added torch training scripts

CustomPreprocessor.py → RLLib_training/CustomPreprocessor.py

@@ -36,7 +36,7 @@ def norm_obs_clip(obs, clip_min=-1, clip_max=1):
     :param obs: Observation that should be normalized
     :param clip_min: min value where observation will be clipped
     :param clip_max: max value where observation will be clipped
-    :return: returnes normalized and clipped observatoin
+    :return: returns normalized and clipped observation
     """
     max_obs = max(1, max_lt(obs, 1000))
     min_obs = max(0, min_lt(obs, 0))
@@ -53,7 +53,11 @@ class CustomPreprocessor(Preprocessor):
         return (105,)
 
     def transform(self, observation):
-        return norm_obs_clip(observation)  # return the preprocessed observation
+        if len(observation) == 105:
+            return norm_obs_clip(observation)
+        else:
+            return observation
+
 
 
 

torch_training/dueling_double_dqn.py

+import copy
+import os
+import random
+from collections import namedtuple, deque, Iterable
+
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.optim as optim
+
+from baselines.torch_training.model import QNetwork, QNetwork2
+
+BUFFER_SIZE = int(1e5)  # replay buffer size
+BATCH_SIZE = 512  # minibatch size
+GAMMA = 0.99  # discount factor 0.99
+TAU = 1e-3  # for soft update of target parameters
+LR = 0.5e-4  # learning rate 5
+UPDATE_EVERY = 10  # how often to update the network
+double_dqn = True  # If using double dqn algorithm
+input_channels = 5  # Number of Input channels
+
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+device = torch.device("cpu")
+print(device)
+
+
+class Agent:
+    """Interacts with and learns from the environment."""
+
+    def __init__(self, state_size, action_size, net_type, seed, double_dqn=True, input_channels=5):
+        """Initialize an Agent object.
+
+        Params
+        ======
+            state_size (int): dimension of each state
+            action_size (int): dimension of each action
+            seed (int): random seed
+        """
+        self.state_size = state_size
+        self.action_size = action_size
+        self.seed = random.seed(seed)
+        self.version = net_type
+        self.double_dqn = double_dqn
+        # Q-Network
+        if self.version == "Conv":
+            self.qnetwork_local = QNetwork2(state_size, action_size, seed, input_channels).to(device)
+            self.qnetwork_target = copy.deepcopy(self.qnetwork_local)
+        else:
+            self.qnetwork_local = QNetwork(state_size, action_size, seed).to(device)
+            self.qnetwork_target = copy.deepcopy(self.qnetwork_local)
+
+        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
+
+    def save(self, filename):
+        torch.save(self.qnetwork_local.state_dict(), filename + ".local")
+        torch.save(self.qnetwork_target.state_dict(), filename + ".target")
+
+    def load(self, filename):
+        if os.path.exists(filename + ".local"):
+            self.qnetwork_local.load_state_dict(torch.load(filename + ".local"))
+        if os.path.exists(filename + ".target"):
+            self.qnetwork_target.load_state_dict(torch.load(filename + ".target"))
+
+    def step(self, state, action, reward, next_state, done, train=True):
+        # Save experience in replay memory
+        self.memory.add(state, action, reward, next_state, done)
+
+        # Learn every UPDATE_EVERY time steps.
+        self.t_step = (self.t_step + 1) % UPDATE_EVERY
+        if self.t_step == 0:
+            # If enough samples are available in memory, get random subset and learn
+            if len(self.memory) > BATCH_SIZE:
+                experiences = self.memory.sample()
+                if train:
+                    self.learn(experiences, GAMMA)
+
+    def act(self, state, eps=0.):
+        """Returns actions for given state as per current policy.
+
+        Params
+        ======
+            state (array_like): current state
+            eps (float): epsilon, for epsilon-greedy action selection
+        """
+        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):
+
+        """Update value parameters using given batch of experience tuples.
+
+        Params
+        ======
+            experiences (Tuple[torch.Tensor]): tuple of (s, a, r, s', done) tuples
+            gamma (float): discount factor
+        """
+        states, actions, rewards, next_states, dones = experiences
+
+        # Get expected Q values from local model
+        Q_expected = self.qnetwork_local(states).gather(1, actions)
+
+        if self.double_dqn:
+            # Double DQN
+            q_best_action = self.qnetwork_local(next_states).max(1)[1]
+            Q_targets_next = self.qnetwork_target(next_states).gather(1, q_best_action.unsqueeze(-1))
+        else:
+            # DQN
+            Q_targets_next = self.qnetwork_target(next_states).detach().max(1)[0].unsqueeze(-1)
+
+            # Compute Q targets for current states
+
+        Q_targets = rewards + (gamma * Q_targets_next * (1 - dones))
+
+        # Compute loss
+        loss = F.mse_loss(Q_expected, Q_targets)
+        # Minimize the loss
+        self.optimizer.zero_grad()
+        loss.backward()
+        self.optimizer.step()
+
+        # ------------------- update target network ------------------- #
+        self.soft_update(self.qnetwork_local, self.qnetwork_target, TAU)
+
+    def soft_update(self, local_model, target_model, tau):
+        """Soft update model parameters.
+        θ_target = τ*θ_local + (1 - τ)*θ_target
+
+        Params
+        ======
+            local_model (PyTorch model): weights will be copied from
+            target_model (PyTorch model): weights will be copied to
+            tau (float): interpolation parameter
+        """
+        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:
+    """Fixed-size buffer to store experience tuples."""
+
+    def __init__(self, action_size, buffer_size, batch_size, seed):
+        """Initialize a ReplayBuffer object.
+
+        Params
+        ======
+            action_size (int): dimension of each action
+            buffer_size (int): maximum size of buffer
+            batch_size (int): size of each training batch
+            seed (int): random 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):
+        """Add a new experience to memory."""
+        e = self.experience(np.expand_dims(state, 0), action, reward, np.expand_dims(next_state, 0), done)
+        self.memory.append(e)
+
+    def sample(self):
+        """Randomly sample a batch of experiences from memory."""
+        experiences = random.sample(self.memory, k=self.batch_size)
+
+        states = torch.from_numpy(self.__v_stack_impr([e.state for e in experiences if e is not None])) \
+            .float().to(device)
+        actions = torch.from_numpy(self.__v_stack_impr([e.action for e in experiences if e is not None])) \
+            .long().to(device)
+        rewards = torch.from_numpy(self.__v_stack_impr([e.reward for e in experiences if e is not None])) \
+            .float().to(device)
+        next_states = torch.from_numpy(self.__v_stack_impr([e.next_state for e in experiences if e is not None])) \
+            .float().to(device)
+        dones = torch.from_numpy(self.__v_stack_impr([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 the current size of internal memory."""
+        return len(self.memory)
+
+    def __v_stack_impr(self, states):
+        sub_dim = len(states[0][0]) if isinstance(states[0], Iterable) else 1
+        np_states = np.reshape(np.array(states), (len(states), sub_dim))
+        return np_states

