Duration 31:12
16+
Play
Video

TensorFlow Dev Summit 2020 Keynote

Megan Kacholia
Engineering Director at Google
+ 2 speakers
  • Video
  • Table of contents
  • Video
TensorFlow Dev Summit 2020
March 11, 2020, Sunnyvale, USA
TensorFlow Dev Summit 2020
Request Q&A
Video
TensorFlow Dev Summit 2020 Keynote
Available
In cart
Free
Free
Free
Free
Free
Free
Add to favorites
38.99 K
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About speakers

Megan Kacholia
Engineering Director at Google
Manasi Joshi
Director of Software Engineering at Google
Kemal El Moujahid
Product Director, Tensor Flow at Google

Megan Kacholia is an Engineering Director on TensorFlow/Brain, and a long-time Googler. She specializes in working on large-scale, distributed systems, and finding ways to tune and improve performance in such environments.

View the profile

Building teams that empower machine learning modelers to move quickly and with confidence.We are a machine learning infrastructure team that works with focus on design for reliability.

View the profile

Entrepreneur, passionate about solving big problems with AI. Currently head of Product for TensorFlow, the world's leading open source machine learning platform.Previously at Facebook, leading the Messenger platform, connecting 20M businesses with 1.3B people, Wit.ai, the leading NLU developer platform, and M, Facebook's 200M MAU assistant.Prior to Google and Facebook, I built 3 companies in Fleet Management, Edtech, and Enterprise collaboration. I sold my last startup, LiveMinutes, to Fuze, a leading videoconferencing vendor, where I led Product and BD.

View the profile

About the talk

Join the TensorFlow team as they kick-off the 2020 TensorFlow Dev Summit. The keynote will feature new product updates for the TensorFlow ecosystem.

Share

Hi everyone. Welcome to The 20/20 TensorFlow developer Summit livestream. I'm Megan kacholia VP of engineering for TensorFlow. Thanks for tuning in to our 4th annual developers summit and our first-ever virtual event. With the recent developments of the coronavirus for wishing all of you good health safety and well-being. While we can't meet in person we're hoping the devsum is more accessible than ever to all of you. We have a lot of great talks along with exciting announcement. So let's get started. When we first open source TensorFlow, our goal was to give everyone a platform to build

AI to solve real-world problems. I'd like to share an example of one of those people. Erwin is a radiologist in the Philippines and no stranger to bone fracture images like the ones that you see here. He's a self-proclaimed AI enthusiast and wanted to learn how AI could be applied to radiology but was discouraged because he didn't have a computer science background, but then he discovered tensorflow.js which allowed him to build this machine learning application they could classify bone fracture images. Now he hopes to inspire his fellow radiologist to actively participate in building an

AI to ultimately help their patients. Erwen's not alone. Tensorflow has been downloaded millions of times with new stories like Erwin's popping up every day and it's a testament to your hard work and contributions to making TensorFlow what it is today. So on behalf of the team, I want to say a big thank you to everyone in our community. Taking a look back 2019 was an incredible year for TensorFlow. We certainly accomplished a lot together. We kicked off the year with our death Summit launch several new libraries and online educational courses hosted our first Google

summer of code went to 11 different cities for the TensorFlow Roadshow and hosted the first TensorFlow world last fall. 2019 was also a very special year for TensorFlow because we launched version 2.0. It was an important milestone for the platform because we looked at TensorFlow at the end and ask ourselves. How can we make it easy to use? Some of the changes were simplifying the API settling on kerris and eager execution and enabling production two more devices. The community really took the changes to heart and we've been amazed by what the community has built.

Hear some great examples from winners of our 2.0 death post challenge. Like disaster watch a crisis mapping platform that aggregates data and predicts physical constraints caused by a natural disaster or Deep Pavlov and NLP library for dialogue systems. And like always you told us what you liked about the latest version but more importantly what you wanted to see improved. Your feedback has been loud and clear. You told us that building models is easier, but that performance can be improved. You also are excited about the changes but migrating your

1.X system to 2.0 it's hard. We heard you and that's why were excited to share the latest version TensorFlow 2.2. We're building off of the momentum from 2.0 last year. You've told us Speed and Performance is important. That's why we've established a new bass lines so we can measure performance in a more structured way for people who had trouble migrating to 2 work making the rest of the ecosystem compatible. So your favorite Library the models work with 2.X. Finally, we're committed to 2.X core library so we won't be making any major changes.

