remove img2img app

app-img2img.py

@@ -1,271 +0,0 @@
-import asyncio
-import json
-import logging
-import traceback
-from pydantic import BaseModel
-
-from fastapi import FastAPI, WebSocket, HTTPException, WebSocketDisconnect
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi.responses import (
-    StreamingResponse,
-    JSONResponse,
-    HTMLResponse,
-    FileResponse,
-)
-
-from diffusers import AutoPipelineForImage2Image, AutoencoderTiny
-from compel import Compel
-import torch
-
-try:
-    import intel_extension_for_pytorch as ipex
-except:
-    pass
-from PIL import Image
-import numpy as np
-import gradio as gr
-import io
-import uuid
-import os
-import time
-import psutil
-
-MAX_QUEUE_SIZE = int(os.environ.get("MAX_QUEUE_SIZE", 0))
-TIMEOUT = float(os.environ.get("TIMEOUT", 0))
-SAFETY_CHECKER = os.environ.get("SAFETY_CHECKER", None)
-TORCH_COMPILE = os.environ.get("TORCH_COMPILE", None)
-
-WIDTH = 512
-HEIGHT = 512
-# disable tiny autoencoder for better quality speed tradeoff
-USE_TINY_AUTOENCODER = True
-
-# check if MPS is available OSX only M1/M2/M3 chips
-mps_available = hasattr(torch.backends, "mps") and torch.backends.mps.is_available()
-xpu_available = hasattr(torch, "xpu") and torch.xpu.is_available()
-device = torch.device(
-    "cuda" if torch.cuda.is_available() else "xpu" if xpu_available else "cpu"
-)
-torch_device = device
-
-# change to torch.float16 to save GPU memory
-torch_dtype = torch.float32
-
-print(f"TIMEOUT: {TIMEOUT}")
-print(f"SAFETY_CHECKER: {SAFETY_CHECKER}")
-print(f"MAX_QUEUE_SIZE: {MAX_QUEUE_SIZE}")
-print(f"device: {device}")
-
-if mps_available:
-    device = torch.device("mps")
-    torch_device = "cpu"
-    torch_dtype = torch.float32
-
-if SAFETY_CHECKER == "True":
-    pipe = AutoPipelineForImage2Image.from_pretrained(
-        "SimianLuo/LCM_Dreamshaper_v7",
-    )
-else:
-    pipe = AutoPipelineForImage2Image.from_pretrained(
-        "SimianLuo/LCM_Dreamshaper_v7",
-        safety_checker=None,
-    )
-
-if USE_TINY_AUTOENCODER:
-    pipe.vae = AutoencoderTiny.from_pretrained(
-        "madebyollin/taesd", torch_dtype=torch_dtype, use_safetensors=True
-    )
-pipe.set_progress_bar_config(disable=True)
-pipe.to(device=torch_device, dtype=torch_dtype).to(device)
-pipe.unet.to(memory_format=torch.channels_last)
-
-if psutil.virtual_memory().total < 64 * 1024**3:
-    pipe.enable_attention_slicing()
-
-if TORCH_COMPILE:
-    pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True)
-    pipe.vae = torch.compile(pipe.vae, mode="reduce-overhead", fullgraph=True)
-
-    pipe(prompt="warmup", image=[Image.new("RGB", (512, 512))])
-
-compel_proc = Compel(
-    tokenizer=pipe.tokenizer,
-    text_encoder=pipe.text_encoder,
-    truncate_long_prompts=False,
-)
-user_queue_map = {}
-
-
-class InputParams(BaseModel):
-    seed: int = 2159232
-    prompt: str
-    guidance_scale: float = 8.0
-    strength: float = 0.5
-    steps: int = 4
-    lcm_steps: int = 50
-    width: int = WIDTH
-    height: int = HEIGHT
-
-
-def predict(
-    input_image: Image.Image, params: InputParams, prompt_embeds: torch.Tensor = None
-):
-    generator = torch.manual_seed(params.seed)
-    results = pipe(
-        prompt_embeds=prompt_embeds,
-        generator=generator,
-        image=input_image,
-        strength=params.strength,
-        num_inference_steps=params.steps,
-        guidance_scale=params.guidance_scale,
-        width=params.width,
-        height=params.height,
-        original_inference_steps=params.lcm_steps,
-        output_type="pil",
-    )
-    nsfw_content_detected = (
-        results.nsfw_content_detected[0]
-        if "nsfw_content_detected" in results
-        else False
-    )
-    if nsfw_content_detected:
-        return None
-    return results.images[0]
-
-
-app = FastAPI()
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],
-    allow_credentials=True,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-
-@app.websocket("/ws")
-async def websocket_endpoint(websocket: WebSocket):
-    await websocket.accept()
-    if MAX_QUEUE_SIZE > 0 and len(user_queue_map) >= MAX_QUEUE_SIZE:
-        print("Server is full")
-        await websocket.send_json({"status": "error", "message": "Server is full"})
-        await websocket.close()
-        return
-
-    try:
-        uid = str(uuid.uuid4())
