hello everyone welcome to AI anytime channel. In this video, we are going to build an application that will help you generate automated visualizations. And we are going to use something called L, which is a new release by Microsoft. They have built a system that will help you automate this entire data analysis process, and that too powered by large language models. So you can see Lita as a pandas AI killer. I will say that to be honest pandas AI, I think they were a little slow when it comes to working on the active issues or on their road maps. Now we have a new system that's called L by Microsoft, and it really looks promising. If I talk about Lita, the meaning of L is like loved by the people. And when I show you the demo, you will believe that this might get a lot of love from the community. Okay, and let me first show you a quick demo of that. What we are going to build in this video, guys. So if you look at

here I have a examit application, and that's what we are going to develop in this video. I have two options: one is summarize, and the other is question-based graph. So, the good thing about L is that it also helps you query your granular data. You have CSVs right now and the excels and the other on the road map app. You can upload a CSV file, and you can generate different graphs based on different queries or questions. So, it can look at your queries, and then it can generate a graph. How it does it, I will explain that in a very high level in a bit. Now, if you look at here, we have something called summarize, and the other is question-based graph. Let's first look at summarize here. In the summarize, I'm just going to browse a file name, and this is the data I uploaded. Let me first show you a data. It is the same data file that I have used earlier to build a CSV chatbot using llama 2. I'll give that video in.

the description. As well, you can have a look. That video is completely open source. Doesn't work that good, but now we have a lot of workarounds with pandas, AI, and CSV loaders, blah blah blah. Right now, this is my source data. And if I work as a data analyst, or you are a data analyst, and you want to augment your entire workflow using a large language model, this is the agenda in this video, guys. Now, you can look at here. It, this is a 2019 world, different countries and their happiness score. So, you can find out the happiness score, which is based out of 10. Then we have GDP per capita, we have social support, healthy life expectancy, freedom to make life choices, and I hope North Korea is below in this generosity and perceptions of corruption. Now, these are some variables, the metrics, and based on this, this final score has been calculated. And you can look at here, Finland, you know, scores the chart here. Finland, Denmark, Norway, the Nordic countries are doing really good when it comes to

happiness score uh at least according to the data. Now, this is my source data, and I want to generate some visualizations or infographics using lead, and that's what this tool is going to help. Now, once I uploaded here, I can find out all the details about the data set. That's what we do when we use pandas, right? DF do info, DF do sa, DF do type, whatever, whatever, right? We do some basic analysis of our data frame when you use FAS to analyze your data frames or csbs or excels. Now, if you look at here, these are some information, uh, first, it talks about the F name main column, which is overall rank. Uh, it gives you some sample that how does it look like, and it has a number of unique value. It has country. Uh, you can look at some countries like Bulgaria, Mongolia, ETC, number of unique values, 156 countries are there in that CSV that I was showing you. Then we have score. We're talking about their score, and we have some that's really good. We also getting standard deviation, mean, and maximum of

that particular column, that's fantastic, right? That's what we used to do with DF do info or DF do mean or Max, whatever. And then we also have different columns and some information over here. For that, I will just, I'll minimize it here. That's how you minimize. Once click here. Now, the interesting thing, guys, it also gives you questions, and I will cover what goal means, by the way, you know, in a bit. It has different, it has four major features: goal, summarize, visualize, and there's one more that we'll cover in detail. Now, if you look at the goal, the first question it gives me, how does the score of countries vary with their GDP per capita? Now, suppose if I am a beginner in a data an analytics field or even if an experienced person, I, when I see data for the first time, I might not visualize in my mind that okay, what kind of chart or what kind of graphs I can generate, what are the two, three different variables that I can consider, can I do a scatter plot, can I

do a bar plot between two variables or two columns. Now, this will give you a very high-level quick overview. Okay, these are the possibilities that you can consider, and you can draw it. Okay, you can use this tool. Or now we have built this app, right? You can just upload your data and play around it. Now, if you look at the first goal, it gives you the question. It also gives you how the visualization will look like. It says, Scatter plot of score versus GDP per capita. That's the heading of your graph. Scatter plot of score versus GDP per capita. But it also gives you rationale, which is really important. Now, if you look at here, we have one goal, we have two goals, and then we also have the chart here. The chart, which I only get it for the first rule, where we just, we did chart and the first index of the list. That's what we're looking at, the first element of the list. So we'll see that in the code when we're writing.