torch_training/model.py

+import torch.nn as nn
+import torch.nn.functional as F
+
+
+class QNetwork(nn.Module):
+    def __init__(self, state_size, action_size, seed, hidsize1=128, hidsize2=128):
+        super(QNetwork, self).__init__()
+
+        self.fc1_val = nn.Linear(state_size, hidsize1)
+        self.fc2_val = nn.Linear(hidsize1, hidsize2)
+        self.fc3_val = nn.Linear(hidsize2, 1)
+
+        self.fc1_adv = nn.Linear(state_size, hidsize1)
+        self.fc2_adv = nn.Linear(hidsize1, hidsize2)
+        self.fc3_adv = nn.Linear(hidsize2, action_size)
+
+    def forward(self, x):
+        val = F.relu(self.fc1_val(x))
+        val = F.relu(self.fc2_val(val))
+        val = self.fc3_val(val)
+
+        # advantage calculation
+        adv = F.relu(self.fc1_adv(x))
+        adv = F.relu(self.fc2_adv(adv))
+        adv = self.fc3_adv(adv)
+        return val + adv - adv.mean()
+
+
+class QNetwork2(nn.Module):
+    def __init__(self, state_size, action_size, seed, input_channels, hidsize1=128, hidsize2=64):
+        super(QNetwork2, self).__init__()
+        self.conv1 = nn.Conv2d(input_channels, 16, kernel_size=3, stride=1)
+        self.bn1 = nn.BatchNorm2d(16)
+        self.conv2 = nn.Conv2d(16, 32, kernel_size=5, stride=3)
+        self.bn2 = nn.BatchNorm2d(32)
+        self.conv3 = nn.Conv2d(32, 64, kernel_size=5, stride=3)
+        self.bn3 = nn.BatchNorm2d(64)
+
+        self.fc1_val = nn.Linear(6400, hidsize1)
+        self.fc2_val = nn.Linear(hidsize1, hidsize2)
+        self.fc3_val = nn.Linear(hidsize2, 1)
+
+        self.fc1_adv = nn.Linear(6400, hidsize1)
+        self.fc2_adv = nn.Linear(hidsize1, hidsize2)
+        self.fc3_adv = nn.Linear(hidsize2, action_size)
+
+    def forward(self, x):
+        x = F.relu(self.conv1(x))
+        x = F.relu(self.conv2(x))
+        x = F.relu(self.conv3(x))
+
+        # value function approximation
+        val = F.relu(self.fc1_val(x.view(x.size(0), -1)))
+        val = F.relu(self.fc2_val(val))
+        val = self.fc3_val(val)
+
+        # advantage calculation
+        adv = F.relu(self.fc1_adv(x.view(x.size(0), -1)))
+        adv = F.relu(self.fc2_adv(adv))
+        adv = self.fc3_adv(adv)
+        return val + adv - adv.mean()

