Code

[CUDA] New CUDA version Part 1#4630

Code

Merged

shiyu1994 merged 176 commits intomicrosoft:mastermicrosoft/LightGBM:masterfrom shiyu1994:cuda-tree-learner-subsetshiyu1994/LightGBM:cuda-tree-learner-subsetCopy head branch name to clipboardMar 23, 2022

+10,276-59

Commits

Commits on Apr 20, 2021

• new cuda framework

  shiyu1994

  committed
  94aed50

  Copy full SHA for 94aed50

Commits on Apr 22, 2021

• add histogram construction kernel

  

  Yu Shi

  committed
  18df6b2

  Copy full SHA for 18df6b2

Commits on Apr 29, 2021

• before removing multi-gpu

  

  Yu Shi

  committed
  9b21d2b

  Copy full SHA for 9b21d2b

• new cuda framework

  

  Yu Shi

  committed
  634a4f1

  Copy full SHA for 634a4f1

Commits on May 6, 2021

• tree learner cuda kernels

  

  Yu Shi

  committed
  23bcaa2

  Copy full SHA for 23bcaa2

Commits on May 7, 2021

• single tree framework ready

  

  Yu Shi

  committed
  6c14cd9

  Copy full SHA for 6c14cd9

Commits on May 9, 2021

• single tree training framework

  

  Yu Shi

  committed
  aa0b3de

  Copy full SHA for aa0b3de

• remove comments

  

  Yu Shi

  committed
  bc85ced

  Copy full SHA for bc85ced

Commits on May 10, 2021

• boosting with cuda

  

  Yu Shi

  committed
  18d957a

  Copy full SHA for 18d957a

Commits on May 13, 2021

• optimize for best split find

  

  Yu Shi

  committed
  28186c0

  Copy full SHA for 28186c0

Commits on May 14, 2021

• data split

  

  Yu Shi

  committed
  60c7e4e

  Copy full SHA for 60c7e4e

Commits on May 17, 2021

• move boosting into cuda

  

  Yu Shi

  committed
  57547fb

  Copy full SHA for 57547fb

Commits on May 27, 2021

• parallel synchronize best split point

  

  Yu Shi

  committed
  608fd70

  Copy full SHA for 608fd70

Commits on Jun 1, 2021

• merge split data kernels

  

  Yu Shi

  committed
  277be8b

  Copy full SHA for 277be8b

• before code refactor

  

  Yu Shi

  committed
  ffcf765

  Copy full SHA for ffcf765

Commits on Jun 2, 2021

• use tasks instead of features as units for split finding

  

  Yu Shi

  committed
  a58c1e1

  Copy full SHA for a58c1e1

• refactor cuda best split finder

  

  Yu Shi

  committed
  72d41c9

  Copy full SHA for 72d41c9

• fix configuration error with small leaves in data split

  shiyu1994

  committed
  f7a7658

  Copy full SHA for f7a7658

• skip histogram construction of too small leaf

  shiyu1994

  committed
  b6efd10

  Copy full SHA for b6efd10

Commits on Jun 3, 2021

• skip split finding of invalid leaves

  Show description for 6f4e39d
  shiyu1994

  committed
  6f4e39d

  Copy full SHA for 6f4e39d

Commits on Jun 4, 2021

• support row wise with CUDA

  shiyu1994

  committed
  4072bb8

  Copy full SHA for 4072bb8

• copy data for split by column

  shiyu1994

  committed
  88ecde9

  Copy full SHA for 88ecde9

Commits on Jun 8, 2021

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

  shiyu1994

  committed
  dec7501

  Copy full SHA for dec7501

• add synchronize best splits for one leaf from multiple blocks

  shiyu1994

  committed
  2dccb7f

  Copy full SHA for 2dccb7f

Commits on Jun 9, 2021

• partition dense row data

  shiyu1994

  committed
  0168d2c

  Copy full SHA for 0168d2c

• fix sync best split from task blocks

  shiyu1994

  committed
  0570fe0

  Copy full SHA for 0570fe0

• add support for sparse row wise for CUDA

  shiyu1994

  committed
  374018c

  Copy full SHA for 374018c

• remove useless code

  shiyu1994

  committed
  40c49cc

  Copy full SHA for 40c49cc

Commits on Jun 10, 2021

• add l2 regression objective

  shiyu1994

  committed
  dc41a00

  Copy full SHA for dc41a00

Commits on Jun 11, 2021

• sparse multi value bin enabled for CUDA

  shiyu1994

  committed
  bd065b7

  Copy full SHA for bd065b7

Commits on Jun 16, 2021

• fix cuda ranking objective

  shiyu1994

  committed
  a5fadfb

  Copy full SHA for a5fadfb

• support for number of items <= 2048 per query

  