But the latest version is only part of what we'd like to talk about today. Today we want to spend the time talking about the TensorFlow ecosystem. You've told us that a big reason why you love TensorFlow is the ecosystem. It's made up of libraries and extensions to help you accomplish your end in all goals whether it's to do cutting-edge research or applied amount in the real world. There's a tool for everyone. If you're a researcher, the ecosystem gives you control and flexibility for experimentation for Applied amount Engineers or data scientist. You get tools that help your models

have real-world impact finally their libraries in the ecosystem that can help create better AI experiences for your users. No matter where they are. All of this is underscored by what all of you the community bring to the ecosystem and our common goal of building AI responsibly will touch upon all of these areas today. Let's start first with talking about the 10th of low ecosystem for research. Tensorflow is being used to push the state of the art of machine learning in many different subfields. For example, natural language processing is an

area where we've seen tensorflow really help push the limits in model architecture the T5 model on the left uses the latest and transfer learning to convert every language problem into a text to text format. The model has over 11 billion parameters and was trained off of the Colossal clean crawled Corpus data side. Meanwhile Mina the conversational model on the right as / 2.6 billion parameters, and it's flexible enough to respond sensitively to conversational contact. Both of these models were built using tensorflow. And these are just a couple examples of what times are flow is being used

for in research. There are hundreds of papers and posters that were presented at nerves last year that use tensorflow. We're really impressed with the research produce a tensorflow everyday at Google and outside of it, and we're humbled that you trust tensorflow with your experiments. So, thank you. For always looking for ways to make your experience better. I want to highlight a few features in the ecosystem that will help you and your experiments first. We've gotten a lot of positive feedback from researchers on the board. Gov a tool. We launched last year that lets you upload and

share your experiment results by URL the URL allows for quickly visualizing hyperparameter Suites, and we were happy to see paper starting to site tensorboard. Deb's URL so that other researchers could share experiment results. Second we're excited to introduce a new performance profiler tools hat in tensorboard that provides consistent monitoring of model performance. We're hoping researchers will love the tool set because it gives you a clear view of how your model is performing including in-depth the bugging guidance. You'll get to hear more about 10. So bored. Death and

the new profiler from Gaul and shuman's talks later today. Researchers have also told us at the changes into. X make it easy for them to implement new ideas changes like eager execution in the core. It supports numpy arrays directly just like all the packages in the pydata ecosystem, you know and love the TF. Data pipelines. We rolled out are all reusable. Make sure you don't miss. Rohan's TF. Data talk today for the latest updates and tensorflow datasets already right out of the box many of the data sets. You'll find were added by our Google summer

of code students. So I want to thank all of them for contributing. This is a great example of how the TF ecosystem is powered by the community. Finally, I want to round out the tensorflow ecosystem for research by highlighting some of the add-ons and extensions that researchers love libraries like TF probability and TF agents work with the latest version an experimental libraries like Jax from Google research a composable with tensorflow, like using tensorflow data pipeline to input data into Jack. But tensorflow has never just been about pushing the state-of-the-art in deep

learning a model is only as good as the impact it has in the real world. This is one of tensorflow score strength. It is Health AI scale to billions of users. We've seen incredible ml applications being built with tensorflow. We're really humbled by all the companies big and small who trusts tensorflow with their ml work clothes. Going from an idea to having your AI create real-world impact can be hard but our users rely on tensorflow to help him accomplish this that's because the tensorflow ecosystem is built to fit your needs. It makes having to go from training to

deployment less of a hassle because you have the libraries and resources all-in-one platform. There's no switching costs involved. I want to highlight a few new things that will help you get to production faster. First you've told us that you love working with Keras and tensorflow because it's easy to build and train custom models. So we're committed to keeping TF. Caris the default high-level API. But if you're not looking to build models from scratch tensorflow Hub hosts, all the ready-to-use pre-trained models in the ecosystem. There are more than a thousand models available in TF Hub with

documentation code snippet demos and interactive collapsed all ready to be used. When you're ready to put your model into production, you can build production-ready. Ml pipelines in tensorflow extended to make sure your ml engineering just work from data validation to ml metadata tracking and today. I'm very excited to announce that using tensorflow in production is getting even easier with an exciting launch Google Cloud AI platform Pipelines. We've partnered with Google Cloud to make it easy to build end in production pipelines using cool flow and

tensorflow extended hosted by Google cloud cloud. AI platform pipelines are available today in your Google Cloud console. And if you're running tensorflow on Google Cloud tensorflow Enterprise, which we announced last year at TF World gives you the long-term support and the Enterprise scale that you need. Finally you can train and deploy your models and pipelines on custom Hardware specifically designed for a I work clothes cuz he refused and the latest version tensorflow is now optimized for cloud TPU using Keras. This means the same API you started with now helps you scaled

