Learn Docker with me

Key Takeaways

all right ladies and gentlemen welcome we are uh going to be looking into Docker today which is something that I've always postponed in learning let me see if this this chat thing I have no idea how I'm supposed to bring up every time like I always open up um uh let's see here there should be like a popup here of some kind but let me do this hello sorry you said why what even is Docker all right let's um let's uh break this down okay because uh it is important that we uh should even reason about why first right so um yeah okay so welcome to the stream it's going to be uh pretty um just beginner I mean I am a beginner when it comes to Docker um I'm ashamed to admit it but yes I don't know anything about this uh but it's a very useful tool that I've um realized that I need to learn about So like um so okay for now right when I do any kind of machine learning development I always use uh cond environment and that's how I it works right but I think a problem is when you start to share things uh and uh sort of if you're working or or you're trying to productionize something then you need to containerize it that's sort of the the way that the cloud is set up and you know how we deploy things so I'm curious first of all like do learning Docker is something that we need like you need to know because you're going to have to do it at some point when you're production l right unless you're sort of only doing things on your computer if you're never going to put it into deployment then I guess you don't need to learn it but that's the goal right for us to be able to deploy things and then um the the the second question is do we need to act so is it beneficial to use Docker when you're actually developing as well um that is something that I'm a question mark around uh but U what I'm thinking is that for example example let's say you spin up a VM uh because uh you need uh some more vram or something then it could be very beneficial to just sort of have a a uh you know you could get your Docker image and then you can just run um it directly you don't have to set up miniconda and all of these things so Docker would be an alternative uh instead of using Docker um unfortunately L guys I don't have a so this is something that we I think should learn and I need to learn um I don't know the best way to learn it you should have seening me try to learn it uh live now I thought what we could do is obviously like first we need to install this thing wait oh so first we need to install the thing right and then secondly uh there are they have some um getting started docks and so on so I thought let's take a look at that let's see what they sort of say but okay um so let's say what they like let's see what they actually say Okay accelerate how you build share and run applications Docker helps developer build share and run applications anywhere without TDS environment configuration or management so that's my understanding is it's sort of uh uh it it's its own operating system uh so when you actually get the docker image and you run it you're not running it on your computer you're running it on a completely sort of separate computer right that that is uh set up in this Docker image and all the packages everything you need is is installed in that so build um uh let's see spin up new environments quickly okay integrate with existing tools containerize applications for consistency yeah so you are like if you containerize you can run it in any environment right uh if you share uh let's see so yeah this is a good thing as well like um that's I mean in practice it might I don't know I haven't used it that much right but this is the promise is that how many times do we sort of run into the it runs on my machine kind of a problem where uh it works but then somebody else's has some other type of weird setup and then they're not able to run it um I mean this happens all the time if you are actually uh trying to run someone's open source stuff right uh then this is a common problem so if you can have a Docker image for it then uh it sort of in theory should solve these kinds of problems and it's also I guess more secure is U another thing and then run uh ship your application knowing they'll run the exact same way in any environment locally or in the cloud uh develop with versatility uh deploy with one command so it's kind of a to me it seems like um a way of ensuring that thing is able to get deployed you're able to share it in a in a in a in sort of a way that has the minimal amount of problems associated with it and so that's kind of the the sorry for that's kind of the I guess the reason why we should learn it right so the question to me is how the hell does this work um in terms of not how is it uh like how does it work in terms of how did they build it but actually like how do we use it that's the only thing that I'm kind of uh curious about here so okay they have some walkthroughs here get introduced in these five minute Hands-On guides so okay container let's read about that it seems important all right so a Contin is an isolated environment for your code um this means that a container has no knowledge of your operating system or your files and runs on the environment provided you to you by Docker desktop containers has everything you need down to a base operating system you can use Docker desktop to manage and explore your containers um all right so we can actually we can follow this but uh before we do that right I need to see just um damnn my FPS is so