Watson Machine Learning will migrate the GPU cluster from London to Frankfurt to serve European Union clients.
Users will need to change their Watson Machine Learning service endpoint and credentials for this migration (instructions to find WML credentials). The GPU cluster in Frankfurt will continue to work with data stored in the London data center. Users don’t need to move data.
Users can start migrating GPU training workloads to the Frankfurt cluster starting April 21, 2020, and the current London cluster will be deprecated May 21, 2020. After the deprecation date, all running jobs in the London cluster will be terminated and new training jobs submitted to the London cluster will receive an error.
For a smooth transition, here are the steps we suggest
- Create a Watson Machine Learning instance in Frankfurt if you don’t have one.
- Use the Frankfurt instance for all new training jobs.
- Save intermediate results of current running jobs in London, if possible.
- If it’s not possible to save intermediate results and the estimated job complete time exceeds our deadline, please create a case through the Watson Studio Support Center so that we are aware of your situation and can provide assistance.
Dates you need to know
- Availability of GPU cluster in Frankfurt: April 21, 2020
- Deprecation of GPU cluster in London: May 21, 2020