Does training increase time scale?

[Imakhiil]

I followed this tutorial:

Unity-Technologies/ml-agents/blob/main/docs/Getting-Started.md

The model learnt just fine, but when I loaded it back in the Editor and entered play mode, balls were much, much slower than during training. Was time scale simply increased when the model was learning?

[LexVolkov]

@Imakhiil

I followed this tutorial:
https://github.com/Unity-Technologies/ml-agents/blob/master/docs/Getting-Started.md
The model learnt just fine, but when I loaded it back in the Editor and entered play mode, balls were much, much slower than during training. Was time scale simply increased when the model was learning?

x20

[LexVolkov]

Unity-Technologies/ml-agents/blob/main/docs/Training-ML-Agents.md#command-line-training-options

–time-scale : The time scale of the Unity environment(s). Equivalent to setting Time.timeScale in Unity. (Default 20.0, maximum 100.0)

[Imakhiil]

Thanks

[ChrissCrass]

Training does not necessarily use fixedDeltaTime to govern its updates, so training may not appear to be the same speed as when you inference on the same code. (results and performance should be the same, but you might wonder why training appears to be faster even when timeScale is set to 1. The reason is that training causes Unity to put out fixed updates as fast as possible, which boosts training speed for physics using agents).