Crate rustyasg 

RustyASG: Graph-based Deep Learning Engine in Rust

RustyASG is a modern experimental deep learning framework written in Rust. Its key feature is an architecture built around the Abstract Semantic Graph (ASG).

Usage Example

use std::rc::Rc;
use std::cell::RefCell;
use rustyasg::tensor::{GraphContext, Tensor};
use rustyasg::losses::mse_loss;

// 1. Create graph context
let context = Rc::new(RefCell::new(GraphContext::new()));

// 2. Define symbolic inputs
let input = Tensor::new_input(&context, "input");
let expected = Tensor::new_input(&context, "expected");

// 3. Build computation graph
let prediction = input.relu(); // Just an example operation
let loss = mse_loss(&prediction, &expected);

// Graph is ready for analysis, differentiation, and execution on a backend!

Modules

analysis

Graph Analysis Module

asg

Module defining the core of the Abstract Semantic Graph (ASG).

autograd

Automatic Differentiation Module

data

Data Loading Module

losses

Module containing implementations of loss functions in graph paradigm.

metrics

Metrics module for evaluating model quality.

nn

Neural Network Layers Module

optimizers

Optimizers Module

runtime

Runtime Execution Backends

serialization

Model serialization and deserialization.

tensor

Module defining Tensor and GraphContext.