diff --git a/README.adoc b/README.adoc index 3ccfb40..b072eac 100644 --- a/README.adoc +++ b/README.adoc @@ -1,7 +1,8 @@ # Docker image for ComfyUI -image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-latest.yml/badge.svg["GitHub Workflow Status"] -image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-rocm.yml/badge.svg["GitHub Workflow Status"] +image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-latest.yml/badge.svg["GitHub Workflow Status",link="https://github.com/YanWenKun/ComfyUI-Docker/actions"] +image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-rocm.yml/badge.svg["GitHub Workflow Status",link="https://github.com/YanWenKun/ComfyUI-Docker/tree/main/rocm"] +image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-megapak.yml/badge.svg["GitHub Workflow Status",link="https://github.com/YanWenKun/ComfyUI-Docker/tree/main/megapak"] *link:README.zh.adoc[＞💡中文说明点我💡＜]* @@ -26,7 +27,7 @@ a Stable Diffusion GUI powering node-based workflow. * NVIDIA GPU with ≥6GB VRAM ** For 4GB see <>. -** For AMD GPU see <>. +** For AMD GPU see link:rocm/README.adoc[ROCm]. * Latest NVIDIA GPU driver ** Either Game or Studio edition will work. @@ -35,27 +36,12 @@ a Stable Diffusion GUI powering node-based workflow. * Docker Installed ** Linux user may need to install https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html[NVIDIA Container Toolkit] (only on host OS). It will enable containers' GPU access. ** Windows user could use https://www.docker.com/products/docker-desktop/[Docker Desktop] with WSL2 enabled. +** For Podman users see: <>. ## Usage - NVIDIA GPU -.A. Using `docker compose` -[source,sh] ----- -git clone https://github.com/YanWenKun/ComfyUI-Docker.git - -cd ComfyUI-Docker - -docker compose up --detach - -# Update image (only when Python components are outdated) -git pull -docker compose pull -docker compose up --detach --remove-orphans -docker image prune ----- - -.B. Using `docker run` +.A. Using `docker run` [source,sh] ---- mkdir -p storage @@ -65,65 +51,43 @@ docker run -it \ --gpus all \ -p 8188:8188 \ -v "$(pwd)"/storage:/home/runner \ - --env CLI_ARGS="" \ + -e CLI_ARGS="" \ yanwk/comfyui-boot:latest +---- +[source,sh] +---- # Update image (only when Python components are outdated) docker rm comfyui docker pull yanwk/comfyui-boot:latest -# Then re-run 'docker run' above again +# Then re-run above 'docker run' again ---- -Once the app is loaded, visit http://localhost:8188/ - - -[[rocm]] -## Usage - AMD GPU (Experimental) - -NOTE: Recommend AMD users to run ComfyUI on Windows with ZLUDA. Running on Linux needs some experience. - -NOTE: If you are using WSL2 with AMD/Intel GPU, consider *link:docs/wsl-directml.adoc[Run ComfyUI on WSL2 with DirectML]*. - -NOTE: On Linux hosts, install -https://rocm.docs.amd.com/projects/radeon/en/latest/docs/install/install-radeon.html[Radeon software for Linux with ROCm] -first. - -.C. Using `docker compose` +.B. Using `docker compose` [source,sh] ---- git clone https://github.com/YanWenKun/ComfyUI-Docker.git cd ComfyUI-Docker -docker compose -f docker-compose-rocm.yml up --detach +docker compose up --detach +---- +[source,sh] +---- # Update image (only when Python components are outdated) git pull -docker compose -f docker-compose-rocm.yml pull -docker compose -f docker-compose-rocm.yml up --detach --remove-orphans +docker compose pull +docker compose up --detach --remove-orphans docker image prune ---- -.D. Using `docker run` (as root) -[source,sh] ----- -mkdir -p storage +Once the app is loaded, visit http://localhost:8188/ -docker run -it --rm \ - --name comfyui-root \ - --device=/dev/kfd --device=/dev/dri \ - --group-add=video --ipc=host --cap-add=SYS_PTRACE \ - --security-opt seccomp=unconfined \ - --security-opt label=disable \ - -p 8188:8188 \ - --env CLI_ARGS="--use-pytorch-cross-attention" \ - --user root --workdir /root \ - -v "$(pwd)"/storage:/root \ - yanwk/comfyui-boot:rocm \ - /bin/bash /home/scripts/root-wrapper.sh ----- -Once the app is loaded, visit http://localhost:8188/ +## Usage - AMD GPU + +See link:rocm/README.adoc[ROCm]. [[q-n-a]] @@ -147,7 +111,7 @@ https://github.com/comfyanonymous/ComfyUI/blob/master/comfy/cli_args.py[ComfyUI] ## Add some custom nodes -.Install a bunch of custom nodes that may save your time +.Install a bunch of custom nodes that may save your time [%collapsible] ==== Note that most dependencies are bundled in the image, you don't need to manually install them. @@ -208,7 +172,8 @@ https://github.com/WASasquatch/was-node-suite-comfyui[WAS Node Suite], because it has some deps version https://github.com/WASasquatch/was-node-suite-comfyui/blob/main/requirements.txt[fixed], and is not under active development. + -However, the image's script only install one custom node (ComfyUI-Manager). You can safely install WAS NS via ComfyUI-Manager in a new deployment. If anything conflicts, just delete unwanted custom nodes and `.local` (or `local`) folder, and update/try-fix/reinstall custom nodes in ComfyUI-Manager. +However, the image's script only install one custom node (ComfyUI-Manager). You can safely install WAS NS via ComfyUI-Manager in a new deployment. + +If anything conflicts, just delete unwanted custom nodes and `.local` (or `local`) folder, and update/try-fix/reinstall custom nodes in ComfyUI-Manager. ==== @@ -226,12 +191,13 @@ Also, if you need to setup proxy, this script will start before `pre-start` and ---- +[[podman]] ## Run with Podman Podman is root-less by default. And https://www.tutorialworks.com/podman-rootless-volumes/[it's tricky] to bind-mount volume as non-root user inside rootless container. https://docs.podman.io/en/latest/markdown/podman-run.1.html#mount-type-type-type-specific-option[Options] provided by Podman will chown files on host space, which probably is undesirable. -I recommend two different ways to workaround: +Recommend three different ways to workaround: ### 1. Go "root-ful" just like Docker @@ -282,6 +248,8 @@ podman run -it --rm \ ---- ==== +### 3. Use link:megapak/README.adoc[megapak] image + ## Debugging diff --git a/README.zh.adoc b/README.zh.adoc index ce2631e..c9a672a 100644 --- a/README.zh.adoc +++ b/README.zh.adoc @@ -1,7 +1,8 @@ # 用于 ComfyUI 的 Docker 镜像 -image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-latest.yml/badge.svg["GitHub Workflow Status"] -image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-rocm.yml/badge.svg["GitHub Workflow Status"] +image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-latest.yml/badge.svg["GitHub Workflow Status",link="https://github.com/YanWenKun/ComfyUI-Docker/actions"] +image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-rocm.yml/badge.svg["GitHub