h2oai/tabh2o-v1.1

TabH2O v1.1

Tabular foundation model by H2O.ai for classification and regression. It predicts the rows of a test table from the labeled rows of a training table in one forward pass (in-context learning); nothing is trained or tuned.

Usage

pip install h2ofm

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split

from h2ofm import TabH2O

X, y = load_breast_cancer(return_X_y=True, as_frame=True)
X_train, X_test, y_train, y_test = train_test_split(X, y, random_state=0)

model = TabH2O.from_pretrained("h2oai/tabh2o-v1.1")
result = model.classify(X_train, y_train, X_test)

result.probabilities  # (n_test, 2), columns follow result.classes
result.labels  # (n_test,), the most probable class

model.regress(X_train, y_train, X_test) returns mean, median and quantiles (default levels 0.1 through 0.9) for a numeric target. See the h2ofm documentation for inputs and options.

Model Details

Parameters	29M
Tasks	Classification, regression
Classes	10 natively; more through a class hierarchy
Format	config.json + model.safetensors (float32)

License

Non-commercial. See LICENSE.