DQN C51/Rainbow

Copyright 2018 The TF-Agents Authors.

Introduction

This example shows how to train a Categorical DQN (C51) agent on the Cartpole environment using the TF-Agents library.

Make sure you take a look through the DQN tutorial as a prerequisite. This tutorial will assume familiarity with the DQN tutorial; it will mainly focus on the differences between DQN and C51.

Setup

If you haven't installed tf-agents yet, run:

sudo apt-get install -y xvfb ffmpeg
pip install -q 'gym==0.10.11'
pip install -q 'imageio==2.4.0'
pip install -q PILLOW
pip install -q 'pyglet==1.3.2'
pip install -q pyvirtualdisplay
pip install -q tf-agents

ffmpeg is already the newest version (7:3.4.8-0ubuntu0.2).
xvfb is already the newest version (2:1.19.6-1ubuntu4.4).
The following packages were automatically installed and are no longer required:
  dconf-gsettings-backend dconf-service dkms freeglut3 freeglut3-dev
  glib-networking glib-networking-common glib-networking-services
  gsettings-desktop-schemas libcairo-gobject2 libcolord2 libdconf1
  libegl1-mesa libepoxy0 libglu1-mesa libglu1-mesa-dev libgtk-3-0
  libgtk-3-common libice-dev libjansson4 libjson-glib-1.0-0
  libjson-glib-1.0-common libproxy1v5 librest-0.7-0 libsm-dev
  libsoup-gnome2.4-1 libsoup2.4-1 libxi-dev libxmu-dev libxmu-headers
  libxnvctrl0 libxt-dev linux-gcp-headers-5.0.0-1026
  linux-headers-5.0.0-1026-gcp linux-image-5.0.0-1026-gcp
  linux-modules-5.0.0-1026-gcp pkg-config policykit-1-gnome python3-xkit
  screen-resolution-extra xserver-xorg-core-hwe-18.04
Use 'sudo apt autoremove' to remove them.
0 upgraded, 0 newly installed, 0 to remove and 85 not upgraded.

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

import base64
import imageio
import IPython
import matplotlib
import matplotlib.pyplot as plt
import PIL.Image
import pyvirtualdisplay

import tensorflow as tf

from tf_agents.agents.categorical_dqn import categorical_dqn_agent
from tf_agents.drivers import dynamic_step_driver
from tf_agents.environments import suite_gym
from tf_agents.environments import tf_py_environment
from tf_agents.eval import metric_utils
from tf_agents.metrics import tf_metrics
from tf_agents.networks import categorical_q_network
from tf_agents.policies import random_tf_policy
from tf_agents.replay_buffers import tf_uniform_replay_buffer
from tf_agents.trajectories import trajectory
from tf_agents.utils import common

tf.compat.v1.enable_v2_behavior()


# Set up a virtual display for rendering OpenAI gym environments.
display = pyvirtualdisplay.Display(visible=0, size=(1400, 900)).start()

Hyperparameters

env_name = "CartPole-v1" # @param {type:"string"}
num_iterations = 15000 # @param {type:"integer"}

initial_collect_steps = 1000  # @param {type:"integer"} 
collect_steps_per_iteration = 1  # @param {type:"integer"}
replay_buffer_capacity = 100000  # @param {type:"integer"}

fc_layer_params = (100,)

batch_size = 64  # @param {type:"integer"}
learning_rate = 1e-3  # @param {type:"number"}
gamma = 0.99
log_interval = 200  # @param {type:"integer"}

num_atoms = 51  # @param {type:"integer"}
min_q_value = -20  # @param {type:"integer"}
max_q_value = 20  # @param {type:"integer"}
n_step_update = 2  # @param {type:"integer"}

num_eval_episodes = 10  # @param {type:"integer"}
eval_interval = 1000  # @param {type:"integer"}

Environment

Load the environment as before, with one for training and one for evaluation. Here we use CartPole-v1 (vs. CartPole-v0 in the DQN tutorial), which has a larger max reward of 500 rather than 200.

train_py_env = suite_gym.load(env_name)
eval_py_env = suite_gym.load(env_name)

train_env = tf_py_environment.TFPyEnvironment(train_py_env)
eval_env = tf_py_environment.TFPyEnvironment(eval_py_env)

Agent

C51 is a Q-learning algorithm based on DQN. Like DQN, it can be used on any environment with a discrete action space.

The main difference between C51 and DQN is that rather than simply predicting the Q-value for each state-action pair, C51 predicts a histogram model for the probability distribution of the Q-value:

By learning the distribution rather than simply the expected value, the algorithm is able to stay more stable during training, leading to improved final performance. This is particularly true in situations with bimodal or even multimodal value distributions, where a single average does not provide an accurate picture.

In order to train on probability distributions rather than on values, C51 must perform some complex distributional computations in order to calculate its loss function. But don't worry, all of this is taken care of for you in TF-Agents!

To create a C51 Agent, we first need to create a CategoricalQNetwork. The API of the CategoricalQNetwork is the same as that of the QNetwork, except that there is an additional argument num_atoms. This represents the number of support points in our probability distribution estimates. (The above image includes 10 support points, each represented by a vertical blue bar.) As you can tell from the name, the default number of atoms is 51.

categorical_q_net = categorical_q_network.CategoricalQNetwork(
    train_env.observation_spec(),
    train_env.action_spec(),
    num_atoms=num_atoms,
    fc_layer_params=fc_layer_params)

We also need an optimizer to train the network we just created, and a train_step_counter variable to keep track of how many times the network was updated.

Note that one other significant difference from vanilla DqnAgent is that we now need to specify min_q_value and max_q_value as arguments. These specify the most extreme values of the support (in other words, the most extreme of the 51 atoms on either side). Make sure to choose these appropriately for your particular environment. Here we use -20 and 20.

optimizer = tf.compat.v1.train.AdamOptimizer(learning_rate=learning_rate)

train_step_counter = tf.compat.v2.Variable(0)

agent = categorical_dqn_agent.CategoricalDqnAgent(
    train_env.time_step_spec(),
    train_env.action_spec(),
    categorical_q_network=categorical_q_net,
    optimizer=optimizer,
    min_q_value=min_q_value,
    max_q_value=max_q_value,
    n_step_update=n_step_update,
    td_errors_loss_fn=common.element_wise_squared_loss,
    gamma=gamma,
    train_step_counter=train_step_counter)
agent.initialize()

One last thing to note is that we also added an argument to use n-step updates with $n$ = 2. In single-step Q-learning ($n$ = 1), we only compute the error between the Q-values at the current time step and the next time step using the single-step return (based on the Bellman optimality equation). The single-step return is defined as:

$G_t = R_{t + 1} + \gamma V(s_{t + 1})$

where we define $V(s) = \max_a{Q(s, a)}$.

N-step updates involve expanding the standard single-step return function $n$ times:

$G_t^n = R_{t + 1} + \gamma R_{t + 2} + \gamma^2 R_{t + 3} + \dots + \gamma^n V(s_{t + n})$

N-step updates enable the agent to bootstrap from further in the future, and with the right value of $n$, this often leads to faster learning.

Although C51 and n-step updates are often combined with prioritized replay to form the core of the Rainbow agent, we saw no measurable improvement from implementing prioritized replay. Moreover, we find that when combining our C51 agent with n-step updates alone, our agent performs as well as other Rainbow agents on the sample of Atari environments we've tested.

Metrics and Evaluation

The most common metric used to evaluate a policy is the average return. The return is the sum of rewards obtained while running a policy in an environment for an episode, and we usually average this over a few episodes. We can compute the average return metric as follows.

def compute_avg_return(environment, policy, num_episodes=10):

  total_return = 0.0
  for _ in range(num_episodes):

    time_step = environment.reset()
    episode_return = 0.0

    while not time_step.is_last():
      action_step = policy.action(time_step)
      time_step = environment.step(action_step.action)
      episode_return += time_step.reward
    total_return += episode_return

  avg_return = total_return / num_episodes
  return avg_return.numpy()[0]


random_policy = random_tf_policy.RandomTFPolicy(train_env.time_step_spec(),
                                                train_env.action_spec())

compute_avg_return(eval_env, random_policy, num_eval_episodes)

# Please also see the metrics module for standard implementations of different
# metrics.

27.6

Data Collection

As in the DQN tutorial, set up the replay buffer and the initial data collection with the random policy.

replay_buffer = tf_uniform_replay_buffer.TFUniformReplayBuffer(
    data_spec=agent.collect_data_spec,
    batch_size=train_env.batch_size,
    max_length=replay_buffer_capacity)

def collect_step(environment, policy):
  time_step = environment.current_time_step()
  action_step = policy.action(time_step)
  next_time_step = environment.step(action_step.action)
  traj = trajectory.from_transition(time_step, action_step, next_time_step)

  # Add trajectory to the replay buffer
  replay_buffer.add_batch(traj)

for _ in range(initial_collect_steps):
  collect_step(train_env, random_policy)

# This loop is so common in RL, that we provide standard implementations of
# these. For more details see the drivers module.

