A quick guide for getting set for fine-tuning or inference using a jupyter notebook.
You have a few options for running fine-tuning notebooks:
- Hosted service (Recommended), e.g. Runpod or Vast.ai:
- Runpod one-click template here - easier setup.
- To support the Trelis Research YouTube channel, you can first sign up for an account with this link.
- Vast.ai one-click template here - offers smaller GPUs (which are cheaper to run).
- To support the Trelis Research YouTube channel, you can first sign up for an account with this affiliate link.
- Google Colab (free and good for 7B models or smaller):
- Upload the .ipynb notebook
- Select a T4 GPU from Runtime -> Change Runtime Type.
- make sure to comment out flash attention when loading the model.
- Your own computer (assuming you have an AMD or Nvidia GPU) - ADVANCED:
- Set up jupyter lab and a virtual environment using the instructions in the 'jupyter-lab-setup.md' file of this repo. see here.