Setting up Watson Machine Learning Accelerator

When you set up and configure Watson Machine Learning Accelerator to work with Watson Studio, you can train models by using deep neural networks and use Watson Studio experiments.

Watson Machine Learning Accelerator provides GPU accelerated training to dramatically shorten training times. You must install Watson Studio and Watson Machine Learning before setting up Watson Machine Learning Accelerator.

In order to associate Watson Studio with a Watson Machine Learning Accelerator instance, you must register it using the updateWMLClusterdetails.sh command line utility. This allows Watson Studio to locate a Watson Machine Learning Accelerator instance and identify which Watson Machine Learning Accelerator details should be used.

Locate the updateWMLClusterdetails.sh script locally after installing wml service:

1. Change directories to cpd-cli-workspace/olm-utils-workspace/work/offline/4.5.x/wml directory
2. Extract the wml case, for example: tar -zxvf ibm-wml-cpd-4.5.x.tgz
3. Run the script from ibm-wml-cpd/inventory/wmlOperatorSetup/files/updateWMLClusterdetails.sh

Alternatively, use these steps to locate and download the script:

1. Download the wml case that contains the script from this link https://github.com/IBM/cloud-pak/tree/master/repo/case/ibm-wml-cpd
2. Extract the wml case, for example tar -zxvf ibm-wml-cpd-4.5.x.tgz
3. Run the script from ibm-wml-cpd/inventory/wmlOperatorSetup/files/updateWMLClusterdetails.sh

Without proper registration, Watson Machine Learning Accelerator features cannot be used by Watson Studio and Watson Machine Learning.

Use the following parameters and details to register Watson Studio using the updateWMLClusterdetails.sh command line utility:

updateWMLClusterdetails.sh wmla_dlpd_service 9243 wml-ig wml-ig-edt wml_routename

where:

- `wmla_dlpd_service` is the WML Accelerator dlpd service information. Set the `wmla_dlpd_service` value to `wmla-dlpd.wmla-namespace.svc.cluster.local` where `wmla-namespace` is the name of the WML Accelerator tethered namespace.

- `port` is the required network that points to the Watson Machine Learning Accelerator instance. For example, `9243`

- `wml-ig` is the required name of the instance group that will be used for training. For example: `wml-ig`

- `wml-ig-edt` is the required name of the instance group that will be used for elastic distributed training. For example: `wml-ig-edt`

- `wml_routename` is the WML route name. To obtain the route name, issue the following command:

``` {: .codeblock}
oc get route/cpd -o jsonpath={.spec.host} -n cpd-namespace
Copy
```
where `cpd-namespace` is the Cloud Pak for Data namespace.

Learn more about installing IBM Watson Machine Learning Accelerator and connecting it to IBM Watson Studio and IBM Watson Machine Learning.

Parent topic: Administering Watson Machine Learning