Duration 12:56
16+
Play
Video

Using GPUs in TensorFlow, TensorBoard in notebooks, finding new datasets, & more!

Paige Bailey
Product Manager (TensorFlow) at Google
+ 1 speaker
  • Video
  • Table of contents
  • Video
TensorFlow Dev Summit 2019
March 7, 2019, Sunnyvale, CA, United States
TensorFlow Dev Summit 2019
Video
Using GPUs in TensorFlow, TensorBoard in notebooks, finding new datasets, & more!
Available
In cart
Free
Free
Free
Free
Free
Free
Add to favorites
5.33 K
I like 0
I dislike 0
Available
In cart
Free
Free
Free
Free
Free
Free
  • Description
  • Transcript
  • Discussion

About speakers

Paige Bailey
Product Manager (TensorFlow) at Google
Laurence Moroney
Staff Developer Advocate at Google

Laurence is a developer advocate at Google working on machine learning and artificial intelligence. He's the author of dozens of programming books, and hundreds of articles. When not Googling, he's author of a best-selling Science Fiction book series, and a produced screenwriter.

View the profile

About the talk

In a special live episode from the TensorFlow Dev Summit, Paige (@DynamicWebPaige) and Laurence (@lmoroney) answer your #AskTensorFlow questions. Learn about using GPU in TensorFlow, saving models as a SavedModel, running TensorBoard on Colab, using feature columns with Keras, and where to find new datasets.

Share

Welcome back and Paige Bailey and misses and we are here to answer all of your ask tensorflow questions of potential play Deb Summitt. So if you have any questions, please submit them to social media with the hashtag ask tensorflow answer many of them as we can. But any that we can't get through today will try to reach out to you later to ask them. So should I try this one on this when I think him on Twitter from you and it said once I install tensorflow GPU important work for me than trying to use it but has been unable to do so if it fails to load native fence

for runs, I remember seeing this one and icicles on YouTube. I was in response to one of my videos about pip install tensorflow GPU. I in collab because once upon a time and cola have you have to pick up install tensorflow GPU to be on to use it and now if you try to do that you end up having some issues and the reason for that is actually really good and it's good news, and it's because you don't need to do it anymore. My laptop I can show you this was The Notebook that are showing earlier on and all you have to do if you want to use GPU in collab is just change the runtime type pick

gpus the hardware accelerator and now you don't need to pick up install tensorflow GPU. It actually doesn't for you under the hood behind the scene. So it's really really cool. And that's why I earlier I was able to train this so quickly cuz I was actually using the GPU and as you can see, there's no pip install GPU on excellent whenever we were initially testing tensorflow to data we had a kind of similar issue as well with the GPU install and that you needed specific Cuda drivers, but now I could attend is supported in collab as well. So I'm so

new GPU stuff cuz this was something that I ran into a number of times when trying to use the GPU was that you always have to carefully take a look at the version of Cuda and cudnn that you're using because I made Mistake that I just went to the like a vendor's website. I downloaded the latest versions. I install them and then I saw tensorflow was actually supporting it's like the earlier version. So if you do get an error when you're trying to use GPU just take a look at the version of the

driver that is looking to support and then from the vendor's website download that specific version. If I get it it's one of the things that makes our job interesting. All right, so excellent questions on Twitter and end and we'll get to all of them not in this not in this sort of asked stagnant, but let's focus on just one for today and then answer the rest offline. How do I use different file for my social we drill into that? Which is the preferred format for saving the model going forwards saved model or something else and if we look at the laptop we can take a

gander at one of the slides from the keynote this morning. I'm really showing that Charis is a first-class citizen in tensorflow 2.0 and saved model is at the heart of every of every deployment. So here you can see saved model be used for tensorflow serving for and lots of other language findings. So really where it where it were pushing for the safe model and it's a lot easier to use than some of the other some of the other deployment options that we've seen before the

models take a look at save model consider using save model and as a result, it's like not only has the advantage of the file format, but just how it supported across all of these things. Going to keep selling it or is that we have a number of code samples in tutorials available today about how you can deploy your models with safe model white stuff into. O. But I saving as a safe model and then going through the conversion to the CFI process. It was a lot easier for me than an egg in previous iterations where I have to use like

tohko converter and all that kind of stuff. So it's really been fine with radiator rating on that and I think it's really cool. So, all right. So thanks for all of those questions are some great stuff in there. We will try to answer some of the rest of them understood that most of them are focused on the file format. Hopefully save model will help you. So this next one comes from is it possible to run pencil board on color labs and I noticed made things really happy cuz going to be

so delighted, you know, we were like talking about it. We really want to know. Tell painful like that and you open if you wanted to get it working in a lab notebook orange. If you ended up using a tool like and Grog and that was kind of not approved by our asses are in general and set but the but yes, so the good news is that you can run tensorboard and collapse. We all got this email from page and it was full of all these smiley emoji. Thank you for the question running through in and sort of

