Use notebooks

Supported notebook images

Notebook instances in KakaoCloud Kubeflow provide a web-based machine learning development environment.
These notebook instances run based on Docker images, and the following images are currently supported in KakaoCloud Kubeflow.

info

The image registry endpoint is mlops.kr-central-2.kcr.dev/kc-kubeflow-registry/(image-name).
For example, to pull the jupyter-scipy:v1.8.0.py38.1a image, use:
mlops.kr-central-2.kcr.dev/kc-kubeflow-registry/jupyter-scipy:v1.8.0.py38.1a.

Supported notebook images

Kubeflow 1.10 Version

Image	Python Version	Platform	Framework (Version)	GPU Support	HDE Support
kc-kubeflow-registry/jupyter-scipy:v1.10.0.py310.1a	python 3.10.16	JupyterLab	scipy(1.15.1)	X	X
kc-kubeflow-registry/jupyter-scipy:v1.10.0.py311.1a	python 3.11.9	JupyterLab	scipy(1.15.1)	X	X
kc-kubeflow-registry/jupyter-pytorch-full:v1.10.0.py310.1a	python 3.10.16	JupyterLab	pytorch(2.5.1)	X	X
kc-kubeflow-registry/jupyter-pytorch-full:v1.10.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.5.1)	X	X
kc-kubeflow-registry/jupyter-pytorch-cuda-full:v1.10.0.py310.1a	python 3.10.16	JupyterLab	pytorch(2.5.1)	O	X
kc-kubeflow-registry/jupyter-pytorch-cuda-full:v1.10.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.5.1)	O	X
kc-kubeflow-registry/jupyter-tensorflow-full:v1.10.0.py310.1a	python 3.10.16	JupyterLab	tensorflow(2.17.1)	X	X
kc-kubeflow-registry/jupyter-tensorflow-full:v1.10.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.17.1)	X	X
kc-kubeflow-registry/jupyter-tensorflow-cuda-full:v1.10.0.py310.1a	python 3.10.16	JupyterLab	tensorflow(2.17.1)	O	X
kc-kubeflow-registry/jupyter-tensorflow-cuda-full:v1.10.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.17.1)	O	X
kc-kubeflow-registry/jupyter-pyspark-pytorch:v1.10.0.py310.1a	python 3.10.16	JupyterLab	pytorch(2.5.1)	X	O
kc-kubeflow-registry/jupyter-pyspark-pytorch:v1.10.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.5.1)	X	O
kc-kubeflow-registry/jupyter-pyspark-pytorch-cuda:v1.10.0.py310.1a	python 3.10.16	JupyterLab	pytorch(2.5.1)	O	O
kc-kubeflow-registry/jupyter-pyspark-pytorch-cuda:v1.10.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.5.1)	O	O
kc-kubeflow-registry/jupyter-pyspark-tensorflow:v1.10.0.py310.1a	python 3.10.16	JupyterLab	tensorflow(2.17.1)	X	O
kc-kubeflow-registry/jupyter-pyspark-tensorflow:v1.10.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.17.1)	X	O
kc-kubeflow-registry/jupyter-pyspark-tensorflow-cuda:v1.10.0.py310.1a	python 3.10.16	JupyterLab	tensorflow(2.17.1)	O	O
kc-kubeflow-registry/jupyter-pyspark-tensorflow-cuda:v1.10.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.17.1)	O	O
kc-kubeflow-registry/jupyter-llm-dev-kit:v1.10.0.py310.1a	python 3.10.16	JupyterLab	pytorch(2.5.1)	O	O
kc-kubeflow-registry/codeserver-python:v1.10.0.py310.1a	python 3.10.16	CodeServer		X	X
kc-kubeflow-registry/codeserver-python:v1.10.0.py311.1a	python 3.11.9	CodeServer		X	X
kc-kubeflow-registry/rstudio-tidyverse:v1.10.0.kbm.1a		RStudio		X	X