-        print(f"New user connected: {uid}")
-        await websocket.send_json(
-            {"status": "success", "message": "Connected", "userId": uid}
-        )
-        user_queue_map[uid] = {"queue": asyncio.Queue()}
-        await websocket.send_json(
-            {"status": "start", "message": "Start Streaming", "userId": uid}
-        )
-        await handle_websocket_data(websocket, uid)
-    except WebSocketDisconnect as e:
-        logging.error(f"WebSocket Error: {e}, {uid}")
-        traceback.print_exc()
-    finally:
-        print(f"User disconnected: {uid}")
-        queue_value = user_queue_map.pop(uid, None)
-        queue = queue_value.get("queue", None)
-        if queue:
-            while not queue.empty():
-                try:
-                    queue.get_nowait()
-                except asyncio.QueueEmpty:
-                    continue
-
-
-@app.get("/queue_size")
-async def get_queue_size():
-    queue_size = len(user_queue_map)
-    return JSONResponse({"queue_size": queue_size})
-
-
-@app.get("/stream/{user_id}")
-async def stream(user_id: uuid.UUID):
-    uid = str(user_id)
-    try:
-        user_queue = user_queue_map[uid]
-        queue = user_queue["queue"]
-
-        async def generate():
-            last_prompt: str = None
-            prompt_embeds: torch.Tensor = None
-            while True:
-                data = await queue.get()
-                input_image = data["image"]
-                params = data["params"]
-                if input_image is None:
-                    continue
-                # avoid recalculate prompt embeds
-                if last_prompt != params.prompt:
-                    print("new prompt")
-                    prompt_embeds = compel_proc(params.prompt)
-                    last_prompt = params.prompt
-
-                image = predict(
-                    input_image,
-                    params,
-                    prompt_embeds,
-                )
-                if image is None:
-                    continue
-                frame_data = io.BytesIO()
-                image.save(frame_data, format="JPEG")
-                frame_data = frame_data.getvalue()
-                if frame_data is not None and len(frame_data) > 0:
-                    yield b"--frame\r\nContent-Type: image/jpeg\r\n\r\n" + frame_data + b"\r\n"
-
-                await asyncio.sleep(1.0 / 120.0)
-
-        return StreamingResponse(
-            generate(), media_type="multipart/x-mixed-replace;boundary=frame"
-        )
-    except Exception as e:
-        logging.error(f"Streaming Error: {e}, {user_queue_map}")
-        traceback.print_exc()
-        return HTTPException(status_code=404, detail="User not found")
-
-
-async def handle_websocket_data(websocket: WebSocket, user_id: uuid.UUID):
-    uid = str(user_id)
-    user_queue = user_queue_map[uid]
-    queue = user_queue["queue"]
-    if not queue:
-        return HTTPException(status_code=404, detail="User not found")
-    last_time = time.time()
-    try:
-        while True:
-            data = await websocket.receive_bytes()
-            params = await websocket.receive_json()
-            params = InputParams(**params)
-            pil_image = Image.open(io.BytesIO(data))
-
-            while not queue.empty():
-                try:
-                    queue.get_nowait()
-                except asyncio.QueueEmpty:
-                    continue
-            await queue.put({"image": pil_image, "params": params})
-            if TIMEOUT > 0 and time.time() - last_time > TIMEOUT:
-                await websocket.send_json(
-                    {
-                        "status": "timeout",
-                        "message": "Your session has ended",
-                        "userId": uid,
-                    }
-                )
-                await websocket.close()
-                return
-
-    except Exception as e:
-        logging.error(f"Error: {e}")
-        traceback.print_exc()
-
-
-@app.get("/", response_class=HTMLResponse)
-async def root():
-    return FileResponse("./static/img2img.html")

static/img2img.html

@@ -1,383 +0,0 @@
-<!doctype html>
-<html>
-
-<head>
-    <meta charset="UTF-8">
-    <title>Real-Time Latent Consistency Model</title>
-    <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <script
