def eval(w2v_model):
# Evaluation评估
checkpoint_file = tf.train.latest_checkpoint(
FLAGS.checkpoint_dir) #查找最新保存的checkpoint文件的文件名
graph = tf.Graph() #创建图层
# #定义属于计算图graph的张量和操作
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加载保存的元图并恢复变量
saver = tf.train.import_meta_graph(
"{}.meta".format(checkpoint_file))
saver.restore(sess, checkpoint_file)
# Get the placeholders from the graph by name从图表中按名称获取占位符
input_x = graph.get_operation_by_name("input_x").outputs[0]
dropout_keep_prob = graph.get_operation_by_name(
"dropout_keep_prob").outputs[0]
# Tensors we want to evaluate我们要计算的张量
predictions = graph.get_operation_by_name(
"output/predictions").outputs[0]
x_test, y_test = load_data(w2v_model, 5)
# Generate batches for one epoch为一个epoch生成批处理
batches = data_helpers.batch_iter(list(x_test),
FLAGS.batch_size,
1,
shuffle=False)
# Collect the predictions here在这里收集预测值
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]) #数据拼接操作
# Print accuracy if y_test is defined如果定义了y_test，则打印精度
if y_test is not None:
correct_predictions = float(sum(all_predictions == y_test))
print("Total number of test examples: {}".format(len(y_test))) #测试实例总数
print("Accuracy: {:g}".format(correct_predictions /
float(len(y_test)))) #得出精确值
# Save the evaluation to a csv将评估保存到csv
predictions_human_readable = np.column_stack(
all_predictions) #column_stack()连接矩阵，以行的方式
out_path = os.path.join(FLAGS.checkpoint_dir, "..",
"prediction.csv") #保存评估数据
print("Saving evaluation to {0}".format(out_path)) #输出保存路径
with open(out_path, 'w') as f:
csv.writer(f).writerows(predictions_human_readable)
def dev_test():
batches_dev = data_helpers.batch_iter(list(zip(x_dev, y_dev)),
FLAGS.batch_size, 1)
for batch_dev in batches_dev:
x_batch_dev, y_batch_dev = zip(*batch_dev)
dev_step(x_batch_dev,
y_batch_dev,
writer=dev_summary_writer)
def eval(w2v_model):
# Evaluation
# ==================================================
checkpoint_file = tf.train.latest_checkpoint(FLAGS.checkpoint_dir)
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_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]
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]
x_test, y_test = load_data(w2v_model, 1290)
# Generate batches for one epoch
batches = data_helpers.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 accuracy if y_test is defined
if y_test is not None:
correct_predictions = float(sum(all_predictions == y_test))
print("Total number of test examples: {}".format(len(y_test)))
print("Accuracy: {:g}".format(correct_predictions /
float(len(y_test))))
# Save the evaluation to a csv
predictions_human_readable = np.column_stack(all_predictions)
out_path = os.path.join(FLAGS.checkpoint_dir, "..", "prediction.csv")
print("Saving evaluation to {0}".format(out_path))
with open(out_path, 'w') as f:
csv.writer(f).writerows(predictions_human_readable)
def dev_test():
batches_dev = batch_iter(list(zip(x_dev, y_dev)),
FLAGS.batch_size, 1)
prediction_all = []
y_true_all = []
for batch_dev in batches_dev:
x_batch_dev, y_batch_dev = zip(*batch_dev)
predictions = dev_step(x_batch_dev,
y_batch_dev,
writer=dev_summary_writer)
prediction_all.extend(predictions)
y_true_all.extend(np.argmax(y_batch_dev, axis=1).tolist())
p, r, micro_p, micro_r, micro_f1, macro_f1, accuracy1 = fastF1(
y_true_all, prediction_all, FLAGS.num_class)
print(
"test P: {:.2f}%, R: {:.2f}%, micro_p: {:.2f}%, micro_r: {:.2f}%,Micro_f1: {:.2f}%, Macro_f1: {:.2f}%, Accuracy: {:.2f}%"
.format(p, r, micro_p, micro_r, micro_f1, macro_f1,
accuracy1))
def train(w2v_model):
# Training
# ==================================================
x_train, x_dev, y_train, y_dev, vocab_size = load_data(w2v_model)
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(w2v_model,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
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())
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,(w,idx) = sess.run(
# [train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy,cnn.get_w2v_W()],
# feed_dict)
_, 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))
# print w[:2],idx[:2]
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
batches = data_helpers.batch_iter(list(zip(x_train, y_train)),
FLAGS.batch_size,
FLAGS.num_epochs)
def dev_test():
batches_dev = data_helpers.batch_iter(list(zip(x_dev, y_dev)),
FLAGS.batch_size, 1)
for batch_dev in batches_dev:
x_batch_dev, y_batch_dev = zip(*batch_dev)
dev_step(x_batch_dev,
y_batch_dev,
writer=dev_summary_writer)
