def __init__(self, args, gpu_id, strict_done=False): super(BasicEpisode, self).__init__() self._env = None self.gpu_id = gpu_id self.strict_done = strict_done self.task_data = None self.glove_embedding = None self.actions = get_actions(args) self.done_count = 0 self.duplicate_count = 0 self.failed_action_count = 0 self._last_action_embedding_idx = 0 self.target_object = None self.prev_frame = None self.current_frame = None self.det_frame = None self.last_det = False self.current_det = False self.det_gt = None self.optimal_actions = None self.scene_states = [] self.detections = [] if args.eval: random.seed(args.seed)
def __init__(self, args, gpu_id, strict_done=False): super(BasicEpisode, self).__init__() self._env = None self.gpu_id = gpu_id self.strict_done = strict_done self.task_data = None self.glove_embedding = None self.prototype = None self.actions = get_actions(args) self.done_count = 0 self.duplicate_count = 0 self.failed_action_count = 0 self._last_action_embedding_idx = 0 self.target_object = None self.prev_frame = None self.current_frame = None self.grid_size = args.grid_size self.goal_success_reward = args.goal_success_reward self.step_penalty = args.step_penalty self.step_penalty_table = [] self.episode_id = "" step_penalty = args.step_penalty for _ in range(0, args.max_ep, args.num_ep_per_stage): self.step_penalty_table.append(step_penalty) step_penalty = step_penalty * args.penalty_decay self.scene_states = [] self.episode_trajectories = [] self.actions_taken = [] if args.eval: random.seed(args.seed)
def __init__(self, args, gpu_id, strict_done=False): super(BasicEpisode, self).__init__() self._env = None self.gpu_id = gpu_id self.strict_done = strict_done self.task_data = None self.glove_embedding = None self.actions = get_actions(args) self.done_count = 0 self.duplicate_count = 0 self.failed_action_count = 0 self._last_action_embedding_idx = 0 self.target_object = None self.prev_frame = None self.current_frame = None self.scene = None self.scene_states = [] if args.eval: random.seed(args.seed) self._episode_times = 0 self.seen_percentage = 0 self.state_reps = [] self.state_memory = [] self.action_memory = [] self.obs_reps = [] self.episode_length = 0 self.target_object_detected = False # tools self.states = [] self.actions_record = [] self.action_outputs = [] self.detection_results = [] # imitation learning self.imitation_learning = args.imitation_learning self.action_failed_il = False self.action_probs = [] self.meta_learning = args.update_meta_network self.meta_predictions = [] self.visual_infos = {} self.match_score = [] self.indices_topk = []