- 4 months ago
In this hands-on demo, we explore how to use Azure OpenAI Service to unlock the power of AI and integrate advanced capabilities like natural language processing and code generation into your applications. This session is part of the AI-900 Microsoft Azure AI Fundamentals course and provides practical insights into utilizing OpenAI models on the Azure platform.
What you’ll learn in this demo:
How to set up and access Azure OpenAI Service
Walkthrough of integrating GPT models, Codex, and DALL·E into applications
Real-world use cases of Azure OpenAI capabilities in text generation, conversation, and content creation
Best practices for leveraging Azure OpenAI Service to build intelligent applications
Hands-on examples and tips for developers and businesses
🎯 Perfect For:
Beginners in AI & Machine Learning
IT Professionals
Students & Developers
Azure Certification aspirants
📚 Explore Our Other Azure Courses: https://www.youtube.com/@skilltechclub
What you’ll learn in this demo:
How to set up and access Azure OpenAI Service
Walkthrough of integrating GPT models, Codex, and DALL·E into applications
Real-world use cases of Azure OpenAI capabilities in text generation, conversation, and content creation
Best practices for leveraging Azure OpenAI Service to build intelligent applications
Hands-on examples and tips for developers and businesses
🎯 Perfect For:
Beginners in AI & Machine Learning
IT Professionals
Students & Developers
Azure Certification aspirants
📚 Explore Our Other Azure Courses: https://www.youtube.com/@skilltechclub
Category
🤖
TechTranscript
00:00with this it's the time for a demo of Azure OpenAI so let's see that thing now
00:14I'm there in my Azure portal this is my portal.azure.com I have already signed
00:19in with my account and as you can see here we have a service called Azure
00:23OpenAI service when I try to create the service I want you to understand one
00:29thing that even if you have an active Azure subscription it's not necessary
00:34that you will be able to create an Azure OpenAI service with that when I create
00:39an Azure OpenAI service and if I choose a subscription in which Azure OpenAI
00:44service is activated then only you'll be able to create the service as you can
00:49see here in my case I have one more subscription in my same account if I
00:53choose the other subscription immediately is showing me that in this subscription
00:57you do not have a configuration which is allowed for OpenAI and this is
01:02something which is blocked in the quota ID feature or something like that if you
01:07want to unblock this you have to request to Microsoft so there is a form you can
01:11click here on this particular link and that's going to take you to this
01:15particular URL aka.ms.oai access OpenAI access this is a nothing but a Microsoft
01:23form which you have to fill up if you fill up the form with the proper details
01:26configuration inside that Microsoft will take 24 to 48 hours to either approve or
01:32reject your request if it is approved in that subscription then only you'll be
01:37able to use Azure OpenAI otherwise you won't now in my case in one of my
01:42subscription I have OpenAI configured so I'm just going to select that I'm going to
01:47specify a new resource group like AI 900 RG this is going to be a name of my OpenAI
01:56service so I'm giving Maruti AZ OpenAI with some number at the end of this the
02:05pricing tier is default standard at zero we're okay with this everything else is
02:10fine I'll just go with next next and review plus create once this is your OpenAI
02:15services provision successfully it is going to have all the pre-built models
02:21configured from OpenAI now when I say all the these are going to be those
02:27models which are approved at the time of the approval process from Microsoft site
02:32if there are few models which are legacy models if there are few models which are
02:37still in beta and it is not coming into your Azure OpenAI service you can
02:41further request to Microsoft for adding those models and then they will do the
02:45needful things to allow you to do that let's wait for a few moments and then it
02:52will be done okay the deployment is completed let's click on go to resource
02:57okay I'm going to click on go to Azure OpenAI studio this is going to help me to
03:05directly open the Azure OpenAI studio in the separate tab okay we are inside the
03:12Azure OpenAI studio I'm just going to configure certain things here but before
03:17that let's check few things this is a section which is known as playground the
03:22playground has the chat completion dali and assistant kind of a four options
03:26inside that we have a management section where we can have deployments which are
03:32nothing but our own customized models which we have deployed inside this then
03:38we have model section which is going to provide all the available pre-built models
03:42from your Azure OpenAI site you can see we have DALI 3.0 we have GPT 35 turbo with
03:49the different kind of model versions available with that in same way we have
03:53some other versions of GPT and we have some embedding models now if I click on data
03:59files this is going to help me to upload my own data set onto on top of this so
04:05you can fine tune your pre-built models with your own data set and then you can
04:10use it for your application and then we have a quota configuration which is
04:14available here now maybe there are chances in a specific region in the
04:19specific subscription you will not have a proper quota configurations done then
04:23you won't be able to use the model so the capacity unit for each model is
