NightFlow
Manage, run, and analyze deep learning experiments from a single native desktop app. Private by default — your data never leaves your machine.
Auto-detecting version...
Manage, run, and analyze deep learning experiments from a single native desktop app. Private by default — your data never leaves your machine.
Auto-detecting version...
Everything you need to manage your ML workflow — no cloud required.
Create and organize ML projects with dataset, model, and training configurations in one place.
View runs, metrics, and training history in a structured, sortable table with full run detail views.
Interactive loss curves, accuracy plots, and custom metric charts rendered in real time.
Built-in GradCAM, attention maps, and interpretability tools to understand what your model learns.
Visualize neural network architectures directly inside the app — no browser needed.
Full PTY terminal with SSH support for remote training servers — no context switching.
No telemetry, no tracking, no cloud. All your data is stored locally on your machine.
Native performance with a tiny footprint. No Electron, no bloated runtimes.
Runs natively on macOS (Intel & Apple Silicon), Windows, and Linux.
Browse your dataset visually with thumbnails, class distribution charts, and automatic train/val/test split detection.
Compare multiple training runs side-by-side with overlaid metric charts to identify winning configurations.
Queue multiple training runs for sequential execution. NightFlow manages the pipeline while you focus on research.
Push trained models directly to Hugging Face Hub with auto-generated model cards, metadata, and security-stripped checkpoints.
A clean, focused workspace for your experiments.
NightFlow is Apache 2.0 licensed. Star on GitHub, report issues, or contribute.
Star on GitHub Buy Me a Coffee