Kumo Tabular

Kumo Tabular is NVIDIA's pretrained tabular foundation model for classification and regression.

Getting Started

Install structured-data-models for inference:

pip install structured-data-models

Use labeled examples as context to predict class probabilities for new data:

import torch
from sklearn.datasets import load_breast_cancer

import sdm

df = load_breast_cancer(as_frame=True).frame
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

table = sdm.TableTensor.from_pandas(
    df=df,
    stypes=sdm.infer_stypes(df, overrides={"target": "categorical"}),
    device=device,
)
model = sdm.models.KumoTabular(task="classification", device=device)

with torch.amp.autocast(
    device.type,
    dtype=torch.float16,
    enabled=device.type == "cuda",
):
    probs = model(
        x_context=table[:300].drop_columns("target"),
        y_context=table[:300, "target"],
        x_query=table[300:].drop_columns("target"),
        num_estimators=8,
    )

print(probs)

To learn more, visit structured-data-models.

License

Kumo Tabular weights are released under OpenMDW 1.1.