Training stops immediately with error: SubprocessEnvManager had workers that didn't signal shutdown

Hi,

When I run 'mlagents-learn --force', I get the Unity logo as expected, but once I press play, it loads for a couple seconds and gives me an error (see full printout below).

I am using the 'ML Agents 2.2.1-exp1' package.

Full printout:

Code (CSharp):

1. (venv) C:\Users\hjalte\Desktop\Szrot\Szrot ML>mlagents-learn --force
2.  
3.             ┐  ╖
4.         ╓╖╬│╡  ││╬╖╖
5.     ╓╖╬│││││┘  ╬│││││╬╖
6. ╖╬│││││╬╜        ╙╬│││││╖╖                               ╗╗╗
7. ╬╬╬╬╖││╦╖        ╖╬││╗╣╣╣╬      ╟╣╣╬    ╟╣╣╣             ╜╜╜  ╟╣╣
8. ╬╬╬╬╬╬╬╬╖│╬╖╖╓╬╪│╓╣╣╣╣╣╣╣╬      ╟╣╣╬    ╟╣╣╣ ╒╣╣╖╗╣╣╣╗   ╣╣╣ ╣╣╣╣╣╣ ╟╣╣╖   ╣╣╣
9. ╬╬╬╬┐  ╙╬╬╬╬│╓╣╣╣╝╜  ╫╣╣╣╬      ╟╣╣╬    ╟╣╣╣ ╟╣╣╣╙ ╙╣╣╣  ╣╣╣ ╙╟╣╣╜╙  ╫╣╣  ╟╣╣
10. ╬╬╬╬┐     ╙╬╬╣╣      ╫╣╣╣╬      ╟╣╣╬    ╟╣╣╣ ╟╣╣╬   ╣╣╣  ╣╣╣  ╟╣╣     ╣╣╣┌╣╣╜
11. ╬╬╬╜       ╬╬╣╣      ╙╝╣╣╬      ╙╣╣╣╗╖╓╗╣╣╣╜ ╟╣╣╬   ╣╣╣  ╣╣╣  ╟╣╣╦╓    ╣╣╣╣╣
12. ╙   ╓╦╖    ╬╬╣╣   ╓╗╗╖            ╙╝╣╣╣╣╝╜   ╘╝╝╜   ╝╝╝  ╝╝╝   ╙╣╣╣    ╟╣╣╣
13.    ╩╬╬╬╬╬╬╦╦╬╬╣╣╗╣╣╣╣╣╣╣╝                                             ╫╣╣╣╣
14.       ╙╬╬╬╬╬╬╬╣╣╣╣╣╣╝╜
15.           ╙╬╬╬╣╣╣╜
16.              ╙
17.  
18. Version information:
19.   ml-agents: 0.28.0,
20.   ml-agents-envs: 0.28.0,
21.   Communicator API: 1.5.0,
22.   PyTorch: 1.7.1+cu110
23. [INFO] Listening on port 5004. Start training by pressing the Play button in the Unity Editor.
24. [INFO] Connected to Unity environment with package version 2.2.1-exp.1 and communication version 1.5.0
25. [INFO] Connected new brain: MoveToGoal?team=0
26. [WARNING] Behavior name MoveToGoal does not match any behaviors specified in the trainer configuration file. A default configuration will be used.
27. [WARNING] Deleting TensorBoard data events.out.tfevents.1648050845.DESKTOP-FHQ6GMJ.25336.0 that was left over from a previous run.
28. [INFO] Hyperparameters for behavior name MoveToGoal:
29.         trainer_type:   ppo
30.         hyperparameters:
31.           batch_size:   1024
32.           buffer_size:  10240
33.           learning_rate:        0.0003
34.           beta: 0.005
35.           epsilon:      0.2
36.           lambd:        0.95
37.           num_epoch:    3
38.           learning_rate_schedule:       linear
39.           beta_schedule:        linear
40.           epsilon_schedule:     linear
41.         network_settings:
42.           normalize:    False
43.           hidden_units: 128
44.           num_layers:   2
45.           vis_encode_type:      simple
46.           memory:       None
47.           goal_conditioning_type:       hyper
48.           deterministic:        False
49.         reward_signals:
50.           extrinsic:
51.             gamma:      0.99
52.             strength:   1.0
53.             network_settings:
54.               normalize:        False
55.               hidden_units:     128
56.               num_layers:       2
57.               vis_encode_type:  simple
58.               memory:   None
59.               goal_conditioning_type:   hyper
60.               deterministic:    False
61.         init_path:      None
62.         keep_checkpoints:       5
63.         checkpoint_interval:    500000
64.         max_steps:      500000
65.         time_horizon:   64
66.         summary_freq:   50000
67.         threaded:       False
68.         self_play:      None
69.         behavioral_cloning:     None
70. [INFO] Exported results\ppo\MoveToGoal\MoveToGoal-0.onnx
71. [INFO] Copied results\ppo\MoveToGoal\MoveToGoal-0.onnx to results\ppo\MoveToGoal.onnx.
72. [ERROR] SubprocessEnvManager had workers that didn't signal shutdown
73. [ERROR] A SubprocessEnvManager worker did not shut down correctly so it was forcefully terminated.
74. Traceback (most recent call last):
75.  File "C:\Users\hjalte\AppData\Local\Programs\Python\Python37\lib\runpy.py", line 193, in _run_module_as_main
76.    "__main__", mod_spec)
77.  File "C:\Users\hjalte\AppData\Local\Programs\Python\Python37\lib\runpy.py", line 85, in _run_code
78.    exec(code, run_globals)
79.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\Scripts\mlagents-learn.exe\__main__.py", line 7, in <module>
80.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents\trainers\learn.py", line 260, in main
81.    run_cli(parse_command_line())
82.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents\trainers\learn.py", line 256, in run_cli
83.    run_training(run_seed, options, num_areas)
84.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents\trainers\learn.py", line 132, in run_training
85.    tc.start_learning(env_manager)
86.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents_envs\timers.py", line 305, in wrapped
87.    return func(*args, **kwargs)
88.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents\trainers\trainer_controller.py", line 176, in start_learning
89.    n_steps = self.advance(env_manager)
90.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents_envs\timers.py", line 305, in wrapped
91.    return func(*args, **kwargs)
92.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents\trainers\trainer_controller.py", line 234, in advance
93.    new_step_infos = env_manager.get_steps()
94.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents\trainers\env_manager.py", line 124, in get_steps
95.    new_step_infos = self._step()
96.  File "C:\Users\hjalte\Desktop\Szrot\Szrot ML\venv\lib\site-packages\mlagents\trainers\subprocess_env_manager.py", line 417, in _step
97.    step: EnvironmentResponse = self.step_queue.get_nowait()
98.  File "C:\Users\hjalte\AppData\Local\Programs\Python\Python37\lib\multiprocessing\queues.py", line 126, in get_nowait
99.    return self.get(False)
100.  File "C:\Users\hjalte\AppData\Local\Programs\Python\Python37\lib\multiprocessing\queues.py", line 109, in get
101.    self._sem.release()
102. ValueError: semaphore or lock released too many times

