Model has a different behaviour on relaunching Unity.

These are the steps that I followed :

  1. Started training without any NNModel and ran it for (11 hours approx and 70k steps)
  2. Stopped training and generated “.nn” file.
  3. Relaunched Unity.
  4. Started training again with the previously generated file.
  5. The behaviour was different in both cases.

Is it because I didn’t let the agent complete the max_steps configured in the config? or is there some other reason for this.

Hi,

What was the scene you used?

What saved frequency did you specified ? (If not specified it’s every 50,000 steps)

By “4. Started training again with the previously generated file.” Do you mean you launched training again with the --load flag?