[[["容易理解","easyToUnderstand","thumb-up"],["確實解決了我的問題","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["難以理解","hardToUnderstand","thumb-down"],["資訊或程式碼範例有誤","incorrectInformationOrSampleCode","thumb-down"],["缺少我需要的資訊/範例","missingTheInformationSamplesINeed","thumb-down"],["翻譯問題","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["上次更新時間：2025-08-19 (世界標準時間)。"],[],[],null,["# Introduction to Vertex AI Workbench managed notebooks\n\nIntroduction to managed notebooks\n=================================\n\n\n| Vertex AI Workbench managed notebooks is\n| [deprecated](/vertex-ai/docs/deprecations). On\n| April 14, 2025, support for\n| managed notebooks will end and the ability to create managed notebooks instances\n| will be removed. Existing instances will continue to function\n| but patches, updates, and upgrades won't be available. To continue using\n| Vertex AI Workbench, we recommend that you\n| [migrate\n| your managed notebooks instances to Vertex AI Workbench instances](/vertex-ai/docs/workbench/managed/migrate-to-instances).\n\n\u003cbr /\u003e\n\nVertex AI Workbench managed notebooks instances\nare Google-managed environments\nwith integrations and capabilities that help you set up and work in\nan end-to-end Jupyter notebook-based production environment.\n\nManaged notebooks instances are prepackaged with\n[JupyterLab](https://jupyterlab.readthedocs.io/en/stable/getting_started/overview.html)\nand have a preinstalled suite of deep learning packages,\nincluding support for the TensorFlow and PyTorch\nframeworks. Managed notebooks instances support GPU accelerators and\nthe ability to sync with a\n[GitHub](https://github.com/) repository.\nYour managed notebooks instances are protected\nby Google Cloud authentication and authorization.\n\nGoogle-managed compute infrastructure\n-------------------------------------\n\nA Vertex AI Workbench managed notebooks instance\nis a Google-managed, Jupyter notebook-based, compute infrastructure.\n\nWhen you create a managed notebooks instance,\nit is deployed as a Google-managed virtual machine (VM) instance in a\n[tenant project](/service-infrastructure/docs/manage-tenant-projects).\n\nYour managed notebooks instance includes many common\ndata science framework environments, such as TensorFlow\nand PyTorch. You can also add your own custom container images to\nyour managed notebooks instance. These environments\nare available as kernels that you can run your\nnotebook file in.\n\nWhen you run a notebook in one of the kernels, Vertex AI Workbench\nstarts the corresponding container, creates a Jupyter session on it, and\nuses that Jupyter session to run your notebook on the container.\n\nThis Google-managed compute infrastructure includes integrations\nand capabilities that help you implement data science and machine learning\nworkflows from start to finish. See the following sections for details.\n\nUse custom containers\n---------------------\n\nYou can add custom Docker container images to\nyour managed notebooks instance\nto run your notebook code in an environment customized for your needs.\n\nThese custom containers are available to use directly from the\nJupyterLab user interface, alongside the preinstalled frameworks.\nFor more information, see [Add a custom container to\na managed notebooks instance](/vertex-ai/docs/workbench/managed/custom-container).\n\nNotebook-based workflow\n-----------------------\n\nManaged notebooks instances let you\nperform workflow-oriented tasks without leaving the JupyterLab user interface.\n\n### Control your hardware and framework from JupyterLab\n\nIn a managed notebooks instance, your JupyterLab user interface\nis where you specify what compute resources your code will run on. For example,\nyou can configure how many vCPUs or GPUs you want, how much RAM you want, and\nwhat framework you want to run the code in. You can write your code first, and\nthen choose how to run it without leaving JupyterLab\nor restarting your instance.\nFor quick tests of your code, you can scale your hardware down and then scale it\nback up to run your code against more data.\n\n### Access to data\n\nYou can access your data without leaving the JupyterLab user interface.\n\nIn JupyterLab's navigation menu on\na managed notebooks instance, you can use the\nCloud Storage integration\nto browse data and other files that you have access to.\nSee [Access Cloud Storage buckets and files\nfrom within JupyterLab](/vertex-ai/docs/workbench/managed/cloud-storage).\n\nYou can also use the\nBigQuery integration\nto browse tables that you have access to,\nwrite queries, preview results, and load data into your notebook.\nSee [Query data in BigQuery tables\nfrom within JupyterLab](/vertex-ai/docs/workbench/managed/bigquery).\n\nExecute notebook runs\n---------------------\n\nUse the executor to run a notebook file\nas a one-time execution or on a schedule.