注:当前项目为 Serverless Devs 应用,由于应用中会存在需要初始化才可运行的变量(例如应用部署地区、函数名等等),所以不推荐直接 Clone 本仓库到本地进行部署或直接复制 s.yaml 使用,强烈推荐通过
s init ${模版名称}
的方法或应用中心进行初始化,详情可参考部署 & 体验 。
部署 ComfyUI 到阿里云函数计算
使用该项目,您需要有开通以下服务并拥有对应权限:
服务 | 备注 |
---|---|
函数计算 FC | 提供 CPU、GPU 等计算资源 |
- 🔥 通过 Serverless 应用中心 , 该应用。
- 通过 Serverless Devs Cli 进行部署:
- 安装 Serverless Devs Cli 开发者工具 ,并进行授权信息配置 ;
- 初始化项目:
s init fc-comfyui -d fc-comfyui
- 进入项目,并进行项目部署:
cd fc-comfyui && s deploy -y
本案例展示了如何将开源项目 ComfyUI 部署到阿里云函数计算上,从而实现 ComfyUI 云端快速部署,实现文生图和图生图等 AIGC 创作活动。
ComfyUI 是一个为 Stable Diffusion 模型设计的,功能强大且高度模块化的图形用户界面(GUI)。它允许用户基于节点构建 AIGC 创作流程,非常适合那些想要摆脱传统编程方法、采用更直观操作流程的用户。该工具由 Comfyanonymous 在 2023 年 1 月创建,初衷是深入了解 Stable Diffusion 模型的运作机制。由于其易用性,Stable Diffusion 的开发者 Stability AI 也采用了 ComfyUI 进行内部测试,并聘请 Comfyanonymous 协助开发内部工具。目前,ComfyUI 在 Github 上的 Fork 数超过 3000,Star 数超过 30000。
Stable Diffusion 是一款由 CompVis、Stability AI 和 LAION 的研究人员及工程师共同开发的开源扩散模型,凭借其开源和高扩展性特点,赢得了全球众多 AIGC 爱好者的支持。据 Civital 模型网站统计,目前最热门的模型下载次数已超过 100 万,有超过 70 个模型下载次数超过 10 万,提供各种风格和功能的模型总数超过 12 万。
在国内,ComfyUI 也受到广泛欢迎。通过 ComfyUI 创作文生图的教程已多次在各大平台热搜榜和排行榜上出现,掀起一阵又一阵的热潮。通过 Serverless 开发平台,您只需要几步,就可以体验Comfyui,并享受Serverless 架构带来的降本提效的技术红利。
- 打开 ComfyUI,点击“Queue Prompt”开始。
- 稍等片刻后,您将得到第一张图片。若需要恢复默认工作流,请使用“Load Default”,并记得保存您的工作流以避免丢失。
为使用自定义模型和节点,需先绑定文件管理 NAS。通过函数控制台的网络配置,绑定专有网络/交换机。若无相关资源,需先创建。
- 进入函数计算控制台:通过应用详情,跳转到函数控制台
- 网络配置:完成专有网络和交换机配置(如果没有相关资源,您可以点击 “创建新的 VPC”、“创建新的交换机”,前往相关产品创建资源。)
- NAS 挂载:进行 NAS 挂载设置,绑定对应专有网络、交换机下存在的 NAS 挂载点。
函数本地目录请填写
/mnt/auto
或/mnt/auto/comfyui
;如果您曾经在当前 NAS 中使用过 Stable Diffusion 应用,可以考虑将远端目录设置为/fc-stable-diffusion-plus
,本地目录设置为/mnt/auto
。 大模型对文件 IO 要求较高,建议创建 通用性能型 NAS 实例,NAS 会根据存储的文件大小进行计费,不通规格的 NAS 计费单价不一致,请参考相关文档。
函数计算支持登入运行中的函数实例,您可以在终端中执行需要的操作(如手动安装自定义节点、依赖等)
注意,在 Serverless 环境下,您的所有改动都不会真正保存,您需要将改动的文件放置在 NAS 中以持久化
借助 NAS 文件浏览器功能,您可以方便地进行云上文件管理
以安装中文翻译插件 AIGODLIKE-COMFYUI-TRANSLATION 为例,使用 ComfyUI-Manager 进行安装。
搜索要安装的节点名称,点击 install
注意
- 安装过程中请不要关闭页面。安装完成后,除去需要点击 restart 外,还需要刷新页面
- 安装过程中可能会访问 Github、HuggingFace 等境外网站,由于网络问题可能会导致访问较慢或失败,您可以在网络上检索如何解决类似的问题。 )
使用国内 pypi 镜像加速依赖下载。编辑 /mnt/auto/comfyui/root/.pip/pip.conf
文件,设置镜像源为阿里云。
[global]
index-url = http://mirrors.aliyun.com/pypi/simple/
[install]
trusted-host = https://mirrors.aliyun.com
导入第三方的工作流,可能会遇到节点不存在的报错,可以借助 ComfyUI Manager 安装缺失的节点
部分节点升级后,可能仍然提示未安装,可参考 ComfyUI Guides 的相关讨论解决。
How to fix: A red node for “IPAdapterApply”? You must already follow our instructions on how to install IP-Adapter V2, and it should all working properly. Now you see a red node for “IPAdapterApply”.