petaflops of CPU compute. All of these libraries are within the tensorflow ecosystem are 2.2 compatible and help you scale. So you're in the application can reach your users. But for a I too have that kind of impact it needs to be where your users are which means getting your models on device. Now we all know this requires working and some constraints like low-latency working with for network connectivity all will trying to preserve privacy. You can do all of this by using tools within the tempo flow in ecosystem. Like tensorflow Lite which

can help make your models run as fast as possible. Whether it's on CPUs gpus dsp's are other accelerators. Here's an example of how we optimize performance for mobile not be one from May last year to today. It's a big reduction in latency and something you get right out of the box with Tia flight. We're also adding Android Studio integration. So you can deploy modeled easily just simply drag and drop into Android studio and automatically generate the Java classes for the TF late model with just a few clicks. When network connectivity is a problem and you need these

power intensive models to work while still offline. You can convert them to run better on device using tensorflow light in the latest version We rebuilt the TF light converter from the ground-up to provide support for more models more intuitive error messages when conversions fail and support for control flow operations. The browser has become an exciting place for interactive amount and tensorflow. JS is allowing JavaScript and web developers to build some incredible applications. There's some exciting new models that are Now supported like facemash and mobile bird

hugging face introduce a new npm package for TF. JS, which allows you to do question answering directly in nodejs. Finally. The new webassembly back-end is available for improve CPU performance. The next two years will see an explosion of platforms and devices for machine learning and the industry needs a way to keep up m l i r is a solution to this rapidly changing landscape. It's compiler infrastructure 40-f and other Frameworks and it's backed by 95% of the world's Hardware accelerator manufacturers, and it's

helping to move the industry forward. We see how important infrastructure of like Emily are is to the future of a mouth which is why we're investing in the future of tensorflow Zone infrastructure. The new tensorflow run time is something you won't be exposed to whether developer a researcher, but it will be working under the covers to give you the best performance possible across a wide variety of the main specific Hardware. We're planning on integrating the new run time later this year, but you'll hear more from mingxing later today. Set a recap everything you've seen so far whether you're

pushing the state-of-the-art and research applying ml to real-world problems are looking to deploy AI wherever your users are. There is a tool for you in the transfer flow ecosystem now, I'd like to invite manasi on stage the hawk talk about how the ecosystem is helping Empower responsible AI. Thank you. Thanks Megan. Hi everyone. My name is manasi Joshi, and I'm an engineering director on tensorflow team. I'm making mention and you saw tensorflow ecosystem is composed of number of useful

libraries and tools that are used that are useful for a divorce in a few cases for the Dead coming from do you put yours on practitioners? I like the feeling of the question whether we are Building Systems in the most inclusive and secure pay I'm here to tell you how thankful for ecosystem empowers. Its users to build suspense responsibly and moreover what type of tools and resources are available to all users to accomplish those goals. Good morning, deep dive into the details of what tensorflow has to offer. Its users. Let's take a

step back and Define what we mean by responsible AI. How do you know what that machine learning has tremendous power for solving lots of challenging real problems. We have to do this responsibly now in the reigning Define responsible AI is based on a 5-minute strategy. Number one General recommended practices for DIY. This is all about reliability all the way from making sure that your mother is not over fitting to your training data. It is more generalized and that making sure you're aware of

limitations of your training data when it comes to different feature distributions can use for example ensuring that the model output are robust when the training day target sports bras all your model to determine its quality because different metrics matter to different context how much is used for promotion demotion filtering rankings for insta pot the second principle of fairness evolving thing in AI for us we Define it as not to Pho reinforce unwanted bias.

Venice can be extremely subjective candy context-sensitive and his associate socio-technical challenge interpretability understanding the mechanics all behind models prediction and showing that you understand what features really matter to the final output which features were important which features were not fourth privacy for this is all about taking into account sensitivity of your painting data and features and its Security in the email security really means that you understand what level it is in your system and the threat models that are associated. No, but I typically user of

tensorflow. This is how the overall developer busload of like an objective. You want to build a system. Can you go about Gathering relevant data for training your model as we understand data is gold for machine learning model training. And so you have to prepare the date of well, you have to transfer to cleanse if sometimes and then once you get your data is ready you go about training your model. Once the model is built its converged you then go about the blow the model in production systems that want to make use of him out. Deployment face is not a

one-time task. You have to continuously keep iterating in an ml workflow and improving the quality of the model now developers box loaded on many many different moments at which you as a modeler needs to be asking all of these questions. Questions like who is the audience for my machine learning model who are the stakeholders and what are their individual objectives for the stakeholders going on today is my day. Vidi. Presenting. Do you buy a seized our distribution skills and do understand.