\nChoose the specific environment and hardware that you want\nyour execution to run on. Your notebook's code will run on\nVertex AI custom training, which can make it easier\nto do distributed training, optimize hyperparameters, or\nschedule continuous training jobs. See [Run notebook files\nwith the executor](/vertex-ai/docs/workbench/managed/executor).\n\nYou can [use parameters in\nyour execution](/vertex-ai/docs/workbench/managed/executor-parameters)\nto make specific changes to each run.\nFor example, you might specify a different dataset to use,\nchange the learning rate on your model, or change the version\nof the model.\n\nYou can also [set a notebook to run on a recurring\nschedule](/vertex-ai/docs/workbench/managed/quickstart-schedule-execution-console).\nEven while your instance is shut down, Vertex AI Workbench will\nrun your notebook file and save the results\nfor you to look at and share with others.\n\nShare insights\n--------------\n\nExecuted notebook runs are stored in a Cloud Storage bucket,\nso you can share your insights with others by granting access\nto the results. See the [previous section on executing\nnotebook runs](#executor).\n\nSecure your instance\n--------------------\n\nYou can deploy your managed notebooks instance\nwith the default Google-managed network,\nwhich uses a default VPC network and subnet.\nInstead of the default network, you can specify a\nVPC network to use with your instance.\nFor more information, see\n[Set up a network](/vertex-ai/docs/workbench/managed/networking). You can use\n[VPC Service Controls](/vpc-service-controls/docs/overview)\nto provide additional security for your\nmanaged notebooks instances.\n\nTo use managed notebooks within a service perimeter, see [Use\na managed notebooks instance within a service\nperimeter](/vertex-ai/docs/workbench/managed/service-perimeter).\n\nBy default, Google Cloud automatically [encrypts data when it is at\nrest](/security/encryption/default-encryption) using encryption keys\nmanaged by Google. If you have specific compliance or regulatory requirements\nrelated to the keys that protect your data, you can use customer-managed\nencryption keys (CMEK) with your managed notebooks instances.\nFor more information,\nsee [Use customer-managed encryption keys](/vertex-ai/docs/workbench/managed/cmek).\n\nAutomated shutdown for idle instances\n-------------------------------------\n\nTo help manage costs,\nmanaged notebooks instances\nshut down after being idle for a specific time period by default.\nYou can change the amount of time or turn this feature off.\nFor more information,\nsee [Idle shutdown](/vertex-ai/docs/workbench/managed/idle-shutdown).\n\nDataproc integration\n--------------------\n\nYou can process data quickly by running a notebook\non a Dataproc cluster.\nAfter your cluster is set up, you can run\na notebook file on it without leaving the JupyterLab user interface.\nFor more information, see [Run a managed notebooks instance\non a Dataproc cluster](/vertex-ai/docs/workbench/managed/dataproc).\n\nLimitations\n-----------\n\nConsider the following limitations of\nmanaged notebooks when planning your project:\n\n- Managed notebooks instances are Google-managed\n and therefore less customizable than Vertex AI Workbench\n user-managed notebooks instances.\n User-managed notebooks instances can be\n more ideal for users who need a lot of control over their environment.\n For more information, see [Introduction to\n user-managed notebooks](/vertex-ai/docs/workbench/user-managed/introduction).\n\n- Third party JupyterLab extensions are not supported.\n\n- The Dataproc JupyterLab plugin isn't supported for\n managed notebooks, but you can use the plugin in\n Vertex AI Workbench instances. See [Create a\n Dataproc-enabled\n instance](/vertex-ai/docs/workbench/instances/create-dataproc-enabled).\n\n- Managed notebooks instances do not allow users to\n have `sudo` access.\n\n- When you use [Access Context Manager](/access-context-manager/docs/create-basic-access-level#corporate-network-example)\n and [Chrome Enterprise Premium](/chrome-enterprise-premium/docs/access-levels)\n to protect managed notebooks instances with\n context-aware access controls, access is evaluated each time\n the user authenticates to the instance. For example, access\n is evaluated the first time the user accesses JupyterLab and\n whenever they access it thereafter if their web browser's\n cookie has expired.\n\n- To use accelerators with managed notebooks instances,\n the accelerator type that you want must be available in your instance's\n zone. To learn about accelerator availability by zone, see\n [GPU regions and zones availability](/compute/docs/gpus/gpu-regions-zones).\n\nWhat's next\n-----------\n\n- [Create a managed notebooks\n instance](/vertex-ai/docs/workbench/managed/create-instance).\n\n- Learn more about the [networking options available for your\n managed notebooks instance](/vertex-ai/docs/workbench/managed/networking)."]]
