How to add camera to the soccer game?

Hi,

I am playing around with the soccer game. I replaced the ray sensor with camera:


This is my training configuration:

Code (CSharp):

1. behaviors:
2.   SoccerTwosVisual:
3.     trainer_type: ppo
4.     hyperparameters:
5.       batch_size: 64
6.       buffer_size: 1024
7.       learning_rate: 0.0003
8.       beta: 0.005
9.       epsilon: 0.2
10.       lambd: 0.95
11.       num_epoch: 3
12.       learning_rate_schedule: linear
13.     network_settings:
14.       normalize: true
15.       hidden_units: 256
16.       num_layers: 2
17.       vis_encode_type: resnet
18.     reward_signals:
19.       extrinsic:
20.         gamma: 0.8
21.         strength: 1.0
22.     keep_checkpoints: 5
23.     max_steps: 50000000
24.     time_horizon: 1000
25.     summary_freq: 10000
26.     threaded: false
27.     self_play:
28.       save_steps: 50000
29.       team_change: 200000
30.       swap_steps: 2000
31.       window: 10
32.       play_against_latest_model_ratio: 0.5
33.       initial_elo: 1200.0
34.  

However I am getting the following error:

Code (CSharp):

1. Traceback (most recent call last):
2.   File "/usr/local/bin/mlagents-learn", line 33, in <module>
3.     sys.exit(load_entry_point('mlagents==0.24.0.dev0', 'console_scripts', 'mlagents-learn')())
4.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/learn.py", line 274, in main
5.     run_cli(parse_command_line())
6.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/learn.py", line 270, in run_cli
7.     run_training(run_seed, options)
8.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/learn.py", line 149, in run_training
9.     tc.start_learning(env_manager)
10.   File "/usr/local/lib/python3.8/dist-packages/mlagents_envs-0.24.0.dev0-py3.8.egg/mlagents_envs/timers.py", line 305, in wrapped
11.     return func(*args, **kwargs)
12.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/trainer_controller.py", line 172, in start_learning
13.     n_steps = self.advance(env_manager)
14.   File "/usr/local/lib/python3.8/dist-packages/mlagents_envs-0.24.0.dev0-py3.8.egg/mlagents_envs/timers.py", line 305, in wrapped
15.     return func(*args, **kwargs)
16.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/trainer_controller.py", line 230, in advance
17.     new_step_infos = env_manager.get_steps()
18.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/env_manager.py", line 112, in get_steps
19.     new_step_infos = self._step()
20.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/subprocess_env_manager.py", line 264, in _step
21.     self._queue_steps()
22.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/subprocess_env_manager.py", line 257, in _queue_steps
23.     env_action_info = self._take_step(env_worker.previous_step)
24.   File "/usr/local/lib/python3.8/dist-packages/mlagents_envs-0.24.0.dev0-py3.8.egg/mlagents_envs/timers.py", line 305, in wrapped
25.     return func(*args, **kwargs)
26.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/subprocess_env_manager.py", line 378, in _take_step
27.     all_action_info[brain_name] = self.policies[brain_name].get_action(
28.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/policy/torch_policy.py", line 207, in get_action
29.     run_out = self.evaluate(decision_requests, global_agent_ids)
30.   File "/usr/local/lib/python3.8/dist-packages/mlagents_envs-0.24.0.dev0-py3.8.egg/mlagents_envs/timers.py", line 305, in wrapped
31.     return func(*args, **kwargs)
32.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/policy/torch_policy.py", line 173, in evaluate
33.     action, log_probs, entropy, memories = self.sample_actions(
34.   File "/usr/local/lib/python3.8/dist-packages/mlagents_envs-0.24.0.dev0-py3.8.egg/mlagents_envs/timers.py", line 305, in wrapped
35.     return func(*args, **kwargs)
36.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/policy/torch_policy.py", line 135, in sample_actions
37.     actions, log_probs, entropies, memories = self.actor_critic.get_action_stats(
38.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/torch/networks.py", line 500, in get_action_stats
39.     action, log_probs, entropies, actor_mem_out = super().get_action_stats(
40.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/torch/networks.py", line 303, in get_action_stats
41.     encoding, memories = self.network_body(
42.   File "/usr/local/lib/python3.8/dist-packages/torch-1.7.1-py3.8-linux-x86_64.egg/torch/nn/modules/module.py", line 727, in _call_impl
43.     result = self.forward(*input, **kwargs)
44.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/torch/networks.py", line 87, in forward
45.     processed_obs = processor(obs_input)
46.   File "/usr/local/lib/python3.8/dist-packages/torch-1.7.1-py3.8-linux-x86_64.egg/torch/nn/modules/module.py", line 727, in _call_impl
47.     result = self.forward(*input, **kwargs)
48.   File "/usr/local/lib/python3.8/dist-packages/mlagents-0.24.0.dev0-py3.8.egg/mlagents/trainers/torch/encoders.py", line 270, in forward
49.     before_out = hidden.view(batch_size, -1)
50. RuntimeError: view size is not compatible with input tensor's size and stride (at least one dimension spans across two contiguous subspaces). Use .reshape(...) instead.
51.  

My guess would be that something is wrong with my camera, how can I fix it?

Regards,
RUn

 

Hi, I think you found a bug on master. I tried to make a fix on this branch : fix-contiguous-resnet (https://github.com/Unity-Technologies/ml-agents/tree/fix-contiguous-resnet) Do you think you can try this branch out and tell us if your problem is resolved ?

 

Thanks. This solves the problem.
Can two camera sensor added for an Agent?

 

> Can two camera sensor added for an Agent?

Yes, although there is no demo environment to show it. Please do let us know if it does not work for you.