https://github.com/cybershadow/drain

https://github.com/cybershadow/drain

Last synced: about 1 month ago
JSON representation

• Host: GitHub
• URL: https://github.com/cybershadow/drain
• Owner: CyberShadow
• Created: 2021-05-21T08:22:57.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2021-06-01T19:32:22.000Z (almost 5 years ago)
• Last Synced: 2025-11-30T00:10:05.290Z (3 months ago)
• Language: D
• Size: 102 KB
• Stars: 2
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
Drain

=====

Deep-learning and autograd package in D.

Work in progress.

Tenets

======

- No mandatory external dependencies.

  - Utilizing this library in your project should be as simple as including it like any other D library.

  - Once your project is built, it can be deployed anywhere like any other D application, with no additional runtime dependencies.

  - Future additions (e.g. for hardware acceleration) may include *optional* dependencies.

- `nothrow`, `@nogc`.

  - Memory is pre-allocated statically, there is no dynamic memory allocation.

- Utilize the full strength of the D programming language where appropriate.

  - Instead of copying an existing API (such as TensorFlow or Keras), the library API is designed around D features (metaprogramming / templates / template constraints).

- Take advantage of powerful optimizations provided by optimizing D compilers such as LDC.

  - The library may use several layers of abstraction to facilitate expressiveness and code quality. 

  - Debug/DMD builds are much slower that LDC/release builds, however DMD performance is a non-goal.

- Minimize runtime overhead.

  - Indirections are avoided when possible. Instead, memory requirements are pre-computed for entire graphs and are made available to implementations.