Skip to content

Tuning & Scaling

With your data configured, the remaining YAML blocks control how the foundation model trains and how it uses your hardware. Basic Configuration introduces a minimal training_params block with batch limits for a smoke test — the guides below expand on that.

In this section

Guide Description
Training Behavior Control optimization, sampling, early stopping, and more as you refine your model
Scaling & Memory Adjust distribution, batch sizing, and memory when defaults under- or over-utilize your infrastructure