Crate dfdx

source · []
Expand description

Ergonomics & safety focused deep learning in Rust. Main features include:

  1. Tensor library, complete with const generic shapes, activation functions, and more.
  2. Safe & Easy to use neural network building blocks.
  3. Standard deep learning optimizers such as Sgd and Adam.
  4. Reverse mode auto differentiation[1] implementation.

Modules

Collection of traits to describe Nd arrays.

A collection of data utility classes such as one_hot_encode() and SubsetIterator.

Provides implementations for modifying Nd arrays on the Cpu.

Implementations of GradientTape and generic Nd array containers via Gradients.

Standard loss functions such as mse_loss(), cross_entropy_with_logits_loss(), and more.

High level neural network building blocks such as Linear, activations, and tuples as Modules. Also includes .save() & .load() for all Modules.

Provides some generic functions to load & save Nd arrays in the .npy format. See load() and save()

Optimizers such as Sgd and Adam that can optimize neural networks.

Contains all public exports.

The struct definitions for all TensorXD, Tensor trait, and more.

Implementations of all ops for tensors including activations like relu(), binary operations like matmul(), and more.