def twitter_bot():
# Only allocate part of the gpu memory when predicting.
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.2)
tf_config = tf.ConfigProto(gpu_options=gpu_options)
with tf.Session(config=tf_config) as sess:
train.show_progress("Creating model...")
model = train.create_or_restore_model(sess,
train.buckets,
forward_only=True)
model.batch_size = 1
train.show_progress("done\n")
enc_vocab, _ = data_processer.initialize_vocabulary(
config.VOCAB_ENC_TXT)
_, rev_dec_vocab = data_processer.initialize_vocabulary(
config.VOCAB_DEC_TXT)
auth = tweepy.OAuthHandler(consumer_key, consumer_secret)
auth.set_access_token(access_token, access_token_secret)
api = tweepy.API(auth)
while True:
try:
stream = tweepy.Stream(auth=api.auth,
listener=StreamListener(
api, sess, model, enc_vocab,
rev_dec_vocab))
stream.userstream()
except Exception as e:
print(e.__doc__)
print(e.message)
def __init__(self, session):
self.session = session
train.show_progress("Creating model...")
self.model = train.create_or_restore_model(self.session, config.buckets, forward_only=True, beam_search=config.beam_search, beam_size=config.beam_size)
self.model.batch_size = 1
train.show_progress("done\n")
self.enc_vocab, _ = data_processer.initialize_vocabulary(config.VOCAB_ENC_TXT)
_, self.rev_dec_vocab = data_processer.initialize_vocabulary(config.VOCAB_DEC_TXT)
def run():
'''
Training function to run the training process after specifying parameters
'''
preprocessing.config.paths = ['./training_data/depressive1.json',
'./training_data/depressive2.json',
'./training_data/depressive3.json',
'./training_data/depressive4.json',
'./training_data/depressive5.json',
'./training_data/depressive6.json',
'./training_data/non-depressive1.json',
'./training_data/non-depressive2.json',
'./training_data/non-depressive3.json',
'./training_data/non-depressive4.json',
'./training_data/non-depressive5.json',
'./training_data/non-depressive6.json']
preprocessing.config.save_path = './training_data/all_training_data.csv'
preprocessing.config.labels = ['depressive', 'depressive', 'depressive', 'depressive', 'depressive', 'depressive',
'not-depressive', 'not-depressive', 'not-depressive', 'not-depressive',
'not-depressive', 'not-depressive']
preprocessing.config.keywords = ['depressed', 'lonely', 'sad', 'depression', 'tired', 'anxious',
'happy', 'joy', 'thankful', 'health', 'hopeful', 'glad']
preprocessing.config.nr_of_tweets = [1000, 1000, 1000, 1000, 1000, 1000,
1000, 1000, 1000, 1000, 1000, 1000]
history, early_stop_check = train.train_rnn(save_path='./weights/lstm_model_2.pth', collect=True) # Collect=False if already collected data
train.show_progress(history=history, save_name='./plots/training_progress.png')
train.animate_progress(history=history, save_path='./plots/training_animation_progress_REAL.gif',
early_stop_check=early_stop_check)
def twitter_bot():
# Only allocate part of the gpu memory when predicting.
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.2)
tf_config = tf.ConfigProto(gpu_options=gpu_options)
auth = tweepy.OAuthHandler(consumer_key, consumer_secret)
auth.set_access_token(access_token, access_token_secret)
api = tweepy.API(auth)
with tf.Session(config=tf_config) as sess:
train.show_progress("Creating model...")
model = train.create_or_restore_model(sess,
train.buckets,
forward_only=True)
model.batch_size = 1
train.show_progress("done\n")
enc_vocab, _ = data_processer.initialize_vocabulary(
config.VOCAB_ENC_TXT)
_, rev_dec_vocab = data_processer.initialize_vocabulary(
config.VOCAB_DEC_TXT)
for tweet in tweets():
status_id, status = tweet
print("Processing {0}...".format(status.text))
screen_name = status.author.screen_name
reply_body = predict.get_predition(sess, model, enc_vocab,
rev_dec_vocab,
status.text.encode('utf-8'))
if reply_body is None:
print("No reply predicted")
else:
reply_body = reply_body.replace('_UNK', '💩')
reply_text = "@" + screen_name + " " + reply_body
print("Reply:{0}".format(reply_text))
api.update_status(status=reply_text,
in_reply_to_status_id=status_id)
mark_tweet_processed(status_id)
def predict():
# Only allocate part of the gpu memory when predicting.
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=0.2)
tf_config = tf.ConfigProto(gpu_options=gpu_options)
with tf.Session(config=tf_config) as sess:
train.show_progress("Creating model...")
model = train.create_or_restore_model(sess, train.buckets, forward_only=True)
model.batch_size = 1
train.show_progress("done\n")
enc_vocab, _ = data_processer.initialize_vocabulary(config.VOCAB_ENC_TXT)
_, rev_dec_vocab = data_processer.initialize_vocabulary(config.VOCAB_DEC_TXT)
sys.stdout.write("> ")
sys.stdout.flush()
line = sys.stdin.readline()
while line:
line = line.encode('utf-8')
predicted = get_predition(sess, model, enc_vocab, rev_dec_vocab, line)
print(predicted)
print("> ", end="")
sys.stdout.flush()
line = sys.stdin.readline()