Training a Neural Network ATARI Pong agent with Policy Gradients from raw pixels

pg-pong.py

""" Trains an agent with (stochastic) Policy Gradients on Pong. Uses OpenAI Gym. """
import numpy as np
import cPickle as pickle
import gym

# hyperparameters
H = 200 # number of hidden layer neurons
batch_size = 10 # every how many episodes to do a param update?
learning_rate = 1e-4
gamma = 0.99 # discount factor for reward
decay_rate = 0.99 # decay factor for RMSProp leaky sum of grad^2
resume = False # resume from previous checkpoint?
render = False

# model initialization
D = 80 * 80 # input dimensionality: 80x80 grid
if resume:
  model = pickle.load(open('save.p', 'rb'))
else:
  model = {}
  model['W1'] = np.random.randn(H,D) / np.sqrt(D) # "Xavier" initialization
  model['W2'] = np.random.randn(H) / np.sqrt(H)
  
grad_buffer = { k : np.zeros_like(v) for k,v in model.iteritems() } # update buffers that add up gradients over a batch
rmsprop_cache = { k : np.zeros_like(v) for k,v in model.iteritems() } # rmsprop memory

def sigmoid(x): 
  return 1.0 / (1.0 + np.exp(-x)) # sigmoid "squashing" function to interval [0,1]

def prepro(I):
  """ prepro 210x160x3 uint8 frame into 6400 (80x80) 1D float vector """
  I = I[35:195] # crop
  I = I[::2,::2,0] # downsample by factor of 2
  I[I == 144] = 0 # erase background (background type 1)
  I[I == 109] = 0 # erase background (background type 2)
  I[I != 0] = 1 # everything else (paddles, ball) just set to 1
  return I.astype(np.float).ravel()

def discount_rewards(r):
  """ take 1D float array of rewards and compute discounted reward """
  discounted_r = np.zeros_like(r)
  running_add = 0
  for t in reversed(xrange(0, r.size)):
    if r[t] != 0: running_add = 0 # reset the sum, since this was a game boundary (pong specific!)
    running_add = running_add * gamma + r[t]
    discounted_r[t] = running_add
  return discounted_r

def policy_forward(x):
  h = np.dot(model['W1'], x)
  h[h<0] = 0 # ReLU nonlinearity
  logp = np.dot(model['W2'], h)
  p = sigmoid(logp)
  return p, h # return probability of taking action 2, and hidden state

def policy_backward(eph, epdlogp):
  """ backward pass. (eph is array of intermediate hidden states) """
  dW2 = np.dot(eph.T, epdlogp).ravel()
  dh = np.outer(epdlogp, model['W2'])
  dh[eph <= 0] = 0 # backpro prelu
  dW1 = np.dot(dh.T, epx)
  return {'W1':dW1, 'W2':dW2}

env = gym.make("Pong-v0")
observation = env.reset()
prev_x = None # used in computing the difference frame
xs,hs,dlogps,drs = [],[],[],[]
running_reward = None
reward_sum = 0
episode_number = 0
while True:
  if render: env.render()

  # preprocess the observation, set input to network to be difference image
  cur_x = prepro(observation)
  x = cur_x - prev_x if prev_x is not None else np.zeros(D)
  prev_x = cur_x

  # forward the policy network and sample an action from the returned probability
  aprob, h = policy_forward(x)
  action = 2 if np.random.uniform() < aprob else 3 # roll the dice!

  # record various intermediates (needed later for backprop)
  xs.append(x) # observation
  hs.append(h) # hidden state
  y = 1 if action == 2 else 0 # a "fake label"
  dlogps.append(y - aprob) # grad that encourages the action that was taken to be taken (see http://cs231n.github.io/neural-networks-2/#losses if confused)

  # step the environment and get new measurements
  observation, reward, done, info = env.step(action)
  reward_sum += reward

  drs.append(reward) # record reward (has to be done after we call step() to get reward for previous action)

  if done: # an episode finished
    episode_number += 1

    # stack together all inputs, hidden states, action gradients, and rewards for this episode
    epx = np.vstack(xs)
    eph = np.vstack(hs)
    epdlogp = np.vstack(dlogps)
    epr = np.vstack(drs)
    xs,hs,dlogps,drs = [],[],[],[] # reset array memory

