mind-cluster

module

v6.0.0+incompatible Latest Latest Go to latest Published: Dec 31, 2024 License: Apache-2.0

Details

Valid go.mod file
Redistributable license
Tagged version
Stable version
Learn more about best practices

Repository

gitee.com/ascend/mind-cluster

Links

Open Source Insights

README ¶

MindX DL

免责说明
组件介绍
支持的产品形态
编译
组件安装
说明
更新日志
版本配套说明

免责说明

本仓库代码中包含多个开发分支，这些分支可能包含未完成、实验性或未测试的功能。在正式发布前，这些分支不应被应用于任何生产环境或者依赖关键业务的项目中。请务必使用我们的正式发行版本，以确保代码的稳定性和安全性。使用开发分支所导致的任何问题、损失或数据损坏，本项目及其贡献值概不负责。
正式版本请参考release版本 https://gitee.com/ascend/mindxdl/releases

介绍

MindX DL（AI集群系统软件）是支持NPU（昇腾AI处理器）训练和推理硬件的深度学习组件，使能构建集群全流程运行，提供NPU集群作业调度、运维监控、故障恢复等功能。深度学习平台开发厂商可以减少底层资源调度相关软件开发工作量，快速使能合作伙伴基于MindX DL开发深度学习平台。

支持的产品形态

支持以下产品使用资源监测
- Atlas 训练系列产品
- Atlas A2 训练系列产品
- Atlas A3 训练系列产品
- 推理服务器（插Atlas 300I 推理卡）
- Atlas 推理系列产品（Ascend 310P AI处理器）
- Atlas 800I A2 推理服务器

编译

拉取mindxdl整体源码放在/usr1目录下
修改组件版本配置文件service_config.ini中mindxdlversion字段值为所需编译版本，默认值如下，
```
mindxdlversion=6.0.RC3
```
执行以下命令，进入/usr1/mindxdl/build目录，选择构建脚本执行

cd /usr1/mindxdl/build
```
dos2unix *.sh && chmod +x *.sh

./build_all.sh $GOPATH
```
执行完成后进入$GOPATH/目录在各组件“output“目录下生成编译完成的文件, 其中ascend-for-volcano组件编译完成文件在output目录中。

组件安装

请参考昇腾社区《MindX DL用户指南》

入口地址：https://www.hiascend.com/software/mindx-dl

说明

当前容器方式部署本组件，本组件的认证鉴权方式为ServiceAccount，该认证鉴权方式为ServiceAccount的token明文显示，建议用户自行进行安全加强。
当前特权容器方式部署，该容器权限具有一定风险，建议用户自行进行安全加强。

更新日志

该仓库融合DL不同组件内容。6.0.0之前的版本见各组件仓库：

组件	链接	说明
NodeD	https://gitee.com/ascend/ascend-noded	-
HCCL-Controller	https://gitee.com/ascend/ascend-hccl-controller	此组件功能已被Ascend-Operator收编，不建议使用
Ascend-Device-Plugin	https://gitee.com/ascend/ascend-device-plugin	-
NPU-Exporter	https://gitee.com/ascend/ascend-npu-exporter	-
Ascend-for-Volcano	https://gitee.com/ascend/ascend-for-volcano	-
Ascend-Docker-Runtime	https://gitee.com/ascend/ascend-docker-runtime	-

