Working with Watson Machine Learning Accelerator notebooks in IBM Cloud Pak for Data

If you are upgrading from IBM Cloud Pak for Data Version 4.8 to Version 5.0.x, you will need to update your notebooks. After upgrading, you must:

• Edit the notebook runtime

From IBM Cloud Pak for Data, follow these steps to find and edit a notebook runtime:

1. In the project navigation trail, click your project name to return to your project.

2. Click the Assets tab to find the notebook.

3. When you click the notebook, it will open in READ ONLY mode.

4. To edit the notebook, click the pencil icon .

5. Click the Information icon to open the Information panel.

6. On the General tab, edit the name and description of the notebook.

7. Click the Environment tab to change the environment used to run the notebook and update the runtime to either Runtime 24.1 on Python 3.11 (recommended) or GPU Runtime 24.1 on Python 3.11.

If you are upgrading from IBM Cloud Pak for Data Version 4.7 to Version 5.0.x, you will need to update your notebooks. After upgrading, you must:

• Edit the notebook runtime
• Edit the runtime code

From IBM Cloud Pak for Data, follow these steps to find and edit a notebook runtime:

1. In the project navigation trail, click your project name to return to your project.

2. Click the Assets tab to find the notebook.

3. When you click the notebook, it will open in READ ONLY mode.

4. To edit the notebook, click the pencil icon .

5. Click the Information icon to open the Information panel.

6. On the General tab, edit the name and description of the notebook.

7. Click the Environment tab to change the environment used to run the notebook and update the runtime to either Runtime 24.1 on Python 3.11 (recommended) or GPU Runtime 24.1 on Python 3.11.

To ensure that your notebook code runs using notebook runtime Runtime 24.1 on Python 3.11 or GPU Runtime 24.1 on Python 3.11, you must:

• Update import FabricModel statement
• Remove references to DATA_DIR

Change the following import statement from:

from fabric_model import FabricModel

To:

from pyedt.fabric_model import FabricModel

For example, change the following code from:

"DATA_DIR = os.getenv("DATA_DIR")"
"print("DATA_DIR: " + os.getenv("DATA_DIR"))"

To:

"DATA_DIR = os.getenv("DATA_DIR", "/gpfs/mydatafs")"
"print("DATA_DIR: " + DATA_DIR)"

Example: to download data from a directory, replace the following by removing the following code:

datasets.MNIST(DATA_DIR, download=True)

def getDatasets():
    return (datasets.MNIST(DATA_DIR, train=True, download=False, transform=transforms.Compose([
                           transforms.ToTensor()
                       ])),
            datasets.MNIST(dataset_dir, train=False, download=False, transform=transforms.Compose([
                           transforms.ToTensor()
                       ]))
            )

Replacing the code referencing DATA_DIR with the following code:

dataset_dir = <your_storage_volume_path>
def getDatasets():
    return (datasets.MNIST(dataset_dir, train=True, download=True, transform=transforms.Compose([
                           transforms.ToTensor()
                       ])),
            datasets.MNIST(dataset_dir, train=False, download=Ture, transform=transforms.Compose([
                           transforms.ToTensor()
                       ]))
            )