Ervin T.

I've been able to use both T4 and P4 GPUs for training Unity environments (including Obstacle Tower). @Sohojoe do you have the `/etc/X11/xorg.conf` for the problematic machine?

Hi @Sohojoe, the obstacle-tower-env doesn't have a TensorFlow requirement as it doesn't install ml-agents. You can check the GPU usage with `nvidia-smi`. What type of GPU are you running locally?

@Sohojoe you are correct, the Readme is wrong (and we'll fix it). The newest versions of OTC no longer uses ML-Agents in its entirety, and doesn't require TensorFlow. Dopamine does...

@Sohojoe thanks for pointing out these issues! For the config, I've put it to more than 10M intentionally, since depending on environment (e.g. for a fixed seed) the agent may...

This is definitely something we want to do - there aren't currently fixed APIs with the Trainers as they're not really meant to be public. I've logged it to our...

Hmm, this isn't something that's currently supported. If I understand correctly, you'd want to constrain the observation in Unity, and have the Box that's returned have those values? I think...

Hmm, the main reason we haven't done this is it's not standard on the Python/training side to have the Alpha channel as input to the CNN. Still, we _do_ support...

That's very slow. Have you tried increasing the time scale? This can be done through the engine configuration side-channel: https://github.com/Unity-Technologies/ml-agents/blob/main/docs/Python-API.md#engineconfigurationchannel

We use 20 for most of our example environments. If you go too fast, the physics gets kind of wonky, and sometimes objects/agents will go through each other. The default...

The raycast thing is a known issue, but it's also not a bug. If it wasn't the case, then the rays would never leave the agent! You're right, the --force...