SwanHubX/SwanLab AI training tracking and visualization platform

SwanLab is an open-source, modern-design AI training tracking and visualization tool that supports both cloud and self-hosted deployments, seamlessly integrating with over 30 frameworks including PyTorch, Transformers, LLaMA Factory, Swift, Ultralytics, veRL, MMEngine, and Keras to help researchers rapidly log, compare, and share their experiments. Licensed under Apache-2.0 and boasting 1.5k+ stars, SwanLab benefits from an active community and regular updates.

Official website：https://swanlab.cn/
Source code：https://github.com/SwanHubX/SwanLab
Documentation：https://docs.swanlab.cn/

Here is a list of its core features:

1. 📊 Experimental metrics and hyperparameter tracking : Embed minimal code into your machine learning pipeline to track key training metrics

• ☁️ Supports cloud usage (similar to Weights & Biases), and you can view the training progress anytime, anywhere. How to watch the experiment on your mobile phone

• 📝 Supports hyperparameter recording , indicator summary , and table analysis

• 🌸Visualize the training process : Visualize the experimental tracking data through the UI interface, so that the trainer can intuitively see the results of each step of the experiment, analyze the indicator trends, and determine which changes lead to the improvement of the model effect, thereby improving the overall efficiency of model iteration.

• Supported metadata types : scalar metrics, images, audio, text, 3D point clouds, biochemical molecules…

• Supported chart types : line chart, media chart (image, audio, text), 3D point cloud, biochemical molecule…

• Background automatic recording : logging, hardware environment, Git repository, Python environment, Python library list, project running directory

2. ⚡️ Comprehensive framework integration : 30+ frameworks including PyTorch, 🤗HuggingFace Transformers, PyTorch Lightning, 🦙LLaMA Factory, MMDetection, Ultralytics, PaddleDetetion, LightGBM, XGBoost, Keras, Tensorboard, Weights&Biases, OpenAI, Swift, XTuner, Stable Baseline3, Hydra

3. 💻 Hardware monitoring : Supports real-time recording and monitoring of system-level hardware indicators of CPU, NPU ( Ascend ), GPU ( Nvidia ), MLU ( Cambricon ), XLU ( Kunlunxin ) and memory

4. 📦 Experiment Management : Through a centralized dashboard designed specifically for training scenarios, you can quickly manage multiple projects and experiments with an overall view.

5. 🆚 Compare results : Compare the hyperparameters and results of different experiments through online tables and comparison charts to explore iterative inspiration

6. 👥 Online collaboration : You can conduct collaborative training with your team, and support real-time synchronization of experiments under one project. You can view the team’s training records online and express your opinions and suggestions based on the results.

7. ✉️ Share results : Copy and send a persistent URL to share each experiment, easily send to a partner, or embed in an online notebook

8. 💻 Support self-hosting : Support offline environment use. The self-hosted community version can also view the dashboard and manage experiments. Use guides

9. 🔌 Plugin expansion : Supports expanding the usage scenarios of SwanLab through plugins, such as Feishu notification , Slack notification , CSV logger , etc.