I can't use mlagents_trainer_plugin dqn in 3DBall and Match3 unity env（btw，a2c is work）

[RangRun]

Describe the bug
3DBall env，It seems that there is an incompatibility issue with the shape when obtaining the action due to some torch version reasons

• shell

mlagents-learn dqn_basic.yaml --env="3DBall" --run-id=3DBall-dqn 

• config.yaml
  behaviors:
  3DBall:
  trainer_type: dqn
  hyperparameters:
  learning_rate: 0.0003
  learning_rate_schedule: constant
  batch_size: 64
  buffer_size: 50000
  tau: 0.005
  steps_per_update: 10.0
  save_replay_buffer: false
  exploration_schedule: linear
  exploration_initial_eps: 0.8
  exploration_final_eps: 0.05
  network_settings:
  normalize: false
  hidden_units: 20
  num_layers: 2
  vis_encode_type: simple
  reward_signals:
  extrinsic:
  gamma: 0.99
  strength: 1.0
  keep_checkpoints: 5
  max_steps: 500000
  time_horizon: 10
  summary_freq: 1000

• envs
  Version information:
  ml-agents: 1.0.0,
  ml-agents-envs: 1.0.0,
  Communicator API: 1.5.0,
  PyTorch: 2.2.1

• error
  in get_greedy_action
  return torch.argmax(all_q.sum(dim=0), dim=1, keepdim=True)
  IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)

Match3 env，It seems like Match3VectorObs is only receiving 2 inputs instead of the expected 4 inputs. Additionally, I have a question: why is Match3VectorObs in Unity for 4 agents, but env.behavior_specs["Match3VectorObs?team=0"].observation_specs only retrieves two?

• shell

mlagents-learn dqn_basic.yaml --env="Match3" --run-id=Match3-dqn 

• config.yaml
  default_settings:
  trainer_type: dqn
  hyperparameters:
  learning_rate: 0.0003
  learning_rate_schedule: constant
  batch_size: 64
  buffer_size: 50000
  tau: 0.005
  steps_per_update: 10.0
  save_replay_buffer: false
  exploration_schedule: linear
  exploration_initial_eps: 0.8
  exploration_final_eps: 0.05
  network_settings:
  normalize: false
  hidden_units: 20
  num_layers: 2
  vis_encode_type: match3
  reward_signals:
  extrinsic:
  gamma: 0.99
  strength: 1.0
  keep_checkpoints: 5
  max_steps: 500000
  time_horizon: 10
  summary_freq: 1000

• envs
  Version information:
  ml-agents: 1.0.0,
  ml-agents-envs: 1.0.0,
  Communicator API: 1.5.0,
  PyTorch: 2.2.1

• error
  line 374, in _assert_worker_can_restart
  raise exception
  mlagents_envs.exception.UnityActionException: The behavior Match3VectorObs?team=0 needs a continuous input of dimension (4, 0) for (, ) but received input of dimension (2, 0)

[github-actions]

This issue is stale because it has been open for 30 days with no activity.

[github-actions]

This issue was closed because it has been inactive for 14 days since being marked as stale. Please open a new issue for related bugs.