SwanLab
An AI experiment tracking platform with training charts, run comparisons and offline dashboards, available self-hosted or in the cloud.
SwanLab is an open-source AI experiment tracking platform for researchers and model training teams. It brings training results into one place so you can spot problems and compare runs before choosing your next experiment.
- Run comparison: Track metrics and hyperparameters, search and filter experiment tables, and compare results against a baseline.
- Training visualization: Inspect metric curves alongside images, audio, video and text, including Markdown-rendered LLM outputs and custom charts.
- Logs and resource monitoring: Capture terminal logs, Python and Git details, and hardware usage during training, with monitoring for CPUs, NVIDIA and AMD GPUs, and Ascend NPUs.
- Framework integrations: Track work in PyTorch, Hugging Face Transformers, LLaMA Factory, Unsloth, veRL, Keras and Ultralytics without replacing your training framework.
- Cloud collaboration: Share experiment links and organize team training projects through the hosted service, with browser access to progress on mobile devices.
You can self-host SwanLab with Docker or Kubernetes, and the community edition supports offline dashboard viewing. The hosted service stores and synchronizes experiments in the cloud for remote access and sharing. The project uses the Apache 2.0 license.
Logotyper