-        src="https://cdnjs.cloudflare.com/ajax/libs/iframe-resizer/4.3.1/iframeResizer.contentWindow.min.js"></script>
-    <script src="https://cdn.jsdelivr.net/npm/[email protected]/piexif.min.js"></script>
-    <script src="https://cdn.tailwindcss.com"></script>
-    <style type="text/tailwindcss">
-        .button {
-          @apply bg-gray-700 hover:bg-gray-800 text-white font-normal p-2 rounded disabled:bg-gray-300 dark:disabled:bg-gray-700 disabled:cursor-not-allowed dark:disabled:text-black
-        }
-    </style>
-    <script type="module">
-        const getValue = (id) => {
-            const el = document.querySelector(`${id}`)
-            if (el.type === "checkbox")
-                return el.checked;
-            return el.value;
-        }
-
-        const startBtn = document.querySelector("#start");
-        const stopBtn = document.querySelector("#stop");
-        const videoEl = document.querySelector("#webcam");
-        const imageEl = document.querySelector("#player");
-        const queueSizeEl = document.querySelector("#queue_size");
-        const errorEl = document.querySelector("#error");
-        const snapBtn = document.querySelector("#snap");
-        const webcamsEl = document.querySelector("#webcams");
-
-        function LCMLive(webcamVideo, liveImage) {
-            let websocket;
-
-            async function start() {
-                return new Promise((resolve, reject) => {
-                    const websocketURL = `${window.location.protocol === "https:" ? "wss" : "ws"
-                        }:${window.location.host}/ws`;
-
-                    const socket = new WebSocket(websocketURL);
-                    socket.onopen = () => {
-                        console.log("Connected to websocket");
-                    };
-                    socket.onclose = () => {
-                        console.log("Disconnected from websocket");
-                        stop();
-                        resolve({ "status": "disconnected" });
-                    };
-                    socket.onerror = (err) => {
-                        console.error(err);
-                        reject(err);
-                    };
-                    socket.onmessage = (event) => {
-                        const data = JSON.parse(event.data);
-                        switch (data.status) {
-                            case "success":
-                                break;
-                            case "start":
-                                const userId = data.userId;
-                                initVideoStream(userId);
-                                break;
-                            case "timeout":
-                                stop();
-                                resolve({ "status": "timeout" });
-                            case "error":
-                                stop();
-                                reject(data.message);
-
-                        }
-                    };
-                    websocket = socket;
-                })
-            }
-            function switchCamera() {
-                const constraints = {
-                    audio: false,
-                    video: { width: 1024, height: 1024, deviceId: mediaDevices[webcamsEl.value].deviceId }
-                };
-                navigator.mediaDevices
-                    .getUserMedia(constraints)
-                    .then((mediaStream) => {
-                        webcamVideo.removeEventListener("timeupdate", videoTimeUpdateHandler);
-                        webcamVideo.srcObject = mediaStream;
-                        webcamVideo.onloadedmetadata = () => {
-                            webcamVideo.play();
-                            webcamVideo.addEventListener("timeupdate", videoTimeUpdateHandler);
-                        };
-                    })
-                    .catch((err) => {
-                        console.error(`${err.name}: ${err.message}`);
-                    });
-            }
-
-            async function videoTimeUpdateHandler() {
-                const dimension = getValue("input[name=dimension]:checked");
-                const [WIDTH, HEIGHT] = JSON.parse(dimension);
-
-                const canvas = new OffscreenCanvas(WIDTH, HEIGHT);
-                const videoW = webcamVideo.videoWidth;
-                const videoH = webcamVideo.videoHeight;
