Introduction
PeekML turns your trained machine learning model into a page anyone can try: they change the inputs or upload a picture, press Submit and see the prediction. Put the link on your résumé, LinkedIn or GitHub README.
How it works
- You export your model to the ONNX format, a standard file format that scikit-learn, XGBoost, LightGBM, Keras, PyTorch and others can save to.
- You upload it, optionally with a short
peekml.yamlfile that describes the form: field names, dropdown options, sliders, picture uploads, class names. - PeekML checks it and publishes a page at
/your-username/model-name. - Visitors try it. The model runs in their own browser, so it's fast, and what they type never leaves their device.
What you need
- A GitHub account, to log in and publish models. Trying models needs no account.
- A trained model that takes table-like data (numbers and categories) or a picture (an image model). See Supported models.
- About 10 minutes.
Where to start
- Quick start: the whole process, step by step.
- Export guides for scikit-learn, XGBoost, LightGBM, Keras / TensorFlow, PyTorch and image models.
- peekml.yaml: everything you can put in the file.