Github Copilot Does NOT Like This... (100% Local Free VSCode CoPilot)

Key Takeaways

write a python code with XG boost cross validation if you're not familiar with XG boost it's a machine learning algorithm and also a library that helps you build classification or regression algorithms and as you can see this is completely local no data are sent to the internet you don't need any subscription all you have to do is use olama isn't it amazing show you how to use deep seek coder within your Visual Studio code for your co-pilot it's an AI co-pilot just like GI up copilot only the difference is you don't have to pay any money first thing is you need to have ol installed then the next thing is you need to download the Deep seek coder model to download the Deep seek coder model all you have to do is AMA Run Deep seek Das coder and that will technically download the base the basic version of Deep seek coder which requires about like 800 MB storage and that requires the least Ram our download is successfully completed let's test the model write a python code for a simple addition so this is the simplest question that I can ask and as you can see that it has successfully given you the answer time for us to go ahead and install the visual studio extension which in this case is continue continue lets you use any llm by default it comes with standard set of llms but I'm also going to show you how to do the same thing with ol Lama so which means the Deep SE coder that we just downloaded with wama we're going to use within continue and then use it within our Visual Studio code which is a very popular ID that a lot of people use you can either click install from the continue which is part of the market place of Visual Studio code or you can open your Visual Studio code and click extensions and click search for continue and then start installing either way this would install continue in its vanilla format once the installation is successful we'll be able to access config.js where we can make the changes that we want so before that first thing that they suggest you to do is they suggest you to move move the continuous sidebar to the right hand side of yours so that it doesn't block your Explorer so depending upon your Visual Studio code like the layout you can make it or keep it wherever you want so I've dragged it to the right of mine so that you can see that the right it says ask a question slash for Slash and there are a bunch of things that you need to know control or command M will select the code control shift M will select code for followup and control shift L is for quick edit and control shift R is to to debug the terminal in itself now it's time for you to select the right model that you want they come with a set of models which we are not definitely going to use but once you go click the settings then you'll be able to access the config.js where you can go add your custom model for example I'm going to copy the existing code Lama code and then create a new model which I'm going to call it deep C coder the title doesn't matter a lot what matters is the provider in this case our provider is Ol Lama but I think you can do the same thing with LM Studio once you you have the provider AMA and then select the model which in this case is a deep seek coder model so you have the Deep seek coder model and the title provider model once you have all these things ready then you have to just select the model and then you are good to go ask any question that you want in the right hand side this is pretty much chat mode uh I'm not sure how much you will use it like I can ask a question write a python code for addition and then it creates an answer the answer is a little bit gibberish at this point when you see that is primarily because we are using the very base model of DC coder so if you want better result you need to use at least the 7 billion parameter model which in this case would take about like 8 GB of your space and you need to have appropriate Ram also you don't want to use DC coder you want to try some other model so in this case you can try AMA run stable LM Zer to try the particular model one of the ways you can explore whether the model is good for you or not is first of all you need to understand how the model has been built and second of all you need to make sure that the model is a chat following model not just a full uh next word prediction model so we're going to take a stable LM Zer so all you have to do is copy go to your terminal that you have got and then just run stable or AMA run stable LM Zer so at this point this will start downloading the model which in this case is a 1.6 gigs of model so once this model is downloaded then you should be able to directly use it within your Visual Studio code using continue which I'll quickly show you how to edit the config.js file once the model download is successful the one thing that you need to make sure that you have ol serve on either open the ol software or just run ol serve just to make sure that all the models that we're discussing about are available through the ol endpoint locally within your computer so that is one thing for you to keep in mind the second thing for you to keep in mind is if you don't like sharing Anonymous data with continue. Dev go to the extensions and click the settings that you have got and once you click the settings you'll see extension settings and inside extension settings you will see an option to send the Telemetry data if you don't feel comfortable in sending Telemetry data disable this that would not send any data from your computer to continue at least that's what they're promising so now coming back to the terminal or wherever that you downloaded the model so just you can send any question then ask for a quick check write a