def evaluate(model,session,data,global_steps=None,summary_writer=None):
correct_num=0
total_num=len(data[0])
for step, (x,y,mask_x) in enumerate(data_helper.batch_iter(data,batch_size=FLAGS.batch_size)):
fetches = model.correct_num
feed_dict={}
feed_dict[model.input_data]=x
feed_dict[model.target]=y
feed_dict[model.mask_x]=mask_x
model.assign_new_batch_size(session,len(x))
state = session.run(model._initial_state)
for i , (c,h) in enumerate(model._initial_state):
feed_dict[c]=state[i].c
feed_dict[h]=state[i].h
count=session.run(fetches,feed_dict)
correct_num+=count
accuracy=float(correct_num)/total_num
dev_summary = tf.scalar_summary('dev_accuracy',accuracy)
dev_summary = session.run(dev_summary)
if summary_writer:
summary_writer.add_summary(dev_summary,global_steps)
summary_writer.flush()
return accuracy
def run_epoch(model,session,data,global_steps,valid_model,valid_data,train_summary_writer,valid_summary_writer=None):
# for step, (x,y,mask_x) in enumerate(data_helper.batch_iter(data,batch_size=FLAGS.batch_size)):
for step, (x, y, seq_length) in enumerate(data_helper.batch_iter(FLAGS.max_len, data, batch_size=FLAGS.batch_size, usemydata=FLAGS.using_mydata)):
feed_dict={}
feed_dict[model.input_data] = x # 输入数据位置！
# feed_dict[model.seq_length] = seq_length # 提前准备好长度
if FLAGS.using_mydata:
feed_dict[model.seq_length] = seq_length
else:
feed_dict[model.mask_x] = seq_length
feed_dict[model.target] = y
model.assign_new_batch_size(session, len(x)) # 每次更新不同的len
fetches = [model.cost, model.accuracy, model.train_op, model.summary]
state = session.run(model._initial_state)
for i, (c,h) in enumerate(model._initial_state):
feed_dict[c]=state[i].c
feed_dict[h]=state[i].h
cost,accuracy,_,summary = session.run(fetches,feed_dict)
train_summary_writer.add_summary(summary,global_steps)
train_summary_writer.flush()
valid_accuracy=evaluate(valid_model,session,valid_data,global_steps,valid_summary_writer)
if(global_steps%1==0):
print("the %i step, train cost is: %f and the train accuracy is %f and the valid accuracy is %f"%(global_steps,cost,accuracy,valid_accuracy))
global_steps += 1
return global_steps
def run_one_epoch(self, sess, train, dev, tag2label, epoch, saver):
num_batches = (len(train) + self.batch_size - 1) // self.batch_size
start_time = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
batches = batch_iter(train, self.batch_size, shuffle=self.shuffle)
step = -1
for batch in batches:
step+=1
seqs, labels = zip(*batch)
if step % 20 == 0:
print(' processing: {} batch / {} batches.'.format(step + 1, num_batches) + '\r')
step_num = epoch * num_batches + step + 1
feed_dict, _ = self.get_feed_dict(seqs, labels, self.lr, self.dropout_keep_prob) #
_, loss_train, summary, step_num_ = sess.run([self.train_op, self.loss, self.merged, self.global_step],
feed_dict=feed_dict)
if step + 1 == 1 or (step + 1) % 300 == 0 or step + 1 == num_batches:
print('{} epoch {}, step {}, loss: {:.4}, global_step: {}'.format(
start_time, epoch + 1, step + 1,
loss_train, step_num))
self.file_writer.add_summary(summary, step_num)
if step + 1 == num_batches:
saver.save(sess, self.model_path, global_step=step_num)
label_list_dev, seq_len_list_dev = self.dev_one_epoch(sess, dev)
self.evaluate(label_list_dev, seq_len_list_dev, dev, epoch)
def predict(model_file, vocab_processor, params, labels, text_list):
x_test = np.array(list(vocab_processor.transform(text_list)))
batches = data_helper.batch_iter(list(x_test),
params['batch_size'],
1,
shuffle=False)
all_predictions = []
graph = tf.Graph()
with graph.as_default():
session_conf = tf.ConfigProto(allow_soft_placement=True,
log_device_placement=False)
sess = tf.Session(config=session_conf)
with sess.as_default():
saver = tf.train.import_meta_graph("{}.meta".format(model_file))
saver.restore(sess, model_file)
input_x = graph.get_operation_by_name("input_x").outputs[0]
dropout_keep_prob = graph.get_operation_by_name(
"dropout_keep_prob").outputs[0]
predictions = graph.get_operation_by_name(
"output/predictions").outputs[0]
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])
to_return = []
for inc in all_predictions:
to_return.append(label_dict[int(inc)])
return to_return
def train_graph(sess, textRNN, x_train, y_train, batch_size, dropout_prob, trainMode=True):
"""
Train model on training set
"""
loss_list, acc_list = [], []
loss, acc, ct = 0, 0, 0
batches = data_helper.batch_iter(list(zip(x_train, y_train)), batch_size)
for batch in batches:
x_batch, y_batch = zip(*batch)
feed = {
textRNN.x: x_batch,
textRNN.y: y_batch,
textRNN.dropout_keep_prob: dropout_prob
}
if trainMode:
sess.run([textRNN.train_step], feed_dict=feed)
curr_loss, curr_acc = sess.run([textRNN.loss, textRNN.accuracy], feed_dict=feed)
# Contain loss/acc per batch
loss_list.append(curr_loss)
acc_list.append(curr_acc)
# Contain loss/acc per epoch
loss, acc, ct = loss + curr_loss, acc + curr_acc, ct + 1
return loss/float(ct), acc/float(ct), loss_list, acc_list
def run(train_x,
visible_size,
hidden_szie=30,
num_epoch=100,
batch_size=16,
lr=1e-3,
test_rate=0.1):
print("training begin")
rbm = RBM.RBM_Model(visible_size=visible_size,
hidden_size=hidden_szie,
lr=lr)
if os.path.exists(W_path):
rbm.W = load_param(W_path)
rbm_b_v = load_param(b_v_path)
rbm_b_h = load_param(b_h_path)
iter_time = 0
for data in dh.batch_iter(train_x, num_epoch, batch_size):
#rbm.lr = lr*math.pow(10,-1*(iter_time/5000))
iter_time += 1
rbm.train(data, iter_time)
print("saving parameters...")
