app.py

import gradio as gr
from transformers import pipeline, set_seed
import os

# Set environment variables for CPU optimization
os.environ["TOKENIZERS_PARALLELISM"] = "false"

# Load model once at startup
generator = pipeline(
    "text-generation",
    model="openai-community/openai-gpt",
    device=-1  # Auto-detect CPU
)

def generate_text(prompt, max_length=30, num_return_sequences=3, seed=42):
    """Generate text with configurable parameters and seed"""
    set_seed(seed)
    results = generator(
        prompt,
        max_length=max_length,
        num_return_sequences=num_return_sequences,
        truncation=True,
        pad_token_id=generator.tokenizer.eos_token_id
    )
    # Format results with numbering
    return "\n\n".join([f"{i+1}. {r['generated_text']}" for i, r in enumerate(results)])

# Define UI
with gr.Blocks(theme="soft", title="GPT-1 Text Generator") as demo:
    gr.Markdown("""
    # 📜 GPT-1 Text Generation Demo
    Run on free HuggingFace CPU • Model: [openai-community/openai-gpt](https://huggingface.co/openai-community/openai-gpt )
    
    *Note: This is the original 2018 GPT model with known limitations and biases. For production use, consider newer models.*
    """)
    
    with gr.Row():
        with gr.Column():
            prompt = gr.Textbox(
                label="Input Prompt", 
                placeholder="Enter your prompt here...",
                lines=4
            )
            seed = gr.Number(value=42, label="Random Seed")
        
        with gr.Column():
            max_length = gr.Slider(10, 100, value=30, step=5, label="Max Output Length")
            num_return_sequences = gr.Slider(1, 5, value=3, step=1, label="Number of Variants")
    
    generate_btn = gr.Button("✨ Generate Text")
    output = gr.Textbox(label="Generated Results", lines=10)
    
    generate_btn.click(
        fn=generate_text,
        inputs=[prompt, max_length, num_return_sequences, seed],
        outputs=output
    )

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