def do_train(args):
# Set up some parameters.
config = Config()
helper, train, dev, train_raw, dev_raw = load_and_preprocess_data(args)
embeddings = load_embeddings(args, helper)
config.embed_size = embeddings.shape[1]
helper.save(config.output_path)
handler = logging.FileHandler(config.log_output)
handler.setLevel(logging.DEBUG)
handler.setFormatter(logging.Formatter('%(asctime)s:%(levelname)s: %(message)s'))
logging.getLogger().addHandler(handler)
report = None # Report(Config.eval_output)
with tf.Graph().as_default():
logger.info("Building model...", )
start = time.time()
model = WindowModel(helper, config, embeddings)
logger.info("took %.2f seconds", time.time() - start)
init = tf.global_variables_initializer()
saver = tf.train.Saver()
with tf.Session() as session:
session.run(init)
model.fit(session, saver, train, dev)
if report:
report.log_output(model.output(session, dev_raw))
report.save()
else:
# Save predictions in a text file.
output = model.output(session, dev_raw)
sentences, labels, predictions = zip(*output)
predictions = [[LBLS[l] for l in preds] for preds in predictions]
output = zip(sentences, labels, predictions)
with open(model.config.conll_output, 'w') as f:
write_conll(f, output)
with open(model.config.eval_output, 'w') as f:
for sentence, labels, predictions in output:
print_sentence(f, sentence, labels, predictions)
def do_test2(args):
logger.info("Testing implementation of WindowModel")
config = Config()
helper, train, dev, train_raw, dev_raw = load_and_preprocess_data(args)
embeddings = load_embeddings(args, helper)
config.embed_size = embeddings.shape[1]
with tf.Graph().as_default():
logger.info("Building model...", )
start = time.time()
model = WindowModel(helper, config, embeddings)
logger.info("took %.2f seconds", time.time() - start)
init = tf.global_variables_initializer()
saver = None
with tf.Session() as session:
session.run(init)
model.fit(session, saver, train, dev)
logger.info("Model did not crash!")
logger.info("Passed!")
def do_test2(args):
logger.info("Testing implementation of RNNModel")
config = Config(args)
helper, train, dev, train_raw, dev_raw = load_and_preprocess_data(args)
embeddings = load_embeddings(args, helper)
config.embed_size = embeddings.shape[1]
with tf.Graph().as_default():
logger.info("Building model...", )
start = time.time()
model = RNNModel(helper, config, embeddings)
logger.info("took %.2f seconds", time.time() - start)
init = tf.global_variables_initializer()
saver = None
with tf.Session() as session:
session.run(init)
model.fit(session, saver, train, dev)
logger.info("Model did not crash!")
logger.info("Passed!")
def do_train(args):
# Set up some parameters.
config = Config()
helper, train, dev, train_raw, dev_raw = load_and_preprocess_data(args)
embeddings = load_embeddings(args, helper)
config.embed_size = embeddings.shape[1]
helper.save(config.output_path)
handler = logging.FileHandler(config.log_output)
handler.setLevel(logging.DEBUG)
handler.setFormatter(
logging.Formatter('%(asctime)s:%(levelname)s: %(message)s'))
logging.getLogger().addHandler(handler)
report = None #Report(Config.eval_output)
logger.info("Building model...", )
start = time.time()
model = WindowModel(helper, config, embeddings)
logger.info("took %.2f seconds", time.time() - start)
model.apply(init_weights)
model.fit(train, dev)
if report:
report.log_output(model.output(dev_raw))
report.save()
else:
# Save predictions in a text file.
output = model.output(dev_raw)
sentences, labels, predictions = zip(*output)
predictions = [[LBLS[l] for l in preds] for preds in predictions]
output = zip(sentences, labels, predictions)
with open(model.config.conll_output, 'w') as f:
write_conll(f, output)
with open(model.config.eval_output, 'w') as f:
for sentence, labels, predictions in output:
print_sentence(f, sentence, labels, predictions)
print 'What should the confusion matrix file name be?'
cmFileName = raw_input()
'''
print 'What should the output file name be?'
outputFileName = raw_input()
######################################
## get the data ##
######################################
# load in the data
debug = False
if len(sys.argv) > 2 and sys.argv[2] == "debug":
debug = True
helper, train_final_data, dev_final_data, test_final_data, train, dev, test = load_and_preprocess_data(
debug)
pretrained_embeddings = load_embeddings(
helper,
vocabPath="../Vectors/gloveVocab.txt",
vectorPath="../Vectors/glove.6B.200d.txt",
wordFirst=True,
embed_size=200)
Config.embed_size = pretrained_embeddings.shape[1]
# for later
neverOpened_gold = True
neverOpened_test = True
######################################
## define graph ##
def __init__(self, config, pretrained_embeddings):
self.pretrained_embeddings = pretrained_embeddings
self.config = config
self.encoder_inputs = None
self.decoder_inputs = None
self.decoder_targets = None
self.grad_norm = None
self.build()
if __name__ == '__main__':
# Get data and embeddings
start = time.time()
print("Loading data...")
