All Segments
TensorFlow Tutorial (Sherry Moore, Google Brain)
40 segments available
The talks at the Deep Learning School on September 24/25, 2016 were amazing. I clipped out individual talks from the full live streams and provided links to each below in case that's useful for people who want to watch specific talks several times (like I do). Please check out the official website (http://www.bayareadlschool.org) and full live streams below. Having read, watched, and presented deep learning material over the past few years, I have to say that this is one of the best collection of introductory deep learning talks I've yet encountered. Here are links to the individual talks and the full live streams for the two days: 1. Foundations of Deep Learning (Hugo Larochelle, Twitter) - https://youtu.be/zij_FTbJHsk 2. Deep Learning for Computer Vision (Andrej Karpathy, OpenAI) - https://youtu.be/u6aEYuemt0M 3. Deep Learning for Natural Language Processing (Richard Socher, Salesforce) - https://youtu.be/oGk1v1jQITw 4. TensorFlow Tutorial (Sherry Moore, Google Brain) - https://youtu.be/Ejec3ID_h0w 5. Foundations of Unsupervised Deep Learning (Ruslan Salakhutdinov, CMU) - https://youtu.be/rK6bchqeaN8 6. Nuts and Bolts of Applying Deep Learning (Andrew Ng) - https://youtu.be/F1ka6a13S9I 7. Deep Reinforcement Learning (John Schulman, OpenAI) - https://youtu.be/PtAIh9KSnjo 8. Theano Tutorial (Pascal Lamblin, MILA) - https://youtu.be/OU8I1oJ9HhI 9. Deep Learning for Speech Recognition (Adam Coates, Baidu) - https://youtu.be/g-sndkf7mCs 10. Torch Tutorial (Alex Wiltschko, Twitter) - https://youtu.be/L1sHcj3qDNc 11. Sequence to Sequence Deep Learning (Quoc Le, Google) - https://youtu.be/G5RY_SUJih4 12. Foundations and Challenges of Deep Learning (Yoshua Bengio) - https://youtu.be/11rsu_WwZTc Full Day Live Streams: Day 1: https://youtu.be/eyovmAtoUx0 Day 2: https://youtu.be/9dXiAecyJrY Go to http://www.bayareadlschool.org for more information on the event, speaker bios, slides, etc. Huge thanks to the organizers (Shubho Sengupta et al) for making this event happen. CONNECT: - If you enjoyed this video, please subscribe to this channel. - AI Podcast: https://lexfridman.com/ai/ - Show your support: https://www.patreon.com/lexfridman - LinkedIn: https://www.linkedin.com/in/lexfridman - Twitter: https://twitter.com/lexfridman - Facebook: https://www.facebook.com/lexfridman - Instagram: https://www.instagram.com/lexfridman - Slack: https://deep-mit-slack.herokuapp.com
Segments Timeline
Introduction to TensorFlow
Sherry Moore introduces herself and the purpose of the tutorial on TensorFlow. She outlines the goals of the session, which include building models to solve classic machine learning problems and ensuring participants have the necessary tools to create applications like image recognition and music generation.
"so I'm going to take a picture so I remember how many of you are here smile um like Sami says my name is Sher Moore I work in the Google brain team um so today I'll be giving a tutorial on tensor flow..."
What is TensorFlow?
Moore explains TensorFlow as a machine learning library developed by Google, highlighting its popularity on GitHub with over 32,000 stars and 14,000 forks. She discusses its flexible data flow infrastructure, making it suitable for various applications beyond machine learning.
"you'll be going home with all the tools that you have to build all the wonderful things that you have watched today like all the image recognition the training of different colors Arts making music so..."
TensorFlow's Design Philosophy
Sherry discusses the design philosophy behind TensorFlow, emphasizing its modularity and the collaboration with researchers to ensure it meets their needs. She highlights the importance of transitioning from research to production seamlessly, allowing for efficient code reuse.
"tensorflow is a machine learning library that we developed uh at Google and we open sourced it last November and ever since then we have become the most most popular machine learning library on GitHub..."
Understanding Tensors
Moore introduces the concept of tensors, describing them as multi-dimensional arrays that hold data in TensorFlow. She explains how neurons in a neural network process data and how these computations are represented in a graph structure, facilitating efficient data flow.
