def create_policy(self, brain_parameters: BrainParameters) -> TFPolicy:
"""
Creates a PPO policy to trainers list of policies.
:param brain_parameters: specifications for policy construction
:return policy
"""
if self.multi_gpu and len(get_devices()) > 1:
policy: PPOPolicy = MultiGpuPPOPolicy(
self.seed,
brain_parameters,
self.trainer_parameters,
self.is_training,
self.load,
)
else:
policy = PPOPolicy(
self.seed,
brain_parameters,
self.trainer_parameters,
self.is_training,
self.load,
)
for _reward_signal in policy.reward_signals.keys():
self.collected_rewards[_reward_signal] = defaultdict(lambda: 0)
return policy
def __init__(
self,
brain,
reward_buff_cap,
trainer_parameters,
training,
load,
seed,
run_id,
multi_gpu,
):
"""
Responsible for collecting experiences and training PPO model.
:param trainer_parameters: The parameters for the trainer (dictionary).
:param reward_buff_cap: Max reward history to track in the reward buffer
:param training: Whether the trainer is set for training.
:param load: Whether the model should be loaded.
:param seed: The seed the model will be initialized with
:param run_id: The identifier of the current run
"""
super(PPOTrainer, self).__init__(brain, trainer_parameters, training,
run_id, reward_buff_cap)
self.param_keys = [
"batch_size",
"beta",
"buffer_size",
"epsilon",
"hidden_units",
"lambd",
"learning_rate",
"max_steps",
"normalize",
"num_epoch",
"num_layers",
"time_horizon",
"sequence_length",
"summary_freq",
"use_recurrent",
"summary_path",
"memory_size",
"model_path",
"reward_signals",
]
self.check_param_keys()
if multi_gpu and len(get_devices()) > 1:
self.ppo_policy = MultiGpuPPOPolicy(seed, brain,
trainer_parameters,
self.is_training, load)
else:
self.ppo_policy = PPOPolicy(seed, brain, trainer_parameters,
self.is_training, load)
self.policy = self.ppo_policy
for _reward_signal in self.policy.reward_signals.keys():
self.collected_rewards[_reward_signal] = {}