the code now. If you look at the goal one, we have a detailed information about how are we going to execute that goal and get our chart or Graphics or infographics, whatever. Now, if you look at here, this is the question: how does the score of countries vary with their GDP per capita? It gives you a detailed information that this visualization will help us understand the relationship between a country's GDP per capita and score. Not only you just go and randomly generate a chart, but there should be a business logic or there should be some logic behind that chart, that what kind of value we are going to create or what kind of impact it can create from a decision-making standpoint. That's what also giving you from rationally here. You can see the rationally on the string here is the visualization, and the other one is which countries have the highest and lowest score. So, this is the second chart that I can generate, right? I can generate a pie chart or some other charts like that, and you can find it over here, bar chart.

of country or region versus score. And then we get a chart here. For the first one, because we are only using the first element. But you can do a for loop and you can also visualize other charts. I will show in the code when we writing it. How does the score of countries vary with their GDP per capita? Now, this is a summarization of your data. You have a large number of CSV where you have a lot of rows inside a CSV file. Or how can you just upload it here on an application and get a quick summary of your data? And also with some visuals and charts. That's the purpose of this. And also gives you with a lot of questions. And you can do. Right now we have only two goals. You can do an number of goals. You know that depends on how. What is the purpose of your use case. Now this is the summarized part. If you go back to. Let me just copy this here guys. I'll just copy this here. And I will go to question based graph. Now when you look at question.

based graph or the query. We have a subheader here. It says query your data to generate graph and here I'm just going to browse files. Let browse the same file. It's already there and here I'm just going to write that question that I've copied from there. You can also do query-based graph generation and that's why its lead is really, I mean it's looking promising because it provides rich contextual information to a large language model. Once you upload a data, right, a CSV file which is granular in sense, it looks at the context behind your entire CSVs and then it gives to a large language model which generates a chart. And it's again, it's a multi-generation steps, it's not a single generation step and that's what they have provided a research paper. We'll look at it in a bit. Now here, once I click on generate graph, you can look at here the app starts running and here for this particular query, I get the

response from a visualization execution. There's an Executor that works in the back end, and I have kept it veros because I want to explain you that it first gives you a b 64 uh string. Over here, it's a b 64 uh encoder value that you get. Uh, you gets it here. You have to take it and convert it to an image. That's what I have done. Now you can look at it. Also, it gives you the code, by the way. Now, if you want to use this code or run inside a backend process as well, you can also do that. And here we got our image for this particular query that we have asked. Now imagine how easy it makes it for me if I'm not good in a data analytics field. That's the one uh area I'm looking at. Even if I'm a very good data analytics guy, this can augment my entire day-to-day uh work that I am doing at my workplace. That's what the power of leader guys uh that Microsoft has released. Now this is on generating the graph. Here you can also set up a lot.

of uh instructions. For example, here you can pass an instruction uh in back end that okay give me graph of red color. It will set the color labels for you. So this kind of customization you can also do it or I will say a personalization you can also do it here uh if you if you are you know if you find some interest in different kind of color labels etc. you can do it. So this is the application that we are going to develop but before we go into the code guys you know uh let me first show you a research paper. Over here you can read the paper here about l so l in French it basically I said right loved by all. That's what l means in French okay. Now if you look at here it says automatic creation of visualizations and it understand the semantics of data the contextual u meaning behind your granular data which is really complex. A lot of people will have question that hey Lama 2 is not performing well for CS SB data and when we building a chat it will not do it to