# 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)
# Training loop. For each batch...
if current_step % FLAGS.evaluate_every == 0:
print("\nEvaluation:")
dev_test()
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))
# Get the placeholders from the graph by name
input_x = graph.get_operation_by_name("input_x").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]
scores = graph.get_operation_by_name("output/scores").outputs[0]
x_test, y_test = load_data(w2v_model, 1290)
# Generate batches for one epoch
batches = data_helpers.batch_iter(list(x_test),
FLAGS.batch_size,
1,
shuffle=False)
# Collect the predictions here
all_predictions = []
all_probabilities = None
for index, x_test_batch in enumerate(batches):
batch_predictions = sess.run([predictions, scores], {
input_x: x_test_batch,
dropout_keep_prob: 1.0
})
all_predictions = np.concatenate(
[all_predictions, batch_predictions[0]])
probabilities = softmax(batch_predictions[1])
def dev_test():
batches_dev = data_helpers.batch_iter(list(zip(x_dev, y_dev)),
batch_size, 1)
for batch_dev in batches_dev:
x_batch_dev, y_batch_dev = zip(*batch_dev)
dev_step(x_batch_dev, y_batch_dev, writer=None)
def train():
# Training
# ==================================================
# x_train, x_dev, y_train, y_dev ,vocab_size= load_data(w2v_model)
with tf.Graph().as_default():
session_conf = tf.ConfigProto(
allow_soft_placement=allow_soft_placement,
log_device_placement=log_device_placement)
sess = tf.Session(config=session_conf)
with sess.as_default():
if (nn_type == "text_cnn"):
nn = TextCNN(model_type=model_type,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
embedding_size=embedding_dim,
filter_sizes=list(
map(int, filter_sizes.split(","))),
num_filters=num_filters,
l2_reg_lambda=l2_reg_lambda)
elif nn_type == "text_birnn":
nn = TextBiRNN(model_type=model_type,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
embedding_size=embedding_dim,
rnn_size=128,
num_layers=3,
l2_reg_lambda=l2_reg_lambda)
elif nn_type == "text_rnn":
nn = TextRNN(model_type=model_type,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
embedding_size=embedding_dim,
rnn_size=128,
num_layers=3,
l2_reg_lambda=l2_reg_lambda)
elif nn_type == "text_rcnn":
nn = TextBiRNN(model_type=model_type,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
embedding_size=embedding_dim,
rnn_size=128,
num_layers=3,
l2_reg_lambda=l2_reg_lambda)
elif nn_type == "text_dnn":
nn = TextDNN(model_type=model_type,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
embedding_size=embedding_dim,
hidden_layes=2,
hidden_size=128,
l2_reg_lambda=l2_reg_lambda)
elif nn_type == "text_fasttext":
nn = TextFast(model_type=model_type,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
embedding_size=embedding_dim,
l2_reg_lambda=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(nn.loss)
train_op = optimizer.apply_gradients(grads_and_vars,
global_step=global_step)
# Initialize all variables
sess.run(tf.global_variables_initializer())
def train_step(x_batch, y_batch):
"""
A single training step
"""
feed_dict = {
nn.input_x: x_batch,
nn.input_y: y_batch,
nn.dropout_keep_prob: dropout_keep_prob
}
_, step, loss, accuracy = sess.run(
[train_op, global_step, nn.loss, nn.accuracy], feed_dict)
time_str = datetime.datetime.now().isoformat()
print("{}: step {}, loss {:g}, acc {:g}".format(
time_str, step, loss, accuracy))
# print w[:2],idx[:2]
# train_summary_writer.add_summary(summaries, step)
def dev_step(x_batch, y_batch, writer=None):
"""
Evaluates model on a dev set
"""
feed_dict = {
nn.input_x: x_batch,
nn.input_y: y_batch,
nn.dropout_keep_prob: 1.0
}
step, loss, accuracy = sess.run(
[global_step, nn.loss, nn.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 = data_helpers.batch_iter(list(zip(x_train, y_train)),
batch_size, num_epochs)
def dev_test():
batches_dev = data_helpers.batch_iter(list(zip(x_dev, y_dev)),
batch_size, 1)
for batch_dev in batches_dev:
x_batch_dev, y_batch_dev = zip(*batch_dev)
dev_step(x_batch_dev, y_batch_dev, writer=None)
