app.py

# 1. spacesを最初にインポート
import spaces
# 2. その後で他のGPU関連のライブラリをインポート
import torch
import transformers
import gradio as gr
import numpy as np
import random
#from diffusers import DiffusionPipeline
from diffusers import StableDiffusionXLPipeline, TCDScheduler
from huggingface_hub import hf_hub_download
from peft import LoraConfig, get_peft_model

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1216

base_model_id = "Laxhar/noobai-XL-1.0"
repo_name = "ByteDance/Hyper-SD"
# Take 2-steps lora as an example
ckpt_name = "Hyper-SDXL-8steps-lora.safetensors"

# Load model.
#pipe = DiffusionPipeline.from_pretrained("stabilityai/sdxl-turbo", torch_dtype=torch.float16, variant="fp16", use_safetensors=True)
pipe = StableDiffusionXLPipeline.from_pretrained(
    base_model_id, 
    torch_dtype=torch.float16, 
    use_safetensors=True,
    custom_pipeline="lpw_stable_diffusion_xl",
    add_watermarker=False
)
pipe.to('cuda')

#pipe = DiffusionPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16, variant="fp16").to("cuda")
pipe.load_lora_weights(hf_hub_download(repo_name, ckpt_name))
pipe.fuse_lora()
# Ensure ddim scheduler timestep spacing set as trailing !!!
pipe.scheduler = TCDScheduler.from_config(pipe.scheduler.config)
# lower eta results in more detail

pipe.load_lora_weights(repo_name)
pipe.fuse_lora()


prompt = "1girl, solo, upper body, v, smile, looking at viewer, outdoors, night, masterpiece, best quality, very aesthetic, absurdres"
negative_prompt = "(worst quality),(low quality),lowres,(bad anatomy),(deformed anatomy),(deformed fingers),(blurry),(extra finger),(extra arms), (extra legs),(monochrome:1.4),(grayscale:1.4),((watermark)),(overweight female:1.6),((pointy ears)),mascot,stuffed human, stuffed animal,chibi,english text, chinese text, korean text"

@spaces.GPU
def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps):

    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
        
    generator = torch.Generator().manual_seed(seed)
    
    image = pipe(
        prompt = prompt+", masterpiece, best quality, very aesthetic, absurdres", 
        negative_prompt = negative_prompt,
        guidance_scale = guidance_scale, 
        num_inference_steps = num_inference_steps, 
        width = width, 
        height = height,
        generator = generator
    ).images[0] 
    
    return image

css="""
#col-container {
    margin: 0 auto;
    max-width: 520px;
}
"""

with gr.Blocks(css=css) as demo:
    
    with gr.Column(elem_id="col-container"):
        gr.Markdown(f"""
        # Text-to-Image Demo
        using [noobai XL 1.0](https://huggingface.co/Laxhar/noobai-XL-1.0)
        """)
        
        with gr.Row():
            
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )
            
            run_button = gr.Button("Run", scale=0)
        
        result = gr.Image(label="Result", show_label=False)

        with gr.Accordion("Advanced Settings", open=False):
            
            negative_prompt = gr.Text(
                label="Negative prompt",
                max_lines=1,
                placeholder="Enter a negative prompt",
                visible=False,
            )
            
            seed = gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=0,
            )
            
            randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
            
            with gr.Row():
                
                width = gr.Slider(
                    label="Width",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=832,
                )
                
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=1216,
                )
            
            with gr.Row():
                
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=20.0,
                    step=0.1,
                    value=7,
                )
                
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=28,
                    step=1,
                    value=28,
                )

    run_button.click(
        fn = infer,
        inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps],
        outputs = [result]
    )

demo.queue().launch()