Developed by the Kubeflow community, TF-Operator is an add-on used to help deploy and execute TensorFlow distributed training jobs in a Kubernetes cluster.
After deployment, you can create, view, and delete TF jobs.
Kubernetes cluster (v1.16 or later)
During Helm deployment, all configuration items are included in values.yaml
.
Some fields may need to be customized, as listed below:
Parameter | Description | Default Value |
---|---|---|
image.repository |
The repository where the TF-Operator image resides | ccr.ccs.tencentyun.com/kubeflow-oteam/tf-operator |
image.tag |
TF-Operator image version | "latest" |
namespace.create |
Whether to create a separate namespace for TF-Operator | true |
namespace.name |
The namespace where TF-Operator is to be deployed | "tf-operator" |
Apakah halaman ini membantu?