torch_training/training_navigation.py

+import random
+from collections import deque
+
+import numpy as np
+import torch
+
+from baselines.torch_training.dueling_double_dqn import Agent
+from flatland.envs.generators import complex_rail_generator
+from flatland.envs.rail_env import RailEnv
+from flatland.utils.rendertools import RenderTool
+
+random.seed(1)
+np.random.seed(1)
+
+# Example generate a rail given a manual specification,
+# a map of tuples (cell_type, rotation)
+transition_probability = [15,  # empty cell - Case 0
+                          5,  # Case 1 - straight
+                          5,  # Case 2 - simple switch
+                          1,  # Case 3 - diamond crossing
+                          1,  # Case 4 - single slip
+                          1,  # Case 5 - double slip
+                          1,  # Case 6 - symmetrical
+                          0,  # Case 7 - dead end
+                          1,  # Case 1b (8)  - simple turn right
+                          1,  # Case 1c (9)  - simple turn left
+                          1]  # Case 2b (10) - simple switch mirrored
+
+# Example generate a random rail
+"""
+env = RailEnv(width=20,
+              height=20,
+              rail_generator=random_rail_generator(cell_type_relative_proportion=transition_probability),
+              number_of_agents=1)
+"""
+env = RailEnv(width=15,
+              height=15,
+              rail_generator=complex_rail_generator(nr_start_goal=10, nr_extra=10, min_dist=10, max_dist=99999, seed=0),
+              number_of_agents=5)
+
+"""
+env = RailEnv(width=20,
+              height=20,
+              rail_generator=rail_from_list_of_saved_GridTransitionMap_generator(
+                      ['../notebooks/temp.npy']),
+              number_of_agents=3)
+
+"""
+env_renderer = RenderTool(env, gl="QTSVG")
+handle = env.get_agent_handles()
+
+state_size = 105 * 2
+action_size = 4
+n_trials = 15000
+eps = 1.
+eps_end = 0.005
+eps_decay = 0.9995
+action_dict = dict()
+final_action_dict = dict()
+scores_window = deque(maxlen=100)
+done_window = deque(maxlen=100)
+time_obs = deque(maxlen=2)
+scores = []
+dones_list = []
+action_prob = [0] * 4
+agent_obs = [None] * env.get_num_agents()
+agent_next_obs = [None] * env.get_num_agents()
+agent = Agent(state_size, action_size, "FC", 0)
+agent.qnetwork_local.load_state_dict(torch.load('./baselines/torch_training/Nets/avoid_checkpoint15000.pth'))
+
+demo = True
+
+
+def max_lt(seq, val):
+    """
+    Return greatest item in seq for which item < val applies.
+    None is returned if seq was empty or all items in seq were >= val.
+    """
+    max = 0
+    idx = len(seq) - 1
+    while idx >= 0:
+        if seq[idx] < val and seq[idx] >= 0 and seq[idx] > max:
+            max = seq[idx]
+        idx -= 1
+    return max
+
+
+def min_lt(seq, val):
+    """
+    Return smallest item in seq for which item > val applies.
+    None is returned if seq was empty or all items in seq were >= val.
+    """
+    min = np.inf
+    idx = len(seq) - 1
+    while idx >= 0:
+        if seq[idx] > val and seq[idx] < min:
+            min = seq[idx]
+        idx -= 1
+    return min
+
+
+def norm_obs_clip(obs, clip_min=-1, clip_max=1):
+    """
+    This function returns the difference between min and max value of an observation
+    :param obs: Observation that should be normalized
+    :param clip_min: min value where observation will be clipped
+    :param clip_max: max value where observation will be clipped
+    :return: returnes normalized and clipped observatoin
+    """
+    max_obs = max(1, max_lt(obs, 1000))
+    min_obs = max(0, min_lt(obs, 0))
+    if max_obs == min_obs:
+        return np.clip(np.array(obs) / max_obs, clip_min, clip_max)
+    norm = np.abs(max_obs - min_obs)
+    if norm == 0:
+        norm = 1.