np.save(W_path, rbm.W)
np.save(b_v_path, rbm.b_v)
np.save(b_h_path, rbm.b_h)
print("save done!")
return rbm
def valid_model(sess, lstm, valid_ori_quests, valid_cand_quests, labels,
results, test_cat_ids):
total_loss, idx = 0, 0
total_ori_cand = []
#total_right, total_wrong, step = 0, 0, 0, 0
for ori_valid, cand_valid, neg_valid, cat_ids_test in batch_iter(
valid_ori_quests,
valid_cand_quests,
test_cat_ids,
FLAGS.batch_size,
1,
isvalid=True):
loss, ori_cand = run_step(sess, ori_valid, cand_valid, cand_valid,
cat_ids_test, lstm, FLAGS.dropout, False,
False)
total_loss += loss
total_ori_cand.extend(ori_cand)
#total_right += right
#total_wrong += wrong
idx += 1
acc, MAP, MRR = cal_acc(labels, results, total_ori_cand)
logger.info("evaluation acc:%s" % (acc))
logger.info("evaluation MAP:%s" % (MAP))
logger.info("evaluation MRR:%s" % (MRR))
return acc, MAP, MRR
def evaluate(model, session, data, global_steps=None, summary_writer=None):
correct_num = 0
total_num = len(data[0])
for step, (x, y, mask_x) in enumerate(
data_helper.batch_iter(data, batch_size=FLAGS.batch_size)):
fetches = model.correct_num
feed_dict = {}
feed_dict[model.input_data] = x
feed_dict[model.target] = y
feed_dict[model.mask_x] = mask_x
model.assign_new_batch_size(session, len(x))
state = session.run(model._initial_state)
for i, (c, h) in enumerate(model._initial_state):
feed_dict[c] = state[i].c
feed_dict[h] = state[i].h
count = session.run(fetches, feed_dict)
correct_num += count
accuracy = float(correct_num) / total_num
dev_summary = tf.summary.scalar('dev_accuracy', accuracy)
dev_summary = session.run(dev_summary)
if summary_writer:
summary_writer.add_summary(dev_summary, global_steps)
summary_writer.flush()
return accuracy
def validate_model(sess, cnn, valid_x, valid_y):
start_time = time.time()
batches = batch_iter(zip(valid_x, valid_y),
FLAGS.batch_size,
shuffle=False)
total_loss, total_acc, total_elapsed_time = 0, 0, 0
idx = 0
pred_labels = list()
act_labels = list()
for batch in batches:
batch_x, batch_y = zip(*batch)
step, cur_loss, cur_acc, predicts, elapsed_time = valid_step(
sess, cnn, batch_x, batch_y, is_optimizer=False)
total_loss += cur_loss
total_acc += cur_acc
total_elapsed_time += elapsed_time
idx += 1
pred_labels.extend(predicts)
act_labels.extend(batch_y)
aver_loss = 1. * total_loss / idx
aver_acc = 1. * total_acc / idx
aver_elapsed_time = 1. * total_elapsed_time / idx
validate_elapsed_time = time.time() - start_time
logger.info(
"validation loss:%s, acc:%s, %6.7f secs/batch_size, total elapsed time: %6.7f"
% (aver_loss, aver_acc, aver_elapsed_time, validate_elapsed_time))
return pred_labels, act_labels
def run_epoch(model, session, data, global_steps, valid_model, valid_data,
train_summary_writer):
for step, batch in enumerate(batch_iter(data,
batch_size=FLAGS.batch_size)):
x, y, mask_x = zip(*batch)
feed_dict = {
model.input_data: x,
model.target: y,
model.mask_x: np.transpose(mask_x)
}
fetches = [model.cost, model.accuracy, model.train_op, model.summary]
cost, accuracy, _, summary = session.run(fetches, feed_dict)
train_summary_writer.add_summary(summary, global_steps)
train_summary_writer.flush()
timestr = datetime.datetime.now().isoformat()
logging.info(
"%s, the %i step, train cost is:%f and the train accuracy is %6.7f"
% (timestr, global_steps, cost, accuracy))
if (global_steps % FLAGS.evaluate_every == 0):
valid_accuracy = evaluate(valid_model, session, valid_data,
global_steps)
logging.info("%s, the valid accuracy is %f" %
(timestr, valid_accuracy))
global_steps += 1
return global_steps
def evaluate(model,session,data,global_steps=None,summary_writer=None):
correct_num=0
if FLAGS.using_mydata:
total_num = len(data[0][0]) # 这里也要修改相应的维数
else:
total_num = len(data[0])
for step, (x, y, seq_length) in enumerate(data_helper.batch_iter(FLAGS.max_len, data,batch_size=FLAGS.batch_size, usemydata=FLAGS.using_mydata)):
fetches = [model.correct_num, model.accuracy]
feed_dict = {}
feed_dict[model.input_data] = x
# feed_dict[model.seq_length] = seq_length # 提前准备好长度
if FLAGS.using_mydata:
feed_dict[model.seq_length] = seq_length
else:
feed_dict[model.mask_x] = seq_length
feed_dict[model.target] = y
model.assign_new_batch_size(session, len(x))
state = session.run(model._initial_state)
for i , (c,h) in enumerate(model._initial_state):
feed_dict[c]=state[i].c
feed_dict[h]=state[i].h
count, acc =session.run(fetches, feed_dict)
correct_num+=count
# print('step:', step, 'count:', count, 'correct_num:', correct_num, 'acc:', acc)
accuracy=float(correct_num)/total_num
dev_summary = tf.summary.scalar('dev_accuracy',accuracy)
dev_summary = session.run(dev_summary)
if summary_writer:
summary_writer.add_summary(dev_summary,global_steps)
summary_writer.flush()
return accuracy
def run():
# training parameters
batch_size = 128
num_epochs = 100
maxlen = 8
step = 1
next_n = 1
# model parameters
num_units = 300
num_rnn_layers = 1
vocab_size = 10000
model = RNN(batch_size, maxlen, num_units, num_rnn_layers, vocab_size)
texts = batch_iter("poetry.txt", batch_size, num_epochs, maxlen, vocab_size, step, next_n)
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
