guest-oo
guest-oo
def find_near_neighbor(self, node_new): n = len(self.vertex) + 1 r = min(self.search_radius * math.sqrt((math.log(n) / n)), self.step_len) dist_table = [math.hypot(nd.x - node_new.x, nd.y - node_new.y) for nd in self.vertex] dist_table_index =...
save_path = saver.save(sess, checkpointFile) File "D:\anconda\envs\pytorch\lib\site-packages\tensorflow\python\training\saver.py", line 1326, in save self.export_meta_graph( File "D:\anconda\envs\pytorch\lib\site-packages\tensorflow\python\training\saver.py", line 1373, in export_meta_graph return export_meta_graph( File "D:\anconda\envs\pytorch\lib\site-packages\tensorflow\python\training\saver.py", line 1713, in export_meta_graph meta_graph_def, _ = meta_graph.export_scoped_meta_graph( File...
May I ask the author, after the training of a2c algorithm, shouldn't the randomness be eliminated when testing?
May I ask the author why def forward is pass? Shouldn't there be forward propagation?
May I ask the author why I used test.py to test at the end of training a2c, but when the state did not change, it produced a different action.
I built a maze environment myself, but I couldn't find results when I used the weight file at the end of training to test it. Do I need to train...
请问 作者在设计DQN的迷宫环境时，每次reset可以产生随机生成环境吗？ 智能体与障碍物发生碰撞的时候，可以选择让其保持在原有位置吗？ 请问可以将self.observation_space = spaces.Box(low=0, high=3, shape=(529,), dtype=np.int32) 迷宫环境转为一维的吗？例如这样self.observation_space = spaces.Box(low=0, high=3, shape=(529,), dtype=np.int32) 0为可移动 1为智能体位置 2 为终点 3 为障碍物
当使用demo_DQN_Dueling_Double_DQN 训练结束的的pt文件无法作为测试时的权重文件 ，是否需要将保存pt文件 由torch.save(actor, actor_path) 更改为torch.save(actor.state_dict(), actor_path)
D:\anconda\envs\pytorch\python.exe C:\Users\user\Desktop\ElegantRL-master\examples\demo_A2C_PPO.py env_args = {'env_name': 'CartPole-v1', 'num_envs': 1, 'max_step': 500, 'state_dim': 4, 'action_dim': 2, 'if_discrete': True} | Arguments Remove cwd: ./CartPole-v1_DiscreteA2C_0 | Evaluator: | `step`: Number of samples, or total...