RRAM Simulator Neural Network Training

This is an tensorflow implementation of neural network training and testing for RRAM based hardware accelerators. This is a simulator to emulate RRAM device variation, read/write noise and its effect on neural network inference.

Requirements: software

1. Requirements for Tensorflow (see: Tensorflow)

2. Python packages you might not have: cython, easydict

Requirements: hardware

1. For training, 3G of GPU memory is sufficient (using CUDNN)

Installation

1. Clone the rramtraining (make sure to clone with --recursive)

git clone --recursive https://amohant4@bitbucket.org/amohant4/rramtraining.git