I had some issues in the past, so I erased all my python, mlagents etc. This should be a perfectly clean install and even entirely new project.

This is the agent, I'm trying to train (It worked perfectly with MLAgents 16.0):

Code (CSharp):

1. using System.Collections;
2. using System.Collections.Generic;
3. using Unity.MLAgents;
4. using Unity.MLAgents.Actuators;
5. using Unity.MLAgents.Sensors;
6. using UnityEngine;
7.  
8. public class TestAgent : Agent
9. {
10.     public Transform target;
11.     public float speed = 5f;
12.  
13.     private MeshRenderer floorRender;
14.     private MaterialPropertyBlock mpb;
15.  
16.     private void Start()
17.     {
18.         floorRender = transform.parent.Find("Ground").GetComponent<MeshRenderer>();
19.         mpb = new MaterialPropertyBlock();
20.     }
21.  
22.     public override void OnEpisodeBegin()
23.     {
24.         base.OnEpisodeBegin();
25.  
26.         transform.localPosition = new Vector3(Random.Range(-3f, 3f), 0, Random.Range(-3f, 3f));
27.         target.localPosition    = new Vector3(Random.Range(-3f, 3f), 0, Random.Range(-3f, 3f));
28.     }
29.  
30.     public override void CollectObservations(VectorSensor sensor)
31.     {
32.         base.CollectObservations(sensor);
33.  
34.         sensor.AddObservation(transform.localPosition);
35.         sensor.AddObservation(target.localPosition);
36.     }
37.  
38.     public override void OnActionReceived(ActionBuffers actions)
39.     {
40.         base.OnActionReceived(actions);
41.  
42.         float moveX = actions.ContinuousActions[0];
43.         float moveZ = actions.ContinuousActions[1];
44.  
45.         transform.localPosition += new Vector3(moveX, 0, moveZ) * Time.deltaTime * speed;
46.     }
47.  
48.     public override void Heuristic(in ActionBuffers actionsOut)
49.     {
50.         base.Heuristic(actionsOut);
51.  
52.         ActionSegment<float> contActions = actionsOut.ContinuousActions;
53.         contActions[0] = Input.GetAxisRaw("Horizontal");
54.         contActions[1] = Input.GetAxisRaw("Vertical");
55.     }
56.  
57.     private void OnTriggerEnter(Collider other)
58.     {
59.         if (other.CompareTag("Wall"))
60.         {
61.             SetReward(-1f);
62.             mpb.SetColor("_Color", Color.red);
63.             floorRender.SetPropertyBlock(mpb);
64.             EndEpisode();
65.         }
66.         if (other.CompareTag("Goal"))
67.         {
68.             SetReward(1f);
69.             mpb.SetColor("_Color", Color.green);
70.             floorRender.SetPropertyBlock(mpb);
71.             EndEpisode();
72.         }
73.     }
74. }
75.  

 

any update on this ?