utils/log.h> for USE_CUDA_EXP only

* fix include order

* fix include order

* remove extra space

* address review comments

* add warning when cuda_exp is used together with deterministic

* add comment about gpu_use_dp in .ci/test.sh

* revert changing order of included headers
Co-authored-by: Yu Shi <shiyu1994@qq.com>
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* fix dataset binary file alignment

* many fixes

* fix warnings

* fix bug

* Update file_io.cpp

* Update file_io.cpp

* simplify code

* Apply suggestions from code review

* general

* remove unneeded alignment

* Update file_io.h

* int32 to byte8 alignment

* Apply suggestions from code review

* Apply suggestions from code review

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* Initial CUDA work

* redirect log to python console (#3090)

* redir log to python console

* fix pylint

* Apply suggestions from code review

* Update basic.py

* Apply suggestions from code review
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* Update c_api.h

* Apply suggestions from code review

* Apply suggestions from code review

* super-minor: better wording
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>
Co-authored-by: StrikerRUS <nekit94-12@hotmail.com>

* re-order includes (fixes #3132) (#3133)

* Revert "re-order includes (fixes #3132) (#3133)" (#3153)

This reverts commit 656d2676

.

* Missing change from previous rebase

* Minor cleanup and removal of development scripts.

* Only set gpu_use_dp on by default for CUDA. Other minor change.

* Fix python lint indentation problem.

* More python lint issues.

* Big lint cleanup - more to come.

* Another large lint cleanup - more to come.

* Even more lint cleanup.

* Minor cleanup so less differences in code.

* Revert is_use_subset changes

* Another rebase from master to fix recent conflicts.

* More lint.

* Simple code cleanup - add & remove blank lines, revert unneccessary format changes, remove added dead code.

* Removed parameters added for CUDA and various bug fix.

* Yet more lint and unneccessary changes.

* Revert another change.

* Removal of unneccessary code.

* temporary appveyor.yml for building and testing

* Remove return value in ReSize

* Removal of unused variables.

* Code cleanup from reviewers suggestions.

* Removal of FIXME comments and unused defines.

* More reviewers comments cleanup.

* More reviewers comments cleanup.

* More reviewers comments cleanup.

* Fix config variables.

* Attempt to fix check-docs failure

* Update Paramster.rst for num_gpu

* Removing test appveyor.yml

* Add ƒCUDA_RESOLVE_DEVICE_SYMBOLS to libraries to fix linking issue.

* Fixed handling of data elements less than 2K.

* More reviewers comments cleanup.

* Removal of TODO and fix printing of int64_t

* Add cuda change for CI testing and remove cuda from device_type in python.

* Missed one change form previous check-in

* Removal AdditionConfig and fix settings.

* Limit number of GPUs to one for now in CUDA.

* Update Parameters.rst for previous check-in

* Whitespace removal.

* Cleanup unused code.

* Changed uint/ushort/ulong to unsigned int/short/long to help Windows based CUDA compiler work.

* Lint change from previous check-in.

* Changes based on reviewers comments.

* More reviewer comment changes.

* Adding warning for is_sparse. Revert tmp_subset code. Only return FeatureGroupData if not is_multi_val_

* Fix so that CUDA code will compile even if you enable the SCORE_T_USE_DOUBLE define.

* Reviewer comment cleanup.

* Replace warning with Log message. Removal of some of the USE_CUDA. Fix typo and removal of pragma once.

* Remove PRINT debug for CUDA code.

* Allow to use of multiple GPUs for CUDA.

* More multi-GPUs enablement for CUDA.

* More code cleanup based on reviews comments.

* Update docs with latest config changes.
Co-authored-by: Gordon Fossum <fossum@us.ibm.com>
Co-authored-by: ChipKerchner <ckerchne@linux.vnet.ibm.com>
Co-authored-by: Guolin Ke <guolin.ke@outlook.com>
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>
Co-authored-by: StrikerRUS <nekit94-12@hotmail.com>
Co-authored-by: James Lamb <jaylamb20@gmail.com>

* fix typo

* fix typo

* fix zero bin

* some fix

* fix bin mapping

* fix

* fix bug

* use stable sort

* fix cat forced split

* Apply suggestions from code review

* Apply suggestions from code review

* Apply suggestions from code review

This reverts commit 656d2676.