# Training loop. For each batch...
for batch in batches:
x_batch, y_batch = zip(*batch)
train_step(x_batch, y_batch)
def train(w2v_model,
epsilon=8 / 255,
alpha=10 / 255,
K=5,
is_free=False,
mode=None):
# Training
# ==================================================
x_train, x_dev, y_train, y_dev, vocab_size = load_data(w2v_model)
# fgsm = FGSM()
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(w2v_model,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
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,
sess=sess,
mode=mode)
# 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])
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())
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, scores, l2_loss, predictions = sess.run(
[
train_op, global_step, train_summary_op, cnn.loss,
cnn.accuracy, cnn.scores, cnn.l2_loss, cnn.predictions
], feed_dict)
time_str = datetime.datetime.now().isoformat()
p, r, micro_p, micro_r, micro_f1, macro_f1, accuracy1 = fastF1(
np.argmax(y_batch, axis=1), predictions, FLAGS.num_class)
print(
"train {}: step {}, loss {:g}, P: {:.2f}%, R: {:.2f}%, micro_p: {:.2f}%, micro_r: {:.2f}%,Micro_f1: {:.2f}%, Macro_f1: {:.2f}%, Accuracy: {:.2f}%"
.format(time_str, step, loss, p, r, micro_p, micro_r,
micro_f1, macro_f1, accuracy1))
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, predictions = sess.run([
global_step, dev_summary_op, cnn.loss, cnn.accuracy,
cnn.predictions
], feed_dict)
time_str = datetime.datetime.now().isoformat()
# p, r, micro_p, micro_r, micro_f1, macro_f1, accuracy1 = fastF1(np.argmax(y_batch, axis=1), predictions,
# 2)
print("test {}: step {}, loss {:g}".format(
time_str, step, loss))
if writer:
writer.add_summary(summaries, step)
return predictions
# Generate batches
batches = batch_iter(list(zip(x_train, y_train)), FLAGS.batch_size,
FLAGS.num_epochs)
num_batches_per_epoch = int(
(len(list(zip(x_train, y_train))) - 1) / FLAGS.batch_size) + 1
def dev_test():
batches_dev = batch_iter(list(zip(x_dev, y_dev)),
FLAGS.batch_size, 1)
prediction_all = []
y_true_all = []
for batch_dev in batches_dev:
x_batch_dev, y_batch_dev = zip(*batch_dev)
predictions = dev_step(x_batch_dev,
y_batch_dev,
writer=dev_summary_writer)
prediction_all.extend(predictions)
y_true_all.extend(np.argmax(y_batch_dev, axis=1).tolist())
p, r, micro_p, micro_r, micro_f1, macro_f1, accuracy1 = fastF1(
y_true_all, prediction_all, FLAGS.num_class)
print(
"test P: {:.2f}%, R: {:.2f}%, micro_p: {:.2f}%, micro_r: {:.2f}%,Micro_f1: {:.2f}%, Macro_f1: {:.2f}%, Accuracy: {:.2f}%"
.format(p, r, micro_p, micro_r, micro_f1, macro_f1,
accuracy1))
# dev_step(x_dev, y_dev, writer=dev_summary_writer)
# Training loop. For each batch...
num_batches_per_epoch_ = num_batches_per_epoch
if is_free:
num_batches_per_epoch_ = int(num_batches_per_epoch / K)
count_num = 0
for batch in batches:
x_batch, y_batch = zip(*batch)
if count_num == num_batches_per_epoch_:
break
if is_free:
for i in range(K):
train_step(x_batch, y_batch)
else:
train_step(x_batch, y_batch)
current_step = tf.train.global_step(sess, global_step)
