#!/usr/bin/env python3
import torch
from diffusers import AutoPipelineForText2Image
from diffusers.pipelines.wuerstchen import DEFAULT_STAGE_C_TIMESTEPS

pipe = AutoPipelineForText2Image.from_pretrained("warp-ai/wuerstchen", torch_dtype=torch.float16).to("cuda")

caption = "Anthropomorphic cat dressed as a fire fighter"
images = pipe(
    caption, 
    width=1024,
    height=1536,
    prior_timesteps=DEFAULT_STAGE_C_TIMESTEPS,
    prior_guidance_scale=4.0,
    num_images_per_prompt=2,
).images

import ipdb; ipdb.set_trace()