PyCon2024 Workshop

5) Finetuning Large Language Models

🔗 Slides

   

Exercise: Finetune a GPT model

Part 1: Train the model

1. Open and inspect the exercise/01_train-gpt.ipynb notebook
2. Execute the code up to section 5
3. Code the missing lines in the training loop and run the remained of the notebook on a GPU machine (A10G or L4)

 

Part 2: Use the model

1. Open and inspect the exercise/02_use-trained-gpt.ipynb notebook
2. Add missing code to load the model
3. Run the remainder of the notebook on a CPU or GPU machine

 

Tip

If you feel stuck on loading the model, see slides 8 for the starter code