Skip to content

README.md

Latest commit

 

History

History
90 lines (59 loc) · 1.91 KB

README.md

File metadata and controls

90 lines (59 loc) · 1.91 KB
Raw

InlineCoder:

This repository contains the code and data for our * paper. It provides scripts for data preprocessing, model evaluation, and ablation studies on the DevEval benchmark.

framework

Dataset Preparation

Clone the DevEval dataset repository:

cd references
git clone git@github.com:seketeam/DevEval.git

Alternatively, download the original repositories from HuggingFace.
After downloading, extract the contents and place them in:
references/DevEval/Source_Code

Environment Setup

Create and activate the Conda environment:

conda create -n inlineCoder python=3.8 -y
conda activate inlineCoder

Add the project root to your PYTHONPATH:

export PYTHONPATH="/path/to/InlineCoder:$PYTHONPATH"

Install required dependencies:

pip install -r requirements.txt

Workflow

1. Data Preprocessing

Run the following script to preprocess the DevEval dataset:

python inline_coder/Preprocess/dev_eval_preprocess.py

2. API Connectivity Test

Configure your API key and base URL in configs/CONFIGS.json:

{
    "api_key": "****your_API_key****",
    "base_url": "https://your_URL"
}

Test API connectivity:

python inline_coder/Models/services.py

3. InlineCoder Evaluation

Activate the environment and run the main evaluation script:

conda activate inlineCoder
python inline_coder/inline_coder_gen.py

Results will be automatically evaluated using BLEU and other metrics.
Note: Path issues may occur; please ensure all paths are correctly configured.

4. Ablation Study

Run ablation experiments with:

bash inline_coder/Ablation/scripts/ablation_no_downstream.sh

Results will be evaluated automatically.

If you have any questions or encounter issues, please refer to the documentation or contact