There are two incoming changes in Watson Machine Learning.
Upgrade to TensorFlow 1.13 and deprecate older versions
Due to a recent security vulnerability for multiple TensorFlow versions, we decide to upgrade TensorFlow versions to 1.13 for Watson Machine Learning deployment and training runtimes and deprecate all unsecure TensorFlow versions, including 1.5 and 1.11.
If you currently have older TensorFlow models deployed in Watson Machine Learning, you will need to download the model, test it in TensorFlow 1.13 environment, and deploy it back to Watson Machine Learning. You may need to make minor modifications based on the TensorFlow version compatibility guide—in many cases, TensorFlow is backward compatible for your model.
Datetime format change in deployment API
As part of the improvements to the Watson Machine Learning service, we changed the datetime format returned from our API. This change will impact the users who are consuming V4 API-supported WML Python clients for creating deployments or jobs and parsing the datetime fields in the deployment- or jobs-related metadata.
The date format currently being returned in a GET response of /v4/deployments is:
yyyy-MM-dd'T'HH:mm:ssZZZZ
The new format is:
yyyy-MM-dd'T'HH:mm:ss.SSS'Z'
Here are the dates you need to know
- TensorFlow 1.13 runtimes available: August 13, 2019
- Older TensorFlow versions deprecation announcement: August 13, 2019
- End of Life for older TensorFlow versions: September 30, 2019
You can read more about working with Watson Machine Learning runtimes—including the new TensorFlow 1.13 runtime—in our documentation.