PyCoder is an AI-powered chatbot built on a QLoRA-quantized DeepSeek-Coder-1B model fine-tuned using LoRA-based SFT. It generates high-quality Python code and provides clear, human-readable explanations, all while remaining efficient on low-memory GPU setups.

🏗️ Model Architecture & Training

• Base Model:

  deepseek-ai/deepseek-coder-1.3b-instruct
  

• Quantization: QLoRA (4-bit)

• Fine-Tuning Method: Supervised Fine-Tuning (SFT)

• Adapters: LoRA

• Dataset:

  iamtarun/python_code_instructions_18k_alpaca
  
  teknium/OpenHermes-2.5
  

• Primary Task: Python code generation & explanation

• Fine-tuned model

    Hhsjsnns/PyCoder-QLoRA-v1
  

---

⚙️ Setup Instructions

1️⃣ Clone the repository

git clone https://github.com/BhagwatiOracle/PyCoder.git

cd PyCoder

2️⃣ Create a virtual environment

python -m venv venv

source venv/bin/activate     # Windows: venv\Scripts\activate

3️⃣ Install dependencies

pip install -r requirements.txt

4️⃣ Run the app

python app.py

---

🎬Demo

📌 Click to expand Demo