Set Up Your Notebooks

Your Kubeflow deployment includes services for spawning and managing Jupyter notebooks.

You can set up multiple notebook servers per Kubeflow deployment. Each notebook server can include multiple notebooks. Each notebook server belongs to a single namespace, which corresponds to the project group or team for that server.

This guide shows you how to set up a notebook server for your Jupyter notebooks in Kubeflow.

Quick guide

Summary of steps:

1. Follow the Kubeflow getting-started guide to set up your Kubeflow deployment and open the Kubeflow UI.

2. Click Notebooks in the left-hand panel of the Kubeflow UI.

3. Click NEW SERVER to create a notebook server.

4. When the notebook server provisioning is complete, click CONNECT.

5. Click Upload to upload an existing notebook, or click New to create an empty notebook.

The rest of this page contains details of the above steps.

Install Kubeflow and open the Kubeflow UI

Follow the Kubeflow getting-started guide to set up your Kubeflow deployment in your environment of choice (locally, on premises, or in the cloud).

When Kubeflow is running, access the Kubeflow UI as described in the getting-started guide for your chosen environment. For example:

• If you deployed Kubeflow on Google Cloud Platform (GCP), the Kubeflow UI is available at the following URI:

  https://<deployment_name>.endpoints.<project>.cloud.goog/
  

• If you set up port forwarding to the Ambassador service, the Kubeflow UI is available at the following URI:

  http://localhost:8080/
  

• For other environments, see the getting-started guide for your chosen environment.

Create a Jupyter notebook server and add a notebook

1. Click Notebooks in the left-hand panel of the Kubeflow UI to access the Jupyter notebook services deployed with Kubeflow:

2. Sign in:

   • On GCP, sign in using your Google Account. (If you have already logged in to your Google Account you may not need to log in again.)
   • On all other platforms, sign in using any username and password.

3. Click NEW SERVER on the Notebook Servers page:

   You should see the New Notebook Server page:

4. Enter a name of your choice for the notebook server. The name can include letters and numbers, but no spaces. For example, my-first-notebook.

5. Enter a namespace to identify the project group or team to which this notebook server belongs. The default is kubeflow.

6. Select a Docker image for the baseline deployment of your notebook server. You can choose from a range of standard images or specify a custom image:

• Standard: The standard Docker images include typical machine learning (ML) packages that you can use within your Jupyter notebooks on this notebook server. Select an image from the Image dropdown menu. The image names indicate the following choices:

  • A TensorFlow version (for example, tensorflow-1.13.1). Kubeflow offers a CPU and a GPU image for each minor version of TensorFlow.

  • cpu or gpu, depending on whether you want to train your model on a CPU or a GPU.

    • If you choose a GPU image, make sure that you have GPUs available in your Kubeflow cluster. Run the following command to check if there are any GPUs available: kubectl get nodes "-o=custom-columns=NAME:.metadata.name,GPU:.status.allocatable.nvidia\.com/gpu"
    • If you have GPUs available, you can schedule your server on a GPU node in the Extra Resources section at the bottom of the form. For example, to reserve two GPUs, enter the following JSON code: {"nvidia.com/gpu": 2}

  • Kubeflow version (for example, v0.5.0).

• Custom: If you select the custom option, you must specify a Docker image in the form registry/image:tag. For guidelines on creating a Docker image for your notebook, see the guide to creating a custom Jupyter image.

1. Specify the total amount of CPU that your notebook server should reserve. The default is 0.5. For CPU-intensive jobs, you can choose more than one CPU (for example, 1.5).

2. Specify the total amount of memory (RAM) that your notebook server should reserve. The default is 1.0Gi.

3. Specify a workspace volume to hold your personal workspace for this notebook server. Kubeflow provisions a Kubernetes persistent volume (PV) for your workspace volume. The PV ensures that you can retain data even if you destroy your notebook server.