"want you to write all the code that's typically you know just throwing away we want you to write code that can literally cut and paste and saving the file and and prze it immediately so tensor flu is ..."
Visualizing TensorFlow Operations
In this segment, Sherry visualizes how TensorFlow operates through an animation, explaining the roles of computation nodes and how data flows through the network. She emphasizes the interconnectedness of neurons and their functions in processing data.
"tensor is nothing more than a multi-dimensional array for those of you who are familiar with numpy arrays it's very similar to the ND array and the graph I think one of the gentle minority this mornin..."
Modularity in TensorFlow
Moore discusses the modular design of TensorFlow, which allows for innovation and improvement while maintaining consistent APIs. She highlights how this modularity empowers developers to work in parallel and contribute to the library.
"so everybody's like wow this sounds like magic how does it work so who said is it Sir Arthur Clark that says any sufficiently what's the word any sufficiently advanced technology is indistinguishable ..."
Portability of TensorFlow
Sherry highlights the portability of TensorFlow, explaining how it can run on various devices, including laptops, data centers, and mobile phones. She encourages creative applications, such as using TensorFlow for security systems on Raspberry Pi.
"so you write you constru your graph in your favorite language and this graph will be sent to we call the core tens FL execution system that's your run time and that's what you all will be running toda..."
Real-World Applications of TensorFlow
Moore shares examples of how TensorFlow is used at Google, including image recognition and voice search. She discusses the success of Smart Reply and the potential for TensorFlow to enhance user experiences in various applications.
"limit so uh we talked about what tensor flow is how it works how do we use it at Google we use it everywhere I think you have seen some of the examples we use it to recognize pictures this is actually..."
Exploring TensorFlow Models
In this segment, Sherry discusses the various models and libraries available in TensorFlow, including Inception and sequence-to-sequence models. She encourages participants to explore and contribute to the TensorFlow community.
"a I I believe if we have um maybe Zach can collect some stats for me later and maybe by now it'll be like 80% it's actually really funny at the very beginning when we train it the first answer is alwa..."
Getting Started with TensorFlow Labs
Moore prepares the audience for hands-on labs, focusing on classic machine learning problems like linear regression and classification. She encourages participants to engage with exercises to enhance their understanding of TensorFlow.
"developed on top of the core tensor flow apis that we encourage people to do so if whatever is out there does not fit your needs perfectly go for it develop your own and we welcome the contribution we..."
Building Your First Model
Sherry guides the audience through the process of setting up their Jupyter notebooks for the lab. She discusses the essential components needed for building a model, including input data, inference graphs, and training operations.
"and B you know all of you have learned by now W stands for weight and and bias b stands for bias so the idea is that if you're given enough samples if you're given enough X and Y values you should be ..."
Key Components of TensorFlow Models
In this segment, Sherry outlines the four key components required for building any TensorFlow model: input data, inference graph, loss definition, and optimizer. She emphasizes the importance of minimizing loss during training.
"all kinds of little toys so later I'm going to ask this question whoever can answer will get some mystery present so really pay attention Okay so pretty much with when whenever you build any model the..."
Visualizing Your Neural Network
Sherry discusses the importance of visualizing the neural network structure created in TensorFlow. She explains how to identify different nodes and their attributes, which is crucial for understanding the model's architecture.
"to minimize your loss so I'm going to be asking that later all right so now we know what we're going to do so you can go start go to that lab yeah everybody have it so shift return We Run the first on..."
Training Your Model
Sherry explains the process of training the model in TensorFlow, highlighting the difference between TensorFlow and other machine learning libraries. She introduces the concept of sessions and how they interact with the runtime.
"when your X is 0.2 you know your Y is 0.32 so this is basically your input data yeah everybody following if at any point you're kind of lost raise your hand and your buddy next to you will be able to ..."
Defining Loss Functions and Optimizers
In this segment, Sherry emphasizes the need to define loss functions and optimizers for effective training. She encourages participants to explore different options available in TensorFlow.
"you want to build yourself so this um next one is very key it's a very key concept here you see variables so variable intensive flow is how is corresponding to the square any of you remember this slid..."
Exploring TensorFlow Variables
Sherry introduces the concept of variables in TensorFlow, explaining their role in holding data, weights, and biases. She discusses best practices for naming variables and managing them during model training.
