experiment-log

What problem does it solve?

This Skill prevents lost insights and repeated failures by systematically logging experiment parameters, results, and the exact execution environment, ensuring full reproducibility.

Core Features & Use Cases

• Unique Experiment IDs: Assigns a distinct ID to each logged experiment.
• Environment Snapshotting: Captures git commit, package versions, and system info.
• Reproducibility Tracking: Stores all details needed to rerun an experiment precisely.
• Comparison: Automatically compares new runs against historical data to highlight trends and regressions.
• Use Case: After running a machine learning model training with specific hyperparameters, log the experiment to record the exact learning rate, batch size, and resulting accuracy, along with the git commit and Python version, so you can easily recall or reproduce it later.

Quick Start

Use the experiment-log skill to log a new experiment with parameters learning rate 0.001 and batch size 32, reporting an accuracy of 0.847.