TabuLa-8B (Tabular Llama-8B) is an 8 billion parameter foundation model developed by mlfoundations, built upon Meta Llama 3, specifically designed for prediction tasks on tabular data. This model excels at classification and binned regression on structured datasets. It was trained using the rtfm framework and the T4 dataset, offering specialized capabilities for tabular data analysis.
Loading preview...
TabuLa-8B: Foundation Model for Tabular Data
TabuLa-8B is an 8 billion parameter model developed by mlfoundations, based on Meta Llama 3, and is specifically engineered for tasks involving tabular data. Described in the paper "Large Scale Transfer Learning for Tabular Data via Language Modeling," this model focuses on prediction, including classification and binned regression.
Key Capabilities
- Tabular Data Prediction: Optimized for classification and binned regression on structured datasets.
- Llama 3 Foundation: Built upon the robust Llama 3 8B architecture.
- Specialized Training: Trained with the
rtfmframework on the extensiveT4 dataset. - Easy Integration: Compatible with
transformersfor straightforward loading and inference.
When to Use TabuLa-8B
- Tabular Prediction Tasks: Ideal for applications requiring predictions on structured data, such as financial forecasting, customer churn prediction, or medical diagnosis from tabular records.
- Leveraging Llama 3 Ecosystem: Benefits from the underlying Llama 3 architecture and its associated community support.
- Research in Tabular ML: Useful for researchers exploring large-scale transfer learning and language modeling approaches for tabular data.
TabuLa-8B is released under the Llama 3 license, requiring adherence to its Community License Agreement and Acceptable Use Policy.