saver = tf.train.Saver(tf.global_variables())
i = 0
for text in texts:
x_batch, y_batch = zip(*text)
# lr = 0.5 * (0.99 ** i)
lr = 0.01
sess.run(tf.assign(model.learning_rate, lr))
train_loss, _ = sess.run([model.loss, model.train_op],
feed_dict={model.xs: x_batch, model.ys: y_batch})
i += 1
if i % 500 == 0:
print("Epoch: %s, loss: %s" % (i, train_loss))
saver.save(sess, 'model/rnn.ckpt', global_step=i)
def run_epoch(model,
session,
data,
global_steps,
valid_model,
valid_data,
train_summary_writer,
valid_summary_writer=None):
for step, (x, y, mask_x) in enumerate(
data_helper.batch_iter(data, batch_size=FLAGS.batch_size)):
feed_dict = {}
feed_dict[model.input_data] = x
feed_dict[model.target] = y
feed_dict[model.mask_x] = mask_x
model.assign_new_batch_size(session, len(x))
fetches = [model.cost, model.accuracy, model.train_op, model.summary]
state = session.run(model._initial_state)
for i, (c, h) in enumerate(model._initial_state):
feed_dict[c] = state[i].c
feed_dict[h] = state[i].h
cost, accuracy, _, summary = session.run(fetches, feed_dict)
train_summary_writer.add_summary(summary, global_steps)
train_summary_writer.flush()
valid_accuracy = evaluate(valid_model, session, valid_data,
global_steps, valid_summary_writer)
if global_steps % 100 == 0:
print(
"the %i step, train cost is: %f and the train accuracy is %f and the valid accuracy is %f"
% (global_steps, cost, accuracy, valid_accuracy))
global_steps += 1
return global_steps
def evaluate(session, test_x, test_y, corpus, global_steps=None):
total_correct_num=0
total_busi_num = 0
total_busi_correct_num = 0
total_other_num = 0
total_other_correct_num = 0
data = zip(test_x, test_y, corpus)
total_num=len(data)
for step, batch in enumerate(batch_iter(data, batch_size=FLAGS.batch_size, shuffle=False)):
x, input_y, batch_corpus = zip(*batch)
input_x, sess_len, sent_len = format_input_x(x)
fetches = [model_prediction, model_correct_num, model_accuracy]
feed_dict={
model.input_data:input_x,
model.target:input_y,
model.session_lengths:sess_len,
model.sent_lengths:sent_len,
model.dropout_ratio:1.0
}
prediction, correct_num, acc = session.run(fetches, feed_dict)
other_correct_num, other_num, busi_correct_num, busi_num = cal_detail_acc(prediction, correct_num, input_y, batch_corpus)
total_correct_num += correct_num
total_busi_num += busi_num
total_busi_correct_num += busi_correct_num
total_other_num += other_num
total_other_correct_num += other_correct_num
accuracy=float(total_correct_num)/total_num
busi_acc = float(total_busi_correct_num) / total_busi_num
other_acc = float(total_other_correct_num) / total_other_num
logger.info("validation success")
return accuracy, busi_acc, other_acc
def train():
x = tf.placeholder(tf.int32, [None, None], name='x')
y = tf.placeholder(tf.int32, [None], name='y')
lr = TextCNN.INIT_LEARNING_RATE
embedding = tf.Variable(embedding_table, dtype=tf.float32, trainable=False)
# embedding = tf.Variable(tf.random_uniform([TextCNN.VOCAB_SIZE, TextCNN.EMBED_FEATURE], -1.0, 1.0))
input = tf.nn.embedding_lookup(embedding, x)
model = TextCNN.TextCNN()
logits_train = model.inference(input, Training=True)
loss = tf.nn.sparse_softmax_cross_entropy_with_logits(logits=logits_train,
labels=y,
name='loss')
loss_ = tf.reduce_mean(loss) + tf.nn.l2_loss(model.fc.get_weights()[0])
train_op = tf.train.AdamOptimizer(lr).minimize(loss_)
logits = model.inference(input)
correct_pred = tf.equal(tf.argmax(logits, axis=1), tf.cast(y, tf.int64))
accuracy = tf.reduce_mean(tf.cast(correct_pred, dtype=tf.float32))
sum_correct_pred = tf.reduce_sum(tf.cast(correct_pred, dtype=tf.float32))
with tf.Session() as sess:
init_op = tf.global_variables_initializer()
sess.run(init_op)
for epoch in range(TextCNN.EPOCH):
for step, (x_, y_) in enumerate(
data_helper.batch_iter(x_train, y_train,
TextCNN.BATCH_SIZE)):
# print(sess.run(input , feed_dict={x:x_}))
_ = sess.run(train_op, feed_dict={x: x_, y: y_})
if step % 64 == 0:
# print(y_)
# print(sess.run(tf.argmax(logits, axis=1), feed_dict={x: x_}))
print('epoch :', epoch, 'step :', step, ' train_acc = ',
sess.run(accuracy, feed_dict={
x: x_,
y: y_
}))
sum_ = 0
for (x__, y__) in data_helper.batch_iter(x_test, y_test,
TextCNN.BATCH_SIZE):
tmp = sess.run(sum_correct_pred, feed_dict={x: x__, y: y__})
sum_ += tmp
print('epoch ', epoch, 'acc = ', sum_ / len(y_test))
if epoch % 30 == 0:
lr /= 2
def test(test_x, rbm, batch_size=16, topK=30):
test_recomendation = []
test_s = []
for test_data in dh.batch_iter(test_x, num_epoch=1, batch_size=batch_size):
v1_state = rbm.recomendation(test_data, topK)
test_s.extend(test_data)
test_recomendation.extend(v1_state)
return test_s, test_recomendation
def predict_unseen_data():
"""Step 0: load trained model and parameters"""
params = json.loads(open('./parameters.json').read())
checkpoint_dir = sys.argv[1]
if not checkpoint_dir.endswith('/'):
checkpoint_dir += '/'
checkpoint_file = tf.train.latest_checkpoint(checkpoint_dir + 'checkpoints')
logging.critical('Loaded the trained model: {}'.format(checkpoint_file))
"""Step 1: load data for prediction"""
test_file = sys.argv[2]
