Get sb3 checkpoints saving during training

src/primaite/config/_package_data/example_config.yaml

@@ -2,7 +2,7 @@ training_config:
   rl_framework: SB3
   rl_algorithm: PPO
   seed: 333
-  n_learn_steps: 2560
+  n_learn_episodes: 25
   n_eval_episodes: 5
   max_steps_per_episode: 128
   deterministic_eval: false
@@ -10,6 +10,10 @@ training_config:
   agent_references:
     - defender
 
+io_settings:
+  save_checkpoints: true
+  checkpoint_interval: 5
+
 
 game_config:
   ports:

src/primaite/game/io.py

@@ -32,8 +32,8 @@ class SessionIO:
     """
 
     def __init__(self, settings: SessionIOSettings = SessionIOSettings()) -> None:
-        self.settings = settings
-        self.session_path = self.generate_session_path()
+        self.settings: SessionIOSettings = settings
+        self.session_path: Path = self.generate_session_path()
 
     def generate_session_path(self, timestamp: Optional[datetime] = None) -> Path:
         """Create a folder for the session and return the path to it."""

src/primaite/game/policy/policy.py

@@ -45,12 +45,12 @@ class PolicyABC(ABC):
         """Reference to the session."""
 
     @abstractmethod
-    def learn(self, n_episodes: int, n_time_steps: int) -> None:
+    def learn(self, n_episodes: int, timesteps_per_episode: int) -> None:
         """Train the agent."""
         pass
 
     @abstractmethod
-    def eval(self, n_episodes: int, n_time_steps: int, deterministic: bool) -> None:
+    def eval(self, n_episodes: int, timesteps_per_episode: int, deterministic: bool) -> None:
         """Evaluate the agent."""
         pass
 

src/primaite/game/policy/sb3.py

@@ -4,6 +4,7 @@ from typing import Literal, Optional, TYPE_CHECKING, Union
 
 from stable_baselines3 import A2C, PPO
 from stable_baselines3.a2c import MlpPolicy as A2C_MLP
+from stable_baselines3.common.callbacks import CheckpointCallback
 from stable_baselines3.common.evaluation import evaluate_policy
 from stable_baselines3.ppo import MlpPolicy as PPO_MLP
 
@@ -36,9 +37,17 @@ class SB3Policy(PolicyABC, identifier="SB3"):
             seed=seed,
         )
 
-    def learn(self, n_time_steps: int) -> None:
+    def learn(self, n_episodes: int, timesteps_per_episode: int) -> None:
         """Train the agent."""
-        self._agent.learn(total_timesteps=n_time_steps)
+        if self.session.io_manager.settings.save_checkpoints:
+            checkpoint_callback = CheckpointCallback(
+                save_freq=timesteps_per_episode * self.session.io_manager.settings.checkpoint_interval,
+                save_path=self.session.io_manager.generate_model_save_path("sb3"),
+                name_prefix="sb3_model",
+            )
+        else:
+            checkpoint_callback = None
+        self._agent.learn(total_timesteps=n_episodes * timesteps_per_episode, callback=checkpoint_callback)
 
     def eval(self, n_episodes: int, deterministic: bool) -> None:
         """Evaluate the agent."""
@@ -60,12 +69,10 @@ class SB3Policy(PolicyABC, identifier="SB3"):
         Therefore, this method is only used to save the final model.
         """
         self._agent.save(save_path)
-        pass
 
-    def load(self) -> None:
+    def load(self, model_path: Path) -> None:
         """Load agent from a checkpoint."""
-        self._agent_class.load("temp/path/to/save.pth", env=self.session.env)
-        pass
+        self._agent = self._agent_class.load(model_path, env=self.session.env)
 
     def close(self) -> None:
         """Close the agent."""

src/primaite/game/session.py

@@ -112,11 +112,12 @@ class TrainingOptions(BaseModel):
 
     rl_framework: Literal["SB3", "RLLIB"]
     rl_algorithm: Literal["PPO", "A2C"]
-    seed: Optional[int]
-    n_learn_steps: int
+    n_learn_episodes: int
     n_eval_episodes: Optional[int] = None
     max_steps_per_episode: int
+    # checkpoint_freq: Optional[int] = None
     deterministic_eval: bool
+    seed: Optional[int]
     n_agents: int
     agent_references: List[str]
 
@@ -188,13 +189,18 @@ class PrimaiteSession:
     def start_session(self) -> None:
         """Commence the training session."""
         self.mode = SessionMode.TRAIN
-        self.training_progress_bar = progress_bar_manager.counter(
-            total=self.training_options.n_learn_steps, desc="Training steps"
-        )
-        n_learn_steps = self.training_options.n_learn_steps
+        n_learn_episodes = self.training_options.n_learn_episodes
         n_eval_episodes = self.training_options.n_eval_episodes
+        max_steps_per_episode = self.training_options.max_steps_per_episode
+        self.training_progress_bar = progress_bar_manager.counter(
+            total=n_learn_episodes * max_steps_per_episode, desc="Training steps"
+        )
+
         deterministic_eval = self.training_options.deterministic_eval
-        self.policy.learn(n_time_steps=n_learn_steps)
+        self.policy.learn(
+            n_episodes=n_learn_episodes,
+            timesteps_per_episode=max_steps_per_episode,
+        )
         self.save_models()
 
         self.mode = SessionMode.EVAL
@@ -513,6 +519,6 @@ class PrimaiteSession:
 
         # READ IO SETTINGS
         io_settings = cfg.get("io_settings", {})
-        sess.io_manager.settings = SessionIO(settings=SessionIOSettings(**io_settings))
+        sess.io_manager = SessionIO(settings=SessionIOSettings(**io_settings))
 
         return sess