Install

Welcome to Caffe2! Get started with deep learning today by following the step by step guide on how to download and install Caffe2.

Select your preferred platform and install type.

This build is confirmed for:

  • Ubuntu 14.04
  • Ubuntu 16.04

Required Dependencies#

1
2
3
4
5
6
7
8
9
10
11
sudo apt-get update
sudo apt-get install -y --no-install-recommends \
build-essential \
cmake \
git \
libgoogle-glog-dev \
libprotobuf-dev \
protobuf-compiler \
python-dev \
python-pip
sudo pip install numpy protobuf

Optional GPU Support#

If you plan to use GPU instead of CPU only, then you should install NVIDIA CUDA 8 and cuDNN v5.1 or v6.0, a GPU-accelerated library of primitives for deep neural networks. NVIDIA’s detailed instructions or if you’re feeling lucky try the quick install set of commands below.

Update your graphics card drivers first! Otherwise you may suffer from a wide range of difficult to diagnose errors.

For Ubuntu 14.04

1
2
3
4
5
sudo apt-get update && sudo apt-get install wget -y --no-install-recommends
wget "http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/cuda-repo-ubuntu1404_8.0.61-1_amd64.deb"
sudo dpkg -i cuda-repo-ubuntu1404_8.0.61-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda

For Ubuntu 16.04

1
2
3
4
5
sudo apt-get update && sudo apt-get install wget -y --no-install-recommends
wget "http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_8.0.61-1_amd64.deb"
sudo dpkg -i cuda-repo-ubuntu1604_8.0.61-1_amd64.deb
sudo apt-get update
sudo apt-get install cuda

Install cuDNN (all Ubuntu versions)#

Version 5.1

1
2
3
4
CUDNN_URL="http://developer.download.nvidia.com/compute/redist/cudnn/v5.1/cudnn-8.0-linux-x64-v5.1.tgz"
wget ${CUDNN_URL}
sudo tar -xzf cudnn-8.0-linux-x64-v5.1.tgz -C /usr/local
rm cudnn-8.0-linux-x64-v5.1.tgz && sudo ldconfig

Version 6.0 Visit NVIDIA’s cuDNN download to register and download the archive. Follow the same instructions above switching out for the updated library.

Optional Dependencies#

Note libgflags2 is for Ubuntu 14.04. libgflags-dev is for Ubuntu 16.04.

1
2
# for Ubuntu 14.04
sudo apt-get install -y --no-install-recommends libgflags2
1
2
# for Ubuntu 16.04
sudo apt-get install -y --no-install-recommends libgflags-dev
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
# for both Ubuntu 14.04 and 16.04
sudo apt-get install -y --no-install-recommends \
libgtest-dev \
libiomp-dev \
libleveldb-dev \
liblmdb-dev \
libopencv-dev \
libopenmpi-dev \
libsnappy-dev \
openmpi-bin \
openmpi-doc \
python-pydot
sudo pip install \
flask \
future \
graphviz \
hypothesis \
jupyter \
matplotlib \
pydot python-nvd3 \
pyyaml \
requests \
scikit-image \
scipy \
setuptools \
six \
tornado

Clone & Build#

1
2
3
git clone --recursive https://github.com/caffe2/caffe2.git && cd caffe2
make && cd build && sudo make install
python -c 'from caffe2.python import core' 2>/dev/null && echo "Success" || echo "Failure"

Run this command below to test if your GPU build was a success. You will get a test output either way, but it will warn you at the top of the output if CPU was used instead along with other errors like missing libraries.

1
python -m caffe2.python.operator_test.relu_op_test

Environment Variables#

These environment variables may assist you depending on your current configuration. When using the install instructions above on the AWS Deep Learning AMI you don’t need to set these variables. However, our Docker scripts built on Ubuntu-14.04 or NVIDIA’s CUDA images seem to benefit from having these set. If you ran into problems with the build tests above then these are good things to check. Echo them first and see what you have and possibly append or replace with these directories. Also visit the Troubleshooting section.

