def get_state_space(self): if self.processor is None: return Space.convert_openai_space(self.env.observation_space) else: self.reset() shape = self.current_state.shape min = np.zeros(shape, dtype=np.uint8) max = np.full(shape, 255, dtype=np.uint8) return Space(min, max, True)
def get_state_space(self): if self.__processor is None: shape = self.__current_buffer.shape else: shape = self.__processor.process(self.__current_buffer).shape min_value = np.zeros(shape, dtype=np.uint8) max_value = np.full(shape, 255) return Space(min_value, max_value, True)
def get_state_space(self): from fruit.types.priv import Space state_spec = self.env.observation_spec() for d in state_spec: if d['name'] == self.mode: shape = d['shape'] return Space(np.full(shape, 0), np.full(shape, 255), True) return None
def get_state_space(self): from fruit.types.priv import Space if self.processor is None: return self.game.get_state_space() else: min = np.zeros([84, 84], dtype=np.uint8) max = np.full([84, 84], 255) return Space(min, max, True)
def get_action_space(self, act_space=None): from fruit.types.priv import Space if act_space is None: act_space = self.environment.actions() print(act_space) if 'num_values' in act_space: return Space(0, act_space['num_values'] - 1, True) elif 'num_actions' in act_space: return Space(0, act_space['num_actions'] - 1, True) elif 'shape' in act_space: if 'min_value' in act_space and 'max_value' in act_space: min_value = act_space['min_value'] max_value = act_space['max_value'] if np.asarray( act_space['min_value']).shape != act_space['shape']: min_value = np.full(act_space['shape'], act_space['min_value']) max_value = np.full(act_space['shape'], act_space['max_value']) elif 'type' in act_space: if isinstance(act_space['type'], bool) or act_space['type'] == 'bool': min_value = np.full(act_space['shape'], 0) max_value = np.full(act_space['shape'], 1) return Space(min_value, max_value, True) else: min_value = np.full(act_space['shape'], 0) max_value = np.full(act_space['shape'], 1) else: min_value = np.full(act_space['shape'], 0) max_value = np.full(act_space['shape'], 1) return Space(min_value, max_value, False) elif 'gymtpl0' in act_space: values = [] for i in range(len(act_space)): key = 'gymtpl' + str(i) values.append(self.get_action_space(act_space[key])) return tuple(values) else: raise ValueError( 'Action space {} is not supported !'.format(act_space))
def get_action_space(self): from fruit.types.priv import Space return tuple([Space(0.0, 1.0, False), Space(-1.0, 1.0, False), Space(0.0, 1.0, False), Space(0, 1, True), Space(0, 1, True), Space(0, 1, True), Space(0, 2, True)])
def get_state_space(self, st_space=None): from fruit.types.priv import Space if st_space is None: st_space = self.environment.states() print(st_space) if 'num_values' in st_space: return Space(0, st_space['num_values'] - 1, True) elif 'shape' in st_space: if self.processor is None: shape = st_space['shape'] else: shape = self.current_state.shape min_value = np.zeros(shape) max_value = np.full(shape, 1.) return Space(min_value, max_value, True) elif 'gymtpl0' in st_space: values = [] for i in range(len(st_space)): key = 'gymtpl' + str(i) values.append(self.get_state_space(st_space[key])) return tuple(values) else: raise ValueError('State space {} is not supported !'.format(st_space))
def get_action_space(self): if self.__action_reduction >= 1: return Space(0, self.__action_reduction - 1, True) else: return Space(0, len(self.__action_set) - 1, True)
def get_action_space(self): return Space.convert_openai_space(self.env.action_space)
def get_action_space(self): from fruit.types.priv import Space return Space(0, len(self.game.get_action_space()) - 1, True)
def get_state_space(self): from fruit.types.priv import Space shape = (20, 1) min_value = np.zeros(shape) max_value = np.full(shape, 100) return Space(min_value, max_value, True)
def get_action_space(self): from fruit.types.priv import Space action_spec = self.env.action_spec() min_values = [d['min'] for d in action_spec] max_values = [d['max'] for d in action_spec] return Space(min_values, max_values, True)
def get_state_space(self): from fruit.types.priv import Space return Space(0, 3**(self.size * self.size), True)
def get_state_space(self): if self.graphical_state: return [self.screen_width, self.screen_height] else: from fruit.types.priv import Space return Space(0, self.num_of_rows * self.num_of_columns - 1, True)
def get_action_space(self): from fruit.types.priv import Space return Space(0, 5, True)
def get_state_space(self): from fruit.types.priv import Space shape = (self.width, self.height, 3) min_value = np.zeros(shape) max_value = np.full(shape, 255) return Space(min_value, max_value, True)
def get_state_space(self): if self.graphical_state: return [self.screen_size, self.screen_size] else: from fruit.types.priv import Space return Space(0, self.max_states_per_dim * self.max_states_per_dim - 1, True)