Low precision neural networks

Low precision neural networks

This project aims to implement an optimized framework for XNOR-nets

Artificial Intelligence, Modern Code

  • 0 Collaborators

  • 0 Followers

    Follow

Description

https://github.com/akhauriyash/XNOR-convolution

This project hopes to implement an optimized model of a quantized neural network.

TO DO: - [x] Add support for variable input sizes - [x] Add support for 3D convolution - [x] Parallelize per convolution - [ ] Maximize shared memory usage - balance channel parallelization - [ ] Create a full precision verification kernel - [ ] Add full support for custom kernel sizes - [ ] Build a parser to take in shape arguments - [ ] Add in proof of concept for densely connected networks

Gallery

Medium pic

Yash A. created project Low precision neural networks

Medium 8d04aea1 6c84 461c 8501 d3c0e23b0837

Low precision neural networks

https://github.com/akhauriyash/XNOR-convolution

This project hopes to implement an optimized model of a quantized neural network.

TO DO: - [x] Add support for variable input sizes - [x] Add support for 3D convolution - [x] Parallelize per convolution - [ ] Maximize shared memory usage - balance channel parallelization - [ ] Create a full precision verification kernel - [ ] Add full support for custom kernel sizes - [ ] Build a parser to take in shape arguments - [ ] Add in proof of concept for densely connected networks

No users to show at the moment.

No users to show at the moment.