SwanLab is an open-source AI training tracking and visualization tool with cloud/self-hosted options, integrating with PyTorch, Transformers, LLaMA Factory, and more. It provides experiment tracking, visualization, metadata logging, and multi-framework support.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
SwanLab is an AI training analysis and metric visualization platform. It supports cloud and self-hosted use and integrates with PyTorch, Transformers, LLaMA Factory, MMDetection, Ultralytics, and other frameworks. It tracks hyperparameters, logs, and provides comparison and visualization features for experiments.
How it works
The project offers a centralized dashboard to track metrics, hyperparameters, logs, and hardware monitoring data across multiple experiments/projects. It supports visual charts (line charts, media charts, 3D point clouds, molecular data, and custom charts) and online collaboration. It also provides API/SDK integration hooks to log metrics programmatically.
Getting started
Installation:
pip install swanlab
Details for source installation are in the README details:
# 方式一
git clone https://github.com/SwanHubX/SwanLab.git
pip install -e .
# 方式二
pip install git+https://github.com/SwanHubX/SwanLab.git
離线看板拓展安裝:
pip install 'swanlab[dashboard]'
Login and API Key:
swanlab login
Code integration example:
import swanlab
swanlab.init(
project="my-first-ml",
config={'learning-rate': 0.003},
)
for i in range(10):
swanlab.log({"loss": i, "acc": i})
Recent releases
- v0.9.2 (2026-07-29): fix: remove safe.block from upload_log to enable transport-layer retry; refactor: add centralized MIME guesser
- v0.9.1 (2026-07-24): docs updates; fix: skip SIGINT handler registration outside the main thread
- v0.9.0 (2026-07-16): Support for views and multi-view project API with backward compatibility
- v0.9.0-beta.2 (2026-07-14): feat/house v2 adapter
- v0.8.5 (2026-07-14): fix: skip py39 ut error; feat: add api_key prompt mask
Traction
Stars: 4113 Forks: 214 Open issues: 58
Behind the repo
Not provided in the README excerpt.
Caveats
License: Apache-2.0 Created: 2023-11-24 Last push: 2026-08-04