Help Center> RES> API Reference> APIs (Old Version)> Online Service-related APIs> Deploying Inference Services
View PDF

Deploying Inference Services

Function

This API is used to deploy inference services.

URI

POST /v1/{project_id}/infer-services

Table 1 describes the URI parameters.

Table 1 URI parameters

Parameter

Mandatory

Type

Description

project_id

Yes

String

Project ID, which is used for resource isolation. For details about how to obtain the project ID, see Obtaining a Project ID.

Request

Table 2 describes the request parameters.

Table 2 Request parameters

Parameter

Mandatory

Type

Description

service_name

Yes

String

Service name. The value can contain a maximum of 48 characters. Only digits, letters, underscores (_), and hyphens (-) are allowed. It must start with res- and cannot end with a hyphen.

service_type

Yes

String

Service type. The options are infer, rank, or nlp.

description

No

String

Service description

flows

Yes

List

List of flows. For details, see Table 3.
Table 3 flows parameters

Parameter

Mandatory

Type

Description

name

Yes

String

Flow name. The value can contain a maximum of 60 characters. Only digits, letters, underscores (_), hyphens (-), and spaces are allowed.

rules

Yes

List

Configuration of different traffic and candidate set rules for each flow. For details, see Table 4.

backup_uuid

No

String

UUID generated when the retrieval strategy is manually imported

need_preferences

No

Boolean

Whether to enable the Preference function. This parameter must match global_features_info_path, profile_uuid, and attr_weight. For example, in a car purchase scenario, a candidate set generated by the ItemCF Recommendation strategy functions as items2; a collection of items suitable for a user functions as items1. To ensure that the recommended items match the user's age, gender, and other attributes as much as possible, select items that appear in items2 for the final candidate set generation. If the selected quantity is insufficient, choose from the remaining items to fill the vacancies.

global_features_info_path

Yes

String

Path storing the global feature files

profile_uuid

No

String

UUIDs generated by the User Profile-Item Profile-Standard Wide Table Generation operator

  • This parameter is mandatory when filter job is performed on candidate sets for online prediction

attr_weight

No

List

Parameter used for attribute weight ranking or preferred item search. Table 11 describes attribute weight lists.

filter_conf

No

Object

List of filter rules. For details, see Table 6.

rank_conf

No

Object

List of ranking configurations. For details, see Table 9.

config

Yes

JSON

Configuration parameter. For details, see Table 5.
Table 4 rules parameters

Parameter

Mandatory

Type

Description

candidate_uuid

Yes

String

Candidate set ID

rule_ratio

Yes

Integer

Proportion of data with the same priority. The value is an integer ranging from 0 to 100. The proportion of data with the same level is 100.

priority

Yes

String

Priority. The value is an integer ranging from 1 to 10. The value 1 indicates the highest priority. The value can contain a maximum of 32 characters.
Table 5 config parameters

Parameter

Mandatory

Type

Description

model_name

Yes

String

Model name

model_version

Yes

String

Model version

weight

Yes

Integer

A/B test traffic percentage. The value is an integer ranging from 0 to 100. The sum of all flows is 100.

specification

Yes

String

Node specification. Currently, only c2.m8.g0 is supported.

instance_count

Yes

Integer

Number of node instances. A maximum of two instances are supported.
Table 6 filter_conf parameters

Parameter

Mandatory

Type

Description

filter_uuid

No

String

UUID generated by using filter rules

  • This parameter is mandatory when attribute filter is performed on candidate sets for online prediction

duplicate_info

No

Objcet

Duplication configurations. For details, see Table 7.

attribute_rules

No

List

Attribute filter configurations. For details, see Table 8.
Table 7 duplicate_info parameters

Parameter

Mandatory

Type

Description

item_properties

No

List

A list of attributes to be deduplicated. For example: If product_color is entered, the items with the same color in the candidate set are deduplicated. Items left in the candidate set have a high retrieval score.

item_id_ignore_length

No

String

Length of an item ID to be ignored. For example, if SKU_A000123 and SKU_A000456 are in a candidate set, and the ignored length is set to 3, then the SKU_A000 with a higher retrieval score is left.
Table 8 attribute_rules parameters

Parameter

Mandatory

Type

Description

user_attributes

Yes

List

User attribute to be filtered. Currently, only attributes of the string type and numerical type are supported. For example, for a user from tier-1 city, the attribute name is city, and the attribute value is tier-1 city.

item_attributes

Yes

List

Item attribute to be filtered. Currently, only attributes of the string type and numerical type are supported. For example, For an item whose weight is 0, the attribute name is weight and the attribute value is 0.00.
Table 9 rank_conf parameters

Parameter

Mandatory

Type

Description

model_path

No

String

OBS file path where models generated by sorting policy are located

etl_uuid

No

String

UUIDs of the ranking training samples generated in Feature Engineering

is_attrWeight_rank

No

Boolean

Whether to perform attribute ranking
Table 10 config parameters

Parameter

Mandatory

Type

Description

weight

Yes

Integer

Traffic proportion of the current model version

specification

Yes

String

Specification of the compute node

instance_count

Yes

Integer

Number of compute nodes, or number of instances

image_id

No

String

Image ID. image_id and model_id are mutually exclusive. You can select only one of them.

model_id

Yes

String

Model ID. image_id and model_id are mutually exclusive. You can select only one of them.
Table 11 attr_weight parameters

Parameter

Mandatory

Type

Description

attribute

Yes

String

Attribute used in attribute weight ranking, such as age.

weight

Yes

Double

Weight used in attribute weight ranking, such as 0.8.

Response

Table 12 describes the response parameters.

Table 12 Response parameters

Parameter

Mandatory

Type

Description

is_success

Yes

Boolean

Whether the request is successful

service_id

Yes

String

Service ID

Example

  • Example request 
    {
    "service_name": "res-service",
    "service_type": "infer",
    "description": "",
    "flows": [
        {
            "name": "lc",
            "ratio": 100,
            "rules": [
                {
                    "rule_ratio": 100,
                    "priority": "1",
                    "candidate_uuid": "d7f23ad959bb409a89ec53b0e7346982"
                }
            ],
            "backup_uuid": "d7djdad959bb409a89ec53b0e7346982",
            "global_features_info_path": "<Path for storing the global feature files>",
            "profile_uuid": "593f1466661e40f58a7bda615b588d2e",
            "filter_conf": {
                "filter_uuid": "ca2b02b83338498da3b551b66d9654c3",
                "duplicate_info": {
                    "item_properties": [
                        "product_name"
                    ],
                    "item_id_ignore_length": ""
                },
                "attribute_rules": [
                    {
                        "user_attributes": [
                            {
                                "name": "gender",
                                "value": "male"
                            }
                        ],
                        "item_attributes": [
                            {
                                "name": "color",
                                "value": "red"
                            }
                        ]
                    }
                ]
            },
            "rank_conf": {
                "model_path": "<OBS path for storing the models generated by the ranking strategies>"
            },
            "config": {
		"weight": 100,
		"specification": "modelarts.vm.cpu.2u",
		"instance_count": 1,
		"model_id": "5bbd6911-02d8-4109-80a1-e3dafc5a2cec"
	    }
        }
    ]
}
  • Example of a successful response 
    {
    "is_success": true,
    "service_id": "bf024976-5ade-49bf-83bf-5600191922db"
}
  • Example of a failed response 
    { 
    "is_success": false, 
    "error_code": "res.1001", 
    "error_msg": "The value of request parameter(service_name) is invalid." 
}

Status Code

For details about status codes, see Status Codes.