Appendix

• Model details
  
  • Annotation models
• Model training and validation configurations
  
  • Training configuration
  • Validation configuration
• Components
• Converting from previous TLT
  
  • Transforms
  • ImagePipeline
  • New variables in config_train.json
  • Other changes
• Training with multiple GPUs
  
  • How it works
  • Training parameters
• Improving inference performance
  
  • Training with dynamic shape
  • Inference models trained with dynamic shape
  • Objectivity of trained models
• Clara Train - Getting started with a Cloud Service Provider
  
  • Recommended system and instance configuration
  • CSP-specific Configuration
  • Getting Started with Clara Train Jupyter Notebooks