Closed greydanus closed 4 years ago
Aha fixed. I was following the Colab tutorial which performs simulations via simulator.simulate_single(...)
. According to the docs this API was deprecated on v0.2.0 in favor of simulator.simulate_action(...)
. This updated method lets the user set need_featurized_objects: bool = True
Is there an easy way to obtain non-pixel state information eg. (x,y,r) information, for each object? I'd like to use ground truth object information for early proof-of-concept experiments before working directly with pixels