Cannot use Azure Free Trial (no GPU machines available)


#1

I am trying to set-up my environment.

In the Lesson HOW TO SET UP YOUR AWS DEEP LEARNING SERVER it is mentioned > We’ve just learned that Azure now allows new users to access GPU machines without requiring a special request, and will provide $200 of free credits <

Unfortunately I do not think this is possible when you go for a free trial subscription, because all GPU machines have 6+ cores and the core quota for a free trial account is 4.

Maybe it works if you choose a Pay-As-You-Go subscription. I tried to change my Azure subscription from Free trial to Pay-As-You-Go but I could not.

I have been following the steup guide by @complancoder (mentioned in the lesson).

Any insight on the topic?


(Neil Swinton) #2

It’s not unusual to have multiple subscriptions in Azure. You can add one via this link: https://portal.azure.com/#blade/Microsoft_Azure_Billing/BillingBlade/Overview. I added an additional subscription instead of upgrading my free one. (Oddly, the web page was English with a little Portuguese mixed in – avaliação means “evaluation”, that’s the free one.)


(Jeremy Howard) #3

I’ll pass on this issue to our contact at Azure - I know they’re interested in debugging any problems people see.

Have you checked that:

  • You have chosen ‘HDD’ and not ‘SSD’
  • You have selected US East

#4

Yes @jeremy, I chose HHD as well as US east.

The option “NC6” shows up, but I cannot select it and the unavailability message mentions that it’s because it has 6 cores, while free trial only allows to use up to 4 cores.


(Jeremy Howard) #5

@elena294 what happened when you tried to switch to PAYG?

(BTW, I had exactly the same problem as you, so it’s nice to know I’m not alone! Hopefully we’ll get this sorted with the Azure team’s help…)


#6

@jeremy
When I tried to switch to PAYG, the only option I was given was to accept to start paying with my credit card any expense I would accrue after the 200$ credit expired.
Once I accepted this condition, my subscription was automatically renamed ‘Pay As You Go’ and I got a confirmation email, but my subscription information was still showing ‘free trial’ and NC6 was still unavailable.


(Abhik Mitra) #7

@elena294 Ugh! I am a Microsoft Employee so I am on the Microsoft Employee Subscription hence I did not get into this issue . I didn’t know employees had it any different . I will updated my post with this ! :frowning:


(Matt) #8

Hi @elena294 @Jeremy, I have finally established that when you convert a free trial to PAYG with a CC, it still retains the same core allocation policy. While we try to fix this, please raise a support ticket to request additional cores (to the number you require). It is not ideal and will add some time to your deployment, but is the only work around in the short term. Apologies in advance.


(Jeremy Howard) #9

I’m glad we resolved this - I will remove the recommendation to use Azure from our site for now. Let us know when it’s fixed. :slight_smile:


(Matt) #10

I would recommend that if people currently wish to use the N-Series GPU to either proceed with a PAYG account from the start and skip the free trial, or if they have started a free trial, raise a support ticket through the Azure Portal to have to core allocation increased.


(Carl) #11

Hi,

Noticed that Azure had a Windows alternative called DSVM (Data Science Virtual Machine) and was wondering if any of you knew if that might be appropriate for this course, can’t see if it have NVidia but I think it comes prepared with most of the tools used in the course.

Is the Ubuntu (NC6 option) otherwise still an good option, or are there any other (would prefer Windows) alternatives on Azure?

Thanks!


(Mukul Gupta) #12

Hi,

I am facing the same issue after upgrading from free trial to Pay-As-You-Go. Could you please help me on how you raise the ticket for raise the number of cores. When I am trying to create a ticket for increasing cores under the category ‘quota’, I must enter SKU family, but NC is not listed there.

Thanks!


(Pierre Guillou) #13

Hi @jeremy,

I would like to send a GPU credit application to AWS, Microsoft Azur, Google Cloud and NVIDIA GPU Cloud in favor of the Fastai study group in Brasilia that I lead with Brazilian colleagues (I had the list of their profiles in the forum: 69 to today on the 80 participants in our course meetings and a larger number through the list of emails).

Could you help me send this request to them ? (or could I get name and emails from you ?)

Thanks if you can help.

Pierre