CmdID / az / az cognitiveservices / az cognitiveservices account / az cognitiveservices account adapter-deployment

az cognitiveservices account adapter-deployment create

Create (or replace / re-target) a LoRA adapter deployment attached to a managed compute deployment. Azure CLI reference

The adapter must reference a protected Foundry fine-tuning LoRA output as its source model. The target deployment is the name of an existing managed compute deployment in the same account. Re-issuing create with the same adapter name and a different --target-deployment-name re-targets the adapter; --source-model-id is immutable.

1 example 2 options Its page in the Azure CLI reference

Examples

from the Azure CLI, each command read by CmdID

Create an adapter deployment attached to an existing managed compute deployment.

az cognitiveservices account adapter-deployment create -g myResourceGroup -n myAccount --adapter-deployment-name my-lora-adapter --source-model-id "azureai://accounts/myAccount/projects/myProject/models/my-fine-tuned-lora/versions/1" --target-deployment-name gpt-oss-120b-gpu

Identify

-gName of resource group.

-ncognitive service account name

--adapter-deployment-nameAdapter deployment name.

--source-model-idImmutable Foundry fine-tuning LoRA output model URI (e.g., azureai://accounts/{account}/projects/{project}/models/{model}/versions/{version}).

--target-deployment-nameName of the existing managed compute parent deployment (resolved within the same account) that will serve this adapter.

Options

2 in the Azure CLI reference

--source-model-idImmutable Foundry fine-tuning LoRA output model URI (e.g., azureai://accounts/{account}/projects/{project}/models/{model}/versions/{version}).

--target-deployment-nameName of the existing managed compute parent deployment (resolved within the same account) that will serve this adapter.