Missed TensorFlow Dev Summit? Check out the video playlist. Watch recordings

Train a Deep Q Network with TF-Agents

View on TensorFlow.org Run in Google Colab View source on GitHub Download notebook

Introduction

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

Cartpole environment

It will walk you through all the components in a Reinforcement Learning (RL) pipeline for training, evaluation and data collection.

To run this code live, click the 'Run in Google Colab' link above.

Setup

try:
  %tensorflow_version 2.x
except:
  pass

If you haven't installed the following dependencies, 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 --upgrade tensorflow-probability
pip install -q tf-agents



ffmpeg is already the newest version (7:3.4.6-0ubuntu0.18.04.1).
xvfb is already the newest version (2:1.19.6-1ubuntu4.4).
0 upgraded, 0 newly installed, 0 to remove and 63 not upgraded.
from __future__ import absolute_import, division, print_function

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

import tensorflow as tf

from tf_agents.agents.dqn import 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 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()
xdpyinfo was not found, X start can not be checked! Please install xdpyinfo!
tf.version.VERSION
'2.1.0'

Hyperparameters

num_iterations = 20000 # @param {type:"integer"}

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

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

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

Environment

In Reinforcement Learning (RL), an environment represents the task or problem to be solved. Standard environments can be created in TF-Agents using tf_agents.environments suites. TF-Agents has suites for loading environments from sources such as the OpenAI Gym, Atari, and DM Control.

Load the CartPole environment from the OpenAI Gym suite.

env_name = 'CartPole-v0'
env = suite_gym.load(env_name)

You can render this environment to see how it looks. A free-swinging pole is attached to a cart. The goal is to move the cart right or left in order to keep the pole pointing up.

env.reset()
PIL.Image.fromarray(env.render())

png

The environment.step method takes an action in the environment and returns a TimeStep tuple containing the next observation of the environment and the reward for the action.

The time_step_spec() method returns the specification for the TimeStep tuple. Its observation attribute shows the shape of observations, the data types, and the ranges of allowed values. The reward attribute shows the same details for the reward.

print('Observation Spec:')
print(env.time_step_spec().observation)
Observation Spec:
BoundedArraySpec(shape=(4,), dtype=dtype('float32'), name='observation', minimum=[-4.8000002e+00 -3.4028235e+38 -4.1887903e-01 -3.4028235e+38], maximum=[4.8000002e+00 3.4028235e+38 4.1887903e-01 3.4028235e+38])
print('Reward Spec:')
print(env.time_step_spec().reward)
Reward Spec:
ArraySpec(shape=(), dtype=dtype('float32'), name='reward')

The action_spec() method returns the shape, data types, and allowed values of valid actions.

print('Action Spec:')
print(env.action_spec())
Action Spec:
BoundedArraySpec(shape=(), dtype=dtype('int64'), name='action', minimum=0, maximum=1)

In the Cartpole environment:

  • observation is an array of 4 floats:
    • the position and velocity of the cart
    • the angular position and velocity of the pole
  • reward is a scalar float value
  • action is a scalar integer with only two possible values:
    • 0 — "move left"
    • 1 — "move right"
time_step = env.reset()
print('Time step:')
print(time_step)

action = np.array(1, dtype=np.int32)

next_time_step = env.step(action)
print('Next time step:')
print(next_time_step)
Time step:
TimeStep(step_type=array(0, dtype=int32), reward=array(0., dtype=float32), discount=array(1., dtype=float32), observation=array([-0.03225056,  0.04766411, -0.00837008,  0.02836   ], dtype=float32))
Next time step:
TimeStep(step_type=array(1, dtype=int32), reward=array(1., dtype=float32), discount=array(1., dtype=float32), observation=array([-0.03129728,  0.24290508, -0.00780288, -0.26695198], dtype=float32))

Usually two environments are instantiated: one for training and one for evaluation.

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

The Cartpole environment, like most environments, is written in pure Python. This is converted to TensorFlow using the TFPyEnvironment wrapper.

The original environment's API uses Numpy arrays. The TFPyEnvironment converts these to Tensors to make it compatible with Tensorflow agents and policies.

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

Agent

The algorithm used to solve an RL problem is represented by an Agent. TF-Agents provides standard implementations of a variety of Agents, including:

The DQN agent can be used in any environment which has a discrete action space.

At the heart of a DQN Agent is a QNetwork, a neural network model that can learn to predict QValues (expected returns) for all actions, given an observation from the environment.