* commit

* fix msvc

* fix format

* speed up for const hessian

* rename template

* fix clang build

* template init

* add comment

* speed up multi-val bin subset for bagging

* remove the duplicated codes

* code refine

* some codes refactoring

* move `is_constant_hessian` into `TrainingShareStates`

* refine

* fix bug

* fix bug when num_groups_ < 0

* fix gpu

* fix gpu bagging

* fix gpu bug

* typo

* Update src/treelearner/serial_tree_learner.h

* fix bug when nan bin is most freq bin

* fix naming

* fix bug

* add test

* Apply suggestions from code review

* fix more bugs

* various minor style, docs and cpplint improvements

* fixed typo in warning

* fix recently added cpplint errors

* move note for params upper in description for consistency

* commit

* fix a bug

* fix bug

* reset to track changes

* refine the auto choose logic

* sort the time stats output

* fix include

* change  multi_val_bin_sparse_threshold

* add cmake

* add _mm_malloc and _mm_free for cross platform

* fix cmake bug

* timer for split

* try to fix cmake

* fix tests

* refactor DataPartition::Split

* fix test

* typo

* formating

* Revert "formating"

This reverts commit 5b8de4f7fb9d975ee23701d276a66d40ee6d4222.

* add document

* [R-package] Added tests on use of force_col_wise and force_row_wise in training (#2719)

* naming

* fix gpu code

* Update include/LightGBM/bin.h
Co-Authored-By: James Lamb <jaylamb20@gmail.com>

* Update src/treelearner/ocl/histogram16.cl

* test: swap compilers for CI

* fix omp

* not avx2

* no aligned for feature histogram

* Revert "refactor DataPartition::Split"

This reverts commit 256e6d9641ade966a1f54da1752e998a1149b6f8.

* slightly refactor data partition

* reduce the memory cost
Co-authored-by: James Lamb <jaylamb20@gmail.com>
Co-authored-by: Nikita Titov <nekit94-08@mail.ru>

* try to use _mm_prefetch anywhere

* refine

* fix bug

* remove the unneeded prefetch

* implement

* fix warning

* fix bug

* fix a bug

* remove unneed function

* fix data push bug

* fix valid data push

* fix bug for missing_type=zero

* refine split

* renames

* typo

* add prefetch for dense bin

* prefetch for ordered bin

* Update meta.h

* Update meta.h

* Update dense_bin.hpp

* fix many cpp lint errors

* indent

* fix bug

* fix more

* fix gpu

* more fixes

* added all necessary includes - fixed build/include_what_you_use error

* fixed the order of includes (build/include_order)

* Initial attempt to implement appending features in-memory to another data set

The intent is for this to enable munging files together easily, without needing to round-trip via numpy or write multiple copies to disk.
In turn, that enables working more efficiently with data sets that were written separately.

* Implement Dataset.dump_text, and fix small bug in appending of group bin boundaries.

Dumping to text enables us to compare results, without having to worry about issues like features being reordered.

* Add basic tests for validation logic for add_features_from.

* Remove various internal mapping items from dataset text dumps

These are too sensitive to the exact feature order chosen, which is not visible to the user.
Including them in tests appears unnecessary, as the data dumping code should provide enough coverage.

* Add test that add_features_from results in identical data sets according to dump_text.

* Add test that booster behaviour after using add_features_from matches that of training on the full data

This checks:
- That training after add_features_from works at all
- That add_features_from does not cause training to misbehave

* Expose feature_penalty and monotone_types/constraints via get_field

These getters allow us to check that add_features_from does the right thing with these vectors.

* Add tests that add_features correctly handles feature_penalty and monotone_constraints.

* Ensure add_features_from properly frees the added dataset and add unit test for this

Since add_features_from moves the feature group pointers from the added dataset to the dataset being added to, the added dataset is invalid after the call.
We must ensure we do not try and access this handle.

* Remove some obsolete TODOs

* Tidy up DumpTextFile by using a single iterator for each feature

This iterators were also passed around as raw pointers without being freed, which is now fixed.

* Factor out offsetting logic in AddFeaturesFrom

* Remove obsolete TODO

* Remove another TODO

This one is debatable, test code can be a bit messy and duplicate-heavy, factoring it out tends to end badly.
Leaving this for now, will revisit if adding more tests later on becomes a mess.

* Add documentation for newly-added methods.

* Fix whitespace issues identified by pylint.

* Fix a few more whitespace issues.

* Fix doc comments

* Implement deep copying for feature groups.

* Replace awkward std::move usage by emplace_back, and reduce vector size to num_features rather than num_total_features.

* Copy feature groups in addFeaturesFrom, rather than moving them.

* Fix bugs in FeatureGroup copy constructor and ensure source dataset remains usable

* Add reserve to PushVector and PushOffset

* Move definition of Clone into class body

* Fix PR review issues

* Fix for loop increment style.

* Fix test failure

* Some more docstring fixes.

