app.py

import gradio as gr
import torch
from transformers import pipeline
import os

#load_dotenv()
key=os.environ["HF_KEY"]


def load_model():
    pipe=pipeline(task="fill-mask",model="atlasia/xlm-roberta-large-ft-alatlas",token=key)
    return pipe

print("[INFO] load model ...")
pipe=load_model()
print("[INFO] model loaded")

# def predict(text):
#     predictions=pipe(text)
#     return predictions[0]["sequence"],predictions


def predict(text):
    
    # Get prediction
    with torch.no_grad():
        outputs = pipe(text)
        scores= [x["score"] for x in outputs]
        tokens= [x["token_str"] for x in outputs]
       # scores= [x["score"] for x in outputs]
    # Convert to percentages and create label-probability pairs
    #probs = probabilities[0].tolist()
    return {label: float(prob) * 100 for label, prob in zip(tokens, scores)}

# Create Gradio interface
with gr.Blocks() as demo:
    with gr.Row():
        with gr.Column():
            # Input text box
            input_text = gr.Textbox(
                label="Input",
                placeholder="Enter text here..."
            )
            
            # Button row
            with gr.Row():
                clear_btn = gr.Button("Clear")
                submit_btn = gr.Button("Submit", variant="primary")
            
            # Examples section
            gr.Examples(
                examples=["Hugging Face is the AI community, working together, to [MASK] the future."],
                inputs=input_text
            )
        
        with gr.Column():
            # Output label
            gr.Label("Classification")
            
            # Output probabilities
            output_labels = gr.Label(
                label="Classification Results",
                show_label=False
            )

    # Button actions
    submit_btn.click(
        predict,
        inputs=input_text,
        outputs=output_labels
    )
    
    clear_btn.click(
        lambda: "",
        outputs=input_text
    )

# Launch the app
demo.launch()