alicloud_pai_workspace_model_version

Provides a PAI Workspace Model Version resource.

For information about PAI Workspace Model Version and how to use it, see What is Model Version.

-> NOTE: Available since v1.248.0.

Example Usage

Basic Usage

variable "name" {
  default = "terraform_example"
}

provider "alicloud" {
  region = "cn-hangzhou"
}

resource "random_integer" "default" {
  min = 10000
  max = 99999
}

resource "alicloud_pai_workspace_workspace" "defaultDI9fsL" {
  description    = "802"
  display_name   = var.name
  workspace_name = "${var.name}_${random_integer.default.result}"
  env_types      = ["prod"]
}

resource "alicloud_pai_workspace_model" "defaultsHptEL" {
  model_name        = var.name
  workspace_id      = alicloud_pai_workspace_workspace.defaultDI9fsL.id
  origin            = "Civitai"
  task              = "text-to-image-synthesis"
  accessibility     = "PRIVATE"
  model_type        = "Checkpoint"
  order_number      = "1"
  model_description = "ModelDescription."
  model_doc         = "https://eas-***.oss-cn-hangzhou.aliyuncs.com/s**.safetensors"
  domain            = "aigc"
  labels {
    key   = "base_model"
    value = "SD 1.5"
  }
  extra_info = {
    test = "15"
  }
}

resource "alicloud_pai_workspace_model_version" "default" {
  version_description = "VersionDescription."
  source_type         = "TrainingService"
  source_id           = "region=$${region},workspaceId=$${workspace_id},kind=TrainingJob,id=job-id"
  extra_info = {
    test = "15"
  }
  training_spec = {
    test = "TrainingSpec"
  }
  uri = "oss://hz-example-0701.oss-cn-hangzhou-internal.aliyuncs.com/checkpoints/"
  inference_spec = {
    test = "InferenceSpec"
  }
  model_id        = alicloud_pai_workspace_model.defaultsHptEL.id
  format_type     = "SavedModel"
  approval_status = "Pending"
  framework_type  = "PyTorch"
  version_name    = "1.0.0"
  metrics = {
  }
  labels {
    key   = "k1"
    value = "vs1"
  }
}

Argument Reference

The following arguments are supported:

• approval_status - (Optional) The approval status. Valid values:
  • Pending: To be determined.
  • Approved: Allow to go online.
  • Rejected: Online is not allowed.
• extra_info - (Optional, Map) Other information.
• format_type - (Optional, ForceNew) The format of the model. Valid values:
  • OfflineModel
  • SavedModel
  • Keras H5
  • Frozen Pb
  • Caffe Prototxt
  • TorchScript
  • XGBoost
  • PMML
  • AlinkModel
  • ONNX
• framework_type - (Optional, ForceNew) The framework of the model. Valid values:
  • Pytorch
  • XGBoost
  • Keras
  • Caffe
  • Alink
  • Xflow
  • TensorFlow
• inference_spec - (Optional, Map) Describes how to apply to downstream inference services.
• labels - (Optional, List) List of model version labels. See labels below.
• metrics - (Optional, Map) The metrics for the model. The serialized length is limited to 8192.
• model_id - (Required, ForceNew) The model ID.
• options - (Optional) The extended field. This is a JSON string.
• source_id - (Optional) The source ID.
• source_type - (Optional) The type of the model source. Valid values:
  • Custom: Custom.
  • PAIFlow:PAI workflow.
  • TrainingService:PAI training service.
• training_spec - (Optional, Map) The training configurations. Used for fine-tuning and incremental training.
• uri - (Required, ForceNew) The URI of the model version.
• version_description - (Optional) The version descriptions.
• version_name - (Optional, ForceNew, Computed) The Model version.

labels

The labels supports the following:

• key - (Optional) label key.
• value - (Optional) label value.

Attributes Reference

The following attributes are exported:

• id - The ID of the resource supplied above.The value is formulated as <model_id>:<version_name>.

Timeouts

The timeouts block allows you to specify timeouts for certain actions:

• create - (Defaults to 5 mins) Used when create the Model Version.
• delete - (Defaults to 5 mins) Used when delete the Model Version.
• update - (Defaults to 5 mins) Used when update the Model Version.

Import

PAI Workspace Model Version can be imported using the id, e.g.

$ terraform import alicloud_pai_workspace_model_version.example <model_id>:<version_name>