What's the hardware spec for Google Colaboratory?

Question

I load some (not so) big data into it. A compressed file at 9GB. It can't decompress because the disk space is not enough.

What's the current hardware spec? What's the disk size?

Bob Smith · Accepted Answer · 2020-07-16T16:25:06.093

44

You can see the disk information using a command like !df -h. To see CPU specs, !cat /proc/cpuinfo. For memory, !cat /proc/meminfo.

Edit: Colab now offers a Pro version which offers double the amount of disk available in the free version.

edited Jul 16 '20 at 16:25

answered Dec 14 '17 at 18:10

Bob Smith

korakot · Answer 2 · 2020-05-03T05:53:15.067

27

Here I run some test.

In short:

2020 Update:

edited May 03 '20 at 05:53

answered Dec 14 '17 at 02:54

korakot

what does idle cut-off and maximum time mean? – laltin Oct 15 '18 at 08:21
1

@laltin If you close the browser, it will stop in 90 minutes. If you open the browser, it will stop in 12 hours. – korakot Oct 15 '18 at 09:03
I can use it only 12 hours? – Mikhail_Sam Nov 13 '18 at 21:13
3

@Mikhail_Sam you can save the result (like every hour), then restart training from your last checkpoint. – korakot Nov 14 '18 at 04:40
@KorakotChaovavanich kindly provide the code or something to store and restore the model in case my model takes more than 12 hours to train... – yamini goel Feb 15 '19 at 09:16
Sometimes I see 2.2GHz chip speed and sometimes I see 2.3GHz. – jeffhale Mar 12 '19 at 22:12
1

"GPU instance upgraded to 350 GB" of free space? – SantoshGupta7 Apr 13 '19 at 22:17
@KorakotChaovavanich How can I save my model in tensorflow every 30 minutes? Any sample code? – chikitin Nov 09 '19 at 23:51
@KorakotChaovavanich How did you found the idle cut-off time? Your colab notebook does not return a value for that? – chikitin Nov 09 '19 at 23:53
@KorakotChaovavanich It seems it is not always the case. I am on MacOS and have my browser Chrome open and have my computer set to never sleep ( But I lock it ), however, after 90 minutes it tells me Runtime disconnected. – chikitin Nov 09 '19 at 23:55
Thanks @korakot for these updated specs. When you say “GPU instance upgraded to 350 GB” is that the GPU RAM or hard disk? – Leockl May 03 '20 at 03:19

Anirudh Chandupatla · Answer 3 · 2018-08-03T16:39:28.500

23

I ran few tests and found ,

GPU: 1xTesla K80 , compute 3.7, having 2496 CUDA cores , 12GB GDDR5 VRAM

CPU: 1xsingle core hyper threaded Xeon Processors @2.3Ghz i.e(1 core, 2 threads)

RAM: ~12.6 GB Available

Disk: ~33 GB Available

edited Aug 03 '18 at 16:39

answered Aug 03 '18 at 02:12

Thanks for the code. Is it possible to find the information about OS if the VM? – chikitin Nov 09 '19 at 23:51
Hi @Anirudh Chandupatla, when you say “2496 CUDA cores”, does that mean there are 2496 GPU cores? I have read somewhere that the free version of Google Colab only has a single (ie. 1) GPU core, though I am not sure how updated this is – Leockl May 03 '20 at 03:22
@Leockl Single GPU has multiple CUDA cores. It's like single CPU has multiple cores (around 4). Also, using single CUDA core simply does not make sense, as that would make GPU slower than CPU. – quartzsaber Jun 28 '20 at 02:26

3 Answers3