LBPnet

Implementation of "Resource Efficient and Error Resilient Neural Networks" dissertation
using pytorch and cuda

This implementation is a little bit different (might not be efficient) than what was explained in the dissertation [1], but it is more straightforward to understand.

I also did some small changes that you could fix to be as original as the paper suggest.

[1]: Resource Efficient and Error Resilient Neural Networks

How to use

• Please run on a GPU node!

• Make sure that ninja is inside your environmnet path:

  • Install ninja from its github source and
    add this line to your .bash_profile file in your home (~) directory

      export PATH=$PATH:$YourHOMEpath/ninja
    

  • Run it on Linux, not Windows! or change the torch header files to at in .cpp and .cu files.

• module load:

    cuda/11.4
    
    pytorch-gpu/py38/1.8  
  

• Then execute python myLBP.py

License

Please feel free to use this repository. For other purposes (e.g. commercial purposes) please contact me beforehand.

If you find this code useful in your research, please consider citing the original paper: Resource Efficient and Error Resilient Neural Networks