Skip to main content

Using QLora to fine-tune a model

https://github.com/artidoro/qlora

Related: QLoRA

Key-take aways​

  • Fine tuning a model can be performed in about 30 hours on a MacBook and allows you to make a generative model do what you want it to do without any prompt required. For example, complete the text as if it was a quote from a famous person.
  • Choosing the right parameters matters (Choosing parameters for Lora)
  • The fine-tuning can be done in a matter of minutes using Google Collab

Use the How do you fine-tune a model using peft to see how to use QLora in practice.