low let's see if this helps no all right anyways let me know if it's slow or anything um karpati starred a build Docker from scratch repo so I got curious could you send that maybe we could take a look at that one as well okay so uh setup walkth through the first thing you need is a running container yeah so we need to install this thing that's what I'm trying to see how we do uh is there an install [Music] somewhere oh like it was actually a um build Docker like how it works from scratch if you find it then please send it and we can at least check take a look at it and uh thanks pratique for uh for becoming a member so yeah we should install this thing is this is a little bit confusing you have you have sort of Docker Let's uh go here and see okay so um what else a Docker engine Docker desktop Docker Scout um parque um there so at the moment there's not very much difference we use Docker for work and essentially in both development and deployment okay that's interesting yeah so I mean I I I do think there's value in using it for for development as well because then you're sort of you don't have to do an additional step in the end right you're at every intermediate step you are um guaranteeing that it's sort of uh you can share it uh at any point in time but uh I wonder if you're using sort of GPU and stuff if that's going to be a hassle of uh and and stuff like that so that's one question question I have uh unless you want to set everything up every time you run you're going to create custom documents to set up tooling for you yeah that makes sense okay what is it that you're typing on at the moment um so uh right now I'm only using the the Macbook so actually like I have I've bought a keyboard like this one and it's like nice like blue caps uh it's a vermo it's like a wireless keyboard and it's amazing but uh I don't know like for some reason I'm I use the MacBook keyboard um mostly because of the trackpad like if I use a mouse and a keyboard I can't do all these gestures right I like these kinds of it's a kind of a smooth interface anyways okay Docker engine this is the uh okay this is this is the thing we need to install what's dark your desktop and this is an easy okay I mean we like it easy so let's install this then oh G oh okay sorry I misunderstood yeah this is GPT for um I just use the API instead it's sort of a I like it this way because I can customize it a little bit more I can set up like temperature I can set up the system prompt and things like that so like what is the actual goal for us okay like install and run docker like create a Docker container um run script with Docker see if it works with GPU um maybe look at kubernetes up B uh get an understanding like these are I don't know this is like the goal I guess wait is this a custom GPT um so uh this is so the the reason I don't like I I mean I like the chat gbt plus but I don't like that it's limiting me uh in terms of I'm not able to set the temperature I'm not able to set the the custom I guess they have special instructions but I don't know if it's the same thing here I can set the system prompt and I can also run this is the main thing I can run more than 30 messages per three hours uh I can run as many as I want uh that's like the main thing I don't know if it's a huge difference though to be honest yeah I know there's a um uh no um so yeah I guess um so the the question is if you create an assistance uh or you create a a custom GPT can you also get a API from it because if you can't do that then it's uh it's good so this is very easy guys uh here you you go to this one I can make a video on this if you want if it's helpful but you go to this one this is the best interface I've found so far uh or the yeah the yeah I've used another interface as well but this is the best one it's this one so so I'm streaming now in 1080 p.m my computer is still lagging I it's weird oh guys I can't even I can read the chat but I can't send stuff in the chat oh that's annoying anyways it's this one guys just search big AGI and you can deploy it on versel that's how I did it just one thing that's incredibly stupid is uh and somebody messaged me about this is that you can actually go to this website right I mean you can go to this if you go to this and you just use your own API it's going to be set locally and you can use it maybe that's the easiest honestly just try do this okay uh thanks for reminding us uh so the do the the the goal here is to learn docker you need a subscription I thought this was free it should be free right all right let me log into this thing hello guys nice to see that so many people are in the chat so I also set the the the latency to ultra low so it should be the best sort of uh I mean continue without signing on then what's your role I am data scientist upgrade your plan all right uh so we watched this one right uh let's see now okay open select search specify Docker SL welcome to Docker Docker pool so okay what do I do guys I do this I guess I in the terminal or and okay oh maybe I shouldn't be on the VPN here it seems to work so I guess it's okay dropped frames 4.6% so should I maybe disconnect this is it fine guys by the way or should I like is this has the stream been working fine till now yes I'm learning Docker for mlops okay so okay cool so we pulled that one what happened exactly um Docker run Docker run okay so where did this actually come like Docker welcome to do it doesn't it didn't go to my Docker pool all right Docker run first thing you need is to run a container