Python SVM with CUDA support.
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
examples Update svm_example.py Mar 14, 2016
pyKMLib Update __init__.py Mar 14, 2016
tests
.gitignore Update .gitignore Mar 14, 2016
LICENSE Update LICENSE Mar 14, 2016
README.md Update README.md Mar 14, 2016

README.md

pyKMLib

Python Kernel SVM library accelerated with CUDA. Library allows for classification sparse and big dataset with use of different sprase storage (matrix) format. CUDA SVM in python.

It is a partial python port of .net KMLib project https://github.com/ksirg/KMLib

author: Krzysztof Sopyła (krzysztofsopyla@gmail.com)

Prerequisits

  • Python 2.7
  • pycuda 2013.1.1
  • Numpy 1.7 MKL
  • Scipy
  • Numba

Ubuntu 13.10 prerequisits installation

##numba installation

  • llvm - This install llvm 3.4
 sudo apt-get install llvm
  • llvmpy - python llvm wrapper
wget https://github.com/llvmpy/llvmpy/releases/tag/0.12.3
tar zxvf 0.12.3.tar.gz
cd 0.12.3
sudo LLVM_CONFIG_PATH=/usr/bin/llvm-config python setup.py install
  • numba -
sudo pip install numba

pycuda installation

Warning!

sudo apt-get install pycuda - probably override your nvidia driver installation, so If you install nvidia driver and cuda toolkit previously than it is not recomended. (I have install cuda toolkit and driver with help http://askubuntu.com/questions/380609/anyone-has-successfully-installed-cuda-5-5-on-ubuntu-13-10-64-bit )

vim ~/.bashrc 
export CUDA_HOME=/usr/local/cuda
export CUDA_ROOT=${CUDA_HOME}
export LD_LIBRARY_PATH=${CUDA_HOME}/lib64

sudo PATH=$PATH LD_LIBRARY_PATH=$LD_LIBRARY_PATH pip install pycuda