Raman-Raje

class BilinearSimilarity(tf.keras.Model):
    """
    This function calcualtes biliear term used for answer span prediction.
    Referance took from --> https://github.com/kellywzhang/reading-comprehension/blob/master/attention.py
    
    """
    
    def __init__(self,hidden_size):
        
        super(BilinearSimilarity,self).__init__()

Raman-Raje

class Attention(tf.keras.Model):
    
    """
    This class implements a weighted sum attention.

    """
    def __init__(self, units = 1):
        
        super(Attention, self).__init__()
        self.W = tf.keras.layers.Dense(units)

Raman-Raje

# getting train data and dev data into csv file

# this snippet of code is for data extranction from json file.
contexts = []
questions = []
answers_text = []
answers_start = []
title = []
for i in range(train.shape[0]):
    topic = train.iloc[i,0]['paragraphs']

Raman-Raje

def countOccurences(word,context_window): 
    
    """
    This function returns the count of context word.
    """ 
    return context_window.count(word)



# https://stackoverflow.com/a/41663359/9371069

Raman-Raje

# Initiliazing the sequential model
model = Sequential()

#batch_size not given i.e None
#input_shape = (timesteps, input_dim)

model.add(LSTM(32,return_sequences=True,input_shape=(timesteps, input_dim)))
model.add(LSTM(16,input_shape=(timesteps, 32)))
model.add(Dropout(0.5))
model.add(Dense(n_classes, activation='sigmoid'))

Raman-Raje

model = Sequential()
model.add(Dense(32, input_dim=32))
# now: model.output_shape == (None, 32)
# note: `None` is the batch dimension

model.add(RepeatVector(3))
# now: model.output_shape == (None, 3, 32)

Raman-Raje

#for sequential modeling
model = Sequential()

# when batch size is not known
model.add(LSTM(5,input_shape = (128,1)))

# if batch size is known
model.add(LSTM(5,batch_input_shape=(16,128,1)))

Raman-Raje

# startseq kick starts caption generation proces
in_text = 'startseq'
for i in range(MAX_LENGTH):
        sequence = [word_to_ix[w] for w in in_text.split() if w in word_to_ix]
        sequence = pad_sequences([sequence], maxlen=MAX_LENGTH)
        yhat = model.predict([photo,sequence], verbose=0)
        # get word corresponding to maximum probability
        yhat = np.argmax(yhat)
        word = ix_to_word[yhat]
        in_text += ' ' + word

Raman-Raje

# data generator, intended to be used in a call to model.fit_generator()

def data_generator(descriptions, photos, MAX_LENGTH,VOCAB_SIZE, num_photos_per_batch):
    X1, X2, y = list(), list(), list()
    n=0
    # loop for ever over images
    while 1:
        for key, desc_list in train_descriptions.items():
            n+=1
            # retrieve the photo feature

Raman-Raje

# iterate for all images in train
for key, desc_list in train_descriptions.items():
    n+=1
    # retrieve the photo feature
    photo = photos[key]
    for desc in desc_list:
        seq = token.texts_to_sequences([desc])
        seq = seq[0]
        for i in range(1,len(seq)):
            in_seq , op_seq = seq[:i],seq[i]