saver = tf.train.import_meta_graph("{}.meta".format(checkpoint_file))
saver.restore(sess, checkpoint_file)
# Get the placeholders from the graph by name
input_x = graph.get_operation_by_name("input_x").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]
# Generate batches for one epoch
batches = processData.batch_iter(list(x_test),
FLAGS.batch_size,
1,
shuffle=False)
# Collect the predictions here
all_predictions = []
for x_test_batch in batches:
batch_predictions = sess.run(predictions, {
input_x: x_test_batch,
dropout_keep_prob: 1.0
})
all_predictions = np.concatenate(
[all_predictions, batch_predictions])
print all_predictions
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: 1.0
}
step, summaries, loss, accuracy = sess.run(
[global_step, dev_summary_op, cnn.loss, cnn.accuracy],
feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(
time_str, step, loss, accuracy))
if writer:
writer.add_summary(summaries, step)
# Generate batches
batches = processData.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_step(x_batch, y_batch)
current_step = tf.train.global_step(sess, global_step)
if current_step % FLAGS.evaluate_every == 0:
print("\nEvaluation:")
dev_step(x_dev, y_dev, writer=dev_summary_writer)
print("")
if current_step % FLAGS.checkpoint_every == 0:
path = saver.save(sess,
checkpoint_prefix,
global_step=current_step)
print("Saved model checkpoint to {}\n".format(path))
def train_tensorflow(x_train, x_test, y_train, y_test):
# Training
with tf.Graph().as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=False,
log_device_placement=False)
sess = tf.Session(config=session_conf)
with sess.as_default():
cnn = TextCNN(
sequence_length=100,
num_classes=2,
vocab_size=len(x_train),
embedding_size=100,
filter_sizes=list(map(int, filter_sizes.split(","))),
num_filters=128,
l2_reg_lambda=0)
# Define Training procedure
global_step = tf.Variable(0, name="global_step", trainable=False)
optimizer = tf.train.AdamOptimizer(1e-3)
grads_and_vars = optimizer.compute_gradients(cnn.loss)
train_op = optimizer.apply_gradients(grads_and_vars, global_step=global_step)
# Keep track of gradient values and sparsity (optional)
grad_summaries = []
for g, v in grads_and_vars:
if g is not None:
grad_hist_summary = tf.summary.histogram("{}/grad/hist".format(v.name), g)
sparsity_summary = tf.summary.scalar("{}/grad/sparsity".format(v.name), tf.nn.zero_fraction(g))
grad_summaries.append(grad_hist_summary)
grad_summaries.append(sparsity_summary)
grad_summaries_merged = tf.summary.merge(grad_summaries)
# 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", cnn.loss)
acc_summary = tf.summary.scalar("accuracy", cnn.accuracy)
# Train Summaries
train_summary_op = tf.summary.merge([loss_summary, acc_summary, grad_summaries_merged])
train_summary_dir = os.path.join(out_dir, "summaries", "train")
# out_dir --> runs
#train_summary_writer = tf.summary.FileWriter("tensorboard_train/", graph=tf.get_default_graph())
train_summary_writer = tf.summary.FileWriter(train_summary_dir, graph=sess.graph)
# Dev summaries
dev_summary_op = tf.summary.merge([loss_summary, acc_summary])
dev_summary_dir = os.path.join(out_dir, "summaries", "dev")
# out_dir --> runs
dev_summary_writer = tf.summary.FileWriter(dev_summary_dir, graph=sess.graph)
#dev_summary_writer = tf.summary.FileWriter("tensorboard_dev/", graph=tf.get_default_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=5)
# Write vocabulary
#vocab_processor.save(os.path.join(out_dir, "vocab"))
# Initialize all variables
sess.run(tf.global_variables_initializer())
def train_step(x_batch, y_batch):
"""
A single training step
"""
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: 1.0
}
_, step, summaries, loss, accuracy = sess.run(
[train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy],
feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
train_summary_writer.add_summary(summaries, step)
def dev_step(x_batch, y_batch, writer=None):
"""
Evaluates model on a dev set
"""
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: 1.0
}
step, summaries, loss, accuracy = sess.run(
[global_step, dev_summary_op, cnn.loss, cnn.accuracy],
feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
if writer:
writer.add_summary(summaries, step)
# Generate batches
train_batches = batch_iter(
list(zip(x_train, y_train)), batch_size, training_epochs) #batch size, training_epochs
dev_batches = batch_iter(
list(zip(x_test, y_test)), batch_size, training_epochs)