  Yu Shi

  committed
  3202b79

  Copy full SHA for 3202b79

Commits on Jun 23, 2021

• speedup histogram construction by interleaving global memory access

  

  Yu Shi

  committed
  cd687c9

  Copy full SHA for cd687c9

Commits on Jun 28, 2021

• split optimization

  

  Yu Shi

  committed
  320c449

  Copy full SHA for 320c449

Commits on Jul 2, 2021

• add cuda tree predictor

  

  Yu Shi

  committed
  eb1d7fa

  Copy full SHA for eb1d7fa

Commits on Jul 18, 2021

• remove comma

  

  Yu Shi

  committed
  dd177f5

  Copy full SHA for dd177f5

Commits on Jul 19, 2021

• refactor objective and score updater

  

  Yu Shi

  committed
  ee836d6

  Copy full SHA for ee836d6

Commits on Jul 21, 2021

• before use struct

  

  Yu Shi

  committed
  0467fce

  Copy full SHA for 0467fce

• use structure for split information

  

  Yu Shi

  committed
  f05da3c

  Copy full SHA for f05da3c

• use structure for leaf splits

  

  Yu Shi

  committed
  400622a

  Copy full SHA for 400622a

• return CUDASplitInfo directly after finding best split

  

  Yu Shi

  committed
  d9d3aa9

  Copy full SHA for d9d3aa9

Commits on Jul 22, 2021

• split with CUDATree directly

  

  Yu Shi

  committed
  45cf7a7

  Copy full SHA for 45cf7a7

Commits on Jul 26, 2021

• use cuda row data in cuda histogram constructor

  

  Yu Shi

  committed
  9dea18d

  Copy full SHA for 9dea18d

• clean src/treelearner/cuda

  

  Yu Shi

  committed
  572e2b0

  Copy full SHA for 572e2b0

Commits on Jul 27, 2021

• gather shared cuda device functions

  

  Yu Shi

  committed
  fe58d4c

  Copy full SHA for fe58d4c

• put shared CUDA functions into header file

  

  Yu Shi

  committed
  dc461dc

  Copy full SHA for dc461dc

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

  

  Yu Shi

  committed
  ba565c1

  Copy full SHA for ba565c1

Commits on Aug 3, 2021

• add tree predictor

  

  Yu Shi

  committed
  a781ef5

  Copy full SHA for a781ef5

• remove useless cuda_tree_predictor

  

  Yu Shi

  committed
  c8a6fab

  Copy full SHA for c8a6fab

Commits on Aug 4, 2021

• predict on CUDA with pipeline

  

  Yu Shi

  committed
  a7504dc

  Copy full SHA for a7504dc

Commits on Aug 9, 2021

• add global sort algorithms

  

  Yu Shi

  committed
  896d47b

  Copy full SHA for 896d47b

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

  

  Yu Shi

  committed
  fe6ed74

  Copy full SHA for fe6ed74

Commits on Aug 11, 2021

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

  

  Yu Shi

  committed
  7808455

  Copy full SHA for 7808455

Commits on Aug 16, 2021

• add cuda metrics

  

  Yu Shi

  committed
  7a0d218

  Copy full SHA for 7a0d218

Commits on Aug 18, 2021

• fix CUDA AUC

  

  Yu Shi

  committed
  ca42f3b

  Copy full SHA for ca42f3b

• remove debug code

  

  Yu Shi

  committed
  c681102

  Copy full SHA for c681102

Commits on Aug 19, 2021

• add regression metrics

  

  Yu Shi

  committed
  ea60566

  Copy full SHA for ea60566

• remove useless file

  

