Software in Non-Default Locations

Software

Many software packages are available by taking steps to include them in your environment.  There are two well-supported mechanisms for doing this.

Environment Modules are available to take advantage of the following packages (the list is not exhaustive and may not be completely current):

  • Boost 1.55.0
  • Coq 8.13
  • Haskel 2014
  • Intel Compiler (icc) 19.1.1.217
  • Matlab R2020b
  • Ocaml 4.11.1

Software Collections (SCL) are available for the following packages (for which the above disclaimer also applies):

  • DevToolSet 3.1 (Includes GCC v. 4.9.2)
  • DevToolSet 8.1 (includes GCC v 8.3.1)
  • MySQL 5.5
  • Git 2.9
  • Common Java Packages 1.1

Some of the Available Software and Libraries

  • Caffe - Framework for convolutional neutral network algorithms, developed with speed in mind.  It was created by Yangqing Jia, and is in active development by the Berkeley Vision and Learning Center.  

The best approach to running caffe would be to install your own virtual environment or anaconda.  This is due to the fact that caffe requires outdated versions of python modules, which are no longer compatible with the rest of the global python environment.  

$ wget https://raw.githubusercontent.com/BVLC/caffe/master/python/requirements.txt 
$ pip install -r requirements.txt 
$ module load caffe/1.00 
$ python 
Python 2.7.5 (default, Nov 8 2016, 19:11:39) [GCC 4.8.5 20150623 (Red Hat 4.8.5-11)] on linux2 Type "help", "copyright", "credits" or "license" for more information. 
>>> import caffe

  • Caffe2 - A lightweight, module, and scalable deep learning framework.  Building on the original Caffe, Caffe2 is designed with expresion, speed, and modularity in mind. 

$ module load caffe2/0.7.0
$ python -c 'from caffe2.python import core' 2>/dev/null && echo "Success" || echo "Failure"
Success

  • Tensorflow - An open source software library for numerical computation using data flow graphs.

Same as caffe, the best way to install tensorflow with all the correct requirements is either with virtual environment or anaconda. 

$ pip install tensorflow # CPU tensorflow
$ pip install tensorflow-gpu # GPU tensorflow
$ module load cudatoolkit
$ module load cudnn7
$ python Python 2.7.5 (default, Nov 8 2016, 19:11:39) [GCC 4.8.5 20150623 (Red Hat 4.8.5-11)] on linux2 Type "help", "copyright", "credits" or "license" for more information. 
>>> import tensorflow

  • Matlab - a high-performance language for technical computing.