Lesson "0": Practical Deep Learning for Coders (fast.ai)

00:00:00.000 | OK, hi everybody, thanks for joining.

00:00:07.300 | This is an entirely optional presentation, which I'll call a Lesson 0, which is all about

00:00:14.820 | how to fast AI.

00:00:18.220 | It's all about how to get the most out of this course, how to make sure you finish it,

00:00:24.640 | and how to make sure you feel like it's been a productive time.

00:00:31.800 | And the reason I'm doing this is because a lot of people who take the course, when they

00:00:38.780 | get to the end of it, they say to me, "Oh, it wasn't until I got to the end of the course

00:00:42.600 | that I realized how I should have done the whole course, and now I'm going to go back

00:00:45.240 | and redo the whole thing over again."

00:00:48.080 | And so I'm going to tell you about what the messages I've heard are about what people

00:00:54.040 | have found the best approaches to making the course work.

00:00:59.160 | I'm also going to go through the actual mechanics of how to get set up with two systems, Google

00:01:07.200 | Colab and AWS EC2, and I'll talk about why you might use one versus the other.

00:01:14.880 | So a lot of people now, as in many hundreds of thousands, have gone through the fast AI

00:01:20.600 | practical deep learning for coders course, and many, many, many of them have gone on

00:01:26.680 | to create successful startups, to write research papers with high impact factors, to create

00:01:35.400 | new products at their companies.

00:01:38.080 | It's a pretty well proven course at this time, but there's also a lot of people that never

00:01:44.240 | finish the course.

00:01:46.680 | And so if you're watching this, it's because you've decided you do want to learn deep learning,

00:01:53.440 | so I'm going to talk a bit about like what's it going to take for you to be one of the

00:01:57.120 | people that makes this into a great experience.

00:01:59.960 | When I talk about the course, I'm also talking about the book.

00:02:05.360 | So just to be clear, there's a book that Sylvain Goudre and I wrote, which you can either

00:02:11.920 | buy from Amazon, and people like it happily, or believe it or not, you can read the whole

00:02:19.120 | thing for free.

00:02:21.420 | So it's called Fastbook, it's a fastbook repo.

00:02:25.000 | Honestly, I make basically nothing from the book, so I don't feel like you need to buy

00:02:30.320 | it to say thank you or something, buy it if you want the book.

00:02:33.800 | If you're happy using notebooks, use the free one, it's all good.

00:02:39.800 | So the book was actually written as Jupyter notebooks, and we wrote something to turn

00:02:44.360 | it into a book book.

00:02:46.000 | Now the book will also, by the way, actually looks great on Kindle, online, as well as

00:02:50.280 | paper.

00:02:51.280 | I know often technical books don't, this one actually does.

00:02:57.400 | And then the course goes through half of the book.

00:03:02.440 | And so quite soon we'll do a part two, which will go through the other half of the book

00:03:06.840 | plus some other new stuff.

00:03:09.640 | But basically each lesson covers a chapter or so of the book.

00:03:14.980 | So if you're doing this course, you'll be going through the book, at least in the notebooks

00:03:19.100 | and you might want the paper one as well.

00:03:22.480 | So here is the main thing that you should commit to right now, which is to finish the

00:03:30.240 | damn course, right, and or finish at least half of the book.

00:03:37.160 | Because everybody I think who joins comes in thinking, okay, I'm going to do this, I'm

00:03:41.080 | going to do deep learning, but if you, when I look at our YouTube analytics, a lot of

00:03:45.880 | people don't finish, okay?

00:03:48.420 | So you just need to decide what day are you going to watch the course each week?

00:03:54.160 | What day are you going to do the assignment?

00:03:55.840 | What day, like how are you going to stretch your time to finish the course?

00:03:58.920 | And maybe you're coming in deciding, I don't want to finish it, which is fine, right?

00:04:04.880 | If that's your intention up front, no problem.

00:04:07.400 | But if your intention is to be a really effective deep learning practitioner, you need to finish

00:04:15.800 | the damn course, okay?

00:04:17.600 | So put it in your head that that's your goal.

00:04:20.520 | Talk to your friends or your spouse and tell them that's my goal.

00:04:23.600 | Get that social pressure that you're going to finish it.

00:04:27.440 | You're not just going to finish the course, but try to finish a project, right?

00:04:33.760 | So Christine McClavy is one of our fantastic alumni.

00:04:37.740 | She's now at OpenAI, one of the world's top research organizations.

00:04:43.120 | She built a fantastic system for creating new music with deep learning.

00:04:48.800 | She used to be a pianist herself.

00:04:52.200 | And I remember this discussion, I told her, focus on making one project great and polishing

00:05:01.680 | it off and finishing it.

00:05:05.440 | And she did.

00:05:07.200 | And that project has ended up creating music which the BBC Orchestra played, right?

00:05:14.440 | And amongst other things helped her get this extremely exclusive job at OpenAI.

00:05:22.660 | So this is a clip from a podcast with one of our students, Sanyam and Christine, in which

00:05:29.600 | Christine is saying this is one of her key insights.

00:05:33.400 | And so I've got to be giving you a few key insights, some of which are from me or some

00:05:37.560 | of them are from me via students, but they're all like things I've heard a bunch of times.

00:05:43.080 | So this is one example.

00:05:44.860 | So finish the course and finish a project.

00:05:49.160 | The project doesn't have to be something no one's ever built before.

00:05:52.820 | Maybe it's just like, oh, I really love that thing that person built.

00:05:56.360 | Gosh, it would be a real stretch if I could build it too.

00:06:00.840 | You know, great.

00:06:02.940 | Or it doesn't have to be world changing, you know.

00:06:05.620 | So one of our students built something for his fiancé, which was a cousin recognizer.

00:06:13.080 | He had, I think, 14 cousins.

00:06:15.520 | And so his fiancé could take a picture of one of the cousins and it would tell them

00:06:20.500 | which cousin it was, right?

00:06:24.080 | In our first course, one of our students built the app for the Silicon Valley TV show which

00:06:31.740 | did Hot Dog or Not Hot Dog, which was actually a huge smash hit, like millions of downloads

00:06:38.320 | that it was written about in the media.

00:06:40.080 | And it did exactly one thing just to tell you whether or not something was a hot dog.

00:06:45.000 | Anyway, or it could, you know, solve medicine.

00:06:50.160 | That would be fine too.

00:06:51.160 | I mean, whatever.

00:06:55.560 | So finishing the course means being tenacious.

00:07:00.280 | And one of the things I hear a lot is a lot of the approaches people learn as they do

00:07:07.440 | fast AI around how to learn and how to study are useful more generally.

00:07:12.480 | And in fact, this is a quote from our book, the number one thing I see the difference

00:07:18.440 | between successful deep learning practitioners and not is tenacity, okay?

00:07:24.300 | And tenacity is on the whole something you can choose.

00:07:27.860 | Now something you can't choose is whether you find yourself in the middle of a global

00:07:33.000 | pandemic or, you know, somebody in your family dies or you come down with a terrible cold

00:07:39.720 | or whatever, like obstacles happen, right?

00:07:42.840 | And so part of being tenacious is being understanding with yourself, right?

00:07:48.280 | And saying, okay, something's happened.

00:07:50.400 | I can't do what I hope to do right now, but then getting back to it, right?

00:07:54.680 | So part of tenacity is not about ignoring the bumps, but keeping going after the bumps.

00:08:01.600 | And maybe that's, you know, quite often I'll have a bump that's like a year long, right?

00:08:06.880 | But if I've decided to finish something, you know, at the end of that year, I'll go back

00:08:11.240 | and finish it.

00:08:13.120 | So sometimes that involves me emailing somebody more than a year after they've sent me something

00:08:19.200 | and saying, okay, I'm ready to reply now, and they forgot that they even sent me an email.

00:08:25.880 | Okay, so what I'm going to do now is I'm going to share with you a bunch of insights from

00:08:32.520 | this book called Meta-Learning.

00:08:35.120 | If you haven't seen it before, that's okay, it came out yesterday.

00:08:39.840 | And it was written by a guy called Redek, who is one of the top alumni of this course.

00:08:48.320 | And it's a book well worth reading because his journey is extraordinary.

00:08:55.280 | You know, this is a guy without a degree who couldn't code just a few years ago with a

00:09:03.440 | job that he found boring, and he set out to learn deep learning and repeatedly failed

00:09:13.000 | to do so.

00:09:15.560 | But Redek is extremely tenacious, and each time he failed to do so, he tried again.

00:09:22.640 | And eventually he figured out a way to do it.

00:09:25.080 | And the way he did it was very intensely based on fast AI, both the course and the philosophy

00:09:32.800 | of learning.

00:09:34.160 | And he is now a Kaggle competition winner.

00:09:39.400 | He was the only non-San Francisco person at QAI, which is one of the world's top medical

00:09:47.280 | AI startups.