x_test, y_test, revs, labels = data_helper.load_data_and_labels(test_file)
logging.info('The number of x_test: {}'.format(len(x_test)))
logging.info('The number of y_test: {}'.format(len(y_test)))
vocab_path = os.path.join(checkpoint_dir, "vocab.pickle")
vocab_processor = learn.preprocessing.VocabularyProcessor.restore(vocab_path)
x_test = np.array(list(vocab_processor.transform(x_test)))
"""Step 2: compute the predictions"""
graph = tf.Graph()
with graph.as_default():
#配置并初始化
session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
sess = tf.Session(config=session_conf)
with sess.as_default():
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]
# Generate batches for one epoch
batches = data_helper.batch_iter(list(x_test), params['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:
y_test = np.argmax(y_test, axis=1)
correct_predictions = sum(all_predictions == y_test)
logging.critical('The accuracy is: {}'.format(correct_predictions / float(len(y_test))))
# Save the evaluation to a csv
predictions_human_readable = np.column_stack((np.array(x_raw), all_predictions))
out_path = os.path.join(checkpoint_dir, "..", "prediction.csv")
logging.critical("Saving evaluation to {0}".format(out_path))
with open(out_path, 'w') as f:
csv.writer(f).writerows(predictions_human_readable)
def predict_new_data():
"""Step 0: load trained model and parameters"""
params = json.loads(open('./parameters.json').read())
checkpoint_dir = sys.argv[1]
if not checkpoint_dir.endswith('/'):
checkpoint_dir += '/'
checkpoint_file = tf.train.latest_checkpoint(checkpoint_dir + 'checkpoints')
logging.critical('Loaded the trained model: {}'.format(checkpoint_file))
"""Step 1: load data for prediction"""
test_file = sys.argv[2]
test_examples = json.loads(open(test_file).read())
# labels.json was saved during training, and it has to be loaded during prediction
labels = json.loads(open('./labels.json').read())
one_hot = np.zeros((len(labels), len(labels)), int)
np.fill_diagonal(one_hot, 1)
label_dict = dict(zip(labels, one_hot))
x_raw = [example['consumer_complaint_narrative'] for example in test_examples]
x_test = [data_helper.clean_str(x) for x in x_raw]
logging.info('The number of x_test: {}'.format(len(x_test)))
y_test = None
if 'product' in test_examples[0]:
y_raw = [example['product'] for example in test_examples]
y_test = [label_dict[y] for y in y_raw]
logging.info('The number of y_test: {}'.format(len(y_test)))
vocab_path = os.path.join(checkpoint_dir, "vocab.pickle")
vocab_processor = learn.preprocessing.VocabularyProcessor.restore(vocab_path)
x_test = np.array(list(vocab_processor.transform(x_test)))
"""Step 2: compute the predictions"""
graph = tf.Graph()
with graph.as_default():
session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
sess = tf.Session(config=session_conf)
with sess.as_default():
saver = tf.train.import_meta_graph("{}.meta".format(checkpoint_file))
saver.restore(sess, checkpoint_file)
input_x = graph.get_operation_by_name("input_x").outputs[0]
dropout_keep_prob = graph.get_operation_by_name("dropout_keep_prob").outputs[0]
predictions = graph.get_operation_by_name("output/predictions").outputs[0]
batches = data_helper.batch_iter(list(x_test), params['batch_size'], 1, shuffle=False)
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])
if y_test is not None:
y_test = np.argmax(y_test, axis=1)
correct_predictions = sum(all_predictions == y_test)
logging.critical('The accuracy is: {}'.format(correct_predictions / float(len(y_test))))
def train():
x_train, x_test, y_train, y_test = data_process()
print('x_train`s shape:',x_train.shape)
with tf.device('/gpu:0'):
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():
cnn = TextCNN(
sequence_length = x_train.shape[1],
num_classes = y_train.shape[1],
filter_sizes = filter_sizes,
num_filters = num_filters,
l2_reg_lambda = l2_reg_lambda,
embedding_size=embedding_dim)
global_step = 0
train_op = tf.train.AdamOptimizer(learn_rate).minimize(cnn.loss)
timestamp = str(int(time.time()))
out_dir = os.path.abspath(os.path.join(os.path.curdir,timestamp))
print('Writing to {}'.format(out_dir))
checkpoint_dir = os.path.abspath(os.path.join(out_dir,'checkpoints'))
if not os.path.exists(checkpoint_dir):
os.makedirs(checkpoint_dir)
saver = tf.train.Saver()
# initialize all variables
# 全局初始化
sess.run(tf.global_variables_initializer())
batches = data_helper.batch_iter(
list(zip(x_train,y_train)),
batch_size,
num_epochs
)
best_acc = 0
for batch in batches:
global_step += 1
x_batch, y_batch = zip(*batch)
train_step(sess,cnn,x_batch,y_batch,train_op,global_step)
if global_step % evaluate_every == 0:
print('\n Evaluation:')
accurent_acc = dev_step(sess,cnn,x_test,y_test,global_step)
print('')
if global_step % checkpoint_every == 0 and accurent_acc > best_acc:
path = saver.save(sess,checkpoint_dir + '/save_net.ckpt')
best_acc = accurent_acc
print('Saved model checkpoint to {}\n'.format(path))
def valid_model(sess, lstm, valid_ori_quests, valid_cand_quests, labels, results):
logger.info("start to validate model")
total_ori_cand = []