That is because you are working on a workflow with IPAdapter V1 node, simply just replace the V1 node with the V2 ones or uninstall IPA v2 and rollback to V1 if you feel like it.
展示了使用 ControlNet 对比直接输出的差异,提供了工作流 JSON 示例以及对应模型的下载说明。
工作流文件
{
"last_node_id": 29,
"last_link_id": 54,
"nodes": [
{
"id": 16,
"type": "ControlNetLoader",
"pos": [ 264.88020036191443, 1201.535094958983 ],
"size": [ 376.46875, 118.875 ],
"flags": {},
"order": 0,
"mode": 0,
"outputs": [ { "name": "CONTROL_NET", "type": "CONTROL_NET", "links": [ 35 ], "shape": 3, "label": "ControlNet", "slot_index": 0 } ],
"properties": { "Node name for S&R": "ControlNetLoader" },
"widgets_values": [ "control_v11p_sd15_lineart.pth" ]
},
{
"id": 23,
"type": "ControlNetApplyAdvanced",
"pos": [ 856.880200361914, 1308.535094958983 ],
"size": { "0": 315, "1": 166 },
"flags": {},
"order": 6,
"mode": 0,
"inputs": [
{ "name": "positive", "type": "CONDITIONING", "link": 39, "label": "正面条件" },
{ "name": "negative", "type": "CONDITIONING", "link": 40, "label": "负面条件" },
{ "name": "control_net", "type": "CONTROL_NET", "link": 35, "label": "ControlNet" },
{ "name": "image", "type": "IMAGE", "link": 36, "label": "图像" }
],
"outputs": [
{ "name": "positive", "type": "CONDITIONING", "links": [ 41 ], "shape": 3, "label": "正面条件", "slot_index": 0 },
{ "name": "negative", "type": "CONDITIONING", "links": [ 42 ], "shape": 3, "label": "负面条件", "slot_index": 1 }
],
"properties": { "Node name for S&R": "ControlNetApplyAdvanced" },
"widgets_values": [ 1, 0, 1 ]
},
{
"id": 7,
"type": "CLIPTextEncode",
"pos": [ 305.4482288105471, 733.0172076020683 ],
"size": { "0": 425.27801513671875, "1": 180.6060791015625 },
"flags": {},
"order": 5,
"mode": 0,
"inputs": [ { "name": "clip", "type": "CLIP", "link": 5, "label": "CLIP" } ],
"outputs": [ { "name": "CONDITIONING", "type": "CONDITIONING", "links": [ 40, 45 ], "slot_index": 0, "label": "条件" } ],
"properties": { "Node name for S&R": "CLIPTextEncode" },
"widgets_values": [ "nsfw" ]
},
{
"id": 3,
"type": "KSampler",
"pos": [ 1815.3138964843754, 1041.130867919922 ],
"size": { "0": 315, "1": 262 },
"flags": {},
"order": 8,
"mode": 0,
"inputs": [
{ "name": "model", "type": "MODEL", "link": 1, "label": "模型" },
{ "name": "positive", "type": "CONDITIONING", "link": 41, "label": "正面条件" },
{ "name": "negative", "type": "CONDITIONING", "link": 42, "label": "负面条件" },
{ "name": "latent_image", "type": "LATENT", "link": 32, "label": "Latent", "slot_index": 3 }
],
"outputs": [ { "name": "LATENT", "type": "LATENT", "links": [ 16 ], "slot_index": 0, "label": "Latent" } ],
"properties": { "Node name for S&R": "KSampler" },
"widgets_values": [ 516902852614178, "randomize", 20, 8, "euler", "normal", 1 ]
},
{
"id": 8,
"type": "VAEDecode",
"pos": [ 2211.3138964843743, 1091.130867919922 ],
"size": { "0": 210, "1": 46 },
"flags": {},
"order": 10,
