tf.flags.DEFINE_boolean("allow_soft_placement", True,
"Allow device soft device placement")
tf.flags.DEFINE_boolean("log_device_placement", False,
"Log placement of ops on devices")
tf.flags.DEFINE_boolean("new_data_format", False, "Use load_data_and_labels2")
FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")
# Load data.
if FLAGS.new_data_format:
x_raw, y_test = data_helpers.load_data_and_labels2(FLAGS.path_to_json_file)
y_test = np.argmax(y_test, axis=1)
else:
x_raw, y_test = data_helpers.load_data_and_labels(FLAGS.path_to_json_file)
y_test = np.argmax(y_test, axis=1)
# Map data into vocabulary
vocab_path = os.path.join(FLAGS.checkpoint_dir, "vocab/", "vocab")
vocab_processor = learn.preprocessing.VocabularyProcessor.restore(vocab_path)
x_test = np.array(list(vocab_processor.transform(x_raw)))
print("Total score for {} is {}".format(x_raw))
print("What is x_raw {}" % (x_raw))
print("\nEvaluating...\n")
def eval():
with tf.device('/gpu:0'):
# x_text, y = data_helpers.load_data_and_labels(FLAGS.rank0_dir, FLAGS.rank1_dir, FLAGS.rank2_dir, FLAGS.rank3_dir)
x_text, y = data_helpers.load_data_and_labels2(FLAGS.rank1_dir,
FLAGS.rank2_dir,
FLAGS.rank3_dir)
# Map data into vocabulary
text_path = os.path.join(FLAGS.checkpoint_dir_rnn, "..", "text_vocab")
text_vocab_processor = tf.contrib.learn.preprocessing.VocabularyProcessor.restore(
text_path)
x_eval = np.array(list(text_vocab_processor.transform(x_text)))
y_eval = np.argmax(y, axis=1)
# first model
# RNN
checkpoint_dir_rnn = tf.train.latest_checkpoint(FLAGS.checkpoint_dir_rnn)
rnn_logits_list = []
graph = tf.Graph()
with graph.as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=FLAGS.allow_soft_placement,
log_device_placement=FLAGS.log_device_placement)
sess = tf.Session(config=session_conf)
with sess.as_default():
# Load the saved meta graph and restore variables
saver = tf.train.import_meta_graph(
"{}.meta".format(checkpoint_dir_rnn))
saver.restore(sess, checkpoint_dir_rnn)
# Get the placeholders from the graph by name
input_text = graph.get_operation_by_name("input_text").outputs[0]
# input_y = graph.get_operation_by_name("input_y").outputs[0]
dropout_keep_prob = graph.get_operation_by_name(
"dropout_keep_prob").outputs[0]
# Tensors we want to evaluate
predictions = graph.get_operation_by_name(
"output/predictions").outputs[0]
logits = graph.get_operation_by_name("output/logits").outputs[0]
# Generate batches for one epoch
batches = data_helpers.batch_iter(list(x_eval),
FLAGS.batch_size,
1,
shuffle=False)
# Collect the predictions here
rnn_predictions = []
for x_batch in batches:
batch_predictions = sess.run(predictions,
{input_text: x_batch})
batch_logits = sess.run(logits, {input_text: x_batch})
rnn_predictions = np.concatenate(
[rnn_predictions, batch_predictions])
for i in range(len(batch_logits)):
rnn_logits_list.append(batch_logits[i])
correct_predictions = float(sum(rnn_predictions == y_eval))
print("Total number of test examples: {}".format(len(y_eval)))
print("Accuracy: {:g}".format(correct_predictions /
float(len(y_eval))))
# second model
# CNN
# Map data into vocabulary
text_path = os.path.join(FLAGS.checkpoint_dir_cnn, "..", "vocab")
text_vocab_processor = tf.contrib.learn.preprocessing.VocabularyProcessor.restore(
text_path)
x_eval = np.array(list(text_vocab_processor.transform(x_text)))
y_eval = np.argmax(y, axis=1)
checkpoint_dir_cnn = tf.train.latest_checkpoint(FLAGS.checkpoint_dir_cnn)
cnn_logits_list = []
graph2 = tf.Graph()
with graph2.as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=FLAGS.allow_soft_placement,
log_device_placement=FLAGS.log_device_placement)
sess2 = tf.Session(config=session_conf)
with sess2.as_default():
# Load the saved meta graph and restore variables
saver = tf.train.import_meta_graph(
"{}.meta".format(checkpoint_dir_cnn))
saver.restore(sess2, checkpoint_dir_cnn)
# Get the placeholders from the graph by name
input_text = graph2.get_operation_by_name("input_x").outputs[0]
# input_y = graph.get_operation_by_name("input_y").outputs[0]
dropout_keep_prob = graph2.get_operation_by_name(
"dropout_keep_prob").outputs[0]
# Tensors we want to evaluate
predictions = graph2.get_operation_by_name(
"output/predictions").outputs[0]
logits = graph2.get_operation_by_name("output/scores").outputs[0]
# Generate batches for one epoch
batches = data_helpers.batch_iter(list(x_eval),
FLAGS.batch_size,
1,
shuffle=False)
# Collect the predictions here
cnn_predictions = []
for x_batch in batches:
batch_predictions = sess2.run(predictions, {
input_text: x_batch,
dropout_keep_prob: 1.0
})
batch_logits = sess2.run(logits, {
input_text: x_batch,
dropout_keep_prob: 1.0
})
cnn_predictions = np.concatenate(
[cnn_predictions, batch_predictions])
for i in range(len(batch_logits)):
cnn_logits_list.append(batch_logits[i])
correct_predictions = float(sum(cnn_predictions == y_eval))
print("Total number of test examples: {}".format(len(y_eval)))
print("Accuracy: {:g}".format(correct_predictions /
float(len(y_eval))))
return y_eval, rnn_logits_list, cnn_logits_list
print("")
# Data Preparation
# ==================================================
# Load data
print("Loading data...")
