Assignment: Build Your Own GPT Model
- Generative AI
- H2O Cascade
This short lesson sets up a hands on assignment rather than teaching new material: build your own GPT style model. It points learners to install H2O Cascade if they have not already, find a dataset from a source like Kaggle or Hugging Face, and choose a backbone model to train from. The assignment is intentionally open ended, asking learners to experiment with their own choice of backbone and dataset to produce their first fine-tuned model.
Key takeaways
- What the assignment asks for: installing H2O Cascade, loading a dataset from Kaggle or Hugging Face, and training a model
- Why choosing a backbone model to train from is left up to the learner rather than specified in the assignment
- How Kaggle and Hugging Face are suggested as starting points for finding a dataset to train a first GPT style model on
Start owning your AI