| 1 | |
| 00:00:00,430 --> 00:00:00,850 | |
| OK. | |
| 2 | |
| 00:00:00,880 --> 00:00:04,340 | |
| So let's talk about how we save our models that we just trained. | |
| 3 | |
| 00:00:04,380 --> 00:00:08,020 | |
| We spend some time training a model and we don't want to lose this model in clothes. | |
| 4 | |
| 00:00:08,040 --> 00:00:09,000 | |
| I buy the book. | |
| 5 | |
| 00:00:09,000 --> 00:00:13,350 | |
| So let's figure out how we save and load our models. | |
| 6 | |
| 00:00:13,390 --> 00:00:15,790 | |
| OK so we just spend so much time printing our model. | |
| 7 | |
| 00:00:15,810 --> 00:00:17,890 | |
| So let's get to saving it now. | |
| 8 | |
| 00:00:17,930 --> 00:00:24,780 | |
| So if you scroll down below these charts to see if she's part of the model object you just specified | |
| 9 | |
| 00:00:24,830 --> 00:00:27,000 | |
| directory and file him as you want. | |
| 10 | |
| 00:00:27,000 --> 00:00:33,570 | |
| So basically we can just call this chapter 8 which is from 8 here eminence simple CNN Tennie box can | |
| 11 | |
| 00:00:33,570 --> 00:00:36,080 | |
| actually put on display here make it look a little bit nicer. | |
| 12 | |
| 00:00:36,510 --> 00:00:39,780 | |
| And voila it saved instantly. | |
| 13 | |
| 00:00:39,780 --> 00:00:46,770 | |
| So now if you want to redo it on what all this is into and you found him here and we use Chris's model | |
| 14 | |
| 00:00:46,890 --> 00:00:48,720 | |
| which is part of the models library here. | |
| 15 | |
| 00:00:49,200 --> 00:00:55,890 | |
| And this basically this is what's strange to most people loading isn't as instantaneous as saving and | |
| 16 | |
| 00:00:55,890 --> 00:01:00,990 | |
| that's because it has to actually unpack everything now and put it back into the model format. | |
| 17 | |
| 00:01:01,020 --> 00:01:06,150 | |
| It's a bit tricky to understand I actually spend some time trying to figure it out because I had a project | |
| 18 | |
| 00:01:06,150 --> 00:01:11,840 | |
| where I was loading 50 different models at once and it was taking maybe about five 10 minutes to live | |
| 19 | |
| 00:01:11,840 --> 00:01:13,130 | |
| with all of them together. | |
| 20 | |
| 00:01:13,530 --> 00:01:14,940 | |
| So let's just run this quickly. | |
| 21 | |
| 00:01:15,600 --> 00:01:16,510 | |
| And there we go. | |
| 22 | |
| 00:01:16,520 --> 00:01:17,930 | |
| Stick what 15 seconds. | |
| 23 | |
| 00:01:18,340 --> 00:01:22,860 | |
| And so there's no snow or model which was called the model before. | |
| 24 | |
| 00:01:22,890 --> 00:01:24,280 | |
| It's called classifier. | |
| 25 | |
| 00:01:24,690 --> 00:01:30,270 | |
| And now we can use it here classify it just like we were using model before they are interchangeable | |
| 26 | |
| 00:01:30,270 --> 00:01:33,080 | |
| now because our model is effectively classified. | |
| 27 | |
| 00:01:33,080 --> 00:01:34,660 | |
| It's just a different name. | |
| 28 | |
| 00:01:34,680 --> 00:01:40,510 | |
| What I mean by that is you see how we declared this as a model equals sequential No what we've related | |
| 29 | |
| 00:01:40,510 --> 00:01:41,220 | |
| to our model. | |
| 30 | |
| 00:01:41,220 --> 00:01:44,160 | |
| I didn't want to overwrite what we had see if it was a model. | |
| 31 | |
| 00:01:44,250 --> 00:01:48,160 | |
| So effectively model is now as we have seen it and loaded it. | |
| 32 | |
| 00:01:48,180 --> 00:01:52,730 | |
| It's now effectively Plus a classified called classify. | |
| 33 | |
| 00:01:52,770 --> 00:01:53,600 | |
| Should say. | |
| 34 | |
| 00:01:53,880 --> 00:01:57,630 | |
| And we can load it here and basically run some tests. | |
| 35 | |
| 00:01:57,640 --> 00:01:58,800 | |
| This has an error in it. | |
| 36 | |
| 00:01:58,800 --> 00:02:01,280 | |
| Him just to leave this quickly. | |
| 37 | |
| 00:02:02,560 --> 00:02:04,940 | |
| And it should work fine now. | |
| 38 | |
| 00:02:04,980 --> 00:02:05,490 | |
| There we go. | |
| 39 | |
| 00:02:05,570 --> 00:02:07,940 | |
| So we're actually running some test animals here. | |
| 40 | |
| 00:02:07,940 --> 00:02:09,630 | |
| So this is a zero a tree. | |
| 41 | |
| 00:02:09,650 --> 00:02:13,740 | |
| It can see how accurate this model is. | |
| 42 | |
| 00:02:13,760 --> 00:02:14,530 | |
| It's quite cool. | |
| 43 | |
| 00:02:15,500 --> 00:02:19,050 | |
| And it actually looks and what it actually is because it's a bit tilted. | |
| 44 | |
| 00:02:19,430 --> 00:02:20,130 | |
| So that's good. | |
| 45 | |
| 00:02:20,200 --> 00:02:20,660 | |
| It's great. | |
| 46 | |
| 00:02:20,660 --> 00:02:27,740 | |
| Actually we've just trained we've just loaded and trained and tested and plotted and analyzed and done | |
| 47 | |
| 00:02:27,800 --> 00:02:30,470 | |
| all of the settings for this model. | |
| 48 | |
| 00:02:30,470 --> 00:02:34,940 | |
| So if you wanted to put it all together because you don't actually need to have every block separate | |
| 49 | |
| 00:02:34,940 --> 00:02:40,130 | |
| like this this was basically for teaching purposes I actually never separate these things that much | |
| 50 | |
| 00:02:42,060 --> 00:02:47,780 | |
| but let's put it all together so quickly you can see this is everything here and you run this and you | |
| 51 | |
| 00:02:47,780 --> 00:02:49,990 | |
| basically start training instantaneously. | |