x_text, y = data_helpers.load_data_and_labels2([
"./data/hotshare_train/artist.txt",
"./data/hotshare_train/emotion.txt",
"./data/hotshare_train/extend.txt",
"./data/hotshare_train/foods.txt",
"./data/hotshare_train/goplaces.txt",
"./data/hotshare_train/hotnews.txt",
"./data/hotshare_train/human.txt",
"./data/hotshare_train/nature.txt",
"./data/hotshare_train/news.txt",
"./data/hotshare_train/poets.txt",
"./data/hotshare_train/programmer.txt",
"./data/hotshare_train/scene.txt",
"./data/hotshare_train/talk.txt",
"./data/hotshare_train/travel.txt"
])
# Build vocabulary
max_document_length = max([len(x) for x in x_text])
vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length)
x = np.array(list(vocab_processor.fit_transform(x_text)))
# Randomly shuffle data
np.random.seed(10)
tf.flags.DEFINE_boolean("allow_soft_placement", True,
"Allow device soft device placement")
tf.flags.DEFINE_boolean("log_device_placement", False,
"Log placement of ops on devices")
FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")
# CHANGE THIS: Load data. Load your own data here
if FLAGS.eval_train:
#x_raw, y_test = data_helpers.load_data_and_labels(FLAGS.positive_data_file, FLAGS.negative_data_file)
x_raw, y_test = data_helpers.load_data_and_labels2(FLAGS.data_file)
y_test = np.argmax(y_test, axis=1)
else:
x_raw = ["a masterpiece four years in the making", "everything is off."]
y_test = [1, 0]
# Map data into vocabulary
vocab_path = os.path.join(FLAGS.checkpoint_dir, "..", "vocab")
vocab_processor = learn.preprocessing.VocabularyProcessor.restore(vocab_path)
x_test = np.array(list(vocab_processor.transform(x_raw)))
print("\nEvaluating...\n")
# Evaluation
# ==================================================
checkpoint_file = tf.train.latest_checkpoint(FLAGS.checkpoint_dir)
#! /usr/bin/env python
import tensorflow as tf
import data_helpers
#tf.flags.DEFINE_string("positive_data_file", "./data/rt-polaritydata/rt-polarity.pos", "Data source for the positive data.")
#tf.flags.DEFINE_string("positive_data_file", "./data/rt-polaritydata/test.pos", "Data source for the positive data.")
tf.flags.DEFINE_string("data_file", "../nsmc/small.txt",
"Data source for test.")
tf.flags.DEFINE_string("negative_data_file",
"./data/rt-polaritydata/rt-polarity.neg",
"Data source for the negative data.")
FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")
#x_text, y = data_helpers.load_data_and_labels(FLAGS.positive_data_file, FLAGS.negative_data_file)
x_text, y = data_helpers.load_data_and_labels2(FLAGS.data_file)
#x_text, y = data_helpers.load_data_and_labels2(FLAGS.positive_data_file)
print(x_text)
FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")
# CHANGE THIS: Load data. Load your own data here
if FLAGS.eval_train:
x_raw, y_test = data_helpers.load_data_and_labels2([
"./data/hotshare_eval/artist.txt", "./data/hotshare_eval/emotion.txt",
"./data/hotshare_eval/extend.txt", "./data/hotshare_eval/foods.txt",
"./data/hotshare_eval/goplaces.txt",
"./data/hotshare_eval/hotnews.txt", "./data/hotshare_eval/human.txt",
"./data/hotshare_eval/nature.txt", "./data/hotshare_eval/news.txt",
"./data/hotshare_eval/poets.txt",
"./data/hotshare_eval/programmer.txt",
"./data/hotshare_eval/scene.txt", "./data/hotshare_eval/talk.txt",
"./data/hotshare_eval/travel.txt"
])
y_test = np.argmax(y_test, axis=1)
else:
x_raw = ["a masterpiece four years in the making", "everything is off."]
