CEMSolver

class CEMSolver(model: Costable, batch_size: int = 1, num_samples: int = 300, var_scale: float = 1, n_steps: int = 30, topk: int = 30, device='cpu', seed: int = 1234)[source]

Bases: object

Cross Entropy Method Solver.

adapted from https://github.com/gaoyuezhou/dino_wm/blob/main/planning/cem.py

property action_dim: int
configure(*, action_space, n_envs: int, config) None[source]
property horizon: int
init_action_distrib(actions=None)[source]

Initialize the action distribution params (mu, sigma) given the initial condition.

Parameters:

actions (n_envs, T, action_dim) – initial actions, T <= horizon

property n_envs: int
solve(info_dict, init_action=None) dict[source]