6.0.0及之后版本发布如下：

版本	发布日期	修改说明
v6.0.0	2024-12-15	配套MindX 6.0.0版本

版本配套说明

版本配套详情请参考：版本配套详情

Directories ¶

?	: This menu
/	: Search site
f or F	: Jump to
y or Y	: Canonical URL

Path	Synopsis
component
ascend-for-volcano Package main is using for HuaWei Ascend pin affinity schedule.	Package main is using for HuaWei Ascend pin affinity schedule.
ascend-for-volcano/config Package config is using for HuaWei Ascend schedule configuration.	Package config is using for HuaWei Ascend schedule configuration.
ascend-for-volcano/internal/ascend310 Package ascend310 is using for HuaWei A300T pin affinity schedule.	Package ascend310 is using for HuaWei A300T pin affinity schedule.
ascend-for-volcano/internal/ascend310/card310x4 Package card310x4 is using for HuaWei A300T Ascend pin affinity schedule.	Package card310x4 is using for HuaWei A300T Ascend pin affinity schedule.
ascend-for-volcano/internal/ascend310/chip310x4 Package chip310x4 is using for HuaWei 310 Ascend pin affinity schedule.	Package chip310x4 is using for HuaWei 310 Ascend pin affinity schedule.
ascend-for-volcano/internal/ascend310p Package ascend310p is using for HuaWei 310P Ascend pin affinity schedule.	Package ascend310p is using for HuaWei 310P Ascend pin affinity schedule.
ascend-for-volcano/internal/ascend310p/card310px2 Package card310px2 is using for HuaWei 300I Duo Ascend pin affinity schedule.	Package card310px2 is using for HuaWei 300I Duo Ascend pin affinity schedule.
ascend-for-volcano/internal/ascend310p/chip310px2 Package chip310px2 is using for HuaWei 300I Duo Ascend pin affinity schedule.	Package chip310px2 is using for HuaWei 300I Duo Ascend pin affinity schedule.
ascend-for-volcano/internal/ascend910 Package ascend910 is using for HuaWei Ascend pin affinity schedule.	Package ascend910 is using for HuaWei Ascend pin affinity schedule.
ascend-for-volcano/internal/ascend910/ascend910b Package ascend910b is using for HuaWei Ascend 910B pin affinity schedule.	Package ascend910b is using for HuaWei Ascend 910B pin affinity schedule.
ascend-for-volcano/internal/ascend910/ascend910b/card910bx2 Package card910bx2 is using for HuaWei Ascend 910B(Atlas 300T A2) card pin affinity schedule.	Package card910bx2 is using for HuaWei Ascend 910B(Atlas 300T A2) card pin affinity schedule.
ascend-for-volcano/internal/ascend910/ascend910b/card910bx2infer Package card910bx2 is using for HuaWei Ascend 910B(Atlas 300T A2) card pin affinity schedule.	Package card910bx2 is using for HuaWei Ascend 910B(Atlas 300T A2) card pin affinity schedule.
ascend-for-volcano/internal/ascend910/ascend910b/module910bx16 Package module910bx16 is using for HuaWei Ascend910B A+X pin affinity schedule.	Package module910bx16 is using for HuaWei Ascend910B A+X pin affinity schedule.
ascend-for-volcano/internal/ascend910/ascend910b/module910bx8 Package module910bx8 is using for HuaWei Ascend910Bx8 pin affinity schedule.	Package module910bx8 is using for HuaWei Ascend910Bx8 pin affinity schedule.
ascend-for-volcano/internal/ascend910/ascend910b/superpod Package superpod is using for HuaWei Atlas 900 A3 SuperPod affinity schedule.	Package superpod is using for HuaWei Atlas 900 A3 SuperPod affinity schedule.
ascend-for-volcano/internal/ascend910/asend910old/card910x2 Package card910x2 is using for HuaWei Ascend pin affinity schedule.	Package card910x2 is using for HuaWei Ascend pin affinity schedule.
ascend-for-volcano/internal/ascend910/asend910old/half910x4 Package half910x4 is using for HuaWei A800/9000 Ascend910 pin affinity schedule.	Package half910x4 is using for HuaWei A800/9000 Ascend910 pin affinity schedule.
ascend-for-volcano/internal/ascend910/asend910old/module910x8 Package module910x8 is using for HuaWei A800/9000 Ascend910 pin affinity schedule.	Package module910x8 is using for HuaWei A800/9000 Ascend910 pin affinity schedule.
ascend-for-volcano/internal/base Package base is using for HuaWei Ascend pin affinity schedule.	Package base is using for HuaWei Ascend pin affinity schedule.
ascend-for-volcano/internal/rescheduling Package rescheduling is using for HuaWei Ascend pin fault rescheduling.	Package rescheduling is using for HuaWei Ascend pin fault rescheduling.
ascend-for-volcano/internal/test Package test is using for HuaWei Ascend testing.	Package test is using for HuaWei Ascend testing.
ascend-for-volcano/internal/vnpu Package vnpu is using for HuaWei Ascend pin vnpu allocation.	Package vnpu is using for HuaWei Ascend pin vnpu allocation.
ascend-for-volcano/plugin Package plugin is using for HuaWei Ascend pin affinity schedule.	Package plugin is using for HuaWei Ascend pin affinity schedule.
ascend-for-volcano/test Package test is using for HuaWei Ascend pin scheduling test.	Package test is using for HuaWei Ascend pin scheduling test.
ascend-for-volcano/util Package util is using for HuaWei infer common Ascend pin affinity schedule.	Package util is using for HuaWei infer common Ascend pin affinity schedule.