app.py

import gradio as gr
import torch
from torch import nn
from transformers import BertTokenizer, BertModel

# Define the BertClassifier class
class BertClassifier(nn.Module):
    def __init__(self, bert: BertModel, num_classes: int):
        super().__init__()
        self.bert = bert
        self.classifier = nn.Linear(bert.config.hidden_size, num_classes)
        self.criterion = nn.BCELoss()

    def forward(self, input_ids, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, labels=None):
        outputs = self.bert(
            input_ids=input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            head_mask=head_mask
        )
        cls_output = outputs.pooler_output
        cls_output = self.classifier(cls_output)
        cls_output = torch.sigmoid(cls_output)

        loss = 0
        if labels is not None:
            loss = self.criterion(cls_output, labels)
        return loss, cls_output

# Load the tokenizer and model
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
bert_model = BertModel.from_pretrained('bert-base-uncased')
model = BertClassifier(bert_model, num_classes=7)

# Load the model weights from the .pkl file
model.load_state_dict(torch.load('bert_classifier_mltc.pkl', map_location=torch.device('cpu')))
model.eval()

# Define prediction function
def predict(text):
    tokens = tokenizer.encode(text, add_special_tokens=True, max_length=512, truncation=True)
    input_ids = torch.tensor([tokens])
    mask = (input_ids != tokenizer.pad_token_id).float()
    
    with torch.no_grad():
        _, outputs = model(input_ids, attention_mask=mask)
    
    # Assuming outputs[0] contains the probability scores for each class
    confidence_scores = outputs[0].tolist()
    
    # Convert to a dictionary mapping labels to confidence scores
    labels = ['toxic', 'severe_toxic', 'obscene', 'threat', 'insult', 'identity_hate', 'clean']
    prediction = {label: score for label, score in zip(labels, confidence_scores)}
    
    return prediction

# Add example texts
examples = [
    ["This is a positive and friendly comment."],
    ["I'm not happy with your service."],
    ["You are a great person!"],
    ["How can one defame someone who thinks the Fort Hood Shooting is justified by God's law?"],
    ["A Bisexual, like a homosexual or a heterosexual, is not defined by sexual activity. (Much like a 15 year old boy who is attracted to a girl sexually but has never had sex is still straight). A person who is actually sexually attracted/aroused by the same sex as well as the opposite sex is bisexual."],
    ["I'LL SAY NIGGER IF I WANT TO BRO"],
    ["Do you need a verifiable source for that statement?  Check your own user page you fat, miserable balding fuck."],
    ["Fuck you bitch! I will go to your house and kill you with my gun as well! Prepare to die!"],
    ["you're hot, i will rape you ;)"],
    ["WOULDN'T BE THE FIRST TIME BITCH. FUCK YOU I'LL FIND OUT WHERE YOU LIVE, SODOMIZE YOUR WIFE AND THEN BURN YOUR HOUSE DOWN. FUCK YOU YOU FUCKING QUEER."],
    ["What a motherfucking piece of crap those fuckheads for blocking us!"],
    ["Get a life you animal fucker. Shut up you fucking nigger. Fuck off and shit your pants full of all the shit you can fill them with. 144.131.176.126"],
    ["HOPE YOUR HEAD GETS CUT OFF AND SOMEONE WIPS THERE ASS WITH IT AND THEN STABS YOU IN YOUR HEART"],
    ["you people are pretty overzealous with this whole free thing. get a fucking life, you fucking niggers !!!23 16!!!"],
    ["Stupid peace of shit stop deleting my stuff asshole go die and fall in a hole go to hell!"],
    ["Bye! Don't look, come or think of comming back! Tosser."]
]
# Create Gradio interface
iface = gr.Interface(
    fn=predict,
    inputs=gr.Textbox(lines=10, placeholder="Enter a comment here..."),
    outputs=gr.Label(num_top_classes=7),
    examples=examples,
    title="Toxic Comment Classification",
    description="Classify comments into toxic and non-toxic categories using BERT and GNN model.",
)

iface.launch()