🎵 Qwen3-4B-Lyrics-GGUF 🎵
✨ Unleash your creative potential with AI-powered lyric generation ✨
✨ Model Description
This repository contains specialized GGUF versions of the Qwen3-4B model fine-tuned to create exceptional song lyrics. Each model has been carefully trained on a curated dataset of high-quality lyrics, enhancing its reasoning capabilities and creative expression.
📦 Available Models
| 🔮 Model File | 🧠 Base Model | ⚙️ Quantization | 📏 Size | 🚀 Optimization |
|---|---|---|---|---|
qwen3-4bf16.gguf |
Qwen3-4B | F16 | 4B parameters | Standard precision |
qwen3-4b-230-f16.gguf |
Qwen3-4B | F16 | 4B parameters | 230 training steps |
🧪 Training Details
| 📊 Aspect | 📝 Details |
|---|---|
| Dataset | 🎯 High-quality reasoning lyrics dataset (private) |
| Examples | 📚 661 unique examples per model |
| Training Strategy | 🔄 Individual fine-tuning for each model variant |
| Evaluation Metrics | 📈 Loss, Learning Rate, Gradient Norm, Steps/Samples per Second |
📊 Performance Metrics
✨ qwen3-4b-230-f16.gguf
| Metric | Final Value |
|---|---|
| 📉 Training Loss | 0.2627 |
| 📊 Evaluation Loss | 2.0974 |
| 📈 Learning Rate | 0.00018248 |
| 🔍 Gradient Norm | 0.71101 |
| ⏱️ Steps per Second | 1.835 |
| 🚀 Samples per Second | 10.554 |
| 🔄 Epochs | 8.87662 |
✨ qwen3-4bf16.gguf
| Metric | Final Value |
|---|---|
| 📉 Training Loss | 1.0589 |
| 📊 Evaluation Loss | 1.48644 |
| 📈 Learning Rate | 4.3798e-8 |
| 🔍 Gradient Norm | 0.44541 |
| ⏱️ Steps per Second | 2.725 |
| 🚀 Samples per Second | 10.445 |
| 🔄 Epochs | 2.99781 |
🚀 How to Use
Simple Setup
- 📥 Download LM Studio for an easy graphical interface
- ⬇️ Download the model from this repository
- 📂 Load the model in LM Studio and start generating lyrics!
Advanced Usage
- 🧠 For command-line enthusiasts, use llama.cpp
- 💻 Download the model and run it locally for maximum control
👏 Acknowledgements
This project would not have been possible without the following resources:
- Base Model: Qwen/Qwen3-4B - The original foundation model developed by the Qwen team at Alibaba Cloud
- Training Framework: Axolotl - An efficient and user-friendly fine-tuning framework that made the training process possible
We extend our sincere gratitude to these teams for their outstanding work and contributions to the open-source AI community.
🎤 Create songs that inspire, lyrics that move, and words that resonate 🎤
- Downloads last month
- 57
Hardware compatibility
Log In to add your hardware
16-bit
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support