model.py

import torch
from torch import nn 
from torch.distributions import Categorical

class Model(nn.Module):
    def __init__ (self,env):
        super().__init__()

        # Define the networks for actor and critic
        self.actor = nn.Sequential(
            nn.Conv2d(in_channels=4, out_channels=32, kernel_size=8, stride=4),
            nn.ReLU(),
            nn.Conv2d(in_channels=32, out_channels=64, kernel_size=4, stride=2),
            nn.ReLU(),
            nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, stride=1),
            nn.ReLU(),
            nn.Flatten(),
            nn.Linear(3136, 512),
            nn.ReLU(),
            nn.Linear(512, env.action_space.n)
        )

        self.critic = nn.Sequential(
            nn.Conv2d(in_channels=4, out_channels=32, kernel_size=8, stride=4),
            nn.ReLU(),
            nn.Conv2d(in_channels=32, out_channels=64, kernel_size=4, stride=2),
            nn.ReLU(),
            nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, stride=1),
            nn.ReLU(),
            nn.Flatten(),
            nn.Linear(3136, 512),
            nn.ReLU(),
            nn.Linear(512, 1)
        )

    # Forward pass
    def forward(self, obs):
        return Categorical(logits = self.actor(obs)), self.critic(obs).reshape(-1) # Return the actor logits and critic value