Questions about mlagents and tensorboard

How to view Step and episode length when converting a prefab environment into an application in mlagents.
For example, in the DodgeBall sample scene, we create an application for a scene with one or more prefabs, and check the learning results of starting 8 instances of it (--num-envs=8) on tensorboard. The question is, is the step of the axis a numerical value for each prefab, or a total value for each prefab? There are similar questions regarding EpisodeLength and reward on the vertical axis.
Also, are the calculation methods for these steps determined within the script? I would be grateful if you could let me know if you have any reference materials such as manuals.