Limitation. Am I allowed to use certain features in a privacy-preserving? Where are they are just simply not available due to constraints. Do I understand the implications of the data on model outputs or not? I might do during deployments believe blindly or am I being little bit mindful about deploying only reliable and included models and finally when we talk about flow do I understand complex feedback? Loops that could be present in my modeling walkthrough. No alone, all of

these questions. I'm happy to tell you that tensorflow ecosystem has future of tools, which could be helpful to answer some of them. I'm not going to go through everything she or what do sharks give you a few examples starting with famous indicators. It's a very effective way by which you can evaluate your performance across many different some group in a confidence interval powered way such that you can evaluate simple but effective pennis metrics for your models. We have water stools that gives you the notion of interpreting the modest outfit

based on the features and changing those features to see the changes in the models of foot. If it has very compelling visual information associated with your data, and then finally tensorflow Federated its potential for 2. X compatible life that helps you train your models with data that's available on device have a talk later today that dive deep into fairness and privacy pillars of the responsible AI strategy be sure not to miss it. Excited to work on this important part of tensorflow ecosystem

with all of you the tensorflow community and now to talk more about the community. I would like to turn it over to come off. Thank you. Thank you. My bestie everyone my name is come out and I'm the product director for tensorflow. So you've heard a lot of from Megan and Montesquieu about the latest Innovations. Now, I'm going to talk about the most important part of what we're building and that's the community and I want to begin by thanking all of you your feedback contributions. What you build. This is what makes all of this possible.

We have an incredible Global Community. We love hearing from you. And we really appreciate everyone that came out to Roadshow of 4/10 of the world last year and going into 20/20. I want to take some time to highlight more opportunities to get involved in the community and you resources to help you all succeed. Let's start with ways you can get involved locally. One great way to connect is to join a tensile for user needs grass with communities started organically and we now have 73 of them globally. We launched the first to do in Latin America after the Roadshow in San

Paulo. And now we spend our presence in Europe. The Correa group is the biggest Woodforest 6000 members and China has user groups in 16 seed. I'm sure this map can have a lot more dots. So if you want to start User Group, please reach out and will help you get started. another way to get involved are two special interest groups or cigs cigs exist to help you build areas of tensorflow that you care the most about we will have 12 6 with 6 Graphics be our latest addition starting at the end of the month moose figs are led by members of the open source Community such as are

fantastic Google developer expert we love rigidities. We now have a 148 of them and I want to take a moment to recognize all of them. They give Tech talks organize workshops and Ducks Prince. I don't want to give a special shout-out to Hast on picture tube of organized with Doxepin and soul is there reviewed several peers and wrote hundreds of comments in 5 hours again, these are amazing. So, please let us know if you're interested in becoming one. Okay, so tense of for User Group Styx and Judy's are great ways to get involved but we all love a little

competition and we all love kaggle supports to point x we've had over a thousand teen enrolled in our last competition. I want to give a special shout-out for a 2.0 Prize winner deep thought and speaking of competition We saw great project in our last death Bush challenge, including Psychopathology assistant and intelligent assistant that track a patient's response is drink face-to-face and removed section and everybody everyone gets faster and everybody Dance Now video generation Library using STFU and tensorflow 2.0.

Thank you to everyone who participated and today. We have a new challenge to manasi spoke earlier, but had tensorflow can help Empower all users to build AI systems responsible. So we want to challenge you to create something great with tensorflow 2.2 and something that has the ER principals at heart. We can't wait to see what you feel. To another area that we're investing in a lot is education starting with our with supporting our younger community members for the first time we participated in Google code in and it was a success. We were very impressed by the

students and we want to thank all the awesome mentors who made this possible. We hope someday to see the students are summer of code program. I love some of Cooke. It's an awesome opportunity for students to work with tensorflow Engineers. We saw amazing projects. In fact, one of the students worked on data visualization for Swift which is still being used today barking some happy announce. We're doing it again this summer and we're excited to see what new projects students will work on programs. Like this are key to the growth of the developer community. So, please visit

the summer of code website to learn when I apply. Will someone help produce great education content starting with r machine learning crash course a great resource for beginners. So today we launched an updated version of the courts are ended the chief completely revamped the programming exercises 2.0 and made them much simpler in the process. Go check it out on this Lane. And we want to provide resources at every stage of learning at the University level. We want to empower Educators and support them as the design develop