"resemble the target Network that you're are trying to to train for so now you have built it wonderful okay so you can shift return do you see anything no so exactly what have we built that's un commmo..."
Building the Training Graph
Sherry walks through the steps of building the training graph in TensorFlow, reiterating the importance of defining loss functions and optimizers. She encourages participants to experiment with their models.
"weight same thing with my so this would be a good practice because L later bi oh because I ran this so many times every single time you run if you don't restart that it's going to continue to grow you..."
Running the Training Session
In this segment, Sherry demonstrates how to run the training session in TensorFlow. She explains the significance of creating a session to communicate with the runtime and how to execute the training process.
"uncommon all the these things that you have created and see what they are and I can tell you these are different operations so that's how you actually get to learn about the network that you have buil..."
Experimenting with Optimizers
Sherry encourages participants to explore different optimizers available in TensorFlow. She highlights the importance of understanding various optimization techniques and invites contributions from the community.
"devices that's a very important concept because people constantly compare and it's just different okay so now you can also accon that to see what the initial values are but we're not going to do that ..."
Hands-On with Linear Regression
In this interactive segment, Sherry engages the audience in a discussion about linear regression and the consensus to move on to recognizing handwritten digits. She introduces the MNIST dataset and sets the stage for building a neural network to recognize digits using TensorFlow.
"would love to see your code or your models on GitHub so um back to this one how is everybody feeling this is too simple yeah should we go W just yes can I say that one oh is that right heit tap to see..."
Building the Neural Network
Sherry outlines the key components necessary for building a neural network, including data, inference graph, and training graph. She introduces critical concepts such as saving checkpoints and using placeholders, which are essential for efficient training and evaluation of the model.
"consensus to me so let's move if you just go to the bottom you can say click on this one so this is our M this model so before we start the lab so once again what are we trying to do so we have all th..."
Understanding Placeholders and Checkpoints
This segment focuses on the importance of placeholders in TensorFlow, allowing for flexible input during training and evaluation. Sherry explains how to define placeholders and the significance of saving checkpoints to resume training without starting from scratch.
"our goal yeah do everybody see the network on the side so now we can go to the lab so can anybody tell me why are the three or four things that's really important whenever you build a network what's t..."
Creating the Training Graph
Sherry demonstrates how to create a training graph in TensorFlow, defining the loss function and optimizer. She emphasizes the importance of naming variables for clarity and introduces the concept of global steps to track training progress.
"typical border play stuff so so that you hi return you import a bunch of uh libraries the second one this is just for um convenience I Define a set of constants some of them you can play with such as ..."
Visualizing Training Progress
In this segment, Sherry discusses the importance of visualizing training progress by plotting loss over time. She demonstrates how to track the loss during training, providing insights into model performance and helping participants understand the training dynamics.
"important also it's very easy for you to validate that you have indeed build a graph that you thought sometimes people call something repeatedly and they have generated this gigantic graph they were l..."
Checkpointing in TensorFlow
Sherry Moore discusses the importance of checkpointing in TensorFlow, especially when training large networks like Inception. She explains how loading from a previous checkpoint allows for continued training without starting from scratch, which is crucial for efficiency in deep learning.
"you can load from a previous checkpoint and continue training from there rather than always reinitialize all your variables and start from scratch when you're training really big networks such as Ince..."
Visualizing Training Loss
In this segment, Sherry demonstrates how to visualize the training loss in TensorFlow. She emphasizes the significance of monitoring loss metrics during training to assess model performance and ensure effective learning.
"so this is an yes so it's trying to minimize loss we can actually see this so we'll run it once okay where did I go okay very fast it's done but what if I really want to see what it's doing so python ..."
From Research to Production
Sherry explains the streamlined process of moving from research to prototyping to production in TensorFlow. She highlights how researchers can easily share their algorithms with data scientists for implementation, showcasing TensorFlow's versatility.
"down so this is how when you do large scale training this is what we typically do we have a gazillion of this javs running in the morning we would just glance at it and we know oh which one is doing r..."
Understanding Checkpoints
This segment focuses on the significance of checkpoints in TensorFlow. Sherry describes how checkpoints store weights and biases after extensive training, allowing for future model evaluations and applications.
"can do it in no time so for those of you who have run this step can you do an LS in your data path wherever you you saved that where wherever you declare your trainer to be what do you see in there ch..."
