Handbook
Welcome to the RunForge Desktop handbook. This is the complete guide to installing, using, and understanding RunForge Desktop — a Windows-native ML experiment tracker that runs entirely on your machine.
What’s inside
Section titled “What’s inside”- Getting Started — Install the app (MSIX or from source) and launch your first training run
- Features — Training creation, hyperparameter sweeps, live monitoring, run comparison, export, settings, crash recovery, and diagnostics
- Gauntlets — The reliability test suite: parallel enforcement, pause/resume, cancel, crash recovery, GPU scheduling
- Reference — NuGet package, system requirements, project structure, artifact format, execution queue, and security scope
- Beginners — New to ML experiment tracking? Start here for a step-by-step introduction
Core principles
Section titled “Core principles”RunForge Desktop is built on four principles:
- Local-first — all training runs on your machine, no cloud required
- Transparent — live logs, real-time metrics, full process control
- Simple — one workspace, clear presets, no config files to manage
- Auditable — all run artifacts saved to disk for inspection and reproducibility