be honest, it's still not there. Okay, so because even if it's an instruction Ted model, they are not that capable of performing really good, you know, there will be wrong, false projective, there will be hallucination, some GPT-3.5 or other close source model performs a bit on expectations that we have when it comes to granular data. Now, this kind of frameworks, library or systems, whatever you call it, they call, they're calling it a system here, the lead the system, that's what they're saying. Now, this is really going to power that LLM workflow when it comes to performing well on CSVs or Excels. So, this is why it's important, it understands the semantics of data, innumerate relevant visualization goals first, create the goal and then generate it. It's not directly generating it, it's a multi-step generation process, and you can see it over, it says multi-stage generation problem, and they have different pipelines. And these are the four things that their features on their modules. One is a summarizer that

converts the data into a rich contextual meaning. Okay, of your CSV file. That's what it does for the first module. In a very natural Language summary that an llm can understand directly feeding an llm and hell hell amount of a hell number of uh your numbers in your data frame. It's not going to work. LM struggles with numbers and Mathematics. There's no doubt about it. Now, if you look at the next part, which is goal Explorer. It creates the goal and it find out the relevant query that you can ask to generate some graph. That's what the goal is. You can find it out over here. Then it has the visual generator that generates refine and the visualization code. And then it has infog grapher generate infographics. That's the four modules it provides. I'll give the link in the description, but again you can find it out yourself as well. It also has a GitHub repository that you can also read it. You can see here the four thing that we are talking about. Okay, if you look at

the four things summarizer which is Rule plus llm combined, and then the other is llm only. Goal Explorer once you give a natural language to an llm, it will create some goals for you based on that uh natural language information that's available to llm, and then it generates the graph and the infographics. There are the four models and L one of my subscribers today just mentioned right when I'm creating this video just 30 minutes before I wanted to create a video on L, and somebody I think if you are watching it I forgot your name by the way somebody has just suggested me to create a video on L. What I also want to thank him as well thanks for suggesting me to create a video on this now let me just close this uh let me just close this as well let me just close this as well now let's let's start developing this application guys and we'll explain a bit that how you can do it and not only with close model how you can also use models from hugging face

and I know that some of you might be interested in using a local model. So let's now start developing this application, guys. To develop this application, guys, you can see I have a folder here that's called le. demo, and I will open this in VS Code. Let me just open this in Code Dot, and then it will open VS Code for me. Here you can find it out. Let me just open it. Now here you'll see I have a requirements.txt. Now in requirements.txt, I have few libraries and packages so lad that is Cap compatible with four different types of model providers. I will say so you can use OpenAI models, and again OpenAI is combined there as well. Then you can use Coher model. If you have Coare model, you can also use Coare model. Then you can also use Palm models like, for example, Vertic AI models. You can use Text Bison, etc. And then you can use HF model, which is Hugging Face model provider. So it does not support or it's not compatible

with quantized model right now. So if you want to use GGF or GPTQ model right now, you cannot do it. You can only use the Transformer-based models, which are not quantized. Okay, the autocausal model you can use now. Here, if you look at I have Streamlit. That's the dependency of the Streamlit application I have to because I will show you. I'm not going to use HF model because that will probably take a bit of time. I will explain that how you can replace the OpenAI with an HF model. There it's just a single line of change that you have to make. And you might get an error that provider has an extra keyword or something think if you are using LLMX because it uses LLMX. I'll show you when I'm writing the code. Let's you can see I have an ENV file. One thing that you have to do you need Python. ENV as well because I'm going to use OpenAI model here. In this case, I'm using GPT.

