Verified:GitHub

weights-and-biases

Log ML experiment metrics, configurations, artifacts, and lineage via W&B SDK.

Updated May 16, 2026

Search Tags:#experiment tracking #wandb #model registry #hyperparameter sweeps #ml experimentation #dashboard visualization #artifacts versioning

Ready to run

npx skills add https://github.com/jaredcroxton/hermes-mirror --skill weights-and-biases-jaredcroxton

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: weights-and-biases
Source: https://github.com/jaredcroxton/hermes-mirror/tree/main/agents/skills/mlops/evaluation/weights-and-biases
Command: npx skills add https://github.com/jaredcroxton/hermes-mirror --skill weights-and-biases-jaredcroxton

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires wandb, and includes references (resource) components.

What problem does it solve?

Weights & Biases (W&B) solves the problem of losing visibility into ML experiments by centralizing metrics, configs, artifacts, and run lineage so you can reliably compare results and reproduce decisions.

Core Features & Use Cases

Experiment Tracking: log scalars, losses, accuracies, media, histograms, and tables while training runs progress.
Hyperparameter Sweeps: run grid/random/Bayesian sweeps and optimize with a single sweep configuration.
Artifacts & Model Registry: version datasets and model checkpoints with lineage, aliases (latest/best/production), and governance-friendly promotion workflows.
Framework Integration: work across common training stacks (e.g., PyTorch, TensorFlow/Keras, HuggingFace Transformers, PyTorch Lightning) using native logging or callbacks.

Quick Start

Initialize W&B by setting up your environment with pip install wandb, then run wandb login and start logging your training metrics from your first wandb.init() session.

Frequently Asked Questions about weights-and-biases

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I track ML experiment metrics during training runs?▼

ML experiment tracking works by logging metrics, configurations, and artifacts to a centralized dashboard, allowing you to reliably compare training runs, evaluate hyperparameter sweeps, and reproduce decisions.

How do I run hyperparameter sweeps with Weights and Biases?▼

You run hyperparameter sweeps by defining a single sweep configuration to execute grid, random, or Bayesian optimization runs that automatically log and evaluate structured metrics across iterations.

Does W&B work with PyTorch and HuggingFace Transformers?▼

W&B works with common training stacks like PyTorch, TensorFlow, Keras, PyTorch Lightning, and HuggingFace Transformers using native logging callbacks to capture metrics and model lineage.

What is the best way to version datasets and model checkpoints?▼

The best way to version datasets and model checkpoints is using artifacts and the model registry to track lineage, assign aliases like latest or production, and enable governance-friendly promotion workflows.

Do I need to install the W&B SDK to log evaluation runs?▼

Yes, you need to install the W&B SDK via pip install wandb, run wandb login, and initialize a session with wandb.init to log metrics, configurations, and artifacts for evaluation runs.