Welcome to DATE, a LLM-based tabular data generator for heterogenous data. We present our Source Code and Appendix for users.
For detailed proofs and external experiments in the original paper, please read our appendix.
To obtain the public data used in our experiment, please refer to the IEEE Dataport. We also public it on Huggingface as Heterogenous_Synthesis_Benchmark
For running Module 1: DGR-based Prompt Designing, please refer to ModelShare_with_DSR_final.py.
For running Module 2: Distribution-Specific Generation, please refer to dataGeneration.py.
Files with the test_ prefix are test files. For example, (test_forward)[test_forward.py] is a testing file based on forward-greedy selection.