app.py

import gradio as gr

#gr.load("models/K00B404/tiny_image_gen").launch()

# GenerativeApp.py
import os
import torch
from PIL import Image
from diffusers import StableDiffusionPipeline
from transformers import AutoTokenizer

def load_model():
    device = "cuda" if torch.cuda.is_available() else "cpu"
    model_path = "K00B404/tiny_image_gen"
    pipe = StableDiffusionPipeline.from_pretrained(model_path, revision='main', torch_dtype=torch.float16)
    pipe = pipe.to(device)
    return pipe

def generate_image(prompt):
    pipe = load_model()
    output = pipe(prompt, num_inference_steps=50, height=512, width=512)
    image = output["samples"][0]
    return Image.fromarray((image.transpose(1, 2, 0) * 255).detach().cpu().numpy().astype(np.uint8))

if __name__ == "__main__":
    gradio_app = gr.Interface(generate_image, inputs="text", outputs="image")
    gradio_app.launch()