app.py

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

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


def load_model():
    print("[INFO] Loading model... This may take a minute on Spaces")
    pipe = pipeline(
        task="fill-mask",
        model="BounharAbdelaziz/XLM-RoBERTa-Morocco",
        token=key,
        device=0,
        torch_dtype=torch.float16  # Use half precision
    )
    print("[INFO] Model loaded successfully!")
    return pipe

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

@spaces.GPU
@gr.cache(persist=True)  # Add persistent caching
def predict(text):
    outputs = pipe(text)
    scores= [x["score"] for x in outputs]
    tokens= [x["token_str"] for x in outputs]
    return {label: float(prob) 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...",
                rtl=True
            )
            
            # Button row
            with gr.Row():
                clear_btn = gr.Button("Clear")
                submit_btn = gr.Button("Submit", variant="primary")
            
            # Examples section with caching
            gr.Examples(
                examples=["العاصمة د <mask> هي الرباط","المغرب <mask> زوين","انا سميتي مريم، و كنسكن ف<mask> العاصمة دفلسطين"],
                inputs=input_text,
                cache_examples=True,
                preprocess=True  # Precompute examples
            )
        
        with gr.Column():
            # Output probabilities
            output_labels = gr.Label(
                label="Prediction Results",
                show_label=False,
                num_top_classes=5  # Limit to top 5 predictions
            )

    # Button actions
    submit_btn.click(
        predict,
        inputs=input_text,
        outputs=output_labels,
        show_progress=True  # Show a progress indicator
    )
    
    clear_btn.click(
        lambda: "",
        outputs=input_text
    )

# Launch the app with queue
demo.queue(concurrency_count=3)  # Allow 3 concurrent predictions
demo.launch(show_api=False)  # Disable API tab if not needed