How to setup virtual machine on Azure for running course lessons?

[Chris_Palmer]

Hi @vrajjshah

How did you get in with running Azure - I also have access to the VS subscription and so am wondering if its a worthwhile resource for our Pytorch learning once the AWS credit runs out.

[manikanta_s (Manikanta Yadunanda Sangu)]

Hi @Chris_Palmer, I am able to get Azure running with our fast ai using the Visual Studio Subscription. If you need any info on the same, let me know.

[vrajjshah (Vraj Shah)]

Hey @manikanta_s could you share the steps here ? So even I can get to know how to get azure running(for fast ai) with my visual studio enterprise subscription? Even others can benefits who have the same question

[manikanta_s (Manikanta Yadunanda Sangu)]

Sure, will put up a small post and share it.

[Chris_Palmer]

Just saw this @manikanta_s - I sure would like the information - please send it along or post for general consumption!

[manikanta_s (Manikanta Yadunanda Sangu)]

I have created the below post with necessary details.

Guide to setup fastai on Azure

Hi, I have just created a post on setting up fastai on Azure. I hope this will be helpful for people facing issues with Azure and people who already have visual studio subscription. The process is relatively simple now with recent Azure Deep Learning Virtual Machine offering. https://medium.com/@manikantayadunanda/setting-up-deeplearning-machine-and-fast-ai-on-azure-a22eb6bd6429 Feedback is appreciated.

[skbisoi (Susant Bisoi)]

I followed the Azure GPU setup process and i am stuck here in the red colored portion in below image.

my question SSH to ubuntu VM means connect to the ubuntu VM hosted in Azure cloud through linux bash(here cygin has some issue so using github bash) like below

So while connection it is asking password .So what is password i should give ?
Is it asking key in the “id_rsa”?
any help will be appreciated.

[skbisoi (Susant Bisoi)]

I gave my password to azure account and connected .After that i need to install the script in the linux/ubuntu vm .So how do that?I am new to linux.(DO i have to connect through remote desktop to the linux VM and then install the script or through terminal (here for me GIT bash)?

[Jefferson (Jefferson Braga)]

Hello Chris,
Were you able to get Azure to run? I got stuck when setting up the Deep Learning DC VM.

[Chris_Palmer]

Hi Jefferson

No, I haven’t even had time to try it yet, but I do intend to try soon. I’ll let you know if I make any progress…

[LeNDuyA]

Hi, I managed to create a VM (Deep Learning template).

So now I start a jupyter server using jupyter notebook --no-browser which gives me a link. I click on that link but the browser (Chrome) says The site can't be reached. I tried replacing localhost with the IP address but the output is still the same.

Any help?

[Chris_Palmer]

Finally got around to trying this - but unfortunately all options were grayed out and were very expensive anyway - even the NC6 was over $800 a month. I made a couple of requests for access but so far nothing has been granted. Might be a waste of time anyway with those prices…

[skbisoi (Susant Bisoi)]

You have to send email to MS office for enabling them the NC6 machines.Although it is is showing the 800$ …u will not that much .If u have leaready have MSDN account then it comes 150$ of monthly free credit and no credit card nor bank account details required.So dont worry just send amail to MS team.I am using azure through MSDN Visual Studio Enterprise account.

[skbisoi (Susant Bisoi)]

NC6 Standard option greyed out-Microsoft Azure-MSDN Professional

Just correcting myself here. GPU resources are available if you have it enabled through support as @skbisoi points out. Just raise a request and they will help you out. Also select HDD as the disk type when provisioning NC6.

follow the above link

[Chris_Palmer]

Thanks Susant,

I did send a message, but they haven’t granted it to me! Not sure why - I will try again.

But the other issue I could see is that these use Python 3.5 and CUDA 8, whereas the fast.ai requirements is 3.6 and CUDA 9. Are you able to upgrade the VM?

[skbisoi (Susant Bisoi)]

They will reply …it will take may 2-3 days .for me it took 3 days.

I am here in python 2.7 and heard that 2018 v1 is using python 3.6 but when i took the batch script for azure it is automatically took python 2.7 and i want to update also python 3.6
I am stuck with other issue like “import utils; reload(utils)
from utils import plots” although i ama bale to run the lession.ipnb and able to go upto some level.\

[Chris_Palmer]

Thanks Susant

Did you build the fast.ai environment using conda env?

[skbisoi (Susant Bisoi)]

No i am using Git Bash for terminal

[Chris_Palmer]

What I meant was, after doing a git clone to get the fast.ai library, did you then execute a conda env create -f environment.yml while in the newly created fastai directory, so that you could use the environment.yml file supplied by fastai?

[skbisoi (Susant Bisoi)]

In Short answer YES…

Long answer below steps i followed…

After connecting to SSH to ubuntu VM(DL VM is hosted in azure cloud and using MS DL VM ) through GIT Bash terminal …
I followed below steps…

Now, move into a directory where you are comfortable installing the Fastai repo, with its libraries and required packages.I did with default directory so default command i am using …
Now you got to clone that repo as follows:

git clone https://github.com/fastai/fastai
Once the cloning process finishes, be sure to be in the directory created by git for the Fastai repository, and type:
conda env create -f fastai/environment.yml

why absolute path of environment.yml(ie. fastai/environment.yml) required because of the below error. error-https://github.com/conda/conda/issues/3847

then activate conda environment like using below command

source activate fastai
why prefix ‘source’ is required in above command due to below error

Permission denied when activating venv

now after activation of fastai conda environment type below command to open the jupyter notebook

(fastai)> jupyter notebook