Creating a Version of a Training Job

Function

This API is used to create a version of a training job.

Calling this API is an asynchronous operation. The job status can be obtained by calling the APIs described in Querying a Training Job List and Querying the Details About a Training Job Version.

URI

POST /v1/{project_id}/training-jobs/{job_id}/versions

Table 1 describes the required parameters.

**Table 1** Parameters
Parameter	Mandatory	Type	Description
project_id	Yes	String	Project ID. For details about how to obtain a project ID, see Obtaining a Project ID and Name.
job_id	Yes	Long	ID of a training job

Request Body

Table 2 describes the request parameters.

**Table 2** Request parameters
Parameter	Mandatory	Type	Description
job_desc	No	String	Description of a training job. The value must contain 0 to 256 characters. By default, this parameter is left blank.
config	Yes	Object	Parameters for creating a training job For details, see Table 3.

**Table 3** **config** parameters
Parameter	Mandatory	Type	Description
worker_server_num	Yes	Integer	Number of workers in a training job. Obtain the maximum value from Querying Job Resource Specifications.
app_url	Yes	String	Code directory of a training job, for example, /usr/app/. This parameter must be used together with boot_file_url. After setting model_id, you do not need to set app_url or boot_file_url, and engine_id.
boot_file_url	Yes	String	Boot file of a training job, which needs to be stored in the code directory. Example value: /usr/app/boot.py This parameter must be used together with app_url. After setting model_id, you do not need to set app_url or boot_file_url, and engine_id.
parameter	No	Array<Object>	Running parameters of a training job. It is a collection of label-value pairs. For details, see the sample request. This parameter is a container environment variable when a training job uses a custom image. For details, see Table 5.
data_url	No	String	OBS URL of the dataset required by a training job. By default, this parameter is left blank. For example, /usr/data/. This parameter cannot be used together with data_source or dataset_id and dataset_version_id. However, one of the parameters must exist.
dataset_id	No	String	Dataset ID of a training job. This parameter must be used together with dataset_version_id, but cannot be used together with data_url or data_source.
dataset_version_id	No	String	Dataset version ID of a training job. This parameter must be used together with dataset_id, but cannot be used together with data_url or data_source.
data_source	No	JSON Array	Dataset of a training job. This parameter cannot be used with data_url, dataset_id, or dataset_version_id. For details, see Table 4.
spec_id	Yes	Long	ID of the resource specifications selected for a training job. Obtain the ID by calling the API described in Querying Job Resource Specifications. When creating a public pool job, ensure that spec_id is mandatory and cannot be used with pool_id.
pool_id	Yes	String	ID of a dedicated resource pool. To obtain the ID, do as follows: Log in to the ModelArts management console, choose Dedicated Resource Pools in the navigation pane on the left, and view the resource pool ID in the dedicated resource pool list. When creating a dedicated pool job, ensure that pool_id is mandatory and cannot be used with spec_id.
engine_id	Yes	Long	ID of the engine selected for a training job. The default value is 1. After setting model_id, you do not need to set app_url or boot_file_url, and engine_id. Obtain the ID by calling the API described in Querying Job Engine Specifications.
model_id	Yes	Long	ID of the built-in model of a training job. Obtain model_id by calling the API described in Querying a Built-in Algorithm. After setting model_id, you do not need to set app_url or boot_file_url, and engine_id.
train_url	Yes	String	OBS URL of the output file of a training job. By default, this parameter is left blank. Example value: /bucket/trainUrl/
log_url	No	String	OBS URL of the logs of a training job. By default, this parameter is left blank. Example value: /usr/train/
pre_version_id	Yes	Long	ID of the previous version of a training job. You can obtain the value of version_id by calling the API described in Obtaining Training Job Versions.
user_image_url	No	String	SWR URL of a custom image used by a training job. Example value: 100.125.5.235:20202/jobmng/custom-cpu-base:1.0
user_command	No	String	Boot command used to start the container of a custom image of a training job. The format is bash /home/work/run_train.sh python /home/work/user-job-dir/app/train.py {python_file_parameter}.

**Table 4** **data_source** parameters
Parameter	Mandatory	Type	Description
dataset_id	No	String	Dataset ID of a training job. This parameter must be used together with dataset_version_id, but cannot be used together with data_url.
dataset_version	No	String	Dataset version ID of a training job. This parameter must be used together with dataset_id, but cannot be used together with data_url.
type	No	String	Dataset type. The value can be obs or dataset. obs and dataset cannot be used at the same time.
data_url	No	String	OBS bucket path. This parameter cannot be used together with dataset_id or dataset_version.

**Table 5** **parameter** parameters
Parameter	Mandatory	Type	Description
label	No	String	Parameter name
value	No	String	Parameter value

Response Body

Table 6 describes the response parameters.

**Table 6** Parameters
Parameter	Type	Description
is_success	Boolean	Whether the request is successful
error_message	String	Error message of a failed API call. This parameter is not included when the API call succeeds.
error_code	String	Error code of a failed API call. For details, see Error Codes. This parameter is not included when the API call succeeds.
job_id	Long	ID of a training job
job_name	String	Name of a training job
status	Int	Status of a training job. For details about the job statuses, see Job Statuses.
create_time	Long	Timestamp when a training job is created
version_id	Long	Version ID of a training job
version_name	String	Version name of a training job

Sample Request

The following shows how to create a job whose job_id is 10 and pre_version_id is 20.

POST    https://endpoint/v1/{project_id}/training-jobs/10/versions/
{
    "job_desc": "This is a ModelArts job",
    "config": {
        "worker_server_num": 1,
        "app_url": "/usr/app/",
        "boot_file_url": "/usr/app/boot.py",
        "parameter": [
            {
                "label": "learning_rate",
                "value": "0.01"
            },
            {
                "label": "batch_size",
                "value": "32"
            }
        ],
        "dataset_id": "38277e62-9e59-48f4-8d89-c8cf41622c24",
        "dataset_version_id": "2ff0d6ba-c480-45ae-be41-09a8369bfc90",
        "spec_id": 1,
        "engine_id": 1,
        "train_url": "/usr/train/",
        "log_url": "/usr/log/",
        "pre_version_id": 20,
        "model_id": 1,
        "pool_id": "test-pool"
    }
}

Sample Response

Successful response

{
    "is_success": true,
    "job_id": 10,
    "job_name": "TestModelArtsJob",
    "status": 1,
    "create_time": 1524189990635,
    "version_id": 10,
    "version_name":"V0001"
}

Failed response

{
    "is_success": false,
    "error_message": "Error string",
    "error_code": "ModelArts.0105"
}

Status Code

For details about the status code, see Status Code.

Parent topic: Training Jobs

Previous topic: Obtaining Training Job Versions

Next topic: Stopping a Training Job

Feedback

Was this page helpful?

Helpful Not helpful

Provide feedback

Thank you very much for your feedback. We will continue working to improve the documentation.See the reply and handling status in My Cloud VOC.

The system is busy. Please try again later.

Which of the following issues have you encountered?

Content is inconsistent with the product UI

Unclear descriptions

Lack of examples or code

Incorrect steps

Can't find what I need

Lack of best practices

Feedback (optional)

0/500

Select at least one type of issue, and enter your comments or suggestions.

Enter a maximum of 500 characters.

Submit Cancel

For any further questions, feel free to contact us through the chatbot.

Chatbot