# Dataset generates trajectories with shape [BxTx...] where
# T = n_step_update + 1.
dataset = replay_buffer.as_dataset(
    num_parallel_calls=3, sample_batch_size=batch_size,
    num_steps=n_step_update + 1).prefetch(3)

iterator = iter(dataset)

WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.6/site-packages/tensorflow/python/autograph/operators/control_flow.py:1004: ReplayBuffer.get_next (from tf_agents.replay_buffers.replay_buffer) is deprecated and will be removed in a future version.
Instructions for updating:
Use `as_dataset(..., single_deterministic_pass=False) instead.

Training the agent

The training loop involves both collecting data from the environment and optimizing the agent's networks. Along the way, we will occasionally evaluate the agent's policy to see how we are doing.

The following will take ~7 minutes to run.

try:
  %%time
except:
  pass

# (Optional) Optimize by wrapping some of the code in a graph using TF function.
agent.train = common.function(agent.train)

# Reset the train step
agent.train_step_counter.assign(0)

# Evaluate the agent's policy once before training.
avg_return = compute_avg_return(eval_env, agent.policy, num_eval_episodes)
returns = [avg_return]

for _ in range(num_iterations):

  # Collect a few steps using collect_policy and save to the replay buffer.
  for _ in range(collect_steps_per_iteration):
    collect_step(train_env, agent.collect_policy)

  # Sample a batch of data from the buffer and update the agent's network.
  experience, unused_info = next(iterator)
  train_loss = agent.train(experience)

  step = agent.train_step_counter.numpy()

  if step % log_interval == 0:
    print('step = {0}: loss = {1}'.format(step, train_loss.loss))

  if step % eval_interval == 0:
    avg_return = compute_avg_return(eval_env, agent.policy, num_eval_episodes)
    print('step = {0}: Average Return = {1:.2f}'.format(step, avg_return))
    returns.append(avg_return)

WARNING:tensorflow:From /tmpfs/src/tf_docs_env/lib/python3.6/site-packages/tensorflow/python/util/dispatch.py:201: calling foldr_v2 (from tensorflow.python.ops.functional_ops) with back_prop=False is deprecated and will be removed in a future version.
Instructions for updating:
back_prop=False is deprecated. Consider using tf.stop_gradient instead.
Instead of:
results = tf.foldr(fn, elems, back_prop=False)
Use:
results = tf.nest.map_structure(tf.stop_gradient, tf.foldr(fn, elems))
step = 200: loss = 3.344022750854492
step = 400: loss = 2.50473690032959
step = 600: loss = 2.1661343574523926
step = 800: loss = 1.963805913925171
step = 1000: loss = 1.8388065099716187
step = 1000: Average Return = 51.30
step = 1200: loss = 1.7277296781539917
step = 1400: loss = 1.5035001039505005
step = 1600: loss = 1.4892648458480835
step = 1800: loss = 1.425370693206787
step = 2000: loss = 1.7112807035446167
step = 2000: Average Return = 255.30
step = 2200: loss = 1.4020955562591553
step = 2400: loss = 1.0620348453521729
step = 2600: loss = 1.097517490386963
step = 2800: loss = 1.3673020601272583
step = 3000: loss = 0.9485558867454529
step = 3000: Average Return = 280.10
step = 3200: loss = 1.0419788360595703
step = 3400: loss = 1.2028640508651733
step = 3600: loss = 1.0199285745620728
step = 3800: loss = 0.7367687225341797
step = 4000: loss = 1.0523639917373657
step = 4000: Average Return = 291.70
step = 4200: loss = 0.7523018717765808
step = 4400: loss = 0.7796685695648193
step = 4600: loss = 0.7155728340148926
step = 4800: loss = 0.8101846575737
step = 5000: loss = 0.6271236538887024
step = 5000: Average Return = 291.80
step = 5200: loss = 0.60479736328125
step = 5400: loss = 0.6058151721954346
step = 5600: loss = 0.6204308271408081
step = 5800: loss = 0.4376189708709717
step = 6000: loss = 0.6457669734954834
step = 6000: Average Return = 390.70
step = 6200: loss = 0.664910078048706
step = 6400: loss = 0.5933699011802673
step = 6600: loss = 0.5152082443237305
step = 6800: loss = 1.1136772632598877
step = 7000: loss = 0.5354762673377991
step = 7000: Average Return = 216.90
step = 7200: loss = 0.5583940744400024
step = 7400: loss = 0.5272458791732788
step = 7600: loss = 0.43842631578445435
step = 7800: loss = 0.4991070032119751
step = 8000: loss = 0.5343080759048462
step = 8000: Average Return = 236.50
step = 8200: loss = 0.5989910960197449
step = 8400: loss = 0.55279541015625
step = 8600: loss = 0.4894791543483734
step = 8800: loss = 0.4605693817138672
step = 9000: loss = 0.7197680473327637
step = 9000: Average Return = 211.10
step = 9200: loss = 0.50169438123703
step = 9400: loss = 0.4047691226005554
step = 9600: loss = 0.3715749680995941
step = 9800: loss = 0.41730958223342896
step = 10000: loss = 0.5585692524909973
step = 10000: Average Return = 446.30
step = 10200: loss = 0.545777440071106
step = 10400: loss = 0.6296675801277161
step = 10600: loss = 0.5013774633407593
step = 10800: loss = 0.18949614465236664
step = 11000: loss = 0.5994815826416016
step = 11000: Average Return = 254.10
step = 11200: loss = 0.558006763458252
step = 11400: loss = 0.4349510669708252
step = 11600: loss = 0.4242212772369385
step = 11800: loss = 0.43511703610420227
step = 12000: loss = 0.31298211216926575
step = 12000: Average Return = 227.00
step = 12200: loss = 0.45687440037727356
step = 12400: loss = 0.4060841500759125
step = 12600: loss = 0.47229182720184326
step = 12800: loss = 0.511258065700531
step = 13000: loss = 0.452497661113739
step = 13000: Average Return = 179.70
step = 13200: loss = 0.2760928273200989
step = 13400: loss = 0.48157840967178345
step = 13600: loss = 0.3031355142593384
step = 13800: loss = 0.6613827347755432
step = 14000: loss = 0.37960049510002136
step = 14000: Average Return = 355.10
step = 14200: loss = 0.42874500155448914
step = 14400: loss = 0.6264122128486633
step = 14600: loss = 0.34130242466926575
step = 14800: loss = 0.43740400671958923
step = 15000: loss = 0.19556452333927155
step = 15000: Average Return = 339.00

