Clara Training Framework

Clara provides a training framework to help accelerate deep learning training and inference for medical imaging use cases. It allows medical imaging researchers and developers to quickly implement new models using a high-level intuitive API. This guide describes installation, essential concepts, and tutorials to help you get started with Clara.

User Guide

• Overview
  
  • Key features
• Essential concepts
  
  • ShapeFormat and MedicalImage
  • Image Pipeline
  • Components
  • Data pipelines
  • Transforms
  • Placeholders
  • Metrics
• Installation
  
  • System requirements
  • Download the docker container
  • Running the container
  • Downloading the models
• Getting started with Clara
  
  • Supervised training
  • Using the data converter
  • Working with classification models
  • Working with segmentation models
• Medical Model Archive (MMAR)
  
  • MMAR Commands
  • Configuration
  • Using model.ckpt or model_final.ckpt
  • Cloning MMAR
• Bring your own components (BYOC)
  
  • Components for training workflow
  • Bring your own Model
  • Bring your own Transformation
  • Bring your own Data Loader
  • Bring your own Loss
  • Bring your own Metric
• API reference
  
  • API reference: Clara Train API
  • API reference: Clara Train specific code for AutoML and Federated Learning
  • API reference: AutoML
  • API reference: Federated Learning
• Clara Train FAQ
  
  • 1. Why should I use Clara Train?
  • 2. How can I debug my data preparation / augmentation output?
  • 3. For multi-label segmentation problems, how do you make the network focus on the labels and not the background?
  • 4. What is the difference between pre-transforms in the train and validate sections in the config_train.json training configuration?
  • 5. Should image_pipeline be the same in the train and validate sections in the config_train.json training configuration?
  • 6. What are the differences between the different cropping transformations?
  • 7. What are some supported 3D data transforms?
  • 8. How do I use batches_to_gen_at_once to speed up training in the CropByPosNegRatio cropping transformations?
  • 9. How can I fix the error below?
  • 10. How can I show train / validation dice per label on tensorboard?
  • 11. How can I have a metric just on the loss?
  • 12. What can I do if AMP doesn’t show me any difference in the model memory footprint?
  • 13. How can I fix the failure below in training which occurs right after loading the data list file?
• Appendix
  
  • Model details
  • Model training and validation configurations
  • Components
  • Converting from previous TLT
  • Training with multiple GPUs
  • Improving inference performance
  • Clara Train - Getting started with a Cloud Service Provider