too-far-away

I maybe wrong, but as I understand they close their current mediation platform, moving applications with DAU>25K to a new platform they call Fyber FairBid. Unless they are shutting down...

@kingsharaman note, that the loss returned by the `step()` is actually a sum of two losses (value and nerd) as seen [here](https://github.com/google-deepmind/open_spiel/blob/82b5aac85c577b6911f9a912544e2a589dacc2f1/open_spiel/python/algorithms/rnad/rnad.py#L845). It looks that the value loss is always...

@marton-avrios I used the learning rate of 1e-7, didn't touch the clip. In my case the trajectory_max=100 so that the games are much longer and it takes more time to...

> Do you have a pointer to code? It's the call to `state.observation_tensor()` in [rnad.py:968](https://github.com/deepmind/open_spiel/blob/ad4de766b702b5e3d1d2be247a11638f3473e59b/open_spiel/python/algorithms/rnad/rnad.py#L968) The `observer.set_from()` is not called directly from `rnad.py` (sorry, I mistakenly used `state.set_from()` instead). I...

You are right, R-NaD was just a trigger for the issue. Thank you very much for suggestions on where to look in the code. I've spent some more time investigating...

Sorry for the delayed answer. As you guys are more familiar with the code, thus you may want to make the change. As a quick and dirty workaround in my...