app.py

import streamlit as st
from diffusers import DiffusionPipeline, DPMSolverMultistepScheduler
import torch

st.title("Image Generation 2")

@st.cache_resource
def load_pipeline():
    pipe = DiffusionPipeline.from_pretrained(
        "stable-diffusion-v1-5/stable-diffusion-v1-5",
        torch_dtype=torch.float16,
    )
    pipe.to("cuda" if torch.cuda.is_available() else "cpu")

    pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipeline.scheduler.config)

    pipe.enable_attention_slicing()
  
    return pipe

pipeline = load_pipeline()

prompt = st.text_input("Enter a prompt to generate an image:", value="pipeline under the sea")

if st.button("Generate Image"):
    with st.spinner("Generating image..."):
        image = pipeline(prompt).images[0]
        st.image(image, caption="Generated Image", use_column_width=True)