def main(unused_argv):
config = alpha_zero.Config(
game=FLAGS.game,
path=FLAGS.path,
learning_rate=FLAGS.learning_rate,
weight_decay=FLAGS.weight_decay,
train_batch_size=FLAGS.train_batch_size,
replay_buffer_size=FLAGS.replay_buffer_size,
replay_buffer_reuse=FLAGS.replay_buffer_reuse,
max_steps=FLAGS.max_steps,
checkpoint_freq=FLAGS.checkpoint_freq,
actors=FLAGS.actors,
evaluators=FLAGS.evaluators,
uct_c=FLAGS.uct_c,
max_simulations=FLAGS.max_simulations,
policy_alpha=FLAGS.policy_alpha,
policy_epsilon=FLAGS.policy_epsilon,
temperature=FLAGS.temperature,
temperature_drop=FLAGS.temperature_drop,
evaluation_window=FLAGS.evaluation_window,
eval_levels=FLAGS.eval_levels,
nn_model=FLAGS.nn_model,
nn_width=FLAGS.nn_width,
nn_depth=FLAGS.nn_depth,
observation_shape=None,
output_size=None,
quiet=FLAGS.quiet,
)
alpha_zero.alpha_zero(config)
def main(unused_argv):
config = alpha_zero.Config(
game="tic_tac_toe",
path=FLAGS.path,
learning_rate=0.01,
weight_decay=1e-4,
train_batch_size=128,
replay_buffer_size=2**14,
replay_buffer_reuse=4,
max_steps=25,
checkpoint_freq=25,
actors=4,
evaluators=4,
uct_c=1,
max_simulations=20,
policy_alpha=0.25,
policy_epsilon=1,
temperature=1,
temperature_drop=4,
evaluation_window=50,
eval_levels=7,
nn_model="resnet",
nn_width=128,
nn_depth=2,
observation_shape=None,
output_size=None,
quiet=True,
)
alpha_zero.alpha_zero(config)