From 1b4b5ec343db7dba58094c75aeca5633be5f0e47 Mon Sep 17 00:00:00 2001 From: Matthieu Courbariaux Date: Thu, 3 Sep 2015 21:29:28 -0400 Subject: [PATCH] cleaned up for push --- README.md | 7 +++++++ cifar10_cnn.py => cifar10.py | 0 mnist_mlp.py => mnist.py | 0 3 files changed, 7 insertions(+) rename cifar10_cnn.py => cifar10.py (100%) rename mnist_mlp.py => mnist.py (100%) diff --git a/README.md b/README.md index 61f4676..81af092 100644 --- a/README.md +++ b/README.md @@ -5,6 +5,7 @@ * Python, Numpy, Scipy * Theano 0.6 (Bleeding edge version) * Pylearn2 0.1 +* Lasagne 0.1 (only for the "lasagne" branch) * PyTables (only for the SVHN dataset) * a fast GPU or a large amount of patience @@ -13,6 +14,12 @@ This repository allows to easily reproduce the experimental results reported in: "BinaryConnect: Training Deep Neural Networks with binary weights during propagations". +## Branches + +There are two branches in this repository: +* The "master" branch is a pure Theano version of BinaryConnect. The results reported in our paper were obtained using this branch. +* The "lasagne" branch is a Lasagne (and thus Theano) version of BinaryConnect. The results are very similar to those of the "master" branch. + ## MNIST python mnist.py diff --git a/cifar10_cnn.py b/cifar10.py similarity index 100% rename from cifar10_cnn.py rename to cifar10.py diff --git a/mnist_mlp.py b/mnist.py similarity index 100% rename from mnist_mlp.py rename to mnist.py