Method __init__

1-grid-world/6-reinforce.py:33–41

(self, state_size, action_size)

# Softmax is applied where we need probabilities (sampling / log-prob).
class PolicyNetwork(nn.Module):
    def __init__(self, state_size, action_size):
        super().__init__()
        self.net = nn.Sequential(
            nn.Linear(state_size, 24),
            nn.ReLU(),
            nn.Linear(24, 24),
            nn.ReLU(),
            nn.Linear(24, action_size),
        )

    def forward(self, x):
        return self.net(x)