[ANN] Arhat: an experimental deep learning framework implemented in Go

[fraga...@gmail.com]

I would like to announce the first public release of Arhat, an experimental deep learning framework implemented in Go.

Unlike most mainstream frameworks that perform training and inference computations directly, Arhat translates neural network descriptions into standalone lean executable platform-specific code. This approach allows direct embedding of the generated code into user applications and does not require deployment of sophisticated machine learning software stacks on the target platforms.

We at Fragata use Arhat as a research platform for our larger ongoing project aiming at construction of a production-grade machine learning framework optimized for use on embedded platforms and massively parallel supercomputers.

The project repository is located here:

https://github.com/fragata-ai/arhat

[Tong Sun]

On Monday, December 2, 2019 at 12:56:14 PM UTC-5, Alexey wrote:

I would like to announce the first public release of Arhat, an experimental deep learning framework implemented in Go.

Unlike most mainstream frameworks that perform training and inference computations directly, Arhat translates neural network descriptions into standalone lean executable platform-specific code.

That's a super good idea. Just curious, what type of neural network descriptions input does it take?