• The default is to create a new volume for your workspace with the following configuration:

  • Name: The volume name is synced with the name of the notebook server. When you start typing the notebook server name, the volume name takes the same value. You can edit the volume name, but if you later edit the notebook server name, the volume name changes to match the notebook server name.
  • Size: 10Gi
  • Mount path: /home/jovyan
  • Access mode: ReadWriteOnce. This setting means that the volume can be mounted as read-write by a single node. See the Kubernetes documentation for more details about access modes.

• Alternatively, you can point the notebook server at an existing volume by specifying the name, mount path, and access mode for the existing volume.

1. (Optional) Specify one or more data volumes if you want to store and access data from the notebooks on this notebook server. You can add new volumes or specify existing volumes. Kubeflow provisions a Kubernetes persistent volume (PV) for each of your data volumes.

2. Click SPAWN and wait a while. You should see an entry for your new notebook server on the Notebook Servers page, with a spinning indicator in the Status column. It can take a few minutes to set up the notebook server.

   • You can check the status of your Pod by hovering your mouse cursor over the icon in the Status column next to the entry for your notebook server. For example, if the image is downloading then the status spinner has a tooltip that says ContainerCreating.

     Alternatively, you can check the Pod status by entering the following command:

     kubectl -n <NAMESPACE> describe pods jupyter-<USERNAME>
     

     Where <NAMESPACE> is the namespace you specified earlier (default kubeflow) and <USERNAME> is the name you used to log in. A note for GCP users: If you have IAP turned on, the Pod has a different name. For example, if you signed in as USER@DOMAIN.EXT the Pod has a name of the following form:

     jupyter-accounts-2egoogle-2ecom-3USER-40DOMAIN-2eEXT
     

3. When the notebook server provisioning is complete, you should see an entry for your server on the Notebook Servers page, with a check mark in the Status column:

4. Click CONNECT to start the notebook server.

5. When the notebook server is running, you should see the Jupyter dashboard interface. If you requested a new workspace, the dashboard should be empty of notebooks:

6. Click Upload to upload an existing notebook, or click New to create an empty notebook. You can read about using notebooks in the Jupyter documentation.

Experiment with your notebook

The default notebook image includes all the plugins that you need to train a TensorFlow model with Jupyter, including Tensorboard for rich visualizations and insights into your model.

To test your Jupyter installation, you can run a basic ‘hello world’ program (adapted from mnist_softmax.py) as follows:

1. Use the Jupyter dashboard to create a new Python 3 notebook.

2. Copy the following code and paste it into a code block in your notebook:

   from tensorflow.examples.tutorials.mnist import input_data
   mnist = input_data.read_data_sets("MNIST_data/", one_hot=True)
   
   import tensorflow as tf
   
   x = tf.placeholder(tf.float32, [None, 784])
   
   W = tf.Variable(tf.zeros([784, 10]))
   b = tf.Variable(tf.zeros([10]))
   
   y = tf.nn.softmax(tf.matmul(x, W) + b)
   
   y_ = tf.placeholder(tf.float32, [None, 10])
   cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_ * tf.log(y), reduction_indices=[1]))
   
   train_step = tf.train.GradientDescentOptimizer(0.05).minimize(cross_entropy)
   
   sess = tf.InteractiveSession()
   tf.global_variables_initializer().run()
   
   for _ in range(1000):
     batch_xs, batch_ys = mnist.train.next_batch(100)
     sess.run(train_step, feed_dict={x: batch_xs, y_: batch_ys})
   
   correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))
   accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
   print("Accuracy: ", sess.run(accuracy, feed_dict={x: mnist.test.images, y_: mnist.test.labels}))
   

3. Run the code. You should see a number of WARNING messages from TensorFlow, followed by a line showing a training accuracy something like this:

   Accuracy:  0.9012
   

Please note that when running on most cloud providers, the public IP address is exposed to the internet and is an unsecured endpoint by default.

Next steps

• Explore Kubeflow Fairing for a complete solution to building, training, and deploying an ML model from a notebook.
• Learn the advanced features available from a Kubeflow notebook, such as submitting Kubernetes resources or building Docker images.

Feedback

Was this page helpful?

Glad to hear it! Please tell us how we can improve.

Sorry to hear that. Please tell us how we can improve.