BUILDING SCALABLE DEEP LEARNING PIPELINES ON AWS : develop, train, and deploy deep learning models

Table of Contents About the Author About the Technical Reviewer Acknowledgments Introduction Chapter 1: Overview of Scalable Deep Learning Pipelines on AWS Components of a Deep Learning Workflow on AWS Data Source (S3) Data Preprocessing (PySpark) Model Building (PyTorch and TensorFlow) Model Training (EC2) Model Evaluation (EC2) Model Deployment (Airflow) Project Directory Structure Virtual Environment Development Environment Summary Chapter 2: Setting Up a Deep Learning Environment on AWS Creating an AWS Account Provisioning Amazon EC2 Instances Setting Up Amazon S3 Creating a Project Directory Creating a Virtual Environment Installing and Configuring Dependencies Installing PySpark Installing PyTorch Installing TensorFlow Installing Boto3 Installing Airflow Standalone Installation Docker-Based Setup Installing JupyterLab Setting Up a Databricks Account and Workspace Summary Chapter 3: Data Preparation with PySpark for Deep Learning The Dataset PySpark’s Parallel Processing Data Preparation with PySpark for PyTorch Data Preparation with PySpark for TensorFlow Bringing It All Together Data Exploration with PySpark Parallel Processing in PySpark Data Preparation with PySpark for PyTorch Data Preparation with PySpark for Tensorflow Summary Chapter 4: Deep Learning with PyTorch for Regression The Dataset Predicting Tesla Stock Price with PyTorch Bringing It All Together Exploring the Tesla Stock Dataset Predicting Tesla Stock Price with PySpark and PyTorch Comparing and Plotting Actual and Predicted Values Summary Chapter 5: Deep Learning with TensorFlow for Regression The Dataset Predicting Tesla Stock Price with TensorFlow TensorFlow vs. PyTorch Bringing It All Together Layers and Activation Functions environment.yml TensorFlow Regression Code Summary Chapter 6: Deep Learning with PyTorch for Classification The Dataset Predicting Diabetes with PyTorch Enhancing Model Evaluation with Cross-Validation Bringing It All Together Exploring the Pima Diabetes Dataset Model Building, Training, and Evaluation with PyTorch Diabetes Classification Without K-Fold Cross-Validation Diabetes Classification With K-Fold Cross-Validation Summary Chapter 7: Deep Learning with TensorFlow for Classification The Dataset Predicting Diabetes with TensorFlow TensorFlow vs. PyTorch Optimizing Model Performance with Hyperparameter Tuning Bringing It All Together Building and Training a TensorFlow Model with Fixed Hyperparameters Optimizing the TensorFlow Model with Hyperparameter Tuning Summary Chapter 8: Scalable Deep Learning Pipelines with Apache Airflow An Airflow Pipeline for Tesla Stock Price Prediction Tesla Stock Price Prediction Without Airflow DAG Tesla Stock Price Prediction with Airflow DAG An Airflow Pipeline for Diabetes Prediction Diabetes Prediction Without Airflow DAG Diabetes Prediction with Airflow DAG Bringing It All Together Tesla Stock Price Prediction Without Airflow DAG Tesla Stock Price Prediction with Airflow DAG Diabetes Prediction Without Airflow DAG Diabetes Prediction with Airflow DAG Summary Chapter 9: Techniques for Improving Model Performance The Baseline Model Early Stopping Dropout L1 and L2 Regularization Learning Rate Model Capacity Automating Hyperparameter Optimization with Keras Tuner Bringing It All Together Baseline Model Early Stopping Dropout L1 Regularization L2 Regularization Learning Rate Model Capacity Automating Hyperparameter Tuning Using Keras Tuner Summary Chapter 10: Deploying and Monitoring Deep Learning Models Steps in Deploying and Monitoring Deep Learning Models Step 1: Setting Up the Environment Step 2: Developing the DAG Step 3: Uploading to S3 Step 4: Configuring the MWAA Environment Step 5: Triggering DAG Execution Step 6: Monitoring Execution Summary