Back to Resources
🦥

Unsloth

Training & Fine-tuning

Fast, memory-efficient fine-tuning for LLMs on consumer GPUs.

48kstars3.5kforksPython

About

Unsloth rewrites core training kernels to fine-tune models like Llama, Qwen, and Gemma up to 2x faster with significantly less VRAM, making local fine-tuning practical on a single consumer GPU.

Key Features

  • 2x faster training
  • Low VRAM usage
  • LoRA / QLoRA
  • Single-GPU fine-tuning

Tags

Fine-tuningLoRAEfficientLocal training

Related Resources