kishankumar09ei

from sklearn.model_selection import train_test_split
X= data['comment'].values
y= data['label'].values
train_df, test_df, y_train, y_test = train_test_split(X, y, stratify=y, test_size=0.2,random_state=42)

TEST_SPLIT = 0.2
BATCH_SIZE = 64

train_size = int(len(train_df) * (1-TEST_SPLIT))

kishankumar09ei

#creating encoding from bert tokenizer for test data
encodings_test = construct_encodings(list(test_df.values.flatten()), tkzr, max_len=MAX_LEN)

#Constructing the dataset for test
tfdataset_test = construct_tfdataset(encodings_test, y_test)
BATCH_SIZE = 64

tfdataset_test = tfdataset_test.shuffle(len(test_df))

tfdataset_test = tfdataset_test.batch(BATCH_SIZE)

kishankumar09ei

N_EPOCHS = 2

model_no_clean = TFDistilBertForSequenceClassification.from_pretrained(MODEL_NAME,num_labels=2)
optimizer = optimizers.Adam(learning_rate=3e-5)
loss = losses.SparseCategoricalCrossentropy(from_logits=True)
modelcompile(optimizer=optimizer, loss=loss, metrics=['accuracy'])

model.fit(tfdataset_train1, batch_size=BATCH_SIZE, epochs=N_EPOCHS)

kishankumar09ei

TEST_SPLIT = 0.2
BATCH_SIZE = 64

train_size = int(len(train_df) * (1-TEST_SPLIT))

tfdataset1 = tfdataset1.shuffle(len(train_df))
tfdataset_train1 = tfdataset1.take(train_size)
tfdataset_cv1 = tfdataset1.skip(train_size)

tfdataset_train1 = tfdataset_train1.batch(BATCH_SIZE)

kishankumar09ei

def construct_tfdataset(encodings, y=None):
    if y.any():
        return tf.data.Dataset.from_tensor_slices((dict(encodings),y))
    else:
        # this case is used when making predictions on unseen samples after training
        return tf.data.Dataset.from_tensor_slices(dict(encodings))
    
tfdataset1 = construct_tfdataset(encodings, y_train)

kishankumar09ei

def construct_encodings(x, tkzr, max_len, trucation=True, padding=True):
    return tkzr(x, max_length=max_len, truncation=trucation, padding=padding)


encodings = construct_encodings(list(train_df.flatten()), tkzr1, max_len=MAX_LEN)

kishankumar09ei

%%capture
#!python3 -m venv venv
#!source venv/bin/activate
#!pip install tensorflow transformers

import tensorflow as tf
from tensorflow.keras import activations, optimizers, losses
from transformers import DistilBertTokenizer, TFDistilBertForSequenceClassification
MODEL_NAME = 'distilbert-base-uncased'
MAX_LEN = 40

kishankumar09ei

#getting Vocab file
vocab_file = bert_layer.resolved_object.vocab_file.asset_path.numpy()
do_lower_case = bert_layer.resolved_object.do_lower_case.numpy()

from bert.tokenization import bert_tokenization
tokenizer=bert_tokenization.FullTokenizer(vocab_file,do_lower_case)


def create_tokens_mask_segment(text,tokenizer,max_seq_length):
    

kishankumar09ei

import tensorflow_hub as hub

## Loading the Pretrained Model from tensorflow HUB
tf.keras.backend.clear_session()

# maximum length of a seq in the data we have, for now i am making it as 55. You can change this
max_seq_length = 40

#BERT takes 3 inputs

kishankumar09ei

import os
import datetime
from tensorflow.keras.callbacks import ModelCheckpoint,TensorBoard,EarlyStopping,ReduceLROnPlateau
import warnings
warnings.filterwarnings("ignore") 
#creating tensboard call back object
log_dir = os.path.join('logs1','fits', datetime.datetime.now().strftime("%Y%m%d-%H%M%S"))
filepath = "best_lstm_model.hdf5"
#history_loss= LossHistory(validation_data=validation_data)
#logdir = os.path.join("logs", datetime.datetime.now().strftime("%Y%m%d-%H%M%S"))