for ori_valid, cand_valid, neg_valid in batch_iter(valid_ori_quests, valid_cand_quests, FLAGS.batch_size, 1, is_valid=True):
ori_cand = valid_run_step(sess, ori_valid, cand_valid, lstm)
total_ori_cand.extend(ori_cand)
data_len = len(total_ori_cand)
acc = cal_acc(labels[:data_len], results[:data_len], total_ori_cand)
timestr = datetime.datetime.now().isoformat()
logger.info("%s, evaluation acc:%s"%(timestr, acc))
def valid_model_train_format(sess, cnn, valid_ori_quests, valid_cand_quests,
valid_neg_quests, test_cat_ids):
for ori_valid, cand_valid, neg_valid, cat_ids_test in batch_iter(
valid_ori_quests,
valid_cand_quests,
test_cat_ids,
FLAGS.batch_size,
1,
neg_quests=valid_neg_quests):
run_step(sess, ori_valid, cand_valid, neg_valid, cat_ids_test, cnn,
FLAGS.dropout, False)
def train(self, dropout, check_step, save_step, batch_size, epoch_num,
model_name, train_word, train_vds, train_reg, train_y, dev_word,
dev_vds, dev_reg, dev_y, test_word, test_vds, test_reg, test_y,
id2label):
root_path = './save/%d_%d_%d/' % (self.word, self.vds, time.time())
print 'ROOT_PATH: %s' % root_path
os.mkdir(root_path)
curr_step = 0
batches = dh.batch_iter(
list(zip(train_word, train_vds, train_reg, train_y)), batch_size,
epoch_num, True)
dev_feed_dict = {
self.x_word: dev_word,
self.x_vds: dev_vds,
self.x_reg: dev_reg,
self.y: dev_y,
self.dropout_keep: 1.0
}
sess = tf.InteractiveSession()
sess.run(self.init)
max_devacc = 0
step_max_devacc = 0
# Training
for batch in batches:
if len(batch) == 0:
continue
word_batch, vds_batch, reg_batch, y_batch = zip(*batch)
train_feed_dict = {
self.x_word: word_batch,
self.x_vds: vds_batch,
self.x_reg: reg_batch,
self.y: y_batch,
self.dropout_keep: dropout
}
self.train_step.run(feed_dict=train_feed_dict)
# print 'Step %d, %s' % (curr_step, time.time())
curr_step += 1
if curr_step % check_step == 0:
dev_acc = self.accuracy.eval(dev_feed_dict)
train_acc = self.accuracy.eval(train_feed_dict)
print 'Step %d, Train: %.03f' % (curr_step, train_acc)
print ' Dev Accuracy: %.03f' % dev_acc
if curr_step % save_step == 0:
save_model_path = os.path.join(
root_path,
"model_%d_devacc_%.3f" % (curr_step, dev_acc))
saver = tf.train.Saver(tf.global_variables())
saver.save(sess, save_model_path)
if dev_acc > max_devacc:
step_max_devacc = curr_step
max_devacc = dev_acc
return max_devacc, step_max_devacc, root_path
def dev_one_epoch(self, sess, dev):
"""
:param sess:
:param dev:
:return:
"""
label_list, seq_len_list = [], []
for batch in batch_iter(dev, self.batch_size, shuffle=False):
seqs, labels = zip(*batch)
label_list_, seq_len_list_ = self.predict_one_batch(sess, seqs)
label_list.extend(label_list_)
seq_len_list.extend(seq_len_list_)
return label_list, seq_len_list
def evaluate(sess, x_, y_):
"""评估在某一数据上的准确率和损失"""
data_len = len(x_)
batch_eval = batch_iter(x_, y_, 128)
total_loss = 0.0
total_acc = 0.0
for x_batch, y_batch in batch_eval:
batch_len = len(x_batch)
feed_dict = feed_data(x_batch, y_batch, 1.0)
loss, acc = sess.run([model.loss, model.acc], feed_dict=feed_dict)
total_loss += loss * batch_len
total_acc += acc * batch_len
return total_loss / data_len, total_acc / data_len
def predict_unseen_data(userMessage):
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
#logging.getLogger().setLevel(logging.INFO)
"""Step 0: load trained model and parameters"""
mainDir = "C:/Users/sid/Desktop/Darwin/DarwinBot/process_message/intentClassifier/"
params = json.loads(open(mainDir+'parameters.json').read())
#checkpoint_dir = mainDir+"trained_model_1519274258/"
checkpoint_dir = mainDir+"trained_model_1522994422"
if not checkpoint_dir.endswith('/'):
checkpoint_dir += '/'
checkpoint_file = tf.train.latest_checkpoint(checkpoint_dir + 'checkpoints')
#logging.critical('Loaded the trained model: {}'.format(checkpoint_file))
"""Step 1: load data for prediction"""
# labels.json was saved during training, and it has to be loaded during prediction
labels = json.loads(open(mainDir+'labels.json').read())
one_hot = np.zeros((len(labels), len(labels)), int)
np.fill_diagonal(one_hot, 1)
label_dict = dict(zip(labels, one_hot))
x_raw = userMessage
x_test = [data_helper.clean_str(x_raw)]
#logging.info('The number of x_test: {}'.format(len(x_test)))
vocab_path = os.path.join(checkpoint_dir, "vocab.pickle")
vocab_processor = learn.preprocessing.VocabularyProcessor.restore(vocab_path)
x_test = np.array(list(vocab_processor.transform(x_test)))
"""Step 2: compute the predictions"""
graph = tf.Graph()
with graph.as_default():
session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
sess = tf.Session(config=session_conf)
with sess.as_default():
saver = tf.train.import_meta_graph("{}.meta".format(checkpoint_file))
saver.restore(sess, checkpoint_file)
input_x = graph.get_operation_by_name("input_x").outputs[0]
dropout_keep_prob = graph.get_operation_by_name("dropout_keep_prob").outputs[0]
predictions = graph.get_operation_by_name("output/predictions").outputs[0]
batches = data_helper.batch_iter(list(x_test), params['batch_size'], 1, shuffle=False)
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])
all_predictions = all_predictions.tolist()
return labels[int(all_predictions[0])]