04:29something which is pre-built and configured from this side and you can request for
04:33a higher quota increase if you want to try certain more number of requests you
04:39want to try in the particular region let's try a few things let me go to my
04:45model section we have one model which is this one GPT 35 turbo 16k now I want to
04:51use this model and for that I have to go to my deployments I'm going to create a new
04:57deployment and I'm going to choose that model from the list which is GPT 35 turbo
05:0216k while creating the model I'm going to select that whatever version which is
05:08automatically update to default is perfectly fine the deployment name I'm
05:12specifying something like Maruti GPT 35 T 16 K now remember this is the name which
05:23you have to use if you want to use your own model in your application while developing the
05:28application so not the not the model which is pre-built you have to use the deployment
05:33name which is configured here and that's the reason I'm giving some unique name here it
05:38is also advisable while configuring the deployment you choose a proper rate limit for that so we
05:43have tokens per minute kind of a rate limit I'm going to set up to around 5000 tokens per minute which
05:50is I think more than enough for my demo purpose right now and do you want to enable dynamic quota
05:55I do not want to enable right now so I'm just going to click on create this deployment will
06:01be done in few seconds once your deployment is visible here you can use this deployment in your
06:07playground so I can just select this and I can choose open in playground which will directly take
06:13me to a particular playground associated with that model because it's a GPT model is directly taking me
06:18to the chat playground where I can use this model in the right side configuration section and you can
06:24see my deployed model is coming in the drop down I also have some parameters configuration here like
06:30temperature top P frequency penalty this all configuration things are going to help you to
06:37randomize your content and if you are getting a proper response or if you're getting some randomized
06:43response it's all going to be configured with this parameters but I'm not going into this parameters
06:48right now in the left side section we have a system message and they are saying that you are an AI
06:54assistant that helps people find information now while learning this thing we discuss this that we
07:00have a system message where we can customize things now maybe I can say that you are an AI assistant but I'm
07:08saying you are a helpful assistant AI assistant that helps people in finding information and so I'm specifying your
07:19helpful assistant that response finding information and always response in a cheerful manner now when I specify this thing let me start
07:37it's saying like hello now before I do this thing I have to just apply changes continue so this changes are applied system message changes are applied I can also use some examples if I want I can add more examples here but I'm just going to say hello
07:55when I say hello when I say hello it's going to wait for some time and if the API model is ready it's going to give me a response but there are chances that there is something wrong which happened here and it's not responding here they are saying that if you have created the deployment within last five minutes you have to wait for a moment now this is a common issue with Azure OpenAI service deployments when you initially deploy them so you have to wait for some time you have to wait at least for five minutes until and unless the model is ready and
08:25and then if it is ready with the deployment then you can try that so this is how it is I'll wait for some time and then I'll try it again after some time so let's wait while this model is getting ready I'm also going to create some more deployments for me so that I can use it in the demo so let me create a new deployment which is going to be using which is going to be using DALI 3.0 domain name I'm giving deployment name Maruti
08:54DALI 3 and then in this we have capacity unit and then in this we have capacity unit and I'm going to click on create now I'm just trying to create multiple deployments here so that by the time I want to use it it should be ready let me try the chat playground once again if that is working fine otherwise I'll try with the DALI I'm again sending hi or hello if it is ready it's going to show me a response yeah you can see now it is ready
09:24it's showing me hello how can I assist you today and now I can provide my kind of an input I can provide an input which is going to be generating some kind of list or I can provide an input which is showing me some configurational code generation kind of thing it's all up to me how I want to generate it so I can see that let's say I am traveling to New York tomorrow
09:53tomorrow I want you to give me a list of places I can visit with my family
10:15now when I say this it's going to give me a proper response based on the generation it's showing me that's wonderful New York City has so much to offer for the families here are some of the popular places you can visit and they're showing me lists like Central Park Statue of Liberty Times Square American Museum of National History and likewise they're also showing me remember to check the opening hour specific guidelines due to COVID-19 so it's showing me all the response with the proper consideration
