Practical overview
When it fits
Use GReaT when a Python workflow needs Transformer-based synthetic tabular samples, conditional generation or imputation from a trained model.
Who it is for
- Python data practitioners creating synthetic tables for prototyping, experimentation or testing.
- Teams exploring model-based generation, conditional sampling or missing-value imputation for tabular data.
Documented uses
- Generate additional tabular rows from an existing dataset.
- Fill missing values in a DataFrame using a model trained on the relevant data distribution.
- Generate rows that meet numeric or categorical conditions.
- Save a model checkpoint and load it again for later use.
Requirements & limits
The repository is public and identifies the project as MIT-licensed. Installation instructions require Python 3.9 or later and use pip.
- The documented workflow fits a model on a table before sampling or imputing.
- Conditional sampling supports numeric comparisons and categorical equality or inequality conditions.
- Guided sampling can help with complex feature sets but may be slower than default sampling.
- For small datasets or tables with many features, the README describes tuning batch size, epochs and floating-point precision.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install GReaT and generate sample rows
Set up the library and follow its documented workflow for generating synthetic rows from a table.
Before you start
- Python 3.9 or later
- A tabular dataset available as a pandas DataFrame
Steps
Install the package with pip install be-great.
Import GReaT and prepare a DataFrame.
Create a GReaT model with a pretrained language model, then call fit with the DataFrame.
Call sample with the number of rows to generate.
Expected outcome
A synthetic DataFrame containing the requested number of sampled rows.
Guide limitations
- The quickstart uses a pretrained model and demonstrates the California Housing dataset; the README does not specify hardware requirements.
Impute missing values with a trained model
Use a trained GReaT model to fill missing values in a DataFrame.
Before you start
- A trained GReaT model for the data distribution
- A pandas DataFrame with missing entries represented as NaN
Steps
Prepare the DataFrame to be imputed and represent missing values as NaN.
Call the trained model's impute method with the DataFrame; the example also supplies max_length.
Expected outcome
An imputed DataFrame returned by the model.
Guide limitations
- The documented imputation interface requires a trained model.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- tabularis-ai/be_great · Official homepage title · retained in its source language