OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Developer Tools & SDKs

Betty

Betty is an open-source Python library for automatic differentiation in generalized meta-learning and multilevel optimization with PyTorch. It represents each optimization level as a Problem and uses an Engine to manage dependencies, gradient calculations and execution order. Its documented applications include hyperparameter optimization, neural architecture search, data reweighting, domain adaptation and model-agno

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Betty — OpenViST project profile in AI Developer Tools & SDKs.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use Betty when an experiment has linked optimization levels and benefits from a shared way to define problems, dependencies and execution. Its examples offer concrete starting points for research workflows such as hyperparameter optimization and implicit meta-learning.

Who it is for

  • Machine-learning researchers implementing bilevel or multilevel optimization experiments
  • PyTorch developers building meta-learning or hyperparameter-optimization workflows
  • Researchers prototyping neural architecture search or data-reweighting methods

Documented uses

  • Define optimization levels with model components, loss functions and configurations.
  • Express hierarchical dependencies between optimization problems.
  • Build reference-style experiments for hyperparameter optimization, architecture search, data reweighting or meta-learning.
  • Run a multilevel program through Betty’s Engine and include a global validation stage.
Before you choose

Requirements & limits

The repository provides the library as the `betty-ml` Python package and identifies its license as Apache 2.0.

  • The library is built around PyTorch and Python; the user supplies the task-specific models, data, losses and optimizers.
  • The Problem and Engine abstractions are relevant when an experiment has hierarchical optimization dependencies.
  • The README identifies Distributed Data Parallel and ZeRO support; Fully Sharded Data Parallel is labeled experimental.
  • The repository includes application examples and tutorials for extending the basic structure.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up a Betty multilevel optimization program

Install Betty and define the problem and dependency structure needed to call its optimization engine.

Before you start

  • Python environment for the betty-ml package
  • PyTorch problem components, including modules, optimizers and data loaders
  • Loss functions and problem configurations for each optimization level

Steps

  1. Install the package with the command below.

    Documented command
    pip install betty-ml

  2. Define each optimization level as a Problem subclass, including its training-step loss and required components.

  3. Create an Engine configuration and specify upper-to-lower and lower-to-upper dependencies between the problems.

  4. Instantiate the engine with the problems, dependencies and configuration, then call engine.run().

Expected outcome

A Betty engine is configured to execute the defined multilevel optimization program.

Guide limitations

  • The developer must supply the model, data, losses, problem configurations and dependency structure; installation alone does not create an experiment.
  • The example documents the programming pattern, not a verified run or a guarantee that a particular project configuration will work.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
14 October 2026
Original website title
leopard-ai/betty · Official homepage title · retained in its source language
How we review sourcesSuggest a correction