3.5 turbo so I'm just storing it in an ENV file and just using it in a load. ENV let's me create an app.py quickly. Here, now here we're going to write the code so let do import streamlet as STD and then I'm going to use from Lea so from Le import manager, text generation config, and then the llm. These are the required dependencies, the modules from L. Okay, so we need a manager that will manage your llm and then the text generation config where we will have some configuration related to a large language model and then LM is something that basically you know that will. So llm is something that where you pass your the name of the llm. Okay, so let's let's see that. Now here I'm going to also use from Dov import load. ENV and I also need import OS. I need import open AI and I need input. Let me just do input IUI as well because we want to use P pillow here to save the

image we're going to get a cbond type of image now from P import image, and from IO import byes. IO because we're going to handle the csbs, the uploaded csb, and who going to need import base64. Okay, I really okay fine. Now let's do a load. EnV here, so let me just do a load. EnV, and I'm just going to do open AI do API key, and I'm just going to get that from EnV file that we have using get EnV function, and this is the name that how I have defined in that en EnV file. Now open AI API key. Now we are done with this, guys. Okay, so let's write a function, or I will not probably write a function. Let me just go to my GitHub gist GitHub, and I will just. This is my GitHub, by the way. If you want any code of any of my video, you can just go and take it from there. Now in just, I'm just going to copy this thing from here. Okay, so let me just copy this thingy.

copy and paste now what this function does. Okay, it takes your base 64 string, okay, which is an encoded string, and then it gives you the image. That's what it does. It returns an image, okay, so you can see it says use by S to convert the Bu data to image because uh, uh, leader provides you an image, which is like it contains the Bas 64 of that image. So I, I, I have to use that on a stimulate application, and that's why I'm going to convert that to an image now. Now what we next we are doing here, let's have something called a variable here, L, and here I'm going to use manager, and inside this manager, I'm going to write, let's see, text gen, or you can also define it later, but this is how you define it, by the way, even if you're not using it, we will overwrite this variable later on, but now here I'm going to write, LLM OpenAI. Now if you're using OpenAI, you're going to give OpenAI like this. If you're not

Using OpenAI, in that case, what you have to do is you have to use LLM. And then you have to provide a provider. You can look at this, right? It says param provider. It takes a parameter that's called provider. In this provider, you have to give HF. And then you give model. Yes, model. And then you give the model name. So, for example, if you are using a model from Meta, or for example, if you're using from together computer a Lama 232B model, then this is how it works looks like. Okay, so something like this Lama 32 key. And yeah. And then you can also pass a device map if you have a limited compute. Okay, so this is how you load an open-source model. Okay, but that will take a bit of time on inference as well if you have limited compute power. Like if you have a single GPU, or if you have very limited CPU memory. There. Now, this is what I'm going to do. Now, let's have a text gen config. So, how do you define a text gen config now?

in text, just text generation config. What I'm going to do is let's write text generation config. And here, this is just for we are not going to use both of this that we are writing here, but this is just to show you, okay. So here, you can define a temperature, and for example, 0.5, and then model, and the model is, for example, GPT. This is model. I'm just going to use 33.5 Turbo. And let's use a 0301 model of Turbo, okay. And then let's also use cach equal true. Okay, so you can also pass it. Also has a caching mechanism. So let's do use cach equal to true. Now, this is how you define. So if you're using a, if you want to use an open-source model, these are two things that you have to worry about. Apart from that, rest of your code will remain same, okay. The model name will not be model in that tation config. Will look different. That's you have to pass your provider model, etc., etc. Now let's keep it.

like this now I'm just going to create a stream Le thingy here s. sidebar. select select box is it right yeah and then here let's write uh choose and choose an option or something okay and here let's uh it's a list so let me just write uh summarize and then the second one is uh question based graph now this is the two menu that I'm going to have okay now if menu so if menu equals equal summarize so let's if menu equals equal summarize then let's have a subheader here hd. subheader and let's call this as summarization of your data so a bit of you know decoration on the app okay and you can extend this further further guys okay now let's have a file uploader and in this file uploader what I'm going to do is we're going to use s. file uploader uh Define a label here in not enter upload your file and you can also Define a type I'm just going to uh let's keep csb for now

