Skip to main content
Version: v0.23.x

Experiment (model) flow

Overview​

An H2O Cascade experiment moves through five stages, from raw data to a deployed model:

  • Datasets: Upload your own data, or start from a demo dataset.
  • Data Generation (optional): Generate synthetic rows or labels with a large hosted model before fine-tuning.
  • Experiments: Configure and run the fine-tuning job. The default settings work for most cases, and advanced options are available if you need them.
  • AutoML (optional): Automatically iterate on hyperparameters and backbone selection, or use Ask KGM for a single recommendation.
  • Deployments: Deploy the trained model behind a REST endpoint and a chat interface for testing.

For a hands-on walkthrough with concrete UI steps, see Quick start: fine-tune your first model.


Feedback