  Yu Shi

  committed
  5c84788

  Copy full SHA for 5c84788

• don't use mask in shuffle reduce

  

  Yu Shi

  committed
  c2c2407

  Copy full SHA for c2c2407

Commits on Sep 2, 2021

• add more regression objectives

  

  Yu Shi

  committed
  b43d367

  Copy full SHA for b43d367

Commits on Sep 3, 2021

• fix cuda mape loss

  Show description for 951aa37
  

  Yu Shi

  committed
  951aa37

  Copy full SHA for 951aa37

• use template for different versions of BitonicArgSortDevice

  

  Yu Shi

  committed
  b50ce5b

  Copy full SHA for b50ce5b

Commits on Sep 6, 2021

• add multiclass metrics

  

  Yu Shi

  committed
  f51fd70

  Copy full SHA for f51fd70

Commits on Sep 7, 2021

• add ndcg metric

  

  Yu Shi

  committed
  35c742d

  Copy full SHA for 35c742d

Commits on Sep 10, 2021

• fix cross entropy objectives and metrics

  

  Yu Shi

  committed
  510d878

  Copy full SHA for 510d878

• fix cross entropy and ndcg metrics

  

  Yu Shi

  committed
  95f4612

  Copy full SHA for 95f4612

• add support for customized objective in CUDA

  

  Yu Shi

  committed
  bb997d0

  Copy full SHA for bb997d0

• complete multiclass ova for CUDA

  

  Yu Shi

  committed
  17b78d1

  Copy full SHA for 17b78d1

Commits on Sep 13, 2021

• merge master

  

  Yu Shi

  committed
  72aa863

  Copy full SHA for 72aa863

• separate cuda tree learner

  

  Yu Shi

  committed
  8537b8c

  Copy full SHA for 8537b8c

• use shuffle based prefix sum

  

  Yu Shi

  committed
  8fb8562

  Copy full SHA for 8fb8562

• clean up cuda_algorithms.hpp

  

  Yu Shi

  committed
  883ed15

  Copy full SHA for 883ed15

Commits on Sep 15, 2021

• add copy subset on CUDA

  

  Yu Shi

  committed
  e7ffc3f

  Copy full SHA for e7ffc3f

• add bagging for CUDA

  

  Yu Shi

  committed
  d7c4bb4

  Copy full SHA for d7c4bb4

• clean up code

  

  Yu Shi

  committed
  d9bf3e5

  Copy full SHA for d9bf3e5

• copy gradients from host to device

  

  Yu Shi

  committed
  95fd61a

  Copy full SHA for 95fd61a

• support bagging without using subset

  

  Yu Shi

  committed
  285c2d6

  Copy full SHA for 285c2d6

Commits on Sep 17, 2021

• add support of bagging with subset for CUDAColumnData

  

  Yu Shi

  committed
  1a09c19

  Copy full SHA for 1a09c19

Commits on Sep 18, 2021

• add support of bagging with subset for dense CUDARowData

  

  Yu Shi

  committed
  740f853

  Copy full SHA for 740f853

Commits on Sep 24, 2021

• refactor copy sparse subrow

  

  Yu Shi

  committed
  f42e87e

  Copy full SHA for f42e87e

Commits on Sep 26, 2021

• use copy subset for column subset

  

  Yu Shi

  committed
  0b9ca24

  Copy full SHA for 0b9ca24

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

  Show description for 9a94240
  

  Yu Shi

  committed
  9a94240

  Copy full SHA for 9a94240

• add USE_CUDA ifdef to cuda tree learner files

  

  Yu Shi

  committed
  1f6dd90

  Copy full SHA for 1f6dd90

• check that dataset doesn't contain CUDA tree learner

  

  Yu Shi

  committed
  4ca7586

  Copy full SHA for 4ca7586

• remove printf debug information

  

  Yu Shi

  committed
  25f57e3

  Copy full SHA for 25f57e3

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

  

  Yu Shi

  committed
  12794b0

  Copy full SHA for 12794b0

• Merge branch 'master' of https://github.com/microsoft/LightGBM into cuda-tree-learner-subset

  

  Yu Shi

  committed
  44e47ec

  Copy full SHA for 44e47ec

Commits on Sep 27, 2021

• disable all CUDA code when using CPU version

  