Kubeflow 1.8 Version

Image	Python Version	Platform	Framework (Version)	GPU Support	HDE Support
kc-kubeflow/jupyter-scipy:v1.8.0.py38.1a	python 3.8.10	JupyterLab	scipy(1.10.1)	X	X
kc-kubeflow/jupyter-scipy:v1.8.0.py311.1a	python 3.11.9	JupyterLab	scipy(1.11.3)	X	X
kc-kubeflow/jupyter-pytorch-full:v1.8.0.py38.1a	python 3.8.10	JupyterLab	pytorch(2.3.0)	X	X
kc-kubeflow/jupyter-pytorch-full:v1.8.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.3.0)	X	X
kc-kubeflow/jupyter-pytorch-cuda-full:v1.8.0.py38.1a	python 3.8.10	JupyterLab	pytorch(2.3.0)	O	X
kc-kubeflow/jupyter-pytorch-cuda-full:v1.8.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.3.0)	O	X
kc-kubeflow/jupyter-tensorflow-full:v1.8.0.py38.1a	python 3.8.10	JupyterLab	tensorflow(2.13.1)	X	X
kc-kubeflow/jupyter-tensorflow-full:v1.8.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.15.1)	X	X
kc-kubeflow/jupyter-tensorflow-cuda-full:v1.8.0.py38.1a	python 3.8.10	JupyterLab	tensorflow(2.13.1)	O	X
kc-kubeflow/jupyter-tensorflow-cuda-full:v1.8.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.15.1)	O	X
kc-kubeflow/jupyter-pyspark-pytorch:v1.8.0.py38.1a	python 3.8.10	JupyterLab	pytorch(2.3.0)	X	O
kc-kubeflow/jupyter-pyspark-pytorch:v1.8.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.3.0)	X	O
kc-kubeflow/jupyter-pyspark-pytorch-cuda:v1.8.0.py38.1a	python 3.8.10	JupyterLab	pytorch(2.3.0)	O	O
kc-kubeflow/jupyter-pyspark-pytorch-cuda:v1.8.0.py311.1a	python 3.11.9	JupyterLab	pytorch(2.3.0)	O	O
kc-kubeflow/jupyter-pyspark-tensorflow:v1.8.0.py38.1a	python 3.8.10	JupyterLab	tensorflow(2.13.1)	X	O
kc-kubeflow/jupyter-pyspark-tensorflow:v1.8.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.15.1)	X	O
kc-kubeflow/jupyter-pyspark-tensorflow-cuda:v1.8.0.py38.1a	python 3.8.10	JupyterLab	tensorflow(2.13.1)	O	O
kc-kubeflow/jupyter-pyspark-tensorflow-cuda:v1.8.0.py311.1a	python 3.11.9	JupyterLab	tensorflow(2.15.1)	O	O
kc-kubeflow/codeserver-python:v1.8.0.py38.1a	python 3.8.10	CodeServer		X	X
kc-kubeflow/codeserver-python:v1.8.0.py311.1a	python 3.11.9	CodeServer		X	X

Create notebook instance

Create a notebook instance with desired specifications to build a computing environment for running machine learning code and processing data.

1. Access the Kubeflow dashboard.

   • For more information on accessing the Kubeflow dashboard, refer to the Access Kubeflow dashboard guide.

2. Select the Notebooks tab on the left.

3. Click the [New notebook] button at the top right.

4. On the New notebook screen, enter the required information and click [Launch] to create the notebook instance.

   • To create a notebook based on a specific image, refer to the guide below.

   Item	Category	Description
   Name	Name	Used to identify the notebook instance in the Kubeflow dashboard
   Platform type		Web development platform for the notebook instance Types: JupyterLab, VisualStudio Code, RStudio VisualStudio Code and RStudio may not function correctly if the domain is not connected to Kubeflow
   Custom notebook	Image	Specify a KakaoCloud-supported Kubeflow notebook image
   	Custom image	When checked, allows input of custom image - Example: bigdata-150.kr-central-2.kcr.dev/kc-kubeflow/jupyter-pytorch-full:v1.0.1.py38
   	Image pull policy	Policy for pulling image when the notebook instance starts
   CPU / RAM	Minimum CPU	Minimum CPU allocated to the notebook instance - Displayed as "Requested CPUs" in version 1.6
   	Minimum memory Gi	Minimum memory allocated to the notebook instance - Displayed as "Requested memory in Gi" in version 1.6
   	Maximum CPU	Maximum CPU allocated to the notebook instance - Displayed as "CPU limit" in version 1.6
   	Maximum memory Gi	Maximum memory allocated to the notebook instance - Displayed as "Memory limit in Gi" in version 1.6
   GPUs	Number of GPUs	Number of GPU resources used in the notebook instance
   	GPU vendor	Available GPU instance types
   Workspace volume		Volume mounted to the notebook instance's Home directory - Can create a new volume or use an existing one
   Data volume		Register additional volumes to be mounted to the notebook instance - Can create a new volume or use an existing one, and specify the mount path
   Configurations		Register PodDefault resources already defined in the current namespace
   Affinity / Tolerations	Affinity config	Specify the node where the notebook instance will run
   	Tolerations group	Specify tolerations to apply to the notebook instance - Available tolerations are defined in the ConfigMap inside the Kubeflow-installed KE cluster
   Miscellaneous settings	Enable shared memory	Enable workaround for shared memory required by frameworks like Torch - Mounts an empty directory volume to /dev/shm

caution

Notebook instance will not be created if Minimum CPU or Memory Gi exceeds the capacity of the node pool selected in Affinity config.

