PyTorch快速安装-基于JupyterHub并运行K8s

PyTorch快速安装-基于JupyterHub并运行K8s

运行PyTorch能够直接逻辑运行、容器中运行、KubeFlow中运行以及基于JupyterHub（独立运行或运行在K8s之上）等多种模式。这里介绍运行在K8s上基于JupyterHub的PyTorch方法，这也是运行在云计算环境的推荐方法。若是须要使用GPU，则须要安装NVidia或AMD的Kubernetes下容器GPU支持，宿主机也必须同时安装GPU驱动。

• pytorch官网资源
  • https://pytorch.org
  • https://pytorch.org/tutorials
  • https://github.com/pytorch/pytorch
• 深度学习引擎-PyTorch资源集锦
  • http://www.javashuo.com/article/p-hoyneiey-cd.html

安装 kubernetes 的支持

• PyTorch支持Kubernetes集群, https://my.oschina.net/u/2306127/blog/1817835
• Kubeflow 使用指南，http://www.javashuo.com/article/p-nppedcay-hp.html
• pytorch-operator 项目源码，https://github.com/kubeflow/pytorch-operator
• pytorch-operator on Kubernetes，http://www.javashuo.com/article/p-afvqxpqf-dm.html
• 镜像构建-iObjects Python with JupyterHub for K8s,http://www.javashuo.com/article/p-uenogwzq-cc.html
• GPU 支持的安装
  • Kubernetes集群升级NVidia GPU驱动版本
  • Kubernetes中调度GPU资源
  • Ubuntu快速玩起色器学习
  • Ubuntu18.04上安装RTX 2080Ti显卡驱动

安装 JupyterHub/JupyterLab支持

conda install numpy
conda install scikit-image

conda install pytorch torchvision cudatoolkit=10.0 -c pytorch
# or: conda install pytorch-cpu torchvision-cpu -c pytorch

conda update --all

开始使用

获取教程数据：

• https://download.pytorch.org/tutorial/faces.zip

使用Notebook：

# 导入支持库
import torch 

# 确认CUDA支持及其版本
print(torch.version.cuda)

10.0.130

# 查看pytorch帮助
help(torch)

Help on package torch:

NAME
    torch

DESCRIPTION
    The torch package contains data structures for multi-dimensional
    tensors and mathematical operations over these are defined.
    Additionally, it provides many utilities for efficient serializing of
    Tensors and arbitrary types, and other useful utilities.
    
    It has a CUDA counterpart, that enables you to run your tensor computations
    on an NVIDIA GPU with compute capability >= 3.0.

PACKAGE CONTENTS
    _C
    _dl
    _jit_internal
    _nvrtc
    _ops
    _six
    _storage_docs
    _tensor_docs
    _tensor_str
    _thnn (package)
    _torch_docs
    _utils
    _utils_internal
    autograd (package)
    backends (package)
    contrib (package)
    cuda (package)
    distributed (package)
    distributions (package)
    for_onnx (package)
    functional
    hub
    jit (package)
    multiprocessing (package)
    nn (package)
    onnx (package)
    optim (package)
    random
    serialization
    sparse (package)
    storage
    tensor
    testing (package)
    utils (package)
    version
	......