so I guess we like we pulled the container from the I don't know um from the repository of different Docker containers so what happened nothing expand the optional run so what's this next step after this like we run I mean we can run it here I guess optional settings we can specify these right but um all right you just ran a container okay cool you can view it in the containers tab of Docker desktop this Docker this container runs a simple web server that displays a simple website when working with more complex projects you'll run different parts in different containers for for example different container for the front end backend and database Okay cool so um where should we go then uh you can view it in the containers oh okay so we're running this okay cool I guess we're already running it right because we're running it here then we should go to where this one no I mean I didn't specify the the port or anything so maybe you would need to do that right probably um let's if we go back here we need to run this is already in use but the way do you guys use a CLI so wa uh can you solve container name let's see if it's good welcome to Docker already in use oh I don't have it oh that's fail okay I thought I had set it up anyways um okay error response container name welcome is already in used all right let's just run it here then I would like to run it from okay so then my question is where is it running because it's oh there it is Maybe uh containers tab so the containers here there should be one here right so let's delete them cool and then let's go to back to this one and now let's go and run this okay and then let's go to Local Host 8 88 nice so so the cool thing I guess is that we didn't need need to do any setup right everything is in that image so like it does a good job of displaying what what's actually like useful then uh how do we check so one thing that I'm also wondering is uh let's see um CU I thought I had set up this yes so source zich see this is what I tried to do before uh okay okay I will try to do something risky okay are we back hopefully we're back I believe so so I was wondering just if that's the reason no sir okay so uh check and find name find all Docker containers yeah it was probably because of that so there we go okay so then we can do uh close uh close all Docker containers running Docker kill is used to kill a container okay so if we do um PSA we can see what's running and then we can do Docker kill uh kill Docker container with IDC 256 D e587 14f okay okay and then Docker PSA it's still running exited okay exited so it's not running because before it was up 2 minutes now it's exited cool okay and then explore your container stop your container oh runs until it you stop it all right so when I want to to after let's say I ran a Docker container with the following following command oh I would like to exit that uh so that it is not continuing to run in the background how do I do that Docker stop okay so you actually have a name for it as well you can stop it and then you can remove U what does the kill command do okay so it's like if you want to be graceful guys you do uh please stop I guess if you want to be extra you do Docker please stop and then you dockor stop and if it doesn't work then you just kill it that's how it works cool so okay we saw how to run a container uh um cool next is how do I run a container didn't we just learn this okay so how do I run what what was the difference here we saw what is a container here we see okay I guess like how to create one perhaps because we can do get clone this thing okay so Docker file here is how we specify I guess like how we what is in the container so I guess in the previous one they set up every kind of package we need in someone hello AA and uh okay that's cool you can kill first few characters of ID that's nice thanks and um yeah makes sense that was a long ID as well I guess you can if you use names on the Dockers it's a little bit easier perhaps so okay what happens here we run from this image we start from a node node base image then we have work directory we copy copy okay okay so we create Docker image there I would like to have a python instead like [Music] alternative so can we do this this um here for example could we do Docker file oh Docker file and then all right guys so we created we have pytorch here what is this uh how do I specify the versions and so on I mean this I don't want okay latest compress size 3 gigabytes here we have 2.1 Cuda CN cudan n version and so on and then we can specify I guess here is just the latest is this CPU version version or is it GPU version okay uh let's see here so I mean we can figure that out at a later point I guess let's just do this first and see if we can just create it um un let maybe just create okay so you actually like you specified the name Docker build t uh and then you specify like okay here is where the docker file is located so you can do I guess Docker build oh okay all right so now it's installing from that pytorch the latest one and it's 4 gigabyte 3.4 gigabyt um all right and then we can so okay let's see let's go back here so then um so what are we doing uh Docker run remove my my py python screw run I automatically remove the container okay so let's see it's still building it I'm I'm really hoping right after we build it like it's only a one time type of a thing it should not be taking this long yeah I think guys I don't know if you're not using I don't know like gb4 seems to be getting a little bit nerfed lately but