1 00:00:00,390 --> 00:00:01,110 Welcome back. 2 00:00:01,110 --> 00:00:09,450 So the last video we made sure that our CO lab notebook runtime or the machine that lives on Google's 3 00:00:09,630 --> 00:00:16,130 cloud server that our notebook has been run on is using a GP you to power its computing. 4 00:00:16,200 --> 00:00:22,860 Now I just want to show you something pretty cool about Ecolab as well because these notebooks are shareable. 5 00:00:22,860 --> 00:00:25,260 You can run them almost instantaneously. 6 00:00:25,260 --> 00:00:30,330 So this is optional video if you want to just go back into getting our data ready go skip this one and 7 00:00:30,330 --> 00:00:37,200 going into the next video to show you an example of a shareable Google lab notebook as well as how much 8 00:00:37,260 --> 00:00:39,960 a GP you actually speeds up compute. 9 00:00:39,960 --> 00:00:46,560 So if we go to a new tab collab researched up Google dot com if we were to cancel out of this the home 10 00:00:46,560 --> 00:00:50,730 page of Google collab is actually a notebook in itself. 11 00:00:50,730 --> 00:00:56,640 So that's the amazing thing you can copy that to your google drive or we can connect this to a runtime. 12 00:00:57,150 --> 00:01:02,460 And now a little bit of extra curriculum would be to go through this notebook here and just get familiar 13 00:01:02,460 --> 00:01:09,150 with Google Ecolab like there's a lot here but you just saw that I connected that to a runtime and I 14 00:01:09,150 --> 00:01:14,220 can immediately start running code in this notebook. 15 00:01:14,220 --> 00:01:20,190 How beautiful is that so if we go here this is going to input number hi Matt Gottlieb input pipeline 16 00:01:20,200 --> 00:01:28,070 let us take no setup I just connected this to a runtime that is amazing and now there's a little section 17 00:01:28,070 --> 00:01:30,050 in here we go. 18 00:01:30,150 --> 00:01:33,820 Years and years tend to flow with GP use. 19 00:01:33,830 --> 00:01:37,790 So this is what we just did in the last video in our runtime we're using a GPO. 20 00:01:37,790 --> 00:01:40,180 You can see that if I move the mouse is going to go away. 21 00:01:40,220 --> 00:01:43,560 We're using a GP you on this dog vision notebook. 22 00:01:43,560 --> 00:01:49,190 But if we click here this little guide this is gonna take us to this Web site here all the touch. 23 00:01:49,190 --> 00:01:53,330 Both of these by the way to the resources section. 24 00:01:53,330 --> 00:02:00,420 Again we just connect our runtime up here is gonna give us a little insight into how fast tensor flow 25 00:02:00,420 --> 00:02:09,940 is with the GPO and without a GPO so if we go here run this cell it's gonna make sure we use intensive 26 00:02:09,950 --> 00:02:11,570 flow to point something. 27 00:02:11,990 --> 00:02:15,110 And this is just a line of code to make sure it's very similar. 28 00:02:15,410 --> 00:02:17,480 Oh JP you're not found. 29 00:02:17,600 --> 00:02:19,240 Why is this. 30 00:02:19,310 --> 00:02:23,140 Let's make sure it is change runtime type. 31 00:02:23,310 --> 00:02:29,490 It may be because I'm using a GP you in double vision you might be limited to only one but let's say 32 00:02:30,180 --> 00:02:34,150 connecting how beautiful is that. 33 00:02:34,720 --> 00:02:39,410 So if we run this says that we've got a GP you beautiful. 34 00:02:39,410 --> 00:02:42,010 So it's found a GP you at device GP use zero. 35 00:02:42,910 --> 00:02:47,720 So if we go here this is observed speed up on GP relative to CPR. 36 00:02:47,720 --> 00:02:50,210 I'm gonna close this table of contents we don't need that. 37 00:02:50,260 --> 00:02:52,270 So we've got a couple of functions here. 38 00:02:52,340 --> 00:02:58,970 What this is basically saying is run some code on the CPA and run the same code on the GP use. 39 00:02:58,990 --> 00:03:04,600 There you go with TAF device with a device GPO see you up there. 40 00:03:05,620 --> 00:03:07,380 So let's say shift in a.. 41 00:03:07,390 --> 00:03:08,620 Run this. 42 00:03:08,630 --> 00:03:11,620 Just gonna run some sort of loop a number of times. 43 00:03:11,770 --> 00:03:13,920 Shouldn't take too long convulsive. 44 00:03:14,320 --> 00:03:15,340 There we go. 45 00:03:15,340 --> 00:03:19,400 Just gonna involve a filter over some random images such as processing images. 46 00:03:19,420 --> 00:03:21,120 That's exactly what we want to do right. 47 00:03:21,130 --> 00:03:26,320 Look at that time to involve 32 by 7 by 7 by three filter over. 48 00:03:26,350 --> 00:03:34,190 So this is just basically processed some images of this over one hundred or so images ten times the 49 00:03:34,190 --> 00:03:40,640 CPA took three point four seconds the GPO took zero point one seconds of just about zero point one seconds 50 00:03:41,090 --> 00:03:44,160 so 30 times faster. 51 00:03:44,400 --> 00:03:45,930 Isn't that beautiful. 52 00:03:45,930 --> 00:03:52,030 First of all the fact how instantly we can connect to using such fast computing in Google collab. 53 00:03:52,320 --> 00:04:00,290 And second of all if we use a GP you we get a speed up of 30 x I just think that's amazing. 54 00:04:00,340 --> 00:04:06,090 And now again remember any of these tutorials in Google collab if in doubt run the code connect to a 55 00:04:06,090 --> 00:04:09,870 runtime and try it out run it for yourself. 56 00:04:09,870 --> 00:04:10,650 That's the motto right. 57 00:04:10,680 --> 00:04:12,390 If in doubt run the code. 58 00:04:12,390 --> 00:04:18,240 But again this was just an optional video just to show you the power of Google collab and the power 59 00:04:18,870 --> 00:04:19,970 of a GP you. 60 00:04:20,120 --> 00:04:22,530 Now we know that what a GP you can do. 61 00:04:22,530 --> 00:04:26,100 Let's get back into our dog vision notebook and use one.