def eval():
# load data
eval_size, eval_query, eval_question = eval_data_helpers.load_data(
FLAGS.eval_data_file)
x = word2vec_helpers.SentencesIndex(eval_query, max_document_length)
x = [x[8]] * FLAGS.batch_size
y = word2vec_helpers.SentencesIndex(eval_question, max_document_length)
y = y[0:FLAGS.batch_size]
# eval
print("\nEvaluating...\n")
real_len_x_value = real_len_func(x)
real_len_xpos_value = real_len_func(y)
feed_dict = {
input_x: x,
input_xpos: y,
real_len_x: real_len_x_value,
real_len_xpos: real_len_xpos_value,
dropout_keep_prob: 1.0,
batch_size: FLAGS.batch_size,
}
batch_x_vs_xpos = sess.run([x_vs_xpos], feed_dict)
print(batch_x_vs_xpos)
top1 = np.argmax(batch_x_vs_xpos)
print(top1)
from gensim.models import Word2Vec
import eval_data_helpers
#################### config ###################
modelfile = "../wvmodel/size300window5sg1min_count100negative10iter50.model"
pattern_file = "../data/eval_data.txt"
output_file = "../data/phrase_gen.txt"
############### end of config #################
model = Word2Vec.load(modelfile)
data_size, data, label = eval_data_helpers.load_data(pattern_file)
word_freq = {}
for sentence in data:
for word in sentence.split():
if word in word_freq:
word_freq[word] = word_freq[word] + 1
else:
word_freq[word] = 1
word_mostsim = {}
for word, freq in word_freq.items():
if word in model.wv.vocab:
word_mostsim[word] = model.wv.most_similar(word)
with open(output_file, 'w') as f:
for word, sim_word in word_mostsim.items():
f.write(word)
f.write('\t')
f.write(str([item[0] for item in sim_word]))
def eval():
# Load data
eval_size, x_raw, y_test = eval_data_helpers.load_data(
FLAGS.eval_data_file)
x_test = word2vec_helpers.SentencesIndex(x_raw, max_document_length)
all_predictions = []
all_scores = []
all_softmax = []
# Send predict_request
for x in x_test:
input_x = x.tolist()
# print(input_x)
# print(type(input_x))
dropout_keep_prob = 1.0
predict_request.model_spec.name = "CNN_classifier"
predict_request.model_spec.signature_name = "predict_sentence"
# print(tf.contrib.util.make_tensor_proto([input_x], shape=[1,22], dtype=tf.int32))
predict_request.inputs["input_x"].CopyFrom(
tf.contrib.util.make_tensor_proto([input_x],
shape=[1, 22],
dtype=tf.int32))
predict_request.inputs["dropout_keep_prob"].CopyFrom(
tf.contrib.util.make_tensor_proto(dropout_keep_prob,
shape=[1],
dtype=tf.float32))
result = stub.Predict(predict_request, 10.0) # 10 secs timeout
# print(type(result))
feature_configs = {
"prediction": tf.FixedLenFeature(shape=[], dtype=tf.int64),
"scores": tf.FixedLenFeature(shape=[], dtype=tf.float32),
"softmax": tf.FixedLenFeature(shape=[], dtype=tf.float32),
}
prediction = result.outputs['prediction'].int64_val
scores = np.array(result.outputs['scores'].float_val)
softmax = np.array(result.outputs['softmax'].float_val)
# print(prediction)
# print(scores)
# print(softmax)
all_predictions = np.concatenate([all_predictions, prediction])
all_scores.append(scores)
all_softmax.append(softmax)
#break
# 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))))
for th in np.linspace(0, 0.95, 10):
threshold = th
true_pos = 0
true_neg = 0
false_pos = 0
false_neg = 0
for i in range(len(y_test)):
if all_predictions[i] != 0:
if all_softmax[i][int(all_predictions[i])] > threshold:
if all_predictions[i] == y_test[i]:
true_pos += 1
if all_predictions[i] != y_test[i]:
false_pos += 1
precision = true_pos / (true_pos + false_pos)
print("Precision: {} in {} threshold".format(precision, threshold))
# Misc Parameters
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("")
# Load data
eval_size, eval_query, eval_question = eval_data_helpers.load_data(
FLAGS.eval_data_file)
max_document_length = 50
word2vec_helpers = Word2VecHelper()
x = word2vec_helpers.SentencesIndex(eval_query, max_document_length)
y = word2vec_helpers.SentencesIndex(eval_question, max_document_length)
# Checkpoint
ckpt = tf.train.get_checkpoint_state(
os.path.join(FLAGS.checkpoint_dir, 'checkpoints'))
if ckpt:
print("Read model parameters from %s" % ckpt.model_checkpoint_path)
def visualization(words, attentions, html_file):
i = 0
# Misc Parameters
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("")
# Load data
eval_size, x_raw, y_test = eval_data_helpers.load_data(FLAGS.eval_data_file)
max_document_length = 22
word2vec_helpers = Word2VecHelper()
x_test = word2vec_helpers.SentencesIndex(x_raw, max_document_length)
# Checkpoint
ckpt = tf.train.get_checkpoint_state(
os.path.join(FLAGS.checkpoint_dir, 'checkpoints'))
if ckpt:
print("Read model parameters from %s" % ckpt.model_checkpoint_path)
# Evaluation
# ==================================================
print("\nEvaluating...\n")