OakEmu-RL v0.1.0

Appearance

PPOState ​

python
PPOState(model_state_dict: dict[str, Any], optimizer_state_dict: dict[str, Any], steps: int, learner_iterations: int, rollouts_collected: int, total_time: float, episode_totals: AgentEpisodeTotalsState = AgentEpisodeTotalsState(), total_rollbacks: int = 0, tester_episode_totals: AgentEpisodeTotalsState = AgentEpisodeTotalsState(), trajectory_state_dict: dict[str, Any] | None = None, run_name: str | None = None) -> None

Dataclass for storing the complete state of a PPO experiment.

Source

Values ​

episode_totals ​

python
episode_totals: AgentEpisodeTotalsState = field(default_factory=AgentEpisodeTotalsState)

Source

learner_iterations ​

python
learner_iterations: int

Source

model_state_dict ​

python
model_state_dict: dict[str, Any]

Source

optimizer_state_dict ​

python
optimizer_state_dict: dict[str, Any]

Source

rollouts_collected ​

python
rollouts_collected: int

Source

run_name ​

python
run_name: str | None = None

Source

steps ​

python
steps: int

Source

tester_episode_totals ​

python
tester_episode_totals: AgentEpisodeTotalsState = field(default_factory=AgentEpisodeTotalsState)

Source

total_rollbacks ​

python
total_rollbacks: int = 0

Source

total_time ​

python
total_time: float

Source

trajectory_state_dict ​

python
trajectory_state_dict: dict[str, Any] | None = None

Source