app.py

import gradio as gr
from transformers import pipeline

# Load your fine-tuned model from the Hub
MODEL_ID = "tasal9/ZamAI-mT5-Pashto"
generator = pipeline(
    "text2text-generation", 
    model=MODEL_ID, 
    tokenizer=MODEL_ID,
    device=-1  # CPU only
)

# Prompt template

def generate_prompt(instruction, input_text=""):
    if input_text:
        return (
            f"Below is an instruction that describes a task, paired with an input that provides further context. "
            f"Write a response that appropriately completes the request.\n\n"
            f"### Instruction:\n{instruction}\n\n"
            f"### Input:\n{input_text}\n\n"
            f"### Response:"
        )
    else:
        return (
            f"Below is an instruction that describes a task. Write a response that appropriately completes the request.\n\n"
            f"### Instruction:\n{instruction}\n\n"
            f"### Response:"
        )

# Inference function
def predict(instruction, input_text):
    prompt = generate_prompt(instruction, input_text)
    outputs = generator(
        prompt,
        max_length=256,
        num_beams=5,
        early_stopping=True
    )
    return outputs[0]["generated_text"]

# Gradio interface
iface = gr.Interface(
    fn=predict,
    inputs=[
        gr.inputs.Textbox(lines=2, placeholder="Enter instruction here...", label="Instruction"),
        gr.inputs.Textbox(lines=2, placeholder="Enter optional input here...", label="Input")
    ],
    outputs=gr.outputs.Textbox(label="Response"),
    title="ZamAI mT5 Pashto Demo",
    description="A zero-GPU Gradio demo for the ZamAI-mT5-Pashto model."
)

if __name__ == "__main__":
    iface.launch()