Custom Models

IAM Policy

You need to create a policy to grant OCI Search with OpenSearch access to the Object Storage bucket you upload the custom model to. The following policy example includes the required permissions:

ALLOW ANY-USER to manage object-family in tenancy WHERE ALL {request.principal.type='opensearchcluster', request.resource.compartment.id='<cluster_compartment_id>'}

Regions for Generative AI Connectors

To use OCI Generative AI, your tenancy must be subscribed to the US Midwest (Chicago) region or the Germany Central (Frankfurt) region. You don't need to create the cluster in either of those regions, just ensure that your tenancy is subscribed to one of the regions.

Configure Cluster Settings

Use the settings operation of the Cluster APIs to configure the recommended cluster settings for semantic search. The following example includes the recommended settings:

PUT _cluster/settings
{
  "persistent": {
    "plugins": {
      "ml_commons": {
        "only_run_on_ml_node": "false",
        "model_access_control_enabled": "true",
        "native_memory_threshold": "99",
        "rag_pipeline_feature_enabled": "true",
        "memory_feature_enabled": "true",
        "allow_registering_model_via_local_file": "true",
        "allow_registering_model_via_url": "true",
        "model_auto_redeploy.enable":"true",
        "model_auto_redeploy.lifetime_retry_times": 10
      }
    }
  }
}

Upload Model to Object Storage Bucket

To make a custom model available to register for a cluster, you need to upload the model to an Object Storage bucket in the tenancy. If you don't have an existing Object Storage bucket, you need to create the bucket. For a tutorial that walks you through how to create a bucket, see Creating a Bucket.

Next, you need to upload the custom model to the bucket, see Uploading Files to a Bucket for a tutorial that walks you through how to upload files to a bucket. For the purposes of this walkthrough, you can download any supported hugging face model to upload.

To get the URL of the uploaded model file

1. Open the navigation menu and click Storage. Under Object Storage & Archive Storage, click Buckets.

2. Click the bucket that contains the uploaded model. The bucket's Details page appears.

3. Click the Actions menu () next to the object name, and then select View Object Details. The Object Details dialog box appears.

4. The URL to access the model file is displayed in the URL Path (URI) field. Copy the URL to use in the next step when you register the custom model.

   Important
   
   You might see a warning message indicating that the current URL in the URL Path (URI) field is deprecated, with a new URL specified in the warning message. If you see this warning message, use the new URL in the warning message instead to register the custom model.

Register the custom model

Use the register operation to register the custom model. In the following example, the custom model uploaded to the Object Storage bucket is the huggingface/sentence-transformers/all-MiniLM-L12-v2 model. The values specified in model_config for this example are from the model's config file. Ensure that you're using the applicable model configuration values for the custom model you're registering.

Specify the Object Storage URL in the actions section, this is an OCI Search with OpenSearch API added to support the BYOM scenario.

POST /_plugins/_ml/models/_register
{
  "model_group_id": "<Model_Group_ID>",
	"name": "sentence-transformers/all-MiniLM-L12-v2",
	"version": "1.0.1",
	"description": "This is a sentence-transformers model: It maps sentences & paragraphs to a 384 dimensional dense vector space and can be used for tasks like clustering or semantic search.",
	"model_task_type": "TEXT_EMBEDDING",
	"model_format": "TORCH_SCRIPT",
	"model_content_size_in_bytes": 134568911,
	"model_content_hash_value": "f8012a4e6b5da1f556221a12160d080157039f077ab85a5f6b467a47247aad49",
	"model_config": {
		"model_type": "bert",
		"embedding_dimension": 384,
		"framework_type": "sentence_transformers",
		"all_config": "{\"_name_or_path\":\"microsoft/MiniLM-L12-H384-uncased\",\"attention_probs_dropout_prob\":0.1,\"gradient_checkpointing\":false,\"hidden_act\":\"gelu\",\"hidden_dropout_prob\":0.1,\"hidden_size\":384,\"initializer_range\":0.02,\"intermediate_size\":1536,\"layer_norm_eps\":1e-12,\"max_position_embeddings\":512,\"model_type\":\"bert\",\"num_attention_heads\":12,\"num_hidden_layers\":12,\"pad_token_id\":0,\"position_embedding_type\":\"absolute\",\"transformers_version\":\"4.8.2\",\"type_vocab_size\":2,\"use_cache\":true,\"vocab_size\":30522}"
	},
    "url_connector": {
        "protocol": "oci_sigv1",
        "parameters": {
            "auth_type": "resource_principal"
        },
        "actions": [
            {
                "method": "GET",
                "action_type": "DOWNLOAD",
                "url": "<Object_Storage_URL_Path>"
            }
        ]
    }
}

https://<tenancy_name>.objectstorage.us-ashburn-1.oraclecloud.com/n/<tenancy_name>/b/<bucket_name>/o/sentence-transformers_all-distilroberta-v1-1.0.1-torch_script.zip

Make note of the task_id returned in the response, you can use the task_id to check the status of the operation.

For example, from the following response:

{
  "task_id": "<task_ID>",
  "status": "CREATED"
}

Track the register task and get the model ID

GET /_plugins/_ml/tasks/<task_ID>

When the register operation is complete, the status value in the response to the Get operation is COMPLETED, as shown the following example:

{
  "model_id": "<embedding_model_ID>",
  "task_type": "REGISTER_MODEL",
  "function_name": "TEXT_EMBEDDING",
  "state": "COMPLETED",
  "worker_node": [
    "3qSqVfK2RvGJv1URKfS1bw"
  ],
  "create_time": 1706829732915,
  "last_update_time": 1706829780094,
  "is_async": true
}

Make note of the model_id value returned in the response to use when you deploy the model.