TabPFN

⚡ TabPFN: Foundation Model for Tabular Data ⚡

What is TabPFN?

⚡ TabPFN: Foundation Model for Tabular Data ⚡

Overview of TabPFN

What it is

TabPFN is a foundation model for tabular data. It provides a pre-trained transformer that can be used for classification and regression tasks on small to medium-sized tabular datasets without the need for extensive hyperparameter tuning. The repository contains the core implementation for fast, local inference with PyTorch and CUDA support.

Key features

  • Pre-trained transformer for tabular classification and regression.
  • Supports multiple model versions: TabPFN-3.5 (default), TabPFN-3.5-Fast, TabPFN-3, and older versions.
  • GPU acceleration recommended; works on Apple Silicon, Nvidia, and AMD GPUs.
  • CPU inference possible for moderate datasets (up to 5000 samples for newer versions).
  • Includes a TabPFN Client for cloud-based inference without a local GPU.
  • Community extensions available for interpretability, unsupervised tasks, embeddings, and many-class classification.
  • No-code graphical interface (TabPFN UX) for prototyping.

Use cases

TabPFN is suited for tabular classification and regression tasks where the dataset size is moderate (up to 5000 samples for newer versions). It can be used for binary classification, multiclass classification, and regression. The model is designed to work out-of-the-box without extensive hyperparameter tuning, making it useful for rapid prototyping and small to medium-sized datasets.

Getting started

bashpip install tabpfnPython 3.10+ is required. For GPU support on Linux with Nvidia GPUs, the installation includes CUDA support automatically. For AMD GPUs, install PyTorch with ROCm first. For CPU-only, install CPU-only PyTorch first. For complete usage examples, see the repository README.

License and pricing

The code in this repository is licensed under the Apache License 2.0. Model weights are licensed separately: TabPFN-3.5, TabPFN-3.5-Fast, TabPFN-3, TabPFN-2.5, and TabPFN-2.6 weights are released under non-commercial licenses. TabPFN-2 weights are licensed under the Prior Labs License (Apache 2.0 with an additional attribution requirement). An Enterprise Edition with a commercial license is available for production use; contact [email protected] for details.

Based on the project's public README and repository metadata.

Add this badge to your README

Featured on YunoDB Badge styles & help

Maintain this project? Show that it is featured on YunoDB.

Markdown[![Featured on YunoDB](https://yunodb.com/badge/priorlabstabpfn.svg)](https://yunodb.com/tools/priorlabstabpfn?utm_source=github&utm_medium=badge)
HTML<a href="https://yunodb.com/tools/priorlabstabpfn?utm_source=github&utm_medium=badge"><img src="https://yunodb.com/badge/priorlabstabpfn.svg" alt="Featured on YunoDB" height="20"></a>