# -*- coding: utf-8 -*-
"""Lyrics5.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1eEWnqPJ4BuKnMDL-kK9b2-vvMJ6_HTyS
"""

!pip install keras

!pip install keras_preprocessing
import keras_preprocessing

!pip install pad_sequences
import pad_sequences

import pandas as pd
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from wordcloud import WordCloud, STOPWORDS, ImageColorGenerator
import string, os
import tensorflow as tf

# keras module for building LSTM
from keras_preprocessing.sequence import pad_sequences
from tensorflow.keras.layers import Embedding, Dropout, LSTM, Dense, Bidirectional
from keras.preprocessing.text import Tokenizer
from keras.callbacks import EarlyStopping
from keras.models import Sequential

import matplotlib.pyplot as plt
import seaborn as sns

df = pd.read_csv('/content/drive/MyDrive/Colab Notebooks/lyrics-data.csv')

df.head()

df.drop(['ALink', 'SName','SLink'],axis=1,inplace=True)

df.shape

df['language'].value_counts()

df = df[df['language']=='en']

df = df[:350]

df.shape

df['Number_of_words'] = df['Lyric'].apply(lambda x:len(str(x).split()))
df.head()

df['Number_of_words'].describe()

import matplotlib.pyplot as plt
plt.style.use('ggplot')
plt.figure(figsize=(12,6))
sns.distplot(df['Number_of_words'],kde = False,color="red",bins=200)
plt.title("Frequency distribution of number of words for each text extracted", size=20)

tokenizer = Tokenizer()
tokenizer.fit_on_texts(df['Lyric'].astype(str).str.lower())

total_words = len(tokenizer.word_index)+1
tokenized_sentences = tokenizer.texts_to_sequences(df['Lyric'].astype(str))
tokenized_sentences[0]

input_sequences = list()
for i in tokenized_sentences:
    for t in range(1, len(i)):
        n_gram_sequence = i[:t+1]
        input_sequences.append(n_gram_sequence)

# Pre padding
max_sequence_len = max([len(x) for x in input_sequences])
input_sequences = np.array(pad_sequences(input_sequences, maxlen=max_sequence_len, padding='pre'))

input_sequences[:10]

X, labels = input_sequences[:,:-1],input_sequences[:,-1]
y = tf.keras.utils.to_categorical(labels, num_classes=total_words)

model = Sequential()
model.add(Embedding(total_words, 40, input_length=max_sequence_len-1))
model.add(Bidirectional(LSTM(250)))
model.add(Dropout(0.1))
model.add(Dense(total_words, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
earlystop = EarlyStopping(monitor='loss', min_delta=0, patience=3, verbose=0, mode='auto')
history = model.fit(X, y, epochs=10, verbose=1, callbacks=[earlystop])

plt.plot(history.history['accuracy'], label='train acc')
plt.legend()
plt.show()
plt.savefig('AccVal_acc')

def complete_this_song(seed_text, next_words):
    for _ in range(next_words):
        token_list = tokenizer.texts_to_sequences([seed_text])[0]
        token_list = pad_sequences([token_list], maxlen=max_sequence_len-1, padding='pre')
        #predicted = model.predict_classes(token_list, verbose=0)
        predict_x=model.predict(token_list, verbose=0)
        classes_x=np.argmax(predict_x,axis=1)
        output_word = ""
        for word, index in tokenizer.word_index.items():
            if index == classes_x:
                output_word = word
                break
        seed_text += " " + output_word
    return seed_text

complete_this_song("the sky is blue", 40)

!pip install keras.models

from tensorflow.keras.models import load_model
model.save('/content/drive/MyDrive/Colab Notebooks/song_lyrics_generator.h5')

import tensorflow as tf

from tensorflow.keras.models import load_model
song_lyrics_generator= tf.keras.models.load_model('/content/drive/MyDrive/Colab Notebooks/song_lyrics_generator.h5')

!pip install gradio
import gradio as gr

interface = gr.Interface(fn= complete_this_song,
                         inputs= ['text', gr.inputs.Slider(0,250, label='No. of words')],
                         outputs='text')

interface.launch()