# Training loop. For each batch...
# 每50步一次
if current_step % FLAGS.evaluate_every == 0 and current_step > 0:
print("\nEvaluation:")
dev_test()
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))
if is_free:
count_num += 1
def train(w2v_model):
# Training
x_train, x_dev, y_train, y_dev, vocab_size = load_data(w2v_model)
with tf.Graph().as_default(): #返回值：返回一个上下文管理器，这个上下管理器使用这个图作为默认的图
#tf.ConfigProto()配置tf.Session的运算方式，比如gpu运算或者cpu运算
#allow_soft_placement允许动态分配GPU内存
#log_device_placement打印出设备信息
session_conf = tf.ConfigProto(
allow_soft_placement=FLAGS.allow_soft_placement,
log_device_placement=FLAGS.log_device_placement)
sess = tf.Session(config=session_conf)
#分类算法TextCNN，主要思想是将不同长度的短文作为矩阵输入，
#使用多个不同size的filter去提取句子中的关键信息，并用于最终的分类
with sess.as_default():
cnn = TextCNN(w2v_model,
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=vocab_size,
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 确定训练程序
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)
#跟踪渐变值和稀疏度（可选）
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
#模型和摘要的输出目录
timestamp = str(int(time.time()))
out_dir = os.path.abspath(os.path.join(os.path.curdir, "runs"))
print("Writing to {}\n".format(out_dir))
# Summaries for loss and accuracy
#生成准确率和损失率标量图
loss_summary = tf.summary.scalar(
"loss", cnn.loss) #tf.summary.scalar()用来显示标量信息，
acc_summary = tf.summary.scalar("accuracy", cnn.accuracy)
# Train Summaries训练总结
train_summary_op = tf.summary.merge(
[loss_summary, acc_summary, grad_summaries_merged]) #保存信息
train_summary_dir = os.path.join(out_dir, "summaries",
"train") #文件路径
train_summary_writer = tf.summary.FileWriter(
train_summary_dir, sess.graph) #指定一个文件用来保存图。
#下面同理
# Dev summaries开发总结
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
##检查Checkpoint目录。Tensorflow假设这个目录已经存在，所以我们需要创建它
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) #tf.train.Saver()保存和加载模型
# Write vocabulary
# vocab_processor.save(os.path.join(out_dir, "vocab"))
# Initialize all variables初始化所有变量
#含有tf.Variable的环境下，因为tf中建立的变量是没有初始化的，
#也就是在debug时还不是一个tensor量，而是一个Variable变量类型
sess.run(tf.global_variables_initializer())
#训练步骤
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, (w, idx) = sess.run(
# [train_op, global_step, train_summary_op, cnn.loss, cnn.accuracy, cnn.get_w2v_W()],
# feed_dict)
_, 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))
# print w[:2], idx[:2]
train_summary_writer.add_summary(summaries, step)
#开发步骤
def dev_step(x_batch, y_batch, writer=None):
"""
Evaluates model on a dev set
评估开发集上的模型
"""
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生成批处理
batches = data_helpers.batch_iter(list(zip(x_train, y_train)),
FLAGS.batch_size,
FLAGS.num_epochs)
#测试步骤
def dev_test():
batches_dev = data_helpers.batch_iter(list(zip(x_dev, y_dev)),
FLAGS.batch_size, 1)
for batch_dev in batches_dev:
x_batch_dev, y_batch_dev = zip(*batch_dev)
dev_step(x_batch_dev,
y_batch_dev,
writer=dev_summary_writer)
# Training loop. For each batch...每个批次循环训练
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) # #获得global_step
# Training loop. For each batch...
if current_step % FLAGS.evaluate_every == 0:
print("\nEvaluation:")
dev_test()
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))
