Show HN: Boost TabPFN's performance by fine-tuning on proprietary data

2 points

a year ago

TabPFN is a transformer-based foundation model for tabular datasets. Published in Nature earlier this year, it achieves SOTA performance on classification and regression tasks without any per-task training or hyperparameter tuning.

With the new tabpfn v2.1 release, fine-tuning is now supported. Similar to LLMs, you can fine-tune tabular foundation models on your proprietary data to boost your performance.

A community member wrote a great walkthrough here: https://medium.com/@iivalchev/how-to-fine-tune-tabpfn-on-you...

You can also check out the example files directly at these links:

> Fine-tune classifier: https://github.com/PriorLabs/TabPFN/blob/main/examples/finet...

> Fine-tune regressor: https://github.com/PriorLabs/TabPFN/blob/main/examples/finet...

If you want to chat or see what others are experimenting with, there’s a Discord too: https://discord.com/invite/VJRuU3bSxt

Would love to hear what you think when you try it!