def __init__(self,
network: Any,
env: Any,
create_model: bool = True,
batch_size: int = 64,
gamma: float = 0.99,
policy_layers: Tuple = (64, 64),
value_layers: Tuple = (64, 64),
lr_policy: float = 0.0001,
lr_value: float = 0.001,
**kwargs):
self.network = network
self.env = env
self.create_model = create_model
self.batch_size = batch_size
self.gamma = gamma
self.policy_layers = policy_layers
self.value_layers = value_layers
self.lr_policy = lr_policy
self.lr_value = lr_value
self.seed = kwargs["seed"] if "seed" in kwargs else None
self.render = kwargs["render"] if "render" in kwargs else False
# Assign device
device = kwargs["device"] if "device" in kwargs else "cpu"
if "cuda" in device and torch.cuda.is_available():
self.device = torch.device(device)
else:
self.device = torch.device("cpu")
# Assign seed
if self.seed is not None:
set_seeds(self.seed, self.env)
def __init__(
self,
agent: Any,
env: Union[gym.Env, VecEnv],
log_mode: List[str] = ["stdout"],
log_key: str = "timestep",
log_interval: int = 10,
logdir: str = "logs",
epochs: int = 50,
max_timesteps: int = None,
off_policy: bool = False,
save_interval: int = 0,
save_model: str = "checkpoints",
run_num: int = None,
load_weights: str = None,
load_hyperparams: str = None,
render: bool = False,
evaluate_episodes: int = 25,
seed: Optional[int] = None,
):
self.agent = agent
self.env = env
self.log_mode = log_mode
self.log_key = log_key
self.log_interval = log_interval
self.logdir = logdir
self.epochs = epochs
self.max_timesteps = max_timesteps
self.off_policy = off_policy
self.save_interval = save_interval
self.save_model = save_model
self.run_num = run_num
self.load_weights = load_weights
self.load_hyperparams = load_hyperparams
self.render = render
self.evaluate_episodes = evaluate_episodes
if seed is not None:
set_seeds(seed, self.env)
self.logger = Logger(logdir=logdir, formats=[*log_mode])
def test_set_seeds(self):
set_seeds(42)
sampled = random.sample([i for i in range(20)], 1)[0]
assert sampled == 3