  Yu Shi

  committed
  7e18687

  Copy full SHA for 7e18687

• recover main.cpp

  

  Yu Shi

  committed
  469e992

  Copy full SHA for 469e992

Commits on Sep 30, 2021

• add cpp files for multi value bins

  

  Yu Shi

  committed
  f2812c8

  Copy full SHA for f2812c8

• update LightGBM.vcxproj

  

  Yu Shi

  committed
  8e884b2

  Copy full SHA for 8e884b2

• update LightGBM.vcxproj

  Show description for 9b9a63c
  

  Yu Shi

  committed
  9b9a63c

  Copy full SHA for 9b9a63c

• fix lint errors

  

  Yu Shi

  committed
  e0c9f6f

  Copy full SHA for e0c9f6f

• fix lint errors

  

  Yu Shi

  committed
  3bba6d7

  Copy full SHA for 3bba6d7

• update Makevars

  Show description for 8f9f03e
  

  Yu Shi

  committed
  8f9f03e

  Copy full SHA for 8f9f03e

Commits on Oct 8, 2021

• fix the case with 0 feature and 0 bin

  Show description for 01d772d
  

  Yu Shi

  committed
  01d772d

  Copy full SHA for 01d772d

• fix lint errors

  Show description for e57dd15
  

  Yu Shi

  committed
  e57dd15

  Copy full SHA for e57dd15

• recover default device type to cpu

  

  Yu Shi

  committed
  a5b9f7a

  Copy full SHA for a5b9f7a

Commits on Oct 9, 2021

• fix na_as_missing case

  Show description for 5f03d45
  

  Yu Shi

  committed
  5f03d45

  Copy full SHA for 5f03d45

Commits on Oct 15, 2021

• fix UpdateDataIndexToLeafIndexKernel

  shiyu1994

  committed
  b2aaa9f

  Copy full SHA for b2aaa9f

• create CUDA trees when needed in CUDADataPartition::UpdateTrainScore

  shiyu1994

  committed
  0726d87

  Copy full SHA for 0726d87

Commits on Oct 21, 2021

• add refit by tree for cuda tree learner

  shiyu1994

  committed
  1dea6bc

  Copy full SHA for 1dea6bc

• fix test_refit in test_engine.py

  

  Yu Shi

  committed
  14b9ce9

  Copy full SHA for 14b9ce9

• create set of large bin partitions in CUDARowData

  

