dreamer
dreamer copied to clipboard
Dream to Control: Learning Behaviors by Latent Imagination
mbrl with distractors
Minor bugs: - gym.spaces.Box instance requires shape to be provided or to be inferred from the shapes of low or high. Since both low and high are scalars here, shape...
I believe these are two minor bugs, although if this is the intended behavior I apologize for the unnecessary changes.
Hard-coding the environment variable in code is restrictive, and some environments do not have EGL-compatible drivers (users normally set MUJOCO_GL to osmesa or glfw)
Hello! I was wondering why in the distributions of the RSSM and action model, the standard deviations are parameterized by the neural networks and in the reward model, the observation...