okay so type csb or you can just not not. It's not needed to provide a list because we only giving one value of tion here, which is type equals csb. But if you have multiple, it's better to pass this as a list if you want to give XLS or something like that. Now, file uploader is not none. Let's check that there's a value inside that somebody has uploaded a file. Then only execute rest of the code. Now, here we have to save the file guys because it except. We have to get the value, the F name. That's where we have to pass now path to save. So let's just do pass to save. And here, I'm just going to call it fin name. CSV. And if you don't want to do the coding alongside me, you can skip this part. But in most of my video, I do coding. And if you want to follow it together, you can follow it. Otherwise, you can just take the code from GitHub or you can just go in the last okay of the

last part of the video. Path to save F name. CSV. Now let's save this. So with open path to save path to save Ed Ed F or Ed file. And then here what we're going to do. We're going to write this. So f. Write that we write on the disk. Uh file appr do get value. Yes, we're going to use get value. Ah, not get attribute. Sorry, I'm going to use get value. And yes, now this will get the value of it. Now this is we are okay with this. So this will save the file. Now let's have a variable called summary. And in the summary here, we start using that module. One of the module we'll use here. So that's called leader do summarize. This is the module we using. So you can look at leader that we have defined on line 22. That's the leader that we're going to use here. Okay, that's the basically the variable. The model that we are using it. Now leader. Summarize. And in this what I'm going to

pass is so in this. Let's pass the file name first. So this is the file name that we're going to save. So F name. CSV. And once you save, save it to going to use. It has a couple of methods. So default means it just gives you. You can look at here right. It gives you summary method default. It has a couple of more where you can have different type of summary. But I just need a high level summary of my CSV file. That's what I'm going to use here. So summary equals uh summary method equals to default. And then just pass list. Let's just pass text gen. Excuse me. Okay. Let's pass a text gen config as text gen config that we have defined on top. Let me do an ALT G. And then you also use Library equals Library. So let me just do o. Okay. Now we are done with this guys. Okay. Now we haven't defined the library yet. We will. We need this Library once we are plotting the chart. Not now. Okay. So this will not. This will Library is part

of the visualization, not the summary. Okay, summarization. Now, this variable holds your summary. Okay, the lead. Summarize. Now, let's do that. So here, let's write SD. Write summary. And this will just write the summary. Now, once we write the summary, next step is to define the goals. So first was the summarizer. And the next part is goals. Now, the LLM will have your summary. Okay, in like a contextual manner. And now, it can look at those summaries and can create some questions for you. The goals for the rest of the generation is step to generate the graph. The next module, which is visualization. Now, in this goals, what I'm going to do is I'm going to use leader. Goal. And inside goals, it's not goal. It's goal. Pass the summary. You have to pass the summary as an input parameter. That's the summary. And then number of rule. Number of goals that you want. Okay, so let's keep it two for now because we are using open AI, which is in a paid model, right? So you

have to pay for it. Okay, so Nal to 2 and text gen config equals text gen config. That's what I'm going to write now. This goal is there. Now, once you have the goal, at least as now we have two goals, we have to use a for loop here. So for goal in goals and here we can just again do s. WR and just you can do goal. That's what you can do here. Now this gives us the goal. Now we have summary. Now and we also have goal. Now the next step is to visualize get the graph out of it. Let's do that. So what I'm going to do here is and this is the only uh thing that we have to write here guys. Most of the code for the next L if which is question based gra will be copy-paste. So I equals z and here I'm going to use Library. It provides you different libr. It has Matplotlib or Plotly also or D not Plotly. I think Matplotlib c bond D3. It has D3 as

well so for simplicity let's keep c bond now in cbon. Now after that, I'm going to again have a text gen config and text generation config here and then let's pass for that n equals to 1 only gives the one image to me. I don't want so we are overwriting here guys okay so temperature and temperature let's keep 0.5 and then let's use cach equal true. Now the reason I'm not uh what I'm doing here I only need one image so nals to 1 now what I'm doing next is once I do uh Nal to 1 okay uh the next is charts so let's have a variable called charts and let's now visualize so we started with summary and then we saw goals and now we're going to visualize it so l. visualize and in this visualization I'm going to pass summary equals summary so we are passing the summary and we're going to pass goal equals goals and you're going to pass let's pass I