"mode": 0,
"inputs": [
{ "name": "samples", "type": "LATENT", "link": 16, "label": "Latent" },
{ "name": "vae", "type": "VAE", "link": 43, "label": "VAE" }
],
"outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [ 9 ], "slot_index": 0, "label": "图像" } ],
"properties": { "Node name for S&R": "VAEDecode" }
},
{
"id": 25,
"type": "VAEDecode",
"pos": [ 2285.340647460937, 637.991402648926 ],
"size": { "0": 210, "1": 46 },
"flags": {},
"order": 9,
"mode": 0,
"inputs": [
{ "name": "samples", "type": "LATENT", "link": 48, "label": "Latent" },
{ "name": "vae", "type": "VAE", "link": 50, "label": "VAE" }
],
"outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [ 49 ], "shape": 3, "label": "图像", "slot_index": 0 } ],
"properties": { "Node name for S&R": "VAEDecode" }
},
{
"id": 26,
"type": "SaveImage",
"pos": [ 2566.3138964843743, 631.130867919922 ],
"size": [ 315, 270.00002098083496 ],
"flags": {},
"order": 11,
"mode": 0,
"inputs": [ { "name": "images", "type": "IMAGE", "link": 49, "label": "图像" } ],
"properties": {},
"widgets_values": [ "ComfyUI" ]
},
{
"id": 9,
"type": "SaveImage",
"pos": [ 2626.3138964843743, 1022.1308679199219 ],
"size": [ 210, 270.00002002716064 ],
"flags": {},
"order": 12,
"mode": 0,
"inputs": [ { "name": "images", "type": "IMAGE", "link": 9, "label": "图像" } ],
"properties": {},
"widgets_values": [ "ComfyUI" ] },
{
"id": 6,
"type": "CLIPTextEncode",
"pos": [ 304.118812936719, 462.9874991923831 ],
"size": { "0": 422.84503173828125, "1": 164.31304931640625 },
"flags": {},
"order": 4,
"mode": 0,
"inputs": [ { "name": "clip", "type": "CLIP", "link": 3, "label": "CLIP" } ],
"outputs": [ { "name": "CONDITIONING", "type": "CONDITIONING", "links": [ 39, 44 ], "slot_index": 0, "label": "条件" } ],
"properties": { "Node name for S&R": "CLIPTextEncode" },
"widgets_values": [ "1 girl" ]
},
{
"id": 24,
"type": "KSampler",
"pos": [ 1880.3138964843754, 601.130867919922 ],
"size": { "0": 315, "1": 262 },
"flags": {},
"order": 7,
"mode": 0,
"inputs": [
{ "name": "model", "type": "MODEL", "link": 46, "label": "模型" },
{ "name": "positive", "type": "CONDITIONING", "link": 44, "label": "正面条件" },
{ "name": "negative", "type": "CONDITIONING", "link": 45, "label": "负面条件" },
{ "name": "latent_image", "type": "LATENT", "link": 54, "label": "Latent", "slot_index": 3 }
],
"outputs": [ { "name": "LATENT", "type": "LATENT", "links": [ 48 ], "shape": 3, "label": "Latent", "slot_index": 0 } ],
"properties": { "Node name for S&R": "KSampler" },
"widgets_values": [ 963578161132850, "randomize", 20, 8, "euler", "normal", 1 ]
},
{
"id": 5,
"type": "EmptyLatentImage",
"pos": [ 1412.3138964843754, 744.130867919922 ],
"size": { "0": 315, "1": 106 },
"flags": {},
"order": 1,
"mode": 0,
"outputs": [ { "name": "LATENT", "type": "LATENT", "links": [ 32, 54 ], "slot_index": 0, "label": "Latent" } ],
