Author: | Jérémie DECOCK |
---|---|
Contact: | jd.jdhp@gmail.com |
Revision: | 1 |
Date: | January 4, 2017 |
License: | Creative Commons 4.0 (CC BY-SA 4.0) |
Table of Contents
Warning
Please note that at this time, this document is a work in progress that is not intended to be read by anyone other than its authors.
The opensource Google Brain library for Deeplearning.
Install dependencies:
conda install theano conda install nose-parameterized mkdir ~/pylearn2_data
Set PYLEARN2_DATA_PATH in ~/.bash_profile (MacOS):
export PYLEARN2_DATA_PATH="/Users/jdecock/pylearn2_data" export PYLEARN2_VIEWER_COMMAND="open -Wn"
Then run:
source ~/.bash_profile
Install pylearn2 (see http://deeplearning.net/software/pylearn2/index.html#download-and-installation):
mkdir -p ~/bin/ cd ~/bin/ git clone git://github.com/lisa-lab/pylearn2.git cd pylearn2 python setup.py develop --user
Run the first tutorial (see ~/bin/pylearn2/pylearn2/scripts/tutorials/grbm_smd/README):
# Download the CIFAR10 dataset cd ~/bin/pylearn2/pylearn2/scripts/datasets ./download_cifar10.sh # Make the dataset cd ~/bin/pylearn2/pylearn2/scripts/tutorials/grbm_smd python make_dataset.py # Train the model ~/bin/pylearn2/pylearn2/scripts/train.py cifar_grbm_smd.yaml # Show results ~/bin/pylearn2/pylearn2/scripts/show_weights.py cifar_grbm_smd.pkl ~/bin/pylearn2/pylearn2/scripts/plot_monitor.py cifar_grbm_smd.pkl
Features:
TODO
TODO
TODO
TODO
TODO
Theano a low level fast computation library between numpy and sympy that is often used by ML libraries (Pylearn2, Blocks, Keras, ...). It makes some linear algebra operation very quickly using GPU, dynamic C code generation, parallelization, ...
It's not a ML library but it's useful to mention it here as it's very often referred by ML libraries.
This document is provided under the terms and conditions of the Creative Commons 4.0 (CC BY-SA 4.0) license.