* Remove blank line

* Read and write datsets from hdfs.
* Only enabled when cmake is run with -DUSE_HDFS:BOOL=TRUE
* Introduces VirtualFile(Reader|Writer) to asbtract VFS differences

* finish the data loading part

* allow prediction.

* fix bug for decision type.

* finish split finding part

* fix bugs.

* bug fixed. add a test .

* fix pep8 .

* update documents.

* fix test bugs.

* fix a format

* fix import error in python test.

* disable missing handle in categorial features.

* fix a bug.

* add more tests.

* fix pep8

* fix bugs.

* remove the missing handle code for categorical feature.

* Revert "python-package: support valid_names in scikit-learn API (#420)"

This reverts commit de39dbcf.

* Revert "faster histogram sum up (#418)"

This reverts commit 98c7c2a3.

* some refactor.

* two stage sum up to reduce sum up error.

* add more two-stage sumup.

* some refactor.

* add alignment.

* change name to aligned_allocator.

* remove some useless sumup.

* fix a warning.

* add -march=native .

* remove the padding of gradients.

* no alignment.

* fix test.

* change KNumSumupGroup to 32768.

* change gcc flags.

* add dummy gpu solver code

* initial GPU code

* fix crash bug

* first working version

* use asynchronous copy

* use a better kernel for root

* parallel read histogram

* sparse features now works, but no acceleration, compute on CPU

* compute sparse feature on CPU simultaneously

* fix big bug; add gpu selection; add kernel selection

* better debugging

* clean up

* add feature scatter

* Add sparse_threshold control

* fix a bug in feature scatter

* clean up debug

* temporarily add OpenCL kernels for k=64,256

* fix up CMakeList and definition USE_GPU

* add OpenCL kernels as string literals

* Add boost.compute as a submodule

* add boost dependency into CMakeList

* fix opencl pragma

* use pinned memory for histogram

* use pinned buffer for gradients and hessians

* better debugging message

* add double precision support on GPU

* fix boost version in CMakeList

* Add a README

* reconstruct GPU initialization code for ResetTrainingData

* move data to GPU in parallel

* fix a bug during feature copy

* update gpu kernels

* update gpu code

* initial port to LightGBM v2

* speedup GPU data loading process

* Add 4-bit bin support to GPU

* re-add sparse_threshold parameter

* remove kMaxNumWorkgroups and allows an unlimited number of features

* add feature mask support for skipping unused features

* enable kernel cache

* use GPU kernels withoug feature masks when all features are used

* REAdme.

* REAdme.

* update README

* fix typos (#349)

* change compile to gcc on Apple as default

* clean vscode related file

* refine api of constructing from sampling data.

* fix bug in the last commit.

* more efficient algorithm to sample k from n.

* fix bug in filter bin

* change to boost from average output.

* fix tests.

* only stop training when all classes are finshed in multi-class.

* limit the max tree output. change hessian in multi-class objective.

* robust tree model loading.

* fix test.

* convert the probabilities to raw score in boost_from_average of classification.

* fix the average label for binary classification.

* Add boost_from_average to docs (#354)

* don't use "ConvertToRawScore" for self-defined objective function.

* boost_from_average seems doesn't work well in binary classification. remove it.

* For a better jump link (#355)

* Update Python-API.md

* for a better jump in page

A space is needed between `#` and the headers content according to Github's markdown format [guideline](https://guides.github.com/features/mastering-markdown/)

After adding the spaces, we can jump to the exact position in page by click the link.

* fixed something mentioned by @wxchan

* Update Python-API.md

* add FitByExistingTree.

* adapt GPU tree learner for FitByExistingTree

* avoid NaN output.

* update boost.compute

* fix typos (#361)

* fix broken links (#359)

* update README

* disable GPU acceleration by default

* fix image url

* cleanup debug macro

* remove old README

* do not save sparse_threshold_ in FeatureGroup

* add details for new GPU settings

* ignore submodule when doing pep8 check

* allocate workspace for at least one thread during builing Feature4

* move sparse_threshold to class Dataset

* remove duplicated code in GPUTreeLearner::Split

* Remove duplicated code in FindBestThresholds and BeforeFindBestSplit

* do not rebuild ordered gradients and hessians for sparse features

* support feature groups in GPUTreeLearner

* Initial parallel learners with GPU support

* add option device, cleanup code

* clean up FindBestThresholds; add some omp parallel

* constant hessian optimization for GPU

* Fix GPUTreeLearner crash when there is zero feature

* use np.testing.assert_almost_equal() to compare lists of floats in tests

* travis for GPU

* reduce the sumup cost of constant hessians.

* fix test.

* fix bug when have weights.

* fix a comment.

* reduce branching.