Copyright 2018 The TF-Agents Authors.
View on TensorFlow.org
|
Run in Google Colab
|
View source on GitHub
|
 Download notebook
|
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 ffmpegpip install -q 'gym==0.10.11'pip install -q 'imageio==2.4.0'pip install -q PILLOWpip install -q 'pyglet==1.3.2'pip install -q pyvirtualdisplaypip install -q --upgrade tensorflow-probabilitypip 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 90 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()
xdpyinfo was not found, X start can not be checked! Please install xdpyinfo!
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.
18.8
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)
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)
step = 200: loss = 3.147371768951416 step = 400: loss = 2.721038341522217 step = 600: loss = 2.4021265506744385 step = 800: loss = 1.9094001054763794 step = 1000: loss = 1.8790810108184814 step = 1000: Average Return = 88.90 step = 1200: loss = 1.9356472492218018 step = 1400: loss = 1.8789730072021484 step = 1600: loss = 1.2981504201889038 step = 1800: loss = 1.202945351600647 step = 2000: loss = 1.2120667695999146 step = 2000: Average Return = 154.90 step = 2200: loss = 1.2672768831253052 step = 2400: loss = 0.9420790672302246 step = 2600: loss = 1.0995533466339111 step = 2800: loss = 0.9630142450332642 step = 3000: loss = 1.1435785293579102 step = 3000: Average Return = 331.00 step = 3200: loss = 0.7764396667480469 step = 3400: loss = 0.8617928624153137 step = 3600: loss = 1.115776777267456 step = 3800: loss = 0.6288259625434875 step = 4000: loss = 0.7575881481170654 step = 4000: Average Return = 225.10 step = 4200: loss = 0.6906059980392456 step = 4400: loss = 0.6422734260559082 step = 4600: loss = 0.8901881575584412 step = 4800: loss = 0.6407145261764526 step = 5000: loss = 0.5517676472663879 step = 5000: Average Return = 217.80 step = 5200: loss = 0.6539983749389648 step = 5400: loss = 0.707004964351654 step = 5600: loss = 0.7433463335037231 step = 5800: loss = 0.7891858220100403 step = 6000: loss = 0.612639307975769 step = 6000: Average Return = 284.20 step = 6200: loss = 0.5980986952781677 step = 6400: loss = 0.5268639326095581 step = 6600: loss = 0.5849822163581848 step = 6800: loss = 0.6060997843742371 step = 7000: loss = 0.6804700493812561 step = 7000: Average Return = 406.20 step = 7200: loss = 0.5191953182220459 step = 7400: loss = 0.5209736824035645 step = 7600: loss = 0.3892410397529602 step = 7800: loss = 0.5716577172279358 step = 8000: loss = 0.5449703931808472 step = 8000: Average Return = 308.90 step = 8200: loss = 0.36637866497039795 step = 8400: loss = 0.5640881061553955 step = 8600: loss = 0.4244692623615265 step = 8800: loss = 0.500537633895874 step = 9000: loss = 0.5503699779510498 step = 9000: Average Return = 432.50 step = 9200: loss = 0.3443048894405365 step = 9400: loss = 0.5223272442817688 step = 9600: loss = 0.39678049087524414 step = 9800: loss = 0.33848363161087036 step = 10000: loss = 0.3010343313217163 step = 10000: Average Return = 435.60 step = 10200: loss = 0.39486125111579895 step = 10400: loss = 0.5263923406600952 step = 10600: loss = 0.5462918281555176 step = 10800: loss = 0.5018383264541626 step = 11000: loss = 0.5227107405662537 step = 11000: Average Return = 314.00 step = 11200: loss = 0.4220547378063202 step = 11400: loss = 0.3780883550643921 step = 11600: loss = 0.4848192632198334 step = 11800: loss = 0.5217380523681641 step = 12000: loss = 0.5033898949623108 step = 12000: Average Return = 468.50 step = 12200: loss = 0.3071030378341675 step = 12400: loss = 0.4585180878639221 step = 12600: loss = 0.39223724603652954 step = 12800: loss = 0.5044253468513489 step = 13000: loss = 0.5497949123382568 step = 13000: Average Return = 470.00 step = 13200: loss = 0.2276148647069931 step = 13400: loss = 0.4225274622440338 step = 13600: loss = 0.30521535873413086 step = 13800: loss = 0.25951749086380005 step = 14000: loss = 0.2249259650707245 step = 14000: Average Return = 413.80 step = 14200: loss = 0.389797568321228 step = 14400: loss = 0.3628026247024536 step = 14600: loss = 0.23991143703460693 step = 14800: loss = 0.3873177170753479 step = 15000: loss = 0.20888370275497437 step = 15000: Average Return = 439.90
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)
(-13.840000200271607, 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).