downloading some files in the hope that we could play with it a little bit but here you can see it actually working you should be able to do a different operations like smoothing changing some of the values and then also making also using the embedding visualizer directly from your collab notebook in order to understand kind of your accuracy. Be able to be able to do model performance to buy game another nice thing that the team has been working. Very very hard on is that you don't have to specify ports. You don't have to you don't have to remember if you

wanted to have multiple times reported instances running that that you were using. What is it $6,006 or whatever for another just automatically selects one that would be a good candidate and and creates it for you. So that the team is phenomenal if you have any interest whatsoever in tensorboard at all, I suggest stalking their PR is like the test and the documentation for the video as well as little notes underneath for you to go. They've been doing such great work

everywhere everywhere and we all lift. So time to board really is a collection of these different visualization so you can see scalars like your accuracy. You can see histograms. You can see I'm betting visualizers which allows you to do class ring like that great and missed example from the depths of it a couple years ago to get moving in Google summer of code projects. This summer are focused on getting additional a visualization plugins added to test your bored. So in addition to what is 2 +

addition to You could make your own this all day. I think we should move on to some other questions. So the next question that came in from amirhossein her and how would you use feature columns with Kara? And I know you are watching feature columns are really part of that Tomatoes Ride their way of like really getting your data efficiently in to estimators. I'm with people like to have some great stuff around that I have but I know that your YouTube channel has the series from Carmel who spoke earlier today

director like for Highland maybe eyes and she has this great series around high-level apis for tensorflow until like really really teach me how to use the high-level apis. And Carmel is working actively and her team a working active. Power to your friends like feature columns intense about to I'm not there yet. But yes, and we're also in the process of building out if you wanted to migrate your models from using estimators to being more about episode 2. O format with

Charis. I am currently in the process of building a migration guide. So if you have any interest around that please feel free to reach out and we're excited to get that released pretty soon in tensorflow was without tomatoes before I learn Kara and I really want to go back and kind of start changing them to scare us without rewriting them know that book right as Tomatoes were really give you the power Charis great for beginners. So it's like hopefully we'll get the best of both worlds. So it's like so Super

some simple data sets for testing and comparing different training methods different looking for new dataset. Like this is great fashion mnist is great. But after a while people want something there in fracture, right? So what what time do you think we can say to Jack we can tell them about those data ingestion those data ingestion pipelines for you to be able to easily use a variety of data sets with all of your deep learning and machine learning models with

just a few lines code. So if you're familiar with scikit-learn and all of its it's sort of nifty data ingestion practices. This feels very similar. It's very easy to do training and testing splits and verifications and we have a lot of data sets readily available right now for you to go and explore. Another thing that I would especially like to call out is a member of Arc. Unity so anybody can make your data famous, right? Like you have a dataset that you're using on your research lab if you have like a bright and shiny CSP that used to do you

think would be a cool add to the stress cause an undergrad research at Stanford. He added this birthday to set from his lab 200,000 chest radiogram Stanford, and he was able to do it in like less than it really is just as simple as take the the template format for images are for audio or whatever you're using add some additional potentially the types for some of the features that you're using with that and absolutely I can make your day. I'm one of the really important things about getting started. Is that if you take a look

at a lot of the samples free tensorflow dataset lots and lots of code about download your data unzip it here, you know, we label it like this put these files in these folders or take UCSB and make these features, but when it's in pfds, it's like one or two lines of code and all in the day to adjust get sorted into training and past tense for you that type of thing and it sounded like really really exciting because I didn't have to go through a hundred lines of code before I got to the neuron that work and sort of the

statistical distributions and this this is really helpful.

Cackle comments for the website

Buy this talk

Access to the talk “Using GPUs in TensorFlow, TensorBoard in notebooks, finding new datasets, & more!”
Available
In cart
Free
Free
Free
Free
Free
Free

Access to all the recordings of the event

Get access to all videos “TensorFlow Dev Summit 2019”
Available
In cart
Free
Free
Free
Free
Free
Free
Ticket

Interested in topic “Software development”?

You might be interested in videos from this event

September 28, 2018
Moscow
16
159
app store, apps, development, google play, mobile, soft

Similar talks

Paige Bailey
Product Manager (TensorFlow) at Google
+ 1 speaker
Laurence Moroney
Staff Developer Advocate at Google
+ 1 speaker
Available
In cart
Free
Free
Free
Free
Free
Free
Paige Bailey
Product Manager (TensorFlow) at Google
+ 1 speaker
Laurence Moroney
Staff Developer Advocate at Google
+ 1 speaker
Available
In cart
Free
Free
Free
Free
Free
Free
Gal Oshri
Product Manager at Google
Available
In cart
Free
Free
Free
Free
Free
Free

Buy this video

Video

Access to the talk “Using GPUs in TensorFlow, TensorBoard in notebooks, finding new datasets, & more!”
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
558 conferences
22053 speakers
8194 hours of content