def runModel(sess, keep_prob_pl, predict, features_pl, bow_vectorizer, tfreq_vectorizer, tfidf_vectorizer):
start_time = time.time()
print("Now running predictions...")
# THIS is the info from Henry
userClaims = "ml/claims2.csv"
userBodies = "ml/bodies.csv"
# parse that info
raw_test = util.FNCData(userClaims, userBodies)
# need more stuff for this
test_set = util.pipeline_test(raw_test, bow_vectorizer, tfreq_vectorizer, tfidf_vectorizer)
# idk what this does really
test_feed_dict = {features_pl: test_set, keep_prob_pl: 1.0}
# run predictions
test_pred = sess.run(predict, feed_dict=test_feed_dict)
# timing
print("generate test_set--- %s seconds ---" % (time.time() - start_time))
print("Preditions complete.")
return test_pred
# Initialise hyperparameters r = random.Random() lim_unigram = 5000 target_size = 4 hidden_size = 100 train_keep_prob = 0.6 l2_alpha = 0.00001 learn_rate = 0.01 clip_ratio = 5 batch_size_train = 500 epochs = 90 # Load data sets raw_train = util.FNCData(file_train_instances, file_train_bodies) raw_test = util.FNCData(file_test_instances, file_test_bodies) # n_train = len(raw_train.instances) # TODO OH DUDE JUST LET THIS THING DO IT'S SHIT IN THE INITILIZATION!!! Use the test and train sets provided then just use the vectors created! # Process data sets - THIS TAKES 17 SECONDS! train_set, train_stances, bow_vectorizer, tfreq_vectorizer, tfidf_vectorizer = util.pipeline_train(raw_train, raw_test, lim_unigram=lim_unigram) # feature_size = len(train_set[0]) # fix feature_size at 10001 feature_size = 10001 # Define model # Create placeholders
def loadML():
# Set file names
file_train_instances = "ml/train_stances.csv"
file_train_bodies = "ml/train_bodies.csv"
file_test_instances = "ml/test_stances_unlabeled.csv"
file_test_bodies = "ml/test_bodies.csv"
file_predictions = 'ml/ML_predictions.csv'
# Initialise hyperparameters
r = random.Random()
lim_unigram = 5000
target_size = 4
hidden_size = 100
train_keep_prob = 0.6
l2_alpha = 0.00001
learn_rate = 0.01
clip_ratio = 5
batch_size_train = 500
epochs = 90
# Load data sets
raw_train = util.FNCData(file_train_instances, file_train_bodies)
raw_test = util.FNCData(file_test_instances, file_test_bodies)
# n_train = len(raw_train.instances)
# TODO OH DUDE JUST LET THIS THING DO IT'S SHIT IN THE INITILIZATION!!! Use the test and train sets provided then just use the vectors created!
# Process data sets - THIS TAKES 17 SECONDS!
train_set, train_stances, bow_vectorizer, tfreq_vectorizer, tfidf_vectorizer = util.pipeline_train(
raw_train, raw_test, lim_unigram=lim_unigram)
# feature_size = len(train_set[0])
# fix feature_size at 10001
feature_size = 10001
# Define model
# Create placeholders
features_pl = tf.placeholder(tf.float32, [None, feature_size], 'features')
stances_pl = tf.placeholder(tf.int64, [None], 'stances')
keep_prob_pl = tf.placeholder(tf.float32)
# Infer batch size
batch_size = tf.shape(features_pl)[0]
# Define multi-layer perceptron
hidden_layer = tf.nn.dropout(tf.nn.relu(
tf.contrib.layers.linear(features_pl, hidden_size)),
keep_prob=keep_prob_pl)
logits_flat = tf.nn.dropout(tf.contrib.layers.linear(
hidden_layer, target_size),
keep_prob=keep_prob_pl)
logits = tf.reshape(logits_flat, [batch_size, target_size])
# Define L2 loss
tf_vars = tf.trainable_variables()
l2_loss = tf.add_n(
[tf.nn.l2_loss(v) for v in tf_vars if 'bias' not in v.name]) * l2_alpha
# Define overall loss
loss = tf.reduce_sum(
tf.nn.sparse_softmax_cross_entropy_with_logits(logits, stances_pl) +
l2_loss)
# Define prediction
softmaxed_logits = tf.nn.softmax(logits)
predict = tf.arg_max(softmaxed_logits, 1)
sess = tf.Session()
util.load_model(sess)
return sess, keep_prob_pl, predict, features_pl, bow_vectorizer, tfreq_vectorizer, tfidf_vectorizer