abc123desygn/Analog-Diffusion

1from diffusers import StableDiffusionPipeline, StableDiffusionImg2ImgPipeline, DPMSolverMultistepScheduler2import gradio as gr3import torch4from PIL import Image5 6model_id = 'wavymulder/Analog-Diffusion'7prefix = ''8     9scheduler = DPMSolverMultistepScheduler.from_pretrained(model_id, subfolder="scheduler")10 11pipe = StableDiffusionPipeline.from_pretrained(12  model_id,13  torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,14  scheduler=scheduler)15 16pipe_i2i = StableDiffusionImg2ImgPipeline.from_pretrained(17  model_id,18  torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,19  scheduler=scheduler)20 21if torch.cuda.is_available():22  pipe = pipe.to("cuda")23  pipe_i2i = pipe_i2i.to("cuda")24 25def error_str(error, title="Error"):26    return f"""#### {title}27            {error}"""  if error else ""28 29def inference(prompt, guidance, steps, width=512, height=512, seed=0, img=None, strength=0.5, neg_prompt="", auto_prefix=False):30 31  generator = torch.Generator('cuda').manual_seed(seed) if seed != 0 else None32  prompt = f"{prefix} {prompt}" if auto_prefix else prompt33 34  try:35    if img is not None:36      return img_to_img(prompt, neg_prompt, img, strength, guidance, steps, width, height, generator), None37    else:38      return txt_to_img(prompt, neg_prompt, guidance, steps, width, height, generator), None39  except Exception as e:40    return None, error_str(e)41 42def txt_to_img(prompt, neg_prompt, guidance, steps, width, height, generator):43 44    result = pipe(45      prompt,46      negative_prompt = neg_prompt,47      num_inference_steps = int(steps),48      guidance_scale = guidance,49      width = width,50      height = height,51      generator = generator)52    53    return result.images[0]54 55def img_to_img(prompt, neg_prompt, img, strength, guidance, steps, width, height, generator):56 57    ratio = min(height / img.height, width / img.width)58    img = img.resize((int(img.width * ratio), int(img.height * ratio)), Image.LANCZOS)59    result = pipe_i2i(60        prompt,61        negative_prompt = neg_prompt,62        init_image = img,63        num_inference_steps = int(steps),64        strength = strength,65        guidance_scale = guidance,66        width = width,67        height = height,68        generator = generator)69        70    return result.images[0]71 72css = """.main-div div{display:inline-flex;align-items:center;gap:.8rem;font-size:1.75rem}.main-div div h1{font-weight:900;margin-bottom:7px}.main-div p{margin-bottom:10px;font-size:94%}a{text-decoration:underline}.tabs{margin-top:0;margin-bottom:0}#gallery{min-height:20rem}73"""74with gr.Blocks(css=css) as demo:75    gr.HTML(76        f"""77            <div class="main-div">78              <div>79                <h1>Analog Diffusion</h1>80              </div>81              <p>82               Demo for <a href="https://huggingface.co/wavymulder/Analog-Diffusion">Analog Diffusion</a> Stable Diffusion model.<br>83               {"Add the following tokens to your prompts for the model to work properly: <b>prefix</b>" if prefix else ""}84              </p>85              Running on {"<b>GPU 🔥</b>" if torch.cuda.is_available() else f"<b>CPU 🥶</b>. For faster inference it is recommended to <b>upgrade to GPU in <a href='https://huggingface.co/spaces/akhaliq/Analog-Diffusion/settings'>Settings</a></b>"} after duplicating the space<br><br>86              <a style="display:inline-block" href="https://huggingface.co/spaces/akhaliq/Analog-Diffusion?duplicate=true"><img src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>87            </div>88        """89    )90    with gr.Row():91        92        with gr.Column(scale=55):93          with gr.Group():94              with gr.Row():95                prompt = gr.Textbox(label="Prompt", show_label=False, max_lines=2,placeholder=f"{prefix} [your prompt]").style(container=False)96                generate = gr.Button(value="Generate").style(rounded=(False, True, True, False))97 98              image_out = gr.Image(height=512)99          error_output = gr.Markdown()100 101        with gr.Column(scale=45):102          with gr.Tab("Options"):103            with gr.Group():104              neg_prompt = gr.Textbox(label="Negative prompt", placeholder="What to exclude from the image")105              auto_prefix = gr.Checkbox(label="Prefix styling tokens automatically ()", value=prefix, visible=prefix)106 107              with gr.Row():108                guidance = gr.Slider(label="Guidance scale", value=7.5, maximum=15)109                steps = gr.Slider(label="Steps", value=25, minimum=2, maximum=75, step=1)110 111              with gr.Row():112                width = gr.Slider(label="Width", value=512, minimum=64, maximum=1024, step=8)113                height = gr.Slider(label="Height", value=512, minimum=64, maximum=1024, step=8)114 115              seed = gr.Slider(0, 2147483647, label='Seed (0 = random)', value=0, step=1)116 117          with gr.Tab("Image to image"):118              with gr.Group():119                image = gr.Image(label="Image", height=256, tool="editor", type="pil")120                strength = gr.Slider(label="Transformation strength", minimum=0, maximum=1, step=0.01, value=0.5)121 122    auto_prefix.change(lambda x: gr.update(placeholder=f"{prefix} [your prompt]" if x else "[Your prompt]"), inputs=auto_prefix, outputs=prompt, queue=False)123 124    inputs = [prompt, guidance, steps, width, height, seed, image, strength, neg_prompt, auto_prefix]125    outputs = [image_out, error_output]126    prompt.submit(inference, inputs=inputs, outputs=outputs)127    generate.click(inference, inputs=inputs, outputs=outputs)128 129    gr.HTML("""130    <div style="border-top: 1px solid #303030;">131      <br>132      <p>This space was created using <a href="https://huggingface.co/spaces/anzorq/sd-space-creator">SD Space Creator</a>.</p>133    </div>134    """)135 136demo.queue(concurrency_count=1)137demo.launch()138