configuration = getattr(configurations_0, args.proto)()
# added by Zhaopeng Tu, 2016-05-12
if args.state:
configuration.update(eval(open(args.state).read()))
logger.info("\nModel options:\n{}".format(pprint.pformat(configuration)))
rng = numpy.random.RandomState(1234)
enc_dec = EncoderDecoder(rng, **configuration)
enc_dec.build_sampler()
# added by Zhaopeng Tu, 2016-05-27
# options to use other trained models
if args.model:
enc_dec.load(path=args.model)
else:
enc_dec.load(path=configuration['saveto_best'])
test_align = Align(enc_dec=enc_dec, configuration)
test_src = get_stream(args.source, configuration['vocab_src'],
**configuration)
test_trg = get_stream(args.target, configuration['vocab_trg'],
**configuration)
aligner = Aligner(align_model=test_align,
test_src=test_src,
test_trg=test_trg,
**configuration)
aligner.apply(args.alignment, True)
def get_external_metadata(inputfile):
# define an empty python dictionary
md = {}
# Check whether this file exists.
if not os.path.exists(inputfile):
return md
# Get the other meta data field parameters
md['file_name'] = os.path.basename(inputfile)
md['file_size'] = str(os.path.getsize(inputfile))
md['crc'] = fileEnstoreChecksum(inputfile)
# Quit here if file type is not ".root"
if not inputfile.endswith('.root'):
return md
# Root checks.
file = ROOT.TFile.Open(larbatch_posix.root_stream(inputfile))
if file and file.IsOpen() and not file.IsZombie():
# Root file opened successfully.
# Get number of events.
obj = file.Get('Events')
if obj and obj.InheritsFrom('TTree'):
# This has a TTree named Events.
nev = obj.GetEntriesFast()
md['events'] = str(nev)
# Get runs and subruns fro SubRuns tree.
subrun_tree = file.Get('SubRuns')
if subrun_tree and subrun_tree.InheritsFrom('TTree'):
md['subruns'] = []
nsubruns = subrun_tree.GetEntriesFast()
tfr = ROOT.TTreeFormula('subruns',
'SubRunAuxiliary.id_.run_.run_',
subrun_tree)
tfs = ROOT.TTreeFormula('subruns',
'SubRunAuxiliary.id_.subRun_',
subrun_tree)
for entry in range(nsubruns):
subrun_tree.GetEntry(entry)
run = tfr.EvalInstance64()
subrun = tfs.EvalInstance64()
run_subrun = (run, subrun)
if not run_subrun in md['subruns']:
md['subruns'].append(run_subrun)
# Get stream name.
try:
stream_name = stream.get_stream(inputfile)
md['data_stream'] = stream_name
except:
pass
return md
def build_and_run(save_to,modelconfig,experimentconfig):
""" part of this is adapted from lasagne tutorial"""
n, num_filters, image_size, num_blockstack = modelconfig['depth'], modelconfig['num_filters'], modelconfig['image_size'], modelconfig['num_blockstack']
print("Amount of bottlenecks: %d" % n)
# Prepare Theano variables for inputs and targets
input_var = T.tensor4('image_features')
#target_value = T.ivector('targets')
target_var = T.lmatrix('targets')
target_vec = T.extra_ops.to_one_hot(target_var[:,0],2)
#target_var = T.matrix('targets')
# Create residual net model
print("Building model...")
network = build_cnn(input_var, image_size, n, num_blockstack, num_filters)
get_info(network)
prediction = lasagne.utils.as_theano_expression(lasagne.layers.get_output(network))
test_prediction = lasagne.utils.as_theano_expression(lasagne.layers.get_output(network,deterministic=True))
# Loss function -> The objective to minimize
print("Instanciation of loss function...")
