ddqn_agent.py

import numpy as np
import random
from collections import namedtuple, deque

from model import QNetwork, Dual_QNetwork

import torch
import torch.nn.functional as F
import torch.optim as optim


device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")


class Agent():
    def __init__(self, state_size, action_size, seed, lr, buffer_size, batch_size, update_step,
                 gamma, tau, dual_network=False):

        self.state_size = state_size
        self.action_size = action_size
        self.seed = random.seed(seed)

        # Q-Network
        if dual_network == False:
            self.qnetwork_local = QNetwork(state_size, action_size, seed).to(device)
            self.qnetwork_target = QNetwork(state_size, action_size, seed).to(device)
        else:
            self.qnetwork_local = Dual_QNetwork(state_size, action_size, seed).to(device)
            self.qnetwork_target = Dual_QNetwork(state_size, action_size, seed).to(device)
        self.optimizer = optim.Adam(self.qnetwork_local.parameters(), lr=lr)

        # Replay memory
        self.memory = ReplayBuffer(action_size, buffer_size, batch_size, seed)
        # Initialize time step (for updating every UPDATE_EVERY steps)
        self.t_step = 0

        self.update_step = update_step
        self.batch_size = batch_size
        self.gamma = gamma
        self.tau = tau

    def step(self, state, action, reward, next_state, done):
        # Save experience in replay memory
        self.memory.add(state, action, reward, next_state, done)

        self.t_step = (self.t_step + 1) % self.update_step
        if self.t_step == 0:
            # If enough samples are available in memory, get random subset and learn
            if len(self.memory) > self.batch_size:
                experiences = self.memory.sample()
                self.learn(experiences, self.gamma)

    def act(self, state, eps=0.):
        state = torch.from_numpy(state).float().unsqueeze(0).to(device)
        self.qnetwork_local.eval()
        with torch.no_grad():
            action_values = self.qnetwork_local(state)
        self.qnetwork_local.train()

        # Epsilon-greedy action selection
        if random.random() > eps:
            return np.argmax(action_values.cpu().data.numpy())
        else:
            return random.choice(np.arange(self.action_size))

    def learn(self, experiences, gamma):
        states, actions, rewards, next_states, dones = experiences

        current_values = self.qnetwork_local(states).gather(1, actions)

        target_action_idx = torch.max(self.qnetwork_local(next_states), 1)[1].unsqueeze(1)
        target_values = rewards + (gamma * self.qnetwork_target(next_states).gather(1, target_action_idx) * (1 - dones))

        # Compute loss
        loss = F.mse_loss(current_values, target_values)
        # Minimize the loss
        self.optimizer.zero_grad()
        loss.backward()
        self.optimizer.step()

        # ------------------- update target network ------------------- #
        self.soft_update(self.qnetwork_local, self.qnetwork_target, self.tau)

    def soft_update(self, local_model, target_model, tau):
        for target_param, local_param in zip(target_model.parameters(), local_model.parameters()):
            target_param.data.copy_(tau * local_param.data + (1.0 - tau) * target_param.data)


class ReplayBuffer:
    def __init__(self, action_size, buffer_size, batch_size, seed):
        self.action_size = action_size
        self.memory = deque(maxlen=buffer_size)
        self.batch_size = batch_size
        self.experience = namedtuple("Experience", field_names=["state", "action", "reward", "next_state", "done"])
        self.seed = random.seed(seed)

    def add(self, state, action, reward, next_state, done):
        e = self.experience(state, action, reward, next_state, done)
        self.memory.append(e)

    def sample(self):
        experiences = random.sample(self.memory, k=self.batch_size)
        states = torch.from_numpy(np.vstack([e.state for e in experiences if e is not None])).float().to(device)
        actions = torch.from_numpy(np.vstack([e.action for e in experiences if e is not None])).long().to(device)
        rewards = torch.from_numpy(np.vstack([e.reward for e in experiences if e is not None])).float().to(device)
        next_states = torch.from_numpy(np.vstack([e.next_state for e in experiences if e is not None])).float().to(
            device)
        dones = torch.from_numpy(np.vstack([e.done for e in experiences if e is not None]).astype(np.uint8)).float().to(
            device)

        return (states, actions, rewards, next_states, dones)

    def __len__(self):
        return len(self.memory)