00:09:48.480 | And now he works at a new nonprofit that is literally trying to translate animal language.

00:09:57.560 | And so he's kind of a good example, like I always think it's a good idea to have a role

00:10:02.080 | model.

00:10:03.080 | And in the fast AI community, there's a lot of role models.

00:10:05.760 | And so here's somebody who's like both a role model for like trying, failing, trying, failing,

00:10:11.200 | trying, failing, and then, you know, finding some success.

00:10:15.600 | And so I'm going to show you some things from his book.

00:10:21.080 | And a lot of his book is him taking stuff I say and kind of casting it into what he took

00:10:28.640 | away from it.

00:10:29.640 | That's his ideas.

00:10:32.660 | So one of the things we hear again and again from unsuccessful deep learning students is

00:10:41.440 | they keep preparing to do deep learning.

00:10:44.880 | And they keep preparing to do projects.

00:10:47.120 | So they study linear algebra.

00:10:49.560 | They study calculus.

00:10:51.180 | They study C++.

00:10:53.440 | They study all these different things.

00:10:56.180 | They do a MOOC and then another MOOC and then they read a book and then another book.

00:11:01.280 | You know, and at what point are they actually going to start doing something?

00:11:05.400 | So the fast AI philosophy is you start doing something week one, OK?

00:11:10.940 | So week one, you need to actually train a model, OK?

00:11:17.400 | Which is not to say that you're not going to learn theory.

00:11:23.320 | You will, right?

00:11:25.320 | As needed in the context of getting stuff done, OK?

00:11:29.080 | And so if you do finish it, right, particularly if you finish the full two parts of the course,

00:11:37.000 | right, you'll have implemented basically all of fast AI's library just about from scratch.

00:11:42.840 | You'll know all about batch normalization.

00:11:45.160 | You'll have benchmarked various matrix multiplication approaches.

00:11:49.320 | You'll know how to write bare metal GPU optimized code.

00:11:53.200 | You'll understand how to do back propagation and the calculus of that from scratch.

00:11:59.240 | You'll do all of that, OK?

00:12:01.240 | But it will all be as you go along in the context of like solving a particular problem

00:12:06.000 | or understanding the next piece of the puzzle.

00:12:10.120 | So yeah, really just reading books and watching videos is not going to get you there.

00:12:18.560 | The thing which is going to get you there is writing code, doing experiments and training

00:12:24.040 | models.

00:12:25.800 | Some of you might not be that great at coding.

00:12:29.360 | Fine.

00:12:30.360 | OK.

00:12:31.360 | That's a perfectly OK place to be.

00:12:34.400 | And but you guys are going to find it the most challenging because being good at coding

00:12:40.920 | is the thing that lets you zip through quickly.

00:12:43.160 | So rather than think, oh, that's a shame.

00:12:45.840 | I'm not that good at coding yet.

00:12:47.200 | This is actually an opportunity because now you have a really fun project to learn to

00:12:53.160 | code in.

00:12:54.160 | So a lot of people have become good coders by doing the course.

00:13:00.240 | Because as you do the course, you'll learn about a lot of computer science concepts like

00:13:04.840 | object-oriented programming and functional programming and mapping over a list and list

00:13:09.520 | comprehensions and GPU acceleration and so on and so forth, right?

00:13:15.160 | So the thing is, though, if you come across a computer science concept or a programming

00:13:21.560 | idea or a piece of syntax that you're not that familiar with, that's a place it's worth

00:13:26.800 | pausing for a moment and making sure that you do understand how that code works.

00:13:33.560 | Because the coding is the kind of critical foundational skill.

00:13:38.880 | This is a pretty good course for getting started with basic computer science.

00:13:43.240 | Harvard CS50 course, which everybody at Harvard does for computer science to get started.

00:13:50.600 | And that's all available for free online.

00:13:52.840 | So I would recommend, well, and so would Radek start there.

00:13:58.160 | And so these quotes are all from Radek's book, by the way.

00:14:02.840 | And then the other piece, so Radek talks about this four-legged table of the things that

00:14:08.240 | are going to help you do your deep learning experiments more effectively and efficiently.

00:14:16.200 | And these are the ideas.

00:14:17.600 | Knowing the basic ideas around code, knowing your tools, so an editor, do you put a notebook,

00:14:26.760 | knowing stuff like Git, like how to save your work and pull in other people's work and so

00:14:32.840 | forth, and understanding kind of SSH and Linux, like how to access a server and manipulate

00:14:40.240 | it and do stuff with it.

00:14:43.520 | So there's this great course called the missing semester of your CS education, which was actually

00:14:49.400 | created, I believe, by students at MIT who said, oh, everybody at MIT is assuming we

00:14:57.720 | already know this stuff, but a lot of us don't.

00:15:02.080 | So there's nothing to be ashamed of if you've never used Git or you've never used SSH or

00:15:08.920 | whatever.

00:15:09.920 | They're just tools which, at some point in the journey, most people just kind of have

00:15:16.360 | to figure out.

00:15:17.360 | So this is actually a great time to do it, and this is a great course to use to help

00:15:23.720 | you get there.

00:15:24.720 | And of course, again, the main thing is to practice these tools.

00:15:30.300 | So that's the kind of foundation around coding and your kind of development environment.

00:15:39.640 | The next big piece of advice, which we talk about a lot in the course and that Radek talks

00:15:44.200 | about in his book, is sharing your work, communicating your work, and writing about your work.

00:15:54.780 | This is something that a lot of people feel very uncomfortable, like tweeting or blogging

00:16:00.920 | or whatever.

00:16:01.920 | It's like, who the hell am I to start writing about deep learning?

00:16:08.080 | I've just started.

00:16:09.880 | Well, here's the thing.

00:16:13.400 | No one is better placed than you to write for, like, what would you have wanted to know

00:16:20.280 | six months ago?

00:16:22.200 | So you now know more than you did six months ago, and you'll know more in a week and more

00:16:27.000 | in a week, more in a week.

00:16:28.320 | And so if you've got a background in, say, the hospitality industry, you know, you could

00:16:33.700 | probably write something very interesting for your colleagues in the hospitality industry

00:16:38.440 | about ideas around deep learning, for example.

00:16:43.800 | Or if you teach at high school, you know, you might have ideas that you could write

00:16:48.040 | down about what high school students might find interesting or teachers might find interesting.

00:16:52.460 | So you know, everybody's got something to say.

00:16:55.240 | And the key thing is just to write it down because that is going to help embed your understanding

00:17:01.200 | a lot better, and it's going to start to build up your portfolio.

00:17:09.000 | And so we'll talk more about that in a moment.

00:17:11.080 | But a lot of people have found that this message of sharing their work has been a critical

00:17:17.360 | part of their journey of learning, and of also building up their personal brand that

00:17:22.800 | has ended up getting them a job.

00:17:26.160 | OK.

00:17:27.320 | So what does it mean to do a fast AI lesson?

00:17:33.680 | So a fast AI lesson is basically a chapter of the book or one video from the course.

00:17:45.080 | Or both.

00:17:46.080 | So what does it mean to do one of these lessons?

00:17:50.400 | Assuming you're doing the video, then it means, OK, obviously watching the video.

00:17:55.900 | So there's a couple of hours, right?

00:17:58.800 | And then it means running the notebook, which we'll look at in a moment.

00:18:06.120 | When you run the notebook, you have the whole book with all of its code and all of its outputs

00:18:11.720 | there, you're playing with it.

00:18:16.640 | You should experiment, right?

00:18:19.240 | You should try things out.

00:18:21.140 | So if you wonder, oh, why is this done before that?

00:18:25.680 | Well, try removing it.

00:18:27.640 | Try doing it in a different order.

00:18:29.280 | If you're wondering, you know, what would happen if I did that, but to this other image,

00:18:34.120 | try it, right?

00:18:35.120 | The more you can start to experiment, the more you're feeding your brain with these

00:18:39.680 | kind of like your own deep learning happening in your brain.

00:18:43.200 | Input output patterns.

00:18:44.200 | You try something, what happens?

00:18:45.440 | You try something, what happens?

00:18:49.680 | So after that, the next step is to try to reproduce the notebook from scratch, OK?

00:19:02.140 | Now you're going to have to look things up, obviously.

00:19:05.720 | But the idea is, can you, with a fresh new notebook, can you go back and recreate some

00:19:18.680 | of those models, retrain them, or redo some of that data processing pipeline?

00:19:23.080 | So try to like type it in yourself, you know, you can switch back to the answer as much

00:19:28.600 | as you like, but you're really trying to start to actually, you know, fill in your own, write

00:19:34.800 | your own code.

00:19:38.000 | And then what you really, the point you really want to get to is repeating some parts of

00:19:43.000 | the lesson with a different data set, which you collect or download.

00:19:49.640 | Now this whole process often takes people a number of times through the course, right?

