td3.py

'''
Twin Delayed DDPG (TD3), if no twin no delayed then it's DDPG.
using target Q instead of V net: 2 Q net, 2 target Q net, 1 policy net, 1 target policy net
original paper: https://arxiv.org/pdf/1802.09477.pdf
'''
import math
import random

import gym
import numpy as np

import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from torch.distributions import Normal

from IPython.display import clear_output
import matplotlib.pyplot as plt
from matplotlib import animation
from IPython.display import display
from reacher import Reacher

import argparse
import time

torch.manual_seed(1234)  #Reproducibility

GPU = True
device_idx = 0
if GPU:
    device = torch.device("cuda:" + str(device_idx) if torch.cuda.is_available() else "cpu")
else:
    device = torch.device("cpu")
print(device)

parser = argparse.ArgumentParser(description='Train or test neural net motor controller.')
parser.add_argument('--train', dest='train', action='store_true', default=False)
parser.add_argument('--test', dest='test', action='store_true', default=False)

args = parser.parse_args()


class ReplayBuffer:
    def __init__(self, capacity):
        self.capacity = capacity
        self.buffer = []
        self.position = 0
    
    def push(self, state, action, reward, next_state, done):
        if len(self.buffer) < self.capacity:
            self.buffer.append(None)
        self.buffer[self.position] = (state, action, reward, next_state, done)
        self.position = int((self.position + 1) % self.capacity)  # as a ring buffer
    
    def sample(self, batch_size):
        batch = random.sample(self.buffer, batch_size)
        state, action, reward, next_state, done = map(np.stack, zip(*batch)) # stack for each element
        ''' 
        the * serves as unpack: sum(a,b) <=> batch=(a,b), sum(*batch) ;
        zip: a=[1,2], b=[2,3], zip(a,b) => [(1, 2), (2, 3)] ;
        the map serves as mapping the function on each list element: map(square, [2,3]) => [4,9] ;
        np.stack((1,2)) => array([1, 2])
        '''
        return state, action, reward, next_state, done
    
    def __len__(self):
        return len(self.buffer)

class NormalizedActions(gym.ActionWrapper):
    def _action(self, action):
        low  = self.action_space.low
        high = self.action_space.high
        
        action = low + (action + 1.0) * 0.5 * (high - low)
        action = np.clip(action, low, high)
        
        return action

    def _reverse_action(self, action):
        low  = self.action_space.low
        high = self.action_space.high
        
        action = 2 * (action - low) / (high - low) - 1
        action = np.clip(action, low, high)
        
        return action


class ValueNetwork(nn.Module):
    def __init__(self, state_dim, hidden_dim, init_w=3e-3):
        super(ValueNetwork, self).__init__()
        
        self.linear1 = nn.Linear(state_dim, hidden_dim)
        self.linear2 = nn.Linear(hidden_dim, hidden_dim)
        self.linear3 = nn.Linear(hidden_dim, hidden_dim)
        self.linear4 = nn.Linear(hidden_dim, 1)
        # weights initialization
        self.linear4.weight.data.uniform_(-init_w, init_w)
        self.linear4.bias.data.uniform_(-init_w, init_w)
        
    def forward(self, state):
        x = F.relu(self.linear1(state))
        x = F.relu(self.linear2(x))
        x = F.relu(self.linear3(x))
        x = self.linear4(x)
        return x
        
        
class QNetwork(nn.Module):
    def __init__(self, num_inputs, num_actions, hidden_size, init_w=3e-3):
        super(QNetwork, self).__init__()
        
        self.linear1 = nn.Linear(num_inputs + num_actions, hidden_size)
        self.linear2 = nn.Linear(hidden_size, hidden_size)
        self.linear3 = nn.Linear(hidden_size, hidden_size)
        self.linear4 = nn.Linear(hidden_size, 1)
        
        self.linear4.weight.data.uniform_(-init_w, init_w)
        self.linear4.bias.data.uniform_(-init_w, init_w)
        
    def forward(self, state, action):
        x = torch.cat([state, action], 1) # the dim 0 is number of samples
        x = F.relu(self.linear1(x))
        x = F.relu(self.linear2(x))
        x = F.relu(self.linear3(x))
        x = self.linear4(x)
        return x
        
        
class PolicyNetwork(nn.Module):
    def __init__(self, num_inputs, num_actions, hidden_size, action_range=1., init_w=3e-3, log_std_min=-20, log_std_max=2):
        super(PolicyNetwork, self).__init__()
        
