samuelinferences/TabPFN · Apply for community grant: Academic project

Hi there,

We build a model that competes with the best other models for tabular classification. The catch: while other models need to perform training and hyperparameter tuning per dataset, our model just does a simple forward pass on a dataset.

Our model is a transformer that accepts the full training set as well as the inputs for the test set and returns all predictions for the test set at once.

It runs much faster on GPU, i.e. going down to as little as 100ms to train and predict. We would love to be able to show this to the world. We cannot afford to pay for a GPU per hour here sadly, though.

See our blog post for more information: https://www.automl.org/tabpfn-a-transformer-that-solves-small-tabular-classification-problems-in-a-second/

Thanks,
Sam