AHPUymhd
AHPUymhd
> Hi, For the highway-v0 run, I think that the problem is that the observation is configures with absolute coordinates (`absolute: True`) instead of relative (`absolute: False`). This means that...
@eleurent Thank you very much
@eleurent Thank you very much
> Hi, This is because this config["observation"]["features_range"]["vx"/"vy"] does not control the range of speeds that can be reached by the vehicle, but only how the observed speed values get normalised...
@eleurent
> Hi, DQN (and its variants Dueling DQN and Double Dueling DQN) is the _learning algorithm_ (just like PPO is), while Attention/Transformer is the _network architecture_, which is being trained...
@zijinoier
@zijinoier
> When I run `pretrain.sh`, the following error occurs. Why doesn't the class `MotionEpicMetaModel` in `motionepic_arch.py` have a `get_sg_encoder` method? Could you provide this method? > > ``` > Traceback...