ConnectionResetError(104) in KitchenLowdimRunner during environment initialization

[mr20010112]

Thank you for the great work on this project. I’ve encountered an issue when running KitchenLowdimRunner during environment initialization. Specifically, I’m seeing a ConnectionResetError that causes the process to fail when trying to initialize the AsyncVectorEnv.

Here is a summary of the problem:

• Error: ConnectionResetError(104, 'Connection reset by peer')
• Relevant file: diffusion_policy/env_runner/kitchen_lowdim_runner.py

• Error stack trace:

  Error in call to target 'diffusion_policy.env_runner.kitchen_lowdim_runner.KitchenLowdimRunner':
  ValueError("XML Error: top-level default class 'main' cannot be renamed\nElement 'default', line 0\n")
  full_key: task.env_runner
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env_runner/kitchen_lowdim_runner.py", line 180, in __init__
    env = SyncVectorEnv(env_fns)
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/gym_util/sync_vector_env.py", line 30, in __init__
    self.envs = [env_fn() for env_fn in env_fns]
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/gym_util/sync_vector_env.py", line 30, in <listcomp>
    self.envs = [env_fn() for env_fn in env_fns]
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env_runner/kitchen_lowdim_runner.py", line 58, in env_fn
    env = KitchenAllV0(use_abs_action=abs_action)
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env/kitchen/base.py", line 61, in __init__
    super(KitchenBase, self).__init__(use_abs_action=use_abs_action, **kwargs)
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env/kitchen/relay_policy_learning/adept_envs/adept_envs/franka/kitchen_multitask_v0.py", line 193, in __init__
    super(KitchenTaskRelaxV1, self).__init__(
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env/kitchen/relay_policy_learning/adept_envs/adept_envs/utils/configurable.py", line 132, in __init__
    base_init(self, *args, **kwargs)
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env/kitchen/relay_policy_learning/adept_envs/adept_envs/franka/kitchen_multitask_v0.py", line 51, in __init__
    super().__init__(
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env/kitchen/relay_policy_learning/adept_envs/adept_envs/robot_env.py", line 80, in __init__
    super(RobotEnv, self).__init__(
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env/kitchen/relay_policy_learning/adept_envs/adept_envs/mujoco_env.py", line 62, in __init__
    self.sim_robot = MujocoSimRobot(
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/env/kitchen/relay_policy_learning/adept_envs/adept_envs/simulation/sim_robot.py", line 62, in __init__
    self.sim = dm_mujoco.Physics.from_xml_path(model_file)
  ValueError: XML Error: top-level default class 'main' cannot be renamed
  Element 'default', line 0
  
  The above exception was the direct cause of the following exception:
  
  File "/home/mrq/project/diffusion_policy-main/diffusion_policy/workspace/train_diffusion_transformer_lowdim_workspace.py", line 113, in run
    env_runner = hydra.utils.instantiate(
  File "/home/mrq/project/diffusion_policy-main/train.py", line 34, in main
    workspace.run()
  File "/home/mrq/project/diffusion_policy-main/train.py", line 37, in <module>
    main()
  hydra.errors.InstantiationException: Error in call to target 'diffusion_policy.env_runner.kitchen_lowdim_runner.KitchenLowdimRunner':
  ValueError("XML Error: top-level default class 'main' cannot be renamed\nElement 'default', line 0\n")
  full_key: task.env_runner

The error seems to occur when AsyncVectorEnv is initialized, and the observation space check is performed via communication with child processes (pipe.recv()), resulting in the connection being reset.

[ShouZhengAI]

@mr20010112 hi bro , have you solved it?

[mr20010112]

Yes, by setting the gym version to 0.21.0 I guess.

[ShouZhengAI]

@mr20010112 thx,bro,I will try it

[Luca-Wiehe]

Yes, by setting the gym version to 0.21.0 I guess.

Unfortunately, this doesn't solve the issue for me. Have you changed anything else additionally? Can you tell us your python version?

[mr20010112]

Yes, by setting the gym version to 0.21.0 I guess.

Unfortunately, this doesn't solve the issue for me. Have you changed anything else additionally? Can you tell us your python version?

https://drive.google.com/file/d/1yhYW4pBkRQlkUtFwvHIGS3H4J--HMkva/view?usp=sharing

[mr20010112]

Yes, by setting the gym version to 0.21.0 I guess.

Unfortunately, this doesn't solve the issue for me. Have you changed anything else additionally? Can you tell us your python version?

3.9.15