Use tf_agents.networks.q_network to create a QNetwork, passing in the observation_spec, action_spec, and a tuple describing the number and size of the model's hidden layers.

fc_layer_params = (100,)

q_net = q_network.QNetwork(
    train_env.observation_spec(),
    train_env.action_spec(),
    fc_layer_params=fc_layer_params)

Now use tf_agents.agents.dqn.dqn_agent to instantiate a DqnAgent. In addition to the time_step_spec, action_spec and the QNetwork, the agent constructor also requires an optimizer (in this case, AdamOptimizer), a loss function, and an integer step counter.

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

train_step_counter = tf.Variable(0)

agent = dqn_agent.DqnAgent(
    train_env.time_step_spec(),
    train_env.action_spec(),
    q_network=q_net,
    optimizer=optimizer,
    td_errors_loss_fn=common.element_wise_squared_loss,
    train_step_counter=train_step_counter)

agent.initialize()

Policies

A policy defines the way an agent acts in an environment. Typically, the goal of reinforcement learning is to train the underlying model until the policy produces the desired outcome.

In this tutorial:

  • The desired outcome is keeping the pole balanced upright over the cart.
  • The policy returns an action (left or right) for each time_step observation.

Agents contain two policies:

  • agent.policy — The main policy that is used for evaluation and deployment.
  • agent.collect_policy — A second policy that is used for data collection.
eval_policy = agent.policy
collect_policy = agent.collect_policy

Policies can be created independently of agents. For example, use tf_agents.policies.random_tf_policy to create a policy which will randomly select an action for each time_step.

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

To get an action from a policy, call the policy.action(time_step) method. The time_step contains the observation from the environment. This method returns a PolicyStep, which is a named tuple with three components:

  • action — the action to be taken (in this case, 0 or 1)
  • state — used for stateful (that is, RNN-based) policies
  • info — auxiliary data, such as log probabilities of actions
example_environment = tf_py_environment.TFPyEnvironment(
    suite_gym.load('CartPole-v0'))
time_step = example_environment.reset()
random_policy.action(time_step)
PolicyStep(action=<tf.Tensor: shape=(1,), dtype=int64, numpy=array([1])>, state=(), info=())

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. Several episodes are run, creating an average return.

The following function computes the average return of a policy, given the policy, environment, and a number of episodes.

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]


# See also the metrics module for standard implementations of different metrics.
# https://github.com/tensorflow/agents/tree/master/tf_agents/metrics

Running this computation on the random_policy shows a baseline performance in the environment.

compute_avg_return(eval_env, random_policy, num_eval_episodes)
20.5

Replay Buffer

The replay buffer keeps track of data collected from the environment. This tutorial uses tf_agents.replay_buffers.tf_uniform_replay_buffer.TFUniformReplayBuffer, as it is the most common.

The constructor requires the specs for the data it will be collecting. This is available from the agent using the collect_data_spec method. The batch size and maximum buffer length are also required.

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

For most agents, collect_data_spec is a named tuple called Trajectory, containing the specs for observations, actions, rewards, and other items.

agent.collect_data_spec
Trajectory(step_type=TensorSpec(shape=(), dtype=tf.int32, name='step_type'), observation=BoundedTensorSpec(shape=(4,), dtype=tf.float32, name='observation', minimum=array([-4.8000002e+00, -3.4028235e+38, -4.1887903e-01, -3.4028235e+38],
      dtype=float32), maximum=array([4.8000002e+00, 3.4028235e+38, 4.1887903e-01, 3.4028235e+38],
      dtype=float32)), action=BoundedTensorSpec(shape=(), dtype=tf.int64, name='action', minimum=array(0), maximum=array(1)), policy_info=(), next_step_type=TensorSpec(shape=(), dtype=tf.int32, name='step_type'), reward=TensorSpec(shape=(), dtype=tf.float32, name='reward'), discount=BoundedTensorSpec(shape=(), dtype=tf.float32, name='discount', minimum=array(0., dtype=float32), maximum=array(1., dtype=float32)))
agent.collect_data_spec._fields
('step_type',
 'observation',
 'action',
 'policy_info',
 'next_step_type',
 'reward',
 'discount')

Data Collection

Now execute the random policy in the environment for a few steps, recording the data in the replay buffer.

def collect_step(environment, policy, buffer):
  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
  buffer.add_batch(traj)

def collect_data(env, policy, buffer, steps):
  for _ in range(steps):
    collect_step(env, policy, buffer)

collect_data(train_env, random_policy, replay_buffer, steps=100)

# This loop is so common in RL, that we provide standard implementations. 
# For more details see the drivers module.
# https://github.com/tensorflow/agents/blob/master/tf_agents/docs/python/tf_agents/drivers.md

The replay buffer is now a collection of Trajectories.