uh to I mean it's it's still worth it I would say oh so you can use it inside vs code as well I guess uh makes it easy to build man deploy okay we we'll see what this is doing but uh yeah this is taking a while guys holy crap like creating it's a ver this is like the most simplistic image as well uh right we're or I guess not right because it's pie torch maybe it's maybe it's a lot if it's Cuda and stuff I'm not sure [Music] and what is this uh okay cool so how do I use this thing as well and expose Port from uh install Docker blah blah blah [Music] run okay that's I guess is cool so I mean we can I guess it's supposed that we can specify right instead of this thing where I specify my like cond environment I could specify like I am able to specify to run it through the docker instead that's the thing I guess y5 from scratch you trying to kill me it's going to take me a long time and why YOLO V5 if we're going to do it let's do YOLO V8 or I mean uh it's YOLO V8 that's the latest one right I mean these YOLO thing have gone uh crazy though I'm much more do so do any one of you know um how uh so last week right I did a stream on the Cog vlm so Vision language models and it's able to do this call thing called grounding where you can actually ask for bounding box and so on uh are you able to has anyone seen a comparison of let's say YOLO V V7 V8 or YOLO Nas or whatever or any state-of-the-art object detection model compared to the grounding vlms so let's say like gp4 if it's I mean it's not Ed trained for that specifically but let's say Cog VM or any of those it seems to me that the visual language models that's we're hitting on something that's much more generalized and seems very promising in general right I mean the you can now use one model to do you don't need to do any like very specific OCR kind of model and so on it's a bit Overkill I agree but you can use gp4 just out of the box instantly you don't have to think about anything uh and so imagine you would at some point right you would have a similar to gp4 Vision in your local desktop so that's something like Cog VM for example all right so cool so now we installed this so now um check what Docker containers we have Docker PS that's how I did it Docker PS what what is why PS PS Docker PS is it just me like why would you name it PS can someone like make understand that process status okay yeah I guess I'm I'm bad at those things so yeah that makes make sense process status okay um so then if we do Docker PS we get nothing what so what did I just do I did something all right so it's actually when the container is running I guess we just created the container let's say if we were to I mean we can do just a test test script so let's do import torch do like I don't know run a for Loop and do some Matrix um print that it was SU successful all right and then Docker run my py torch image that's what we created right and then test.py so this is an interesting thing right if we can run this now uh it is here I think by the way the latency now is much I like can somebody write something and then we can like tell what latency is so right right now uh wait sorry let's actually do redo that let me see the timing where's the time like in seconds whatever I wanted to do like a test of how the latency is yeah okay so that is like 15 seconds that's pretty good so why does this not work the are you're seeing indicates you're trying to unable to find this could be so I am in the directory of the test of the test.py um if you're in m to run you need to mount director as a volume what the hell all right so what is this uh so this is so okay we've down we have a Docker image now with all the packages right the idea is that we should never be able we should never need to use a cond environment those are sort of the two paths or you use a cond or you use a Docker um and I guess like I don't know there might be too simplistic but that's in my my mind how I'm thinking so if we're using a Docker then now what I want to do is after I've created it I want it to be there and I want to be able to run all the scripts through it so how do we do this in an easy way and I guess I have created a Docker image right and named it my pytorch image um can now I am in a directory with a script and test. P that that I want to run how do I run it using using uh Docker you'll need to mount the current directory into a Docker container and then execute okay open navigate use the following run I mean this is a little bit complicated but I guess let's say let's see if it's like a onetime type of a thing so version okay what are we seeing um I had problem with mounting also uh the easiest is to mount your folder in your local in the [Music] container so I'm just trying to understand like here we are also removing the container after we're set setting to be we're mounting this mounts the current directory to access this seems very complicated uh I don't know maybe it's not but it seems like is there not a way to do this easier I I would like to do something like Docker run test. p and then name my pytorch image you know like I should run this script using this thing what is like what the hell do I need to do all this for um all right anyways let's try it then so here we're doing RMV workspace python all right whatever let's see so one thing right it needs to be fast because if it's not fast then you're going to lose development speed so it needs to be just as fast as if we were to run it using just aronda uh that's in my like thought process here otherwise