# Training loop. For each batch...
for train_batch, dev_batch in zip(train_batches, dev_batches):
x_train_batch, y_train_batch = zip(*train_batch)
train_step(x_train_batch, y_train_batch)
current_step = tf.train.global_step(sess, global_step)
print current_step
if current_step % 100 == 0:
print("\nEvaluation:")
x_dev_batch, y_dev_batch = zip(*dev_batch)
dev_step(x_dev_batch, y_dev_batch, writer=dev_summary_writer)
print("")
if current_step % 100 == 0:
path = saver.save(sess, checkpoint_prefix, global_step=current_step)
print("Saved model checkpoint to {}\n".format(path))
def oneLayer():
train_X, dev_X, test_X, train_y, dev_y, test_y = get_glove_data()
# Parameters
RANDOM_SEED = 42
tf.set_random_seed(RANDOM_SEED)
learning_rate = 0.01
training_epochs = 100
batch_size = 500
display_step = 1
evaluate_every = 25
# Network Parameters
n_hidden_1 = 256
n_hidden_2 = 128
n_input = train_X.shape[1] # MNIST data input (img shape: 28*28)
n_classes = train_y.shape[1] # MNIST total classes (0-9 digits)
global_step = tf.Variable(0, name="global_step", trainable=False)
# tf Graph Input
x = tf.placeholder(tf.float32, [None, n_input], name='X')
# 0-9 digits recognition => 10 classes
y = tf.placeholder(tf.float32, [None, n_classes], name='y')
# Store layers weight & bias
weights = {
'w1': tf.Variable(tf.random_normal([n_input, n_classes]), name='W1'),
}
biases = {
'b1': tf.Variable(tf.random_normal([n_classes]), name='b1'),
}
pred = single_perceptron(x, weights, biases)
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y))
# Gradient Descent
optimizer = tf.train.GradientDescentOptimizer(learning_rate)
# Op to calculate every variable gradient
grads = tf.gradients(loss, tf.trainable_variables())
grads = list(zip(grads, tf.trainable_variables()))
# Op to update all variables according to their gradient
apply_grads = optimizer.apply_gradients(grads_and_vars=grads, global_step=global_step)
acc = tf.equal(tf.argmax(pred, axis=1), tf.argmax(y, axis=1))
acc = tf.reduce_mean(tf.cast(acc, tf.float32))
# Initializing the variables
init = tf.global_variables_initializer()
# Launch the graph
with tf.Session() as sess:
sess.run(init)
# Output directory for models and summaries
timestamp = str(int(time.time()))
out_dir = os.path.abspath(os.path.join(os.path.curdir, "oneLayer_runs", timestamp))
print("Writing to {}\n".format(out_dir))
# Create a summary to monitor cost tensor
loss_summary = tf.summary.scalar("loss", loss)
# Create a summary to monitor accuracy tensor
acc_summary = tf.summary.scalar("accuracy", acc)
# Summarize all gradients
grad_summaries = []
for g, v in grads:
if g is not None:
grad_hist_summary = tf.summary.histogram("{}/grad/hist".format(v.name), g)
sparsity_summary = tf.summary.scalar("{}/grad/sparsity".format(v.name), tf.nn.zero_fraction(g))
grad_summaries.append(grad_hist_summary)
grad_summaries.append(sparsity_summary)
grad_summaries_merged = tf.summary.merge(grad_summaries)
# Train Summaries
train_summary_op = tf.summary.merge([loss_summary, acc_summary, grad_summaries_merged])
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)
# Generate batches
train_batches = batch_iter(
list(zip(train_X, train_y)), batch_size, training_epochs)
dev_batches = batch_iter(
list(zip(test_X, test_y)), batch_size, training_epochs)
# Training cycle
for train_batch, dev_batch in zip(train_batches, dev_batches):
x_train_batch, y_train_batch = zip(*train_batch)
_, step, summaries, loss_1, accuracy = sess.run([apply_grads, global_step, train_summary_op, loss, acc],
feed_dict={x: x_train_batch, y: y_train_batch})
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss_1, accuracy))
train_summary_writer.add_summary(summaries, step)
train_summary_writer.flush()
current_step = tf.train.global_step(sess, global_step)
if current_step % evaluate_every == 0:
print("\nEvaluation:")
x_dev_batch, y_dev_batch = zip(*dev_batch)
step, summaries, y_pred, loss_1, accuracy = sess.run([global_step, dev_summary_op, pred, loss, acc],
feed_dict={x: x_dev_batch, y: y_dev_batch})
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss_1, accuracy))
dev_summary_writer.add_summary(summaries, step)
dev_summary_writer.flush()
print("")
# Report precision, recall, F1
print("Precision: {}%".format(100*metrics.precision_score(np.argmax(y_dev_batch, axis=1), np.argmax(y_pred, axis=1), average="weighted")))
print("Recall: {}%".format(100*metrics.recall_score(np.argmax(y_dev_batch, axis=1), np.argmax(y_pred, axis=1), average="weighted")))
print("f1_score: {}%".format(100*metrics.f1_score(np.argmax(y_dev_batch, axis=1), np.argmax(y_pred, axis=1), average="weighted")))
train_summary_writer.close()
dev_summary_writer.close()
def train_tensorflow(x_train, x_test, y_train, y_test, vocab_processor):
# Training
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():
cnn = TextCNN(