        self.log_std_min = log_std_min
        self.log_std_max = log_std_max
        
        self.linear1 = nn.Linear(num_inputs, hidden_size)
        self.linear2 = nn.Linear(hidden_size, hidden_size)
        self.linear3 = nn.Linear(hidden_size, hidden_size)
        self.linear4 = nn.Linear(hidden_size, hidden_size)

        self.mean_linear = nn.Linear(hidden_size, num_actions)
        self.mean_linear.weight.data.uniform_(-init_w, init_w)
        self.mean_linear.bias.data.uniform_(-init_w, init_w)
        
        self.log_std_linear = nn.Linear(hidden_size, num_actions)
        self.log_std_linear.weight.data.uniform_(-init_w, init_w)
        self.log_std_linear.bias.data.uniform_(-init_w, init_w)

        self.action_range = action_range
        self.num_actions = num_actions

        
    def forward(self, state):
        x = F.relu(self.linear1(state))
        x = F.relu(self.linear2(x))
        x = F.relu(self.linear3(x))
        x = F.relu(self.linear4(x))

        mean  = F.tanh(self.mean_linear(x))
        # mean = F.leaky_relu(self.mean_linear(x))
        # mean = torch.clamp(mean, -30, 30)

        log_std = self.log_std_linear(x)
        log_std = torch.clamp(log_std, self.log_std_min, self.log_std_max) # clip the log_std into reasonable range
        
        return mean, log_std
    
    def evaluate(self, state, deterministic, eval_noise_scale, epsilon=1e-6):
        '''
        generate action with state as input wrt the policy network, for calculating gradients
        '''
        mean, log_std = self.forward(state)
        std = log_std.exp() # no clip in evaluation, clip affects gradients flow
        
        normal = Normal(0, 1)
        z      = normal.sample() 
        action_0 = torch.tanh(mean + std*z.to(device)) # TanhNormal distribution as actions; reparameterization trick
        action = self.action_range*mean if deterministic else self.action_range*action_0
        log_prob = Normal(mean, std).log_prob(mean+ std*z.to(device)) - torch.log(1. - action_0.pow(2) + epsilon) -  np.log(self.action_range)
        # both dims of normal.log_prob and -log(1-a**2) are (N,dim_of_action); 
        # the Normal.log_prob outputs the same dim of input features instead of 1 dim probability, 
        # needs sum up across the features dim to get 1 dim prob; or else use Multivariate Normal.
        log_prob = log_prob.sum(dim=1, keepdim=True)
        ''' add noise '''
        eval_noise_clip = 2*eval_noise_scale
        noise = normal.sample(action.shape) * eval_noise_scale
        noise = torch.clamp(
        noise,
        -eval_noise_clip,
        eval_noise_clip)
        action = action + noise.to(device)

        return action, log_prob, z, mean, log_std
        
    
    def get_action(self, state, deterministic, explore_noise_scale):
        '''
        generate action for interaction with env
        '''
        state = torch.FloatTensor(state).unsqueeze(0).to(device)
        mean, log_std = self.forward(state)
        std = log_std.exp()
        
        normal = Normal(0, 1)
        z      = normal.sample().to(device)
        
        action = mean.detach().cpu().numpy()[0] if deterministic else torch.tanh(mean + std*z).detach().cpu().numpy()[0]

        ''' add noise '''
        noise = normal.sample(action.shape) * explore_noise_scale
        action = self.action_range*action + noise.numpy()

        return action


    def sample_action(self,):
        a=torch.FloatTensor(self.num_actions).uniform_(-1, 1)
        return self.action_range*a.numpy()


class TD3_Trainer():
    def __init__(self, replay_buffer, hidden_dim, action_range, policy_target_update_interval=1):
        self.replay_buffer = replay_buffer


        self.q_net1 = QNetwork(state_dim, action_dim, hidden_dim).to(device)
        self.q_net2 = QNetwork(state_dim, action_dim, hidden_dim).to(device)
        self.target_q_net1 = QNetwork(state_dim, action_dim, hidden_dim).to(device)
        self.target_q_net2 = QNetwork(state_dim, action_dim, hidden_dim).to(device)
        self.policy_net = PolicyNetwork(state_dim, action_dim, hidden_dim, action_range).to(device)
        self.target_policy_net = PolicyNetwork(state_dim, action_dim, hidden_dim, action_range).to(device)
        print('Q Network (1,2): ', self.q_net1)
        print('Policy Network: ', self.policy_net)