-                const aspectRatio = WIDTH / HEIGHT;
-
-                const ctx = canvas.getContext("2d");
-                ctx.drawImage(webcamVideo, videoW / 2 - videoH * aspectRatio / 2, 0, videoH * aspectRatio, videoH, 0, 0, WIDTH, HEIGHT)
-                const blob = await canvas.convertToBlob({ type: "image/jpeg", quality: 1 });
-                websocket.send(blob);
-                websocket.send(JSON.stringify({
-                    "seed": getValue("#seed"),
-                    "prompt": getValue("#prompt"),
-                    "guidance_scale": getValue("#guidance-scale"),
-                    "strength": getValue("#strength"),
-                    "steps": getValue("#steps"),
-                    "lcm_steps": getValue("#lcm_steps"),
-                    "width": WIDTH,
-                    "height": HEIGHT,
-                }));
-            }
-            let mediaDevices = [];
-            async function initVideoStream(userId) {
-                liveImage.src = `/stream/${userId}`;
-                await navigator.mediaDevices.enumerateDevices()
-                    .then(devices => {
-                        const cameras = devices.filter(device => device.kind === 'videoinput');
-                        mediaDevices = cameras;
-                        webcamsEl.innerHTML = "";
-                        cameras.forEach((camera, index) => {
-                            const option = document.createElement("option");
-                            option.value = index;
-                            option.innerText = camera.label;
-                            webcamsEl.appendChild(option);
-                            option.selected = index === 0;
-                        });
-                        webcamsEl.addEventListener("change", switchCamera);
-                    })
-                    .catch(err => {
-                        console.error(err);
-                    });
-                const constraints = {
-                    audio: false,
-                    video: { width: 1024, height: 1024, deviceId: mediaDevices[0].deviceId }
-                };
-                navigator.mediaDevices
-                    .getUserMedia(constraints)
-                    .then((mediaStream) => {
-                        webcamVideo.srcObject = mediaStream;
-                        webcamVideo.onloadedmetadata = () => {
-                            webcamVideo.play();
-                            webcamVideo.addEventListener("timeupdate", videoTimeUpdateHandler);
-                        };
-                    })
-                    .catch((err) => {
-                        console.error(`${err.name}: ${err.message}`);
-                    });
-            }
-
-            async function stop() {
-                websocket.close();
-                navigator.mediaDevices.getUserMedia({ video: true }).then((mediaStream) => {
-                    mediaStream.getTracks().forEach((track) => track.stop());
-                });
-                webcamVideo.removeEventListener("timeupdate", videoTimeUpdateHandler);
-                webcamsEl.removeEventListener("change", switchCamera);
-                webcamVideo.srcObject = null;
-            }
-            return {
-                start,
-                stop
-            }
-        }
-        function toggleMessage(type) {
-            errorEl.hidden = false;
-            errorEl.scrollIntoView();
-            switch (type) {
-                case "error":
-                    errorEl.innerText = "To many users are using the same GPU, please try again later.";
-                    errorEl.classList.toggle("bg-red-300", "text-red-900");
-                    break;
-                case "success":
-                    errorEl.innerText = "Your session has ended, please start a new one.";
-                    errorEl.classList.toggle("bg-green-300", "text-green-900");
-                    break;
-            }
-            setTimeout(() => {
-                errorEl.hidden = true;
-            }, 2000);
-        }
-        function snapImage() {
-            try {
-                const zeroth = {};
-                const exif = {};
-                const gps = {};
-                zeroth[piexif.ImageIFD.Make] = "LCM Image-to-Image";