def valid_model(sess, cnn, valid_ori_quests, valid_cand_quests, labels, results):
total_loss, idx = 0, 0
total_ori_cand = []
#total_right, total_wrong, step = 0, 0, 0, 0
for ori_valid, cand_valid, neg_valid in batch_iter(valid_ori_quests, valid_cand_quests, FLAGS.batch_size, 1, is_valid=True):
loss, ori_cand = run_step(sess, ori_valid, cand_valid, cand_valid, cnn, FLAGS.dropout, False)
total_loss += loss
total_ori_cand.extend(ori_cand)
#total_right += right
#total_wrong += wrong
idx += 1
acc = cal_acc(labels, results, total_ori_cand)
timestr = datetime.datetime.now().isoformat()
logging.info("%s, evaluation loss:%s, acc:%s"%(timestr, total_loss/idx, acc))
def evaluate(rnn, sess, x, y):
"""在其他数据集上评估模型的准确率"""
data_len = len(x)
total_loss = 0.0
total_acc = 0.0
for x_batch, y_batch in batch_iter(x, y):
batch_len = len(x_batch)
feed_dict = {
rnn.input_x: x_batch,
rnn.input_y: y_batch,
rnn.keep_prob: 1.0
}
loss, acc = sess.run([rnn.loss, rnn.acc], feed_dict)
total_loss += loss * batch_len
total_acc += acc * batch_len
return total_loss / data_len, total_acc / data_len
def prediction(x, y):
batches = batch_iter(x, y, BATCH_SIZE, 1)
outputs = []
predictions = []
logits = []
for batch_x, batch_y in batches:
logit, prediction = sess.run([model.logits, model.predictions],
feed_dict={
model.x: batch_x,
model.y: batch_y,
model.keep_prob: 1.0
})
logits.extend(logit)
predictions.extend(prediction.tolist())
outputs.extend(batch_y.tolist())
return logits, predictions, outputs
def evaluate(model, session, data, global_steps=None, summary_writer=None):
correct_num = 0
total_num = len(data[0])
# state = session.run(model.initial_state)
state_fw = session.run(model.initial_state_fw)
state_bw = session.run(model.initial_state_bw)
for step, (x, y, mask_x) in enumerate(
data_helper.batch_iter(data, batch_size=FLAGS.batch_size)):
# fetches = [model.correct_num, model.final_state]
fetches = [
model.correct_num, model.final_state_fw, model.final_state_bw
]
feed_dict = {}
feed_dict[model.input_data] = x
feed_dict[model.targets] = y
feed_dict[model.mask_x] = mask_x
# if step == 0:
# model.assign_new_batch_size(session, len(x))
# state = session.run(model.initial_state)
# for i, (c, h) in enumerate(model.initial_state):
# feed_dict[c] = state[i].c
# feed_dict[h] = state[i].h
for i, (c, h) in enumerate(model.initial_state_fw):
feed_dict[c] = state_fw[i].c
feed_dict[h] = state_fw[i].h
for i, (c, h) in enumerate(model.initial_state_bw):
feed_dict[c] = state_bw[i].c
feed_dict[h] = state_bw[i].h
# count, state = session.run(fetches, feed_dict)
count, state_fw, state_bw = session.run(fetches, feed_dict)
correct_num += count
# print("-----------------------------------------")
# print(correct_num)
# print(total_num)
accuracy = float(correct_num) / total_num
dev_summary = tf.summary.scalar('dev_accuracy', accuracy)
dev_summary = session.run(dev_summary)
if summary_writer:
summary_writer.add_summary(dev_summary, global_steps)
summary_writer.flush()
return accuracy
def only_test(test_x, test_y, kernal_initilizer):
config = tf.ConfigProto(gpu_options=tf.GPUOptions(
per_process_gpu_memory_fraction=0.5))
config.gpu_options.allow_growth = True
with tf.Session(config=config) as sess:
# with tf.Session() as sess:
tf.set_random_seed(seed)
BATCH_SIZE = args.batch_size
model = biLSTM(max_input_length=args.max_input_len,
num_class=len(args.hidden_ratio),
input_dim=args.input_dim,
hidden_layer_num=args.hidden_layers,
bi_direction=args.bi_directional,
use_attention=args.use_attention,
attention_size=args.attention_size,
num_hidden=args.num_hidden,
fc_num_hidden=args.fc_num_hidden,
hidden_layer_num_bi=args.hidden_layers_bi,
num_hidden_bi=args.num_hidden_bi)
# Define training procedure
global_step = tf.Variable(0, trainable=False)
# Initialize all variables
sess.run(tf.global_variables_initializer())
# Load variables from pre-trained model
if not args.pre_trained == "none":
pre_trained_variables = [
v for v in tf.global_variables() if "Adam" not in v.name
]
saver_pre = tf.train.Saver(pre_trained_variables)
ckpt = tf.train.get_checkpoint_state(args.summary_dir)
saver_pre.restore(sess, ckpt.model_checkpoint_path)
batches = batch_iter(test_x, test_y, BATCH_SIZE, 1)
outputs = []
predictions = []
for batch_x, batch_y in batches:
prediction, = sess.run([model.predictions],
feed_dict={
model.x: batch_x,
model.y: batch_y,
model.keep_prob: 1.0
})
predictions.extend(prediction.tolist())
outputs.extend(batch_y.tolist())
accuracy = sum(np.equal(predictions, outputs)) / len(outputs)
print("test accuracy: %f" % accuracy)
def evaluate(model, session, data, global_steps=None):
correct_num=0
total_num=len(data)
for step, batch in enumerate(batch_iter(data, batch_size=FLAGS.batch_size)):
x, y, mask_x = zip(*batch)
fetches = model.correct_num
feed_dict={
model.input_data:x,
model.target:y,
model.mask_x:np.transpose(mask_x)
}
count=session.run(fetches, feed_dict)
correct_num += count
accuracy=float(correct_num)/total_num
return accuracy
def run_epoch(model,session,data,global_steps,valid_model,valid_data,train_summary_writer,valid_summary_writer=None):