        self.target_q_net1 = self.target_ini(self.q_net1, self.target_q_net1)
        self.target_q_net2 = self.target_ini(self.q_net2, self.target_q_net2)
        self.target_policy_net = self.target_ini(self.policy_net, self.target_policy_net)
        

        q_lr = 3e-4
        policy_lr = 3e-4
        self.update_cnt = 0
        self.policy_target_update_interval = policy_target_update_interval

        self.q_optimizer1 = optim.Adam(self.q_net1.parameters(), lr=q_lr)
        self.q_optimizer2 = optim.Adam(self.q_net2.parameters(), lr=q_lr)
        self.policy_optimizer = optim.Adam(self.policy_net.parameters(), lr=policy_lr)
    
    def target_ini(self, net, target_net):
        for target_param, param in zip(target_net.parameters(), net.parameters()):
            target_param.data.copy_(param.data)
        return target_net

    def target_soft_update(self, net, target_net, soft_tau):
    # Soft update the target net
        for target_param, param in zip(target_net.parameters(), net.parameters()):
            target_param.data.copy_(  # copy data value into target parameters
                target_param.data * (1.0 - soft_tau) + param.data * soft_tau
            )

        return target_net
    
    def update(self, batch_size, deterministic, eval_noise_scale, reward_scale=10., gamma=0.9,soft_tau=1e-2):
        state, action, reward, next_state, done = self.replay_buffer.sample(batch_size)
        # print('sample:', state, action,  reward, done)

        state      = torch.FloatTensor(state).to(device)
        next_state = torch.FloatTensor(next_state).to(device)
        action     = torch.FloatTensor(action).to(device)
        reward     = torch.FloatTensor(reward).unsqueeze(1).to(device)  # reward is single value, unsqueeze() to add one dim to be [reward] at the sample dim;
        done       = torch.FloatTensor(np.float32(done)).unsqueeze(1).to(device)

        predicted_q_value1 = self.q_net1(state, action)
        predicted_q_value2 = self.q_net2(state, action)
        new_action, log_prob, z, mean, log_std = self.policy_net.evaluate(state, deterministic, eval_noise_scale=0.0)  # no noise, deterministic policy gradients
        new_next_action, _, _, _, _ = self.target_policy_net.evaluate(next_state, deterministic, eval_noise_scale=eval_noise_scale) # clipped normal noise

        reward = reward_scale * (reward - reward.mean(dim=0)) / (reward.std(dim=0) + 1e-6) # normalize with batch mean and std; plus a small number to prevent numerical problem

    # Training Q Function
        target_q_min = torch.min(self.target_q_net1(next_state, new_next_action),self.target_q_net2(next_state, new_next_action))

        target_q_value = reward + (1 - done) * gamma * target_q_min # if done==1, only reward

        q_value_loss1 = ((predicted_q_value1 - target_q_value.detach())**2).mean()  # detach: no gradients for the variable
        q_value_loss2 = ((predicted_q_value2 - target_q_value.detach())**2).mean()
        self.q_optimizer1.zero_grad()
        q_value_loss1.backward()
        self.q_optimizer1.step()
        self.q_optimizer2.zero_grad()
        q_value_loss2.backward()
        self.q_optimizer2.step()

        if self.update_cnt%self.policy_target_update_interval==0:
        # This is the **Delayed** update of policy and all targets (for Q and policy). 
        # Training Policy Function
            ''' implementation 1 '''
            # predicted_new_q_value = torch.min(self.q_net1(state, new_action),self.q_net2(state, new_action))
            ''' implementation 2 '''
            predicted_new_q_value = self.q_net1(state, new_action)

            policy_loss = - predicted_new_q_value.mean()

            self.policy_optimizer.zero_grad()
            policy_loss.backward()
            self.policy_optimizer.step()
        
        # Soft update the target nets
            self.target_q_net1=self.target_soft_update(self.q_net1, self.target_q_net1, soft_tau)
            self.target_q_net2=self.target_soft_update(self.q_net2, self.target_q_net2, soft_tau)
            self.target_policy_net=self.target_soft_update(self.policy_net, self.target_policy_net, soft_tau)

        self.update_cnt+=1

        return predicted_q_value1.mean()

    def save_model(self, path):
        torch.save(self.q_net1.state_dict(), path+'_q1')
        torch.save(self.q_net2.state_dict(), path+'_q2')
        torch.save(self.policy_net.state_dict(), path+'_policy')

    def load_model(self, path):
        self.q_net1.load_state_dict(torch.load(path+'_q1'))
        self.q_net2.load_state_dict(torch.load(path+'_q2'))
        self.policy_net.load_state_dict(torch.load(path+'_policy'))
        self.q_net1.eval()
        self.q_net2.eval()
        self.policy_net.eval()

def plot(rewards):
    clear_output(True)
    plt.figure(figsize=(20,5))
    plt.plot(rewards)
    plt.savefig('td3.png')
    # plt.show()