10:45configuration with that now I'm saying that this is something which is a proper response now I can maybe add on some more examples some more configuration in the left side and then I can change this configuration now let's say right now this is my helpful assistant so you can just help me with this kind of thing what normally chat GPT can do now let's say I'm saying you are a helpful
11:07coding assistant so I'm just saying code assistant
11:14to generate
11:19code in Python
11:24and C sharp
11:26C sharp
11:28and C sharp
11:29now I'm intentionally specifying two languages here and now I'm going to say apply changes
11:35continue
11:37so this time we are saying you are a helpful code assistant not a normal AI assistant and I'm saying that
11:43hello
11:45can you generate
11:48a code
11:50for
11:52a loop
11:54now I'm just specifying this if that configuration is working it can maybe show me some code okay they are showing me how I can generate a Python loop example
12:04so it's a for loop which they are showing me in the C sharp also they are showing me and yeah that syntax is identified showing me that correctly
12:12now let's say I'm saying I want to generate a configuration for
12:16for
12:17okay
12:18give me
12:19give me
12:20a sample code
12:23for
12:24Fibonacci
12:25in
12:26C sharp
12:27now I'm not using C hash this time I'm using C sharp intentionally
12:31now
12:32maybe while typing this also I have made a mistake in Fibonacci spelling but still they understood that thing so in my input I specified a wrong spelling intentionally so I specified Fibonacci
12:44but this is the correct spelling actually they are they also understand the spelling mistakes they also understand what you are trying to say inside that and they have given me a class Fibonacci with the C sharp main method with some kind of a configuration logic inside that with a loop and they are saying in this code we define the calculate Fibonacci function that recursively calculate the Fibonacci number for a given index now this is how a code generator or a helper is going to help me
13:09now same way if I want to try some other models I can do that and I can play with the playground if I have the deployments done with this
13:16let me click on DALI now the DALI is going to use my DALI 3.0 model so you can see the deployment it is coming because I have already deployed it in the prompt I can provide a prompt generation for this like I can say generate
13:36generate image for Azure AI conference and community meet now I just provided something like Azure AI conference and community meet maybe we are planning some community meet next month and I want to generate some
14:06image for that now when I provide this kind of insufficient prompt most of the time the generation of the image is not going to be proper so I have to properly provide more and more details inside that like you can see here they have generated an image for Azure AI conference and community meet now this is like a cartoon right now they're showing me a cartoonized image where a conference where multiple developers are sitting there and then they have laptops they have a screen they have a projector and this is the Azure AI conference which they are seeing okay
14:36now this is something which I can download if I want I can generate a new image if I'm not happy with that I can copy this prompt or I can see the code configuration of this all these options are available let's say I'm going to click on generate a new image
14:50the moment I generate a new image this time I'm not changing the prompt the prompt is same and I'm just saying no I'm not happy with the first image I wanted to generate a new image with this
15:01when this is going to generate a new image this time it is maybe much better than the first one it's showing me a proper sitting in the different style it's not a this kind of a roundtable conference thing this is something which is different and this is maybe an image which I got a little improvised now I'm just going to say okay generate an image for Azure AI conference and community meet with real people
15:27and Azure cloud logo visible on projector visible on wall that's it now this time it's showing me Azure cloud logo with AI on the wall and then showing me a lot of people are meeting and greeting each other
15:56now I can play with this I can improvise the prompt I can generate more and more images like this I can also specify the tile size configuration here I can also have a settings configuration where I can change the image size image style now by default is a vivid image style which is selected here let's say I'm going to change it to natural and then in the image quality is showing me standard I'm keeping it standard and I'm going to save now just image style I have changed and I'm again going to click on generate
16:23now this time it's not vivid it's something which was which is natural now so this one is vivid which was generating this kind of an image even though when I said with real people they are showing me this thing
16:34now this time I said it is going to be natural so you can see these are like humans real humans are now visible with the Azure logo and maybe this is the image which I'm more able to relate with so this is something
16:47how you can customize this thing you can see the view code part of this and you can just directly choose your choice of language association with that mostly it's going to be visible with Python you can also take the JSON format of that prompt configuration and then you can associate with that so this is how it is this is what Azure OpenAI Playground is all about
17:07this is how it is this is what Azure OpenAI Playground is all about
Recommended
28:43
|
Up next
23:28
20:05
20:09
29:21
12:03
1:15