y_test = [1, 0]
# Map data into vocabulary
vocab_path = os.path.join(FLAGS.checkpoint_dir, "..", "vocab")
vocab_processor = learn.preprocessing.VocabularyProcessor.restore(vocab_path)
x_test = np.array(list(vocab_processor.transform(x_raw)))
tf.flags.DEFINE_boolean("log_device_placement", False,
"Log placement of ops on devices")
FLAGS = tf.flags.FLAGS
FLAGS._parse_flags()
print("\nParameters:")
for attr, value in sorted(FLAGS.__flags.items()):
print("{}={}".format(attr.upper(), value))
print("")
# Data Preparation
# ==================================================
# Load data
print("Loading data...")
x_text, y = data_helpers.load_data_and_labels2(FLAGS.data_file,
FLAGS.num_classes)
# Build vocabulary
max_document_length = max([len(x) for x in x_text])
vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length)
print(x_text)
x = np.array(list(vocab_processor.fit_transform(x_text)))
# Randomly shuffle data
np.random.seed(10)
shuffle_indices = np.random.permutation(np.arange(len(y)))
x_shuffled = x[shuffle_indices]
y_shuffled = y[shuffle_indices]
# Split train/test set
# TODO: This is very crude, should use cross-validation
def train():
with tf.device('/gpu:0'):
# x_text, y = data_helpers.load_data_and_labels(FLAGS.rank0_dir, FLAGS.rank1_dir, FLAGS.rank2_dir, FLAGS.rank3_dir)
x_text, y = data_helpers.load_data_and_labels2(FLAGS.rank1_dir,
FLAGS.rank2_dir,
FLAGS.rank3_dir)
max_document_length = max([len(x.split(" ")) for x in x_text])
text_vocab_processor = tf.contrib.learn.preprocessing.VocabularyProcessor(
max_document_length)
x = np.array(list(text_vocab_processor.fit_transform(x_text)))
print("Text Vocabulary Size: {:d}".format(
len(text_vocab_processor.vocabulary_)))
print("x = {0}".format(x.shape))
print("y = {0}".format(y.shape))
print("")
# Randomly shuffle data
np.random.seed(10)
shuffle_indices = np.random.permutation(np.arange(len(y)))
x_shuffled = x[shuffle_indices]
y_shuffled = y[shuffle_indices]
x_train, y_train, x_dev, y_dev = make_dataset(x_shuffled, y_shuffled, 10)
# print("Train/Dev split: {:d}/{:d}\n".format(len(y_train), len(y_dev)))
with tf.Graph().as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=FLAGS.allow_soft_placement,
log_device_placement=FLAGS.log_device_placement)
sess = tf.Session(config=session_conf)
with sess.as_default():
rnn = RNN(sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=len(text_vocab_processor.vocabulary_),
embedding_size=FLAGS.embedding_dim,
cell_type=FLAGS.cell_type,
hidden_size=FLAGS.hidden_size,
l2_reg_lambda=FLAGS.l2_reg_lambda)
# Define Training procedure
global_step = tf.Variable(0, name="global_step", trainable=False)
train_op = tf.train.AdamOptimizer(FLAGS.learning_rate).minimize(
rnn.loss, global_step=global_step)
# Output directory for models and summaries
timestamp = str(int(time.time()))
out_dir = os.path.abspath(
os.path.join(os.path.curdir, "runs", timestamp))
print("Writing to {}\n".format(out_dir))
# Summaries for loss and accuracy
loss_summary = tf.summary.scalar("loss", rnn.loss)
acc_summary = tf.summary.scalar("accuracy", rnn.accuracy)
# Train Summaries
train_summary_op = tf.summary.merge([loss_summary, acc_summary])
train_summary_dir = os.path.join(out_dir, "summaries", "train")
train_summary_writer = tf.summary.FileWriter(
train_summary_dir, sess.graph)
# Dev summaries
dev_summary_op = tf.summary.merge([loss_summary, acc_summary])
dev_summary_dir = os.path.join(out_dir, "summaries", "dev")
dev_summary_writer = tf.summary.FileWriter(dev_summary_dir,
sess.graph)
# Checkpoint directory. Tensorflow assumes this directory already exists so we need to create it
checkpoint_dir = os.path.abspath(
os.path.join(out_dir, "checkpoints"))
checkpoint_prefix = os.path.join(checkpoint_dir, "model")
