book

Mastering PyTorch - Second Edition

Name: Mastering PyTorch - Second Edition
Author: Ashish Ranjan Jha
ISBN: 9781801074308

by Ashish Ranjan Jha

May 2024

Intermediate to advanced

554 pages

14h 21m

English

Packt Publishing

Read now

Unlock full access

Preface
Who this book is forWhat this book coversTo get the most out of this bookGet in touchMaking the Most Out of This Book – Get to Know Your Free BenefitsUnlock Your Book’s Exclusive BenefitsHow to unlock these benefits in three easy stepsNeed help?
Overview of Deep Learning Using PyTorch
A refresher on deep learningActivation functionsOptimization scheduleExploring the PyTorch library in contrast to TensorFlowTensor modulesPyTorch modulestorch.nntorch.optimtorch.utils.dataTraining a neural network using PyTorchSummaryReference list
Deep CNN Architectures
Why are CNNs so powerful?Evolution of CNN architecturesDeveloping LeNet from scratchUsing PyTorch to build LeNetTraining LeNetTesting LeNetFine-tuning the AlexNet modelUsing PyTorch to fine-tune AlexNetRunning a pretrained VGG modelExploring GoogLeNet and Inception v3Inception modules1x1 convolutionsGlobal average poolingAuxiliary classifiersInception v3Discussing ResNet and DenseNet architecturesResNetDenseNetUnderstanding EfficientNets and the future of CNN architecturesSummaryReferences
Combining CNNs and LSTMs
Building a neural network with CNNs and LSTMsText encoding demoBuilding an image caption generator using PyTorchDownloading the image captioning datasetsPreprocessing caption (text) dataPreprocessing image dataDefining the image captioning data loaderDefining the CNN-LSTM modelTraining the CNN-LSTM modelGenerating image captions using the trained modelSummaryReferences
Deep Recurrent Model Architectures
Exploring the evolution of recurrent networksTypes of recurrent neural networksRNNsBidirectional RNNsLSTMsExtended and bidirectional LSTMsMulti-dimensional RNNsStacked LSTMsGRUsGrid LSTMsGated orthogonal recurrent unitsTraining RNNs for sentiment analysisLoading and preprocessing the text datasetInstantiating and training the modelBuilding a bidirectional LSTMLoading and preprocessing the text datasetInstantiating and training the LSTM modelDiscussing GRUs and attention-based modelsGRUs and PyTorchAttention-based modelsSummaryReferences
Advanced Hybrid Models
Building a transformer model for language modelingReviewing language modelingUnderstanding the transformer model architectureDefining a transformer model in PyTorchLoading and processing the datasetTraining the transformer modelDeveloping a RandWireNN model from scratchUnderstanding RandWireNNsDeveloping RandWireNNs using PyTorchDefining a training routine and loading dataDefining the randomly wired graphDefining RandWireNN model modulesTransforming a random graph into a neural networkTraining the RandWireNN modelEvaluating and visualizing the RandWireNN modelSummaryReferences
Graph Neural Networks
Introduction to GNNsUnderstanding the intuition behind GNNsUsing regular NNs on graph data – a thought experimentUnderstanding the power of GNNs with computational graphsTypes of graph learning tasksUnderstanding node-level tasksUnderstanding edge-level tasksUnderstanding graph-level tasksReviewing prominent GNN modelsUnderstanding graph convolutions with GCNsUsing attention in graphs with GATPerforming graph sampling with GraphSAGEBuilding a GCN model using PyTorch GeometricLoading and exploring the citation networks datasetBuilding a simple NN-based node classifierBuilding a GCN model for node classificationTraining a GAT model with PyTorch GeometricSummaryReference list
Music and Text Generation with PyTorch
Building a transformer-based text generator with PyTorchTraining the transformer-based language modelSaving and loading the language modelUsing the language model to generate textUsing GPT models as text generatorsOut-of-the-box text generation with GPT-2Text generation strategies using PyTorchGreedy searchBeam searchTop-k and top-p samplingText generation with GPT-3Generating MIDI music with LSTMs using PyTorchLoading the MIDI music dataDefining the LSTM model and training routineTraining and testing the music generation modelSummaryReferences
Neural Style Transfer
Understanding how to transfer style between imagesImplementing neural style transfer using PyTorchLoading the content and style imagesLoading and trimming the pretrained VGG19 modelBuilding the neural style transfer modelTraining the style transfer modelExperimenting with the style transfer systemSummaryReferences
Deep Convolutional GANs
Defining the generator and discriminator networksUnderstanding the DCGAN generator and discriminatorTraining a DCGAN using PyTorchDefining the generatorDefining the discriminatorLoading the image datasetTraining loops for DCGANsUsing GANs for style transferUnderstanding the pix2pix architectureExploring the pix2pix generatorExploring the pix2pix discriminatorSummaryReferences

