def test_states_dict(self):
gym_spaces = Dict({
'history':
Box(low=0.0, high=1.0, shape=(6, 50, 3), dtype=np.float32),
'weights':
Box(low=1.0, high=1.0, shape=(6, ), dtype=np.float32),
'weights2':
Box(low=1.0, high=1.0, shape=(6, ), dtype=np.float32)
})
tensorforce_spaces = {
'history': {
'shape': (6, 50, 3),
'type': 'float'
},
'weights': {
'shape': (6, ),
'type': 'float'
},
'weights2': {
'shape': (6, ),
'type': 'float'
}
}
env = OpenAIGymCustom(gym_id="CartPole-v0")
self.assertEqual(env.state_from_space(gym_spaces), tensorforce_spaces,
"conversion of states is not correct")
def test_actions_dict(self):
gym_spaces = Dict({
'history':
Box(low=0.0, high=1.0, shape=(6, 50, 3), dtype=np.float32),
'weights':
Box(low=1.0, high=1.0, shape=(6, ), dtype=np.float32),
'weights2':
Box(low=1.0, high=1.0, shape=(6, ), dtype=np.float32)
})
tensorforce_spaces = {
'history': {
'shape': (6, 50, 3),
'type': 'float'
},
'weights': {
'shape': (6, ),
'type': 'float'
},
'weights2': {
'shape': (6, ),
'type': 'float'
}
}
env = OpenAIGymCustom(gym_id="CartPole-v0")
env.action_from_space(gym_spaces)
def test_actions_box(self):
gym_spaces = Box(low=-10.0, high=10.0, shape=(1, ), dtype=np.float32)
tensorforce_spaces = {
'type': 'float',
'shape': (1, ),
'min_value': -10.0,
'max_value': 10.0
}
env = OpenAIGymCustom(gym_id="CartPole-v0")
self.assertEqual(env.action_from_space(gym_spaces), tensorforce_spaces,
"conversion of actions is not correct")
def test_actions_discrete(self):
gym_spaces = Discrete(10)
tensorforce_spaces = {'type': 'int', 'num_actions': 10}
env = OpenAIGymCustom(gym_id="CartPole-v0")
self.assertEqual(env.action_from_space(gym_spaces), tensorforce_spaces,
"conversion of actions is not correct")
def test_states_box(self):
gym_spaces = Box(low=-10.0, high=10.0, shape=(1, ), dtype=np.float32)
tensorforce_spaces = {'shape': (1, ), 'type': 'float'}
env = OpenAIGymCustom(gym_id="CartPole-v0")
self.assertEqual(env.state_from_space(gym_spaces), tensorforce_spaces,
"conversion of states is not correct")
def test_states_discrete(self):
gym_spaces = Discrete(10)
tensorforce_spaces = {'shape': (), 'type': 'int'}
env = OpenAIGymCustom(gym_id="CartPole-v0")
self.assertEqual(env.state_from_space(gym_spaces), tensorforce_spaces,
"conversion of states is not correct")