Azure OpenAI Service - Getting Started

hello everyone welcome to the first video in my series on Azure open AI service in this video I'll introduce you to Azure binary service and its features and I'll show you how to get started with it now I have three goals in mind for this series of videos I'm going to cover the features of azure open AI service the basics of prompt engineering and then I'm going to show you how to build intelligent applications with these large language models I'm not going to go into the history of openai of the theoretical side of this large language models like how Transformers work because there are plenty of resources that covers those things now Azure open AI is a powerful combination of Microsoft Azure cloud computing platform and open ai's large language models this service went generally available on January this year 2023 now with this service we can easily integrate open ai's powerful natural language models into our applications using a rest API and you can use these models for Content generation summarization some search natural language to code translations and many more things Azure open AI service host sand provides compute power to run open ai's advanced language models they're basically hosting them and if you're already familiar with Azure services this service will be very easy for you to work with and learn even though I'll be using Azure open AI in this series of videos the knowledge you will gain here will be applicable to open AI as well thanks to compatibility of these two apis if we compare Azure open Ai and openai Azure always offers security and Enterprise grade offerings inherent to Microsoft Azure like identity and access management and private networking and network security now these AI models they're becoming a major part of the applications that we are going to develop they will be a part of their architecture so it's really important we have this private networking and network security features and also Azure has their data encryption and compliance and disaster recovery and Regional availability these things in place so the idea here is that they're providing the Enterprise grade reliability for its uses and the next major feature of azure open AI is the responsible AI Microsoft is doing a lot to ensure that they are providing these open AI Services responsibly and we're using these houses responsibly there are future just like content filters we have a lot of features to filter out the content that is generated by these AIS I'm going to cover these things later in the series of videos now in addition to its language models Azure open AI service provides access to image generation models as well like Dolly as you can see I'm going to show you how to use this service and apis later in this series of videos and the next important thing for us is that the customer data the data actually we sent to these models they will not be used for training the models or they will not be sent to open AI all the customer data sent to Azure open AI remains within the Azure open AI service within that region now that you understand what Azure open AI is let me show you a quick demo to understand this service better I'm gonna build a clone of chat GPT using Azure open AI let me show you how to do that alright now I'm in my Azure portal as you can see we have Azure open AI service I'm going to go into Azure opening a service and if I can search here this is under cognitive Services we have speech language a lot of other services under this and we have Azure open AI service on the top now if I go into Azure open AI now let me click create here to provision our service the provisioning is no different than provisioning any other Azure service basically you have to select the subscription and you have to specify a resource Group now let me call it RG oai service all right now we have to specify the region in which we are creating the service currently we have only four options is the US Frost Central UK South and West Europe I'm just gonna go with UK South and now we have to name the resource I'm going to call it o a i experiments let's see whether it is available and this should be a unique name and now we have to select the price tier this is the only price tier we have now if you want to view the full pricing details you can click on this link and scroll down as you can see this page contains the the what you have to pay to use these models now after creating this service you have to deploy a module to use and these are the available models as you can see we have the basic two Advanced models and core generation models as well this this is the amount you have to pay per thousand tokens now if I scroll down we have gpt4 as well you have to submit a form to get gpd4 access and we have image models like Dali and you have to pay two dollars per 100 images with the service and we have fine-tuned models as well for some of these models it is the same as the default models but if you look at DaVinci you have to pay 10 times more to use this fine-tuned models and to train these models this is the amount you have to pay per computer it's it's expensive and if I scroll bit down also we have embedded models as well I will be touching these embeddings as well in my upcoming videos now if I go back and click next I can specify virtual networks so that the service will be accessible only by a certain set of networks I'm going to keep this publicly accessible for now and then I'm going to click on create as you can see we have a validation error here now it's saying that the subdomain oai experiments is not available even though it's not it is not mentioned it earlier let me just call it uh yeah something like this now let me go ahead and try creating this service again all right now validation is passed now let me click create to create the service even though I have used the Azure portal to create these resources you can still use the CLI to provision these resources as well as you can see with this page you can even deploy these models as well I will be looking into this in my later videos the deployment is now complete now if I go into the resource this is standard Azure Resource as you can see it shows the subscription information we have this service specific parameters like the API type and the endpoint and keys as well now if I go into the keys we have the keys to this service this endpoint and the key is used to access the azure AI API so this is the same with openai as well the non-azure version in this video I'm not going to cover all these things now if I click here we have the Azure open AI service the Azure studio now if I go into chat I can't do any chatting or I can't use any completions or anything because I have no deployments first I have to deploy a model for that I'm going to go into the deployment section and create a new model and I have to select a base model I'm going to go with GPT 3.5 turbo and the deployment name also I'm going to call it gpt35 turbo and I'm not going to use any advanced options here like content filters I'm just going to keep it as default and click on create as you can see the model has been successfully deployed and if I go into chat as you can see I can do chatting here if I just zoom out a bit I can set the system message and I can start chatting with the model now the goal of this video is to introduce you to this and also to build a clone of tech GPT to build the UI for this clone I'm not going to code it from scratch because I already found a nice UI for it I'll put the link Down Below in the chat and if I scroll down as you can see we can easily build and do a Docker run and this will work now we have a model deployment in place and a repository we can use for deploying the UI let me just go into my vs code and show you how it's done I have actually cloned this repository and built that you can use this script to do the same so basically you just clone this repository and you can run this Docker build command to build it and you can use Docker run command and when you run it you can use this open API host to inject the open API host and key as well let me just go into Azure portal and now if I go into keys and endpoint section here we have the endpoint and the keys as well now let me just show you how to run this locally all you have to do is copy this endpoint URL and replace that here copy the key and then update it here all right now if you do one small change and that is since I have pushed this image I'm going to use that to run this now let me just paste it here the image is running and if I go into this link here let me just click on it we have chat GPT running locally now let's see let's say hi all right it works now let's deploy this to Azure for that what I'm going to do is I'm going to use the same script and if I scroll down as you can see I have a script to deploy an Azure container app with the same image and the same parameters if you're using open AI apis you don't have to specify the open AI API type default model and a lot of these things you don't have to specify but if you're using Azure you need to use these as environment variables now let me just paste the open air host and open API key here I'm just going to run this script to deploy the Azure container app the services are getting created the container up the environment is in place now I'm deploying the container wrap alright now let me go into Azure portal again as you can see I'm in that Resource Group in that Resource Group we have the openai service we have the content wrap as well now let me go into that as you can see the Ingress has been disabled probably I should have made it public anyway what I can do is I can just go into Ingress I can just enable public HTTP access accepting traffic from everywhere Target Port I'm gonna enter 3000 here because that is the default Port the container is listening on let me just click on Save now if I go into all view and we have the application URL now I'm going to click on it as you can see our chatbot is running on cloud now what I wanted to give you an introduction to Azure open AI service in this video and I've shown you how to work with deployments and we have deployed a chat Deputy clone as well in my upcoming videos as I told you earlier I'm going to cover the features of azure opening AI basics of prompt engineering and how to integrate GPT models or large language models into your applications if you have any questions comments or video suggestions you can let me know in the comments down below if you haven't subscribed already please subscribe I will see you in my next video thanks for watching

Channel: Meet Kamal Today - Cloud Mastery

Views: 13,892

Keywords: azure openai, openai, gpt, gpt4, azure, chatgpt clone, chatgpt custom ui, Azure gpt, large language models, Azure large language models, Openai tutorial, azure openai turorial

Id: B9IXOJtEKAA

Length: 10min 47sec (647 seconds)

Published: Sun Jun 25 2023