python code to add two numbers that's it that's what the question is and you can see that it has given you much cleaner answer than what deep deep seek coders base model give so now all you have to do is go back to your Visual Studio code open the config.js file and inside that you have to add a small block within the models so you would see models and within that you will have to add the small block title is table def zer provider is AMA and the model is stable zeper once again this model name comes from the AMA library in itself which you can just go and click here and then see from the library or we just downloaded it so we don't need it because we just copied it literally from there once you have the save the file and go back to any file that you have got and if you do not see the right side bar where you got continue all you have to do is control shift M and then you would get this thing once to get this particular screen then you can either like click new session with control command M and then you can select the model that you want so in this case you can select either stable Zer or you can select stable deep C coder I'm going to select stable zeper and I'm going to start writing a new python code so I'm going to use command shift M which is to ask any question or you can open a new file and then do command shift L which just has been indicated here so I'm going to just go go here and then say write a python code that would take an input name and then reverse it and print it so this is all I want all I want is to get the name and then print it back somehow this particular model did not work so I'm going to go back to the Deep SE coder model that we already have and then I'm going to do the same thing command shift L command shift L write a python code to get the name and then return the reverse of it so I'm not saying that it should get the string all I'm saying is this so it just wrote this python code and as you can see one of the problems that I face with a deep C coder is that it keeps on adding this um little com markdowns which I don't like it but anyways you can accept it or you can do all these shortcuts if you want I'm going to just simply accept it save it and then run it once I run it you can see that it takes the input and then it just gives you the output but it doesn't take the input from me so I'm going to just select all and then command shift let me close this so I don't need any of this please can you go away that's good so I can go here and then ask the same question or the other option is I can select all command shift l once again just to edit the highlighted code can you modify the code to take the user input so I'm going to ask this thing so that this code is modified to take the user input not necessarily give me whatever it is already there so it does it work let's try to accept everything and then see the code doesn't seem very clean and that is probably the problem with lower size models that you would find so let me delete the entire code come back once again here command shift L and then say write a python code to reverse the string and get the string as an input from the user send this and then see if it actually responds back okay it has created a code I'm going to accept it once again I'm going to remove this crappy markdown at the end and I have one more marck down here which I have to remove it and save the file and run the file so if I run the file as you can see it says raw input is not defined so that is a problem with python because probably python did not have anything called raw input and if you know a little bit of python you would know that that was probably there in the previous one and I'm going to just save it as input and run this once again it says please enter your text I'm going to say one little coder here and once you see it it has successfully reversed it 1 l i t t l e c o d e r so this way you can use it with any model so if youve got a different model all you have to do is control shift M and then you can just click the settings config file and then go here and then say for example I want mistal and I can say mistal and I can save this once again and come back here and then run this so for example I can go here and then say mistal and then control command shift L write a new simple python code what do you want a python code maybe I don't want a python code write a simple JavaScript application that can add two numbers and that's it send it and then technically it should write a JavaScript code even though I've uh saved the file as a p file maybe that's the dumbest thing to do so it says replace python with JavaScript and it does okay accept accept and uh once you accept everything you have the file ready and you can just make the changes that you want the point is that this may not be 100% replacement of your GitHub qu pilot because the quality of model depends a lot but if you have been wondering okay I have been part of a company I've got sensitive data which I don't want to send it to anybody in the cloud all I want is a local model that is successfully running and the then serving my own purpose and this is exactly the local model and you can call this model and you can see the locks also how much time it takes so that you can play with different models of different sizes of different varieties there are more to this thing than what I just showed you but the idea here is that to introduce AMA and also extension like continue that you can use to play with this kind of different local models I hope this tutorial was helpful to you in learning how to run a local co-pilot as an alternative to gith up co-pilot without any Cloud interface if you have any question let me know in the comment section otherwise see you in another video Happy prompting