# choose env
ENV = ['Reacher', 'Pendulum-v0', 'HalfCheetah-v2'][1]
if ENV == 'Reacher':
    NUM_JOINTS=2
    LINK_LENGTH=[200, 140]
    INI_JOING_ANGLES=[0.1, 0.1]
    SCREEN_SIZE=1000
    SPARSE_REWARD=False
    SCREEN_SHOT=False
    action_range = 10.0

    env=Reacher(screen_size=SCREEN_SIZE, num_joints=NUM_JOINTS, link_lengths = LINK_LENGTH, \
    ini_joint_angles=INI_JOING_ANGLES, target_pos = [369,430], render=True, change_goal=False)
    action_dim = env.num_actions
    state_dim  = env.num_observations
else:
    env = NormalizedActions(gym.make(ENV))
    action_dim = env.action_space.shape[0]
    state_dim  = env.observation_space.shape[0]
    action_range=1.

replay_buffer_size = 5e5
replay_buffer = ReplayBuffer(replay_buffer_size)


# hyper-parameters for RL training
max_episodes  = 1000
max_steps   = 20 if ENV ==  'Reacher' else 150  # Pendulum needs 150 steps per episode to learn well, cannot handle 20
frame_idx   = 0
batch_size  = 300
explore_steps = 0  # for random action sampling in the beginning of training
update_itr = 1
hidden_dim = 512
policy_target_update_interval = 3 # delayed update for the policy network and target networks
DETERMINISTIC=True  # DDPG: deterministic policy gradient
explore_noise_scale = 0.5  # 0.5 noise is required for Pendulum-v0, 0.1 noise for HalfCheetah-v2
eval_noise_scale = 0.5
reward_scale = 1.
rewards     = []
model_path = './model/td3'

td3_trainer=TD3_Trainer(replay_buffer, hidden_dim=hidden_dim, policy_target_update_interval=policy_target_update_interval, action_range=action_range )

if __name__ == '__main__':
    if args.train:
        # training loop
        for eps in range(max_episodes):
            if ENV == 'Reacher':
                state = env.reset(SCREEN_SHOT)
            else:
                state =  env.reset()
            episode_reward = 0
            
            for step in range(max_steps):
                if frame_idx > explore_steps:
                    action = td3_trainer.policy_net.get_action(state, deterministic = DETERMINISTIC, explore_noise_scale=explore_noise_scale)
                else:
                    action = td3_trainer.policy_net.sample_action()
                if ENV ==  'Reacher':
                    next_state, reward, done, _ = env.step(action, SPARSE_REWARD, SCREEN_SHOT)
                else:
                    next_state, reward, done, _ = env.step(action) 
                    # env.render()

                replay_buffer.push(state, action, reward, next_state, done)
                
                state = next_state
                episode_reward += reward
                frame_idx += 1
                
                if len(replay_buffer) > batch_size:
                    for i in range(update_itr):
                        _=td3_trainer.update(batch_size, deterministic=DETERMINISTIC, eval_noise_scale=eval_noise_scale, reward_scale=reward_scale)
                
                if done:
                    break
              
            if eps % 20 == 0 and eps>0:
                plot(rewards)
                np.save('rewards_td3', rewards)
                td3_trainer.save_model(model_path)

            print('Episode: ', eps, '| Episode Reward: ', episode_reward)
            rewards.append(episode_reward)
        td3_trainer.save_model(model_path)
        
    if args.test:
        td3_trainer.load_model(model_path)
        for eps in range(10):
            if ENV == 'Reacher':
                state = env.reset(SCREEN_SHOT)
            else:
                state =  env.reset()
                env.render()   
            episode_reward = 0

            for step in range(max_steps):
                action = td3_trainer.policy_net.get_action(state, deterministic = DETERMINISTIC, explore_noise_scale=0.0)
                if ENV ==  'Reacher':
                    next_state, reward, done, _ = env.step(action, SPARSE_REWARD, SCREEN_SHOT)
                else:
                    next_state, reward, done, _ = env.step(action)
                    env.render() 

                episode_reward += reward
                state=next_state

            print('Episode: ', eps, '| Episode Reward: ', episode_reward)