here if you have n number of thingy uh the goals, and then text generation config equals text generation config, and then here we're going to pass our library, which is cbon. Okay, so let's pass the library. Now if you want to use D3, if you want to use M plot live, you have feel free to use that as well. Now we have the charts. Now, this charts basically holds a a base 64 uh encoding, as I said. Now we have to handle that, so let's do it. So I'm going to call it image base 64 and and image base 64, or let's call it uh image base 64 string, a little bit of self-explanatory variable, and then here I'm going to call it charts, and we only have the uh, and that's called R. I have done that exercise, so I know that's called that's has been called raster that holds your base 64, so I'm just iterating, I'm just you know I have to look at the subscript of that uh Json.

that I have the list that I have okay instead of Json now charts. Rest it's not a Json by the way. It's a list okay now charts restore now here I'm just going to use image and you can see on top we have this function so let's utilize that function so base 64 to image and here I'm going to pass that IMG thingy uh B 64 string and then just st. Image that's it and this will give me the image now we are done with the first part the first menu now the next is L if so let me just write it quickly uh L if menu equals equals uh it is question based so let's just write question based graph okay question based graph let me just copy this thing from uh I'll just copy it from at least this part okay now let's just copy it to make it little quick come down to okay and let's not call it summarization let's call it query your data to generate graph query your data to generate.

graph this is what I'm going to do and file uploader. It looks good. Upload your CSV. Okay, upload your CSV. And file uploader is not n. Let's keep the same name. There's no problem. Let's change this at least because it will get overwritten. Otherwise, file name 1. CSV. File. This looks nice. This looks nice. This looks nice. Now let's come out of width. Okay, and we'll have a text area. And in text area, I'm going to use h. Text area. And I'm going to write query e your data to generate graph. And let's have a height of that text area as 20000. Looks nice. And if St do button. S button generate graph. Let's keep that. And here if text area is not none or length of text area. Let's do if length of text area is greater than zero. There should be a value. There should be value inside it. Otherwise, don't execute the rest of the code. Okay. Now s. Info. S. Info. Your query. Your query. And it's a concatenation of text.

so text area. Okay, this looks good. Now we have text area as well. And now again the leader. So let's have a leader, which is manager. Let's copy this guys here. Okay, where? Where we? Okay. And we can just use it, by the way. Also, but we should have just called Le equals L leader manager. Uh, text. This looks nice. Text generation config. Uh, text generation config. Let's use this text generation config. Okay. I'm just going to use text gen config from here. And I'm just going to paste. Let's call it 0.2 for this case. Probably let also keep this. This really too high. You know, 0.5, 0.2. And we are okay with that. Now the summary thingy. So let's get the summary thing from here. Uh, you can look at the summary thing. So we first need the summary. And this will become fin name 1. CSB. This is default is okay. Tech gen config is fine. And once we have the summary, now let's have a user query. So your user query is nothing but what is

that text area right. Okay, uh, user query. You can directly pass it as well. Now let's have the charts. So in the charts, I'm going to pass leader to visualize. And here we have to make a change, guys. Here summary will be summary. So let's do that. But your summary is summary, but your goal is not. Now the goals, which is like, which is by default given by leader. Now this goal is a user-based query. Okay, so the query which user has given, right? So that's what you have to do here. So your goal becomes user query. That's what your goal becomes now. And then text gen config becomes your text gen config. That's it. And we are done with now. Okay, so let's do. And this is how if you want to do it in terminal, you can find out charts and the first of the list. And then you have, let's, for example, image base 64. And then you can e chart z. Rest again, just to get the B 64 string from there. And then just have an image.

