Structured Data Models
A tabular foundation model library introduced by NVIDIA that wraps multiple pretrained tabular models, provides a unified PyTorch interface, and supports in-context learning predictions with no training.
Timeline (last 90 days)
NVIDIA has launched the Structured Data Models library, which wraps multiple pretrained tabular foundation models in a single PyTorch interface and supports zero-training prediction.
The Structured Data Models library currently wraps three types of models: TabICLv2 and KumoTabular for single tables, and KumoRelational for relational databases.
A unified PyTorch interface means teams do not need to maintain separate dependencies and calling conventions for each model; switching models requires only changing configuration rather than rewriting code, which can lower trial-and-error costs.
All Facts (3)
NVIDIA has launched the Structured Data Models library, which wraps multiple pretrained tabular foundation models in a single PyTorch interface and supports zero-training prediction.
50%UnverifiedThe Structured Data Models library currently wraps three types of models: TabICLv2 and KumoTabular for single tables, and KumoRelational for relational databases.
50%UnverifiedA unified PyTorch interface means teams do not need to maintain separate dependencies and calling conventions for each model; switching models requires only changing configuration rather than rewriting code, which can lower trial-and-error costs.
50%