app.py

import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch
from PIL import Image

# Load tokenizer and model
tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neo-2.7B")
model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-neo-2.7B")

def generate_image(text):
    # Tokenize input text
    input_ids = tokenizer.encode(text, return_tensors="pt")

    # Generate image conditioned on input text
    output = model.generate(input_ids, do_sample=True, max_length=128, num_return_sequences=1)

    # Decode generated image tokens to get image
    image_bytes = tokenizer.decode(output[0], skip_special_tokens=True)

    # Convert image bytes to PIL image
    image = Image.open(image_bytes)

    return image

# Create Gradio interface
iface = gr.Interface(
    fn=generate_image,
    inputs=gr.inputs.Textbox(lines=3, label="Input Text"),
    outputs="image",
    title="Text-to-Image Generator",
    description="Generate images from text using Hugging Face's GPT-Neo model.",
    theme="huggingface"
)

# Launch Gradio interface
iface.launch()