NVIDIA'S NEW OFFLINE GPT! Chat with RTX | Crash Course Guide

what's up I'm troubleshoot in this quick guide I'll be showing you a brand new piece of software from Nvidia called chat with RTX you'll find a link to this in the description down below I'll show you how to install it and of course how to use it what exactly is it well youve probably used chat models before like chat gbt Gemini Etc but this runs on your PC if you've been following my channel for a while youve probably seen generating images on your PC and even text using other free downloadable things this is a demo from Nvidia themselves using models that are pretty highly respected today to give you a chat interface that you can interact with and of course this thing can interact with your documents too so you can question it on a bunch of documents in a folder it's simple to install and simple to get going but there are a few requirements on this page you can click system requirements where scrolling down you'll see you'll need Windows 16 gigs or more of ram Windows 11 driver 535 or later and you'll need an RTX 30 or 40 series graphics card or RTX Amper or a generation with at least 8 GB of vram but most most people with newer graphics cards for gaming and stuff like that you'll probably be able to fill this requirement if you've bought one of these graphics cards I luckily have so I'll be able to use this do keep in mind if you want to use a AI chat bot that you can chat with hosted on your PC you can do so even without a graphics card using just your CPU and your RAM also completely for free not to mention on top of this you can use all the graphics cards too just as long as you have enough RAM anyways that's for a different video or of course one that you'll find linked down below this one is focusing on with RTX in order to download it and install it head across to the page in the description down below and choose download now it'll be a giant zip file that you download at around 35 GB when this massive zip file is done downloading extract the folder inside of it and of course delete the actual zip itself to save yourself a bunch of space when it's done opening the folder we have setup in here simply open it and you can install it much like you do any other Nvidia piece of software you need to install chat with RTX and optionally you can install mile 7B in form this is one of the models that we can use I'd recommend you keep this one chosen as it's a really good model we'll choose next choose a place to install it now you can choose where it goes for me I'm fine putting it here next and we'll wait for it to install after it's finished installing which can take quite a while this whole process up until now has taken about 40 minutes for me I'll be able to launch it and click close sweet now the program will fire up of course you can delete your downloaded folder here to save some extra space and wait for it to start up this will start downloading the final little bits of it and in another around 3 gigs of downloading it'll prepare everything and eventually start up that is of course unless you get some kind of catastrophic error for me I got some weird Cuda error the first time running it through and restarting the app didn't help I had to reboot my entire PC but when I did I was able to launch up the program and now you can see that it's working as it should be so if we go ahead and control click this here in terminal otherwise you can select it by just dragging around it right clicking to copy we can then open it in a browser once it opens you'll see something like this we can choose an AI model currently there's just mistal in this I've seen llama 2 somewhere not sure why that's not included here but anyways let's quickly run through what this thing is capable of after we choose a model and choose a data set folder we can ask it some questions about this data set if I open this up you'll see a bunch of text documents in here all with information about different things for NVIDIA we can go ahead and ask it anything such as some of the prompts here and it'll very shortly after give us a response using just our graphics card on our system now I'm running a 3080ti and it seems to be relatively quick so it'll type out the response and even give you a quote to where it found the information you can click on it to open it up Etc sweet we can undo regenerate response clear chats Etc it's everything you'd expect what is NVIDIA cooking up apparently dlss 3.5 blah blah blah this is where it came from sweet you can punch in any URL over here just by clicking change folder path and you can choose anywhere on your system this is of course explanatory choose a folder with a bunch of text files PDFs or even dot doc files and you can query your system to learn more about what's in that folder the best part is is that the files are linked for you to click them and find out exactly where it found that information sweet we can also try YouTube URL and this is quite a fancy one we can head across to a YouTube video in my case I'll pick one of my latest uploads and we can copy the YouTube link it needs to be this long fashion not right click copy this shorten youtube.be needs to be the full link as such but this we can punch it in here click the transcript button and it'll process everything learn about the video then shortly after we can query it so what is the video about it tells me that it's about an antiche for hdis 2 Etc and links back to the video as reference if we say how was it done should hopefully give us some information about it yeah it seems like it's telling us about the antiche which is pretty cool obviously whatever you put in is what you get out it's a chat model much like any others of course you can put in anything here about the video say summarize the video into 10 points and it should hopefully give us 10 points about it it misspelled and you cheat there but I'm pretty sure that's probably how it's spelled in the transcript that I got from YouTube obviously YouTube transcripts aren't perfect and this is probably one example of it but anyways you get the point you can put on a YouTube link and learn about it obviously there's tools like this such as chat gbt you can use for free Etc but this runs on your own system that's great and finally AI model default is just the AI model so what is AI model default and it'll tell us basically it's just the model itself so we can interact with Mell as is so is this great is it bad well to be honest it's okay you can already download in Chad with tons of AI models using other existing pieces of software such as uoga desktop which I've covered previously it's easier to download extra models into that more customizable uncensored models even but the way that things are packaged here as a demo it's a great place to get started if you have no idea what you're doing with python and Etc it does it all for you sets it all up download one big file and poof there you go you can chat with Invidia RTX I suppose having the ability to interact with documents in different folders understand YouTube transcripts and things like that it's cool but of course it is done Elsewhere for example YouTube transcripts can be read through and processed on chat GPT for free Google Gemini probably and tons of other websites that do exactly this for free but of course this is using your own hardware and of course completely private for that reason while you could technically copy a transcript into UB Boogie desktop and work with it there I suppose it's just a little bit nicer to punch in a URL here for example it complained about using short links in console somewhere here though I can't find it now something about couldn't download let's see up here yep couldn't do it reject search didn't understand what this short link was a bit disappointing is right click copying links is kind of a basic thing from YouTube so all in all it's a cool demo but that's pretty much all that it is you can get much more usability out of different models using Uki desktop for example it's a cool Tech demo it's nice that it's made by in video so hopefully it'll be supported with a huge amount of money behind it but that's all to be seen to be honest this is just a demo in its current state and it's pretty good for what it is I suppose if you want something more functional more functional things do exist and you'll find them linked down below for example we can close out of it by clicking the shutdown button it tells us we could restart to chat with it and that's pretty much that anyways do chat down below I'd love to hear what you think about Nvidia chat with RTX is it just branded garbage or is there some more use to it that maybe not seeing I think it's pretty cool to have something like this put as simply as possible for people who've never messed around with this kind of thing that's a great way to get started if you'd like to learn more about using AI on your own system if you've never done it before but anyways thank you all for watching my name is being troubleshoot and I'll see you all next time CIA

Channel: TroubleChute

Views: 67,810

Keywords: chat with rtx, chat with rtx review, chat with rtx install, nvidia chat with rtx, nvidia, ai, machine learning, how to, easy, chatbot, llm, artificial intelligence, ai news, gaming, artificial technology, geforce, pc gaming, geforce rtx, nvidia geforce, nvidia rtx, chatrtx, chatwithrtx, amd, amd vs nvidia, ai news channel, openai, chatgpt, mistral, wizard, llama2, int4

Id: V-gXPb2lNrE

Length: 7min 53sec (473 seconds)

Published: Thu Feb 15 2024