[QUOTE=kriesel;532614]And I've been having a hard time getting a K80 on the sessions I'm using to benchmark gpuowl P-1 run time scaling and limit. When I started, K80 was all I was getting; now it's the exception.
It would be great for that if Google made it possible to specify what gpu model was required or preferred, instead of just none, gpu, or tpu.[/QUOTE]

Maybe they are taking feature requests on some forum or issue tracker?

gpu model in colab question
 

[QUOTE=preda;532619]Maybe they are taking feature requests on some forum or issue tracker?[/QUOTE]After a bit of searching it appears they are using stackoverflow. So:

[URL]https://stackoverflow.com/questions/59287301/in-google-colaboratory-is-there-a-way-to-indicate-preference-for-or-to-require[/URL]

5 views in first half hour but no replies yet.

[QUOTE=kriesel;532628]After a bit of searching it appears they are using stackoverflow. So:

[URL]https://stackoverflow.com/questions/59287301/in-google-colaboratory-is-there-a-way-to-indicate-preference-for-or-to-require[/URL]

5 views in first half hour but no replies yet.[/QUOTE]

If they do implement that, probably no one will use the K80 or P4 - and the P100 and T4's general availability might be significantly less due to everyone wanting those.

[QUOTE=kracker;532632]If they do implement that, probably no one will use the K80 or P4 - and the P100 and T4's general availability might be significantly less due to everyone wanting those.[/QUOTE]
Imagine the following choices, instead of the generic None, GPU, TPU:

None

GPU: P100 preferred (falls back to other gpu if no P100 available)
GPU: P100 required (backend availability fails if no P100 available)
GPU: T4 preferred (falls back to other gpu if no T4 available)
GPU: T4 required (backend availability fails if no T4 available)
GPU: K80 preferred (falls back to other gpu if no K80 available)
GPU: K80 required (backend availability fails if no K80 available)
GPU: P4 preferred (falls back to other gpu if no P4 available)
GPU: P4 required (backend availability fails if no P4 available)
etc. as new GPU models are added
maybe even P100 wait (waits until a P100 is available) etc or
GPU: Any required
GPU: Any optional (run the gpu task if one is available, run the cpu task in any event)
GPU: P100 optional (etc for other models too)

TPU

This is a new message to me:

[CODE]Beginning GPU Trial Factoring Environment Bootstrapping...
Please see https://www.gpu72.com/ for additional details.

20191212_091736: GPU72 TF V0.33 Bootstrap starting...
20191212_091736: Working as "...................."...

20191212_091736:
20191212_091736: *** FATAL ERROR ***
20191212_091736: GPU not found. Make sure GPU is enabled (Runtime menu -> Change runtime type).
[/CODE]

[QUOTE=bayanne;532705]This is a new message to me:[/QUOTE]

OK. I just restarted one of my running instances to make sure this wasn't something new. It isn't. (Was given a P100 instead of K80 during the exercise.)

What this message means is you tried running the TF Notebook in an instance without a GPU. This can happen if you accidentally hit "Connect" when prompted with the "No GPU Backend available, connect anyway?" message.

[QUOTE=chalsall;532717]What this message means is you tried running the TF Notebook in an instance without a GPU. This can happen if you accidentally hit "Connect" when prompted with the "No GPU Backend available, connect anyway?" message.[/QUOTE]

It also happened to me when Kaggle for some reason turned off my GPU selection and I had to reenable it.

[QUOTE=kriesel;532655]Imagine the following choices, instead of the generic None, GPU, TPU:

[/QUOTE]


It would also be nice if we could specify how many CPU cores we would like to deploy in our instances, depending on available resources.

[QUOTE=bayanne;532532]Sessions on Colab seem to have reduced from 12 hours to 10 hours now[/QUOTE]

But only for gpus, as far as I can see. I am currently getting 10 hours on gpus and 12 on cpus.

Anyone else seeing a blank page instead of your Kaggle kernels results? I can see my kernels list and I can edit / commit them, but I can't see the results of a commited run, it's just a blank page. It was working for me a few hours ago...

Same here. Blank page.

UPDATE: Working again now.