#loss = CategoricalCrossEntropy().apply(target_var.flatten(), prediction)
#test_loss = CategoricalCrossEntropy().apply(target_var.flatten(), test_prediction)
# loss = lasagne.objectives.categorical_crossentropy(prediction, target_var.flatten()).mean()
# test_loss = lasagne.objectives.categorical_crossentropy(test_prediction, target_var.flatten()).mean()
loss = lasagne.objectives.squared_error(prediction,target_vec).mean()
test_loss = lasagne.objectives.squared_error(test_prediction,target_vec).mean()
# loss = tensor.nnet.binary_crossentropy(prediction, target_var).mean()
# test_loss = tensor.nnet.binary_crossentropy(test_prediction, target_var).mean()
test_loss.name = "loss"
# loss.name = 'x-ent_error'
# loss.name = 'sqr_error'
layers = lasagne.layers.get_all_layers(network)
#l1 and l2 regularization
#pondlayers = {x:0.000025 for i,x in enumerate(layers)}
#l1_penality = lasagne.regularization.regularize_layer_params_weighted(pondlayers, lasagne.regularization.l2)
#l2_penality = lasagne.regularization.regularize_layer_params(layers[len(layers)/4:], lasagne.regularization.l1) * 25e-6
#reg_penalty = l1_penality + l2_penality
#reg_penalty.name = 'reg_penalty'
#loss = loss + reg_penalty
loss.name = 'reg_loss'
error_rate = MisclassificationRate().apply(target_var.flatten(), test_prediction).copy(
name='error_rate')
# Load the dataset
print("Loading data...")
istest = 'test' in experimentconfig.keys()
if istest:
print("Using test stream")
train_stream, valid_stream, test_stream = get_stream(experimentconfig['batch_size'],image_size,test=istest)
# Defining step rule and algorithm
if 'step_rule' in experimentconfig.keys() and not experimentconfig['step_rule'] is None :
step_rule = experimentconfig['step_rule'](learning_rate=experimentconfig['learning_rate'])
else :
step_rule=Scale(learning_rate=experimentconfig['learning_rate'])
params = map(lasagne.utils.as_theano_expression,lasagne.layers.get_all_params(network, trainable=True))
print("Initializing algorithm")
algorithm = GradientDescent(
cost=loss, gradients={var:T.grad(loss,var) for var in params},#parameters=cg.parameters, #params
step_rule=step_rule)
#algorithm.add_updates(extra_updates)
grad_norm = aggregation.mean(algorithm.total_gradient_norm)
grad_norm.name = "grad_norm"
print("Initializing extensions...")
plot = Plot(save_to, channels=[['train_loss','valid_loss'],
['train_grad_norm'],
#['train_grad_norm','train_reg_penalty'],
['train_error_rate','valid_error_rate']], server_url='http://hades.calculquebec.ca:5042')
checkpoint = Checkpoint('models/best_'+save_to+'.tar')
# checkpoint.add_condition(['after_n_batches=25'],
checkpoint.add_condition(['after_epoch'],
predicate=OnLogRecord('valid_error_rate_best_so_far'))
#Defining extensions
extensions = [Timing(),
FinishAfter(after_n_epochs=experimentconfig['num_epochs'],
after_n_batches=experimentconfig['num_batches']),
TrainingDataMonitoring([test_loss, error_rate, grad_norm], # reg_penalty],
prefix="train", after_epoch=True), #after_n_epochs=1
DataStreamMonitoring([test_loss, error_rate],valid_stream,prefix="valid", after_epoch=True), #after_n_epochs=1
plot,
#Checkpoint(save_to,after_n_epochs=5),
#ProgressBar(),
# Plot(save_to, channels=[['train_loss','valid_loss'], ['train_error_rate','valid_error_rate']], server_url='http://hades.calculquebec.ca:5042'), #'grad_norm'
# after_batch=True),
Printing(after_epoch=True),
TrackTheBest('valid_error_rate',min), #Keep best
checkpoint, #Save best
FinishIfNoImprovementAfter('valid_error_rate_best_so_far', epochs=5)] # Early-stopping
# model = Model(loss)
# print("Model",model)
main_loop = MainLoop(
algorithm,
train_stream,
# model=model,
extensions=extensions)
print("Starting main loop...")
