7 Ways to Use the ChatGPT Code Interpreter Plugin

Key Takeaways

what's up guys welcome back to the channel in this video you're going to learn seven ways to use the chegebt code interpreter plugin so let's get started okay so the first application we're going to look at is the idea of EX doing exploratory data analysis from a CSV file this is one of the most common ways to use the charger pretty code interpreter plugin and it's one that I really really like so essentially what we're going to be doing is we're going to be loading a CSV file and we're going to be writing a prompt to ask for a exploratory data analysis on that data set so we're going to write something like perform we're going to write something like perform data analysis perform exploratory data analysis on this data set showcasing the main Trends and insights from the data and the data I'm going to use is one that I already have it stored here so we're going to be loading up this climate data set that I pulled from cargo actually let's do a cooler one we're going to be pulling a solar radiation data set that I have on a project that I've been working on so I'm just going to go into my folder and try to put interpret the plugin and there we go solar prediction.csv and now I'm going to upload this prompt plus the CSV file and we can look at the results and see what happens so essentially now it's generating a bunch of explanations on what each column and what information is contained in each column right and it's gonna the cool thing about this is it's going to explain what it's going to do and then it's going to be doing and we're going to be looking at the complete explanation of the data which for me is just mind-blowing how easy this thing is and how easy you can get something really really good so let's take a look at the final results so now it's explaining stuff like okay so these are the columns available you have you know time of the data point radiation this is a data set of for solar prediction so you're predicting solar radiation in a given region so I have a bunch of information like pressure humidity wind direction speed Etc and there you go so now we can get a distribution of radiation distribution of temperature uh distribution of pressure humidity and you can get a sense for what the data contains extremely easy you can get look at that a correlation heat map of the variables in the data set and notice that it's only showing a correlation for the variables that make sense to a correlation it's not doing a correlation for everything it could have done something done like doing correlation from Unix style and data but no it's doing correlation for radiation temperature pressure humidity wind direction and speed which makes complete sets so yeah this is one awesome application of the code code interpreter which is to use it as a first step in your data analysis pipeline where you're doing exploratory data analysis and just saying taking a look at the data here's showing a Time series of radiation over time temperature over time and pressure over time so you can start seeing some some patterns some repeater patterns some Trends Etc uh so this is a great way to use the charity code interpreter plugin another thing we can do is fit models to data that we load so from the solar prediction data that we have right we could have we could say something like fit a few times series models and showcase the um quality of the fit We compare the best performance the best the best performing notice and plot a comparison of the best performing models so let's see what happens so now I'm just saying the filter is modeled through this data that we uh to data that we uploaded and show us the results you know comparison right now the mod is gonna doing so it's going to be using some basic time series uh time series forecasting model so arima sarima and profit from Facebook which are widely used time series models and let's take a look so now it's explained it's going to use the stats model library to provide a comprehensive suit of statistical models like the arima and sarima models so now it's fitting evaluating the performance and now I got a mirror and this is one of the coolest things about this it got an error when it was running the code but it just learned from the error that I got so that we don't have to you know manually go through that process so that we can just look at the actual results at the end so this is another awesome result and we're gonna see the results in a bit okay so now it's doing root mean squared error so essentially it's plotty results getting these these results in these values and it's analyzing and comparing because that's the thing that I asked for uh we're gonna wait for a second while it finishes there you go so it plotted it in it fit the arima model it fit the setima model and now it's fitting The Profit model from Facebook and we should see the results black guys so this is cool so try to fit the Facebook profit model and it couldn't so now it's fitting a different model exponential smoothing and then uh it's just gonna try to fit that get the result and now it's gonna summarize in compared to the models and finally we're going to get a plot of the fits for the solar radiation prediction all right there we go so we have the actual data and then I mean the colors could have been a little bit better but we have the idea of the sediment the ETS predictions and so and it's comparing I mean this is really cool so it's comparing the results and discussing a little bit on which ones were a little bit better so as you can see it had these differs also the best performing one for the root mean Squad error was the edema model and of course this is not going to be the best fit ever that's something that you have to do yourself where you got fit different models and try different parameters but this is a great way to get started on a Time series data analysis project where you can just get something fit to your model and you can take a look and see which ones are doing better right off the bat another one another use case that's really interesting is getting summary reports from Jupiter notebooks now imagine you're uh you have a Jupiter notebook in cargo or whatever you found and it's doing some exploratory data analysis on a data set that you care about but uh let's imagine that you don't really understand everything that's in there and you want to get just a summary of what's going on in that Jupiter notebook uh so that's something you can do with uh what you can do with judge PD code interpreter you can say something like create a summary uh create a summary report from this Jupiter notebook and illustrating uh explain the main Concepts and giving the necessary contexts and for that I'm gonna fit in a Jupiter notebook that I have prepared here which is this brain tumor detection notebook that I got from kaggle so essentially I'm gonna upload a Jupiter notebook and I'm gonna ask for a summarized report from the Jupiter notebook explaining everything that's going on um sure um okay so let's bow so essentially now we can just ask for create a summary create a cell report from this Jupiter notebook explaining the main Concepts and give them the necessary context and I'm going to upload a Jupiter notebook that I have stored here for a brain tumor detection problem that I uh that I downloaded from cargo just as an example of something that you can do so now I'm uploading and now the model is working