Thanks to visit codestin.com
Credit goes to github.com

Skip to content

riyanshibohra/TuneKit

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TuneKit

The fine-tuning workflow you wish existed.

License: MIT Python 3.10+ FastAPI Powered by Unsloth Free to Use

Try it now · Report Bug · Request Feature

P.S. TuneKit trending at #19 on Product Hunt (Day of Launch) with 88 followers & 50+ Upvotes - check it out!

demo.mp4

See it in action


The Problem

Fine-tuning LLMs is powerful but painful. You need to:

  • Set up CUDA, PyTorch, and a dozen dependencies
  • Rent expensive GPUs or wait hours on slow hardware
  • Write training scripts, manage hyperparameters, handle OOM errors
  • Figure out how to export and deploy your model

Most developers give up before they start.

The Solution

TuneKit eliminates all of that. Upload your data, answer two questions, and get a ready-to-run Colab notebook. Click "Run All" and your fine-tuned model is ready in ~15 minutes.

Your Data → TuneKit → Colab Notebook → Fine-tuned Model

No local GPU. No dependencies. No cost (uses Google's free T4).


How It Works

Step What You Do What TuneKit Does
1 Upload JSONL file Validates format, analyzes patterns
2 Answer 2 questions Recommends optimal model + hyperparameters
3 Click "Get Notebook" Generates pre-configured Colab notebook
4 Hit "Run All" in Colab Trains on free T4 GPU (~15 min)
5 Download your model Export as LoRA, GGUF, or merged weights

Why TuneKit?

Without TuneKit

  • Hours of environment setup
  • $50-200 in GPU costs
  • Hundreds of lines of code
  • Trial and error with hyperparameters
  • OOM errors and debugging

With TuneKit

  • Zero setup
  • $0 (free Colab GPU)
  • Zero code to write
  • AI-optimized configuration
  • Just works

Powered by Unsloth - 2x faster training, 70% less VRAM.


Supported Models

Model Parameters Best For
Phi-4 Mini 3.8B Classification, extraction, structured output
Llama 3.2 1B, 3B Q&A, conversational AI, context tracking
Mistral 7B 7B Long-form generation, complex reasoning
Qwen 2.5 1.5B, 3B Multilingual, JSON output, structured data
Gemma 2 2B Edge deployment, mobile, fast inference

Data Format

TuneKit uses the standard conversation format:

{"messages": [{"role": "user", "content": "What's the capital of France?"}, {"role": "assistant", "content": "Paris is the capital of France."}]}
{"messages": [{"role": "system", "content": "You are a helpful coding assistant."}, {"role": "user", "content": "Write hello world in Python"}, {"role": "assistant", "content": "print('Hello, World!')"}]}

Requirements:

  • JSONL format (one JSON object per line)
  • Each line has a messages array
  • Messages have role (user/assistant/system) and content
  • Minimum 50 examples (100-1000 recommended)

Run Locally

# Clone the repo
git clone https://github.com/riyanshibohra/TuneKit.git
cd TuneKit

# Install dependencies
pip install -r requirements.txt

# Start the server
uvicorn api.main:app --reload

Open http://localhost:8000 in your browser.


Tech Stack

  • Frontend: Vanilla JS, CSS
  • Backend: FastAPI, Python
  • Training: Unsloth, Transformers, PEFT
  • Infrastructure: Google Colab (free T4 GPU)

Contributing

Contributions are welcome! Feel free to:

  1. Fork the repo
  2. Create a feature branch (git checkout -b feature/amazing-feature)
  3. Commit your changes (git commit -m 'Add amazing feature')
  4. Push to the branch (git push origin feature/amazing-feature)
  5. Open a Pull Request

License

MIT License - use it for anything.


tunekit.app

Built with lots of caffeine + curiosity.

About

Upload your data → Get a fine-tuned SLM. Free.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published