如何使用多线程在 OSX 上安装 XGBoost

Question

我正在尝试按照此处的指南在我的 mac（osx 10.12.1）上安装 xgboost，但我遇到了一些问题。

---

第1步

通过以下方式获取具有 openmp 支持的 gcc-6.xx brew install gcc --without-multilib

终端

Ben$ brew install gcc --without-multilib
Error: gcc-5.3.0 already installed
To install this version, first `brew unlink gcc`

Ben$ brew unlink gcc
Unlinking /usr/local/Cellar/gcc/5.3.0... 1288 symlinks removed

Ben$ brew install gcc --without-multilib
[26 minutes later]
==> Summary
  /usr/local/Cellar/gcc/6.2.0: 1,358 files, 238.3M, built in 26 minutes 20 seconds

---

第2步

克隆存储库 git clone --recursive https://github.com/dmlc/xgboost

终端

Ben$ git clone --recursive https://github.com/dmlc/xgboost
Cloning into 'xgboost'...
remote: Counting objects: 18754, done.
remote: Compressing objects: 100% (21/21), done.
remote: Total 18754 (delta 1), reused 0 (delta 0), pack-reused 18733
Receiving objects: 100% (18754/18754), 6.23 MiB | 3.74 MiB/s, done.
Resolving deltas: 100% (11347/11347), done.
Checking connectivity... done.
Submodule 'dmlc-core' (https://github.com/dmlc/dmlc-core) registered for path 'dmlc-core'
Submodule 'rabit' (https://github.com/dmlc/rabit) registered for path 'rabit'
Cloning into '/Users/Ben/xgboost/dmlc-core'...
Cloning into '/Users/Ben/xgboost/rabit'...
Submodule path 'dmlc-core': checked out 'f35f14f30835af238257b979cc1fac3e41ff3291'
Submodule path 'rabit': checked out 'a9a2a69dc1144180a43f7d2d1097264482be7817'

---

Step3

使用以下命令构建 cd xgboost; cp make/config.mk ./config.mk; make -j4

注意：如果您使用 OSX El Capitan，brew 会安装 gcc 最新版本的 gcc-6。所以你可能需要修改 Makefile#L46 并将 gcc-5 更改为 gcc-6。之后在 make/config.mk 中将 gcc-5/g++-5 更改为 gcc-6/g++-6 然后使用以下命令构建

嗯。不完全确定在这里做什么。我看到里面有一个名为“Makefile”的文件xgboost/。在里面我看到了很多我不明白的东西，但这些行看起来很相关

生成文件（片段）

# on Mac OS X, force brew gcc-6, since the Xcode c++ fails anyway
# it is useful for pip install compiling-on-the-fly
OS := $(shell uname)
ifeq ($(OS), Darwin)
export CC = $(if $(shell which gcc-6),gcc-6,$(if $(shell which gcc-mp-6), gcc-mp-6, clang))
export CXX = $(if $(shell which g++-6),g++-6,$(if $(shell which g++-mp-6),g++-mp-6, clang++))
endif

将忽略这一点并继续......

终端

Ben$ cd xgboost; cp make/config.mk ./config.mk; make -j4
[Tons of log output. Appears to finish succesfully]
...
g++-6 -std=c++0x -Wall -O3 -msse2  -Wno-unknown-pragmas -funroll-loops -Iinclude   -Idmlc-core/include -Irabit/include -fPIC -fopenmp -shared -o lib/libxgboost.so build/learner.o build/logging.o build/c_api/c_api.o build/c_api/c_api_error.o build/common/common.o build/data/data.o build/data/simple_csr_source.o build/data/simple_dmatrix.o build/data/sparse_page_dmatrix.o build/data/sparse_page_raw_format.o build/data/sparse_page_source.o build/data/sparse_page_writer.o build/gbm/gblinear.o build/gbm/gbm.o build/gbm/gbtree.o build/metric/elementwise_metric.o build/metric/metric.o build/metric/multiclass_metric.o build/metric/rank_metric.o build/objective/multiclass_obj.o build/objective/objective.o build/objective/rank_obj.o build/objective/regression_obj.o build/tree/tree_model.o build/tree/tree_updater.o build/tree/updater_colmaker.o build/tree/updater_histmaker.o build/tree/updater_prune.o build/tree/updater_refresh.o build/tree/updater_skmaker.o build/tree/updater_sync.o dmlc-core/libdmlc.a rabit/lib/librabit.a -pthread -lm  -fopenmp 