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
saver = tf.train.Saver(tf.global_variables(),
max_to_keep=FLAGS.num_checkpoints)
# Write vocabulary
text_vocab_processor.save(os.path.join(out_dir, "text_vocab"))
# Initialize all variables
sess.run(tf.global_variables_initializer())
# Pre-trained word2vec
if FLAGS.word2vec:
# initial matrix with random uniform
initW = np.random.uniform(-0.25, 0.25, (len(
text_vocab_processor.vocabulary_), FLAGS.embedding_dim))
# load any vectors from the word2vec
print("Load word2vec file {0}".format(FLAGS.word2vec))
with open(FLAGS.word2vec, "rb") as f:
header = f.readline()
vocab_size, layer1_size = map(int, header.split())
binary_len = np.dtype('float32').itemsize * layer1_size
for line in range(vocab_size):
word = []
while True:
ch = f.read(1).decode('latin-1')
if ch == ' ':
word = ''.join(word)
break
if ch != '\n':
word.append(ch)
idx = text_vocab_processor.vocabulary_.get(word)
if idx != 0:
initW[idx] = np.fromstring(f.read(binary_len),
dtype='float32')
else:
f.read(binary_len)
sess.run(rnn.W_text.assign(initW))
print("Success to load pre-trained word2vec model!\n")
for train_index, test_index in KFold(10).split(x):
x_train, x_dev = x[train_index], x[test_index]
y_train, y_dev = y[train_index], y[test_index]
# Generate batches
batches = data_helpers.batch_iter(list(zip(x_train, y_train)),
FLAGS.batch_size,
FLAGS.num_epochs)
# Training loop. For each batch...
for batch in batches:
x_batch, y_batch = zip(*batch)
# Train
feed_dict = {
rnn.input_text: x_batch,
rnn.input_y: y_batch,
rnn.dropout_keep_prob: FLAGS.dropout_keep_prob
}
_, step, summaries, loss, accuracy, = sess.run([
train_op, global_step, train_summary_op, rnn.loss,
rnn.accuracy
], feed_dict)
train_summary_writer.add_summary(summaries, step)
# Training log display
if step % FLAGS.display_every == 0:
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(
time_str, step, loss, accuracy))
# Evaluation
if step % FLAGS.evaluate_every == 0:
print("\nEvaluation:")
feed_dict_dev = {
rnn.input_text: x_dev,
rnn.input_y: y_dev,
rnn.dropout_keep_prob: FLAGS.dropout_keep_prob
}
summaries_dev, loss, accuracy = sess.run(
[dev_summary_op, rnn.loss, rnn.accuracy],
feed_dict_dev)
dev_summary_writer.add_summary(summaries_dev, step)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}\n".format(
time_str, step, loss, accuracy))
# Model checkpoint
if step % FLAGS.checkpoint_every == 0:
path = saver.save(sess,
checkpoint_prefix,
global_step=step)
print("Saved model checkpoint to {}\n".format(path))
# Export model
export_path = 'models/1'
builder = tf.saved_model.builder.SavedModelBuilder(export_path)
g = tf.get_default_graph()
inp = g.get_tensor_by_name(rnn.input_text.name)
out_pred = g.get_tensor_by_name(rnn.predictions.name)
out_logit = g.get_tensor_by_name(rnn.logits.name)
# Build the signature_def_map.
classification_inputs = tf.saved_model.utils.build_tensor_info(inp)
classification_outputs_classes = tf.saved_model.utils.build_tensor_info(
out_pred)
classification_outputs_logits = tf.saved_model.utils.build_tensor_info(
out_logit)
classification_signature = (
tf.saved_model.signature_def_utils.build_signature_def(
inputs={
tf.saved_model.signature_constants.CLASSIFY_INPUTS:
classification_inputs
},
outputs={
tf.saved_model.signature_constants.CLASSIFY_OUTPUT_CLASSES:
classification_outputs_classes,
tf.saved_model.signature_constants.CLASSIFY_OUTPUT_SCORES:
classification_outputs_logits
},
method_name=tf.saved_model.signature_constants.
CLASSIFY_METHOD_NAME))
builder.add_meta_graph_and_variables(
sess, [tf.saved_model.tag_constants.SERVING],
signature_def_map={
tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY:
classification_signature,
},
main_op=tf.tables_initializer(),
strip_default_attrs=True)
builder.save()