train, dev, test, _, _, _, max_x, max_y, E, voc = load_and_preprocess_data(
output='tokens_debug.txt', debug=True)
print("Took {} seconds to load data".format(time.time() - start))
# Set up some parameters.
print(80 * "=")
print("INITIALIZING")
print(80 * "=")
config = Config()
config.voc_size = len(voc)
config.embedding_size = E.shape[1]
config.max_length_x = 250
config.max_length_y = 11
config.voc = voc
UNK_IDX = voc[UNK_TOKEN]
START_IDX = voc[START_TOKEN]
print 'What should the confusion matrix file name be?'
cmFileName = raw_input()
print 'What should the output file name be?'
outputFileName = raw_input()
######################################
## get the data ##
######################################
# load in the data
debug = False
if len(sys.argv) > 2 and sys.argv[2] == "debug":
debug = True
helper, train_final_data, dev_final_data, test_final_data, train, dev, test, country_dict_key_int = load_and_preprocess_data(
debug)
pretrained_embeddings = load_embeddings(
helper,
vocabPath="../Vectors/gloveVocab.txt",
vectorPath="../Vectors/glove.6B.200d.txt",
wordFirst=True,
embed_size=200)
Config.embed_size = pretrained_embeddings.shape[1]
# for later
neverOpened_gold = True
neverOpened_test = True
######################################
## define graph ##
def main():
# load data
print("Loading datasets...")
dataset, dataset_val, dataset_test = load_and_preprocess_data(
'cifar-100-python/')
print("Dataset loading completes.")
# setup metadata
metadata = get_metadata('cifar-100-python', 'meta')
config.data_info = list(dataset['data'].shape[1:])
config.fine_label_names = metadata['fine_label_names']
config.coarse_label_names = metadata['coarse_label_names']
config.label_mapping = get_label_mapping(metadata)
# create log directory
hyper_parameter_str = 'bs_{}_lr_{}'.format(
config.batch_size,
config.learning_rate,
)
if config.apply_batch_norm: hyper_parameter_str += '_batchnorm'
if config.add_more_layers: hyper_parameter_str += '_morelayers'
if config.larger_filter_size: hyper_parameter_str += '_largefilter'
if config.add_dropout: hyper_parameter_str += '_dropout'
train_dir = './train_dir/{}_{}_{}_{}'.format(
'cifar10', config.prefix, hyper_parameter_str,
time.strftime("%Y%m%d-%H%M%S"))
if not os.path.exists(train_dir): os.makedirs(train_dir)
print("Train Dir: {}".format(train_dir))
# reset default graph
tf.reset_default_graph()
# create model
model = Model(config)
# training setups
saver = tf.train.Saver(max_to_keep=100)
summary_writer = tf.summary.FileWriter(train_dir)
session_config = tf.ConfigProto(
gpu_options=tf.GPUOptions(allow_growth=True), device_count={'GPU': 1})
with tf.Session(config=session_config) as session:
session.run(tf.global_variables_initializer())
# buffers for train and val losses
train_losses = []
val_losses = []
# train model
for step in range(config.max_steps):
# run validation step
if step % config.val_step == 0:
val_batch = sample_batch(dataset_val, config.batch_size)
val_stats = run_single_step(session,
model,
val_batch,
summary_writer,
config.label_mapping,
mode='val')
val_losses.append([val_stats['step'], val_stats['loss']])
# run train step
train_batch = sample_batch(dataset, config.batch_size)
train_stats = run_single_step(session,
model,
train_batch,
summary_writer,
config.label_mapping,
mode='train',
log=step % config.log_step == 0)
train_losses.append([train_stats['step'], train_stats['loss']])
# save checkpoint
if step % config.save_checkpoint_step == 0:
print("Saved checkpoint at step {}".format(step))
saver.save(session,
os.path.join(train_dir, 'model'),
global_step=step)
# test model
test_logfile = os.path.join(train_dir, 'test_result.txt')
exp_results = run_single_step(session,
model,
dataset_test,
summary_writer,
config.label_mapping,
mode='test',
test_logfile=test_logfile)
# add loss curves to experiment results
exp_results['train_losses'] = train_losses
exp_results['val_losses'] = val_losses
# log test results
with open(os.path.join(train_dir, 'test_result.p'), 'wb') as f:
pickle.dump(exp_results, f)
print("Logged experiment results to {}".format(f.name))
# flush Tensorboard summaries
summary_writer.flush()
self.config = config
self.encoder_inputs = None
self.decoder_inputs = None
self.decoder_targets = None
self.length_encoder_inputs = None
self.length_decoder_inputs = None
self.grad_norm = None
self.build()
if __name__ == '__main__':
# Get data and embeddings
start = time.time()
print("Loading data...")
train, dev, test, _, _, _, max_x, max_y, E, voc = load_and_preprocess_data(
)
# train, dev, test, _, _, _, max_x, max_y, E, voc = load_and_preprocess_data(output = 'tokens_debug.txt', debug = True)
print("Took {} seconds to load data".format(time.time() - start))
# Set up some parameters.
print(80 * "=")
print("INITIALIZING")
print(80 * "=")
config = Config()
config.voc_size = len(voc)
config.embedding_size = E.shape[1]
config.max_length_x = 250
config.max_length_y = 11
config.voc = voc
config.idx2word = dict([[v, k] for k, v in voc.items()])