Creating the Model entry in Cloudera AI Registry in air-gapped environment

The example outlines how to create the Model entry in Cloudera AI Registry within an air-gapped environment.

AWS
Azure

curl -k https://$MODELREGISTRYDOMAIN/api/v2/models -X POST -H 'Content-Type: application/json' -H "Authorization: Bearer $TOKEN" --data-raw '
    { "name": "llama3-instruct-70b",
    "createModelVersionRequestPayload": {
    "metadata": {
    "model_repo_type": "NGC"
    },
    "downloadModelRepoRequest": {
    "source": "REMOTE",
    "remoteObjectStoragePath": "s3://bucket1/secured-models",
    "repo_id": "nim/meta/llama-3_1-70b-instruct:0.11.1+14957bf8-h100x4-fp8-throughput.1.2.18099809"
    }
    }
    }'

curl -k https://$MODELREGISTRYDOMAIN/api/v2/models -X POST -H 'Content-Type: application/json' -H "Authorization: Bearer $TOKEN" --data-raw '
{ "name": "llama3-instruct-70b",
	"createModelVersionRequestPayload": {
    	"metadata": {
        	"model_repo_type": "NGC"
    	},
    	"downloadModelRepoRequest": {
        	"source": "REMOTE",
		"remoteObjectStoragePath": "abfs://data@datalakeaccount.dfs.core.windows.net/modelregistry/secured-models",
        	"repo_id": "nim/meta/llama-3_1-70b-instruct:0.11.1+14957bf8-h100x4-fp8-throughput.1.2.18099809"
    	}
	}
}'

These curl requests create a new model named llama3-instruct-70b in Cloudera AI Registry, and its initial version.

These requests also trigger the copying of model artifacts from your uploaded, secured model location (specified by remoteObjectStoragePath in your cloud-specific object store) to a preferred Cloudera AI Registry object store location.