Episodes & steps

Reinforcement learning data has a time dimension you control with the scale toggle and slider at the top of the Environments views. Together they let you ask “what did each environment look like at this point in training?”

Scale toggle: RL Step vs Episode

Choose what the slider scrubs through:

• RL Step — position by training gradient-update step. Use this to see the state of environments at a given point in training progress.
• Episode — position by episode number instead.

See RL concepts for what these scales mean.

The slider

The slider moves to a specific point on the chosen scale (for example, RL step 302). As you move it:

• The environments table updates to show each environment’s values at that point (e.g. its current episode number).
• Environment charts reflect the selected position.

This makes it easy to step through training and watch how environments evolved.