Wandb Experiment Logger

What it does

This skill allows AI agents to log experiment runs directly to Weights & Biases (wandb). It automatically tracks metrics, hyperparameters, and artifacts generated during the training process. This provides a centralized platform for visualizing, comparing, and managing machine learning experiments.

When to use it

• Tracking model training: Use this skill when an agent is responsible for training a machine learning model and you want to monitor its progress in wandb.
• Hyperparameter optimization: Log different hyperparameter configurations and their resulting performance metrics to easily compare models.
• Experiment reproducibility: Automatically capture all relevant information about an experiment, ensuring that it can be reproduced later.
• Artifact management: Store model checkpoints, datasets, or other artifacts generated during training within your wandb project.

Key capabilities

• Automatic metric logging
• Hyperparameter tracking
• Artifact storage (model checkpoints, datasets)
• Integration with Weights & Biases platform

Example prompts

• "Log the current loss and accuracy to wandb."
• "Save this model checkpoint as 'best_model' in wandb."
• "Track these hyperparameters: learning rate = 0.01, batch size = 32."

Tips & gotchas

• Requires a Weights & Biases account and API key to function correctly. Ensure the agent has access to this information.