for step, (x,y,mask_x) in enumerate(data_helper.batch_iter(data,batch_size=FLAGS.batch_size)):
feed_dict={}
feed_dict[model.input_data]=x
feed_dict[model.target]=y
feed_dict[model.mask_x]=mask_x
model.assign_new_batch_size(session,len(x))
fetches = [model.cost,model.accuracy,model.train_op,model.summary]
state = session.run(model._initial_state)
for i , (c,h) in enumerate(model._initial_state):
feed_dict[c]=state[i].c
feed_dict[h]=state[i].h
cost,accuracy,_,summary = session.run(fetches,feed_dict)
train_summary_writer.add_summary(summary,global_steps)
train_summary_writer.flush()
valid_accuracy=evaluate(valid_model,session,valid_data,global_steps,valid_summary_writer)
if(global_steps%100==0):
print("the %i step, train cost is: %f and the train accuracy is %f and the valid accuracy is %f"%(global_steps,cost,accuracy,valid_accuracy))
global_steps+=1
return global_steps
def run_epoch(model,session,data,global_steps,valid_model,valid_data,train_summary_writer):
for step, batch in enumerate(batch_iter(data, batch_size = FLAGS.batch_size)):
x, y, mask_x = zip(*batch)
feed_dict={
model.input_data:x,
model.target:y,
model.mask_x:np.transpose(mask_x)
}
fetches = [model.cost, model.accuracy, model.train_op, model.summary]
cost, accuracy, _, summary = session.run(fetches, feed_dict)
train_summary_writer.add_summary(summary,global_steps)
train_summary_writer.flush()
timestr = datetime.datetime.now().isoformat()
logging.info("%s, the %i step, train cost is:%f and the train accuracy is %6.7f"%(timestr, global_steps, cost, accuracy))
if(global_steps % FLAGS.evaluate_every == 0):
valid_accuracy = evaluate(valid_model,session,valid_data,global_steps)
logging.info("%s, the valid accuracy is %f"%(timestr, valid_accuracy))
global_steps += 1
return global_steps
def predict_unseen_data():
"""Step 0: load trained model and parameters"""
params = json.loads(open('./parameters.json').read())
checkpoint_dir = sys.argv[1]
if not checkpoint_dir.endswith('/'):
checkpoint_dir += '/'
checkpoint_file = tf.train.latest_checkpoint(checkpoint_dir + 'checkpoints')
logging.critical('Loaded the trained model: {}'.format(checkpoint_file))
"""Step 1: load data for prediction"""
test_file = sys.argv[2]
test_examples = json.loads(open(test_file).read())
# labels.json was saved during training, and it has to be loaded during prediction
labels = json.loads(open('./labels.json').read())
one_hot = np.zeros((len(labels), len(labels)), int)
np.fill_diagonal(one_hot, 1)
label_dict = dict(zip(labels, one_hot))
x_raw = [example['consumer_complaint_narrative'] for example in test_examples]
x_test = [data_helper.clean_str(x) for x in x_raw]
logging.info('The number of x_test: {}'.format(len(x_test)))
y_test = None
if 'product' in test_examples[0]:
y_raw = [example['product'] for example in test_examples]
y_test = [label_dict[y] for y in y_raw]
logging.info('The number of y_test: {}'.format(len(y_test)))
vocab_path = os.path.join(checkpoint_dir, "vocab.pickle")
vocab_processor = learn.preprocessing.VocabularyProcessor.restore(vocab_path)
x_test = np.array(list(vocab_processor.transform(x_test)))
"""Step 2: compute the predictions"""
graph = tf.Graph()
with graph.as_default():
session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
sess = tf.Session(config=session_conf)
with sess.as_default():
saver = tf.train.import_meta_graph("{}.meta".format(checkpoint_file))
saver.restore(sess, checkpoint_file)
input_x = graph.get_operation_by_name("input_x").outputs[0]
dropout_keep_prob = graph.get_operation_by_name("dropout_keep_prob").outputs[0]
predictions = graph.get_operation_by_name("output/predictions").outputs[0]
batches = data_helper.batch_iter(list(x_test), params['batch_size'], 1, shuffle=False)
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])
if y_test is not None:
y_test = np.argmax(y_test, axis=1)
correct_predictions = sum(all_predictions == y_test)
# Save the actual labels back to file
actual_labels = [labels[int(prediction)] for prediction in all_predictions]
for idx, example in enumerate(test_examples):
example['new_prediction'] = actual_labels[idx]
with open('./data/small_samples_prediction.json', 'w') as outfile:
json.dump(test_examples, outfile, indent=4)
logging.critical('The accuracy is: {}'.format(correct_predictions / float(len(y_test))))
logging.critical('The prediction is complete')
correct_pred = tf.equal(tf.argmax(pred,1), tf.argmax(y,1)) accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32)) print "print " # Initializing the variables init = tf.initialize_all_variables() print "Launch the graph" print len(x_inp) ma = len(x_input) # Launch the graph batches = data_helper.batch_iter(list(zip(x_input, y_input)), 64, 200)
def train_cnn():
"""Step 0: load sentences, labels, and training parameters"""
train_file = sys.argv[1]
x_raw, y_raw, df, labels = data_helper.load_data_and_labels(train_file)
parameter_file = sys.argv[2]
params = json.loads(open(parameter_file).read())
"""Step 1: pad each sentence to the same length and map each word to an id"""
max_document_length = max([len(x.split(' ')) for x in x_raw])
