app.py

from diffusers import KandinskyV22CombinedPipeline
import gradio as gr
from accelerate import Accelerator
import torch, os, random
from transformers import pipeline
from PIL import Image

accelerator = Accelerator()
pipe = accelerator.prepare(KandinskyV22CombinedPipeline.from_pretrained("kandinsky-community/kandinsky-2-2-decoder", torch_dtype=torch.float32, use_safetensors=True, safety_checker=False))
pipe = pipe.to("cpu")
apol=[]
def plex(prompt,negative_prompt,stips,uno):
    apol=[]
    generator = torch.Generator(device="cpu").manual_seed(random.randint(1, 4876364))
    image = pipe(prompt=[prompt]*2, negative_prompt=[negative_prompt]*2,num_inference_steps=stips, prior_guidance_scale=uno, height=512, width=512, generator=generator)
    for i, igs in enumerate(image["images"]):
        apol.append(igs)
    return apol

iface = gr.Interface(fn=plex,inputs=[gr.Textbox(label="prompt"),gr.Textbox(label="negative prompt", value="low quality, bad quality"), gr.Slider(label="inference_steps",minimum=1,step=1,maximum=10,value=10),gr.Slider(label="prior_guidance_scale",minimum=0.1,step=0.1,maximum=1.0,value=0.5)],outputs=gr.Gallery(columns=2), title="Txt2Img_KndskyV22_Cmbnd by JoPmt", description="Running on CPU, very slow!")
iface.queue(max_size=1,api_open=False)
iface.launch(max_threads=1)