and teach machine learning courses last sure. We supported Georgia Tech University of Hong Kong Pace University and many others and this year. We have a commitment to fund schools from underrepresented communities in AI historically black enlightennext colleges and universities. So sure of faculty and you want to teach ML please reach out. And we also want to help people self-study. That's why we partnered with deep learning IDI to give people access to Great educational material today over 200,000 people have enrolled in our courses the date on deployment course is a great

special edition course that covers stencil for a GS and more advanced in Oreos. This is a great option for people who are really looking to build their ml coding skills and you can audit it for free and there's more Imperial College London just really is a getting started with tensorflow course on Coursera. This course was created in part but depends with the funding. I mentioned earlier and were super happy to see this. So you're taking all these courses you're becoming better ml. But how do you show your expertise to the world? This is why I'm excited to announce the

launch of the tensorflow certificate program and assessment created by the sense of protein covering topics such as Tech certification using NLP to build spam filter computer vision using CNN to do image recognition sequences and prediction by passing this foundational certification. You'll be able to share your expertise with the world and display your certificate badge on LinkedIn GitHub or defensive posture tificate Network. And to widen access to people of diverse backgrounds and experiences were excited to offer a limited number of stipends for covering the certification cost. You

can find out more at 10 support our work flash certificate. So a lot of things to do and I want to thank you again for making the tensorflow community. So awesome. As you seen the tense of ecosystem is having an incredible impact in the world today and what it really comes down to is how AI is helping make people's lives better. That's really what inspires it for the team to build all these amazing tool. So I'd like to end by sharing one final story. Results of

hotel is providing care in 70 countries country in the world. We discovered them today, but has 10 million deaths by 2050. Wants to identify that there is a bacteria you do multiple test to know which type of bacteria it is. And then you test the sensitivity of identify the most effective one. There is an additional step that is the interpretation of this result in the majority of the countries where there is a microbiologist and a human resources to do this interpretation step.

Anyway, he is 10 year olds and tooth injury was it exploded inside his leg fanfiction? We Are The Last Hope for Android? I saw a I might be one of the solution to this problem. We developed an application to help let technicians to interpret the results of the analysis test only using their mobile phones. Where do you think tensorflow computer vision and machine learning to detect interactions happening between the bacteria and antibiotics based on an exploded image of the pigeon dish? Our goal is not to

replace electric Nations. The app is really meant to support them and doing their diagnosis test always keeping a human in the loop. We managed to train a muzzle within a matter of days using carrots and 15,000 animals pictures of diagnosis test. Kpi whisper expressive and it was surprisingly really quick and easy to achieve the app using tensorflow Lite so I can be used offline on any branch of mobile devices in all of our clinics. Today we have a prototype.

I'm excited because this application can be a game-changer that's going to help me. Or anywhere there is one antibiotic suitable for disinfection. I think this time it will solve many problems everywhere. That's amazing and the incredibly inspiring when I see something like this. It makes me very proud to be building tensorflow. So go build an amazing things, and we'll be there to help and with that. I will pass it on to page to catch up our day. Thank you.

Cackle comments for the website

Buy this talk

Access to the talk “TensorFlow Dev Summit 2020 Keynote”
Available
In cart
Free
Free
Free
Free
Free
Free

Ticket

Get access to all videos “TensorFlow Dev Summit 2020”
Available
In cart
Free
Free
Free
Free
Free
Free
Ticket

Interested in topic “AI and Machine learning”?

You might be interested in videos from this event

March 11, 2020
Sunnyvale
30
215.89 K
dev, google, js, machine learning, ml, scaling, software , tensorflow, web

Similar talks

Na Li
Software Engineer at Google
Available
In cart
Free
Free
Free
Free
Free
Free
Catherina Xu
Product at Google
+ 1 speaker
Miguel Guevara
Product Manager at Google
+ 1 speaker
Available
In cart
Free
Free
Free
Free
Free
Free
Jacques Pienaar
Software Engineer at Google
+ 1 speaker
Tatiana Shpeisman
Senior Engineering Manager at Google Brain
+ 1 speaker
Available
In cart
Free
Free
Free
Free
Free
Free

Buy this video

Video

Access to the talk “TensorFlow Dev Summit 2020 Keynote”
Available
In cart
Free
Free
Free
Free
Free
Free

Conference Cast

With ConferenceCast.tv, you get access to our library of the world's best conference talks.

Conference Cast
570 conferences
23018 speakers
8623 hours of content
Megan Kacholia
Manasi Joshi
Kemal El Moujahid