6.1 Definitions, acronyms, buzzwords, and abbreviations6.2 Benefits of IBM Spectrum Conductor Deep Learning Impact6.3 Key features of Deep Learning Impact6.3.1 Parallel data set processing6.3.2 Monitoring and Optimization for one training model6.3.3 Hyperparameter optimization and search6.3.4 IBM Fabric for distributed training6.3.5 IBM Fabric and auto-scaling6.3.6 DLI inference model6.3.7 Supporting a shared multi-tenant infrastructure6.4 DLI deployment6.4.1 Deployment consideration6.4.2 DLI single-node mode6.4.3 DLI cluster without a high availability function6.4.4 DLI cluster with a high availability function6.4.5 Binary files installation for the high availability enabled cluster6.4.6 A DLI cluster with a high availability function installation guide6.5 Master node crashed when a workload is running6.6 Introduction to DLI graphic user interface6.6.1 Data set management6.6.2 Model management6.6.3 Deep learning activity monitor and debug management6.7 Supported deep learning network and training engine in DLI6.7.1 Deep learning network samples6.7.2 Integrating with a customer’s network in DLI6.8 Use case: Using a Caffe Cifar-10 network with DLI6.8.1 Data preparation6.8.2 Data set import6.8.3 Model creation6.8.4 Model training6.8.5 Model validation6.8.6 Model tuning6.8.7 Model prediction6.8.8 Training model weight file management