Name: SAP AI ; Predictive Maintenance ; Workflows
Uploaded: 2023-9-7
Duration: 585 s

Share feedback

<p><u><strong>Planned maintenance of Practice Systems</strong></u></p>

Planned maintenance of Practice Systems

Why create an SAP account?

Enhance your learning experience

SAP Learning

Choose your subscription

Manage your subscription

Live Sessions

Hands-on Practice

Certification

My Learning

Ready to start learning?

<div><div><div><div><div>In this playlist by Tahir Hussain Babar (Bob), from the SAP HANA Academy, we use SAP Artificial Intelligence to build a Deep Learning model on Acoustical Sound for Predictive Maintenance with SAP AI Core. In this video, we look at how the SAP AI Core Github Application works together with the serving and workflow YAML files, which are placed in a "workflows" folder locally and in Github. We also look at how our Workflow YAML files are synchronized.<br><br>00:00 - Intro<br>00:50 - Applications<br>01:15 - Local "Workflows" folder <br>02:12 - Training Workflow YAML<br>04:18 - Serving Workflow YAML<br>05:51 - Moving Workflow YAMLs to Github<br>06:44 - Syncing Applications<br>07:48 - Checking Sync Status in Notebook<br><br>SAP AI in BTP: <a href="https://www.sap.com/products/artificial-intelligence.html" rel="nofollow noopener noreferrer" target="_blank">https://www.sap.com/products/artificial-intelligence.html</a><br><br>Please note:<br>                     This video was originally published on the SAP HANA Academy YouTube Channel                      and therefore may contain references to SAP HANA Academy and YouTube.                      The tech world is always evolving and the SAP HANA Academy YouTube channel                      is set to close on December 31, 2023.                      At that time, references will no longer be valid                      and will gradually be removed when the videos are updated.<br>Also, this video includes automatically generated captions                      and therefore might misrepresent the spoken content or display incorrect product names.</div></div></div></div></div>