-                zeroth[piexif.ImageIFD.ImageDescription] = `prompt: ${getValue("#prompt")} | seed: ${getValue("#seed")} | guidance_scale: ${getValue("#guidance-scale")} | strength: ${getValue("#strength")} | lcm_steps: ${getValue("#lcm_steps")} | steps: ${getValue("#steps")}`;
-                zeroth[piexif.ImageIFD.Software] = "https://github.com/radames/Real-Time-Latent-Consistency-Model";
-
-                exif[piexif.ExifIFD.DateTimeOriginal] = new Date().toISOString();
-
-                const exifObj = { "0th": zeroth, "Exif": exif, "GPS": gps };
-                const exifBytes = piexif.dump(exifObj);
-
-                const canvas = document.createElement("canvas");
-                canvas.width = imageEl.naturalWidth;
-                canvas.height = imageEl.naturalHeight;
-                const ctx = canvas.getContext("2d");
-                ctx.drawImage(imageEl, 0, 0);
-                const dataURL = canvas.toDataURL("image/jpeg");
-                const withExif = piexif.insert(exifBytes, dataURL);
-
-                const a = document.createElement("a");
-                a.href = withExif;
-                a.download = `lcm_txt_2_img${Date.now()}.png`;
-                a.click();
-            } catch (err) {
-                console.log(err);
-            }
-        }
-
-
-        const lcmLive = LCMLive(videoEl, imageEl);
-        startBtn.addEventListener("click", async () => {
-            try {
-                startBtn.disabled = true;
-                snapBtn.disabled = false;
-                const res = await lcmLive.start();
-                startBtn.disabled = false;
-                if (res.status === "timeout")
-                    toggleMessage("success")
-            } catch (err) {
-                console.log(err);
-                toggleMessage("error")
-                startBtn.disabled = false;
-            }
-        });
-        stopBtn.addEventListener("click", () => {
-            lcmLive.stop();
-        });
-        window.addEventListener("beforeunload", () => {
-            lcmLive.stop();
-        });
-        snapBtn.addEventListener("click", snapImage);
-        setInterval(() =>
-            fetch("/queue_size")
-                .then((res) => res.json())
-                .then((data) => {
-                    queueSizeEl.innerText = data.queue_size;
-                })
-                .catch((err) => {
-                    console.log(err);
-                })
-            , 5000);
-    </script>
-</head>
-
-<body class="text-black dark:bg-gray-900 dark:text-white">
-    <div class="fixed right-2 top-2 p-4 font-bold text-sm rounded-lg max-w-xs text-center" id="error">
-    </div>
-    <main class="container mx-auto px-4 py-4 max-w-4xl flex flex-col gap-4">
-        <article class="text-center max-w-xl mx-auto">
-            <h1 class="text-3xl font-bold">Real-Time Latent Consistency Model</h1>
-            <h2 class="text-2xl font-bold mb-4">Image to Image</h2>
-            <p class="text-sm">
-                This demo showcases
-                <a href="https://huggingface.co/SimianLuo/LCM_Dreamshaper_v7" target="_blank"
-                    class="text-blue-500 underline hover:no-underline">LCM</a> Image to Image pipeline
-                using
-                <a href="https://github.com/huggingface/diffusers/tree/main/examples/community#latent-consistency-pipeline"
-                    target="_blank" class="text-blue-500 underline hover:no-underline">Diffusers</a> with a MJPEG
-                stream server.
-            </p>
-            <p class="text-sm">
-                There are <span id="queue_size" class="font-bold">0</span> user(s) sharing the same GPU, affecting
-                real-time performance. Maximum queue size is 4. <a
-                    href="https://huggingface.co/spaces/radames/Real-Time-Latent-Consistency-Model?duplicate=true"
-                    target="_blank" class="text-blue-500 underline hover:no-underline">Duplicate</a> and run it on your
-                own GPU.
-            </p>
-        </article>
-        <div>
-            <h2 class="font-medium">Prompt</h2>
-            <p class="text-sm text-gray-500">
-                Change the prompt to generate different images, accepts <a
-                    href="https://github.com/damian0815/compel/blob/main/doc/syntax.md" target="_blank"
-                    class="text-blue-500 underline hover:no-underline">Compel</a> syntax.
