---
tags:
- myenv-v1
- ppo
- deep-reinforcement-learning
- reinforcement-learning
- custom-implementation
- deep-rl-course
model-index:
- name: PPO
  results:
  - task:
      type: reinforcement-learning
      name: reinforcement-learning
    dataset:
      name: myenv-v1
      type: myenv-v1
    metrics:
    - type: mean_reward
      value: -1.10 +/- 0.00
      name: mean_reward
      verified: false
---

      # PPO Agent Playing myenv-v1
    
      This is a trained model of a PPO agent playing callosp.
      
      # Gameplay

      <video controls src="https://huggingface.co/MRNH/ppo-callofsp/resolve/main/replay.mp4"></video>
      
      # Hyperparameters
      ```python
      {'exp_name': 'ppo_no_pbt'
'seed': 1
'torch_deterministic': True
'cuda': True
'track': False
'wandb_project_name': 'cleanRL'
'wandb_entity': None
'capture_video': False
'env_id': 'myenv-v1'
'total_timesteps': 10000
'learning_rate': 0.00025
'num_envs': 1
'num_steps': 2048
'anneal_lr': True
'anneal_ent_coef': False
'anneal_clip_coef': False
'gae': True
'gamma': 0.99
'gae_lambda': 0.95
'num_minibatches': 64
'update_epochs': 3
'norm_adv': True
'clip_coef': 0.2
'clip_vloss': True
'ent_coef': 0.03
'vf_coef': 0.5
'max_grad_norm': 0.5
'target_kl': None
'repo_id': 'MRNH/ppo-callofsp'
'save_path': 'agent.pt'
'save_every': 10
'batch_size': 2048
'minibatch_size': 32}
      ```


Structure Actor-critic:


```
def layer_init(layer, std=np.sqrt(2), bias_const=0.0):
    torch.nn.init.orthogonal_(layer.weight, std)
    torch.nn.init.constant_(layer.bias, bias_const)
    return layer


class Agent(nn.Module):

    def __init__(self, envs):
        super().__init__()
        obs_dim = int(np.array(envs.single_observation_space.shape).prod())
        n_actions = envs.single_action_space.n

        self.critic = nn.Sequential(
            layer_init(nn.Linear(obs_dim, 64)),
            nn.Tanh(),
            layer_init(nn.Linear(64, 64)),
            nn.Tanh(),
            layer_init(nn.Linear(64, 1), std=1.0),
        )
        self.actor = nn.Sequential(
            layer_init(nn.Linear(obs_dim, 64)),
            nn.Tanh(),
            layer_init(nn.Linear(64, 64)),
            nn.Tanh(),
            layer_init(nn.Linear(64, n_actions), std=0.01),
        )

    def get_value(self, x):
        return self.critic(x)

    def get_action_and_value(self, x, action=None):
        logits = self.actor(x)
        probs = Categorical(logits=logits)
        if action is None:
            action = probs.sample()
        return action, probs.log_prob(action), probs.entropy(), self.critic(x)
```