00:19:58.320 | So often the first time through, people might just watch each lecture and try to kind of

00:20:03.080 | run it and, you know, just get to the end to get a kind of a general sense of what's

00:20:07.620 | going on.

00:20:08.620 | So people will often kind of go through the whole thing like three times and then come

00:20:12.280 | back and try to go further and further, right?

00:20:15.360 | So don't worry if you can't do all this right away.

00:20:20.840 | Certainly in lesson one, that's going to be challenging.

00:20:23.200 | Just take it as far as you can, right?

00:20:24.680 | And as you go along, try to push yourself to do more and more, and you could even go

00:20:29.240 | back to an earlier notebook and see if you can understand more and more of it.

00:20:33.760 | So let's take a look at what that looks like.

00:20:38.280 | So here's the course, okay?

00:20:42.120 | And here's the lessons, which you can watch.

00:20:49.000 | And then here are the places you can run the notebooks.

00:20:54.560 | So there's two types of platform for running the notebooks.

00:21:01.840 | There are notebook servers.

00:21:03.080 | These are things that as soon as you click into it, the actual environment we use, Jupyter

00:21:09.440 | notebook, will pop up and you can just start running it pretty much straight away.

00:21:15.460 | So that is obviously the easiest.

00:21:20.300 | Colab is free.

00:21:22.920 | Colab has a free tier.

00:21:25.840 | And SageMaker is not free.

00:21:31.320 | So we're going to look at Colab today.

00:21:35.720 | The other option is to use a full Linux server, and this is something where you're going to

00:21:40.160 | have to basically set up Linux and install the Python system and install notebooks and

00:21:46.560 | get the code from GitHub and run the server and log into SSH and do all that.

00:21:52.040 | That's obviously a lot more work.

00:21:54.360 | You might want to skip it for now in like lesson one.

00:22:01.180 | But I would recommend at some point you go through this path.

00:22:07.480 | And the reason why is that in real life at your workplace or if you do your own startup

00:22:13.280 | or whatever, this is what you'll be doing.

00:22:16.720 | You will be interacting with a Linux server using SSH that's running a GPU.

00:22:22.720 | And you'll want to understand how it all works.

00:22:25.400 | And once you're using your own Linux server, you'll suddenly learn about all these productivity-enhancing

00:22:31.440 | tips and tools that make your life easier.

00:22:36.200 | So I'll be showing how to set up AWS EC2, that's the Amazon platform today.

00:22:46.880 | You'll find Google Cloud looks very, very similar indeed.

00:22:54.320 | Jarvis Labs was created by a Fast AI alum, and this is probably at this stage the best

00:22:58.800 | value of the full Linux servers.

00:23:02.060 | So that would certainly be also very much worth checking out.

00:23:06.920 | One good thing about AWS, so a couple of things, AWS is currently the most popular platform

00:23:13.600 | for cloud computing.

00:23:14.960 | So it's very likely that whatever company you're at or end up at is already using it.

00:23:22.120 | They're also pretty generous with credits for startups and students.

00:23:26.680 | So even though it can set you back 60 or 70 cents an hour, you might well find you can

00:23:33.440 | get a few hundred dollars worth of credits through your school or even a few thousand

00:23:39.680 | dollars worth of credits through their startup programs and so forth.

00:23:45.440 | So let's have a look at what Colab looks like.

00:23:51.920 | So Colab is, it's wonderful how easy it is to get started.

00:23:57.960 | You literally just click on the chapter, so let's do chapter one, and it pops up Colab.

00:24:10.240 | You can pay, I think it's $10 a month for Colab Pro to get like longer sessions and

00:24:17.280 | more likely that you'll get a better GPU, but for most people you'll find the free version

00:24:21.960 | is totally fine.

00:24:23.320 | One of the biggest problems with Colab is that it's not persistent, which is to say

00:24:30.400 | when I go to this notebook, it thinks it's never seen me before.

00:24:35.200 | Nothing's set up for me the way I want it, but we've set up the notebook so that the

00:24:39.720 | very first cell actually installs everything you need.

00:24:44.020 | So if I click this little run cell button here, it will run the cell.

00:24:51.680 | Although what I will do is I'm going to pop over to Colab here, and let's also read the

00:25:04.360 | steps here.

00:25:05.360 | And actually it says here before running anything, you should tell Colab you're interested in

00:25:09.600 | using a GPU.

00:25:11.140 | So if you find that when you run a cell from the course, and it's going to take like half

00:25:20.760 | an hour or an hour or more, it's very likely you forgot to use GPU.

00:25:25.920 | The GPU runs things many hundreds of times faster.

00:25:29.540 | So all you do as it says here is go runtime, change runtime type, and say GPU, okay?

00:25:39.960 | So now I can run this cell.

00:25:50.000 | And this is all Python code except lines that start with an explanation mark actually sent

00:25:56.920 | to a terminal.

00:25:58.560 | So PIP is something that installs Python software, and Fastbook contains all of the Python software

00:26:05.320 | necessary for the course, and so it's going to go away and set it all up.

00:26:11.800 | And so this is this like mildly annoying bit.

00:26:18.000 | You can then connect Colab to Google Drive, and that's going to be how you can save your

00:26:26.200 | notebooks and save your work as you go, okay?

00:26:28.840 | I'm not going to do that right now, but if you go to this link that it says and it'll

00:26:33.440 | give you a code and then that'll connect it up to your Google Drive.

00:26:39.240 | And so at this point now everything from the for the course is now available, and you can

00:26:47.400 | see the whole book is here, okay?

00:26:53.200 | So here's the book, and you can open up sections to read them, okay?

00:27:03.960 | You can go to the table of contents, okay?

00:27:15.940 | And so eventually we'll get to this cell here which contains all the code needed to run

00:27:25.880 | a model.

00:27:27.440 | So if I click run, here is where it goes.

00:27:34.080 | Now this is going to it's amazing how much this little bit of code is going to do.

00:27:40.320 | It's going to download tens of thousands of pictures of dogs and cats.

00:27:46.400 | It's going to use a simple rule to recognize the dogs from the cats based on their file

00:27:52.480 | names.

00:27:53.480 | Basically the way that this has been set up is that you can tell from the file name whether

00:27:57.240 | it's a dog or a cat.

00:28:00.240 | It's then going to download something called a pre-trained model, which is something that

00:28:04.000 | already knows how to recognize various types of images.

00:28:09.000 | It's then going to construct, it's then going to train that model to make it particularly

00:28:14.520 | good at recognizing dogs from cats, and then it's going to validate that model to see how

00:28:19.480 | good it is at recognizing dogs from cats using a set of pictures that it hasn't seen before.

00:28:25.040 | And that's all happening.

00:28:26.880 | So so far it's already downloaded the dataset, it's already downloaded the pre-trained model,

00:28:33.320 | and it's now busily going through the first epoch, which is to look at every picture once

00:28:38.880 | to try to learn how to recognize dogs from cats.

00:28:44.120 | And that's it.

00:28:45.120 | The lines starting with a hash are just comments.

00:28:47.840 | Because this is also the source of an actual book, there's a few like slightly weird comments

00:28:53.360 | that you can ignore.

00:28:54.360 | They're just things that are used for setting up references in the book.

00:28:58.480 | There's the caption, so forth.

00:29:00.080 | Okay, so it's now testing out, I think that first epoch.

00:29:05.480 | Okay, so it's finished in epoch, and so far it's got a 1% error rate.

00:29:11.000 | So after 54 seconds, it has learned to recognize dogs from cats with 99% accuracy.

00:29:19.880 | And so yeah, we're going to let that finish off.

00:29:25.560 | So that's how we get started with Colab.

00:29:27.920 | And there's nothing else to set up.

00:29:34.440 | Now what you can do is you can open Notebook, and you can open a notebook from GitHub.

00:29:57.560 | And here is the Fastbook repository.

00:30:04.600 | And you'll see in the Fastbook repository, for every notebook, there's a second copy

00:30:08.240 | inside the clean folder with the same name.

00:30:11.320 | So I was just looking at 01Intro.

00:30:13.680 | There's also a clean 01Intro.

00:30:15.640 | If I open that up, you'll see that it's got exactly the same thing as the last one I was

00:30:26.440 | just looking at, but all the pros is now missing.

00:30:30.560 | It's just got headings and code.

00:30:34.480 | Also all the outputs are missing.

00:30:37.800 | So the reason that we have this clean version is to help you with these stages here, is

00:30:45.520 | our suggestion is once you've gone through the lesson, and you've run the notebook, and

00:30:52.120 | you feel like, okay, I think I get it, is you open up this clean version.

00:30:57.960 | And before you run each cell, try to think, okay, why is this cell here?

00:31:04.800 | What's it for?

00:31:06.040 | What's it going to do?

00:31:07.320 | What's the output going to look like, right?

00:31:09.640 | So once you remove all that context, this is a good test for you to kind of get your

00:31:15.120 | brain going to think what was actually going on.