---

Step4（此处有说明？）

如果您想使用最新的 xgboost 版本并且已经编译了 xgboost，请使用library(devtools); install('xgboost/R-package') 手动安装 xgboost 包（相应地更改编译 xgboost 的路径）。

新建 R 会话

library(devtools)
install("/Users/Ben/xgboost/R-package")

Installing xgboost
'/Library/Frameworks/R.framework/Resources/bin/R' --no-site-file --no-environ --no-save --no-restore --quiet CMD INSTALL  \
  '/Users/Ben/xgboost/R-package' --library='/Library/Frameworks/R.framework/Versions/3.3/Resources/library' --install-tests 

* installing *source* package ‘xgboost’ ...
** libs
g++-5 -std=c++11 -I/Library/Frameworks/R.framework/Resources/include -DNDEBUG -I../..//include -I../..//dmlc-core/include -I../..//rabit/include -I../../ -DXGBOOST_STRICT_R_MODE=1 -DDMLC_LOG_BEFORE_THROW=0 -DDMLC_ENABLE_STD_THREAD=1 -DDMLC_DISABLE_STDIN=1 -DDMLC_LOG_CUSTOMIZE=1 -DXGBOOST_CUSTOMIZE_LOGGER=1 -DRABIT_CUSTOMIZE_MSG_ -DRABIT_STRICT_CXX98_ -I/usr/local/opt/llvm/include   -fopenmp  -fPIC  -Wall -mtune=core2 -g -O2 -c xgboost_R.cc -o xgboost_R.o
make: g++-5: No such file or directory
make: *** [xgboost_R.o] Error 1
ERROR: compilation failed for package ‘xgboost’
* removing ‘/Library/Frameworks/R.framework/Versions/3.3/Resources/library/xgboost’
* restoring previous ‘/Library/Frameworks/R.framework/Versions/3.3/Resources/library/xgboost’
Error: Command failed (1)

失败:(我哪里出错了？

Answer 1

更新的解决方案

截至 2019 年 3 月，R 版本 3.5.3，xgboost 版本 0.82.0.1，自从我的旧答案以来，情况发生了变化。以下步骤基于此处的安装指南，但略有不同。

1. brew install cmake或brew upgrade cmake（cmake --version为我显示 3.14.0）
2. brew install gcc或brew upgrade gcc（请注意安装的 gcc 版本。它应该位于/usr/local/bin。我得到 gcc 8.3.0 ( gcc-8 --version)）
3. cd some/unimportant/directory
4. git clone --recursive https://github.com/dmlc/xgboost
5. cd xgboost/
6. mkdir build
7. cd build
8. CC=gcc-8 CXX=g++-8 cmake .. -DR_LIB=ON
9. make -j4
10. sudo make install

现在重新启动/刷新 RStudio，它应该被安装

在 R 中测试

set.seed(222)
N <- 2*10^5
p <- 350
x <- matrix(rnorm(N  * p), ncol = p)
y <- rnorm(N)

system.time(mymodel <- xgboost(
  # nthread = 4,
  data = x,
  label = y, 
  nrounds = 5, 
  objective = "reg:linear", 
  tree_method = "exact",
  max_depth = 10,
  min_child_weight = 1, 
  eta = 1, 
  subsample = 0.66, 
  colsample_bytree = 0.33
))

# Tested on 2018 MPB, xgboost version 0.82.0.1, multi-threaded version
# nthread = default:   7.4 seconds (elapsed)
# nthread = 1:        24.0 seconds (elapsed)
# nthread = 2:        13.7 seconds (elapsed)
# nthread = 4:         7.5 seconds (elapsed)

旧答案

令人惊讶的是，在 StackOverflow 上写下您的问题通常会直接引导您找到答案。在花了几个小时之后，我发现我需要改变这三行~/.R/Makevars

CC=gcc-5
CXX=g++-5
CXX1X = g++-5

到

CC=gcc-6
CXX=g++-6
CXX1X = g++-6

另外，我最终从“drat”仓库安装了 xgboost

install.packages("drat", repos="https://cran.rstudio.com")
drat:::addRepo("dmlc")
install.packages("xgboost", repos="http://dmlc.ml/drat/", type = "source")