1
2
3
4
5
echo $PYTHONPATH
# export PYTHONPATH=/usr/local:$PYTHONPATH
# export PYTHONPATH=$PYTHONPATH:/home/ubuntu/caffe2/build
echo $LD_LIBRARY_PATH
# export LD_LIBRARY_PATH=/usr/local/lib:$LD_LIBRARY_PATH

Setting Up Tutorials & Jupyter Server#

If you’re running this all on a cloud computer, you probably won’t have a UI or way to view the IPython notebooks by default. Typically, you would launch them locally with ipython notebook and you would see a localhost:8888 webpage pop up with the directory of notebooks running. The following example will show you how to launch the Jupyter server and connect to remotely via an SSH tunnel.

First configure your cloud server to accept port 8889, or whatever you want, but change the port in the following commands. On AWS you accomplish this by adding a rule to your server’s security group allowing a TCP inbound on port 8889. Otherwise you would adjust iptables for this.

Next you launch the Juypter server.

1
jupyter notebook --no-browser --port=8889

Then create the SSH tunnel. This will pass the cloud server’s Jupyter instance to your localhost 8888 port for you to use locally. The example below is templated after how you would connect AWS, where your-public-cert.pem is your own public certificate and ubuntu@super-rad-GPU-instance.compute-1.amazonaws.com is your login to your cloud server. You can easily grab this on AWS by going to Instances > Connect and copy the part after ssh and swap that out in the command below.

1
ssh -N -f -L localhost:8888:localhost:8889 -i "your-public-cert.pem" ubuntu@super-rad-GPU-instance.compute-1.amazonaws.com

Troubleshooting#

PYTHON ERRORS  
Python version Python is core to run Caffe2. We currently require Python2.7Ubuntu 14.04 and greater have Python built in by default, and that can be used to run Caffe2. To check your version: python --version
Solution If you want the developer version of python, you could install the dev package for Python: sudo apt-get install python-dev
Python environment You may have another version of Python installed or need to support Python version 3 for other projects.
Solution Try virtualenv or Anaconda. The Anaconda platform provides a single script to install many of the necessary packages for Caffe2, including Python. Using Anaconda is outside the scope of these instructions, but if you are interested, it may work well for you.
pip version If you plan to use Python with Caffe2 then you need pip.
Solution sudo apt-get install python-pip and also try using pip2 instead of pip.
BUILDING FROM SOURCE  
OS version Caffe2 requires Ubuntu 14.04 or greater.
git While you can download the Caffe2 source code and submodules directly from GitHub as a zip, using git makes it much easier.
Solution sudo apt-get install git
protobuf You may experience an error related to protobuf during the make step.
Solution Make sure you’ve installed protobuf in both of these two ways: sudo apt-get install libprotobuf-dev protobuf-compiler && sudo pip install protobuf
libgflags2 error This optional dependency is for Ubuntu 14.04.
Solution Use apt-get install libgflags-dev for Ubuntu 16.04.
GPU SUPPORT  
GPU errors Unsupported GPU or wrong version
Solution You need to know the specific deb for your version of Linux. sudo dpkg -i cuda-repo-<distro>_<version>_<architecture>.debRefer to NVIDIA’s installation guide.
Build issues Be warned that installing CUDA and cuDNN will increase the size of your build by about 4GB, so plan to have at least 12GB for your Ubuntu disk size.
CAFFE2 PYTHON  
Module not found Verify that Caffe2 was installed correctly
Solution Run the following: python -c 'from caffe2.python import core' 2>/dev/null && echo "Success" || echo "Failure"An output of Success means you are ready to with Caffe2 - congratulations! An output of Failure usually means you have not installed one of the dependencies correctly.
Dependencies missing It’s possible you’re trying to run something that was using an optional dependency.
Solution sudo pip install setuptools flask jupyter matplotlib scipy pydot tornado python-nvd3 scikit-image pyyaml
matplotlib error Sometimes you need setuptools first: sudo pip install -U pip setuptools && sudo pip install matplotlib
model downloader error If you need to run it as sudo (because it’s trying to install the model in /usr/local/caffe2…), then PYTHONPATH might not be visible in that context.
Solution sudo visudo then add this line: Defaults env_keep += "PYTHONPATH"
“AttributeError: ‘module’ object has no attribute ‘MakeArgument’” Occurs when calling core.CreateOperator
Solution Check your install directory (/usr/local/), and remove the folder /caffe2/python/utils
OTHER ERRORS  
libdc1394 error for some reason once opencv is installed you may get errors with libdc1394
Solution ln /dev/null /dev/raw1394 but that’s not persistent so try sh -c 'ln -s /dev/null /dev/raw1394' or when instantiating the container use: --device /dev/null:/dev/raw1394
caffe2_pybind11_state_gpu WARNING:root:Debug message: No module named caffe2_pybind11_state_gpu
Solution ignore if you’re using CPU-only
Python kernel crashing This happens when you try to call Jupyter server directly (like in a Docker container).
Solution Use sh -c "jupyter notebook ..." to get around this problem.
Exception: “dot” not found in path This happens in some of the tutorials when graphing.
Solution Make sure you have graphviz and pydotsudo apt-get install python-pydot and sudo pip install graphvizor brew install these to fix the problem.