  Yu Shi

  committed
  4b936de

  Copy full SHA for 4b936de

Commits on Oct 23, 2021

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

  shiyu1994

  committed
  4193768

  Copy full SHA for 4193768

Commits on Oct 26, 2021

• add find best split for categorical features on CUDA

  shiyu1994

  committed
  0b6e79e

  Copy full SHA for 0b6e79e

Commits on Oct 27, 2021

• add bitvectors for categorical split

  shiyu1994

  committed
  25f20a7

  Copy full SHA for 25f20a7

Commits on Oct 28, 2021

• cuda data partition split for categorical features

  shiyu1994

  committed
  82c33e4

  Copy full SHA for 82c33e4

Commits on Oct 29, 2021

• fix split tree with categorical feature

  shiyu1994

  committed
  ca16070

  Copy full SHA for ca16070

Commits on Nov 4, 2021

• fix categorical feature splits

  shiyu1994

  committed
  c8716f1

  Copy full SHA for c8716f1

Commits on Nov 5, 2021

• refactor cuda_data_partition.cu with multi-level templates

  shiyu1994

  committed
  4bcaa03

  Copy full SHA for 4bcaa03

• refactor CUDABestSplitFinder by grouping task information into struct

  shiyu1994

  committed
  536f603

  Copy full SHA for 536f603

Commits on Nov 8, 2021

• pre-allocate space for vector split_find_tasks_ in CUDABestSplitFinder

  shiyu1994

  committed
  015e099

  Copy full SHA for 015e099

• fix misuse of reference

  shiyu1994

  committed
  4c260d2

  Copy full SHA for 4c260d2

• remove useless changes

  shiyu1994

  committed
  89d8214

  Copy full SHA for 89d8214

Commits on Nov 9, 2021

• add support for path smoothing

  shiyu1994

  committed
  54bc66a

  Copy full SHA for 54bc66a

• virtual destructor for LightGBM::Tree

  shiyu1994

  committed
  86e208a

  Copy full SHA for 86e208a

Commits on Nov 12, 2021

• fix overlapped cat threshold in best split infos

  shiyu1994

  committed
  d888f1d

  Copy full SHA for d888f1d

Commits on Nov 17, 2021

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

  shiyu1994

  committed
  5efe0fb

  Copy full SHA for 5efe0fb

• merge with LightGBM/master

  shiyu1994

  committed
  559a569

  Copy full SHA for 559a569

• comment useless parameter

  shiyu1994

  committed
  0bb88fb

  Copy full SHA for 0bb88fb

Commits on Nov 18, 2021

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

  shiyu1994

  committed
  0678d9a

  Copy full SHA for 0678d9a

• fix mfb_is_na and mfb_is_zero and is_single_feature_column

  shiyu1994

  committed
  26130d9

  Copy full SHA for 26130d9

Commits on Nov 19, 2021

• remove debug log

  shiyu1994

  committed
  d49e92a

  Copy full SHA for d49e92a

Commits on Nov 23, 2021

• fix cat_l2 when one-hot

  Show description for 3214d68
  shiyu1994

  committed
  3214d68

  Copy full SHA for 3214d68

• merge master

  shiyu1994

  committed
  361d2b0

  Copy full SHA for 361d2b0

Commits on Dec 1, 2021

• switch shared histogram size according to CUDA version

  shiyu1994

  committed
  85ea408

  Copy full SHA for 85ea408

• gpu_use_dp=true when cuda test

  shiyu1994

  committed
  2af0f5d

  Copy full SHA for 2af0f5d

• revert modification in config.h

  shiyu1994

  committed
  d0a628f

  Copy full SHA for d0a628f

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

  shiyu1994

  committed
  e0018ea

  Copy full SHA for e0018ea

• fix linter errors

  shiyu1994

  committed
  e54b51a

  Copy full SHA for e54b51a

• fix linter error

  Show description for 541235f
  shiyu1994

  committed
  541235f

  Copy full SHA for 541235f

• recover main.cpp

  shiyu1994

  committed
  a2ead3c

  Copy full SHA for a2ead3c

Commits on Dec 8, 2021

• separate cuda_exp and cuda

  shiyu1994

  committed
  2a81af6

  Copy full SHA for 2a81af6

• fix ci bash scripts

  Show description for 9881075
  shiyu1994

  committed
  9881075

  Copy full SHA for 9881075

Commits on Dec 14, 2021

• add USE_CUDA_EXP flag

  shiyu1994

  committed
  52b1e88

  Copy full SHA for 52b1e88

• Merge branch 'master' into cuda-tree-learner-subset

  shiyu1994

  authored
  09054a1

  Copy full SHA for 09054a1

Commits on Dec 24, 2021

• switch off USE_CUDA_EXP

  shiyu1994

  committed
  c2a0be8

  Copy full SHA for c2a0be8

• Merge remote-tracking branch 'LightGBM/master' into cuda-tree-learner-subset

  shiyu1994

  committed
  0651cca

  Copy full SHA for 0651cca

• revert changes in python-packages

  shiyu1994

