Hi my name is J.Weathers AWS SolutionsArchitect Professional and MCSA cloud platforms today we're doing and getting started with AWSSageMaker.
Aws and sagemaker
We will deploy AWS Sagemaker instance define IAM roles to work withs3 review the Juniper notebook and algorithms that are actually deployedwith the AWS SageMaker instance at the end we'll overview what we learn andtalk about next steps so first things first let's login to our AWS account re:Invent was this year and was awesome Andy Jazzy looking great guys he did an awesome job of re:invent ok so we'regoing to sign into our console it's a couple ways you can get to AWS Sage makeryou can scroll down under machine learning clicks aja maker it's underyour recently visited items so we'll be here or you can just type in the sagemaker first thing you want to do is create a notebook instance so we'regonna click create notebook instance we're gonna name itdeep learning demo for this demo we're going to use a ml.t2.medium instanceyou can select ml.m4.xlarge or ml.p2.xlarge but for demo purposesthe t2.medium will be just fine. For the execution roles you can actuallycreate a new role or enter a Amazon Resource Name (ARN) from a existing role or using the existing IAM role, but right now we're going to create a new role so for a specific s3 bucket whatyou need to do is actually get the name of a bucket that already exists.
So letme go here well I'm gonna login into AWS again all right we're going to go to s3. Okaywe want to go to s3 that's our storage where our data our csv files will bekept. The bucket I'm going to choose is Deepapp1 so you see this bucket righthere DF one I want to type that here it's called deepapp1 rightokay so this bucket s3 policy will be created for this bucket specificallyand we're going to go through that so Deepapp1 is a bucket that I already have I'mgoing to click create role so now this is the role that is actually with mypolicy when I click on this it takes me to identity access management (IAM) here's myrole and here's the policy that was actually created for this so you can seethis is the resource and s3 deepapp1 and I'm allowed to get objectswhich means pull objects from s3 put objects which means I can put objectsinto the bucket or delete objects for the bucket for deepapp1 so thatlooks great now we're going to talk about the Virtual Private Cloud (VPC) so you have your VPCwe're selecting the default VPC for this particular demonstration Amazonrecommends that if you're doing a production based app that you can youcreate your own VPC and you can do that by utilizing cloud formation and cloudformation is a tool that you can use to automate all your infrastructure. So nowwe're going to select the subnets in this availability zone in this region usthe East one and the security group I'm going to use a default security groupfor encryption keys that's basically you have to protect your data with anencryption key you can do this you can go to encryption keys you can use a keythat you already have like I have this key right or you can create a brand newkey so you can create a key you can just have an Alias for the keydeep learning demo. demo key, advanced options kms only next step you can do itfor lets say an app env you can do dev that's the Tag next step then youcan choose other things that can access it we're going to go to the next stepswell actually we're going to go to our stage maker execution role both of those click next steps and click finish sohere is everything that can actually access with this roll the stage make forexecution rolls both that we just did and you can see all create, describe,enable, list, put, delete click finish we have our app deep learning demo key wecan copy this on go back to sage maker paste the arm here and there we have ourencryption key create a notebook instance. as you can see deep learningdemo is pending that means it's deploying.
I have a instance in service and I havean instance stopped. The great thing is if you're not working on your sage makerinstance or in your Juniper notebook you can stop your instance start it and allyour data still be there.
So while this one was pending we're going to go aheadand open a instance that I've already created. I'm gonna open that and here wesee sample notebooks. so when we actually take a look at the sample notebookswe're going to... you can actually see the models that are deployed with AWS sagemaker.
You have advanced functionality and introduction to Amazon algorithms,introduction to applying machine learning algorithms. these are pre-installed and aresamples with data and algorithms are already there and you can actually startto train and predict breast cancer using this linear learner model with featuresderived from images of breast mass. So you can actually walk through this right now and start learning how to do modelingfor breast cancer which is amazing. Now if we go back here we're going to take alook at a few more so let's go back that was actually introduction to applymachine learning let's look at introduction to Amazon algorithms so ifwe take a look at these we have factorization, imageclassification model, IDA topic modeling linear learner modeling with MS NISTthat's the data set for image recognition and then you have sequenceof sequence and x boost with mnist let's go back a little bit more go backto where we were and now we have the sage maker Python SDK here you have your Gloun on for and mnist. sentiment analysis and then you have your Tensor flow models for distributed mnist you can actually start tolearn tensor flow and work on tensor flow if that is your framework of choiceokay we're going to go ahead and wrap up this particular session for the Amazonalgorithms in part two we'll actually start to look into introduction toAmazon machine learning and the first model of will run is the breast cancerprediction. As promised for the overview we deployed the sage make for instancewe defined the I am role to work with s3 will review thefor notebook and algorithms for our next steps we encourage you to join our slackchannel and you can go to our website here click on our slack channel buttonwe want you to be a part of our community we want to help you learn deep learning and answer any questions that you might need you might want to join ourKaggle team and help us solve and win Kaggle competitions. You might form your own kaggle team from the deep learning teams resources and talent or you mightjust want to become a part of the deep learning team and start teaching how youbuild algorithms. Regardless of what you want to do, we want to hear from you. Wewant to get to know you and we want to create great change with this newtechnology.