OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Evaluation & Observability

Weights & Biases

Weights & Biases is a Python-based developer platform and open-source `wandb` toolkit for tracking, visualizing and managing machine-learning pipelines, including datasets, experiments, hyperparameter tuning and models. Its related Weave toolkit covers tracking, debugging, evaluation and monitoring for LLM applications.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Weights & Biases — OpenViST project profile in AI Evaluation & Observability.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when model development needs a persistent, viewable record of runs, configurations and metrics, or when an LLM application needs the separate tracking and evaluation capabilities described for Weave.

Who it is for

  • Machine-learning developers recording training runs and metrics.
  • Teams comparing experiments, hyperparameters and model-development results.
  • Developers adding data versioning and experiment tracking to existing framework-based projects.
  • LLM application teams needing the documented Weave observability workflow.

Documented uses

  • Record configurations, metrics and other information from training runs.
  • Compare machine-learning experiments and hyperparameters.
  • Visualize metric changes during training.
  • Connect datasets, experiments and models across development.
  • Track, debug, evaluate and monitor LLM applications with Weave.
Before you choose

Requirements & limits

The documented quickstart requires the `wandb` Python library, a W&B account and an API key for the primary run-recording workflow. W&B documents managed cloud and self-managed hosting options. The `wandb` repository is public and MIT-licensed.

  • The core quickstart is Python-based and uses wandb.init() and run.log() to record runs.
  • Framework integrations are documented, with JAX and Keras identified among the supported ecosystems in the project profile.
  • Teams can choose managed cloud hosting or deploy W&B Server on documented cloud or on-premises infrastructure.
  • LLM observability is presented through the related Weave toolkit rather than the basic run-tracking example.
  • The repository is active, public and released under the MIT License.

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

Record a machine-learning training experiment

Set up a Python project to initialize a W&B run and log training configuration and metrics.

Before you start

  • A Python project or notebook.
  • The wandb library installed with pip install wandb.
  • A W&B account and API key; the quickstart also documents configuring the key with the wandb login CLI or responding to the first-use prompt.

Steps

  1. Install the wandb library.

  2. Create or access a W&B account and create an API key, storing it securely.

  3. Initialize a run with wandb.init() and provide a project name and configuration values such as epochs and learning rate.

  4. During training, log values such as accuracy and loss with run.log().

  5. Open the W&B workspace to view the recorded metrics and run entry.

Expected outcome

A recorded W&B run containing the supplied configuration and logged training metrics, viewable in the W&B workspace.

Guide limitations

  • The documented quickstart covers recording and viewing a training run; it does not establish that a particular model, framework or deployment will work without further configuration.
  • API keys are shown only once when created and should be stored securely.
Documentation 1 ↗
Evidence and transparency

Sources & review

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