def __init__(
self,
brain: BrainParameters,
trainer_parameters: dict,
training: bool,
run_id: int,
reward_buff_cap: int = 1,
):
"""
Responsible for collecting experiences and training a neural network model.
:BrainParameters brain: Brain to be trained.
:dict trainer_parameters: The parameters for the trainer (dictionary).
:bool training: Whether the trainer is set for training.
:int run_id: The identifier of the current run
:int reward_buff_cap:
"""
self.param_keys: List[str] = []
self.brain_name = brain.brain_name
self.run_id = run_id
self.trainer_parameters = trainer_parameters
self.summary_path = trainer_parameters["summary_path"]
if not os.path.exists(self.summary_path):
os.makedirs(self.summary_path)
self.cumulative_returns_since_policy_update: List[float] = []
self.is_training = training
self.stats: Dict[str, List] = defaultdict(list)
self.trainer_metrics = TrainerMetrics(path=self.summary_path + ".csv",
brain_name=self.brain_name)
self.summary_writer = tf.summary.FileWriter(self.summary_path)
self._reward_buffer: Deque[float] = deque(maxlen=reward_buff_cap)
self.policy: Policy = None
def test_experience_collection_timer(self):
mock_path = 'fake'
mock_brain_name = 'fake'
trainer_metrics = TrainerMetrics(path=mock_path,
brain_name=mock_brain_name)
trainer_metrics.start_experience_collection_timer()
trainer_metrics.end_experience_collection_timer()
assert trainer_metrics.delta_last_experience_collection == 0
def test_policy_update_timer(self):
mock_path = 'fake'
mock_brain_name = 'fake'
fake_buffer_length = 350
fake_mean_return = 0.3
trainer_metrics = TrainerMetrics(path=mock_path,
brain_name=mock_brain_name)
trainer_metrics.start_experience_collection_timer()
trainer_metrics.end_experience_collection_timer()
trainer_metrics.start_policy_update_timer(number_experiences=fake_buffer_length,
mean_return=fake_mean_return)
trainer_metrics.end_policy_update()
fake_row = [mock_brain_name, 0,0, 0, 350, '0.300']
assert trainer_metrics.rows[0] == fake_row
def __init__(self, brain, trainer_parameters, training, run_id):
"""
Responsible for collecting experiences and training a neural network model.
:BrainParameters brain: Brain to be trained.
:dict trainer_parameters: The parameters for the trainer (dictionary).
:bool training: Whether the trainer is set for training.
:int run_id: The identifier of the current run
"""
self.param_keys = []
self.brain_name = brain.brain_name
self.run_id = run_id
self.trainer_parameters = trainer_parameters
self.summary_path = trainer_parameters['summary_path']
if not os.path.exists(self.summary_path):
os.makedirs(self.summary_path)
self.cumulative_returns_since_policy_update = []
self.is_training = training
self.stats = {}
self.trainer_metrics = TrainerMetrics(path=self.summary_path + '.csv',
brain_name=self.brain_name)
self.summary_writer = tf.summary.FileWriter(self.summary_path)
self.policy = None