https://github.com/ShigekiKarita/grain2

Autograd and GPGPU library for dynamic neural networks in D
https://github.com/ShigekiKarita/grain2

Last synced: 3 months ago
JSON representation

Autograd and GPGPU library for dynamic neural networks in D

• Host: GitHub
• URL: https://github.com/ShigekiKarita/grain2
• Owner: ShigekiKarita
• Created: 2019-07-05T11:30:24.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2020-03-14T17:15:05.000Z (over 4 years ago)
• Last Synced: 2024-05-02T21:45:05.119Z (6 months ago)
• Language: D
• Homepage: http://karita.xyz/grain2/grain.html
• Size: 1.05 MB
• Stars: 6
• Watchers: 3
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-d - grain2 - Autograd and GPGPU library for dynamic neural networks in D (Machine Learning / XML)

README

        
# grain2

[![CircleCI](https://circleci.com/gh/ShigekiKarita/grain2.svg?style=svg)](https://circleci.com/gh/ShigekiKarita/grain2)

[![Build status](https://ci.appveyor.com/api/projects/status/xd6ux9097brdbs61?svg=true)](https://ci.appveyor.com/project/ShigekiKarita/grain2)

[![codecov](https://codecov.io/gh/ShigekiKarita/grain2/branch/master/graph/badge.svg)](https://codecov.io/gh/ShigekiKarita/grain2)

![CI](https://github.com/ShigekiKarita/grain2/workflows/CI/badge.svg)

Autograd and GPGPU library for dynamic neural networks in D.

This project is a successor of [grain](https://github.com/ShigekiKarita/grain).

## features

- @nogc support

- heterogeneous device (CPU/CUDA/OpenCL) support

- multiple CUDA device support

- see also [document]([![CircleCI](https://circleci.com/gh/ShigekiKarita/grain2.svg?style=svg)](https://circleci.com/gh/ShigekiKarita/grain2))

## requirements

- [dpp](https://github.com/atilaneves/dpp) master

- CPU backend

  - BLAS library (e.g., OpenBLAS, MKL)

- CUDA backend

  - CUDA header/library v10-

  - CUDNN header/library v7-

- OpenCL backend

  - OpenCL header/library v1.2

## development

- test `dub test grain2:core` and `dub test grain2:cuda`

- generate ddoc `make doc`

## roadmap

- [x] heterogeneous multi-dimentional structure (grain.tensor.Tensor)

- [x] reference count-based memory management

- [x] multiple CUDA devices support

- [ ] core autograd (forward/backward chain)

- [ ] dataset API (MNIST, PTB, etc)

- [ ] more memory manegement policy (garbage collection, memory pool)

- [ ] JIT compilation of computational graph

- [ ] OpenCL/HIP support

## trouble shooting

- Q. My local CUDA/OpenCL libraries are different from files under `grain2/dpp/build`

  - A. Install [dpp](https://github.com/atilaneves/dpp) (commit 821a5d4) and regenerate them by `$ cd dpp; make -j cuda (or opencl)`