Qwen2.5 QLoRA, LoRA, and Full Fine-tuning on Your Computer

Don't wait for Llama 4 or Qwen3. Qwen2.5 is already a significant progress.

Alibaba released Qwen2.5. The models are available in many sizes: 0.5B, 1.5B, 3B, 7B, 14B, 32B, and 72B, as base and instruct models:

• Hugging Face collection: Qwen2.5

With all these sizes, Qwen2.5 offers flexibility across various hardware configurations. The 7B and 14B models are ideal for 24 GB GPUs, and they support efficient fine-tuning with LoRA and QLoRA. For full fine-tuning, smaller models are a better fit.

In this article, I explain how to fine-tune Qwen2.5, step by step. We will compare QLoRA, LoRA, and full fine-tuning speed and memory consumption. Since many model sizes are available, even if you only have a small 8 GB GPU, you will find a Qwen2.5 model that can be fine-tuned on your hardware.

The code for fine-tuning Qwen2.5 is also implemented in this notebook:

If you need GPUs, I recommend RunPod (referral link). I mainly use RunPod now, especially for long training.