sequence_length=x_train.shape[1],
num_classes=2,
vocab_size=len(vocab_processor.vocabulary_),
embedding_size=FLAGS.embedding_dim,
filter_sizes=list(map(int, FLAGS.filter_sizes.split(","))),
num_filters=FLAGS.num_filters,
l2_reg_lambda=FLAGS.l2_reg_lambda)
# Define Training procedure
global_step = tf.Variable(0, name="global_step", trainable=False)
optimizer = tf.train.AdamOptimizer(1e-3)
grads_and_vars = optimizer.compute_gradients(cnn.loss)
train_op = optimizer.apply_gradients(grads_and_vars, global_step=global_step)
# Keep track of gradient values and sparsity (optional)
grad_summaries = []
for g, v in grads_and_vars:
if g is not None:
grad_hist_summary = tf.summary.histogram("{}/grad/hist".format(v.name), g)
sparsity_summary = tf.summary.scalar("{}/grad/sparsity".format(v.name), tf.nn.zero_fraction(g))
grad_summaries.append(grad_hist_summary)
grad_summaries.append(sparsity_summary)
grad_summaries_merged = tf.summary.merge(grad_summaries)
# 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", cnn.loss)
acc_summary = tf.summary.scalar("accuracy", cnn.accuracy)
# Train Summaries
train_summary_op = tf.summary.merge([loss_summary, acc_summary, grad_summaries_merged])
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
vocab_processor.save(os.path.join(out_dir, "vocab"))
# Initialize all variables
sess.run(tf.global_variables_initializer())
if FLAGS.enable_word_embeddings and cfg['word_embeddings']['default'] is not None:
vocabulary = vocab_processor.vocabulary_
initEmbedding = None
if embedding_name == 'word2vec':
# load embedding vectors from the word2vec
print("Load word2vec file {}".format(cfg['word_embeddings']['word2vec']['path']))
initEmbedding = load_embedding_vectors_word2vec(vocabulary,
cfg['word_embeddings']['word2vec']['path'],
cfg['word_embeddings']['word2vec']['binary'])
print("word2vec file has been loaded")
elif embedding_name == 'glove':
# load embedding vectors from the glove
print("Load glove file {}".format(cfg['word_embeddings']['glove']['path']))
initEmbedding = load_embedding_vectors_glove(vocabulary,
cfg['word_embeddings']['glove']['path'],
embedding_dimension)
print("glove file has been loaded\n")
sess.run(cnn.E.assign(initEmbedding))
def train_step(x_batch, y_batch):
"""
A single training step
"""
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: FLAGS.dropout_keep_prob
}
_, step, summaries, loss, accuracy = sess.run(
[train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy],
feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
train_summary_writer.add_summary(summaries, step)
def dev_step(x_batch, y_batch, writer=None):
"""
Evaluates model on a dev set
"""
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: 1.0
}
step, summaries, loss, accuracy = sess.run(
[global_step, dev_summary_op, cnn.loss, cnn.accuracy],
feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(time_str, step, loss, accuracy))
if writer:
writer.add_summary(summaries, step)
# Generate batches
train_batches = batch_iter(
list(zip(x_train, y_train)), FLAGS.batch_size, FLAGS.num_epochs)
dev_batches = batch_iter(
list(zip(x_test, y_test)), FLAGS.batch_size, FLAGS.num_epochs)
# Training loop. For each batch...
for train_batch, dev_batch in zip(train_batches, dev_batches):
x_train_batch, y_train_batch = zip(*train_batch)
train_step(x_train_batch, y_train_batch)
current_step = tf.train.global_step(sess, global_step)
# if current_step == 1000:
# # Test on Arun's data after 10 epochs.
# x_dev_batch = generateVinodTrain()
# y_dev_batch = generateVinodTest()
# dev_step(x_dev_batch, y_dev_batch, writer=dev_summary_writer)
# break
if current_step % FLAGS.evaluate_every == 0:
print("\nEvaluation:")
x_dev_batch, y_dev_batch = zip(*dev_batch)
dev_step(x_dev_batch, y_dev_batch, writer=dev_summary_writer)
print("")
if current_step % FLAGS.checkpoint_every == 0:
path = saver.save(sess, checkpoint_prefix, global_step=current_step)
print("Saved model checkpoint to {}\n".format(path))
# Save parameters to file
params_file = open(os.path.join(outdir, 'params.pkl'), 'wb')
pkl.dump(params, params_file, True)
params_file.close()
# Simple Cross validation
# TODO use k-fold cross validation
x_train, x_valid, y_train, y_valid, train_lengths, valid_lengths = train_test_split(data,
labels,
lengths,
test_size=FLAGS.test_size,
random_state=22)
# Batch iterator
train_data = processData.batch_iter(x_train, y_train, train_lengths, FLAGS.batch_size, FLAGS.num_epochs)
# Train
# =============================================================================
with tf.Graph().as_default():
with tf.Session() as sess:
if FLAGS.clf == 'clstm':
classifier = clstm_clf(FLAGS)
else:
raise ValueError('clf should be one of [cnn, lstm, blstm, clstm]')
# Train procedure
global_step = tf.Variable(0, name='global_step', trainable=False)
optimizer = tf.train.AdamOptimizer(FLAGS.learning_rate)