00:31:18.480 | So this is a kind of much more active approach to reading and recall.

00:31:23.880 | And so then once you've done that, and you've finished going through this, at the bottom,

00:31:32.000 | one thing that is kept is the questionnaire.

00:31:35.760 | So at the end of every chapter is a questionnaire.

00:31:38.260 | And so then at this point, you should now, as much as you can without looking, go through

00:31:44.880 | and try to answer each of those questions.

00:31:48.480 | They all have answers in the notebook, in the book.

00:31:52.520 | So if you can't remember, you can always look it up.

00:31:56.280 | But if you can't remember, that's a sign to you that like, oh, did I skip over that bit

00:32:01.920 | too quickly?

00:32:02.920 | Like what's happened that I've not remembered?

00:32:06.080 | And then try to remind yourself, and then go back and finish the questionnaire.

00:32:12.180 | So there's a lot of pieces to help take this from a passive, I'm just watching a video,

00:32:19.900 | I'm just reading a book, into a participatory exercise that you're a part of.

00:32:33.680 | So as soon as you can, we want you to create something that's yours.

00:32:41.640 | And so this is the easiest way to do that, is basically at the end of lesson one, once

00:32:46.880 | you're kind of up and running, try to do it with your own data set.

00:32:51.800 | And if you go to forums.fast.ai, which is something that you're going to want to be

00:32:58.040 | deeply familiar with, because this is going to be full of people just like you, other

00:33:02.760 | people who want to learn deep learning.

00:33:05.720 | And these people are all asking questions, and making comments, and you can see there's

00:33:11.760 | like a lot going on all the time.

00:33:14.600 | And so you can see here's the part one course topic.

00:33:22.380 | And you can see there's 1.4 thousand topics there, and each one is going to have lots

00:33:26.300 | and lots of replies.

00:33:29.600 | So this is where, amongst other things, you'll find, if you search for it, something called

00:33:35.600 | Share Your Work Here, which has 2,000 replies, and you can see links to and pictures of lots

00:33:41.360 | of examples of things that other people have done after the first week or two of the course.

00:33:47.120 | And so hopefully that might help give you some inspiration.

00:33:52.800 | And it would be great if you could reply and add a picture or a link to what you build.

00:34:00.720 | And you'll see, you know, everybody is very positive to each other on the forums in general

00:34:08.040 | and in this topic in particular.

00:34:09.800 | Nobody's going to go, "Oh my god, I could have done that years ago," right?

00:34:15.320 | People are going to be excited for you, that you have now joined the ranks of people that

00:34:20.260 | have built their first deep learning model.

00:34:23.320 | And I will be excited for you.

00:34:30.120 | So as I said, Radik, this is again from his book, expresses in his book a way of not doing

00:34:39.600 | fast AI, which I have heard now probably hundreds of times.

00:34:43.960 | I don't know why this is so common, but many, many people do what Radik did, which was basically

00:34:50.080 | to learn all these math things, right?

00:34:54.720 | So he started with calculus, and then once he got to a certain point in calculus, he

00:35:01.080 | found that he had to start understanding real analysis.

00:35:05.080 | And then as he started understanding real analysis, he had found he had to learn set

00:35:11.120 | theory, you know, and you get the idea, right?

00:35:13.740 | If you want to learn all of math, that's going to take a while.

00:35:20.640 | There's a lot of gatekeeping out there that says like, "Oh, if you're going to be a real

00:35:25.440 | deep learning practitioner, you have to finish, you know, a graduate level course in linear

00:35:30.560 | algebra."

00:35:31.560 | Here's the truth, the actual linear algebra you do in basically all deep learning is matrix

00:35:39.880 | multiplication.

00:35:41.600 | And if you've forgotten what that is, that is multiplying things together and then adding

00:35:46.400 | them up, okay?

00:35:48.200 | So what you need to be able to do is multiply things together and add them up, right?

00:35:53.160 | So if you can do that, you're good to go.

00:35:57.460 | So yeah, don't get, you know, you're not going to finish it if A, you never start it because

00:36:04.960 | you keep preparing, or B, you keep thinking, "Oh, I wonder exactly what's happening here,"

00:36:10.280 | and you go all the way down to the bottom until you found yourself in the midst of set

00:36:13.560 | theory, right?

00:36:14.560 | Don't worry, you'll get deeper and deeper over time, but if you're learning mathematical

00:36:22.240 | theory, you're not coding, you're not experimenting, you're not practicing, you're not actually

00:36:27.160 | building deep learning models, and if you're watching this course and your goal is not

00:36:31.680 | to build deep learning models, you're in the wrong course, okay?

00:36:35.560 | But if your goal is to build deep learning models, then don't do this.

00:36:46.360 | So as Rennick says here, it's as you train actual models that you're going to get feedback,

00:36:54.840 | right?

00:36:55.840 | And the feedback that a lot of people get is, "Oh my God, I can already train useful

00:37:01.360 | models," like a lot of people are surprised at how early on they can actually get astonishingly

00:37:09.600 | good results.

00:37:10.600 | Okay, so, you know, jump in and be open to surprising yourself that you can do a bit

00:37:16.360 | more than you thought.

00:37:17.360 | You can't do everything right away, okay, but start that feedback loop of figuring out

00:37:22.080 | what do you know, what can you do, what can you get working, what can't you get working?

00:37:30.240 | So one of the key things that you're going to need to do if you're going to finish all

00:37:34.240 | of the course is become an even better developer than you are now, even better coder than you

00:37:40.240 | are now, wherever you're up to, and so to do this, you need to read code and write code.

00:37:48.720 | The fast.ai source code is designed to be extremely readable, so you can read that code.

00:37:53.300 | You can obviously read the code in the notebooks, but yeah, you want to be spending as much

00:37:59.120 | time as possible reading and writing code, and particularly reading and writing deep learning

00:38:05.240 | code.

00:38:08.080 | All right, how do you find out what's going on in the world of deep learning, and how do

00:38:17.400 | you get yourself on the map of people doing deep learning?

00:38:22.560 | Probably the best answer is Twitter.

00:38:24.600 | For those of you whose only knowledge of Twitter is the Kardashians and Donald Trump, this might

00:38:31.720 | come as a surprise, but actually to create this slide, I opened Twitter and I copied

00:38:39.600 | and pasted the first three tweets that appeared on my screen.

00:38:43.360 | So one of them is somebody has a discussion about costs and impacts of different approaches

00:38:50.600 | to labeling. This is a fast.ai alum who's a 17 year old PhD graduate who's doing well,

00:38:58.680 | who shows how to mix PyTorch and fast.ai, and then Hilary Mason, who's a professor,

00:39:04.840 | and I guess not a professor anymore, but now in industry, talking about organizational

00:39:11.240 | issues in data science.

00:39:13.600 | So there's a whole world out there of machine learning on Twitter, and if you want to get

00:39:23.480 | your work noticed, that's a great place to do it because really everybody's there.

00:39:29.840 | And if you want me to highlight your work, that's where I can see it and I can retweet

00:39:35.680 | it. So yeah, Twitter is a really good place to be. If you're just starting with Twitter

00:39:44.880 | and you don't know who to follow, go to my Twitter, go to my likes, and go through my

00:39:51.800 | likes and find tweets that you think you actually like that tweet to, and then follow the person

00:39:57.320 | who did that tweet. Okay, and pretty quickly you'll have 100 people you're following, okay,

00:40:03.520 | and then they'll retweet things and you'll find other people you like, and before you

00:40:07.800 | know it, hopefully you've got a nice big lot of interesting, deep learning stuff to read

00:40:12.620 | every day. At first you'll understand like 1% of it, which is fine, but you know, you're

00:40:19.080 | there, you're in it, and it'll be all washing over you, and you'll start to find the people

00:40:24.320 | who write stuff you find engaging and interesting, and you'll also find the people that actually

00:40:29.720 | you don't, and make sure you unfollow them so that you don't have your feed, have stuff

00:40:34.720 | you don't care about. So then beyond Twitter, you want to start blogging. Okay, and again,

00:40:44.720 | blogging is not about writing what you had for dinner, okay, it's about writing something

00:40:51.920 | that you of six months ago would have found interesting. Okay, so you know more than you

00:40:56.960 | did six months ago, so write that down. We have something called Fast Pages that makes

00:41:04.000 | it ridiculously easy to start a blog, and so there's no reason for you not to, you know,

00:41:13.440 | at least create a blog. There we go. And one of the nice things about Fast Pages is you

00:41:20.800 | can even turn Jupyter Notebooks into blog posts, so it's great for kind of technical

00:41:24.880 | reasons. So this is what a Fast Pages blog looks like. This is a Fast Pages blog about

00:41:29.400 | Fast Pages. I had to write Fast Pages in order to write the Fast Pages blog about Fast Pages.

