app.py

import gradio as gr
from transformers import BertTokenizer, BertForSequenceClassification
import torch
import torch.nn.functional as F

# Load the tokenizer and model
tokenizer = BertTokenizer.from_pretrained('indobenchmark/indobert-large-p1')
model = BertForSequenceClassification.from_pretrained("hendri/nergrit")

labels = ["LABEL_0", "LABEL_1", "LABEL_2", "LABEL_3", "LABEL_4", "LABEL_5" , "LABEL_6"]

# Map these to your actual labels:
label_mapping = {
    "LABEL_0": "I-PERSON",
    "LABEL_1": "B-ORGANISATION",
    "LABEL_2": "I-ORGANISATION",
    "LABEL_3": "B-PLACE",
    "LABEL_4": "I-PLACE",
    "LABEL_5": "O",
    "LABEL_6": "B-PERSON"
}

# Define a function to process user input and return predictions
def classify_emotion(text):
    inputs = tokenizer(text, return_tensors="pt", truncation=True, padding=True, max_length=128)
    with torch.no_grad():
        outputs = model(**inputs)
        logits = outputs.logits
        probabilities = F.softmax(logits, dim=-1)
        predictions = {label_mapping[labels[i]]: round(float(prob), 4) for i, prob in enumerate(probabilities[0])}
    return predictions

# Create the Gradio interface
interface = gr.Interface(
    fn=classify_emotion,
    inputs=gr.Textbox(label="Enter Text for NER"),
    outputs=gr.Label(label="Predicted NER"),
    title="Emotion Classification",
    description="This application uses an IndoBERT model fine-tuned for NER. Enter a sentence (bahasa Indonesia) to see the predicted NER and their probabilities."
)

# Launch the Gradio interface
interface.launch()