test_control.py File Reference

Classes
class	test_control.Eval

Namespaces
	test_control

Variables
int	test_control.seed = 42
string	test_control.load_path = '/home/giuseppe/src/taxons/experiments/Maze_AE_Mixed/{}'
	test_control.params = parameters.Params()
	test_control.env = gym.make(params.env_tag)
list	test_control.target_pose = [450, 500]
	test_control.obs = env.reset(desired_ball_pose=target_pose)
	test_control.target_image = env.render(mode='rgb_array')
	test_control.qpos = np.zeros(15)
	test_control.qvel = np.zeros(14)
	test_control.pose = target_pose+env.initPos[-1:]
	test_control.p = fs.Posture(*pose)
	test_control.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
	test_control.selector = ae.ConvAE(device=device, encoding_shape=params.feature_size)
	test_control.agent_type = agents.FFNeuralAgent
	test_control.pop = population.Population(agent=agent_type, pop_size=0, shapes=params.agent_shapes)
bool	test_control.done = False
int	test_control.t = 0
	test_control.agent_input = t
	test_control.action = utils.action_formatting(params.env_tag, agent['agent'](agent_input))
	test_control.reward
	test_control.info
	test_control.CoM = np.array([env.env.data.qpos[:2]])
	test_control.state = env.render(mode='rgb_array')
	test_control.surprise
	test_control.bs_point = bs_point.flatten().cpu().data.numpy()
	test_control.y
	test_control.goal = torch.Tensor(np.ascontiguousarray(target_image)).permute(2, 0, 1).unsqueeze(0).to(device)
	test_control.reconstr
	test_control.bs_space = np.stack([a[0] for a in pop['features'].values])
	test_control.diff = np.atleast_2d(bs_space - bs_point)
	test_control.dists = np.sqrt(np.sum(diff * diff, axis=1))
	test_control.closest_agent = np.argmin(dists)
	test_control.selected = pop[closest_agent]
list	test_control.images = []
list	test_control.saved_balls_pose = []
list	test_control.saved_joints_pose = []
list	test_control.saved_robot_pose = []
	test_control.f_pose = CoM
	test_control.SHOW_ARM_IN_ARRAY
	test_control.point_pose
	test_control.final_distance = np.sqrt(np.sum((target_pose - f_pose) ** 2))