main_loop.run()
import sqlite3
from sqlite3 import Error
import stream as st
from time import strftime
stream = st.create_stream()
tweets, dt, future = st.get_stream(stream)
conn = sqlite3.connect("assign2.db")
db = conn.cursor()
db.execute('drop table if exists HASHTAGS;')
db.execute(
'create table HASHTAGS (tag_ID integer primary key autoincrement,hashtag text);'
)
conn.commit()
for tweet in tweets:
if 'entities' in tweet:
hashtags = tweet['entities']['hashtags']
for hashtag in hashtags:
# print "%s" % (hashtag['text'])
db.execute('INSERT INTO HASHTAGS (hashtag) VALUES (?);',
[hashtag['text']])
conn.commit()
cur = db.execute(
'select hashtag, count(*) as count from HASHTAGS group by hashtag order by count desc limit 10;'
)
top10 = cur.fetchall()
fp = open('assign2-report.txt', 'w')
def build_and_run(label, config):
############## CREATE THE NETWORK ###############
#Define the parameters
num_epochs, num_batches, num_channels, image_shape, filter_size, num_filter, pooling_sizes, mlp_hiddens, output_size, batch_size, activation, mlp_activation = config[
'num_epochs'], config['num_batches'], config['num_channels'], config[
'image_shape'], config['filter_size'], config[
'num_filter'], config['pooling_sizes'], config[
'mlp_hiddens'], config['output_size'], config[
'batch_size'], config['activation'], config[
'mlp_activation']
# print(num_epochs, num_channels, image_shape, filter_size, num_filter, pooling_sizes, mlp_hiddens, output_size, batch_size, activation, mlp_activation)
lambda_l1 = 0.000025
lambda_l2 = 0.000025
print("Building model")
#Create the symbolics variable
x = T.tensor4('image_features')
y = T.lmatrix('targets')
#Get the parameters
conv_parameters = zip(filter_size, num_filter)
#Create the convolutions layers
conv_layers = list(
interleave([(Convolutional(filter_size=filter_size,
num_filters=num_filter,
name='conv_{}'.format(i))
for i, (filter_size,
num_filter) in enumerate(conv_parameters)),
(activation),
(MaxPooling(size, name='pool_{}'.format(i))
for i, size in enumerate(pooling_sizes))]))
# (AveragePooling(size, name='pool_{}'.format(i)) for i, size in enumerate(pooling_sizes))]))
#Create the sequence
conv_sequence = ConvolutionalSequence(conv_layers,
num_channels,
image_size=image_shape,
weights_init=Uniform(width=0.2),
biases_init=Constant(0.))
#Initialize the convnet
conv_sequence.initialize()
#Add the MLP
top_mlp_dims = [np.prod(conv_sequence.get_dim('output'))
] + mlp_hiddens + [output_size]
out = Flattener().apply(conv_sequence.apply(x))
mlp = MLP(mlp_activation,
top_mlp_dims,
weights_init=Uniform(0, 0.2),
biases_init=Constant(0.))
#Initialisze the MLP
mlp.initialize()
#Get the output
predict = mlp.apply(out)
cost = CategoricalCrossEntropy().apply(y.flatten(),
predict).copy(name='cost')
error = MisclassificationRate().apply(y.flatten(), predict)
#Little trick to plot the error rate in two different plots (We can't use two time the same data in the plot for a unknow reason)
error_rate = error.copy(name='error_rate')
error_rate2 = error.copy(name='error_rate2')
########### REGULARIZATION ##################
cg = ComputationGraph([cost])
weights = VariableFilter(roles=[WEIGHT])(cg.variables)
biases = VariableFilter(roles=[BIAS])(cg.variables)
# # l2_penalty_weights = T.sum([i*lambda_l2/len(weights) * (W ** 2).sum() for i,W in enumerate(weights)]) # Gradually increase penalty for layer
l2_penalty = T.sum([
lambda_l2 * (W**2).sum() for i, W in enumerate(weights + biases)
]) # Gradually increase penalty for layer
# # #l2_penalty_bias = T.sum([lambda_l2*(B **2).sum() for B in biases])
# # #l2_penalty = l2_penalty_weights + l2_penalty_bias
l2_penalty.name = 'l2_penalty'
l1_penalty = T.sum([lambda_l1 * T.abs_(z).sum() for z in weights + biases])
# l1_penalty_weights = T.sum([i*lambda_l1/len(weights) * T.abs_(W).sum() for i,W in enumerate(weights)]) # Gradually increase penalty for layer
# l1_penalty_biases = T.sum([lambda_l1 * T.abs_(B).sum() for B in biases])
# l1_penalty = l1_penalty_biases + l1_penalty_weights
l1_penalty.name = 'l1_penalty'
costreg = cost + l2_penalty + l1_penalty
costreg.name = 'costreg'
########### DEFINE THE ALGORITHM #############
# algorithm = GradientDescent(cost=cost, parameters=cg.parameters, step_rule=Momentum())
algorithm = GradientDescent(cost=costreg,
parameters=cg.parameters,
step_rule=Adam())
########### GET THE DATA #####################
istest = 'test' in config.keys()
train_stream, valid_stream, test_stream = get_stream(batch_size,
image_shape,
test=istest)
########### INITIALIZING EXTENSIONS ##########
checkpoint = Checkpoint('models/best_' + label + '.tar')
checkpoint.add_condition(
['after_epoch'], predicate=OnLogRecord('valid_error_rate_best_so_far'))
#Adding a live plot with the bokeh server
plot = Plot(
label,
channels=[
['train_error_rate', 'valid_error_rate'],
['valid_cost', 'valid_error_rate2'],
# ['train_costreg','train_grad_norm']], #
[
'train_costreg', 'train_total_gradient_norm',
'train_l2_penalty', 'train_l1_penalty'
]
],
server_url="http://hades.calculquebec.ca:5042")
grad_norm = aggregation.mean(algorithm.total_gradient_norm)
grad_norm.name = 'grad_norm'
extensions = [
Timing(),
FinishAfter(after_n_epochs=num_epochs, after_n_batches=num_batches),
DataStreamMonitoring([cost, error_rate, error_rate2],
valid_stream,
prefix="valid"),
TrainingDataMonitoring([
costreg, error_rate, error_rate2, grad_norm, l2_penalty, l1_penalty
],
prefix="train",
after_epoch=True),
plot,
ProgressBar(),
Printing(),
TrackTheBest('valid_error_rate', min), #Keep best
checkpoint, #Save best
FinishIfNoImprovementAfter('valid_error_rate_best_so_far', epochs=4)
] # Early-stopping
model = Model(cost)
main_loop = MainLoop(algorithm,
data_stream=train_stream,
model=model,
extensions=extensions)
main_loop.run()
def build_and_run(label, config):
############## CREATE THE NETWORK ###############
#Define the parameters
num_epochs, num_batches, num_channels, image_shape, filter_size, num_filter, pooling_sizes, mlp_hiddens, output_size, batch_size, activation, mlp_activation = config['num_epochs'], config['num_batches'], config['num_channels'], config['image_shape'], config['filter_size'], config['num_filter'], config['pooling_sizes'], config['mlp_hiddens'], config['output_size'], config['batch_size'], config['activation'], config['mlp_activation']
# print(num_epochs, num_channels, image_shape, filter_size, num_filter, pooling_sizes, mlp_hiddens, output_size, batch_size, activation, mlp_activation)
lambda_l1 = 0.000025
lambda_l2 = 0.000025
print("Building model")
#Create the symbolics variable
x = T.tensor4('image_features')
y = T.lmatrix('targets')
#Get the parameters
conv_parameters = zip(filter_size, num_filter)
#Create the convolutions layers
conv_layers = list(interleave([(Convolutional(
filter_size=filter_size,
num_filters=num_filter,
name='conv_{}'.format(i))
for i, (filter_size, num_filter)
in enumerate(conv_parameters)),
(activation),
(MaxPooling(size, name='pool_{}'.format(i)) for i, size in enumerate(pooling_sizes))]))
# (AveragePooling(size, name='pool_{}'.format(i)) for i, size in enumerate(pooling_sizes))]))
#Create the sequence
conv_sequence = ConvolutionalSequence(conv_layers, num_channels, image_size=image_shape, weights_init=Uniform(width=0.2), biases_init=Constant(0.))
#Initialize the convnet
conv_sequence.initialize()
#Add the MLP
top_mlp_dims = [np.prod(conv_sequence.get_dim('output'))] + mlp_hiddens + [output_size]
out = Flattener().apply(conv_sequence.apply(x))
mlp = MLP(mlp_activation, top_mlp_dims, weights_init=Uniform(0, 0.2),
biases_init=Constant(0.))