I'm not going to be using it right um uh okay okay let's also go back here so view your Docker build your first image uh we've built this right uh run your container this is I guess like what we're doing view the front end Docker Hub images this is also what we're trying to [Music] do what's going on does anyone know like why this is insanely slow do uh doer PS so it's been up three minute and then I mean we're using this Command right which I don't know if it's correct or not but uh definitely seems why is this not working the hell oh man that was slow okay uh what the hell cond activate rexes okay python test. P I think I'm might have like ruined my um cond environments for some reason it's a bit unclear but uh yeah this was very slow I mean is it same if I run it again because that was I mean insanely slow uh can do 10 we can 50 okay so maybe it was the actual script um okay so that makes sense maybe um I don't know we need to compare it right but uh for some reason this is not yeah for some reason I don't have it here which is strange what the hell cond environment let's see cond environment list yeah no torch here either okay so that's something strange anyways it's running it it's just a little bit strange why it's doing it this way or like why you need to do it this way so okay here we are removing it afterward something we're specifying let's see what it said um okay so we are Docker run remove the container when it exits Mount the current directory this sets the working directory work space so we're mounting the current directory into the workspace directory inside the container and then we're running it so yeah maybe um guess that makes sense hello to you too Peter and Merry Christmas all right uh yeah so what's missing guys we need to learn how to let's see what was my actual like notes for this all right we want to install and run Docker want to create a Docker container we want to run script and want to see if it works with GPU now we can't do this actually because I'm using Mac right but maybe this is something we can do um okay and then we can create a repository on Docker hub I'm trying to see like what there is to uh to learn as well like what is it that we need to uh to know there's so many different things here we did this right um all right so containerize a python app I mean after Docker file start with Docker compose um I don't know have we learned how Docker file works in the best way is this like the are you supposed to run things with I mean we can the docker file we can learn right there's I guess more to learn with the docker file but is then this like are you supposed to run it using these things uh or is there another way uh to do it multicontainer that's I guess what you meant uh with the do compos so I feel though like we should spend a little bit more time actually learning about first the docker file um let's see if there's a good tutorial as well so um I'm more Curious like what the the terminology and all these things so here what is he talking about Docker install he's using it in py charm I guess okay he's creating a script okay let's go to Patrick's video actually this guy is also good I think okay so he creates the docker file from python pip install and then he's running he's asking it afterwards he's asking you to run python main.py so let's see what does so from um start with the base image then run is to execute a command command provides defaults for executing ER provides defaults for executing okay entry point allows the configure that would run as executable okay copy add CMD defines default commands and or parameters for a container these commands can be overwritten by the on the last one guess my brain is a little bit slow I don't think I really fully understand the difference here anyways of the entry point and CMD here is actually specifying a work directory so this simplifies things I guess because now he can just okay um set the working directory um okay so what's the benefit of setting a work copy the script to the folder okay so okay so okay is this why is this like a circum how to circumvent the this problem right of not having this thing because if we can first let's do Docker Pia oh it's not running okay then let's go to the docker file and then let's do a Sim did I just close what he did oh no here it is so then uh we'll set the work directory we don't have any uh requirements here but we want to copy and then start the server so what does this do um for Okay so I don't know if code is better setting up nonroot user um pyour pytor latest work directory install packages expose a certain port create a nonroot user specify the command to run okay cool uh and what was the idea of running let's see so after we have [Music] this create a Docker image from Docker file in current directory Docker build my image okay so Docker build my PCH image oh that was faster so did something change here uh yeah okay so then what happened exactly Docker run test UPI okay so I did this but now when I do Docker test Docker run oh oh sorry Docker run my py torch image test.py okay so this makes a lot of sense now if we specify so we build the docker okay so the only question now is uh right so I my idea is I create a doer image um if I share it then let's see so to the docker container um so I guess the confusion here is are we sharing the script as well or are we just sharing because I'm thinking like this is a I want to use it as a substitute for cond essentially but I am confused if we are also specifying the actual