A lightweight, dependency-minimal library for tracking machine learning experiments — metrics, hyperparameters, and artifacts — with zero setup required (no external servers, just local files).
Most experiment tracking tools (MLflow, Weights & Biases) require running a server or signing up for a SaaS product. This is a minimal alternative for quick local experimentation: log metrics with one line, compare runs with one function call.
🚧 Under active development.
- Core run/metric logging
- Hyperparameter tracking
- Run comparison utilities
- Simple visualization
- CLI for browsing past runs