logging.info('The maximum length of all sentences: {}'.format(max_document_length))
vocab_processor = learn.preprocessing.VocabularyProcessor(max_document_length)
x = np.array(list(vocab_processor.fit_transform(x_raw)))
y = np.array(y_raw)
"""Step 2: split the original dataset into train and test sets"""
x_, x_test, y_, y_test = train_test_split(x, y, test_size=0.1, random_state=42)
"""Step 3: shuffle the train set and split the train set into train and dev sets"""
shuffle_indices = np.random.permutation(np.arange(len(y_)))
x_shuffled = x_[shuffle_indices]
y_shuffled = y_[shuffle_indices]
x_train, x_dev, y_train, y_dev = train_test_split(x_shuffled, y_shuffled, test_size=0.1)
"""Step 4: save the labels into labels.json since predict.py needs it"""
with open('./labels.json', 'w') as outfile:
json.dump(labels, outfile, indent=4)
logging.info('x_train: {}, x_dev: {}, x_test: {}'.format(len(x_train), len(x_dev), len(x_test)))
logging.info('y_train: {}, y_dev: {}, y_test: {}'.format(len(y_train), len(y_dev), len(y_test)))
"""Step 5: build a graph and cnn object"""
graph = tf.Graph()
with graph.as_default():
session_conf = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False)
sess = tf.Session(config=session_conf)
with sess.as_default():
cnn = TextCNN(
sequence_length=x_train.shape[1],
num_classes=y_train.shape[1],
vocab_size=len(vocab_processor.vocabulary_),
embedding_size=params['embedding_dim'],
filter_sizes=list(map(int, params['filter_sizes'].split(","))),
num_filters=params['num_filters'],
l2_reg_lambda=params['l2_reg_lambda'])
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)
timestamp = str(int(time.time()))
out_dir = os.path.abspath(os.path.join(os.path.curdir, "trained_model_" + timestamp))
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()
# One training step: train the model with one batch
def train_step(x_batch, y_batch):
feed_dict = {
cnn.input_x: x_batch,
cnn.input_y: y_batch,
cnn.dropout_keep_prob: params['dropout_keep_prob']}
_, step, loss, acc = sess.run([train_op, global_step, cnn.loss, cnn.accuracy], feed_dict)
# One evaluation step: evaluate the model with one batch
def dev_step(x_batch, y_batch):
feed_dict = {cnn.input_x: x_batch, cnn.input_y: y_batch, cnn.dropout_keep_prob: 1.0}
step, loss, acc, num_correct = sess.run([global_step, cnn.loss, cnn.accuracy, cnn.num_correct], feed_dict)
return num_correct
# Save the word_to_id map since predict.py needs it
vocab_processor.save(os.path.join(out_dir, "vocab.pickle"))
sess.run(tf.global_variables_initializer())
# Training starts here
train_batches = data_helper.batch_iter(list(zip(x_train, y_train)), params['batch_size'], params['num_epochs'])
best_accuracy, best_at_step = 0, 0
"""Step 6: train the cnn model with x_train and y_train (batch by batch)"""
for train_batch in train_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)
"""Step 6.1: evaluate the model with x_dev and y_dev (batch by batch)"""
if current_step % params['evaluate_every'] == 0:
dev_batches = data_helper.batch_iter(list(zip(x_dev, y_dev)), params['batch_size'], 1)
total_dev_correct = 0
for dev_batch in dev_batches:
x_dev_batch, y_dev_batch = zip(*dev_batch)
num_dev_correct = dev_step(x_dev_batch, y_dev_batch)
total_dev_correct += num_dev_correct
dev_accuracy = float(total_dev_correct) / len(y_dev)
logging.critical('Accuracy on dev set: {}'.format(dev_accuracy))
"""Step 6.2: save the model if it is the best based on accuracy on dev set"""
if dev_accuracy >= best_accuracy:
best_accuracy, best_at_step = dev_accuracy, current_step
path = saver.save(sess, checkpoint_prefix, global_step=current_step)
logging.critical('Saved model at {} at step {}'.format(path, best_at_step))
logging.critical('Best accuracy is {} at step {}'.format(best_accuracy, best_at_step))
"""Step 7: predict x_test (batch by batch)"""
test_batches = data_helper.batch_iter(list(zip(x_test, y_test)), params['batch_size'], 1)
total_test_correct = 0
for test_batch in test_batches:
x_test_batch, y_test_batch = zip(*test_batch)
num_test_correct = dev_step(x_test_batch, y_test_batch)
total_test_correct += num_test_correct
test_accuracy = float(total_test_correct) / len(y_test)
logging.critical('Accuracy on test set is {} based on the best model {}'.format(test_accuracy, path))
logging.critical('The training is complete')
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]
# Generate batches for one epoch
batches = data_helper.batch_iter(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])
# Write the predictions to a cvs file
data_helper.create_submission_file(all_predictions, FLAGS.submission_filename)
Max["NN-P"] = prec
Max["NN-R"] = rec
Max["NN-F"] = f1
Max["P"] = precision
Max["R"] = recall
Max["F"] = f1_score
print("Max result")
for key, value in Max.items():
print(key, value)
print()
if writer:
writer.add_summary(summaries, step)
return pred
# Generate batches
batches = data_helper.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:")
pred = dev_step(X_dev, y_dev,
writer=dev_summary_writer)
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))
print("Max result")
print(domain)
for key, value in Max.items():