"properties": { "Node name for S&R": "EmptyLatentImage" },
"widgets_values": [ 512, 512, 1 ]
},
{
"id": 12,
"type": "LoadImage",
"pos": [ 273.88020036191443, 1411.535094958983 ],
"size": { "0": 315, "1": 314 },
"flags": {},
"order": 2,
"mode": 0,
"outputs": [
{ "name": "IMAGE", "type": "IMAGE", "links": [ 36 ], "shape": 3, "label": "图像", "slot_index": 0 },
{ "name": "MASK", "type": "MASK", "links": null, "shape": 3, "label": "遮罩" }
],
"properties": { "Node name for S&R": "LoadImage" },
"widgets_values": [ "example.png", "image" ]
},
{
"id": 4,
"type": "CheckpointLoaderSimple",
"pos": [ -315.8811870632813, 556.9874991923831 ],
"size": { "0": 356.0684509277344, "1": 159.5682373046875 },
"flags": {},
"order": 3,
"mode": 0,
"outputs": [
{ "name": "MODEL", "type": "MODEL", "links": [ 1, 46 ], "slot_index": 0, "label": "模型" },
{ "name": "CLIP", "type": "CLIP", "links": [ 3, 5 ], "slot_index": 1, "label": "CLIP" },
{ "name": "VAE", "type": "VAE", "links": [ 43, 50 ], "slot_index": 2, "label": "VAE" }
],
"properties": { "Node name for S&R": "CheckpointLoaderSimple" },
"widgets_values": [ "AWPortraitv1.1.safetensors" ]
}
],
"links": [
[ 1, 4, 0, 3, 0, "MODEL" ],
[ 3, 4, 1, 6, 0, "CLIP" ],
[ 5, 4, 1, 7, 0, "CLIP" ],
[ 9, 8, 0, 9, 0, "IMAGE" ],
[ 16, 3, 0, 8, 0, "LATENT" ],
[ 32, 5, 0, 3, 3, "LATENT" ],
[ 35, 16, 0, 23, 2, "CONTROL_NET" ],
[ 36, 12, 0, 23, 3, "IMAGE" ],
[ 39, 6, 0, 23, 0, "CONDITIONING" ],
[ 40, 7, 0, 23, 1, "CONDITIONING" ],
[ 41, 23, 0, 3, 1, "CONDITIONING" ],
[ 42, 23, 1, 3, 2, "CONDITIONING" ],
[ 43, 4, 2, 8, 1, "VAE" ],
[ 44, 6, 0, 24, 1, "CONDITIONING" ],
[ 45, 7, 0, 24, 2, "CONDITIONING" ],
[ 46, 4, 0, 24, 0, "MODEL" ],
[ 48, 24, 0, 25, 0, "LATENT" ],
[ 49, 25, 0, 26, 0, "IMAGE" ],
[ 50, 4, 2, 25, 1, "VAE" ],
[ 54, 5, 0, 24, 3, "LATENT" ]
],
"groups": [
{ "title": "ControlNet", "bounding": [ 210, 1105, 1012, 660 ], "color": "#3f789e", "font_size": 24 },
{ "title": "文生图", "bounding": [ -347, 228, 1185, 747 ], "color": "#3f789e", "font_size": 24 },
{ "title": "输出", "bounding": [ 1296, 400, 1615, 951 ], "color": "#3f789e", "font_size": 24 }
],
"config": {},
"extra": {},
"version": 0.4
}
fc-comfyui 是一个第三方工具,旨在帮助用户将 ComfyUI 项目部署到阿里云函数计算服务。请注意,该工具与 ComfyUI 项目及阿里云官方无直接联系。
- 第三方链接:本工具提供的第三方网站或服务链接仅为用户方便,开发者对这些内容、隐私政策或操作不承担责任,亦不代表认可。
- 社区同步:ComfyUI 为活跃的开源社区项目,功能丰富且更新频繁,如果您希望使用更新版本的 ComfyUI,可自行基于 Dockerfile 文件进行构建。
- 费用提示:在阿里云部署 ComfyUI 可能产生费用,请参考阿里云的计费文档。若需持久化存储(如模型、节点),还需开通文件管理 NAS,可能产生额外费用。
- 许可协议:使用 ComfyUI 项目需遵守其开源许可协议。使用前,请确保已阅读并理解 ComfyUI 项目及相关第三方工具的许可协议。
- 遵守服务条款:部署至阿里云函数计算服务,需同意阿里云服务条款和使用政策。
- 无担保声明:本工具“按现状”提供,不包含任何形式的担保。使用风险由用户自担,开发者不负责任何直接或间接损害。
- 资源消耗:ComfyUI 页面建立长连接请求,持续消耗计算资源。为避免不必要费用,请不使用时关闭所有页面。
使用本工具即表示您已理解并同意以上声明。若不同意,请勿使用。
您如果有关于错误的反馈或者未来的期待,您可以在 Serverless Devs repo Issues 中进行反馈和交流。如果您想要加入我们的讨论组或者了解 FC 组件的最新动态,您可以通过以下渠道进行:
微信公众号:serverless |
微信小助手:xiaojiangwh |
钉钉交流群:33947367 |