volcano.sh/volcano@v1.9.0/example/MindSpore-example/README.md (about) 1 # MindSpore Volcano Example 2 3 #### These examples shows how to run MindSpore via Volcano. Since MindSpore itself is relatively new, these examples maybe oversimplified, but will evolve with both communities. 4 5 ## Introduction of MindSpore 6 7 MindSpore is a new open source deep learning training/inference framework that 8 could be used for mobile, edge and cloud scenarios. MindSpore is designed to 9 provide development experience with friendly design and efficient execution for 10 the data scientists and algorithmic engineers, native support for Ascend AI 11 processor, and software hardware co-optimization. 12 13 MindSpore is open sourced on both [Github](https://github.com/mindspore-ai/mindspore ) and [Gitee](https://gitee.com/mindspore/mindspore ). 14 15 ## Prerequisites 16 17 These two examples are tested under below env: 18 19 - Ubuntu: `16.04.6 LTS` 20 - docker: `v18.06.1-ce` 21 - Kubernetes: `v1.16.6` 22 - NVIDIA Docker: `2.3.0` 23 - NVIDIA/k8s-device-plugin: `1.0.0-beta6` 24 - NVIDIA drivers: `418.39` 25 - CUDA: `10.1` 26 27 ## MindSpore CPU example 28 29 Using a modified MindSpore CPU image as the container image which 30 trains LeNet with MNIST dataset. 31 32 pull image: `docker pull lyd911/mindspore-cpu-example:0.2.0` 33 to run: `kubectl apply -f mindspore-cpu.yaml` 34 to check the result: `kubectl logs mindspore-cpu-pod-0` 35 36 ## MindSpore GPU example 37 38 Using a modified image which the openssh-server is installed from 39 the official MindSpore GPU image. To check the eligibility of 40 MindSpore GPU's ability to communicate with other processes, we 41 leverage the mpimaster and mpiworker task spec of Volcano. In this 42 example, we launch one mpimaster and two mpiworkers, the python script 43 is taken from [MindSpore Gitee README](https://gitee.com/mindspore/mindspore/blob/master/README.md ), which is also modified to be 44 able to run parallelly. 45 46 pull image: `docker pull lyd911/mindspore-gpu-example:0.2.0` 47 to run: `kubectl apply -f mindspore-gpu.yaml` 48 to check result: `kubectl logs mindspore-gpu-mpimster-0` 49 50 The expected output should be (2*3) of multi-dimensional array. 51 52 ## Future 53 54 An end to end example of training a network using MindSpore on 55 distributed GPU via Volcano is expected in the future.