00:41:36.240 | But basically, and one of the other nice things, it's all in GitHub, right? So as you're blogging,

00:41:41.920 | you're learning more about Git. It's all written with Markdown, which is something that you're

00:41:46.240 | definitely going to need to know anyway. So as you're blogging, you'll be learning about

00:41:50.800 | a lot of the tools you need to learn about anyway.

00:42:04.560 | So one interesting idea for things to blog about is this example from Aman Arora, who

00:42:11.160 | is an Aussie Fast AI alum who is now working at Weights and Biases, which is one of the

00:42:16.920 | top AI startups in the world. This is a really interesting kind of blog post. What Aman did

00:42:25.240 | was he took a video that I did at the launch here of the Queensland AI Hub, and he wrote

00:42:32.360 | down what I said. And that's an example of something that you could do. If there are

00:42:40.720 | videos out there that you liked and nobody's turned it into a post, be the first to do

00:42:47.520 | so because there's all these benefits. When somebody sends me something saying, "I've

00:42:55.080 | written up this talk you gave," I'm very grateful to that person because now my talk

00:43:00.480 | is now available in a second medium. A lot of people prefer to read rather than listen

00:43:04.280 | to a talk. You know, that person's taken the time to do this. They've taken the time to

00:43:10.280 | have me check their work. And kind of everybody ends up winning from this. So I've seen with

00:43:17.360 | Aman's post about my talk, it's got attention from people that my talk didn't. So for example,

00:43:26.240 | I noticed on my LinkedIn feed, the CEO of Data61, which is the CSIRO, so the top data

00:43:34.440 | science body in Australia, highlighted it and said, "Check out this post from Anamurara."

00:43:41.680 | So this is like an example of the kind of stuff you can do. It's like try to be helpful,

00:43:48.920 | and at the same time you're also learning. So there's an example of an interesting kind

00:43:54.400 | of blog post which very few people are writing, and so there's a huge amount of opportunity

00:43:59.760 | here for you to practice your writing. Okay, now, what is the difference between machine

00:44:12.320 | learning and other kinds of coding? Ezra Dex says in this chapter of his book, "The key

00:44:20.920 | about machine learning is that we can generalize. We can train a model with one set of data

00:44:29.280 | and apply it to a different set of data and still get good results." And everything just

00:44:36.640 | about that we're doing in this course is all about creating models that are going to generalize

00:44:43.320 | well. And we're going to be learning about how you can measure how well your model generalizes.

00:44:53.120 | So answering these questions about can we trust our model to be correct on new data

00:45:00.240 | that we feed it is absolutely critical to every model that you build, whether it be

00:45:07.600 | in a Kaggle competition or a little prototype or a production model you're creating at work.

00:45:18.260 | One of the most important things here is creating a good validation set, and this is something

00:45:22.360 | that you'll hear about in lesson one of the course. But I really wanted to highlight it

00:45:28.040 | here, as did Radek in his book. It's a really important idea is you need a good way to measure

00:45:36.320 | whether your model is any good. So you need a data set that really represents what kind

00:45:42.080 | of data is your model likely to have to deal with in real life. And my partner Rachel wrote

00:45:51.120 | this really great blog post on the Fast.ai blog about this. Actually interestingly, this

00:45:56.760 | was kind of came out of a lesson that I did at the University of San Francisco and then

00:46:03.760 | Rachel turned it into a blog post and Rachel's blog post has ended up much more influential

00:46:08.880 | than my video ever was. So this is actually a good example of what I was talking about.

00:46:13.600 | And she took it a lot further. Okay. The next key thing that Radek mentions and I totally

00:46:25.720 | agree with is it's hard to write correct machine learning code. I always assume that every

00:46:35.360 | line of machine learning code I write is wrong. And I'm normally correct about that. It normally

00:46:40.720 | is wrong because there's lots of ways to be wrong. And unlike creating a you know, a context

00:46:50.720 | management app on the web or whatever, it's much harder to see that you're wrong. You

00:46:55.160 | know, you can't see that the name didn't get stored in the database or you can't see that

00:46:59.600 | the title isn't centered. Right. Often it's wrong that it's going to be like half a percent

00:47:05.120 | less accurate, you know, or your image is upside down, but it's kind of maybe you didn't

00:47:09.440 | even look at it. I got straight into the system and you end up with something that can only

00:47:13.520 | recognize upside down images or whatever. So whenever you're doing, you know, whenever

00:47:22.640 | you're building a project, make sure you start with a simple baseline, right? Like create

00:47:28.240 | the simplest possible model you can that's that, you know, solves the problem so simply

00:47:34.600 | that you can't have made a mistake. So often that'll be like just taking the average of

00:47:40.760 | the data or if there's two groups, take the average of each of the two groups or you know,

00:47:46.320 | something that something really, really simple and then you can gradually build up from there.

00:47:52.080 | So another very common beginner mistake with projects, remember we want you all doing projects

00:47:57.800 | is somebody in a project group will say, Oh, I read about this new Bayesian learning thing

00:48:05.040 | with these clusters and this, you know, advanced transformers pipeline, and we could put all

00:48:10.840 | that together. It's going to be better than anything before. And they then spend months

00:48:16.800 | creating this complex thing. And at the end, it doesn't work. Now, why doesn't it work?

00:48:23.720 | Well, I don't know. It's so big and so complicated. Maybe it's a stupid idea. Maybe there's a

00:48:28.720 | bug in one piece of it. Maybe that one piece there shouldn't be there, but it should be

00:48:32.400 | somewhere else. I don't know, right? That's not how anybody creates successful machine

00:48:38.600 | learning projects. Machine successful machine learning projects are always built, in my

00:48:45.280 | experience, by creating a simplest possible solution that gets something all the way from

00:48:51.760 | end to end first, and then very gradually it makes it incrementally slightly better. Okay,

00:48:59.680 | so keep that in mind, right? You might feel a bit silly when you build that first model

00:49:05.520 | that just takes the average of the data, right? But that's how, that's how the pros do it.

00:49:11.600 | That's how everybody that actually gets it to work does it. So often I've had, you know,

00:49:18.640 | Silicon Valley startup hotshots come to me and ask me to like, check out their amazing

00:49:23.000 | new startup, and I'll ask them, you know, oh, you reckon this can separate, you know,

00:49:30.560 | sick people from well people or whatever. Have you taken the average of each of these

00:49:35.100 | two groups and compared that to your model, for example? And they'll say, oh, no. And

00:49:40.640 | then they try it and they find out their model's worse, right? So you need to know whether

00:49:46.080 | your model's actually doing something useful. For projects, one of the things you might

00:49:57.160 | want to do is join a Kaggle competition. That might be the last thing you see yourself as

00:50:05.680 | doing is being a Kaggle competitor, but actually this is one of the best possible projects you

00:50:12.000 | can do because to enter a Kaggle competition, even to come last, you have to go through

00:50:17.560 | the entire process of downloading a dataset, formatting it into the right method, ready

00:50:23.680 | for a model, getting it through the model, saving the output, getting it into the correct

00:50:28.360 | submission format and submitting it back to Kaggle, right? So getting a model actually

00:50:37.400 | up onto the Kaggle leaderboard is really going to test out your end-to-end understanding,

00:50:43.040 | right? And once you've done that, you can start to iterate. You can start to make it

00:50:47.120 | slightly better, slightly better, slightly better. So although in a lot of ways, Kaggle

00:50:54.760 | is not representative of the real world, you know, you don't have to worry about deployment.

00:50:58.800 | You don't particularly have to worry about kind of inference speed, stuff like that.

00:51:03.400 | In a lot of ways, it is closer to the real world than you might expect and that it really

00:51:08.880 | does force you to go through the whole process and also to think about kind of planning your

00:51:16.560 | project carefully. So enter a competition with your kind of goal that I want to win.