files and stuff then I guess the difference right is you can specify the work directory here if you just have the actual image then we have all the packages and stuff and this is just for um for the code part right so this is just you can do this anytime right if you've installed this then creating updating this is very simple I guess maybe that's the way to think about it okay so let's see what else is there that we need to figure out all right I mean I think let's see guys is there what else is there to figure out here there's a lot right but um maybe just do could you explain what kubernetes is doer engine I run time that allows you to build and run containers Docker images executable packages uh Docker containers instances of Docker images that runs the application's docker Hub registry K8 on the other hand is a container or orchestration system that manages large number of containers deployed across multiple hosts it's designed to help you manage the life cycle of container applications automatic scheduling selfhealing horizontal scaling service Discovery and load balancing automated automated roll outs and roll backs okay uh what else guys what else uh [Music] so to run the same Docker file which could theoretically which could in the future use other packages package versions um and break the and yeah this is what I'm thinking right so if we build it once uh and if it works right with us installing through pip and so on we want to then push it somewhere and ensure that it works right that's kind of the idea so now for example right Docker PS we don't have any okay uh how do I check which Docker images I have Docker images okay so these are my Docker images Let's uh when I run Docker images I get how do I remove the none and welcome to Docker as it is taking up unnecessary space and I only have 200 gig 256 gigabyte on this crappy laptop okay and then if we want to kill or remove remove a specific one let's remove that one dark images Docker RMI Docker welcome to Docker is using its referenced image I didn't know it was actually running I didn't think it was I [Music] remove let's try this Docker images Okay cool so then we have that image right uh Docker uh images and then okay if I have that one how do I share this uh new Docker image so that um that is using very specific very specific cond environments uh um package versions Etc so that people in the [Music] F to push it to a Docker registry the most common but you can also use getop container registry or a private registry Docker T tag uh let's see Docker tag my py torch let's create one then if this works I guess I also need to create this thing let's see doer registry oh that's not actually my [Music] profile uhhuh okay so then we created this and what happened there we go oh what is this four years ago I did something four years ago okay um okay so this one now we should be able [Music] to we should be able to do this to pull this one so can we now use this so [Music] from oh your tag as well okay so then it should be uh testing v0 cool so Docker images now we have that one I guess Doo P torch cool how is it Ducker remove okay so okay like what okay what did we let's see what this guy says as well but I mean so here we're specifying we can specify like a uh the image we want to use right and uh if let's say that we were running very specific uh requirement file and we don't want to run it in the future then we can just save that image people can use that image it will always work kind of so that's a really good thing right and then we use pip and stop we can copy uh script from our current folder uh to a work directory and we can specify that uh work directory and then that means that we don't have to um basically we're giving our current files to the docker um to the docker environment what happens if you modify it locally after you've run it so so that's a question I have uh in the docker file we are specifying a work directory and then copying files from our current directory to the docker uh working directory this allows us uh ability to Simply run Docker run Etc uh the dock run and then uh you image python script for example um if I then modify the test. py or the script up py do I need to re rebuild or rebuild how does this work I guess my question is not that clear yeah so the container will still used that was included so essentially so okay uh what are we doing here so [Music] we're I am a little bit confused now so I mean we're creating the from this work directory copy Etc right uh then however when you want want to override a file that was added you can use the V when running a container this will mount a specified host file effectively bypassing the version okay I mean that makes more sense I would say so Docker run my image name uh so what happens if we do Docker run and then we specify v um directory okay so we can specify V PVD um like this so what are we trying to do we're trying to kind of losing what I'm trying to do even I'm doing um I don't want to constantly be rebuilding the docker right I want to mount so then it's always oh it's just a warning okay so then no [Music] morning so there's something that I'm missing here I think um oh [Music] okay what was it we used before so I'm trying TR to think uh we did Docker build Let's see we can go back a little bit maybe guess this is what we did before right [Music] so okay so that works but then if I change this it's not going to run the new one right so then we would have to Docker build and then rebuild it right I think that's