Modern Keras design patterns | Session

Показать описание

"Keras is the preferred high level machine learning (ML) library for many ML practitioners, and it's now fully integrated into TensorFlow. Keras enables a spectrum of workflows, gradually exposing complexity when you need it to build more elaborate models. Using a simple variational autoencoder as an example, this Session introduces you to standard Keras design patterns that let you add customizations and low level TensorFlow code where it matters, keeping things simple everywhere else.

Resources:

Speakers: Martin Gorner, Francois Chollet

Watch more:

#GoogleIO #ML/AI

product: TensorFlow - General; event: Google I/O 2021; fullname: Martin Gorner, Francois Chollet; re_ty: Premiere;

Рекомендации по теме

Комментарии

François Chollet is a very down to earth yet extremely smart individual. Always love to see him explain a reasonably intuitive yet complex subject. Was intrigued by his take on general intelligence on the Lex Fridman Podcast. He's appeared there twice. Must watch for anyone interested in the field of machine learning and A.I

ER-sfzo

Thank you for creating this video. The explanation is super clear! PS: Functional API is super handy. I've been using it 95% of the time. It is easy to use and provides huge flexibility.

MiroslavNovak

I feel like they should upload this video on the keras documentation :)

marx

Thanks for this very clear introduction to Keras. I am wondering if in the notebook reconstruction_loss should not be reconstruction_loss_val at line 26 and 30 in the code cell defining VAE as reconstruction_loss is a function. Otherwise I get an error that a value and a function cannot be added.

lalalitamtamtum

Fantastic explanation, really good walkthrough of the different ways to use tf.keras

I didn't even know that you could define a train_step method in your model subclass, that is incredibly useful. I was writing external functions to train and test the model but then I was missing keras checkpointing, logging, early stopping, and not to mention distributed training, I'm def not gonna do that anymore. Thank you!!

timogden

Hi, I cannot find Josh Gordon Autoencoder Workshop. Do you have a link to share?

aljjxw

Modern Keras design patterns | Session

Modern Keras design patterns | Session

Deep Learning Design Patterns - Jr Data Scientist - Part 1 - Model Design Patterns

Deep Learning Design Patterns Primer - Part 1 - Introduction to Neural Networks

Architecture of a Modern CNN: the design pattern approach

Data Science Design Patterns

Deep Learning Design Patterns - Jr Data Scientist - Part 2 - Wide Convolutional Networks

Deep Learning Design Patterns - Jr Data Scientist - Part 3 - Alternative Connectivity Patterns

Deep Learning Design Patterns with Andrew Ferlitsch

Deep Learning Design Patterns Primer - Part 2 - Introduction to Convolutional Neural Networks

PyTorch in 100 Seconds

Deep Learning Design Patterns

Machine Learning Design Patterns | Michael Munn, Google

Introduction to Keras

Neural Network In 5 Minutes | What Is A Neural Network? | How Neural Networks Work | Simplilearn

A Tale of 3 Deep Learning Frameworks: TensorFlow, Keras, & PyTorch with Jules Damji & Brooke...

Deep Learning Design Patterns - Jr Data Scientist - Part 6 - Hyperparameter Tuning

How Large Language Models Work

CppCon 2017: Nicolas Guillemot “Design Patterns for Low-Level Real-Time Rendering”

Scale Transform 2021 - The Next Five Years of Keras & Tensor Flow with Francois Chollet

But what is a neural network? | Chapter 1, Deep learning

Tensorflow keras how to create custom layers

Keras Tips, Tricks, & Techniques:Create & Train Convolutional Neural N/W Using Keras | packt...

AI/ML Track Workshop 2: Keras and Modern Covnets, on TPUs

Easier data processing with Keras