app.py

import spaces
import gradio as gr
import numpy as np
import random
from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM, GenerationConfig
import os

from huggingface_hub import hf_hub_download
from safetensors.torch import load_file

device = "cuda" 

token=os.environ["TOKEN"]

model_id="aipicasso/emix-1-0"
scheduler = EulerAncestralDiscreteScheduler.from_pretrained(model_id,subfolder="scheduler",token=token)
pipe = StableDiffusionXLPipeline.from_pretrained(model_id, scheduler=scheduler, torch_dtype=torch.bfloat16,token=token)

negative_ti_file = hf_hub_download(repo_id="Aikimi/unaestheticXL_Negative_TI", filename="unaestheticXLv31.safetensors")
state_dict = load_file(negative_ti_file)
pipe.load_textual_inversion(state_dict["clip_g"], token="unaestheticXLv31", text_encoder=pipe.text_encoder_2, tokenizer=pipe.tokenizer_2)
pipe.load_textual_inversion(state_dict["clip_l"], token="unaestheticXLv31", text_encoder=pipe.text_encoder, tokenizer=pipe.tokenizer)

pipe = pipe.to(device)



MODEL_NAME = "p1atdev/dart-v2-moe-sft"

tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, trust_remote_code=True) # trust_remote_code is required for tokenizer
model = AutoModelForCausalLM.from_pretrained(MODEL_NAME, torch_dtype=torch.bfloat16)
model=model.to(device)

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1344

@spaces.GPU
def infer(seed, randomize_seed, width, height, guidance_scale, num_inference_steps):
    prompt = (
        f"<|bos|>"
        f"<copyright></copyright>"
        f"<character></character>"
        f"<|rating:general|><|aspect_ratio:tall|><|length:long|>"
        f"<general>1girl<|identity:none|><|input_end|>"
    )
    inputs = tokenizer(prompt, return_tensors="pt").input_ids
    with torch.no_grad():
      outputs = model.generate(
        inputs.to(device),
        do_sample=True,
        temperature=1.0,
        top_p=1.0,
        top_k=100,
        max_new_tokens=64,
        num_beams=1,
      )
    
    prompt=", ".join([tag for tag in tokenizer.batch_decode(outputs[0], skip_special_tokens=True) if tag.strip() != ""])
    negative_prompt="unaestheticXLv31, 3d, photo, realism"
    
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
        
    generator = torch.Generator().manual_seed(seed)
    
    image = pipe(
        prompt = prompt, 
        negative_prompt = negative_prompt,
        guidance_scale = guidance_scale, 
        num_inference_steps = num_inference_steps, 
        width = width, 
        height = height,
        generator = generator
    ).images[0] 
    
    return image, prompt

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"""
        # 著作権のないイラスト
        ## Anime image without copyright
        Generateボタンを押し、画像を生成してください。この画像がいくらきれいであろうと著作権は誰にもありません。この画像は時刻を入力とした自然現象によって作られたものです。美しいとは何でしょうか。
        """)
        
        with gr.Row():
            
            run_button = gr.Button("Generate", scale=0)
        
        result = gr.Image(label="Result", show_label=False)
        generated_prompt = gr.Textbox(label="Generated prompt", show_label=False, interactive=False)
        with gr.Accordion("Advanced Settings", open=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=512,
                    maximum=MAX_IMAGE_SIZE,
                    step=64,
                    value=1024,
                )
                
                height = gr.Slider(
                    label="Height",
                    minimum=512,
                    maximum=MAX_IMAGE_SIZE,
                    step=64,
                    value=1024,
                )
            
            with gr.Row():
                
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=1.0,
                    maximum=10.0,
                    step=0.1,
                    value=7.5,
                )
                
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=30,
                    step=1,
                    value=20,
                )

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

demo.queue().launch()