Dependencies#

Try to keep the system and python dependencies at the same version. We’ve encountered issues when the python version is more updated than the system version or vice versa.

SYSTEM DEPENDENCIES  
cmake  
git  
gflags  
glog: Google Logging Module  
NumPy  
protobuf: Google Protocol Buffers version 3.2.0
Build tools for C++ 11 Xcode CLTs & automake (mac/iOS), build-essential (linux), Visual Studio (win), Android Studio (droid)
PYTHON DEPENDENCIES  
gflags  
glog: Google Logging Module  
NumPy  
protobuf: Google Protocol Buffers version 3.2.0

Strictly speaking, the core dependencies above are all you need to run the core Caffe2 successfully. However, for real-world deep learning (e.g., image processing, mathematical operations, etc), there are other dependencies that you will want to install in order to experience the full features of Caffe2.

OPTIONAL SYSTEM DEPENDENCIES  
cuDNN if using GPU, this is needed for Caffe2’s cuDNN operators
Eigen 3  
LevelDB  
Nvidia CUDA v6.5 or greater
OpenCV for image-related operations; requires leveldb <= v1.19
OpenMPI for MPI-related Caffe2 operators
RocksdB for Caffe2’s RocksDB IO backend
ZeroMQ needed for Caffe2’s ZmqDB IO backend (serving data through a socket)
PYTHON OPTIONAL DEPENDENCIES  
There are also various Python libraries that will be valuable in your experience with Caffe2. Many of these are required to run the tutorials.  
Flask  
Graphviz  
Hypothesis  
Jupyter for the Jupyter Notebook
LevelDB  
lmdb  
Matplotlib  
Pydot  
Python-nvd3  
pyyaml  
requests  
Scikit-Image  
SciPy  
setuptools  
Tornado  
ZeroMQ  
WHAT’S IN THIRD PARTY?  
Whether building from source or installing from the Python wheel, you also get complimentary tools installed as well.  
Android cmake  
benchmark  
cnmem  
cub  
eigen  
googletest  
ios-cmake  
nccl  
nervanagpu  
NNPACK requires ninja and confu to build
Google Protocol Buffers (protobuf)  
pybind11  

caffe2--ubuntu16.04--14.04--install的更多相关文章

  1. Ubuntu LTS 系统学习使用体会和实用工具软件汇总 6.04 8.04 10.04 12.04 14.04 16.04

    Ubuntu LTS 系统学习体会和工具软件汇总 6.04 8.04 10.04 12.04 14.04 16.04 ubuntu入门必备pdf:http://download.csdn.net/de ...

  2. CVE-2015-1328 Ubuntu 12.04, 14.04, 14.10, 15.04 overlayfs Local Root

    catalog . 引言 . Description . Effected Scope . Exploit Analysis . Principle Of Vulnerability . Patch ...

  3. 安装ubuntu出现BUG soft lockup的解决方法(16.04 14.04)

    对于16.04而言,当时用的是UtrISO 安装的,导致安装过程用会出现 “not a com32r image” 的错误,解决方法见上文的: boot: live 华硕Z9主板安装16.04以上系统 ...

  4. Ubuntu16.04 --> 14.04

    从16到14 自认为14是比较稳定的.从安装依赖上说. 14安装应用 更多参见[请直接拉到"华丽丽的分割线"下面] Java9 注意,添加源的时候先把lantern打开!!! 添加 ...

