app.py

import gradio as gr
import torch
from torch import autocast
from diffusers import StableDiffusionPipeline

model_id = "CompVis/stable-diffusion-v1-4"
pipe = StableDiffusionPipeline.from_pretrained(model_id, use_auth_token='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', torch_dtype=torch.float32, low_cpu_mem_usage=True)
has_cuda = torch.cuda.is_available()
device = torch.device('cpu' if not has_cuda else 'cuda')    
pipe = pipe.to(device)

def convert(prompt):
    samples = 4
    generator = torch.Generator(device=device)
    torch.cuda.empty_cache()
    with autocast("cuda"):
        images_list = pipe(
            [prompt] * samples,
            height=256, width=384,
            num_inference_steps=50,
        )
        images = []
        for i, image in enumerate(images_list["sample"]):
            images.append(image)
    return images


gr.Interface(convert,
             inputs = [gr.inputs.Textbox(label="Enter text")],    
             outputs = [gr.outputs.Image(label="Generated Image")], 
             title="Text to Image Generation").launch()