Caffe (http://caffe2.ai) is a C++ deep neural network framework originally developed by Yangqing Jia at the Computer Vision department of UC Berkeley, which later became the Berkeley Artificial Intelligence Research Lab. Caffe is now being developed at Facebook and is part of the PyTorch project.

Installing PyTorch on Ubuntu 18.04

After installing miniconda3 and creating an environment for PyTorch,

conda create --name pytorch

you can install PyTorch for CUDA 9.1 by entering

conda install pytorch torchvision cuda91 -c pytorch


Building Caffe on Ubuntu 16.04

On Ubuntu make sure that the dependencies glog and gflags are installed before running cmake. Otherwise the Caffe build script tries to build a local version of glog and gflags and runs into a makefile configuration problem where the build process aborts and complains that it can't find automake-1.14. The version of automake from the Ubuntu repositories has been updated to 1.15. Simply run

sudo apt install libgoogle-glog-dev
sudo apt install libgflags-dev

before running

mkdir build
cd build
cmake -DCMAKE_INSTALL_PREFIX=/usr/local/ ..
make
sudo make install



Debug data: