classifier.py

import torch
import transformers
from transformers import BertModel, BertTokenizer, AutoTokenizer
from torch import nn, optim
from torch.utils.data import Dataset, DataLoader
import torch.nn.functional as F

###########################################################
review_text = "I love you"
###########################################################


PRE_TRAINED_MODEL_NAME = 'nreimers/BERT-Tiny_L-2_H-128_A-2'
class_names = ["Normal", "Abusive"]
MAX_LEN = "max_length"

class CyberbullyingClassifier(nn.Module):

  def __init__(self, n_classes):
    super(CyberbullyingClassifier, self).__init__()
    self.bert = BertModel.from_pretrained(PRE_TRAINED_MODEL_NAME).to("cpu")
    # self.drop = nn.Dropout(p=0.3)
    self.out = nn.Linear(self.bert.config.hidden_size, n_classes)
  
  def forward(self, input_ids, attention_mask):
    bert_out = self.bert(
      input_ids=input_ids,
      attention_mask=attention_mask
    )
    pooled_output = bert_out[1]
    # output = self.drop(pooled_output)
    return self.out(pooled_output)

tokenizer = AutoTokenizer.from_pretrained(PRE_TRAINED_MODEL_NAME)
model = CyberbullyingClassifier(2)
model.load_state_dict(torch.load('./best_model_state.bin', map_location=torch.device('cpu')))

def classify(review_text):
    encoded_review = tokenizer(review_text, padding=MAX_LEN, truncation=True, return_tensors="pt")

    input_ids = encoded_review['input_ids'].to('cpu')
    attention_mask = encoded_review['attention_mask'].to('cpu')

    output = model(input_ids, attention_mask)
    _, prediction = torch.max(output, dim=1)

    print(f'Review text: {review_text}')
    print(f'Sentiment  : {class_names[prediction]}')

    return class_names[prediction]