SAP AI ; Predictive Maintenance ; Workflows

Welcome to the SAP HANA Academy. My name is Bob, and in this video series, we'll be looking at SAP Artificial Intelligence. In this video playlist we'll be building a deep learning model on acoustical sound for predictive maintenance with SAP AI Core. And in this video we're going to be looking at the concept of workflows. Examples of what these workflow files are used for is to instruct SAP AI Core about executing docker images, inputs which can be provided in the form of datasets, and which outputs to generate. So in the previous video what we did is we uploaded two docker images into our docker repository. Now, if I go to my SAP AI Core Administration tab within Applications, this is our application that we built in one of the first videos, specifically when we went through the onboarding recap. Now, we can see here two expanded tabs. One of them is Source Repository. This will of course detail where in the github these files will be located. So if we go to github we can see that in the workflows sounds folder within that repository, this is where we will need to put our files. And if I go to that files folder which Jupiter is using, you'll see that we've got workflow-images, which is what we created for the previous playlist. We're going to create a local folder which will be used for our workflows for sound. And again, this will be our local copy of those workflows. So I'll create myself a new folder and we'll call it workflows-sounds. Again, this will be a local folder. So we're going to have a look at two of these files. They're both going to be in YAML format and I'll highlight some of the key aspects of each of the files. So workflows within SAP AI Core, these are executed using what's called the Argo workflows open source project. Now, this is an open source container native workflow engine for orchestrating parallel jobs on Kubernetes. And this is the file that we'll use for training our model. So what we'll do for this file, we'll save it as the following filename within my workflows-sound folder. And we'll call it training-workflow.yaml, because as you can imagine the files are in YAML format. Now for SAP AI Core, workflow templates and in fact also serving templates are mapped as what are called executables. Mapping requires certain attributes in the metadata section of this template, and the AI API uses the annotations and labels in the template to locate what are called scenarios and executables. Now, something here which is quite important is this concept of resource plans. And these are used to select resources within your workflow and serving templates. Different steps of a workflow can have different resource plans. Now, in general, if your workflow needs or profits from GPU acceleration, you should use one of the GPU enabled resource plans. Otherwise, you can use a resource plan based on the anticipated CPU and memory need of your workloads. So of course this depends on how much computation you're going to be doing to train or serve your model. Again, why are these files used? Well they'll be used for your docker images. And of course this one will be used for your serving image. So here within container image we'll need to specify our image and its tag. So I'm going to change my image here on line 37 and I'm going to put the same name that we used when we pushed our image to our docker repository. So it's going to be the name of my docker account forward slash the name of the repository colon and then the tag. So we don't have to change anything else, we can just save and close this file. What I'll do next is I'll paste in the code for our serving YAML file. And you'll notice that the file contents look quite similar. The difference between this file and our training YAML file is the YAML specification. And it must follow what's called the KF Serving Specification. Again I'll highlight some of the main parts of this file which are different. So if you look on line 18, this artifact is actually the name of your model. Again we're using that resource plan. This is the specific resource plan I'm using which utilizes GPU. And what I'll do is I'll save this file and we'll call it serving underscore workflow dot yaml. So if I scroll down now what we'll need to do is deal with the Docker image. Just like in the previous file, we'll need to define our container image. And this is the one that we're going to be used for serving of course, so I'll paste it in here. And it's called sound seg serve and of course we need to include that tag as well. Additionally, for authentication, this is the Docker registry secret which it's going to use. So now that we've configured our local YAML files, what we'll need to do is put them into our GitHub repository. So all we need to do is actually drag them to that workflows-sounds folder. So I'll move this folder, we'll go back to github. You can see at the moment there's that X within that workflows-sounds folder. I'll upload the files. And we know they're in our Jupiter Workflow Sounds folder. I'll select them both and select Open. And then all we need to do at the bottom is to commit the changes. So if I go into Workflow Sounds, here we've got those two files. And here we've got the extra file x, but now because we've created those two files, so there's objects in that folder, we can delete this file. If you delete that file before you enter those two objects, the folder will also get deleted. So now we have the files in our GitHub repository as well as locally. These are the ones that will be referenced when we're executing Jupyter commands, but this will actually look for those files in the GitHub repository. So all I need to do is refresh and wait for the synchronization to happen. So to recap what we've done is register what's called an application here in SAP AI Core Administration. Now what's going to happen is AI Core will scan that folder and then each workflow file will define what's called the scenario that we'll see within the AI Launchpad. So this synchronization with github takes a few minutes. But here you can see after refreshing and waiting that the sync status is healthy. We've got the date and of course at the bottom we can see the sync details of our various sound files for both training and serving the model. Which of course in this GitHub repository. Now we checked within here within the SAP AI Launchpad the status of our workflow files. But we can also do this back in our Jupyter Notebook. So I'll create myself that header. And then I'll paste in the code. So it's going to search for this app name which is what needs to match of course what we have in the SAP AI Launchpad. Here you can see it's actually called AI Core dash sound dash app. So I need to change this code here to ai-core-sound-app. And then I can execute the code. And here you can see that it's identified the status of those workflows. I mean those workflow templates and those serving templates as well. There's other metadata that you can see within here. So you can see the date the sync started, whether or not it's synchronized, and also the message for each of your templates. So to be clear, now what we've got is two copies of these files. One is the local version that will be used in Jupyter, and the other one is here within our GitHub repository which is referenced by the application for syncing. And again the ones within the workflows sounds folder in the GitHub repository will let SAP AI Core know how to access Docker and which container to actually use. Again, if we go back to Jupyter, we've got those original local files and these are the files that we're going to be referencing in the next couple of videos where we build configurations and executions. and then we're going to go back to Jupyter, we're going to take the original found file. we're going to make a Sneak Peek with a folkspeed which is a nice custom feature. original found file. we're going to make a Sneak Peek with a folkspeed which is a nice custom feature. original found file. we're going to make a Sneak Peek with a folkspeed which is a nice custom feature.