on the text from the Jupiter notebook it's going to read it through and it's going to Output back an explanation of everything that's going on that on that Jupiter notebook so it's explaining cell by cell which is it's fascinating so first it provides an overview of the notebook so the first note the first cell is like CNN did I bring tumors vg16 image net okay importing libraries okay data pre-sauce data preprocessing explaining the data process involves resizing images converting to grayscale normalizing with shaping model building compilation training evaluation okay okay uh I'm not really satisfied with this explanation so I'm going to ask her can you give me some clarity on the concepts behind the analysis that was done sure let's break it down so CNN's and then it will explain what cnns are they can learn features Etc topics no for Simplicity can leverage dependence models learn from you so now it's kind of like giving me the background on the Jupiter notebook so that I can get a sense for all right so why the choices made on the Jupiter notebook why were they made and what was going on uh uh and that's pretty much the application of getting summary report from the Jupiter notebook directly in the judge PT code interpreter which I think is a really cool application because sometimes for like big Jupiter notebooks you can just get a really clear summary of what is going on inside it instead of having to like go through the entire source code inside that Jupiter notebook uh the next application we're going to talk about is applying image processing to an image so I'm going to upload an image and then I'm going to ask judge PT to plot a few to do a few Transformations on that image and in this case I'm going to say Plata I'm going to give it an image so I'm going to upload an image in this case I'm going to upload this pop art image that I have and I'm going to say plot a histogram of the pixel values on this image across the three different channels and now I'm gonna get as a result the histogram of the pixel values on this image across the three different channels so we can take a look and there we go so now I have a histogram with I mean there's better ways to plot it maybe but you get the idea the idea is essentially you you see how pixel values are distributed and now um I want to say something like turn this image into a sketch so now this is another image processing idea that you can apply you can apply all sorts of image processing techniques essentially anything that you know has opencv in it because that's what it's using here you can do in an image from just asking to do it so sketch of an image and there you have it sketch of the image so this is a really cool application of the treasury code interpreter which is to apply image processing techniques from natural language directly to the image you just upload the image and you're done now one of my favorite applications of the judge petite code interpreter is to run OCR on text inside of an image so for example one thing that I can do is I can upload an image of text so I'm going to find my image of text here so I'm going to go to my folder content map shares particle interpreter and now I want to say where's my image and there you go now I have an image with text here and I'm going to say run OCR on this text yeah extract the text inside this image now I'm going to run this and this is what's going to happen so we wait a little second and we'll wait a little bit and now there you go now that's awesome so extracted the text from the image we and this is the output which is perfect yeah it's exactly what the image said and there you go you can extract text from the image of from an image where there's Stacks inside and you just get the final output which is amazing and now this is kind of cool because this is kind of like the um Segway to the next application which is you can extract color from a picture and then make that image into a palette that you can download so the next application can do is as new shark attacks that you guys just saw was suggesting you can create a strip that extracts the top five colors from a picture and then make it into a palette to download and I'm going to add the let's think step by step to make sure it the model knows to be very thorough and do things step by step so I'm going to upload some random image I'm going to upload the same pop art image and I'm gonna run it so now it's briefing it down into load the image extract the colors find the top five dominant colors create a palette image using these colors and then save it as a PNG file and there we go so this is the image I uploaded now we're seeing it you get the point you can get you can extract colors from an image and then save that to another image and then you know have that palette and that's what you can do it okay so the next application that I think it's probably one of my favorites is you can get um image of text and you can extract the code and run the code from that image of text so to Showcase that I have an image here that I'm going to upload and I'm going to add the following prompt I'm gonna say use OCR to extract the code from this image correct any mistakes that may have appear after dualshar and run the code just to make sure that judge PT doesn't get confused I'm going to create a new chat I'm going to access my gpt4 code interpreter and then I'm going to say use those yard to extract the code from this image correct the mistakes that may have appeared and run the code and now I'm going to upload the file I have the code snippet file here um used so first I'm going to say display the image of the code and then use those CR to extract so that you guys can take a look at what we're doing so first we're going to display the image so this is the image that we have there we go so it's a very simple it's an image of a very simple python script where you know we're plotting the synesoid the sin of some data that we created and now we're going to write some OCR code to extract the text from this image make sure that judge petite takes a look at the extracted code so that it fixes any mistakes that UCR may have made and then it will run the code sure and now perfect that's exactly the answer for this code that is exactly the answer so as you can see you can take a picture of code and run that code right inside the charger PT code interpreter which for me is just the coolest way to one of the coolest applications of the charity code interpreter we talked about we talked about Eda we talked about how you can do data visualization and oci and images apply image processing you can do OCR from code and then run the code now just what I think is interesting is how we can start thinking about combining those things into workflows that really improve how we think right so we saw a few examples of how we can use the judge PT code interpreter to do things like doing those sharing images applying any type of image processing technique doing data visualization and exploratory data analysis and in the future I want to explore interesting workflows that where we can integrate these different capabilities of the church the code interpreter plugin to build really fast and effective research workflows you know workflows where you're thinking about bigger picture stuff instead of concerning yourself with like smaller issues let's say all right so thank you guys for watching don't forget to like And subscribe and see you next time cheers