net.py

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.distributions import Categorical


class policy_net(nn.Module):
    def __init__(self, input_dim, output_dim):
        super(policy_net, self).__init__()
        self.input_dim = input_dim
        self.output_dim = output_dim
        self.fc1 = nn.Linear(self.input_dim, 128)
        self.fc2 = nn.Linear(128, 128)
        self.fc3 = nn.Linear(128, self.output_dim)

    def forward(self, input):
        x = F.relu(self.fc1(input))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return F.softmax(x, 1)

    def act(self, input):
        probs = self.forward(input)
        dist = Categorical(probs)
        action = dist.sample()
        action = action.detach().item()
        return action


class value_net(nn.Module):
    def __init__(self, input_dim, output_dim):
        super(value_net, self).__init__()
        self.input_dim = input_dim
        self.output_dim = output_dim

        self.fc1 = nn.Linear(self.input_dim, 128)
        self.fc2 = nn.Linear(128, 128)
        self.fc3 = nn.Linear(128, self.output_dim)

    def forward(self, input):
        x = F.relu(self.fc1(input))
        x = F.relu(self.fc2(x))
        x = self.fc3(x)
        return x