NGC | Catalog
Welcome Guest
CatalogCollectionsClara Train

Clara Train

For contents of this collection and more information, please view on a desktop device.
Logo for Clara Train

Description

Clara Train - domain optimized training framework - includes Clara Train container, models, getting started jupyter notebook, utilitie

Curator

NVIDIA

Modified

June 2, 2022
Containers
Helm Charts
Models
Resources

What is NVIDIA Clara Train SDK for Medical Imaging?

Clara Train is a domain specialized developer application framework that includes a PyTorch based training framework with state of the art pre-trained models to kick start AI development with techniques like Transfer Learning, Federated Learning, and AutoML.

To enable faster creation of AI-Ready data, Clara Train includes APIs for AI-Assisted Annotation, making any medical viewer AI capable.

Getting Started

Check out our GitHub repo for an extensive collection of example Jupyter notebooks that walk you through Clara train SDK features and capabilities:

  1. Get started with Clara Train.
  2. Notebook on Federated Learning.
  3. Notebook on Pathology.

Clara Train runs on-premise and on the cloud. We have a handy Quick Start guide for launching Clara Train on AWS. For other Cloud Service Providers, refer to the documentation here.

What is New?

Clara Train v4.0 is now powered by MONAI, a domain-specialized open-source PyTorch framework, accelerating deep learning in Healthcare imaging.

v4.0 also expands into Digital Pathology and introduces homomorphic encryption for server side aggregation in federated learning.

Clara Train 4.0 includes a digital pathology pipeline for a fully convolutional classification network that works with whole-slide images and performs optimized data loading using cuCIM, which can tile large datasets on-demand and process them through a CUDA-enabled pipeline.

Homomorphic encryption allows the computation of data while the data is still encrypted, and it ensures that each client’s model weight updates to a global federated learning model stays hidden by preventing the server from reverse-engineering the submitted weights and discovering any training data.

You can access new pre-trained models compatible with Clara Train 4.0 here. Please note that v4.0 models are not compatible with previous containers/versions, and these v4.0 model names begin with the prefix “clara_pt”.

Older TensorFlow based models, which do not have the prefix “clara_pt” in their name, can be found under the model entities in this collection. They are compatible with Clara Train containers/tags before v4.0.

What is in this Collection?

  1. Pull the latest Clara Train SDK Container to get started! The “Getting Started” section on this page will lead you through what to do next.
  2. Find here PyTorch based pre-trained models for your use!

Hardware Requirements

Clara supports CUDA compute capability 6.0 and higher. This corresponds to GPUs in the Pascal, Volta, Turing, and Ampere families.

Clara Train recommends NGC-Ready systems, with NVIDIA Tesla V100 GPUs, NVIDIA Tesla T4 GPUs, or NVIDIA Ampere A100 GPUs.

Driver Requirements

Clara is based on NVIDIA CUDA 11.0.3, which requires NVIDIA Driver release 450 or later. However, if you are running on Tesla (for example, T4 or any other Tesla board), you may use NVIDIA driver release 418.40.04+ or 440.33.01+. To see details on compatibility, see CUDA Compatibility.

License

End User License Agreement is included with the product. Licenses are also available along with the model application zip file. By pulling and using the Clara Train SDK container and downloading models, you accept the terms and conditions of these licenses.

Technical Support

Use the NVIDIA Devtalk forum for questions regarding this Software