Skip to content

Latest commit

 

History

History
122 lines (105 loc) · 2.71 KB

step10-install-caffe.md

File metadata and controls

122 lines (105 loc) · 2.71 KB

Caffe installation & quick check setup [back]

Main article with installation instructions caffe.berkeleyvision.org/installation.html

(1) Install Boost 1.55 version:

$ sudo apt-get install libboost1.55-all-dev

(2) Install Protobuf:

$ sudo apt-get install libprotobuf-dev libprotoc-dev libprotobuf-c0-dev python-protobuf protobuf-c-compiler protobuf-compiler

(3) Install Glog:

$ sudo apt-get install libgoogle-glog-dev

(4) Install Gflags:

$ sudo apt-get install libgflags-dev python-gflags

(5) Install HDF5:

$ sudo apt-get install libhdf5-dev hdf5-tools python-h5py

(6) Install lmdb, leveldb, snappy:

$ sudo apt-get install libsnappy-dev python-snappy  liblmdb-dev  libleveldb-dev
$ sudo pip install leveldb -U
$ sudo pip install lmdb -U

(7) Compile Caffe:

$ cd ~/deep-learning
$ cp Makefile.config.example Makefile.config

(7.1) Check flags in config:

  • USE_CUDNN := 1
  • BLAS := open
  • WITH_PYTHON_LAYER := 1

(7.2) Compile:

$ make -j8
$ make pycaffe
$ make distribute

(7.3) Create Caffe Env setuper:

$ cat ~/bin/set-caffe.sh
  #!/bin/bash
  bd="$HOME/deep-learning/caffe.git/distribute"
  LD_LIBRARY_PATH="${bd}/lib:$LD_LIBRARY_PATH"
  PYTHONPATH="${bd}/python:$PYTHONPATH"
  PATH="${bd}/bin:$PATH"
  export PATH LD_LIBRARY_PATH PYTHONPATH

$ echo "
source $HOME/bin/set-caffe.sh
" >> ~/.bashrc
$ echo "
source $HOME/bin/set-caffe.sh
" >> ~/.profile

(8) Check Caffe installation:

(8.1)Load MNIST Data:

$ cd ~/deep-learning/caffe.git/data/mnist/
$ ./get_mnist.sh
$ cd ~/deep-learning/caffe.git/

(8.2) Create dataset:

$ examples/mnist/create_mnist.sh

(8.3) Train on GPU #0

$ caffe.bin train -solver examples/mnist/lenet_solver.prototxt -gpu 0
I0219 17:30:42.840487  7692 caffe.cpp:185] Using GPUs 0
I0219 17:30:43.106034  7692 caffe.cpp:190] GPU 0: GRID K520
I0219 17:30:43.226626  7692 solver.cpp:48] Initializing solver from parameters: 
test_iter: 100
test_interval: 500
base_lr: 0.01
display: 100
max_iter: 10000
lr_policy: "inv"
gamma: 0.0001
power: 0.75
momentum: 0.9
weight_decay: 0.0005
snapshot: 5000
snapshot_prefix: "examples/mnist/lenet"
solver_mode: GPU
device_id: 0
net: "examples/mnist/lenet_train_test.prototxt"
.......
I0219 17:31:27.672592  7692 solver.cpp:338] Iteration 10000, Testing net (#0)
I0219 17:31:27.864341  7692 solver.cpp:406]     Test net output #0: accuracy = 0.9921
I0219 17:31:27.864397  7692 solver.cpp:406]     Test net output #1: loss = 0.0257289 (* 1 = 0.0257289 loss)
I0219 17:31:27.864413  7692 solver.cpp:323] Optimization Done.
I0219 17:31:27.864429  7692 caffe.cpp:222] Optimization Done.

... [Ok]