    # compute the discounted reward backwards through time
    discounted_epr = discount_rewards(epr)
    # standardize the rewards to be unit normal (helps control the gradient estimator variance)
    discounted_epr -= np.mean(discounted_epr)
    discounted_epr /= np.std(discounted_epr)

    epdlogp *= discounted_epr # modulate the gradient with advantage (PG magic happens right here.)
    grad = policy_backward(eph, epdlogp)
    for k in model: grad_buffer[k] += grad[k] # accumulate grad over batch

    # perform rmsprop parameter update every batch_size episodes
    if episode_number % batch_size == 0:
      for k,v in model.iteritems():
        g = grad_buffer[k] # gradient
        rmsprop_cache[k] = decay_rate * rmsprop_cache[k] + (1 - decay_rate) * g**2
        model[k] += learning_rate * g / (np.sqrt(rmsprop_cache[k]) + 1e-5)
        grad_buffer[k] = np.zeros_like(v) # reset batch gradient buffer

    # boring book-keeping
    running_reward = reward_sum if running_reward is None else running_reward * 0.99 + reward_sum * 0.01
    print 'resetting env. episode reward total was %f. running mean: %f' % (reward_sum, running_reward)
    if episode_number % 100 == 0: pickle.dump(model, open('save.p', 'wb'))
    reward_sum = 0
    observation = env.reset() # reset env
    prev_x = None

  if reward != 0: # Pong has either +1 or -1 reward exactly when game ends.
    print ('ep %d: game finished, reward: %f' % (episode_number, reward)) + ('' if reward == -1 else ' !!!!!!!!')

@Grsz That is a perfect question for StackOverflow...

Python3 version of pg-pong.py with the minimum changes to make it work:
https://gist.github.com/haluptzok/d2a3eba5d25d238d6c2cbe847bc58b6b
Still a great policy gradient blog post and python script - but Python2 is so 2016 : )
Most folks reading this now will fire it up in python3 and blow up and not get the fun experience

For those interested in seeing this implemented on top of TensorFlow 2 running entirely on graph mode here's the repo:

https://github.com/WillianFuks/Pong

The AI trained fairly quickly, in a day it already reached average return of ~14 points. But then it stops there and doesn't quite improve much after all. Not sure on how to further improve it then, other than keep tweaking the hyperparams.

May 19th 2022
I modify some lines of pg-pong.py because this is too old (but gold).
In my case rendering option did not work because of openai-gym issue.
pls check this code if you want to train agent playing pong in py38, gym>=0.21.0

In case someone wants to share a cool colab demo still - here is my notebook that ended up achieving level performance with the human opponent
https://colab.research.google.com/drive/1KZeGjxS7OUHKotsuyoT0DtxVzKaUIq4B?usp=sharing

I created a variation of the original demo by adding another layer of hidden variables. This new variations converges much faster than the original solution. I also fixed a bunch of bugs/issues introduced because of new versions of libraries. Have fun!

https://colab.research.google.com/drive/1w1EklesVqWaCOK2KyidJbauarn7kUoaV#scrollTo=TwjiwKisQM19

https://gist.github.com/CPPAlien/91388eb16a85e80ec55689069bda0c25

I implemented this code by pytorch, and it seems the positive reward not increase during the training process. Are there any mistakes of my code?

PyTorch version - https://gist.github.com/xanderex-sid/ae6cd3ea0c3759c1e3f92835ebd6e158

You can use the code above to create various types of CNN or MLP models to train your ATARI Pong agent. The code is compatible with GPU usage as well.

It has been successfully training on Google Colab, so feel free to use it in your experiments. If you find any bugs or have suggestions for improvements, please let me know, as it will help me enhance my skills.

I have also attached a code to test your agent against the default OpenAI Pong agent.