The upcoming chapters of this book will show what we have learned so far in order to implement some practical and real-world use cases of CNNs and RNNs. But before doing that, let's consider DL4J in a production environment. This chapter is divided into four main sections:
- Some considerations about the setup for a DL4J environment in production, with focus in particular on memory management, CPU, and GPU setup, and job submission for training
- Distributed training architecture details (data parallelism and strategies implemented in DL4J)
- The practical way to import, train, and execute Python (Keras and TensorFlow) models in a DL4J (JVM)-based production environment
- A comparison between DL4J and a couple of alternative DL frameworks for the Scala programming language (with particular focus on their readiness for production)
