{"id":23422758,"url":"https://github.com/linonetwo/moss-dockerfile","last_synced_at":"2025-04-12T15:07:50.266Z","repository":{"id":156155866,"uuid":"632386747","full_name":"linonetwo/MOSS-DockerFile","owner":"linonetwo","description":"用于在 Docker 里运行复旦的 MOSS 语言模型，使用 GradIO 提供 WebUI。","archived":false,"fork":false,"pushed_at":"2023-12-15T05:40:05.000Z","size":205,"stargazers_count":15,"open_issues_count":5,"forks_count":3,"subscribers_count":2,"default_branch":"master","last_synced_at":"2025-04-12T15:07:49.246Z","etag":null,"topics":["ai","chatglm","chatgpt","cuda","deeplearning","docker","gpu","moss","pytorch"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/linonetwo.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2023-04-25T09:53:31.000Z","updated_at":"2024-02-29T04:59:38.000Z","dependencies_parsed_at":"2023-12-15T06:49:08.562Z","dependency_job_id":null,"html_url":"https://github.com/linonetwo/MOSS-DockerFile","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/linonetwo%2FMOSS-DockerFile","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/linonetwo%2FMOSS-DockerFile/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/linonetwo%2FMOSS-DockerFile/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/linonetwo%2FMOSS-DockerFile/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/linonetwo","download_url":"https://codeload.github.com/linonetwo/MOSS-DockerFile/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":248586236,"owners_count":21128997,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai","chatglm","chatgpt","cuda","deeplearning","docker","gpu","moss","pytorch"],"created_at":"2024-12-23T03:09:49.828Z","updated_at":"2025-04-12T15:07:50.239Z","avatar_url":"https://github.com/linonetwo.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# MOSS-DockerFile\n\nhttps://hub.docker.com/repository/docker/linonetwo/moss/general\n\n用于在 Docker 里运行复旦的 MOSS 语言模型，使用 GradIO 提供 WebUI。\n\n## Run\n\n### 运行环境\n\nWindows 上 docker 运行环境搭建参考[2022 最新 Docker 和 WSL2 ，炼丹环境配置指南 - 无人知晓的文章 - 知乎](https://zhuanlan.zhihu.com/p/543280130)。然后使用 docker desktop 下载和管理镜像。\n\n最小的 int4 量化版本，需要一个 3090ti，将占用 14GB 显存，并随着上下文增加而增加。\n\n### 下载模型\n\n参考 https://huggingface.co/fnlp/moss-moon-003-sft-int4\n\n```sh\n# Make sure you have git-lfs installed (https://git-lfs.com)\ngit lfs install\ngit clone https://huggingface.co/fnlp/moss-moon-003-sft-int4\n\n# if you want to clone without large files – just their pointers\n# prepend your git clone with the following env var:\nGIT_LFS_SKIP_SMUDGE=1\n```\n\n我个人是先带上 `GIT_LFS_SKIP_SMUDGE=1` 去 `git clone`，来下载 `config.json` 等文件。然后单独用浏览器下载 10GB 的 `pytorch_model.bin` 文件，会比较快。\n\n之后需要将这个 `moss-moon-003-sft-int4` 文件夹作为 Volume 挂载到 docker 的 `/mnt/llm` 文件夹上。\n\n### 运行 docker\n\nExample shell command to run it in windows11 with WSL2 enabled\n\n```sh\ndocker run --gpus all -it --volume=C:\\Users\\linonetwo\\Documents\\model\\LanguageModel:/mnt/llm -d moss:latest\n```\n\nMy models are inside `C:\\Users\\linonetwo\\Documents\\model\\LanguageModel`, which includes `C:\\Users\\linonetwo\\Documents\\model\\LanguageModel\\MOSS\\moss-moon-003-sft-plugin-int4\\pytorch_model.bin`.\n\nIf your model are in different location, change the `--volume=xxx:/mnt/llm` part in the command.\n\nIn Docker Desktop, it looks like this:\n\n![llm-volume](images/llm-volume.png)\n\n## Build\n\n如果修改了 Docker file，需要重新构建镜像，这样构建：\n\n```shell\ndocker build -t moss . -f ./moss-int4-cuda117.dockerfile\n```\n\n## Screenshots\n\n![images/webui-demo.jpg](./images/webui-demo.jpg)\n\n![images/resource-usage.jpg](./images/resource-usage.jpg)\n\n## FAQ\n\n### Killed\n\nDocker 显示 Killed，一般是因为内存（RAM）不足，改 WSL2 的配置改大提供给虚拟机的内存即可：\n\n参考 https://stackoverflow.com/a/72693871/4617295\n\n### failed to solve with frontend dockerfile.v0\n\n```log\n =\u003e ERROR [internal] load metadata for docker.io/nvidia/cuda:11.7.1-cudnn8-devel-ubuntu22.04                                    0.2s\n------\n \u003e [internal] load metadata for docker.io/nvidia/cuda:11.7.1-cudnn8-devel-ubuntu22.04:\n------\nfailed to solve with frontend dockerfile.v0: failed to create LLB definition: unexpected status code [manifests 11.7.1-cudnn8-devel-ubuntu22.04]: 403 Forbidden\n```\n\n可能需要关掉 buildkit ，参考 https://stackoverflow.com/a/70153060/4617295\n\n### Failed to run image\n\n可能是没填 `ENTRYPOINT`，导致尝试基础镜像执行已被删掉的 `/opt/nvidia/nvidia_entrypoint.sh`\n\n```log\nFailed to run image. Error invoking remote method 'docker-run-container': Error: (HTTP code 400) unexpected - failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: exec: \"/opt/nvidia/nvidia_entrypoint.sh\": stat /opt/nvidia/nvidia_entrypoint.sh: no such file or directory: unknown\n```\n\n补上即可\n\n```dockerfile\nENTRYPOINT [\"/usr/bin/python3\"]\nCMD [\"moss_web_demo_gradio.py\"]\n# ENTRYPOINT [\"/usr/bin/env\"]\n# CMD [\"bash\"]\n```\n\n### moss_web_demo_gradio.py\n\n修改自 https://github.com/OpenLMLab/MOSS/blob/main/moss_web_demo_gradio.py ，那个版本未考虑量化用法，普通人电脑跑不了，故适配了量化用法，并自动修复官方运行 demo 会报的 `name 'autotune' is not defined` 错误。\n\n### GPU no found\n\n关键是带上 `--gpus all` ，即 `docker run --gpus all`\n\n## Credit\n\n参考了 [mortals-debuging/pytorch-docker](https://github.com/mortals-debuging/pytorch-docker)\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flinonetwo%2Fmoss-dockerfile","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Flinonetwo%2Fmoss-dockerfile","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Flinonetwo%2Fmoss-dockerfile/lists"}