#Initialisze the MLP
mlp.initialize()
#Get the output
predict = mlp.apply(out)
cost = CategoricalCrossEntropy().apply(y.flatten(), predict).copy(name='cost')
error = MisclassificationRate().apply(y.flatten(), predict)
#Little trick to plot the error rate in two different plots (We can't use two time the same data in the plot for a unknow reason)
error_rate = error.copy(name='error_rate')
error_rate2 = error.copy(name='error_rate2')
########### REGULARIZATION ##################
cg = ComputationGraph([cost])
weights = VariableFilter(roles=[WEIGHT])(cg.variables)
biases = VariableFilter(roles=[BIAS])(cg.variables)
# # l2_penalty_weights = T.sum([i*lambda_l2/len(weights) * (W ** 2).sum() for i,W in enumerate(weights)]) # Gradually increase penalty for layer
l2_penalty = T.sum([lambda_l2 * (W ** 2).sum() for i,W in enumerate(weights+biases)]) # Gradually increase penalty for layer
# # #l2_penalty_bias = T.sum([lambda_l2*(B **2).sum() for B in biases])
# # #l2_penalty = l2_penalty_weights + l2_penalty_bias
l2_penalty.name = 'l2_penalty'
l1_penalty = T.sum([lambda_l1*T.abs_(z).sum() for z in weights+biases])
# l1_penalty_weights = T.sum([i*lambda_l1/len(weights) * T.abs_(W).sum() for i,W in enumerate(weights)]) # Gradually increase penalty for layer
# l1_penalty_biases = T.sum([lambda_l1 * T.abs_(B).sum() for B in biases])
# l1_penalty = l1_penalty_biases + l1_penalty_weights
l1_penalty.name = 'l1_penalty'
costreg = cost + l2_penalty + l1_penalty
costreg.name = 'costreg'
########### DEFINE THE ALGORITHM #############
# algorithm = GradientDescent(cost=cost, parameters=cg.parameters, step_rule=Momentum())
algorithm = GradientDescent(cost=costreg, parameters=cg.parameters, step_rule=Adam())
########### GET THE DATA #####################
istest = 'test' in config.keys()
train_stream, valid_stream, test_stream = get_stream(batch_size,image_shape,test=istest)
########### INITIALIZING EXTENSIONS ##########
checkpoint = Checkpoint('models/best_'+label+'.tar')
checkpoint.add_condition(['after_epoch'],
predicate=OnLogRecord('valid_error_rate_best_so_far'))
#Adding a live plot with the bokeh server
plot = Plot(label,
channels=[['train_error_rate', 'valid_error_rate'],
['valid_cost', 'valid_error_rate2'],
# ['train_costreg','train_grad_norm']], #
['train_costreg','train_total_gradient_norm','train_l2_penalty','train_l1_penalty']],
server_url="http://hades.calculquebec.ca:5042")
grad_norm = aggregation.mean(algorithm.total_gradient_norm)
grad_norm.name = 'grad_norm'
extensions = [Timing(),
FinishAfter(after_n_epochs=num_epochs,
after_n_batches=num_batches),
DataStreamMonitoring([cost, error_rate, error_rate2], valid_stream, prefix="valid"),
TrainingDataMonitoring([costreg, error_rate, error_rate2,
grad_norm,l2_penalty,l1_penalty],
prefix="train", after_epoch=True),
plot,
ProgressBar(),
Printing(),
TrackTheBest('valid_error_rate',min), #Keep best
checkpoint, #Save best
FinishIfNoImprovementAfter('valid_error_rate_best_so_far', epochs=4)] # Early-stopping
model = Model(cost)
main_loop = MainLoop(algorithm,data_stream=train_stream,model=model,extensions=extensions)
main_loop.run()
def build_and_run(experimentconfig, modelconfig, save_to=None): #modelconfig,
""" part of this is adapted from lasagne tutorial"""
# Prepare Theano variables for inputs and targets
input_var = T.tensor4('image_features')
target_var = T.lmatrix('targets')
target_vec = T.extra_ops.to_one_hot(target_var[:,0],2)
# Create vgg model
print("Building model...")
image_size = modelconfig['image_size']
network = vgg16.build_small_model()
prediction = lasagne.utils.as_theano_expression(lasagne.layers.get_output(network["prob"],input_var))
# test_prediction = lasagne.layers.get_output(network["prob"],input_var,deterministic=True)
# Loss function -> The objective to minimize
print("Instanciation of loss function...")