a a problem um I don't want to do that so uh so then I guess like we're kind of back [Music] to okay there is something that I'm missing here I don't think this should be too difficult um it is interactively this is removing it afterwards and this is the thing that does it um Okay cool so okay so then if we just remove the interactive part [Music] then oh okay cool so okay okay so all right so I mean the question I have then is what's the benefit of using what's the benefit of actually using the work directory so is it fair to say that let's say we have something that's sort of complete right we um we just want you to be able to okay I guess it makes sense if you're not using live development right then you obviously would want to mount the volume and so on but let's say that you would want to actually have something that is production ready in terms of we just have the container you just run it you don't have to clone any code and so because the code could be modified right so we we uh have all the code needed all the packages everything in this Docker container and then you just sort of that everything works inside it right there's no external dependencies of any kind so so I'm curious for example with the the the one we did last week which was um Cog vlm right uh in this case it was a little bit complicated of setting it up but I'm curious if we just use Docker instead how easy would it have been uh so this is a question mark as well like why aren't you using Docker here then uh this does anyone have have an answer like why wouldn't they be using Docker so they have all of these pip install requirements and so on why wouldn't you just have a Docker image here with all the code and so on and you just get the weights from in downloading them uh John do set work directory set Mount directory from local to work directory yeah that's what we did here I guess entry point runs on your code yeah we don't have an entry point now but uh so if we do this let's say if we do uh we do it interactively as well oh okay so what oh okay so if we don't do that part and we do it this way no let's go back I mean we ran this before so it should work oh okay it's not okay it should just be one okay cool so I mean if we do this for example then and let's say we have another here and we do here and we devel do some development print um by is this updated then yeah so that's actually really cool right because then we can actually do development this way and that way we avoid a lot of stuff in terms of us having to play around with all uh so we can just get pull this image and then start doing development instantly on any kind of VM and so on that we spin up in instead of having to uh download all the pack like download miniconda if it's a clean install right then we don't have to download miniconda set it up install it and all of these things we just have to install Docker we can get the image and start developing uh developing instantly and while we're developing we have an insurance that this works on uh this would work on any uh sort of uh if we give this to another person it would still work so yeah I think that was a good start uh to Docker uh I did feel we get like we only played around with it for one and a half two hours right but uh we did get an understanding so we did do these three things kubernetes we had an like very brief look at in terms of what it is which is basically a way of managing the Ducker containers and doing um like this horizontal scaling load balancing um scheduling and stuff so it's uh didn't really understand pods uh that kind of thing but I felt like we understood so first like Docker file we understood uh let's see we looked at sort of Docker images uh Docker containers Docker registry and then we had something called Docker compose which we didn't get into right like how to run multiple containers uh or something like that or um we learned like how to some uh attaching volume uh our current directory as a volume to the to Docker container guess like that's what we learned about um in this one so and also the docker like pushing to the docker registry um uh did you have any issues with ML Libs on M1 M2 um in the when I bought it yeah but now uh no I don't I do that the MP MPS right using uh the sort of the the MPS enabled for for pytorch is a little bit untrust I don't trust it that much it seems to be a little bit strange so I don't I don't do anything that requires a lot of um basically I don't use it uh in if I do any training or anything I do it on a a Linux machine but um no I haven't noticed any issues for sort of testing stuff and and that kind of a thing all right guys I think that that was it for this uh short uh stream where we took a look at Docker and tried to understand uh the basics of it uh we didn't obviously we did not do any extensive right we didn't do anything extensive here but um we uh learned the very Basics and at least we now know the essentials of how we can use it for live development uh we didn't see if it it it can be run on a GPU that is something we need to try maybe in a separate in the future but uh just simple stuff we learned how to do so yeah cool all right guys uh let me know if you have any questions before we uh before we uh before I shut down the stream all right guys thanks so much for uh for joining this one uh it was almost exactly two hours I think we learned a lot was a little bit of a struggle in the be or like understanding how this works but I mean that's common um yeah all right have a good day guys