worldbank/REaLTabFormer
A suite of auto-regressive and Seq2Seq (sequence-to-sequence) transformer models for tabular and relational synthetic data generation.
Implements dual architecture using GPT-2 for independent tabular data and Seq2Seq for modeling parent-child table relationships with foreign key constraints. Features built-in validators (e.g., GeoValidator) for filtering invalid synthetic samples and automatic early stopping based on distribution matching between synthetic and real data. Available as a pip-installable PyPI package with native pandas DataFrame integration for seamless workflow incorporation.
244 stars.
Stars
244
Forks
29
Language
Jupyter Notebook
License
MIT
Category
Last pushed
Jan 04, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/transformers/worldbank/REaLTabFormer"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Related models
MagedSaeed/generate-sequences
A python package made to generate sequences (greedy and beam-search) from Pytorch (not...
tlkh/t2t-tuner
Convenient Text-to-Text Training for Transformers
NohTow/PPL-MCTS
Repository for the code of the "PPL-MCTS: Constrained Textual Generation Through...
styfeng/TinyDialogues
Code & data for the EMNLP 2024 paper: Is Child-Directed Speech Effective Training Data for...
saltudelft/codefill
Contains the code and data for our #ICSE2022 paper titled as "CodeFill: Multi-token Code...