C2RustRep: Dependency-Guided Repository-Level C-to-Rust Translation with Reinforcement Alignment

C2RustRep is a framework for automatic repository-level migration from C to Rust. It leverages fine-grained dependency analysis and reinforcement learning to enhance translation accuracy, safety, and idiomaticity.

📁 Directory Structure

• dataset/: Contains datasets for evaluation, multi-task fine-tuning, and reinforcement learning (in JSONL format).
• inference/: Scripts for model inference and evaluation.
• train/: Training scripts for multi-task fine-tuning and reward-based reinforcement learning.

🚀 Quick Start

1. Prepare Data
   Place the source C code and relevant datasets under the dataset/ directory.

2. Train the Model
   Use train/lora_mft.py for multi-task fine-tuning, or train/lora_grpo.py for reinforcement learning with GRPO.

3. Run Inference & Evaluation
   Use inference/main.py to perform translation and evaluate results.

⚙️ Notes

• This project emphasizes the automation workflow and evaluation of repository-level C-to-Rust migration.
• For implementation details, please refer to the corresponding script files.
• Requires basic knowledge of Python, Git, and deep learning frameworks.