Workflow Status",link="https://github.com/YanWenKun/ComfyUI-Docker/tree/main/rocm"] +image:https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-megapak.yml/badge.svg["GitHub Workflow Status",link="https://github.com/YanWenKun/ComfyUI-Docker/tree/main/megapak"] image::docs/chart-concept.zh.svg["布局"] @@ -24,7 +25,7 @@ https://hub.docker.com/r/yanwk/comfyui-boot[Docker 镜像] * NVIDIA 显卡， ≥6GB 显存 ** 4GB 方法见 <> -** AMD 显卡见 <> +** AMD 显卡见 link:rocm/README.zh.adoc[ROCm]. * 安装好最新的 NVIDIA 显卡驱动 ** 游戏驱动或 Studio 驱动均可。 @@ -34,27 +35,12 @@ https://hub.docker.com/r/yanwk/comfyui-boot[Docker 镜像] ** Linux 用户可能需要安装 https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html[NVIDIA Container Toolkit] （使得容器可以访问宿主机的 GPU）。 ** Windows 用户建议使用 https://www.docker.com/products/docker-desktop/[Docker Desktop] 并在安装时启用 WSL2，并 https://zhuanlan.zhihu.com/p/345645621[限制内存用量] 。 ** 下载过程可能需要科学上网，在 Docker 设置中可以找到代理设置。 +** Podman 用户参见： <>。 ## 运行方法 - NVIDIA 显卡 -.方法 A 使用 `docker compose` -[source,sh] ----- -git clone https://github.com/YanWenKun/ComfyUI-Docker.git - -cd ComfyUI-Docker - -docker compose up --detach - -# 更新镜像（仅在 Python 组件版本过期时需要） -git pull -docker compose pull -docker compose up --detach --remove-orphans -docker image prune ----- - -.方法 B 使用 `docker run` +.方法 A 使用 `docker run` [source,sh] ---- mkdir -p storage @@ -64,64 +50,43 @@ docker run -it \ --gpus all \ -p 8188:8188 \ -v "$(pwd)"/storage:/home/runner \ - --env CLI_ARGS="" \ + -e CLI_ARGS="" \ yanwk/comfyui-boot:latest +---- +[source,sh] +---- # 更新镜像（仅在 Python 组件版本过期时需要） docker rm comfyui docker pull yanwk/comfyui-boot:latest # 接下来再运行一遍上述 'docker run' 即可 ---- -启动完成后，访问 http://localhost:8188/ - - -[[rocm]] -## 运行方法 - AMD 显卡（尚在实验） - -NOTE: 推荐 AMD 用户使用 Win11 + ZLUDA 方式运行，安装 AMD HIP SDK 并使用常见整合包运行即可（不要开启 xFormers），初次运行会花 15 分钟左右来编译代码，之后就和 CUDA 体验差不多了。不推荐缺乏经验的用户折腾 Linux 下的 ROCm。 - -NOTE: WSL2 + AMD/Intel GPU 可以考虑： *link:docs/wsl-directml.zh.adoc[在 WSL2 环境下通过 DirectML 运行 ComfyUI]*。 - -NOTE: Linux 宿主机需要安装 -https://rocm.docs.amd.com/projects/radeon/en/latest/docs/install/install-radeon.html[Radeon software for Linux with ROCm] 以让 Docker 支持 GPU，各发行版可能安装方式不同。 - -.方法 C 使用 `docker compose` +.方法 B 使用 `docker compose` [source,sh] ---- git clone https://github.com/YanWenKun/ComfyUI-Docker.git cd ComfyUI-Docker -docker compose -f docker-compose-rocm.yml up --detach +docker compose up --detach +---- +[source,sh] +---- # 更新镜像（仅在 Python 组件版本过期时需要） git pull -docker compose -f docker-compose-rocm.yml pull -docker compose -f docker-compose-rocm.yml up --detach --remove-orphans +docker compose pull +docker compose up --detach --remove-orphans docker image prune ---- -.方法 D 使用 `docker run`（root 身份） -[source,sh] ----- -mkdir -p storage +启动完成后，访问 http://localhost:8188/ -docker run -it --rm \ - --name comfyui-root \ - --device=/dev/kfd --device=/dev/dri \ - --group-add=video --ipc=host --cap-add=SYS_PTRACE \ - --security-opt seccomp=unconfined \ - --security-opt label=disable \ - -p 8188:8188 \ - --env CLI_ARGS="--use-pytorch-cross-attention" \ - --user root --workdir /root \ - -v "$(pwd)"/storage:/root \ - yanwk/comfyui-boot:rocm \ - /bin/bash /home/scripts/root-wrapper.sh ----- -启动完成后，访问 http://localhost:8188/ +## 运行方法 - AMD 显卡 + +参见 link:rocm/README.zh.adoc[ROCm]。 [[q-n-a]] @@ -245,15 +210,16 @@ echo "[INFO] Proxy set to $HTTP_PROXY" ==== +[[podman]] ## 使用 Podman 运行 -Podman 默认是 root-less 的，不需要 sudo，这也给挂载目录 +关于文件权限：Podman 默认是 root-less 的，不需要 sudo，这也给挂载目录 https://www.tutorialworks.com/podman-rootless-volumes/[带来了限制] 。 + Podman 默认挂载文件为 root 身份，而想要在容器内挂载为非 root 用户，Podman 提供的 https://docs.podman.io/en/latest/markdown/podman-run.1.html#mount-type-type-type-specific-option[选项] 则会对主机上的文件执行 `chown`，变成和容器内一样的 uid 和 gid，给文件管理带来混乱。 -这里推荐两种不同的方式绕过： +这里推荐三种不同的方式绕过： ### 1. 像 Docker 一样 "root-ful" @@ -304,6 +270,8 @@ podman run -it --rm \ ---- ==== +### 3. 改用 link:megapak/README.zh.adoc[megapak] 镜像 + ## 一些方便 Debug 的命令 diff --git a/docs/README-DockerHub.md b/docs/README-DockerHub.md index 3cbfd2d..73378e9 100644 --- a/docs/README-DockerHub.md +++ b/docs/README-DockerHub.md @@ -1,7 +1,8 @@ # ComfyUI -[![GitHub Workflow Status](https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-latest.yml/badge.svg)](https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-latest.yml) -[![GitHub Workflow Status](https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-rocm.yml/badge.svg)](https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-rocm.yml) +[![GitHub Workflow Status](https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-latest.yml/badge.svg)](https://github.com/YanWenKun/ComfyUI-Docker/actions) +[![GitHub Workflow Status](https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-rocm.yml/badge.svg)](https://github.com/YanWenKun/ComfyUI-Docker/tree/main/rocm) +[![GitHub Workflow Status](https://github.com/YanWenKun/ComfyUI-Docker/actions/workflows/build-megapak.yml/badge.svg)](https://github.com/YanWenKun/ComfyUI-Docker/tree/main/megapak) **[CHECK THE GITHUB REPO](https://github.com/YanWenKun/ComfyUI-Docker)** @@ -23,27 +24,12 @@ docker run -it --rm \ yanwk/comfyui-boot:latest ``` -## Usage - AMD GPU (Experimental) - -```sh -mkdir -p storage +Once the app is loaded, visit http://localhost:8188/ -docker run -it --rm \ - --name comfyui-root \ - --device=/dev/kfd --device=/dev/dri \ - --group-add=video --ipc=host --cap-add=SYS_PTRACE \ - --security-opt seccomp=unconfined \ - --security-opt label=disable \ - -p 8188:8188 \ - --env CLI_ARGS="--use-pytorch-cross-attention" \ - --user root --workdir /root \ - -v "$(pwd)"/storage:/root \ - yanwk/comfyui-boot:rocm \ - /bin/bash /home/scripts/root-wrapper.sh -``` +## More Usage -## More +- [AMD GPU with ROCm](https://github.com/YanWenKun/ComfyUI-Docker/tree/main/rocm) -Once the app is loaded, visit http://localhost:8188/ +- ["Megapak" (all-in-one bundle)](https://github.com/YanWenKun/ComfyUI-Docker/tree/main/megapak) -For detailed `docker run` commands, check the doc on [GitHub repo](https://github.com/YanWenKun/ComfyUI-Docker). +- For docs and image detail, check [GitHub Page](https://github.com/YanWenKun/ComfyUI-Docker). diff --git a/rocm/README.adoc b/rocm/README.adoc index c27bb48..5f98c0e 100644 --- a/rocm/README.adoc +++ b/rocm/README.adoc @@ -1,5 +1,7 @@ # Run ComfyUI with ROCm on AMD GPU +*link:README.zh.adoc[中文说明]* + ## In case of WSL2 If you are using WSL2 with AMD/Intel GPU, try diff --git a/rocm/README.zh.adoc b/rocm/README.zh.adoc index c8c5e39..a6907af 100644 --- a/rocm/README.zh.adoc +++ b/rocm/README.zh.adoc @@ -1,6 +1,6 @@ # 基于 AMD GPU + ROCm 运行 ComfyUI -## 提醒： WSL2 +## 提醒： WSL2 用户 如果你在用 WSL2 + AMD/Intel GPU， 可以考虑： link:../docs/wsl-directml.zh.adoc[在 WSL2 环境下通过 DirectML 运行 ComfyUI]。