00:51:25.640 | Now obviously on your first one, you're not going to win, but the whole point is it's

00:51:29.800 | a competition. So you've got to try to do your best, right? And so to do your best,

00:51:35.960 | join a competition that's early, right? Give yourself plenty of time. And every single

00:51:44.360 | day try to make a small improvement. And then you'll find that, you know, if you keep reading

00:51:52.680 | the forums on Kaggle and keep trying a bit more every day, you'd be amazed at the end

00:51:57.640 | of the three months how much you've learned, how much of the stuff that at the start you

00:52:01.820 | thought this is, I have no idea what's going on. And then you'll realize, oh, suddenly I

00:52:08.640 | do know what's going on. And you might find you get in the top 50% which might be better

00:52:15.920 | than you expected. So that this is, you know, highly recommended at some point during this

00:52:23.960 | course is have a real go at a Kaggle competition. So at the end of all of this, you might be

00:52:36.920 | looking for a job. Now this could mean a number of things. A lot of people just want to bring

00:52:44.240 | some deep learning into their current job. And so, you know, that's, if your organization's

00:52:51.160 | already doing some deep learning, that might be easier than if it's not. If it's not, you

00:52:55.000 | might just have to start prototyping some things and try to build up some kind of, you

00:52:59.640 | know, proof of concepts internally. Or maybe you're going to try and go out and get a new

00:53:07.120 | role as a researcher or a data scientist or whatever. Most people are not going to be

00:53:17.520 | able to rely on their, you know, Stanford PhD to get them there, right? Most people are

00:53:28.040 | going to rely, have to rely on their portfolio. So your portfolio is going to be all the stuff

00:53:34.920 | you build along the way. It's your footprint on the deep learning community. And that footprint

00:53:43.400 | is going to include, you know, think things like your contributions to the Fast AI forums

00:53:49.440 | and your tweets and your stuff on Discord. I would say pretty much every one of the Fast

00:53:58.760 | AI alumni that have come to my attention as being thoughtful and effective community members

00:54:07.560 | all have very, very, very good jobs now. And so like people really, really notice this

00:54:14.720 | footprint, right? So your blog posts, your GitHub projects, these are the things that

00:54:22.800 | are going to get you a job. They probably won't get you a job at a big company, a big

00:54:34.200 | old company in a, you know, kind of standard established IT job, right? That's going to

00:54:42.920 | go through HR and HR, like they're not going to understand any of your GitHub code or know

00:54:47.680 | any about your community impact. They're just going to know about credentials, right? And

00:54:52.960 | you'll come up against somebody with a Stanford PhD and they'll get the job, right? But startups,

00:54:59.720 | really startups from other people who've got similar backgrounds of which there are many

00:55:04.600 | are going to appreciate you or companies that don't really have an established AI group

00:55:10.280 | yet, or the startup you built yourself will certainly appreciate you, right? So it's the

00:55:22.320 | more you've got a portfolio and that you can show that you've really built stuff, the better.

00:55:27.240 | And so start early. Another reason to finish this first course is that it's going to allow

00:55:45.920 | you to do the second course. And if you're doing this live, part two, we're going to

00:55:52.040 | be doing actually a whole new part two towards, you know, basically shortly after this is finished,

00:55:59.360 | right? So if you finish this and do a good job of it, then you could actually be one

00:56:04.920 | of the first to do part two. Now, we've seen how easy Colab is to get started. We've also

00:56:19.500 | talked about some of the downsides of it, right? It's kind of ephemeral. You start from scratch

00:56:24.080 | every time. You've got this kind of hacky stuff of saving notebooks into your Google

00:56:28.720 | Drive, blah, blah, blah. AWS, on the other hand, is going to give you and Google Cloud

00:56:35.960 | and Java Slabs and so forth are going to give you a real Linux server. Okay. And it's going

00:56:42.400 | to cost you, Java Slabs is the cheapest, about 40 cents, AWS, I think about 60 cents US per

00:56:50.960 | hour. It's not going to send you broke, but it's, you know, it's not nothing. But it's

00:56:58.160 | a good idea to try it if you can. And I'm going to show you how to get started there.

00:57:05.280 | And what we might do, Michael, is I'll do some Q&A while things are running. So I'm going

00:57:16.060 | to head over to AWS EC2. Okay. So one of the tricky things about AWS is they've got hundreds

00:57:25.200 | of products. This is Amazon Web Services, and they all have names that are totally meaningless.

00:57:29.960 | Okay. So you just have to know, EC2 is the name of the thing that you go to, to rent

00:57:35.000 | a computer. Okay. So they don't call it Amazon Computer Rental, they call it EC2. So the first

00:57:45.640 | thing you need to do is you need to sign up to AWS. And one of the things that they get

00:57:54.440 | is a lot of fraud. So a lot of people try to use their GPUs to mine Bitcoin. So you have

00:57:59.720 | to ask them to give you permission to use their GPUs. Now that's called requesting a

00:58:05.120 | service limit increase. So you'll need to follow the steps here to ask them for a limit

00:58:11.640 | increase. If you write these exact words with this exact formatting, it might come through

00:58:17.760 | a little bit quicker. If you're from a country where there's a lot of fraud, you might not

00:58:27.560 | even get this permission. Maybe Java Slabs is going to be easier. I'm not sure Java Slabs

00:58:34.840 | even has the fraud check. So anyway, there's quite a few places you can try to get an instance.

00:58:41.620 | So if AWS has a problem with your quota, try somewhere else. But generally speaking, most

00:58:46.520 | people should get a response pretty quickly saying you've now got approved. So for you

00:58:50.660 | doing this course, if you're going to try out AWS EC2, I suggest you log in and request

00:58:57.160 | this service limit increase right away. So that, you know, by the time you come back

00:59:02.400 | tomorrow or the next day, it'll be done. And so what I'm currently doing is I'm on course

00:59:05.840 | fast.ai and I've gone Linux servers, AWS EC2, and we're following through that project

00:59:11.200 | process. Okay. Now to log in to your server, you're going to need to use something called

00:59:18.680 | SSH, Secure Shell. So this is something where on your computer screen, that server's computer

00:59:26.560 | screen effectively is going to appear and the stuff you type is actually running on that

00:59:31.880 | remote server, not on your computer. Nowadays, pretty much nobody uses usernames and passwords

00:59:39.120 | for SSH. Instead, we use something called public key cryptography, which is where you basically

00:59:46.520 | have a secret number, which only you know. And then there's another public number that

00:59:54.480 | you tell other people. And basically there's a really cool math trick, which allows people

01:00:01.520 | to check whether you have the secret number without actually anybody, without actually

01:00:05.780 | telling them the secret number. And the process, so that's called, so that's what an SSH key

01:00:11.320 | is. So there's this thing called a public key, and that's the number that you're, the

01:00:17.280 | code that you're going to give to anybody you want to be able to log into. And then

01:00:20.680 | there's your private key, which you're going to keep for yourself. So you're going to need

01:00:27.360 | a terminal. So on Windows, in the store, there's something called the Windows Terminal, which

01:00:35.640 | Microsoft provides for free, which is pretty good. Mac has a terminal that comes with it.

01:00:41.520 | Linux has a terminal that comes with it. So I'm using Windows, but it'll basically look

01:00:48.760 | the same for everybody. Now on Windows, you need a Ubuntu Linux shell, not a normal Windows

01:00:59.960 | shell. So to do that, you need something called WSL, Windows Subsystem for Linux. And that

01:01:06.280 | will give you a full Ubuntu system on your Windows computer. Again, it's free. It only

01:01:11.600 | takes a couple of minutes to set up. So there's a link to how to do it here. So once you've

01:01:15.520 | done it, whatever, whether you're on Mac or Linux or Windows, it's going to look basically

01:01:22.320 | the same, right? And so you'll create your SSH key by following the instructions in the

01:01:30.120 | documentation, which is basically you run SSH keygen, and it's just going to go through

01:01:35.880 | and create these two files. So you just run it, it creates these two files. And so this

01:01:45.240 | is the one that we have to give Amazon. This is the one that we're going to keep for ourselves.

01:01:50.680 | So following along in the documentation here, it says to click on services, EC2, find key

01:02:05.720 | pairs. Okay, and then we'll go here, import key pair, and whatever, AWS. And this is where

01:02:23.760 | we're going to find the ID RSA pub that we just created. And you can see this, here it

01:02:37.960 | is, right? It's just a big long code. And it's fine, you can all look at this. This is public,

01:02:42.280 | not secret. This is the call thing, right? There's no passwords. And I say import. And

01:02:47.600 | so now we have an SSH key, and we can use that to log in. Okay, so this is just all this

01:02:56.120 | is. Here's all those steps. So renting a server in AWS speak is called launching an instance.

01:03:07.720 | So to launch an instance, we'll scroll back up to the top to instances, and we will say

01:03:14.560 | launch instance. Okay. And it'll say, okay, what kind of thing do you want to run Amazon

01:03:22.560 | Linux or Windows or Red Hat or whatever? I strongly, strongly suggest you use Ubuntu

01:03:27.520 | and the latest version, which is currently 20. So I'm just going to say select. Okay.

01:03:33.960 | And then it'll say, okay, what kind of server do you want? For playing around, there's actually

01:03:39.520 | one that you can get for free. Now it doesn't do, it's pretty, it's kind of slow, right?

01:03:44.240 | But for learning about SSH and Linux and stuff, this is actually a great one to use. It's

01:03:50.080 | no good for deep learning. It doesn't have a GPU. So if I go to G4DN, that's the cheapest

01:03:57.360 | kind of good GPUs we can get. And I'll get the smallest one there. G4DN xlarge. And then

01:04:05.880 | I'll say next. Next. So how big a hard drive do I want? I normally say about 100 gig. Launch.

01:04:18.840 | And launch. And so now it's going to say, okay, when you log into this, which key pair

01:04:25.040 | are you going to use? Okay. So you just select the one that you just imported and say, yep,

01:04:31.240 | I know that I have that. And then launch. And you'll see, Dell says, this has now been initiated.

