Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

upgrade for TensorFlow 1.x; add granular timing to help estimation #6

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

upgrade for TensorFlow 1.x; add granular timing to help estimation #6

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
41 changes: 23 additions & 18 deletions model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,8 +40,8 @@ def __init__(self, sess, model_path, summary_path, checkpoint_path, restore=Fals
alpha = tf.maximum(0.01, tf.train.exponential_decay(0.1, self.global_step, \
40000, 0.96, staircase=True), name='alpha')

tf.scalar_summary('lambda', lamda)
tf.scalar_summary('alpha', alpha)
tf.summary.scalar('lambda', lamda)
tf.summary.scalar('alpha', alpha)

# describe network size
layer_size_input = 294
Expand All @@ -57,8 +57,8 @@ def __init__(self, sess, model_path, summary_path, checkpoint_path, restore=Fals
self.V = dense_layer(prev_y, [layer_size_hidden, layer_size_output], tf.sigmoid, name='layer2')

# watch the individual value predictions over time
tf.scalar_summary('V_next', tf.reduce_sum(self.V_next))
tf.scalar_summary('V', tf.reduce_sum(self.V))
tf.summary.scalar('V_next', tf.reduce_sum(self.V_next))
tf.summary.scalar('V', tf.reduce_sum(self.V))

# delta = V_next - V
delta_op = tf.reduce_sum(self.V_next - self.V, name='delta')
Expand Down Expand Up @@ -94,12 +94,12 @@ def __init__(self, sess, model_path, summary_path, checkpoint_path, restore=Fals
delta_avg_ema_op = delta_avg_ema.apply([delta_avg_op])
accuracy_avg_ema_op = accuracy_avg_ema.apply([accuracy_avg_op])

tf.scalar_summary('game/loss_avg', loss_avg_op)
tf.scalar_summary('game/delta_avg', delta_avg_op)
tf.scalar_summary('game/accuracy_avg', accuracy_avg_op)
tf.scalar_summary('game/loss_avg_ema', loss_avg_ema.average(loss_avg_op))
tf.scalar_summary('game/delta_avg_ema', delta_avg_ema.average(delta_avg_op))
tf.scalar_summary('game/accuracy_avg_ema', accuracy_avg_ema.average(accuracy_avg_op))
tf.summary.scalar('game/loss_avg', loss_avg_op)
tf.summary.scalar('game/delta_avg', delta_avg_op)
tf.summary.scalar('game/accuracy_avg', accuracy_avg_op)
tf.summary.scalar('game/loss_avg_ema', loss_avg_ema.average(loss_avg_op))
tf.summary.scalar('game/delta_avg_ema', delta_avg_ema.average(delta_avg_op))
tf.summary.scalar('game/accuracy_avg_ema', accuracy_avg_ema.average(accuracy_avg_op))

# reset per-game monitoring variables
game_step_reset_op = game_step.assign(0.0)
Expand All @@ -115,8 +115,8 @@ def __init__(self, sess, model_path, summary_path, checkpoint_path, restore=Fals

# watch the weight and gradient distributions
for grad, var in zip(grads, tvars):
tf.histogram_summary(var.name, var)
tf.histogram_summary(var.name + '/gradients/grad', grad)
tf.summary.histogram(var.name, var)
tf.summary.histogram(var.name + '/gradients/grad', grad)

# for each variable, define operations to update the var with delta,
# taking into account the gradient as part of the eligibility trace
Expand All @@ -127,11 +127,11 @@ def __init__(self, sess, model_path, summary_path, checkpoint_path, restore=Fals
# e-> = lambda * e-> + <grad of output w.r.t weights>
trace = tf.Variable(tf.zeros(grad.get_shape()), trainable=False, name='trace')
trace_op = trace.assign((lamda * trace) + grad)
tf.histogram_summary(var.name + '/traces', trace)
tf.summary.histogram(var.name + '/traces', trace)

# grad with trace = alpha * delta * e
grad_trace = alpha * delta_op * trace_op
tf.histogram_summary(var.name + '/gradients/trace', grad_trace)
tf.summary.histogram(var.name + '/gradients/trace', grad_trace)

grad_apply = var.assign_add(grad_trace)
apply_gradients.append(grad_apply)
Expand All @@ -151,13 +151,13 @@ def __init__(self, sess, model_path, summary_path, checkpoint_path, restore=Fals
self.train_op = tf.group(*apply_gradients, name='train')

# merge summaries for TensorBoard
self.summaries_op = tf.merge_all_summaries()
self.summaries_op = tf.summary.merge_all()

# create a saver for periodic checkpoints
self.saver = tf.train.Saver(max_to_keep=1)

# run variable initializers
self.sess.run(tf.initialize_all_variables())
self.sess.run(tf.global_variables_initializer())

# after training a model, we can restore checkpoints here
if restore:
Expand Down Expand Up @@ -194,15 +194,17 @@ def test(self, episodes=100, draw=False):

def train(self):
tf.train.write_graph(self.sess.graph_def, self.model_path, 'td_gammon.pb', as_text=False)
summary_writer = tf.train.SummaryWriter('{0}{1}'.format(self.summary_path, int(time.time()), self.sess.graph_def))
summary_writer = tf.summary.FileWriter('{0}{1}'.format(self.summary_path, int(time.time()), self.sess.graph_def))

# the agent plays against itself, making the best move for each player
players = [TDAgent(Game.TOKENS[0], self), TDAgent(Game.TOKENS[1], self)]

validation_interval = 1000
episodes = 5000

train_start_ts = time.time()
for episode in range(episodes):
start_ts = time.time()
if episode != 0 and episode % validation_interval == 0:
self.test(episodes=100)

Expand Down Expand Up @@ -233,7 +235,10 @@ def train(self):
], feed_dict={ self.x: x, self.V_next: np.array([[winner]], dtype='float') })
summary_writer.add_summary(summaries, global_step=global_step)

print("Game %d/%d (Winner: %s) in %d turns" % (episode, episodes, players[winner].player, game_step))
end_ts = time.time()
print("Game %d/%d (Winner: %s) in %d turns (%.2f secs)" % (episode, episodes, players[winner].player, game_step, end_ts-start_ts))
if episode in [9, 99, 999, 9999, 99999]:
print("%d games avg time: %.2f secs" % (episode+1, (end_ts - train_start_ts) / (episode+1)))
self.saver.save(self.sess, self.checkpoint_path + 'checkpoint', global_step=global_step)

summary_writer.close()
Expand Down