+    return np.clip((np.array(obs) - min_obs) / norm, clip_min, clip_max)
+
+
+for trials in range(1, n_trials + 1):
+
+    # Reset environment
+    obs = env.reset()
+
+    final_obs = obs.copy()
+    final_obs_next = obs.copy()
+
+    for a in range(env.get_num_agents()):
+        data, distance = env.obs_builder.split_tree(tree=np.array(obs[a]), num_features_per_node=5, current_depth=0)
+
+        data = norm_obs_clip(data)
+        distance = norm_obs_clip(distance)
+        obs[a] = np.concatenate((data, distance))
+
+    for i in range(2):
+        time_obs.append(obs)
+    # env.obs_builder.util_print_obs_subtree(tree=obs[0], num_elements_per_node=5)
+    for a in range(env.get_num_agents()):
+        agent_obs[a] = np.concatenate((time_obs[0][a], time_obs[1][a]))
+
+    score = 0
+    env_done = 0
+    # Run episode
+    for step in range(100):
+        if demo:
+            env_renderer.renderEnv(show=True)
+        # print(step)
+        # Action
+        for a in range(env.get_num_agents()):
+            if demo:
+                eps = 0
+            # action = agent.act(np.array(obs[a]), eps=eps)
+            action = agent.act(agent_obs[a])
+            action_prob[action] += 1
+            action_dict.update({a: action})
+
+        # Environment step
+        next_obs, all_rewards, done, _ = env.step(action_dict)
+        for a in range(env.get_num_agents()):
+            data, distance = env.obs_builder.split_tree(tree=np.array(next_obs[a]), num_features_per_node=5,
+                                                        current_depth=0)
+            data = norm_obs_clip(data)
+            distance = norm_obs_clip(distance)
+            next_obs[a] = np.concatenate((data, distance))
+
+        time_obs.append(next_obs)
+
+        # Update replay buffer and train agent
+        for a in range(env.get_num_agents()):
+            agent_next_obs[a] = np.concatenate((time_obs[0][a], time_obs[1][a]))
+
+            if done[a]:
+                final_obs[a] = agent_obs[a].copy()
+                final_obs_next[a] = agent_next_obs[a].copy()
+                final_action_dict.update({a: action_dict[a]})
+            if not demo and not done[a]:
+                agent.step(agent_obs[a], action_dict[a], all_rewards[a], agent_next_obs[a], done[a])
+            score += all_rewards[a]
+
+        agent_obs = agent_next_obs.copy()
+        if done['__all__']:
+            env_done = 1
+            for a in range(env.get_num_agents()):
+                agent.step(final_obs[a], final_action_dict[a], all_rewards[a], final_obs_next[a], done[a])
+            break
+    # Epsilon decay
+    eps = max(eps_end, eps_decay * eps)  # decrease epsilon
+
+    done_window.append(env_done)
+    scores_window.append(score)  # save most recent score
+    scores.append(np.mean(scores_window))
+    dones_list.append((np.mean(done_window)))
+
+    print('\rTraining {} Agents.\t Episode {}\t Average Score: {:.0f}\tDones: {:.2f}%' +
+          '\tEpsilon: {:.2f} \t Action Probabilities: \t {}'.format(
+              env.get_num_agents(),
+              trials,
+              np.mean(scores_window),
+              100 * np.mean(done_window),
+              eps, action_prob / np.sum(action_prob)), end=" ")
+
+    if trials % 100 == 0:
+        print(
+            '\rTraining {} Agents.\t Episode {}\t Average Score: {:.0f}\tDones: {:.2f}%' +
+            '\tEpsilon: {:.2f} \t Action Probabilities: \t {}'.format(
+                env.get_num_agents(),
+                trials,
+                np.mean(scores_window),
+                100 * np.mean(done_window),
+                eps,
+                action_prob / np.sum(action_prob)))
+        torch.save(agent.qnetwork_local.state_dict(),
+                   '../flatland/baselines/Nets/avoid_checkpoint' + str(trials) + '.pth')
+        action_prob = [1] * 4