TensorFlow Enterprise (TF Dev Summit '20)

Key Takeaways

[Music] hello everyone my name is Jason Mays I'm a developer advocate within the tensorflow org here at Google and today I've got Gonzalo from 10 supplier Enterprise team can you think how are you very good thanks so what's new in terms of their enterprise pentalpha enterprise it's a seamless scalable and supported tend to flow a distribution which is available in a variety of Google Cloud area products today tencel Enterprise provides users with an optimized version of tensorflow which is also which also includes a long-term version support also tensorflow Enterprise contains custom built into flow related packages such as tensor flow data sets tend to flow il10 so probability 10250 intense flow estimator each tension flow in the predestination is anchored to a particular version of tensor flow a no puppies are included in the open source version 10 to flow Enterprise is available in different Google cloud layer products such as deep learning containers air platform notebooks and deep learning VM image tensorflow enterprise also provides optimizations when use with Google cloud or with other Google cloud services such as Google Ad storage and bigquery so what demo is do you have in store today for us today we will be installing a deep learning VM image and the way we're gonna do it is from the UI first and then we also have the option to deploy a VM image from the CLI tend to flow in the presentation when use with TCP provide security fixes and selected bot patches for a period of three years all users of ten to flow open source receive only one year of security fixes in contrast sense of flow Enterprise provides three years let me give you example so tons of flow 1.15 in the open source version will release security patches for each minor version for my release for a year in contrast stencil flow Enterprise will give you three years of security fixes and book patches and all those are part I will be available in the in the github repository as open source it's very katate extra support that write itself and tensorflow Enterprise is not a fork hole it's available in Google Cloud products all the code it's available in the tensor flow open source repository and we also provide white glove service brilliant so what if I show you how to get started definitely let's see how we get started with this I'm gonna do a quick start demo and we will create a deep learning VM image in the Google Cloud console we're also going to launch and they are perform notebooks and we're gonna run notebook that is downloaded from the Thames flood website so you guys can see how tensorflow enterprise is compatible with existing tensorflow and associate populous in this case we're using transfer learning and notebook so we'll execute it and finally we will deploy a deep learning container excellent and a lot of people interested in transfer learning these days this is very exciting yes anymore so if we go to the Google Cloud console we're gonna create a deep learning VM image so you go to compute engine VM instances you're going to click on create then you go to the marketplace and you're gonna look for deep learning VM you're gonna click on the first option and select lunch in the DB learning VM you can select if you want a GPU or you want CPU and you can also select the tensorflow Enterprise version today we support inflow 1.15 and 2.1 so in this case is gonna create tens of the 1.15 version without GPUs we only gonna have a very simple image and just click the poi lovely and then we wait for that to fire up I guess there's a different way to create tensorflow Enterprise deep learning VM image and that's beyond a common line we create a CPU only version but what if we create a beautiful machine with GPU on a more recent version of tensorflow let me show you how to do it so I'm using the Google Cloud console and the only thing that you need to define is the image family in this case we're gonna use a tensorflow to with the latest GPU version we want to define one DP u so we'll find a seller later type in this case at Tesla t4 and we also want to install the Nvidia driver automatically means this installed Nvidia driver and CUDA latest version and if you wanna say reduce the cost you can the preemptable flower which will allow you to create an instance with a lower cost and just click enter so we created a deep learning beam image from the UI and we also had the option to do it from the CLI we create tons of enterprise 1.15 from the UI and a new instance from the CLI with GPU and the latest version of tons of rock which is 2.1 awesome it's great to have so many options to do the same thing right depending what you prefer say less awesome right now we're going to be creating a new a apart from notebook instance with the latest version of tensorflow enterprise 2.1 excellent see that for that you need to go to the Apple menu and then select notebooks create a new instance and you can use the default options which is tensorflow 2.1 with one nvidia test like a 80 or you can also customize customize this instance in this case I'm just gonna go with the default one I'm gonna select to install the GPU driver automatically for me that's all saves me a lot of time and click create well the instance is being creative you will see that here you have the immediate Tesla k80 and you can also have the options to let's say if you want to change it in the future you can change it to a different GPU like a big hundred or a t-4 and once the instance is available you will see they open Jupiter lab link enable this takes like a few seconds now there that the open Jupiter lab is enabled you can just click on it and you will see the Jupiter interface in this case we're gonna download new player notebook from the times flow website and just upload it here we are importing tons of the SDF and you can see it's just the same information as you do with regular tensor flow here we have 10 to flow 2.1 so let's go to the notebook use the transfer learning notebook for image recognition which uses TF care as sometimes of Rojo you don't need to do any modifications to it it will just run right on so I'm just gonna run all cells this notebook basically is downloading some images from a web server then it uses TF hope to use transfer learning the an image that we gonna try to recognize so basically the example here is is using the flowers data set we are using an TF module which is gonna help us to improve the quality of our results and they're gonna be training the model so this model training is actually happening right now you can see how the accuracy is increasing we're running for 280 bucks just for the sake of this demo and then later we will plot the results now we can see how that loss is reduce mentally and the accuracy increase over time we have some nice a flowers predictions and then we can see some of the results here and this is just without any modification so the last product that we have available for tents flow in the price is a deep learning containers so deep learning containers provides docker containers which are already pre-installed with tons of flow and if you want to use stencil flow with GPU we also prints out the Associated GPU drivers actually let me show you how to deploy one so this is my local environment there's no docker container right now so you just need to run it as easy as this you enter docker run proxying the port 8080 and I'm using the tens of float to CPU version because the docker container also uses the jupiter lab i will be able to use my jupiter lab in the in my local computer very similar today today apart from the book when you have deep learning containers you have the option to deploy deep learning containers we transfer Enterprise in other products such as Google Cloud coronaries engine for example so let's go and take a look I'm going to connect to my local host and now you can see I'm I'm there now listen you can see how is to get a start that we tend to flow enterprising Google cloud definitely thank you very much every demo hmm Thanks and where can I learn more information you can go to the google cloud website and look for tensorflow enterprise and also you can get started if you already have an account and go to the Google Cloud console and follow the steps we use for free you see [Music]