What's new and changed in Watson Machine Learning Accelerator

Watson Machine Learning Accelerator updates can include new features, bug fixes, and security updates. Releases are listed in reverse chronological order so that the latest release is at the beginning of the topic.

watsonx.ai and watsonx.governance Version 2.0.2

A new version of Watson Machine Learning Accelerator was released in August 2024 with watsonx.ai and watsonx.governance 2.0.2.

Version 2.0.2 is installed on IBM Cloud Pak for Data 5.0.2 with the service operand version 5.0.2.

This release includes various fixes.

watsonx.ai and watsonx.governance Version 2.0.1

A new version of Watson Machine Learning Accelerator was released in July 2024 with watsonx.ai and watsonx.governance 2.0.1.

Version 2.0.1 is installed on IBM Cloud Pak for Data 5.0.1 with the service operand version 5.0.1.

This release includes the following changes:

New features
This release of Watson Machine Learning Accelerator includes the following features:
New default replica setting
For medium and large sizing of the Watson Machine Learning Accelerator service, the Watson Machine Learning Accelerator pods are now set to scale to a minimum of 2 replicas to improve redundancy, minimize failover, and reduce downtime of the service.
Issues fixed in this release
The following issues were fixed in this release:
Pods continue running after shutdown
  • Issue: When a user shuts down the Watson Machine Learning Accelerator service, some pods continue to run.
  • Resolution: Stop the pods.

watsonx.ai and watsonx.governance Version 2.0.0

A new version of Watson Machine Learning Accelerator was released in June 2024 with watsonx.ai and watsonx.governance 2.0.0.

Version 2.0.0 is installed on IBM Cloud Pak for Data 5.0.0 with the service operand version 5.0.0.

This release includes the following changes:

New features
This release of Watson Machine Learning Accelerator includes the following features:
New deep learning libraries
You can now use the following deep learning libraries with Watson Machine Learning Accelerator:
  • Python 3.11.5
  • PyTorch 2.1.2
  • Tensor Flow 2.14.1
  • NVIDIA CUDA Toolkit 12.2.0

If you have existing models, update and test your models to use the latest supported frameworks. For more information, see Supported deep learning frameworks in the Watson Machine Learning Accelerator documentation.

New NVIDIA GPU Operator version
You can now use the following deep learning libraries with Watson Machine Learning Accelerator:
  • Version 24.3.0
Issues fixed in this release
The following issues were fixed in this release:
A TensorFlow job using multiple GPUs per worker fails
  • Issue: A Tensor Flow elastic distributed training job fails when using 8 GPU per worker.
  • Resolution: A TensorFlow elastic distributed training job run successfully when using multiple GPUs per worker.
Model deployment fails if a shared GPU allocation policy is specified
  • Issue: When deploying a model with a shared GPU allocation policy, that deployment will fail.
  • Resolution: Model deployment is successful.