  committed
  fbc3760

  Copy full SHA for fbc3760

• more careful separation for USE_CUDA_EXP

  shiyu1994

  committed
  c58635b

  Copy full SHA for c58635b

Commits on Jan 3, 2022

• fix CUDARowData::DivideCUDAFeatureGroups

  Show description for 93d5950
  shiyu1994

  committed
  93d5950

  Copy full SHA for 93d5950

• revert config.h

  shiyu1994

  committed
  9f6aa8a

  Copy full SHA for 9f6aa8a

• fix test settings for cuda experimental version

  shiyu1994

  committed
  12d8161

  Copy full SHA for 12d8161

Commits on Jan 4, 2022

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

  shiyu1994

  committed
  354845e

  Copy full SHA for 354845e

• fix lint issue by adding a blank line

  shiyu1994

  committed
  cb49dd1

  Copy full SHA for cb49dd1

• fix lint errors by resorting imports

  shiyu1994

  committed
  2e2c696

  Copy full SHA for 2e2c696

• fix lint errors by resorting imports

  shiyu1994

  committed
  3433674

  Copy full SHA for 3433674

• Merge branch 'master' into cuda-tree-learner-subset

  shiyu1994

  authored
  0c94bdd

  Copy full SHA for 0c94bdd

• fix lint errors by resorting imports

  shiyu1994

  committed
  c72d555

  Copy full SHA for c72d555

Commits on Jan 5, 2022

• merge cuda.yml and cuda_exp.yml

  shiyu1994

  committed
  63a9dc1

  Copy full SHA for 63a9dc1

• update python version in cuda.yml

  shiyu1994

  committed
  31ac33b

  Copy full SHA for 31ac33b

• remove cuda_exp.yml

  shiyu1994

  committed
  5f1f38d

  Copy full SHA for 5f1f38d

Commits on Jan 6, 2022

• remove unrelated changes

  shiyu1994

  committed
  ba22deb

  Copy full SHA for ba22deb

Commits on Feb 16, 2022

• fix compilation warnings

  Show description for b008424
  shiyu1994

  committed
  b008424

  Copy full SHA for b008424

Commits on Feb 21, 2022

• resolve conflicts with master

  shiyu1994

  committed
  fad4b91

  Copy full SHA for fad4b91

• Merge branch 'cuda-tree-learner-subset' of https://github.com/shiyu1994/LightGBM into cuda-tree-learner-subset

  shiyu1994

  committed
  55a94b5

  Copy full SHA for 55a94b5

Commits on Feb 22, 2022

• recover task

  shiyu1994

  committed
  d77dd23

  Copy full SHA for d77dd23

• use multi-level template in histogram construction

  Show description for 6a9d530
  shiyu1994

  committed
  6a9d530

  Copy full SHA for 6a9d530

• ignore NVCC related lines in parameter_generator.py

  shiyu1994

  committed
  4adca58

  Copy full SHA for 4adca58

Commits on Feb 23, 2022

• Merge remote-tracking branch 'LightGBM/master' into cuda-tree-learner-subset

  shiyu1994

  committed
  8d99b2b

  Copy full SHA for 8d99b2b

• update job name for CUDA tests

  shiyu1994

  committed
  1e23342

  Copy full SHA for 1e23342

Commits on Mar 8, 2022

• apply review suggestions

  shiyu1994

  committed
  f44b881

  Copy full SHA for f44b881

Commits on Mar 9, 2022

• Update .github/workflows/cuda.yml

  Show description for d7b65c4
  

  shiyu1994

  and

  StrikerRUS

  authored
  d7b65c4

  Copy full SHA for d7b65c4

• Update .github/workflows/cuda.yml

  Show description for a6a51fd
  

  shiyu1994

  and

  StrikerRUS

  authored
  a6a51fd

  Copy full SHA for a6a51fd

• update header

  shiyu1994

  committed
  9135582

  Copy full SHA for 9135582

• remove useless TODOs

  shiyu1994

  committed
  cd101ae

  Copy full SHA for cd101ae

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

  shiyu1994

  committed
  9af98ac

  Copy full SHA for 9af98ac

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

  shiyu1994

  committed
  e34fcce

  Copy full SHA for e34fcce

• fix include order

  shiyu1994

  committed
  499639d

  Copy full SHA for 499639d

• fix include order

  shiyu1994

  committed
  6fe4874

  Copy full SHA for 6fe4874

• remove extra space

  shiyu1994

  committed
  3cf4c74

  Copy full SHA for 3cf4c74

Commits on Mar 15, 2022

• address review comments

  shiyu1994

  committed
  34fdfe4

  Copy full SHA for 34fdfe4

• add warning when cuda_exp is used together with deterministic

  shiyu1994

  committed
  3bb91ae

  Copy full SHA for 3bb91ae

Commits on Mar 21, 2022

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

  shiyu1994

  committed
  e47d009

  Copy full SHA for e47d009

• revert changing order of included headers

  shiyu1994

  committed
  53430dd

  Copy full SHA for 53430dd