Evaluating Model Predictions
Sherry demonstrates how to evaluate model predictions using TensorFlow. She discusses the importance of validating model performance against a ground truth and how to effectively assess accuracy.
"that's really critical but how do we use it have no fear all right let's move on to 2.8 if you are not already there so can somebody tell me what we're trying to do first that's right first we load th..."
Challenges in Model Recognition
In this segment, Sherry addresses common challenges in model recognition, particularly when the model encounters unfamiliar data. She draws parallels to human learning and emphasizes the need for proper training data.
"this is what the machine sees this is where I sees and this is what the machine sees so in the amness data set all the numbers are between zero and one I believe I could be wrong but I believe it's be..."
Running Evaluations in Parallel
Sherry discusses the practice of running training and evaluation in parallel to monitor model performance continuously. She explains how frequent evaluations can inform training adjustments and improve outcomes.
"so you can play with it you know you can see how well see every time see so far it's 100% other than the first one which I cannot tell either so what are some of the exercises that we can do here what..."
TensorFlow's Evolution
Sherry concludes her tutorial by highlighting TensorFlow's evolution from research to production. She encourages audience members to explore TensorFlow and contribute to its development, emphasizing its potential in machine learning.
"did you notice anything if you try to load from a really early checkpoint how is your uh how how how good is it when it tries to identify the the digits just take a while guess yeah very bad maybe eve..."
Q&A Session on TensorFlow
In this segment, Sherry engages with the audience during a Q&A session, addressing questions about TensorFlow's C++ API, support for Windows, and the availability of TPUs for testing. This interactive discussion highlights community interest and ongoing development.
"have fun [Applause] thank you Sher we have time for questions for those who actually tried it see a so well done everybody feel like they're experts they're all ready to go make arts now right right g..."
TPUs Availability for Testing
A participant inquires about the availability of TPUs for non-Google employees. Sherry responds, indicating that TPUs are not currently available for public testing and discusses the ongoing developments regarding their accessibility.
"just curious because I I mean when I look at the road map I didn't see a clear timeline for Windows but uh thing I know that just like the reason why you cannot support Windows is because of basil bas..."
Integrating TensorFlow with Open Source Frameworks
Sherry outlines plans for integrating TensorFlow with open-source frameworks like MyOS and HDFS to facilitate distributed TensorFlow. She emphasizes the ongoing work and the lack of a solid timeline for these features.
"be available in the Google CL would you like to take that one I'm so glad we have a product boss here so that he can sorry okay thank you nice tutor I have a question are there any like plans to integ..."
Loading Custom Datasets in TensorFlow
In this segment, Sherry explains how users can load their own datasets into TensorFlow. She mentions existing examples, such as the TensorFlow Poet, and encourages users to explore retraining models with their own data.
"saying yes by November you have what you know so thank you but uh if you have this type of question I think Zach is the best person to answer to oh hi I was wondering um does sensor flow have any exam..."
Running TensorFlow on Android
Sherry discusses the capabilities of TensorFlow on Android devices, addressing concerns about model formats and performance. She reassures users that models can run across different platforms, including mobile, while highlighting potential limitations.
"go hello thank you for your presentation uh I have a question concerning the the training you can't you can train using thr tensor for in any uh virtually in any system like Android uh and uh what abo..."
TensorFlow Serving and Language Support
A participant asks about TensorFlow Serving and its language support. Sherry clarifies that models can be written in various languages and discusses the potential for future API support for additional languages beyond Python and C++.
"as well so definitely I encourage you to do that thank you cool yeah hi I have a question related to tensorflow serving so uh I went through the the online documentation and um currently I think it re..."
Creating Executables for TensorFlow Models
Sherry explains the process of creating executables from TensorFlow models for inference. She highlights the efficiency of the current implementation and encourages users to explore packaging their models for easy deployment.
"now because we are we're doing another version of uh serving tensor serving and is there any plan to you know provide uh um apis for other uh languages like you know like mxnet has something called mx..."
Closing Remarks and Resources
In the closing segment, Sherry thanks the audience for their participation and reminds them that all presentation slides will be available online. She also offers TensorFlow stickers as a parting gift, fostering community engagement.
"fast you cool okay let's uh thank Sher [Applause] [Music] again we're going to take a short break of 10 minutes let me remind you for those who haven't noticed yet but all the slides of all the talks ..."