# For the curious:
# Uncomment to peel one of these off and inspect it.
# iter(replay_buffer.as_dataset()).next()

The agent needs access to the replay buffer. This is provided by creating an iterable tf.data.Dataset pipeline which will feed data to the agent.

Each row of the replay buffer only stores a single observation step. But since the DQN Agent needs both the current and next observation to compute the loss, the dataset pipeline will sample two adjacent rows for each item in the batch (num_steps=2).

This dataset is also optimized by running parallel calls and prefetching data.

# Dataset generates trajectories with shape [Bx2x...]
dataset = replay_buffer.as_dataset(
    num_parallel_calls=3, 
    sample_batch_size=batch_size, 
    num_steps=2).prefetch(3)


dataset
<PrefetchDataset shapes: (Trajectory(step_type=(64, 2), observation=(64, 2, 4), action=(64, 2), policy_info=(), next_step_type=(64, 2), reward=(64, 2), discount=(64, 2)), BufferInfo(ids=(64, 2), probabilities=(64,))), types: (Trajectory(step_type=tf.int32, observation=tf.float32, action=tf.int64, policy_info=(), next_step_type=tf.int32, reward=tf.float32, discount=tf.float32), BufferInfo(ids=tf.int64, probabilities=tf.float32))>
iterator = iter(dataset)

print(iterator)


<tensorflow.python.data.ops.iterator_ops.OwnedIterator object at 0x7fd6f0475908>
# For the curious:
# Uncomment to see what the dataset iterator is feeding to the agent.
# Compare this representation of replay data 
# to the collection of individual trajectories shown earlier.

# iterator.next()

Training the agent

Two things must happen during the training loop:

  • collect data from the environment
  • use that data to train the agent's neural network(s)

This example also periodicially evaluates the policy and prints the current score.

The following will take ~5 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, replay_buffer)

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

  step = agent.train_step_counter.numpy()

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

  if step % eval_interval == 0:
    avg_return = compute_avg_return(eval_env, agent.policy, num_eval_episodes)
    print('step = {0}: Average Return = {1}'.format(step, avg_return))
    returns.append(avg_return)
step = 200: loss = 15.113554000854492
step = 400: loss = 7.517986297607422
step = 600: loss = 5.416933059692383
step = 800: loss = 15.555917739868164
step = 1000: loss = 10.316483497619629
step = 1000: Average Return = 12.5
step = 1200: loss = 13.16561508178711
step = 1400: loss = 9.391595840454102
step = 1600: loss = 10.453310012817383
step = 1800: loss = 2.979498863220215
step = 2000: loss = 9.595071792602539
step = 2000: Average Return = 23.399999618530273
step = 2200: loss = 10.297588348388672
step = 2400: loss = 12.953393936157227
step = 2600: loss = 10.64794635772705
step = 2800: loss = 21.513774871826172
step = 3000: loss = 4.553102493286133
step = 3000: Average Return = 31.200000762939453
step = 3200: loss = 36.4605712890625
step = 3400: loss = 19.34351921081543
step = 3600: loss = 48.65455627441406
step = 3800: loss = 34.3037109375
step = 4000: loss = 9.973252296447754
step = 4000: Average Return = 66.0999984741211
step = 4200: loss = 41.363563537597656
step = 4400: loss = 7.632184028625488
step = 4600: loss = 144.8666534423828
step = 4800: loss = 7.129297256469727
step = 5000: loss = 48.949195861816406
step = 5000: Average Return = 162.1999969482422
step = 5200: loss = 65.9273681640625
step = 5400: loss = 4.031966686248779
step = 5600: loss = 35.347267150878906
step = 5800: loss = 104.69065856933594
step = 6000: loss = 7.288619518280029
step = 6000: Average Return = 138.3000030517578
step = 6200: loss = 39.277854919433594
step = 6400: loss = 78.52557373046875
step = 6600: loss = 85.15464782714844
step = 6800: loss = 121.00666046142578
step = 7000: loss = 10.944552421569824
step = 7000: Average Return = 140.0
step = 7200: loss = 67.29304504394531
step = 7400: loss = 182.94839477539062
step = 7600: loss = 69.43492126464844
step = 7800: loss = 130.48533630371094
step = 8000: loss = 158.48220825195312
step = 8000: Average Return = 169.5
step = 8200: loss = 182.46078491210938
step = 8400: loss = 476.6938781738281
step = 8600: loss = 124.17583465576172
step = 8800: loss = 12.144281387329102
step = 9000: loss = 12.156733512878418
step = 9000: Average Return = 172.3000030517578
step = 9200: loss = 16.37725830078125
step = 9400: loss = 72.38121032714844
step = 9600: loss = 20.577919006347656
step = 9800: loss = 156.2805633544922
step = 10000: loss = 51.40230941772461
step = 10000: Average Return = 155.39999389648438
step = 10200: loss = 142.04129028320312
step = 10400: loss = 80.34188079833984
step = 10600: loss = 679.5103759765625
step = 10800: loss = 42.54202651977539
step = 11000: loss = 23.710407257080078
step = 11000: Average Return = 177.10000610351562
step = 11200: loss = 449.2401123046875
step = 11400: loss = 23.078594207763672
step = 11600: loss = 478.2265319824219
step = 11800: loss = 168.52894592285156
step = 12000: loss = 17.50870704650879
step = 12000: Average Return = 187.0
step = 12200: loss = 16.723529815673828
step = 12400: loss = 18.591989517211914
step = 12600: loss = 131.7432861328125
step = 12800: loss = 115.3463363647461
step = 13000: loss = 199.59288024902344
step = 13000: Average Return = 195.60000610351562
step = 13200: loss = 28.509334564208984
step = 13400: loss = 194.09640502929688
step = 13600: loss = 18.391881942749023
step = 13800: loss = 45.12052536010742
step = 14000: loss = 31.816574096679688
step = 14000: Average Return = 178.6999969482422
step = 14200: loss = 685.3082275390625
step = 14400: loss = 452.07391357421875
step = 14600: loss = 384.8614807128906
step = 14800: loss = 39.358741760253906
step = 15000: loss = 23.265565872192383
step = 15000: Average Return = 195.89999389648438
step = 15200: loss = 46.020240783691406
step = 15400: loss = 49.2839469909668
step = 15600: loss = 554.0230712890625
step = 15800: loss = 30.424806594848633
step = 16000: loss = 455.3811950683594
step = 16000: Average Return = 197.39999389648438
step = 16200: loss = 36.63092803955078
step = 16400: loss = 112.33302307128906
step = 16600: loss = 870.1535034179688
step = 16800: loss = 409.5487365722656
step = 17000: loss = 37.29066467285156
step = 17000: Average Return = 196.60000610351562
step = 17200: loss = 66.248291015625
step = 17400: loss = 39.26573944091797
step = 17600: loss = 295.79449462890625
step = 17800: loss = 48.2138671875
step = 18000: loss = 258.0484924316406
step = 18000: Average Return = 192.60000610351562
step = 18200: loss = 41.24483108520508
step = 18400: loss = 52.373321533203125
step = 18600: loss = 46.42143630981445
step = 18800: loss = 760.6484375
step = 19000: loss = 1074.0638427734375
step = 19000: Average Return = 196.6999969482422
step = 19200: loss = 934.6668701171875
step = 19400: loss = 1318.4241943359375
step = 19600: loss = 304.5872802734375
step = 19800: loss = 46.10614776611328
step = 20000: loss = 61.52030944824219
step = 20000: Average Return = 194.3000030517578

Visualization

Plots

Use matplotlib.pyplot to chart how the policy improved during training.

One iteration of Cartpole-v0 consists of 200 time steps. The environment gives a reward of +1 for each step the pole stays up, so the maximum return for one episode is 200. The charts shows the return increasing towards that maximum each time it is evaluated during training. (It may be a little unstable and not increase monotonically each time.)


iterations = range(0, num_iterations + 1, eval_interval)
plt.plot(iterations, returns)
plt.ylabel('Average Return')
plt.xlabel('Iterations')
plt.ylim(top=250)
(-0.20999989509582484, 250)

png

Videos

Charts are nice. But more exciting is seeing an agent actually performing a task in an environment.

First, create a function to embed videos in the notebook.

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)

Now iterate through a few episodes of the Cartpole game with the agent. The underlying Python environment (the one "inside" the TensorFlow environment wrapper) provides a render() method, which outputs an image of the environment state. These can be collected into a video.

def create_policy_eval_video(policy, filename, num_episodes=5, fps=30):
  filename = filename + ".mp4"
  with imageio.get_writer(filename, fps=fps) 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 = policy.action(time_step)
        time_step = eval_env.step(action_step.action)
        video.append_data(eval_py_env.render())
  return embed_mp4(filename)




create_policy_eval_video(agent.policy, "trained-agent")
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.

For fun, compare the trained agent (above) to an agent moving randomly. (It does not do as well.)

create_policy_eval_video(random_policy, "random-agent")
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.