Oh, I could start using git...

example/config.yaml

@@ -0,0 +1,25 @@
+policy:
+  input_dim: 4
+  output_dim: 2
+  hidden_sizes: [64, 64]
+
+ppo:
+  learning_rate: 3e-4
+  n_steps: 2048
+  batch_size: 64
+  n_epochs: 10
+  gamma: 0.99
+  gae_lambda: 0.95
+  clip_range: 0.2
+  total_timesteps: 1000000
+  eval_interval: 2048
+  eval_deterministic: true
+  eval_episodes: 10
+  seed: 42
+
+loggers:
+  - type: terminal
+  - type: wandb
+    project: "PPO_project"
+    entity: "your_entity"
+    push_interval: 10

example/example.py

@@ -0,0 +1,37 @@
+import yaml
+import torch
+from fancy_rl.ppo import PPO
+from fancy_rl.policy import Policy
+from fancy_rl.loggers import TerminalLogger, WandbLogger
+import gymnasium as gym
+
+def main(config_file):
+    with open(config_file, 'r') as file:
+        config = yaml.safe_load(file)
+
+    env_fn = lambda: gym.make("CartPole-v1")
+    env = env_fn()
+
+    policy_config = config['policy']
+    policy = Policy(env=env, hidden_sizes=policy_config['hidden_sizes'])
+
+    ppo_config = config['ppo']
+    loggers_config = config['loggers']
+
+    loggers = []
+    for logger_config in loggers_config:
+        logger_type = logger_config.pop('type')
+        if logger_type == 'terminal':
+            loggers.append(TerminalLogger(**logger_config))
+        elif logger_type == 'wandb':
+            loggers.append(WandbLogger(**logger_config))
+
+    ppo = PPO(policy=policy,
+              env_fn=env_fn,
+              loggers=loggers,
+              **ppo_config)
+
+    ppo.train()
+
+if __name__ == "__main__":
+    main("example/config.yaml")