Code Generation using GPT & CodeSearchNet
Overview
This project focuses on developing a code generation model using a GPT-like architecture trained on the CodeSearchNet dataset. The model is built with Python and PyTorch, leveraging the transformers library for NLP tasks. The goal is to generate code snippets based on natural language descriptions.
Features
Uses the CodeSearchNet dataset for training
Implements a GPT-style language model
**Supports fine-tuning for specific programming languages
Implements tokenization, training, and evaluation pipelines
**Future support for deployment via API (FastAPI/Flask)
**: Tentative
Setup Instructions
Prerequisites
Ensure you have Python and pip installed. Then, install the required dependencies:
pip install torch torchvision torchaudio transformers datasets tokenizers sentencepiece tqdm
Clone the Repository
git clone cd
Running the Project
Load and preprocess the dataset
from datasets import load_dataset dataset = load_dataset("code_search_net", "python")
Train the model
python train.py
Evaluate the model
python evaluate.py
Roadmap
Contributing
Currently, this project is closed for contribution.
License
This project is open-source and available under the MIT License.
Acknowledgments
Hugging Face's transformers library
CodeSearchNet dataset
PyTorch community