unsloth-finetuning - Fine-tune LLMs with Unsloth
Fine-tune LLMs with Unsloth using 4-bit quantization and LoRA
Tags
Updated: 2026-03-26Capabilities
Typical Inputs
Typical Outputs
What this skill does
- load 4-bit quantized models
- configure LoRA parameters
- train on datasets
- export models to GGUF
- export models to Hugging Face
- export models to vLLM
- export models to Ollama
- monitor training progress
- enable gradient checkpointing
- adjust batch size
- set learning rate
Inputs
- training dataset
- pre-trained model
- GPU with CUDA
- Hugging Face token
Outputs
- fine-tuned model
- training logs
- exported model files
- checkpoints
Requirements
- Python 3.10-3.12
- NVIDIA GPU
- CUDA 11.8 or 12.1+
- PyTorch 2.0+
- Transformers 4.37+