This repository contains the implementation of the Planning-after-Trial (PaT) framework, designed for robust and efficient code generation.
The majority of the code in this repository is based on the FunCoder project. We extend our deepest gratitude to the original authors for their foundational work, which served as a crucial basis for our research.
To run experiments, you need to set up
- Environment
conda create -y -n PaT python 3.10 conda activate PaT python -m pip install -r requirements.txt
- Datasets
python -m PaT.eval download-datasets
- Configuration
python -m vllm.entrypoints.openai.api_server --model /path/to/your/model/Qwen3-8B --dtype float16 --api-key token-qwen3_8 --port 28110
python -m Pat.eval draft --results-dir /your/experiment/dir/
python -m Pat.eval judge --results-dir /your/experiment/dir/