FineTuneStudio 🚀

The No-Code AI Model Fine-Tuning, Comparison & Deployment Platform.

FineTuneStudio is an all-in-one open-source studio designed to democratize AI fine-tuning. It allows developers and researchers to upload datasets, clean them with AI-powered suggestions, fine-tune state-of-the-art models (Llama, Mistral, Gemma, Phi) using LoRA/PEFT, and deploy them directly to Hugging Face—all through a beautiful, premium glassmorphism interface.

✨ Features

• 📊 Intelligent Data Studio: Upload CSV, JSON, Audio, or Video. Get AI-powered cleaning suggestions (duplicates, missing values, etc.).
• ⚡ Hardware Orchestrator: Run training on your local machine, Google Colab, or Kaggle. (RunPod & AWS support coming soon).
• 🧠 Multi-Model Support: Fine-tune Llama 3.2, Mistral 7B, Phi-2, Gemma 2B, and Liquid Foundation Models (LFM).
• ⚖️ Side-by-Side Comparison: Evaluate "Before vs After" performance with real-time metrics and Knowledge Transfer Radar charts.
• 🚢 1-Click Deployment: Push your optimized models directly to Hugging Face Hub.
• 🛡️ Secure Credential Management: Store HF and Kaggle keys securely with integration for Colab Secrets.

🛠️ Tech Stack

• Frontend: React.js, Vite, Recharts (Visualizations), Lucide React (Icons).
• Backend: FastAPI (Python), MongoDB (Metadata storage), Pandas (Data processing).
• AI/ML: Hugging Face Transformers, PEFT/LoRA, BitsAndBytes (4-bit quantization).

🚀 Getting Started

Prerequisites

• Python 3.10+
• Node.js 18+
• MongoDB (Running locally or on Atlas)

1. Clone the Repository

git clone https://github.com/maleshkumar/FineTuneStudio.git
cd FineTuneStudio

2. Backend Setup

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt
uvicorn app.main:app --reload --port 8000

3. Frontend Setup

cd ../frontend
npm install
npm run dev

Visit http://localhost:5173 to start fine-tuning!

📖 Roadmap

• WebSocket migration for real-time training logs.
• Multi-Cloud provider integration (RunPod/AWS).
• Shared Community Library for datasets.
• Export to ONNX/TensorRT for edge deployment.

🤝 Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

📄 License

FineTuneStudio is released under the MIT License.

---

Built with ❤️ by Malesh Kumar