No VICE examples are provided #95
Description
Hi, thanks for your great work!
Though I found the state-based SAC and image-based DrQ work well in simulation, no VICE examples are provided, though the VICEAgent definition is provided.
After fixing several bugs, I trained VICEAgent in the PickCubeVision-v0 simulation using the default DrQ hyperparameters and make_vice_agent settings, but it converged to a policy that never succeeds.
Considering that VICE is highlighted in your paper, I wonder if you can provide several VICEAgent examples with fine-grained parameters that can work well in simulation and real world.