🚀 Qwen 1.5 4B LoRA Fine-tuning

Fine-tune Qwen 4B for function calling using LoRA on FREE GPU!

⚡ Features:

  • LoRA (90% less VRAM than full fine-tuning)
  • 3 epochs in ~30-40 minutes
  • Automatic model merging
  • Download trained model

📋 Dataset format: JSONL with instruction/output fields

Dataset Upload

Configuration

1 5
1 8
64 256

Training

📚 Next Steps

  1. Download qwen-4b-finetuned.zip
  2. Extract on your machine
  3. Load to Ollama:
    ollama create qwen-4b-finetuned -f Modelfile
ollama run qwen-4b-finetuned "Wylistuj pliki"
  4. Use in OpenCode/LiteLLM

⚙️ Hardware

  • GPU: Free T4 (15GB VRAM)
  • Storage: ~10GB for output
  • Time: ~40 minutes

Made with ❤️ for Qwen fine-tuning