Google Diet Colab Notebook - Page 74

ric · 2020-01-10, 23:58

Quote:

Originally Posted by Prime95

I also suspect they recently instituted a lifetime cap on GPU time.

Hope they didn't. FWIW, in the last couple of days, I've been receiving a couple of 10 hr sessions per night (down from 4/5 in the golden days), starting around 11PM/midnight my time. Fun thing, on the same accounts I've been squeezing the most during these months, while other and less used accounts do not succeed.

Another factual observation: brand new accounts receive an initial 10/12 hr allotment, and do not reconnect any further.

Whack-a-mole, anyone?

James Heinrich · 2020-01-11, 00:02

Quote:

Originally Posted by Prime95

I suspect Google keeps track of how much total GPU time your account has used. I also suspect they recently instituted a lifetime cap on GPU time.

I could believe that, except I've always been launching Colab sessions near-simultaneously on both accounts (when I remember to reconnect them twice a day or so), and until now I've (almost) never been able to get connected on one but not the other. But for the last 2-3 days my first account can never get a GPU, and the second account always gets a GPU. Strange. Perhaps they're limiting it to one GPU per IP? That would make more sense.

kriesel · 2020-01-11, 15:16

Quote:

Originally Posted by James Heinrich

I could believe that, except I've always been launching Colab sessions near-simultaneously on both accounts (when I remember to reconnect them twice a day or so), and until now I've (almost) never been able to get connected on one but not the other. But for the last 2-3 days my first account can never get a GPU, and the second account always gets a GPU. Strange. Perhaps they're limiting it to one GPU per IP? That would make more sense.

I'm running multiple accounts from different hosts, one account per host, behind a NAT firewall. (although there's no NAT in the IPV6 protocol) Getting multiple simultaneous Colab gpus often.

petrw1 · 2020-01-11, 16:11

Am I the only one?
Could it be because I run it from my Phone?

Uncwilly · 2020-01-12, 00:25

Quote:

Originally Posted by petrw1

Am I the only one?
Could it be because I run it from my Phone?

I have one account that is only attached to a phone that hasn't gotten much in days. The other I use a desktop with a non-"gmail" log-in (it is corporate that uses an enhanced/managed version of g-mail and g-suite), also on a phone that is attached to that same account.

chalsall · 2020-01-12, 16:18

Quote:

Originally Posted by Uncwilly

I have one account that is only attached to a phone that hasn't gotten much in days.

Yup... I'm afraid it's pretty clear the "gig" is up... Only five people have had any instances in the last three days, and all but one (Wayne) lasted for only a few minutes.

Oh, well... It was fun while it lasted.

Dylan14 · 2020-01-12, 18:00

Quote:

Originally Posted by chalsall

Yup... I'm afraid it's pretty clear the "gig" is up... Only five people have had any instances in the last three days, and all but one (Wayne) lasted for only a few minutes.

Oh, well... It was fun while it lasted.

I've actually been reasonably successful in getting GPU's on Colab, but I've been running the tf1G script, not the gpu72 script. Usually I wait a day or two if I get the message that there are no GPU's available.

bayanne · 2020-01-13, 14:05

I have one exponent left unfinished from Colab situation (88.72% completed).

Is there a way to transfer this exponent to my normal crunching so as to not waste that worked percentage?

xx005fs · 2020-01-13, 14:16

Quote:

Originally Posted by bayanne

I have one exponent left unfinished from Colab situation (88.72% completed).

Is there a way to transfer this exponent to my normal crunching so as to not waste that worked percentage?

Just download the checkpoint from your google drive and run it on a local instance.

chalsall · 2020-01-13, 14:44

Quote:

Originally Posted by bayanne

Is there a way to transfer this exponent to my normal crunching so as to not waste that worked percentage?

I just PM'ed you the file's contents.

James Heinrich · 2020-01-13, 17:34

If Colab is truly "over", can someone volunteer to clean up the half-done exponents? I apparently have one at 91.6%, if Chris can send the checkpoint to whoever wants to finish it for me (I care nothing for credit).

2020-01-11, 16:11	#807
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 2·5·7·67 Posts	I continue to get Kaggle GPUs for 30+ hours a week Am I the only one? Could it be because I run it from my Phone?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Alternatives to Google Colab	kriesel	Cloud Computing	11	2020-01-14 18:45
Notebook	enzocreti	enzocreti	0	2019-02-15 08:20
Computer Diet causes Machine Check Exception -- need heuristics help	Christenson	Hardware	32	2011-12-25 08:17
Computer diet - Need help	garo	Hardware	41	2011-10-06 04:06
Workunit diet ?	dsouza123	NFSNET Discussion	5	2004-02-27 00:42

2020-01-13, 14:05	#811
bayanne "Tony Gott" Aug 2002 Yell, Shetland, UK 2²×83 Posts	I have one exponent left unfinished from Colab situation (88.72% completed). Is there a way to transfer this exponent to my normal crunching so as to not waste that worked percentage?

2020-01-13, 17:34	#814
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 11·311 Posts	If Colab is truly "over", can someone volunteer to clean up the half-done exponents? I apparently have one at 91.6%, if Chris can send the checkpoint to whoever wants to finish it for me (I care nothing for credit).