Copyright 2021 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 updatesudo apt-get install -y xvfb ffmpeg freeglut3-devpip install 'imageio==2.4.0'pip install pyvirtualdisplaypip install tf-agentspip install pyglet
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
# 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.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.
25.1
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/python3.9/site-packages/tensorflow/python/autograph/impl/api.py:377: 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/python3.9/site-packages/tensorflow/python/util/dispatch.py:1082: 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 = 3.1652722358703613 step = 400: loss = 2.3220953941345215 step = 600: loss = 1.9085898399353027 step = 800: loss = 1.5957838296890259 step = 1000: loss = 1.373847484588623 step = 1000: Average Return = 466.70 step = 1200: loss = 1.2249349355697632 step = 1400: loss = 1.2289665937423706 step = 1600: loss = 1.225049614906311 step = 1800: loss = 1.4009439945220947 step = 2000: loss = 0.8110367059707642 step = 2000: Average Return = 310.30 step = 2200: loss = 0.8426725268363953 step = 2400: loss = 0.9993857145309448 step = 2600: loss = 0.7408146858215332 step = 2800: loss = 1.0472800731658936 step = 3000: loss = 0.8934259414672852 step = 3000: Average Return = 294.80 step = 3200: loss = 0.67853844165802 step = 3400: loss = 0.9168663024902344 step = 3600: loss = 0.6471030712127686 step = 3800: loss = 0.8118085861206055 step = 4000: loss = 0.7178002595901489 step = 4000: Average Return = 339.10 step = 4200: loss = 0.5277565717697144 step = 4400: loss = 0.6562362909317017 step = 4600: loss = 0.6893218755722046 step = 4800: loss = 0.6171445846557617 step = 5000: loss = 0.6233919262886047 step = 5000: Average Return = 192.00 step = 5200: loss = 0.5258955359458923 step = 5400: loss = 0.6037764549255371 step = 5600: loss = 0.6617163419723511 step = 5800: loss = 0.45471426844596863 step = 6000: loss = 0.5623942017555237 step = 6000: Average Return = 375.00 step = 6200: loss = 0.5260623097419739 step = 6400: loss = 0.5474383234977722 step = 6600: loss = 0.6723802089691162 step = 6800: loss = 0.4168206453323364 step = 7000: loss = 0.6093295812606812 step = 7000: Average Return = 396.20 step = 7200: loss = 0.5631401538848877 step = 7400: loss = 0.5302916765213013 step = 7600: loss = 0.4411312937736511 step = 7800: loss = 0.5489145517349243 step = 8000: loss = 0.4881543517112732 step = 8000: Average Return = 352.20 step = 8200: loss = 0.5519999265670776 step = 8400: loss = 0.4684922993183136 step = 8600: loss = 0.523332953453064 step = 8800: loss = 0.4230990409851074 step = 9000: loss = 0.5511386394500732 step = 9000: Average Return = 169.30 step = 9200: loss = 0.5994375944137573 step = 9400: loss = 0.3859468698501587 step = 9600: loss = 0.3768221437931061 step = 9800: loss = 0.3608618378639221 step = 10000: loss = 0.45109525322914124 step = 10000: Average Return = 159.40 step = 10200: loss = 0.4834355115890503 step = 10400: loss = 0.3417738378047943 step = 10600: loss = 0.42035162448883057 step = 10800: loss = 0.513039231300354 step = 11000: loss = 0.4203823208808899 step = 11000: Average Return = 329.90 step = 11200: loss = 0.532701849937439 step = 11400: loss = 0.34555840492248535 step = 11600: loss = 0.23318243026733398 step = 11800: loss = 0.373273640871048 step = 12000: loss = 0.4745432734489441 step = 12000: Average Return = 484.00 step = 12200: loss = 0.38893377780914307 step = 12400: loss = 0.45256471633911133 step = 12600: loss = 0.2996901571750641 step = 12800: loss = 0.44166380167007446 step = 13000: loss = 0.34164178371429443 step = 13000: Average Return = 329.70 step = 13200: loss = 0.45920413732528687 step = 13400: loss = 0.4424200654029846 step = 13600: loss = 0.48878079652786255 step = 13800: loss = 0.48222893476486206 step = 14000: loss = 0.3798040747642517 step = 14000: Average Return = 433.20 step = 14200: loss = 0.46709728240966797 step = 14400: loss = 0.24153408408164978 step = 14600: loss = 0.28913378715515137 step = 14800: loss = 0.36507582664489746 step = 15000: loss = 0.32009801268577576 step = 15000: Average Return = 141.00
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)
(-14.11999959945679, 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. [swscaler @ 0x56191e3c7880] Warning: data is not aligned! This can lead to a speed loss
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).