Image Generation Using Diffusion
Understanding image generation using diffusionUnderstanding how diffusion worksTraining a forward diffusion modelPerforming reverse diffusion or denoisingTraining a diffusion model for image generationLoading the dataset using Hugging Face datasetsProcessing the dataset using torchvision transformsAdding noise to images using diffusersDefining the UNet modelTraining the UNet modelDefining the optimizer and learning scheduleUsing Hugging Face Accelerate to accelerate trainingRunning the model training loopGenerating realistic anime images using (reverse) diffusionUnderstanding text-to-image generation using diffusionEncoding text input into an embedding vectorIngesting additional text data in the (conditional) UNet modelUsing the Stable Diffusion model to generate images from textSummaryReference list
Deep Reinforcement Learning
Reviewing RL conceptsTypes of RL algorithmsModel-basedModel-FreeDiscussing Q-learningUnderstanding deep Q-learningUsing two separate DNNsExperience replay bufferBuilding a DQN model in PyTorchInitializing the main and target CNN modelsDefining the experience replay bufferSetting up the environmentDefining the CNN optimization functionManaging and running episodesTraining the DQN model to learn PongSummaryReference list
Model Training Optimizations
Distributed training with PyTorchTraining the MNIST model in a regular fashionTraining the MNIST model in a distributed fashionDistributed training on GPUs with CUDAAutomatic mixed precision trainingRegular model training on a GPUMixed precision training on a GPUSummaryReference list
Operationalizing PyTorch Models into Production
Model serving in PyTorchCreating a PyTorch model inference pipelineSaving and loading a trained modelBuilding the inference pipelineBuilding a basic model serverWriting a basic app using FlaskUsing Flask to build our model serverSetting up model inference for Flask servingBuilding a Flask app to serve modelUsing a Flask server to run predictionsCreating a model microserviceServing a PyTorch model using TorchServeInstalling TorchServeLaunching and using a TorchServe serverExporting universal PyTorch models using TorchScript and ONNXUnderstanding the utility of TorchScriptModel tracing with TorchScriptModel scripting with TorchScriptRunning a PyTorch model in C++Using ONNX to export PyTorch modelsServing PyTorch models in the cloudUsing PyTorch with AWSServing a PyTorch model using an AWS instanceUsing TorchServe with Amazon SageMakerServing PyTorch models on Google CloudServing PyTorch models with AzureWorking with Azure’s DSVMsDiscussing Azure Machine Learning ServiceSummaryReference list
PyTorch on Mobile Devices
Deploying a PyTorch model on AndroidConverting the PyTorch model to a mobile-friendly formatSetting up the Android app development environmentUsing the phone camera in the Android app to capture imagesEnabling the camera during app startupHandling camera permissions in AndroidOpening the camera for image captureCapturing images using the phone cameraRunning ML model inference on camera-captured imagesValidating the ML model binary pathPerforming image classification on camera-captured imagesLaunching the app on an Android mobile deviceBuilding PyTorch apps on iOSSetting up the iOS development environmentUsing a phone camera in the iOS app to capture imagesRunning ML model inference on camera-captured imagesSummaryReference list
Rapid Prototyping with PyTorch
Using fastai to set up model training in a few minutesSetting up fastai and loading dataTraining an MNIST model using fastaiEvaluating and interpreting the model using fastaiTraining models on any hardware using PyTorch LightningDefining the model components in PyTorch LightningTraining and evaluating the model using PyTorch LightningProfiling MNIST model inference using PyTorch ProfilerProfiling on a CPUProfiling model inference on the GPUVisualizing model profiling resultsSummaryReference list
PyTorch and AutoML
Finding the best neural architectures with AutoMLUsing Auto-PyTorch for optimal MNIST model searchLoading the MNIST datasetRunning a neural architecture search with Auto-PyTorchVisualizing the optimal AutoML modelUsing Optuna for hyperparameter searchDefining the model architecture and loading the datasetDefining the model training routine and optimization scheduleRunning Optuna’s hyperparameter searchSummaryReference list
PyTorch and Explainable AI
Model interpretability in PyTorchTraining the handwritten digits classifier – a recapVisualizing the convolutional filters of the modelVisualizing the feature maps of the modelUsing Captum to interpret modelsSetting up CaptumExploring Captum’s interpretability toolsSummaryReference List
Recommendation Systems with PyTorch
Using deep learning for recommendation systemsUnderstanding a movie recommendation system datasetUnderstanding embedding-based recommender systemsUnderstanding and processing the MovieLens datasetDownloading the MovieLens datasetLoading and analyzing the MovieLens datasetProcessing the MovieLens datasetCreating the MovieLens dataloaderTraining and evaluating a recommendation system modelDefining the EmbeddingNet architectureTraining EmbeddingNetEvaluating the trained EmbeddingNet modelBuilding a recommendation system using the trained modelSummaryReference list
PyTorch and Hugging Face
Understanding Hugging Face within the PyTorch contextExploring Hugging Face components relevant to PyTorchIntegrating Hugging Face with PyTorchUsing the Hugging Face Hub for pre-trained modelsUsing the Hugging Face Datasets library with PyTorchUsing Accelerate to speed up PyTorch model trainingUsing Optimum to optimize PyTorch model deploymentSummaryReference list
Other Books You May Enjoy
Index

