本頁面說明如何建立已預先安裝 TensorFlow 和其他工具的 TensorFlow 深度學習 VM 映像檔執行個體。您可以在 Google Cloud 主控台中透過 Cloud Marketplace 建立 TensorFlow 執行個體,也可以使用指令列建立。
事前準備
Sign in to your Google Cloud account. If you're new to
Google Cloud,
create an account to evaluate how our products perform in
real-world scenarios. New customers also get $300 in free credits to
run, test, and deploy workloads.
In the Google Cloud console, on the project selector page,
select or create a Google Cloud project.
[[["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:2025-09-04 (世界標準時間)。"],[[["\u003cp\u003eThis page guides users on creating a TensorFlow Deep Learning VM instance with pre-installed TensorFlow and other tools, available through the Google Cloud Marketplace or the command line.\u003c/p\u003e\n"],["\u003cp\u003eUsers can customize their VM instance by selecting machine type, zone, GPU type and number, TensorFlow framework version, boot disk specifications, and networking settings.\u003c/p\u003e\n"],["\u003cp\u003eNVIDIA GPU drivers can be automatically installed upon the first startup if using GPUs, or it can be done manually, though automatic installation will require a few minutes to complete.\u003c/p\u003e\n"],["\u003cp\u003eThe command line method offers flexibility to create instances with or without GPUs and specifies required parameters such as image family, zone, and project, with specific options for GPU and driver setup.\u003c/p\u003e\n"],["\u003cp\u003eUsers have the option to create preemptible instances, which are more cost-effective but subject to termination by Compute Engine, by using the \u003ccode\u003e--preemptible\u003c/code\u003e command in the creation process.\u003c/p\u003e\n"]]],[],null,["# Create a TensorFlow Deep Learning VM instance\n\nThis page shows you how to create\na TensorFlow Deep Learning VM Images instance\nwith TensorFlow and other tools pre-installed. You can create\na TensorFlow instance from Cloud Marketplace within\nthe Google Cloud console or using the command line.\n\nBefore you begin\n----------------\n\n- Sign in to your Google Cloud account. If you're new to Google Cloud, [create an account](https://console.cloud.google.com/freetrial) to evaluate how our products perform in real-world scenarios. New customers also get $300 in free credits to run, test, and deploy workloads.\n- In the Google Cloud console, on the project selector page,\n select or create a Google Cloud project.\n\n | **Note**: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.\n\n [Go to project selector](https://console.cloud.google.com/projectselector2/home/dashboard)\n-\n [Verify that billing is enabled for your Google Cloud project](/billing/docs/how-to/verify-billing-enabled#confirm_billing_is_enabled_on_a_project).\n\n- In the Google Cloud console, on the project selector page,\n select or create a Google Cloud project.\n\n | **Note**: If you don't plan to keep the resources that you create in this procedure, create a project instead of selecting an existing project. After you finish these steps, you can delete the project, removing all resources associated with the project.\n\n [Go to project selector](https://console.cloud.google.com/projectselector2/home/dashboard)\n-\n [Verify that billing is enabled for your Google Cloud project](/billing/docs/how-to/verify-billing-enabled#confirm_billing_is_enabled_on_a_project).\n\n1. If you are using GPUs with your Deep Learning VM, check the [quotas page](https://console.cloud.google.com/quotas) to ensure that you have enough GPUs available in your project. If GPUs are not listed on the quotas page or you require additional GPU quota, [request a\n quota increase](/compute/quotas#requesting_additional_quota).\n\nCreating a TensorFlow Deep Learning VM instance from the Cloud Marketplace\n--------------------------------------------------------------------------\n\nTo create a TensorFlow Deep Learning VM instance\nfrom the Cloud Marketplace, complete the following steps:\n\n1. Go to the Deep Learning VM Cloud Marketplace page in\n the Google Cloud console.\n\n [Go to the Deep Learning VM Cloud Marketplace page](https://console.cloud.google.com/marketplace/details/click-to-deploy-images/deeplearning)\n2. Click **Get started**.\n\n3. Enter a **Deployment name** , which will be the root of your VM name.\n Compute Engine appends `-vm` to this name when naming your instance.\n\n4. Select a **Zone**.\n\n5. Under **Machine type** , select the specifications that you\n want for your VM.\n [Learn more about machine types.](/compute/docs/machine-types)\n\n6. Under **GPUs** , select the **GPU type** and **Number of GPUs** .\n If you don't want to use GPUs,\n click the **Delete GPU** button\n and skip to step 7. [Learn more about GPUs.](/gpu)\n\n 1. Select a **GPU type** . Not all GPU types are available in all zones. [Find a combination that is supported.](/compute/docs/gpus)\n 2. Select the **Number of GPUs** . Each GPU supports different numbers of GPUs. [Find a combination that is supported.](/compute/docs/gpus)\n7. Under **Framework**, select one of the TensorFlow\n framework versions.\n\n8. If you're using GPUs, an NVIDIA driver is required.\n You can install the driver\n yourself, or select **Install NVIDIA GPU driver automatically\n on first startup**.\n\n9. You have the option to select **Enable access to JupyterLab via URL\n instead of SSH (Beta)**. Enabling this Beta feature lets you\n access your JupyterLab\n instance using a URL. Anyone who is in the Editor or Owner role in your\n Google Cloud project can access this URL.\n Currently, this feature only works in\n the United States, the European Union, and Asia.\n\n10. Select a boot disk type and boot disk size.\n\n11. Select the networking settings that you want.