  5. Laptop Ubuntu16.04/14.04 安装Nvidia显卡驱动

    笔记本型号 机械革命(MECHREVO)深海泰坦X6Ti-S(黑曜金)15.6英寸 CPU型号 i5-7300HQ 内存 8G 硬盘容量 128SSD+1T机械硬盘 显卡 GeForce GTX 10 ...

  6. Ubuntu16.04 14.04安装配置Caffe(GPU版)

    caffe配置过程很长啊,坑非常多,没有linux基础的估计会香菇的.我参考了网上很多的帖子,基本上每个帖子都有或多或少的问题,研究很久最终配置成功.参考过的帖子太多,都记不太清来源了.为了对前人的感 ...

  7. Ubuntu16.04 14.04 配置caffe(CPU only)

    1.安装依赖 sudo apt-get install libprotobuf-dev libleveldb-dev libsnappy-dev libopencv-dev libhdf5-seria ...

  8. Ubuntu Server 12.04(14.04) 静态IP简洁配置

    1.配置静态IP地址: # vim /etc/network/interfaces 原内容有如下4行:auto loiface lo inet loopback auto eth0iface eth0 ...

  9. Install Google Pinyin on Ubuntu 14.04

    Install Google Pinyin on Ubuntu 14.04 I've been spending more and more time on Ubuntu and I'm not us ...

  10. How To Install Tinc and Set Up a Basic VPN on Ubuntu 14.04

    Introduction In this tutorial, we will go over how to use Tinc, an open source Virtual Private Netwo ...

随机推荐

  1. 【Luogu】P1430序列取数(DP)

    题目链接 博弈DP太喵了qwq 设f[i][j]表示剩下区间[i,j]要取,先手最大值 明显我们要从这区间里面拿个最大的 就等价于这段区间的前缀和,我们要给对手留下个最小的 就是f[i][j]=sum ...

  2. 刷题总结——拆网线(noip模拟 贪心)

    题目: 给定一颗树··在保证有k个点与其它点连接的情况下问最少保留多少条边···· 树的节点树n和k均小于100000: 题解: 很容易看出来我们要尽量保留那种一条边连两个节点的情况···· 然后考试 ...

  3. Docker部署注册中心、Docker创建私有镜像库、自签名证书、Deploy a registry server

    这是我在内部部署Docker Registry时记录下来的笔记,操作环境是Centos 7.Docker 18.06.1-ce 1.运行registry 我当前所使用的主机的IP是192.168.1. ...

  4. Idea连接服务器docker并部署代码到docker实现一键启动

    好记性不如烂笔头,写笔记是为了回头看的. 谁要是不小心搜了看了,如有不足之处敬请谅解. 一.准备工作 虚拟机centos7.X,docker1.3.X,Win10 Idea2018.1 默认Idea已 ...

  5. gridview中的相关事件操作

    原文发布时间为:2008-07-27 -- 来源于本人的百度文章 [由搬家工具导入] using System;using System.Data;using System.Configuration ...

  6. Codeforces 375D Tree and Queries(DFS序+莫队+树状数组)

    题目链接  Tree and Queries 题目大意  给出一棵树和每个节点的颜色.每次询问$vj, kj$ 你需要回答在以$vj$为根的子树中满足条件的的颜色数目, 条件:具有该颜色的节点数量至少 ...

  7. IntelliJ IDEA提示:Class JavaLaunchHelper is implemented in both的错误解决

    这个错误是Mac下特有的,并且据说是一个老Bug,不影响使用. 修复方法: Help->Edit Custom Properties,没有这个properties文件的话,IDEA会提示创建,然 ...

  8. [转] DataSet的的几种遍历

    1. 多表多行多列的情况 foreach (DataTable dt in YourDataset.Tables) //遍历所有的datatable { foreach (DataRow dr in ...

  9. soursTree新建过程.md

    网上博客 https://www.cnblogs.com/tian-xie/p/6264104.html 主要的推送流程 完成所有项目的远程推送工作 点击git工作流选择第二个建立新的版本; 输入发布 ...

  10. 邁向IT專家成功之路的三十則鐵律 鐵律九:IT人社群互動之道-縮小自己

    身為一位專業的IT人士所要學習的東西實在非常的多,然而對於時間相當有限的我們,最快速的學習方法就是向他人學習,而向他人學習的首要態度就是「縮小自己」.唯有將自己縮小到別人的眼睛裡,才能夠讓他們真心誠意 ...