review_sentiment.py

# -*- coding: utf-8 -*-
"""Review sentiment.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1KJJbg4V-8S1VFHqaj8RT3G-0TJr4ADyz
"""

import pandas as pd
import numpy as np
dataset = pd.read_csv("Reviews.csv")

dataset.head()

def f(row):
  if row['Score']>=1 and row['Score']<=3:
    val = 0
  else:
    val = 1
  return val
dataset['sentiment']= dataset.apply(f,axis = 1)

dataset['sentiment'].value_counts()

grouped = dataset.groupby(dataset.sentiment)
sample_1=grouped.get_group(1)
sample_0 = grouped.get_group(0)
sample_1['sentiment'].value_counts()
sample_1 = sample_1.sample(frac = 0.29)
sample_1['sentiment'].value_counts()
sample_1['sentiment'].value_counts()
frames = [sample_1,sample_0]
last_version_dataset = pd.concat(frames)
last_version_dataset['sentiment'].value_counts()

df = last_version_dataset.sample(frac = 0.1)
df.isnull().sum()

df = df.dropna()

df.isnull().sum()

from transformers import pipeline

generator = pipeline(task = "text-generation")

from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("distilgpt2")
model = AutoModelForCausalLM.from_pretrained("distilgpt2")
better_generator = pipeline(task = "text-generation",model = model, tokenizer = tokenizer)

from sklearn.model_selection import train_test_split
from transformers import DistilBertForSequenceClassification, TrainingArguments, Trainer, DistilBertTokenizerFast

tokenizer = DistilBertTokenizerFast.from_pretrained("distilbert-base-uncased")

labels = df['sentiment']
texts = df['Text']

#Splitting dataset into training and test
train_texts, test_texts, train_labels, test_labels = train_test_split(texts, labels, test_size=.3)

#Tokenization
train_encodings = tokenizer(list(train_texts),truncation=True, padding = True)
test_encodings = tokenizer(list(test_texts),truncation=True, padding = True)

import torch
class IMDbDataset(torch.utils.data.Dataset):
    def __init__(self, encodings, labels):
        self.encodings = encodings
        self.labels = labels

    def __getitem__(self, idx):
        item = {key: torch.tensor(val[idx]) for key, val in self.encodings.items()}
        item['labels'] = torch.tensor(self.labels[idx])
        return item

    def __len__(self):
        return len(self.labels)

train_dataset = IMDbDataset(train_encodings, list(train_labels))
test_dataset = IMDbDataset(test_encodings, list(test_labels))

import evaluate
model = DistilBertForSequenceClassification.from_pretrained("distilbert-base-uncased")

metric = evaluate.load("accuracy")
training_args = TrainingArguments(
    output_dir='./results',          # output directory
    num_train_epochs=1,           # total number of training epochs
    per_device_train_batch_size=16,  # batch size per device during training
    per_device_eval_batch_size=64,   # batch size for evaluation
    warmup_steps=0,                # number of warmup steps for learning rate scheduler
    weight_decay=0                # strength of weight decay
)
def compute_metrics(eval_pred):
    logits, labels = eval_pred
    predictions = np.argmax(logits, axis=-1)
    return metric.compute(predictions=predictions, references=labels)
    trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=train_dataset,
    eval_dataset=test_dataset,
    compute_metrics=compute_metrics,
)
    trainer.train()
model = model.to(device="cpu")
our_predicts = pipeline(model = model, tokenizer=tokenizer, task="text-classification")
print(our_predicts("This is a pretty solid product. It lasted a while and I enjoyed using it. 10/10 would recommend."))
torch.save(our_predicts.model.state_dict(), "model.pt")