-            </p>
-            <div class="flex text-normal px-1 py-1 border border-gray-700 rounded-md items-center">
-                <textarea type="text" id="prompt" class="font-light w-full px-3 py-2 mx-1  outline-none dark:text-black"
-                    title="Prompt, this is an example, feel free to modify"
-                    placeholder="Add your prompt here...">Portrait of The Terminator with , glare pose, detailed, intricate, full of colour, cinematic lighting, trending on artstation, 8k, hyperrealistic, focused, extreme details, unreal engine 5, cinematic, masterpiece</textarea>
-            </div>
-
-        </div>
-        <div class="">
-            <details>
-                <summary class="font-medium cursor-pointer">Advanced Options</summary>
-                <div class="grid grid-cols-3 sm:grid-cols-6 items-center gap-3 py-3">
-                    <label for="webcams" class="text-sm font-medium">Camera Options: </label>
-                    <select id="webcams" class="text-sm border-2 border-gray-500 rounded-md font-light dark:text-black">
-                    </select>
-                    <div></div>
-                    <label class="text-sm font-medium " for="steps">Inference Steps
-                    </label>
-                    <input type="range" id="steps" name="steps" min="1" max="20" value="4"
-                        oninput="this.nextElementSibling.value = Number(this.value)">
-                    <output class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md">
-                        4</output>
-                    <!--  -->
-                    <label class="text-sm font-medium" for="lcm_steps">LCM Inference Steps
-                    </label>
-                    <input type="range" id="lcm_steps" name="lcm_steps" min="2" max="60" value="50"
-                        oninput="this.nextElementSibling.value = Number(this.value)">
-                    <output class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md">
-                        50</output>
-                    <!--  -->
-                    <label class="text-sm font-medium" for="guidance-scale">Guidance Scale
-                    </label>
-                    <input type="range" id="guidance-scale" name="guidance-scale" min="0" max="30" step="0.001"
-                        value="8.0" oninput="this.nextElementSibling.value = Number(this.value).toFixed(2)">
-                    <output class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md">
-                        8.0</output>
-                    <!--  -->
-                    <label class="text-sm font-medium" for="strength">Strength</label>
-                    <input type="range" id="strength" name="strength" min="0.1" max="1" step="0.001" value="0.50"
-                        oninput="this.nextElementSibling.value = Number(this.value).toFixed(2)">
-                    <output class="text-xs w-[50px] text-center font-light px-1 py-1 border border-gray-700 rounded-md">
-                        0.5</output>
-                    <!--  -->
-                    <label class="text-sm font-medium" for="seed">Seed</label>
-                    <input type="number" id="seed" name="seed" value="299792458"
-                        class="font-light border border-gray-700 text-right rounded-md p-2 dark:text-black">
-                    <button
-                        onclick="document.querySelector('#seed').value = Math.floor(Math.random() * Number.MAX_SAFE_INTEGER)"
-                        class="button">
-                        Rand
-                    </button>
-                    <!--  -->
-                    <!--  -->
-                    <label class="text-sm font-medium" for="dimension">Image Dimensions</label>
-                    <div class="col-span-2 flex gap-2">
-                        <div class="flex gap-1">
-                            <input type="radio" id="dimension512" name="dimension" value="[512,512]" checked
-                                class="cursor-pointer">
-                            <label for="dimension512" class="text-sm cursor-pointer">512x512</label>
-                        </div>
-                        <div class="flex gap-1">
-                            <input type="radio" id="dimension768" name="dimension" value="[768,768]"
-                                lass="cursor-pointer">
-                            <label for="dimension768" class="text-sm cursor-pointer">768x768</label>
-                        </div>
-                    </div>
-                    <!--  -->
-                </div>
-            </details>
-        </div>
-        <div class="flex gap-3">
-            <button id="start" class="button">
-                Start
-            </button>
-            <button id="stop" class="button">
-                Stop
-            </button>
-            <button id="snap" disabled class="button ml-auto">
-                Snapshot
-            </button>
-        </div>
-        <div class="relative rounded-lg border border-slate-300 overflow-hidden">
-            <img id="player" class="w-full aspect-square rounded-lg "
-                src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNkYAAAAAYAAjCB0C8AAAAASUVORK5CYII=">
-            <div class="absolute top-0 left-0 w-1/4 aspect-square">
-                <video id="webcam" class="w-full aspect-square relative z-10 object-cover" playsinline autoplay muted
-                    loop></video>
-                <svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 448 448" width="100"
-                    class="w-full p-4 absolute top-0 opacity-20 z-0">
-                    <path fill="currentColor"
-                        d="M224 256a128 128 0 1 0 0-256 128 128 0 1 0 0 256zm-45.7 48A178.3 178.3 0 0 0 0 482.3 29.7 29.7 0 0 0 29.7 512h388.6a29.7 29.7 0 0 0 29.7-29.7c0-98.5-79.8-178.3-178.3-178.3h-91.4z" />
-                </svg>
-            </div>
-        </div>
-    </main>
-</body>
-
-</html>