robertsdionne/deepdream-install.md

## deepdream-install.md

      
    Raw
  

              deepdream-install.md
            
          
    #!/usr/bin/env bash

# Assuming OS X Yosemite 10.10.4

# Install XCode and command line tools
# See https://itunes.apple.com/us/app/xcode/id497799835?mt=12#
# See https://developer.apple.com/library/mac/documentation/Darwin/Reference/ManPages/man1/xcode-select.1.html
xcode-select --install

# Install the Homebrew package manager if you don't already use it; see source http://brew.sh
ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"

# Install homebrew cask
brew install caskroom/cask/brew-cask

# Install Python 2.7 and pip
brew install python pip

# Install Cuda 7
brew cask install cuda

# Optionally install cuDNN from https://developer.nvidia.com/cudnn which requries
# 1) Registering for an NVIDIA account, accepting an end-user-license-agreement
# 2) Downloading and extracting a tar.gz archive
# 3) Copying lib* files to /usr/local/cuda/lib and cudnn.h to /usr/local/cuda/include
#
# robertsdionne:~/$ ls /usr/local/cuda/lib/*dnn*
# /usr/local/cuda/lib/libcudnn.6.5.dylib /usr/local/cuda/lib/libcudnn_static.a
# /usr/local/cuda/lib/libcudnn.dylib
# robertsdionne:~/$ ls /usr/local/cuda/include/*dnn*
# /usr/local/cuda/include/cudnn.h

# Install dependencies; see http://caffe.berkeleyvision.org/install_osx.html
brew install --fresh -vd snappy leveldb gflags glog szip lmdb
# need the homebrew science source for OpenCV and hdf5
brew tap homebrew/science
brew install hdf5 opencv
# with Python pycaffe needs dependencies built from source
brew install --build-from-source --with-python --fresh -vd protobuf
# We need boost{,-python}-1.57.0 due to a bug in 1.58.0 on Yosemite 10.10.4
brew install --build-from-source --fresh -vd \
    https://raw.githubusercontent.com/Homebrew/homebrew/6fd6a9b6b2f56139a44dd689d30b7168ac13effb/Library/Formula/boost.rb \
    https://raw.githubusercontent.com/Homebrew/homebrew/3141234b3473717e87f3958d4916fe0ada0baba9/Library/Formula/boost-python.rb

# Clone caffe
git clone https://github.com/BVLC/caffe.git
cd caffe

# Install python dependencies
pip install --requirement python/requirements.txt

# Adjust Makefile.config
cp Makefile.config.example Makefile.config

########## Make the following adjustments to Makefile.config:
diff --git a/Makefile.config b/../caffe/Makefile.config
index a873502..c6806c2 100644
--- a/Makefile.config
+++ b/../caffe/Makefile.config
@@ -2,7 +2,7 @@
 # Contributions simplifying and improving our build system are welcome!
 
 # cuDNN acceleration switch (uncomment to build with cuDNN).
-# USE_CUDNN := 1
+# USE_CUDNN := 1  ########### Keep it like this to disable cuDNN
+USE_CUDNN := 1    ########### Make it like this to enable cuDNN
 
 # CPU-only switch (uncomment to build without GPU support).
 # CPU_ONLY := 1
@@ -34,7 +34,7 @@ BLAS := atlas
 # Custom (MKL/ATLAS/OpenBLAS) include and lib directories.
 # Leave commented to accept the defaults for your choice of BLAS
 # (which should work)!
-# BLAS_INCLUDE := /path/to/your/blas
+BLAS_INCLUDE := /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX10.10.sdk/System/Library/Frameworks/Accelerate.framework/Versions/Current/Frameworks/vecLib.framework/Headers
 # BLAS_LIB := /path/to/your/blas
 
 # Homebrew puts openblas in a directory that is not on the standard search path
@@ -48,8 +48,8 @@ BLAS := atlas
 
 # NOTE: this is required only if you will compile the python interface.
 # We need to be able to find Python.h and numpy/arrayobject.h.
-PYTHON_INCLUDE := /usr/include/python2.7 \
-		/usr/lib/python2.7/dist-packages/numpy/core/include
+PYTHON_INCLUDE := /usr/local/lib/python2.7/site-packages/numpy/core/include/ \
+		/usr/local/Cellar/python/2.7.10/Frameworks/Python.framework/Versions/2.7/include/python2.7/
 # Anaconda Python distribution is quite popular. Include path:
 # Verify anaconda location, sometimes it's in root.
 # ANACONDA_HOME := $(HOME)/anaconda
@@ -58,7 +58,9 @@ PYTHON_INCLUDE := /usr/include/python2.7 \
 		# $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include \
 
 # We need to be able to find libpythonX.X.so or .dylib.
-PYTHON_LIB := /usr/lib
+PYTHON_LIB := /usr/local/lib/python2.7 \
+		/usr/local/Cellar/python/2.7.10/Frameworks/Python.framework/Versions/2.7/lib/ \
+		/usr/local/lib/python2.7/site-packages/numpy/lib
 # PYTHON_LIB := $(ANACONDA_HOME)/lib
 
 # Homebrew installs numpy in a non standard path (keg only)
########## Make the following adjustments to Makefile:
diff --git a/Makefile b/../caffe/Makefile
index e4e66df..8babaf8 100644
--- a/Makefile
+++ b/../caffe/Makefile
@@ -335,7 +335,7 @@ else
 			LDFLAGS += -framework Accelerate
 		else
 			BLAS_INCLUDE ?= /System/Library/Frameworks/vecLib.framework/Versions/Current/Headers/
-			LDFLAGS += -framework vecLib
+			LDFLAGS += -framework Accelerate
 		endif
 	endif
 endif
# Build caffe
make all -j8; make pycaffe -j8

# Package caffe
make distribute

# Set your environment variables for success
export PYTHONPATH="$HOME/caffe/distribute/python:$PYTHONPATH"
export DYLD_FALLBACK_LIBRARY_PATH=/usr/local/cuda/lib:/usr/local/lib:$HOME/caffe/distribute/lib:$DYLD_FALLBACK_LIBRARY_PATH

# Clone deepdream
cd ..
git clone https://github.com/google/deepdream.git
cd deepdream

# Install python requirements
pip install numpy scipy scikit-image pillow ipython[notebook]

# Start IPython Notebook and select dream.ipynb
ipython notebook dream.ipynb