import os
import pandas as pd
import tensorflow as tf
import numpy as np
data=pd.read_csv('train.csv')
data.head(5)

from tensorflow.keras.layers import TextVectorization

x=data['comment_text']
y=data[data.columns[2:]].values

max_features=200000
vectorizer=TextVectorization(max_tokens=max_features,
                            output_sequence_length=1800,
                            output_mode='int')
vectorizer.get_vocabulary()
vectorizer.adapt(x.values)
vectorizer("have you watched breaking bad")[:5]
vectorized_text=vectorizer(x.values)
dataset=tf.data.Dataset.from_tensor_slices((vectorized_text, y))
dataset=dataset.cache()
dataset=dataset.shuffle(160000)
dataset=dataset.batch(16)
dataset=dataset.prefetch(8)

batch_x, batch_y = dataset.as_numpy_iterator().next()

train=dataset.take(int(len(dataset)*.7))
val=dataset.skip(int(len(dataset)*.7)).take(int(len(dataset)*.2))
test=dataset.skip(int(len(dataset)*.9)).take(int(len(dataset)*.1))

train_generator=train.as_numpy_iterator()
train_generator.next()

from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import LSTM, Dropout, Bidirectional, Dense, Embedding

model=Sequential()
model.add(Embedding(max_features+1, 32))
model.add(Bidirectional(LSTM(32, activation='tanh')))
model.add(Dense(128, activation='relu'))
model.add(Dense(256, activation='relu'))
model.add(Dense(128, activation='relu'))
model.add(Dense(6, activation='sigmoid'))
model.compile(loss='BinaryCrossentropy', optimizer='adam', metrics=['accuracy'])

model.summary()

history=model.fit(train, epochs=10, validation_data=val)

model.evaluate(test)

x_batch, y_batch = test.as_numpy_iterator().next()
(model.predict(x_batch) > 0.5).astype(int)
input_text=vectorizer('I am coming to kill you pal')

input_text[:7]
batch=test.as_numpy_iterator().next()
res=model.predict(np.expand_dims(input_text,0))
res
model.save('finalprojecttoxic.h5')

from transformers import pipeline

import gradio as gr

model=tf.keras.models.load_model('finalprojecttoxic.h5')
input_str=vectorizer('Hey i freaking hate you!. I\'m going to hurt you!')
res=model.predict(np.expand_dims(input_str,0))


translator_hindi = pipeline("translation", model="Helsinki-NLP/opus-mt-hi-en", tokenizer="Helsinki-NLP/opus-mt-hi-en")
hindi_text = "नमस्ते, आप कैसे हैं?"
en_to_hin = translator_hindi(hindi_text)
en_to_hin[0]['translation_text']

def translate_hindi(from_text):
    result2 = translator_hindi(from_text)
    
    return result2[0]['translation_text']


translate_hindi('नमस्ते, आप कैसे हैं?')

def score_comment(comment):
    vectorized_comment = vectorizer([comment])
    results=model.predict(vectorized_comment)
    
    text=''
    for idx, col in enumerate(data.columns[2:]):
        text+= '{}: {}\n'.format(col, results[0][idx]>0.5)
        
    return text

def combined_models(input):
    output1=translate_hindi(input)
    output2=score_comment(input)
    
    return output1, output2

interface = gr.Interface(fn=combined_models, inputs="text", outputs=["text","text"],title="Toxic Comment Analyzer")

interface.launch(share=True)