01:04:42.040 | It's got a code. So this is the thing that I've just launched. So if I click on it, here

01:04:47.160 | it shows me, here's my instance. Okay. So as you, if you haven't done much with servers

01:05:01.640 | and Linux and SSH and stuff, there's going to be this whole world of new stuff for you

01:05:06.400 | to learn about. But this is an opportunity. It's not a problem. So if you're not familiar

01:05:10.560 | with things like IP addresses, that's cool. There's lots of tutorials around at the moment.

01:05:14.760 | But for now, just know this is the unique address, like a street address that your new

01:05:20.840 | computer has. And so we're going to connect to it. So this button here will click, will

01:05:25.080 | copy that address. Okay. So we can then go to our terminal and we can type SSH and paste

01:05:33.960 | in the address. And then the only other thing I do need to do is I need to say, provide

01:05:39.360 | a username and AWS always uses the username Ubuntu for all of its Ubuntu images. So you

01:05:50.440 | say Ubuntu at, and then the IP. And so if I'm now press enter, we're in. Okay. So now

01:05:59.120 | everything I type here is actually being typed on that remote computer. So for example, to

01:06:05.800 | list the contents of a directory, I type LS. Okay. So the thing I'm actually typing into

01:06:11.040 | here is bash, a bash shell. So bash is something, another of these things need to be familiar

01:06:15.320 | with and you can learn about it in that missing semester MIT course I mentioned. You know,

01:06:21.320 | it takes a few weeks to get somewhat comfortable with bash. It's a very different feel to using

01:06:29.120 | a GUI if you're more familiar with explorer or finder or whatever, but you'll find it's,

01:06:33.920 | will be much more productive soon enough because you can replicate things quickly. You can

01:06:38.280 | script things, you can copy and paste things and so forth. Anyway, so here's my, here's

01:06:43.920 | my computer. It's going to sit here running until you tell it not to. Even if you turn

01:06:54.800 | your computer off, your server is still running and that means you're still paying for it.

01:06:58.960 | Okay. So one of the things I guarantee you're going to learn the hard way by wasting money

01:07:04.440 | is that you're going to forget to turn it off. Okay. So to turn it off, you're just going

01:07:10.400 | to go stop instance. Okay. So you make sure you do that. All right. Let's see how we're

01:07:20.480 | going here. So we've launched our instance and we SSH into it. Okay. So keeping a Linux

01:07:38.840 | server up to date and running used to be kind of annoying, but luckily I've created something

01:07:48.960 | called fast set up for you, which makes it easy. And all you need to do is copy this

01:07:55.280 | and paste it into your terminal. And this is one of the really cool things about Linux

01:07:59.260 | and using bash is like in windows or with Mac finder, you'd have pages and pages of

01:08:05.400 | click this and drag that and scroll here. But I've just scripted the whole thing. So

01:08:09.960 | I'm just going to go ahead and paste it over here and it's off. Okay. Now what this is

01:08:16.520 | going to do is it's going to fully set up this Linux server. It's going to make it automatically

01:08:22.720 | update with the latest software. It's going to configure it all correctly. And so forth.

01:08:31.760 | And it's going to ask a minimum number of questions. So I'm just going to show you the

01:08:34.960 | questions it's going to ask you. It's going to ask for a host name. So a host name is just

01:08:39.480 | a more convenient way to access a server. And so you can basically write anything you

01:08:44.000 | like as long as it's got at least two dots in it. So I'm going to call this course test

01:08:50.580 | dot fast dot AI, for example. Okay. And then after an email address. Now the email address

01:08:59.280 | is basically just goes where it's going to send kind of error locks and stuff too. So

01:09:04.480 | maybe we'll say info at fast dot AI. Okay. Do you want to set a password? Probably do.

01:09:12.000 | So hit enter for yes. So I'm going to put in a password. Ask you to type it again. Okay.

01:09:23.720 | Reboot automatically when required. I'll say yes. And that's it. Okay. So that's all the

01:09:29.120 | information that it needed. So behind the scenes. What's actually happening here. Is

01:09:41.280 | it's grabbed the latest get repo from fast set up. And it's running this thing called

01:09:48.800 | Ubuntu initial. And you know this is something you can check out if you're interested. It's

01:09:53.600 | basically 125 lines of bash script, which is going to set up your firewall for you, set

01:10:02.480 | up SSH security for you, set up your swap file for you, set up your SSH configuration

01:10:08.440 | for you, install all the software you need for you, set up your logging and upgrades

01:10:13.960 | for you, set up your password and host name for you. Okay. So it's going to do all that.

01:10:23.200 | And you know this is the kind of thing that if you, you know from time to time you can

01:10:25.880 | just might think oh I'm interested in how X works. And since everything is open source

01:10:31.720 | you can just go in and see how X works. And at first none of this might make any sense.

01:10:37.040 | And so you go oh all right let's pick something and learn about it. Enable firewall. UFW. No

01:10:42.320 | what's UFW? Copy, paste. UFW. Probably not United Farm Workers. Uncomplicated firewall.

01:10:51.720 | Did Jeremy mention firewall? Okay. What the hell's a firewall? And you know you can start

01:10:55.960 | reading right. And then you could be like oh maybe firewall tutorial. Often adding tutorial

01:11:03.560 | can be helpful. Okay. So you know you can start to just jump in here and there. Okay. Don't

01:11:10.240 | get too distracted. We want to spend as much time as possible training models. But this

01:11:14.800 | is how we learn about our tools. Okay. So this is now going and downloading the latest version

01:11:24.640 | of all the software that it's going to need from Linux. So it may be a good time for questions

01:11:28.600 | if we have any Michael. What's your current opinion regarding Swift and Julia as replacements

01:11:34.480 | of Python? So Swift is basically out now. So Google has basically archived the Swift for

01:11:43.440 | TensorFlow project. So you can safely ignore that. Yeah. Julia is interesting. You know

01:11:52.600 | I think it's a lovely language. Nothing has the ecosystem that Python does. So you know

01:12:04.280 | if you use Julia you're going to have to figure out a lot more stuff on your own and you'll

01:12:08.440 | find a lot more hard edges. But I do think at some point Python is going to have to be

01:12:13.640 | replaced and Julia seems like one of if not the most likely thing to replace it. Or maybe

01:12:20.680 | it won't be replaced by Julia. Maybe there will be replaced by something else that's

01:12:25.320 | kind of Python like Jax which actually takes Python and compiles it using something called

01:12:31.840 | XLA into a much faster thing than Python otherwise would be. Okay. Do you think that deep learning

01:12:48.980 | or more traditional ML or stats approaches are more useful for traditional industry applications

01:12:55.400 | right now? So before I answer that question I'm going to press Y which is going to reboot

01:13:01.240 | our computer now that it's all updated. And obviously when we reboot that computer running

01:13:08.200 | at the AWS data center it closes the connection because it's busy rebooting. Okay so we'll

01:13:15.000 | give it a couple of minutes. There's not a single good answer to that question and you

01:13:24.360 | don't really need to answer that question because basically any time you want to try

01:13:31.880 | any kind of machine learning model on a problem you should try a few different algorithms.

01:13:38.920 | And switching from a random forest to a gradient boosting machine to logistic regression to

01:13:43.120 | deep learning is you know an extra half hour. So you should just try a few different approaches.

01:13:53.160 | I find personally for me deep learning is increasingly turning out to be the easiest

01:14:00.920 | thing to get started with and gives me the best results for most projects I seem to do

01:14:07.680 | nowadays. But you know have a look at like Kaggle competitions from time to time there

01:14:12.280 | are still things where gradient boosting machines work better or very often people use both

01:14:18.640 | and ensemble them. But yeah it's not a question that you actually need to answer because it's

01:14:25.000 | you want to get to a point where it just takes you a few minutes to try another algorithm

01:14:30.480 | out and so you don't need to be wedded to one or the other. So I'm just going to see

01:14:36.040 | if I could I don't know how long it's going to take to reboot so I'm just going to I just

01:14:39.040 | pressed up arrow to get back my last SSH command and I'll press enter and we'll see if we're

01:14:42.840 | back we're back. OK so this is finished rebooting. Oh actually this time it says to do something

01:14:56.080 | slightly different which is to add this minus L here. This is the thing that's going to

01:15:00.440 | let us connect to Jupyter Notebook. So I'm going to type excerpt to excerpt from the server

01:15:07.400 | and this time I'm going to add the extra bit of the command. There we go. OK and all right

01:15:17.040 | so the next thing is we're going to install something called MIDI Conda. MIDI Conda is

01:15:20.520 | a very nice distribution of Python the programming language. A lot of people have bad experiences

01:15:30.480 | of their computers getting really confused with Python packages and things conflicting

01:15:37.600 | and all kinds of stuff like that. That's because pretty much all the major operating systems

01:15:43.280 | now come with a version of Python that is used by your computer for you know important

01:15:49.680 | operating system tasks. You should not be using that Python to train your machine learning

01:15:54.680 | models. Leave that Python alone. Right. You should always install many Conda which is

01:16:02.280 | going to give you your own version of Python which is nothing to do with your operating

01:16:05.980 | system as you can play around with as you like. It's really easy just you can delete

01:16:12.720 | the whole folder and create it again in like three minutes. You can create new environments

01:16:17.600 | which is like little testing grounds. You can try different things. This is a very strong

01:16:22.160 | recommendation is to make sure that you install even if you're just playing around on Windows

01:16:27.960 | or a Mac not on a server install many Conda. It's cross platform. You can use it everywhere

01:16:33.880 | and use that Python. OK. So many Conda is now installed. So we now have our own Python

01:16:43.440 | setup. So the last setup step is we have to install drivers for the GPU and Ubuntu actually

01:16:54.080 | comes with something that figures out for you what the best drivers are for your device.

