Unity 机器学习代理工具包 (ML-Agents) 是一个开源项目,它使游戏和模拟能够作为训练智能代理的环境。
您最多选择25个主题 主题必须以中文或者字母或数字开头,可以包含连字符 (-),并且长度不得超过35个字符
 
 
 
 
 

163 行
6.0 KiB

import mock
import numpy as np
import os
import pytest
import socket
import mock
import struct
from unityagents import UnityEnvironment, UnityEnvironmentException, UnityActionException, BrainInfo, BrainParameters
def append_length(input):
return struct.pack("I", len(input.encode())) + input.encode()
dummy_start = '''{
"AcademyName": "RealFakeAcademy",
"resetParameters": {},
"brainNames": ["RealFakeBrain"],
"externalBrainNames": ["RealFakeBrain"],
"apiNumber":"API-2",
"brainParameters": [{
"stateSize": 3,
"actionSize": 2,
"memorySize": 0,
"cameraResolutions": [],
"actionDescriptions": ["",""],
"actionSpaceType": 1,
"stateSpaceType": 1
}]
}'''.encode()
dummy_reset = [
'CONFIG_REQUEST'.encode(),
append_length(
'''
{
"brain_name": "RealFakeBrain",
"agents": [1,2],
"states": [1,2,3,4,5,6],
"rewards": [1,2],
"actions": [1,2,3,4],
"memories": [],
"dones": [false, false]
}'''),
'False'.encode()]
dummy_step = ['actions'.encode(),
append_length('''
{
"brain_name": "RealFakeBrain",
"agents": [1,2,3],
"states": [1,2,3,4,5,6,7,8,9],
"rewards": [1,2,3],
"actions": [1,2,3,4,5,6],
"memories": [],
"dones": [false, false, false]
}'''),
'False'.encode(),
'actions'.encode(),
append_length('''
{
"brain_name": "RealFakeBrain",
"agents": [1,2,3],
"states": [1,2,3,4,5,6,7,8,9],
"rewards": [1,2,3],
"actions": [1,2,3,4,5,6],
"memories": [],
"dones": [false, false, true]
}'''),
'True'.encode()]
def test_handles_bad_filename():
with pytest.raises(UnityEnvironmentException):
UnityEnvironment(' ')
def test_initialialization():
with mock.patch('subprocess.Popen') as mock_subproc_popen:
with mock.patch('socket.socket') as mock_socket:
with mock.patch('glob.glob') as mock_glob:
mock_glob.return_value = ['FakeLaunchPath']
mock_socket.return_value.accept.return_value = (mock_socket, 0)
mock_socket.recv.return_value.decode.return_value = dummy_start
env = UnityEnvironment(' ')
with pytest.raises(UnityActionException):
env.step([0])
assert env.brain_names[0] == 'RealFakeBrain'
env.close()
def test_reset():
with mock.patch('subprocess.Popen') as mock_subproc_popen:
with mock.patch('socket.socket') as mock_socket:
with mock.patch('glob.glob') as mock_glob:
mock_glob.return_value = ['FakeLaunchPath']
mock_socket.return_value.accept.return_value = (mock_socket, 0)
mock_socket.recv.return_value.decode.return_value = dummy_start
env = UnityEnvironment(' ')
brain = env.brains['RealFakeBrain']
mock_socket.recv.side_effect = dummy_reset
brain_info = env.reset()
env.close()
assert not env.global_done
assert isinstance(brain_info, dict)
assert isinstance(brain_info['RealFakeBrain'], BrainInfo)
assert isinstance(brain_info['RealFakeBrain'].observations, list)
assert isinstance(brain_info['RealFakeBrain'].states, np.ndarray)
assert len(brain_info['RealFakeBrain'].observations) == brain.number_observations
assert brain_info['RealFakeBrain'].states.shape[0] == len(brain_info['RealFakeBrain'].agents)
assert brain_info['RealFakeBrain'].states.shape[1] == brain.state_space_size
def test_step():
with mock.patch('subprocess.Popen') as mock_subproc_popen:
with mock.patch('socket.socket') as mock_socket:
with mock.patch('glob.glob') as mock_glob:
mock_glob.return_value = ['FakeLaunchPath']
mock_socket.return_value.accept.return_value = (mock_socket, 0)
mock_socket.recv.return_value.decode.return_value = dummy_start
env = UnityEnvironment(' ')
brain = env.brains['RealFakeBrain']
mock_socket.recv.side_effect = dummy_reset
brain_info = env.reset()
mock_socket.recv.side_effect = dummy_step
brain_info = env.step([0] * brain.action_space_size * len(brain_info['RealFakeBrain'].agents))
with pytest.raises(UnityActionException):
env.step([0])
brain_info = env.step([0] * brain.action_space_size * len(brain_info['RealFakeBrain'].agents))
with pytest.raises(UnityActionException):
env.step([0] * brain.action_space_size * len(brain_info['RealFakeBrain'].agents))
env.close()
assert env.global_done
assert isinstance(brain_info, dict)
assert isinstance(brain_info['RealFakeBrain'], BrainInfo)
assert isinstance(brain_info['RealFakeBrain'].observations, list)
assert isinstance(brain_info['RealFakeBrain'].states, np.ndarray)
assert len(brain_info['RealFakeBrain'].observations) == brain.number_observations
assert brain_info['RealFakeBrain'].states.shape[0] == len(brain_info['RealFakeBrain'].agents)
assert brain_info['RealFakeBrain'].states.shape[1] == brain.state_space_size
assert not brain_info['RealFakeBrain'].local_done[0]
assert brain_info['RealFakeBrain'].local_done[2]
def test_close():
with mock.patch('subprocess.Popen') as mock_subproc_popen:
with mock.patch('socket.socket') as mock_socket:
with mock.patch('glob.glob') as mock_glob:
mock_glob.return_value = ['FakeLaunchPath']
mock_socket.return_value.accept.return_value = (mock_socket, 0)
mock_socket.recv.return_value.decode.return_value = dummy_start
env = UnityEnvironment(' ')
assert env._loaded
env.close()
assert not env._loaded
mock_socket.close.assert_called_once()
if __name__ == '__main__':
pytest.main()