Content preview from Mastering PyTorch - Second Edition

12 Model Training Optimizations

Before serving pre-trained machine learning models, which we will discuss extensively in Chapter 13, Operationalizing PyTorch Models into Production, we need to train them. In Chapters 2 to 6, we saw the vast expanse of increasingly complex deep learning model architectures. Such gigantic models often have millions and even billions of parameters. The recent (at the time of writing) Pathways Language Model (PaLM) can have up to 540 billion parameters, for example using backpropagation to tune these many parameters requires enormous amounts of memory and compute power. And even then, model training can take days to finish.

In this chapter, we will explore ways of speeding up the model training process by distributing ...

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.

Read now

Unlock full access

More than 5,000 organizations count on O’Reilly

O’Reilly covers everything we've got, with content to help us build a world-class technology community, upgrade the capabilities and competencies of our teams, and improve overall team performance as well as their engagement.

Julian F.

Head of Cybersecurity

I wanted to learn C and C++, but it didn't click for me until I picked up an O'Reilly book. When I went on the O’Reilly platform, I was astonished to find all the books there, plus live events and sandboxes so you could play around with the technology.

Addison B.

Field Engineer

I’ve been on the O’Reilly platform for more than eight years. I use a couple of learning platforms, but I'm on O'Reilly more than anybody else. When you're there, you start learning. I'm never disappointed.

Amir M.

Data Platform Tech Lead

I'm always learning. So when I got on to O'Reilly, I was like a kid in a candy store. There are playlists. There are answers. There's on-demand training. It's worth its weight in gold, in terms of what it allows me to do.

Mark W.

Embedded Software Engineer

Publisher Resources

ISBN: 9781801074308

Cloud Computing

Data Engineering

Data Science

AI & ML

Programming Languages

Software Architecture

IT/Ops

Security

Design

Business

Soft Skills

Mastering PyTorch - Second Edition

by Ashish Ranjan Jha

12

Model Training Optimizations

Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.