and image equal to let's use that function. What the function name? This is the function name. And let's pass image base 642. And then hd. Image thingy over here. And in the S do image, I'm going to pass the image. That's it. We are done with our code, guys. Now let's run this and see if we are getting the response. But let me explain you because after that I will cut short the video. Once I show you the demo, I already have shown it. What we are doing here, guys, we have two things. One is summarize. The other is question-based graph. Now we are using a library called L, which is like a system that provides you flexibility of multi-generation steps, which generates automated visuals for your queries. Okay, on top of your CSV file. That's what it does. Okay, and it has some both open source and close source model supports. Now this leader gets open AI. We have some configuration and then we have a couple of conditionals. In the first conditional, what we are doing, we are getting the CSV from end user.

we are storing it, getting the value of it, and then we have three different modules that we are using. Apart from in Graphics, we have used all the three modules. Rest of the three modules, the first is the summary, which provides you the contextual meaning, the semantics behind your data, and then it has goals, the llm, which looks at the semantics or the context of your data, generates some goals, which are the questions probably for the llm to again generate an image or chart, and that's what it does, the visualize. That's what we're doing. The first condition, second condition, is more on the user-based query. If I have some query that I want to generate some chart for, I can use the second part of it. Now let's run this application and see if we are getting any error or it's working fine. Let me open the uh terminal, and here I'm just going to do streamate run app.py. Now once I do streamate run app.py, I'm expecting that it will open the same app I was just showing in the

beginning, let me just quickly give you a high-level walk-through. Again, you upload the CSV. You can see it running. It has given you the summary first. This is a summary step, and you can enhance this further, guys. We can put some more decorations, some various like subheaders, headers, etc. And you can look at the goals now. Okay, the goals that I have got, and you can also look at the chart that I have got. And this is really, really interesting, right? It's really helps you make better decisions on your data analytics workflow. Right? If you want to have different types of charts, graphs, you can set the goal high, right? You can do NALS to 10 to get 10 different types of charts, logic to get it quickly. And then you have question-based graphs. Here, you can ask the question that you know you want to ask. In the question part, you can see it's generating for you. Let's take this. Now we can take this one and try to see if it's like.

generating any graph or something for us. And once I do that, probably it will generate or it will be a messy graph because you know, sometimes if you generate bar graph or plot graph, you can see list index out of range, which is perfectly fine, I think. I guess for me, right? Because you know, it might be the index might be out of range for that particular graph chart. But you got the idea, I hope, right? So, the code is available on GitHub. What I will recommend you to do is I will recommend you to try out the open source models, okay, with that. And I will give the link in the description, that collab notebook that you can follow for that particular thing. Open source model where, but you might get an error if you do pip install leader, because let me just show you quickly what I'm talking about. You know, when you want to use open source, if you want to use open source, you have to look at here. It says lead. I want to show you the real

dependencies behind this particular system that Microsoft has built. Uh, you can, you have to come down and you have to search for, uh, the open-source thingy. You can see with locally hosted LLMs and it uses a library called LLM X. Now, there is an issue with LLM X. You have to downgrade the LLM X version to use Hugging Face models with L. So, if you are facing any error with L and if you're watching my videos, I will recommend you to downgrade the LLM X version. I can give you that in my GitHub link or the YouTube link description somewhere that which version you have to try for if you are facing any error on Windows machine because right now I'm on Windows. So, I'll give the link in the description. Please go ahead and try it out. I really loved it because it's really promising and I'm going to experiment something with open source on some other use cases. So, uh, that's all. I think this is what I wanted to show you in this video. If you have any question, thoughts, feedback for me.

please let me know in the comment box. You can also reach out to me through my social media channels. The credentials are available on my YouTube banner and the channel about us. Uh, if you haven't subscribed to the channel yet, please do subscribe and support the channel so I can create more such videos. Uh, please share the video and Channel with your friends and toer. Thank you so much for watching. See you in the next one.

Query Your CSV using LIDA: Automatic Generation of Visualizations with LLMs

Welcome to AI Anytime, your source for the latest and greatest in Generative AI and ML. From tutorials to expert interviews, I cover it all. Subscribe for cutting-edge content on LLMs, Deep Learning, NLP, etc. Join us on this journey of AI and ML.