Create CPU-based notebook

Follow the steps below to create a notebook instance based on a CPU image:

1. Access the Kubeflow dashboard, click the Notebooks tab, and then click [New notebook].

2. On the New notebook screen, enter the required information and click [Launch] to create the notebook instance.

   Item	Category	Description
   Name	Name	Used to identify the notebook instance in the Kubeflow dashboard
   Platform type		Web development platform for the notebook instance Types: JupyterLab, VisualStudio Code, RStudio
   Custom notebook	Image	Select an image that matches the Kubeflow version
   	Custom image	When checked, allows input of custom image Example: bigdata-150.kr-central-2.kcr.dev/kc-kubeflow/jupyter-pytorch-full:v1.0.1.py38
   	Image pull policy	Policy for pulling image when the notebook instance starts
   CPU / RAM	Minimum CPU	2 - Minimum CPU allocated to the notebook instance
   	Minimum memory Gi	8 - Minimum memory allocated to the notebook instance
   	Maximum CPU	Maximum CPU allocated to the notebook instance
   	Maximum memory Gi	Maximum memory allocated to the notebook instance
   GPUs	Number of GPUs	None - No GPU resource used
   	GPU vendor	Available GPU instance types
   Workspace volume		Volume mounted to the notebook instance's Home directory - Can create a new volume or use an existing one
   Data volume		Register additional volumes to be mounted to the notebook instance - Can create a new volume or use an existing one, and specify the mount path
   Configurations		Register PodDefault resources already defined in the current namespace
   Affinity / Tolerations	Affinity config	Select a node created as a CPU instance type - Specify the node where the notebook instance will run
   	Tolerations group	None - Allow specific node taints
   Miscellaneous settings	Enable shared memory	Enable workaround for shared memory required by frameworks like Torch - Mounts an empty directory volume to /dev/shm

Create GPU-based notebook

Follow the steps below to create a notebook instance based on a GPU image:

1. Access the Kubeflow dashboard, click the Notebooks tab, and then click [New notebook].

2. On the New notebook screen, enter the required information and click [Launch] to create the notebook instance.

   Item	Category	Description
   Name	Name	Used to identify the notebook instance in the Kubeflow dashboard
   Platform type		Web development platform for the notebook instance Types: JupyterLab, VisualStudio Code, RStudio
   Custom notebook	Image	Select an image that matches the Kubeflow version
   	Custom image	When checked, allows input of custom image Example: bigdata-150.kr-central-2.kcr.dev/kc-kubeflow/jupyter-pytorch-full:v1.0.1.py38
   	Image pull policy	Policy for pulling image when the notebook instance starts
   CPU / RAM	Minimum CPU	2 - Minimum CPU allocated to the notebook instance
   	Minimum memory Gi	8 - Minimum memory allocated to the notebook instance
   	Maximum CPU	Maximum CPU allocated to the notebook instance
   	Maximum memory Gi	Maximum CPU allocated to the notebook instance
   GPUs	Number of GPUs	1 - Number of GPU resources used
   	GPU vendor	NVIDIA MIG - 1g.10gb - Available GPU instance types
   Workspace volume		Volume mounted to the notebook instance's Home directory - Can create a new volume or use an existing one
   Data volume		Register additional volumes to be mounted to the notebook instance - Can create a new volume or use an existing one, and specify the mount path
   Configurations		Register PodDefault resources already defined in the current namespace
   Affinity / Tolerations	Affinity config	Select a node created as a GPU instance type - Specify the node where the notebook instance will run
   	Tolerations group	None - Allow specific node taints
   Miscellaneous settings	Enable shared memory	Enable workaround for shared memory required by frameworks like Torch - Mounts an empty directory volume to /dev/shm

Access notebook instance

1. Access the Kubeflow dashboard.

   • For more information on accessing the Kubeflow dashboard, refer to the Access Kubeflow dashboard guide.

2. Select the Notebooks tab on the left.

3. In the list of created notebooks, click [Connect] for the desired notebook.

4. After clicking [Connect], a new tab will open showing the Jupyter Notebook interface for the selected notebook instance.

   • You can write code or run machine learning models in the opened notebook.

   
   Jupyter Notebook interface

Stop notebook instance

Notebook instances are typically stopped to save resources, enhance security, reduce costs, or manage the environment. Leaving a notebook instance running may prevent other tasks from accessing resources and lead to increased costs or security risks.

1. Access the Kubeflow dashboard.

   • For more information on accessing the Kubeflow dashboard, refer to the Access Kubeflow dashboard guide.

2. Select the Notebooks tab on the left.

3. In the list of notebooks, click the ◼ (stop) button for the notebook you wish to stop.

   • When the stop confirmation modal appears, click [Stop].

   
   Notebook stop confirmation

4. In the list of notebooks, verify that the status of the stopped notebook is correctly reflected.

   
   Notebook stop verification

Delete notebook instance

Notebook instances are typically deleted to reclaim resources, enhance security, reduce costs, or manage the environment. Once deleted, the resources associated with the notebook can be reallocated to other tasks.

1. Access the Kubeflow dashboard.

   • For more information on accessing the Kubeflow dashboard, refer to the Access Kubeflow dashboard guide.

2. Select the Notebooks tab on the left.

3. In the list of notebooks, click the delete (trash bin) icon for the notebook you wish to delete.

   • When the delete confirmation popup appears, click [Delete].

   
   Notebook delete confirmation

4. Confirm that the notebook is no longer listed in the notebook list.

   
   Notebook delete verification

caution

• Deleted notebook instance names cannot be reused.
• If you need to reuse a name, delete the {notebook-name}_volume PVC from the Kubernetes cluster in the KakaoCloud Kubernetes Engine service before creating it again.

info

• For more details, refer to the Kubeflow > Notebook official documentation.
• For more usage examples using Kubeflow services, visit Tutorial > Machine learning & AI category.