Skip to content

Latest commit

 

History

History
26 lines (18 loc) · 959 Bytes

README.md

File metadata and controls

26 lines (18 loc) · 959 Bytes

A C++ Library for Convolutional Neural Nets with Parallel Computing(openMP, CUDA, MPI)

Usage:

g++ -std=c++11 -fopenmp lenet.cpp -o lenet
./lenet

  • This is a multi-threaded version of the model (with data parallelism) and you can change the number of threads by using:
    export OMP_NUM_THREADS=4

  • For using the MPI version of code, you need to compile using mpic++:
    mpic++ -std=c++11 -fopenmp lenet.cpp -o lenet

and you can run this on multi-node system!

Create your own Network

  • You can create your own deep neural network class by deriving from the Model class and adding all your layers in order by using addLayer() method.
  • You can also introduce your own Activation layers by extending the ActivationLayer.
  • You can create your custom Loss functions by extending the LossFunction class.

Work in Progress

Optimizations using:
openMP: COMPLETED
MPI: COMPLETED
CUDA: IN PROGRESS