# loss = lasagne.objectives.categorical_crossentropy(prediction, target_var.flatten())
loss = lasagne.objectives.squared_error(prediction,target_vec)
# test_loss = lasagne.objectives.squared_error(test_prediction,target_vec)
loss = loss.mean()
# layers = network.values()
#l1 and l2 regularization
# pondlayers = {x:0.01 for x in layers}
# l1_penality = lasagne.regularization.regularize_layer_params_weighted(pondlayers, lasagne.regularization.l2)
# l2_penality = lasagne.regularization.regularize_layer_params(layers[len(layers)/4:], lasagne.regularization.l1) * 1e-4
# reg_penalty = l1_penality + l2_penality
# reg_penalty.name = 'reg_penalty'
#loss = loss + reg_penalty
loss.name = 'loss'
error_rate = MisclassificationRate().apply(target_var.flatten(), prediction).copy(
name='error_rate')
# Load the dataset
print("Loading data...")
if 'test' in experimentconfig.keys() and experimentconfig['test'] is True:
train_stream, valid_stream, test_stream = get_stream(experimentconfig['batch_size'],image_size,test=True)
else :
train_stream, valid_stream, test_stream = get_stream(experimentconfig['batch_size'],image_size,test=False)
# Defining step rule and algorithm
if 'step_rule' in experimentconfig.keys() and not experimentconfig['step_rule'] is None :
step_rule = experimentconfig['step_rule'](learning_rate=experimentconfig['learning_rate'])
else :
step_rule=Scale(learning_rate=experimentconfig['learning_rate'])
params = map(lasagne.utils.as_theano_expression,lasagne.layers.get_all_params(network['prob'], trainable=True))
algorithm = GradientDescent(
cost=loss, gradients={var:T.grad(loss,var) for var in params},
step_rule=step_rule)
grad_norm = aggregation.mean(algorithm.total_gradient_norm)
grad_norm.name='grad_norm'
print("Initializing extensions...")
plot = Plot(save_to, channels=[['train_loss','valid_loss','train_grad_norm'],['train_error_rate','valid_error_rate']], server_url='http://hades.calculquebec.ca:5042')
checkpoint = Checkpoint('models/best_'+save_to+'.tar')
# checkpoint.add_condition(['after_n_batches=25'],
checkpoint.add_condition(['after_epoch'],
predicate=OnLogRecord('valid_error_rate_best_so_far'))
#Defining extensions
extensions = [Timing(),
FinishAfter(after_n_epochs=experimentconfig['num_epochs'],
after_n_batches=experimentconfig['num_batches']),
TrainingDataMonitoring([loss, error_rate, grad_norm, reg_penalty], prefix="train", after_epoch=True), #after_n_epochs=1
DataStreamMonitoring([loss, error_rate],valid_stream,prefix="valid", after_epoch=True), #after_n_epochs=1
#Checkpoint(save_to,after_n_epochs=5),
#ProgressBar(),
plot,
# after_batch=True),
Printing(after_epoch=True),
TrackTheBest('valid_error_rate',min), #Keep best
checkpoint, #Save best
FinishIfNoImprovementAfter('valid_error_rate_best_so_far', epochs=5)] # Early-stopping
# model = Model(ComputationGraph(network))
main_loop = MainLoop(
algorithm,
train_stream,
# model=model,
extensions=extensions)
print("Starting main loop...")
main_loop.run()
import sqlite3
from sqlite3 import Error
import stream as twitter_stream
from time import strftime
stream = twitter_stream.create_stream()
tweet_buff, date_time, future = twitter_stream.get_stream(stream)
connection = sqlite3.connect("tweets.db")
tweet_db = connection.cursor()
tweet_db.execute('drop table if exists POPULAR_HASHTAG;')
tweet_db.execute('create table POPULAR_HASHTAG (tag_ID integer primary key autoincrement,hashtag text);')
connection.commit()
fp = open('report.txt', 'w')
for tweet in tweet_buff:
if 'entities' in tweet:
hashtags = tweet['entities']['hashtags']
for hashtag in hashtags:
tweet_db.execute('INSERT INTO POPULAR_HASHTAG (hashtag) VALUES (?);', [hashtag['text']])
connection.commit()
result = tweet_db.execute('select hashtag, count(*) as count from POPULAR_HASHTAG group by hashtag order by count desc limit 10;')
popHashtags = result.fetchall()
print('{}\t{}'.format(date_time,future))
fp.write('{}\t{}\n'.format(date_time,future))
for popHashtag in popHashtags:
print(popHashtag[0].encode('utf-8') + '\t{}'.format(popHashtag[1])) #encoding to priduce hashtags in that are not in english
fp.write(popHashtag[0].encode('utf-8') + '\t{}\n'.format(popHashtag[1]))