Canot run the training script successfully

[Revindell]

Hi muyue1202,

Thanks for sharing this project. I have an issue running the training script. After I ran the script, I got an error message which shows

Traceback (most recent call last):
  File "scripts/train.py", line 256, in <module>
    train_go1(headless=False)
  File "scripts/train.py", line 216, in train_go1
    runner.learn(num_learning_iterations=100000, init_at_random_ep_len=True, eval_freq=100)
  File "/home/deyu.fu/quadruped_locomotion_project/go1_gym_learn/ppo_cse/__init__.py", line 141, in learn
    actions_train = self.alg.act(obs[:num_train_envs], privileged_obs[:num_train_envs],
  File "/home/deyu.fu/quadruped_locomotion_project/go1_gym_learn/ppo_cse/ppo.py", line 86, in act
    self.transition.actions = self.actor_critic.act(obs_history).detach()
TypeError: act() missing 1 required positional argument: 'depth_obs'

After I checked the source code, I do see that the method ActorCritic.act requires a parameter called depth_obs, which is missing when this function is called. Can you help me fix this issue? Thanks!

[baoping-ma]

I have the same as your question, if need the insert the camera?

[muye1202]

Sorry guys I just saw this, let me take a look and get back to you ASAP

This might be due to the training and deployment pipeline are using the same actor_critic structure. Deployment requires camera input for depth observation (the depth_ons in the error message); I will try adding depth_obs to the pipeline.

[baoping-ma]

Excuse me, have you made any progress on the solution you mentioned (adding the depth_obs to the pipeline)? Thank you!

[baoping-ma]

Great job. Can you help with that? Appreciate very much.

TypeError: act() missing 1 required positional argument: 'depth_obs'

[baoping-ma]

Hi muyue1202,嗨，muyue1202，

Thanks for sharing this project. I have an issue running the training script. After I ran the script, I got an error message which shows感谢您分享这个项目。我在运行训练脚本时遇到问题。运行脚本后，我收到一条错误消息，显示

Traceback (most recent call last):
  File "scripts/train.py", line 256, in <module>
    train_go1(headless=False)
  File "scripts/train.py", line 216, in train_go1
    runner.learn(num_learning_iterations=100000, init_at_random_ep_len=True, eval_freq=100)
  File "/home/deyu.fu/quadruped_locomotion_project/go1_gym_learn/ppo_cse/__init__.py", line 141, in learn
    actions_train = self.alg.act(obs[:num_train_envs], privileged_obs[:num_train_envs],
  File "/home/deyu.fu/quadruped_locomotion_project/go1_gym_learn/ppo_cse/ppo.py", line 86, in act
    self.transition.actions = self.actor_critic.act(obs_history).detach()
TypeError: act() missing 1 required positional argument: 'depth_obs'

After I checked the source code, I do see that the method ActorCritic.act requires a parameter called depth_obs, which is missing when this function is called. Can you help me fix this issue? Thanks!在我检查源代码后，我确实看到方法 ActorCritic.act 需要一个名为 height_obs 的参数，但在调用该函数时缺少该参数。你能帮我解决这个问题吗？谢谢！

Have you solved this problem? How was it solved? Can you share? Thank you very much!

[muye1202]

Hi guys

Sorry for the delay, the depth_obs in go1_gym_learn/ppo_cse/actor_critic.py is required if you want the robot to have image observation as input during training. If you do not want to use visual info during training you can use og_actor_critic.py in the same folder, that file does not require depth_obs.

Thanks again for looking at our repo!