01:16:59.680 | So this is just what this step here is. And so I'm going to look down. Look here it says

01:17:04.280 | recommended. OK. So here's the driver I want. OK. But what I actually recommend is you use

01:17:10.960 | that but also the one at the dash server to the end. That's going to make like not install

01:17:15.760 | the stuff for playing computer games or whatever. OK. So let's go ahead and run these lines

01:17:23.240 | of code. This is a bit here. See this is 460 depending on your graphics card. When you

01:17:28.360 | run this you might have some different number. OK. But since I wrote this today it's still

01:17:33.200 | 460. So we'll go ahead and do that. And this is going to go ahead and install this. Oh

01:17:43.680 | pseudo pseudo is a special thing you can add to the front of a command that runs it as

01:17:51.040 | an administrator. OK. So some things you know by default commands you run basically can't

01:17:59.720 | break your system. Right. Where else things like installing new software you have to tell

01:18:05.000 | it to run it as an administrator. So when you do that it'll ask you for your password.

01:18:09.480 | This is the password that you put in just just a moment ago with the setup. There we

01:18:16.020 | go. OK. Is there a section of the course that people skip over too quickly. Yes. Part two.

01:18:38.080 | But not enough people do part two. And the difference between part one and part two is

01:18:46.680 | the difference between being a pretty handy practitioner you know who can who can do some

01:18:55.600 | pretty good work as long as it's in reasonably well established kinds of areas and versus

01:19:03.120 | being somebody who understands how everything's put together you could you know if you're

01:19:07.800 | told to create a deep learning model on in a domain that's like there are no published

01:19:12.240 | models you'll be able to create one. If you'll you'll understand how to create models which

01:19:17.560 | combine multiple different data types you know you'll it's it's yeah it's it's a really

01:19:29.800 | big thing to to finish and not enough people realize how much is is there. And just the

01:19:38.480 | later lessons in in general you know it can like after you've done three lessons you you

01:19:47.000 | are pretty handy and you'll feel pretty handy right. But it's pretty easy to stop there

01:19:52.720 | because it feels like OK I get it you know I can train a model I get what's going on

01:20:00.720 | and to be fair it does very dramatically kind of scale up in terms of intensity after that

01:20:07.440 | because in lesson 4 you'll have to write your own optimizer from scratch and you'll be getting

01:20:12.560 | into the calculus and stuff. But you know it it it is a big difference in terms of what

01:20:20.120 | what you can do and what you understand. So I think in general you know not enough people

01:20:25.340 | are getting deeper into the lessons. OK. So this is now finished installing the Nvidia

01:20:34.680 | drivers. Normally at this point people say to reboot but there's actually a magic thing

01:20:41.640 | you can do which means you don't have to reboot. And the Nvidia provides something called Nvidia

01:20:49.920 | SMI which will tell you about your installed GPUs. And so if you run it and it pops up

01:20:56.680 | anything at all other than an error it means that you are successfully have your GPUs installed.

01:21:01.800 | So in this case we have a Tesla T4. It's currently 36 centigrade in there and the most important

01:21:11.720 | thing to know about is that it has 15 gigabytes of memory of which we're using nothing at

01:21:17.200 | all. And there are no processors currently running on the GPU. So if you're finding something's

01:21:25.040 | going very slowly and you're wondering maybe it's not using the GPU you can always run

01:21:30.120 | Nvidia SMI and if it says no running process is found you're not using the GPU. OK. OK.

01:21:39.680 | So one more setup step which is we have to install all of the software all of the Python

01:21:47.040 | libraries needed. So PyTorch, FastAI, Jupyter Notebook and so forth. And so I've created

01:21:53.840 | a package which has that whole lot. It's called Fastbook. If you're if you've used had a condor

01:22:00.320 | or mini condor before you might be surprised who says Mamba rather than condor. You should

01:22:05.660 | definitely use Mamba and not condor. It's way way way faster. So anytime you see something

01:22:11.160 | saying condor install you should instead type Mamba install. It's way faster. OK. So off

01:22:20.760 | it goes. Members now going to install all of the all this Python software getting installed

01:22:26.440 | for us. PyTorch is well over a gigabyte. So this is going to take a few minutes just because

01:22:35.880 | it has to download that that whole thing. And yeah that can take a while. So while this

01:22:41.600 | is going do we got any more questions Michael. Do you recommend any software for experiment

01:22:50.640 | tracking. So the most popular experiment tracking software would be TensorBoard and Weights

01:23:00.480 | and Biases. Experiment tracking software is stuff which will basically you can use a faster

01:23:08.080 | call back and you basically will say train whilst tracking with TensorBoard or train

01:23:14.720 | whilst tracking with Weights and Biases. And what it will do is it will kind of create

01:23:20.920 | a little database showing you all the training results from all the different experiments

01:23:26.160 | you've run and create some little graphs of them and so forth. Personally I don't use any

01:23:35.920 | experiment tracking software. And the reason I don't is I found that many many many people

01:23:43.480 | just about everybody I know who uses them finds it incredibly distracting. So the trick

01:23:49.000 | to training models is don't watch them train. So if you've done any C programming it's like

01:23:54.560 | don't watch it compile right. Go and do something else preferably set up your next experiment.

01:24:03.400 | Experiment tracking software just makes it so tempting to look at all the pretty graphs

01:24:08.240 | in my opinion. So I would suggest get it running leave come back when it's done and there should

01:24:18.160 | be a bit of reason you were running that experiment. So check whatever that reason was right. Having

01:24:26.880 | said that if you're really sure you need the services of experiment tracking software

01:24:33.680 | for what you're doing and there are some things that genuinely need it then I think Weights

01:24:39.120 | and Biases is the best at the moment. I think it's really great. And furthermore they've

01:24:46.360 | hired lots of fast AI alumni and they're super nice people. So definitely recommend that.

01:24:59.880 | So that's all the installation. So the last step is just to grab the book the notebooks.

01:25:06.760 | And so you use something called get clone to grab a repository of code and this is going

01:25:11.640 | to grab the fast book repository. Paste. So you can see it's saying cloning this repository.

01:25:25.680 | And so you'll now find that there is a fast book directory so you can CD into it. And

01:25:34.000 | there is our book. OK. So I think something on Anaconda is going slowly so we're not going

01:25:43.280 | to wait for it to download. But so I want to show the very last step. But the very last

01:25:48.480 | step is to run Jupyter Notebook and then you'll be able to click on the URL that pops up and

01:25:54.000 | it'll bring up something that basically looks just like we saw in Colab. But the nice thing

01:25:59.640 | is everything you save like everything you do will be remembered. So all of your experiments

01:26:07.800 | are going to be there. The data sets you download is still there. So on and so forth. So that's

01:26:17.400 | that. So when you're done it will remind you here to as I mentioned before stop your instance.

01:26:26.840 | So you can either choose stop in this menu or you can choose stop here. Or personally

01:26:35.600 | what I quite like to do is to run pseudo. Remember pseudo is this thing that runs as

01:26:41.800 | an administrator shut down halt now. And so that shuts it down from here without having

01:26:49.760 | to go into the AWS GUI. And there we go. OK. So if we look back at the EC2 here in a moment

01:27:08.120 | this will switch from running state to stop state. OK. So I think that's everything Michael

01:27:17.680 | is there anything else to cover. OK. Great. All right. Well thank you everybody for listening

01:27:25.120 | into Lesson 0. And I look forward to hearing how you go with Lesson 1 and seeing your projects

01:27:38.120 | that you create. And don't forget to get involved in the forums. If you do get stuck with something

01:27:45.160 | the first thing to do is to search the forums because out of the hundreds of thousands of

01:27:50.080 | people that have done this before somebody's probably got stuck in the same way before.

01:27:54.320 | So hopefully they can answer your question. Otherwise feel free to ask your own questions

01:27:58.440 | and hopefully somebody will answer for you. Thanks everybody. Bye.

01:28:03.160 | (audience applauds)

Lesson "0": Practical Deep Learning for Coders (fast.ai)

Chapters