\n\n12. Click **Deploy**.\n\nIf you choose to install NVIDIA drivers, allow 3-5 minutes for installation\nto complete.\n\nAfter the VM is deployed, the page updates with instructions for\naccessing the instance.\n\nCreating a TensorFlow Deep Learning VM instance from the command line\n---------------------------------------------------------------------\n\nTo use the Google Cloud CLI to create\na new Deep Learning VM instance,\nyou must first install and initialize the [Google Cloud CLI](/sdk/docs):\n\n1. Download and install the Google Cloud CLI using the instructions given on [Installing Google Cloud CLI](/sdk/downloads).\n2. Initialize the SDK using the instructions given on [Initializing Cloud\n SDK](/sdk/docs/initializing).\n\nTo use `gcloud` in Cloud Shell, first activate Cloud Shell using the\ninstructions given on [Starting Cloud Shell](/shell/docs/starting-cloud-shell).\n\nYou can create a TensorFlow instance with or without GPUs. \n\n### Without GPUs\n\nTo provision a Deep Learning VM instance without a GPU: \n\n export IMAGE_FAMILY=\"tf-ent-latest-cpu\"\n export ZONE=\"us-west1-b\"\n export INSTANCE_NAME=\"my-instance\"\n\n gcloud compute instances create $INSTANCE_NAME \\\n --zone=$ZONE \\\n --image-family=$IMAGE_FAMILY \\\n --image-project=deeplearning-platform-release\n\nOptions:\n\n- `--image-family` must be one of the following:\n - `tf-ent-latest-cpu` to get the latest [TensorFlow Enterprise](/tensorflow-enterprise/docs) 2 image\n - An earlier TensorFlow or TensorFlow Enterprise image family name (see [Choosing an image](/deep-learning-vm/docs/images))\n- `--image-project` must be `deeplearning-platform-release`.\n\n### With one or more GPUs\n\nCompute Engine offers the option of adding one or more\nGPUs to your virtual machine instances. GPUs offer faster processing\nfor many complex data and machine learning tasks. To learn more about\nGPUs, see [GPUs on Compute Engine](/compute/docs/gpus).\n\nTo provision a Deep Learning VM instance with one or more GPUs: \n\n export IMAGE_FAMILY=\"tf-ent-latest-gpu\"\n export ZONE=\"us-west1-b\"\n export INSTANCE_NAME=\"my-instance\"\n\n gcloud compute instances create $INSTANCE_NAME \\\n --zone=$ZONE \\\n --image-family=$IMAGE_FAMILY \\\n --image-project=deeplearning-platform-release \\\n --maintenance-policy=TERMINATE \\\n --accelerator=\"type=nvidia-tesla-v100,count=1\" \\\n --metadata=\"install-nvidia-driver=True\"\n\nOptions:\n\n- `--image-family` must be one of the following:\n\n - `tf-ent-latest-gpu` to get the latest [TensorFlow Enterprise](/tensorflow-enterprise/docs) 2 image\n - An earlier TensorFlow or TensorFlow Enterprise image family name (see [Choosing an image](/deep-learning-vm/docs/images))\n- `--image-project` must be `deeplearning-platform-release`.\n\n- `--maintenance-policy` must be `TERMINATE`. To learn more, see\n [GPU Restrictions](/compute/docs/gpus#restrictions).\n\n- `--accelerator` specifies the GPU type to use. Must be\n specified in the format\n `--accelerator=\"type=`\u003cvar translate=\"no\"\u003eTYPE\u003c/var\u003e`,count=`\u003cvar translate=\"no\"\u003eCOUNT\u003c/var\u003e`\"`.\n For example, `--accelerator=\"type=nvidia-tesla-p100,count=2\"`.\n See the [GPU models\n table](/compute/docs/gpus#other_available_nvidia_gpu_models)\n for a list of available GPU types and counts.\n\n Not all GPU types are supported in all regions. For details, see\n [GPU regions and zones availability](/compute/docs/gpus/gpu-regions-zones).\n- `--metadata` is used to specify that the NVIDIA driver should\n be installed on your behalf. The value is `install-nvidia-driver=True`.\n If specified, Compute Engine loads the latest stable\n driver on the first boot and performs the necessary steps (including\n a final reboot to activate the driver).\n\nIf you've elected to install NVIDIA drivers, allow 3-5 minutes\nfor installation to complete.\n\nIt may take up to 5 minutes before your VM is fully provisioned. In this\ntime, you will be unable to SSH into your machine. When the installation is\ncomplete, to guarantee that the driver installation was successful, you can\nSSH in and run `nvidia-smi`.\n\nWhen you've configured your image, you can save a snapshot of your\nimage so that you can start derivitave instances without having to wait\nfor the driver installation.\n\n### About TensorFlow Enterprise\n\n[TensorFlow Enterprise](/tensorflow-enterprise/docs) is a\ndistribution of\n[TensorFlow](https://www.tensorflow.org/)\nthat has been optimized to run on Google Cloud and includes\n[Long Term Version\nSupport](/tensorflow-enterprise/docs/overview#long_term_version_support).\n\nCreating a preemptible instance\n-------------------------------\n\nYou can create a preemptible Deep Learning VM instance. A preemptible\ninstance is an instance you can create and run at a much lower price than\nnormal instances. However, Compute Engine might stop (preempt) these\ninstances if it requires access to those resources for other tasks.\nPreemptible instances always stop after 24 hours. To learn more about\npreemptible instances, see [Preemptible VM\nInstances](/compute/docs/instances/preemptible).\n\nTo create a preemptible Deep Learning VM instance:\n\n- Follow the instructions located above to create a new instance using the\n command line. To the `gcloud compute instances create` command, append the\n following:\n\n ```\n --preemptible\n ```\n\nWhat's next\n-----------\n\nFor instructions on connecting to your new Deep Learning VM instance\nthrough the Google Cloud console or command line, see [Connecting to\nInstances](/compute/docs/instances/connecting-to-instance). Your instance name\nis the **Deployment name** you specified with `-vm` appended."]]