完全混淆了Re AML，VS，工作台，工作区，实验，项目，帐户，Arghhhh.... [英] Totally confused Re AML, VS, Workbench, Workspace, Experiment, Project, Account, Arghhhh....

问题描述

大家好，

首先，我不是一名开发人员(我主要从事vb/c#.net的工作，但如果需要的话，总是可以选择python的)，也不是Azure迷，并且会理解v.一点行话.我是一位可以偶然接触大多数代码和软件的分析师，并且对机器感兴趣 学习.

First, I'm not a developer (I mainly do vb/c#.net but could always pick up python if required), nor an Azure aficionado, and will understand v. little jargon.  I'm an analyst who can stumble through most code and software, and has an interest in machine learning.

不过，我对Microsoft的ML产品完全感到困惑.几年前，我曾经在在线Azure ML工作室玩过.我这样做是因为它是有用的GUI，并且在可能的情况下，我通常都喜欢GUI.我最近回来了 机器学习，我想尽我所能来了解Microsoft的产品，但我完全感到困惑.

However I've got myself totally confused with Microsoft's ML offerings.  I used to play with the online Azure ML studio - years ago.  I did this because it was a useful GUI and I generally like GUIs where possible.  I've recently come back to machine learning and I'm trying to get my head around Microsoft's offering(s), but I'm totally confused. 

我已经有了一个新的Azure帐户并创建了一个新的ML Studio工作区，并且对Web GUI(Machine Learning Studio)有了一点点修改，这在很大程度上是我记得的.

I have got myself a new Azure account and created a new ML Studio Workspace, and had a little tinker with the web GUI (Machine Learning Studio) and it's largely as I remember.

但是(以我的旧帐户而言)，我对机器学习工作室的成本和缓慢感到​​不满意，因此我一直在研究Microsoft提供的范围更广的ML产品.但是我对行话的范围感到困惑.之间有什么区别 Azure ML Studio，实验，实验(资源)，项目，工作台，工作区等，等等.? 

However (on my old account), I've been irritated by the cost and slowness of the machine learning studio, so I've been looking at the broader ML offering from Microsoft.  But I'm totally baffled by the range of jargon.  What's the difference between Azure ML Studio, experiment, experimentation (resource), project, workbench, workspace etc. etc. etc.?  

有些事情引起了我的兴趣，我想尝试理解:

A few things have caught my interest, that I want to try to understand: 

这里有CUDA的概念.我有一个合理的CUDA图形卡，因为前一段时间我以为我要做一些手动CUDA编码，但是那个项目从未成功.我记得最近读过，Microsoft ML的某些方面能够 可以使用我自己的本地CUDA卡，但是我再也找不到该文章了，我完全迷失了-有什么想法吗?

There's the concept of CUDA.  I have a reasonable CUDA graphics card because a while ago I thought I was going to do some manual CUDA coding but that project never took off.  I remember reading recently that some aspect of Microsoft ML is able to make use of my own local CUDA card, but I can't find that article any more and I'm totally lost - any ideas?

然后是Azure ML Workbench应用程序.我刚刚安装了此软件，但我并没有真正理解这一点.我很难遵循该文档，因为它假定对其他事物(例如docker)过于熟悉.我看到您可以配置 IDE"但是将其配置为可以做什么?它声称我可以将Visual Studio用作IDE(就我所知，这是我的VS)，但是我不喜欢我的可执行路径(对不起，不支持IDE路径")，并且帮助页面仅处理VS Code.或PyCharm， 我都不知道.

Then there's the Azure ML Workbench App.  I've just installed this, but I'm not really getting the point.  I struggle to follow the documentation because it assumes too much familiarity with other things (eg. docker).  I see you can "Configure an IDE" but configure it to do what?  It claims I can use Visual Studio as an IDE (which is my pref as I know VS) but it doesn't like my executable path ("Sorry, IDE path not supported"), and the help page only deals with VS Code or PyCharm, neither of which I know.

在Visual Studio中，我已经安装了适用于Visual Studio的AI工具，并且可以进入Gallery.但是当我尝试安装示例(例如虹膜分类器)时，对话框似乎不想连接到我的Azure帐户(VS知道是我)，因此我无法 选择一个帐户(为什么该帐户在资源组之后?)或工作区.我感觉它想让我使用CLI登录，但是怎么办呢?它在哪里?工作台应用程序和VS之间有什么联系?

Within Visual Studio, I've installed AI tools for Visual Studio and can get to the Gallery.  But when I try to install an example (e.g. the iris classifier) the dialog doesn't seem to want to connect to my Azure account (VS knows it's me), so I can't select an account (why is account after resource group?) or workspace.  I feel like it wants me to log in with a CLI but how? Where is it? What's the connection between the workbench app and VS?

似乎没有任何东西行之有效".而且我碰到了太多的行话-不知道什么是CNTK，Tensorflow或SPARK或Hadoop或这些东西中的任何一个.只是想训练一下大脑！

Nothing seems to *just work*.  And I run up against so much jargon - no idea what CNTK, Tensorflow or SPARK or Hadoop or any of these things are.  Just wanna train some brains!

最终，我只想要一种简单的方法来创建和训练神经网络/ML模型，希望在GUI中，希望使用我的CUDA卡，希望以一种连接到Azure的方式，因此如果我可以将其转换为Azure API曾经使任何有用的事情，并尽可能接近 尽可能免费. :)

Ultimately, I just want a simple way to create and train neural networks/ML models, hopefully in a GUI, hopefully using my CUDA card, hopefully in a way connected to Azure so I can turn it into an Azure API if I ever make anything useful, and for as close to free as possible.  :)

您能帮我弄清楚微软正在努力提供什么，以及为我充分利用它的最佳方法，以及如何使其正常运行"吗?！

Can you help me unravel what Microsoft is trying to offer, and the best way to make use of it for me, and how to make it *just work*?!

谢谢所有！

推荐答案

也许，来自产品团队的人将能够解决您对Microsoft产品的问题.

Perhaps, someone from the product team will be able to address your questions on the Microsoft product offerings. 

对于这个问题(在下面的斜体中)，您可以考虑使用深度学习虚拟机，这样您就不必担心任何安装: 
https://azuremarketplace.microsoft.com/zh-CN/marketplace/apps/microsoft-ads.dsvm-deep-learning
https://docs.microsoft.com/zh-CN/azure/machine-learning/data-science-virtual-machine/vm-do-ten-things

For this issue (in italics below), you could consider using the Deep Learning Virtual Machine, that way you need not worry about any installation: 
https://azuremarketplace.microsoft.com/en-ca/marketplace/apps/microsoft-ads.dsvm-deep-learning
https://docs.microsoft.com/en-us/azure/machine-learning/data-science-virtual-machine/vm-do-ten-things

Ultimately, I just want a simple way to create and train neural networks/ML models, hopefully in a GUI, hopefully using my CUDA card, hopefully in a way connected to Azure so I can turn it into an Azure API if I ever make anything useful, and for as close to free as possible.  :)

此致，
Jaya

Regards,
Jaya

这篇关于完全混淆了Re AML，VS，工作台，工作区，实验，项目，帐户，Arghhhh....的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！