Fine-Tuning Orchestration Dashboard (FinetuneOrch)

FinetuneOrch is a web-based dashboard that simplifies the process of fine-tuning large language models by orchestrating a suite of powerful open-source tools. It provides a unified interface to monitor easy-dataset, LLaMA-Factory, and the subsequent model conversion and deployment pipeline with llama.cpp and Ollama.

---

Why This Project?

The workflow for fine-tuning LLMs often involves juggling multiple command-line interfaces, services, and log files. FinetuneOrch streamlines this by providing a single "mission control" panel, enabling you to:

• Simplify Your Workflow: Go from dataset preparation to a ready-to-use GGUF model in Ollama through a guided, observable process.
• Centralize Monitoring: Keep an eye on all the moving parts from one clean web interface.
• Deploy with Confidence: The entire stack is containerized, ensuring a reproducible and easy-to-start environment with a single command.

Key Features

• Unified Service Dashboard: Monitor the status of all containerized services at a glance.
• One-Click Access: Provides direct links to the UIs of easy-dataset, LLaMA-Factory, and TensorBoard.
• Automated Model Pipeline: Orchestrates the post-training workflow:
  1. Merges LoRA adapters into the base model.
  2. Converts the merged model to GGUF format using llama.cpp.
  3. Loads the final GGUF model into your local Ollama instance.
• Training Visualization: Includes an integrated TensorBoard service for real-time monitoring of model training metrics.

Technology Stack

• Backend: Python, FastAPI
• Frontend: Vite, React, TypeScript
• Containerization: Docker, Docker Compose
• Reverse Proxy: Caddy
• Core Fine-Tuning Tools: easy-dataset, LLaMA-Factory
• Model Conversion & Serving: llama.cpp, Ollama

---

Getting Started

Prerequisites

• Docker and Docker Compose
• An NVIDIA GPU with the appropriate drivers installed.

Configuration

1. Clone the Repository:

   git clone <repository-url>
   cd <repository-name>

2. Create and Configure the Environment File: Copy the example environment file to a new .env file. This file is used to configure all the services.

   cp .env.example .env
   ```    Open the `.env` file and customize the variables as needed.
   

Environment Variables

The following variables can be configured in the .env file:

• HF_TOKEN: (Required) Your Hugging Face API token. This is necessary for downloading models.
• OLLAMA_API_URL: The internal URL for the Ollama API, used by the backend. Defaults to http://ollama:11434.
• VITE_LLAMA_FACTORY_URL: The public-facing URL for the LLaMA Factory service. Defaults to http://localhost:7002.
• VITE_EASY_DATASET_URL: The public-facing URL for the Easy Dataset service. Defaults to http://localhost:7001.
• VITE_TENSORBOARD_URL: The public-facing URL for the TensorBoard service. Defaults to http://localhost:7003.
• LLAMA_FACTORY_CONTAINER: The name of the LLaMA Factory Docker container. Defaults to finetune-llama-factory.
• LLAMA_CPP_CONTAINER: The name of the llama.cpp Docker container. Defaults to finetune-llama-cpp.
• OLLAMA_CONTAINER: The name of the Ollama Docker container. Defaults to ollama.

Running the Application

1. Build and Start the Services:

   docker compose up --build -d

2. Access the Dashboard: The main orchestration dashboard will be available at http://localhost:3000. From the dashboard, you can access the UIs for all other services.

Usage

(Detailed usage instructions follow, as previously defined)

Acknowledgements

This project would not be possible without the incredible work of the open-source community. We would like to extend our sincere gratitude to the developers and researchers behind the following projects:

• LLaMA Factory: A powerful and flexible framework for fine-tuning large language models.

  @inproceedings{zheng2024llamafactory,
    title={LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models},
    author={Yaowei Zheng and Richong Zhang and Junhao Zhang and Yanhan Ye and Zheyan Luo and Zhangchi Feng and Yongqiang Ma},
    booktitle={Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
    address={Bangkok, Thailand},
    publisher={Association for Computational Linguistics},
    year={2024},
    url={http://arxiv.org/abs/2403.13372}
  }

• Easy Dataset: An extensible framework for synthesizing fine-tuning data from unstructured documents.

  @misc{miao2025easydataset,
    title={Easy Dataset: A Unified and Extensible Framework for Synthesizing LLM Fine-Tuning Data from Unstructured Documents},
    author={Ziyang Miao and Qiyu Sun and Jingyuan Wang and Yuchen Gong and Yaowei Zheng and Shiqi Li and Richong Zhang},
    year={2025},
    eprint={2507.04009},
    archivePrefix={arXiv},
    primaryClass={cs.CL},
    url={https://arxiv.org/abs/2507.04009}
  }

• llama.cpp: A high-performance C++ library for running language models.

• Ollama: A tool for running and managing large language models locally.

• FastAPI: A modern, fast (high-performance) web framework for building APIs with Python.

• React: A JavaScript library for building user interfaces.

• Vite: A fast and lightweight build tool for modern web development.

• Docker: A platform for developing, shipping, and running applications in containers.

License

This project is licensed under the MIT License. See the LICENSE file for details.

Contributing

Contributions are welcome! Please see our Contributing Guide for more details on how to get started.

Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.