Visualization

Plots

We can plot return vs global steps to see the performance of our agent. In Cartpole-v1, the environment gives a reward of +1 for every time step the pole stays up, and since the maximum number of steps is 500, the maximum possible return is also 500.

steps = range(0, num_iterations + 1, eval_interval)
plt.plot(steps, returns)
plt.ylabel('Average Return')
plt.xlabel('Step')
plt.ylim(top=550)

(-12.129999589920047, 550.0)

Videos

It is helpful to visualize the performance of an agent by rendering the environment at each step. Before we do that, let us first create a function to embed videos in this colab.

def embed_mp4(filename):
  """Embeds an mp4 file in the notebook."""
  video = open(filename,'rb').read()
  b64 = base64.b64encode(video)
  tag = '''
  <video width="640" height="480" controls>
    <source src="data:video/mp4;base64,{0}" type="video/mp4">
  Your browser does not support the video tag.
  </video>'''.format(b64.decode())

  return IPython.display.HTML(tag)

The following code visualizes the agent's policy for a few episodes:

num_episodes = 3
video_filename = 'imageio.mp4'
with imageio.get_writer(video_filename, fps=60) as video:
  for _ in range(num_episodes):
    time_step = eval_env.reset()
    video.append_data(eval_py_env.render())
    while not time_step.is_last():
      action_step = agent.policy.action(time_step)
      time_step = eval_env.step(action_step.action)
      video.append_data(eval_py_env.render())

embed_mp4(video_filename)

WARNING:root:IMAGEIO FFMPEG_WRITER WARNING: input image is not divisible by macro_block_size=16, resizing from (400, 600) to (400, 608) to ensure video compatibility with most codecs and players. To prevent resizing, make your input image divisible by the macro_block_size or set the macro_block_size to None (risking incompatibility). You may also see a FFMPEG warning concerning speedloss due to data not being aligned.

C51 tends to do slightly better than DQN on CartPole-v1, but the difference between the two agents becomes more and more significant in increasingly complex environments. For example, on the full Atari 2600 benchmark, C51 demonstrates a mean score improvement of 126% over DQN after normalizing with respect to a random agent. Additional improvements can be gained by including n-step updates.

For a deeper dive into the C51 algorithm, see A Distributional Perspective on Reinforcement Learning (2017).