Reference

Encyclopedia Responsibilia.

All reference documentation includes detailed Examples sections. Please scroll to the bottom of any given reference page to see the examples.

Wherever possible, the features provided by the rAI-toolbox are designed to: 1) adhere to common PyTorch APIs – for natural compatibility with popular machine learning frameworks and libraries – and 2) to be agnostic to the domain of application. E.g., our APIs for optimizing data perturbations are compatible with any tensor-based data; we do not assume that users are interested only in computer vision and natural language processing applications.

Contents:

• Optimizers
  • Base Parameter-Transforming Optimizers
  • Optimizers with Normed Gradients
  • Miscellaneous Gradient-Transforming Optimizers
  • Optimizers with Projections Onto Constraint Sets
  • Frank-Wolfe Optimizers
• Perturbations
  • Models
  • Solvers
  • Initializers
• Datasets
  • CIFAR-Based Datasets
  • ImageNet-Based Datasets
• Augmentations
  • Fourier-Basis Augmentations and Utilities
  • AugMix Augmentations and Utilities
• Losses
  • rai_toolbox.losses.jensen_shannon_divergence
• Mushin
  • Workflows
  • PyTorch-Lightning Utilities
• General Utilities
  • PyTorch Utilities