def test_3_alphago_value(self):
print("TEST 3\n=====================================================")
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
# Restrict TensorFlow to only use the first GPU
try:
tf.config.experimental.set_visible_devices(gpus[0], 'GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)
tf.config.set_soft_device_placement(True)
except RuntimeError as e:
print(e)
rows, cols = 19, 19
encoder = AlphaGoEncoder()
input_shape = (encoder.num_planes, rows, cols)
alphago_value_network = alphago_model(input_shape)
alphago_value = ValueAgent(alphago_value_network, encoder)
experience = load_experience(
h5py.File('test_alphago_rl_experience.h5', 'r'))
alphago_value.train(experience)
with h5py.File('test_alphago_value.h5', 'w') as value_agent_out:
alphago_value.serialize(value_agent_out)
def test_1_supervised_learning(self):
print("TEST 1\n=====================================================")
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
# Restrict TensorFlow to only use the first GPU
try:
tf.config.experimental.set_visible_devices(gpus[0], 'GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)
tf.config.set_soft_device_placement(True)
except RuntimeError as e:
print(e)
rows, cols = 19, 19
encoder = AlphaGoEncoder()
input_shape = (encoder.num_planes, rows, cols)
alphago_sl_policy = alphago_model(input_shape, is_policy_net=True)
alphago_sl_policy.compile('sgd',
'categorical_crossentropy',
metrics=['accuracy'])
alphago_sl_agent = DeepLearningAgent(alphago_sl_policy, encoder)
inputs = np.ones((10, ) + input_shape)
outputs = alphago_sl_policy.predict(inputs)
assert (outputs.shape == (10, 361))
with h5py.File('test_alphago_sl_policy.h5', 'w') as sl_agent_out:
alphago_sl_agent.serialize(sl_agent_out)
def main():
# sl data
encoder = AlphaGoEncoder()
processor = GoDataProcessor(encoder=encoder.name())
# Paraller Processor
generator = processor.load_go_data('train', NUM_GAMES, use_generator=True)
test_generator = processor.load_go_data('test',
NUM_GAMES,
use_generator=True)
# Data Processor
# todo: does not have use_generator capability
# generator = processor.load_go_data('train', NUM_GAMES)
# test_generator = processor.load_go_data('test', NUM_GAMES)
# sl model
input_shape = (encoder.num_planes, ROWS, COLS)
alphago_sl_policy = alphago_model(input_shape=input_shape,
is_policy_net=True)
# read earlier trained bot
bot_filepath = 'alphago/alpha_sl_policy_e13_1k.h5'
alphago_sl_policy.load_weights(bot_filepath)
alphago_sl_policy.compile(optimizer='sgd',
loss='categorical_crossentropy',
metrics=['accuracy'])
# sl train
epochs = 200
batch_size = 128
alphago_sl_policy.fit_generator(
generator=generator.generate(batch_size, NUM_CLASSES),
epochs=epochs,
steps_per_epoch=generator.get_num_samples() / batch_size,
validation_data=test_generator.generate(batch_size, NUM_CLASSES),
validation_steps=test_generator.get_num_samples() / batch_size,
callbacks=[ModelCheckpoint('alphago_sl_policy_load_train_{epoch}.h5')])
alphago_sl_agent = DeepLearningAgent(alphago_sl_policy, encoder)
# save model
with h5py.File('alphago_sl_policy_load_train.h5', 'w') as sl_agent_out:
alphago_sl_agent.serialize(sl_agent_out)
# evaluate
alphago_sl_policy.evaluate_generator(
generator=test_generator.generate(batch_size, NUM_CLASSES),
steps=test_generator.get_num_samples() / batch_size)
def test_3_alphago_value(self):
rows, cols = 19, 19
encoder = AlphaGoEncoder()
input_shape = (encoder.num_planes, rows, cols)
alphago_value_network = alphago_model(input_shape)
alphago_value = ValueAgent(alphago_value_network, encoder)
experience = load_experience(
h5py.File('test_alphago_rl_experience.h5', 'r'))
alphago_value.train(experience)
with h5py.File('test_alphago_value.h5', 'w') as value_agent_out:
alphago_value.serialize(value_agent_out)
def test_1_supervised_learning(self):
rows, cols = 19, 19
encoder = AlphaGoEncoder()
input_shape = (encoder.num_planes, rows, cols)
alphago_sl_policy = alphago_model(input_shape, is_policy_net=True)
alphago_sl_policy.compile('sgd',
'categorical_crossentropy',
metrics=['accuracy'])
alphago_sl_agent = DeepLearningAgent(alphago_sl_policy, encoder)
inputs = np.ones((10, ) + input_shape)
outputs = alphago_sl_policy.predict(inputs)
assert (outputs.shape == (10, 361))
with h5py.File('test_alphago_sl_policy.h5', 'w') as sl_agent_out:
alphago_sl_agent.serialize(sl_agent_out)
# tag::init_value[]
from dlgo.networks.alphago import alphago_model
from dlgo.encoders.alphago import AlphaGoEncoder
from dlgo.rl import ValueAgent, load_experience
import h5py
rows, cols = 19, 19
encoder = AlphaGoEncoder()
input_shape = (encoder.num_planes, rows, cols)
alphago_value_network = alphago_model(input_shape)
alphago_value = ValueAgent(alphago_value_network, encoder)
# end::init_value[]
# tag::train_value[]
experience = load_experience(h5py.File('alphago_rl_experience.h5', 'r'))
alphago_value.train(experience)
with h5py.File('alphago_value.h5', 'w') as value_agent_out:
alphago_value.serialize(value_agent_out)
# end::train_value[]
from dlgo.data.parallel_processor import GoDataProcessor
from dlgo.encoders.alphago import AlphaGoEncoder
from dlgo.agent.predict import DeepLearningAgent
from dlgo.networks.alphago import alphago_model
rows, cols = 19, 19
num_classes = rows * cols
num_games = 10000
encoder = AlphaGoEncoder()
processor = GoDataProcessor(encoder=encoder.name())
generator = processor.load_go_data('train', num_games, use_generator=True)
test_generator = processor.load_go_data('test', num_games, use_generator=True)
input_shape = (encoder.num_planes, rows, cols)
alphago_sl_policy = alphago_model(input_shape, is_policy_net=True)
alphago_sl_policy.compile('sgd',
'categorical_crossentropy',
metrics=['accuracy'])
epochs = 200
batch_size = 128
alphago_sl_policy = fit_generator(
generator=generator.generate(batch_size, num_classes),
epochs=epochs,
steps_per_epoch=generator.get_num_samples() / batch_size,
validation_data=test_generator.generate(batch_size, num_classes),
validation_steps=test_generator.get_num_samples() / batch_size,
callbacks=[ModelCheckpoint('alphago_sl_policy_{epoch}.h5')])
alphago_sl_agent = DeepLearningAgent(alphago_sl_policy, encoder)
