Platform: GCP ✅


(Fernando Melo) #334

@xtra_xtra_medium , You should not believe the error message “Try a different zone” , because I created a new vm from scratch on another zone and I got same error.


#335

I just ran into this issue, too. I’d be fine with choosing <Ok>, but I can’t figure out how in the cloud terminal. Urgh.


(Gaurav Desai) #336

Yes.How to choose OK…I am not able to figure that out.


(Ad Postma) #337

Have you tried upgrading your account?


(Ad Postma) #338

Have you tried to upgrade your account?


(Gaurav Desai) #339

Hi Arunoda,
Thanks for this utility. How to switch the instance location to europe-west1-b?
I do not see this location in fastai list -zones and I am not able to create instances in other zones.


(Arunoda Susiripala) #340

For the start, I only add zones where we’ve at-least 3 GPU options.
That’s why that zone is not available in the list.

I’ll add more zones soon.
Or you can send me a PR: https://github.com/arunoda/fastai-shell/blob/master/fastai.sh#L13


(Fernando Melo) #341

Yes. I have just upgraded it. SAME ERROR!
UNBELIEVABLE !!


(Bruce Holmer) #342

When you upgrade to a paid account, does the $300 credit (what’s left of it) disappear? Or is the credit still used to pay for time?


(Bruce Holmer) #343

I guess another alternative is to use an instance without the preemptible option.

I would like the capability to change this option on my current instance - so each time I start my instance, I choose preemptible or not. A quick google search came up with statements that this can’t be done. OK, so the next best thing would be to have two instances, one preemptible, the other not. But I want these two to share the same disk image. A quick search shows this is not possible.

I’m beginning to wonder if I’ll be able to even start up my current instance to grab all my data before moving on to some other platform.

Anyone have a good solution?


#345

i think you can connect cloud storage bucket to two of your instances and store data and code there

https://cloud.google.com/compute/docs/disks/gcs-buckets


(Robert Salita) #346

Thanks for providing additional credits. Unfortunately, I discovered your offer a day late. :frowning:


(Fadhli Ismail) #347

No. Your usage will be charge on that until it finish. If I’m not mistaken, they will inform you if your credit finish (nearly finish ?) before charging on your card.


(Edward Hiscoke) #348

I’m trying to develop a service as in the the course part 2 using Starlette on GCP and I wonder how to set up an IDE of some sort for editing. I tried using vim, enough said. If i’d been working locally I’d use pycharm or Visual Studio: can they be run on GCP somehow? Or is there some other online IDE like cloud9? Perhaps I’m missing something very obvious like an integration between my desktop IDE and GCP?

Grateful for any tips for where to get a better understanding for GCP cloud development for fastai.


#349

You could add a GUI to your GCP instance and then remote into it with VNC or NoMachine; I use the latter to remote into my home computer. Then you could use whatever IDE you want.


(Edward Hiscoke) #350

That’s an idea. I’ll give it a try!
I’ll also check out python anywhere and run the inference in a separate environment.


(Bruce Holmer) #351

Upgrading seems to have eliminated the quota exceeded error for me.


(Christian Werner) #352

Thanks for the xtra credits. I sure can use them :wink:


(Vijay Narayanan Parakimeethal) #353

Thanks @noskill. I tried this as well but it is not working.


(noskill) #354

Hey hanqi,
Tensorflow-gpu will automatically pull the compatible cuda library version. Export commad has nothing to do with this. In short, you won’t have any problem with this setup. :slight_smile: