Ml-agents low level python api: How do I activate training mode? Is there still something like that?

Hello,
I am used to an earlier version of unity agents, where there was a an train_mode option in the reset function of the environment:
env.reset(train_mode=True)[brain_name]
I have not found anything similar in the documentation for ml-agents.

The only thing I found that goes into this direction is the side-channels option, but I don't know how to configure them to optimally train my model.

Help is greatly appreciated.
Thank you

 

Hi,
train_mode was removed because it was a confusing name. train_mode did not mean that there would be training, it only meant that the engine would try to simulate faster (higher timescale and set captureframerate). You are right, it was replaced with the engine_configuration_side_channel on Python.