Generating training

Article Summary

Share feedback

Thanks for sharing your feedback!

Available in Classic and VPC

Generates training. Once the generated training is completed, you can call the API for generating conversational sentences to see the responses generated by CLOVA Studio.

Requests

Describes the request format. The request format is as follows:

Method	URI
POST	/tuning/v2/tasks

Header

The following describes the header.

Field	Required	Description
X-NCP-APIGW-TIMESTAMP	Y	Time elapsed in milliseconds since January 1, 1970, 00:00:00 UTC If the time differs from that of the API Gateway server by 5 minutes or more, then the request is considered invalid x-ncp-apigw-timestamp:{Timestamp}
X-NCP-IAM-ACCESS-KEY	Y	Access key ID issued from NAVER Cloud Platform portal x-ncp-iam-access-key:{Main Account Access Key}
X-NCP-APIGW-SIGNATURE-V2	Y	Signature encrypted with the access key ID value and secret key x-ncp-apigw-signature-v2:{API Gateway Signature}
X-NCP-CLOVASTUDIO-REQUEST-ID	N	Request ID for each request
Content-Type	Y	multipart/form-data

Body

The following describes the body.

Field	Type	Required	Description
name	string	N	Training name (default: training creation date)
model	string	Y	Tuning model name
tuningType	string	N	Tuning method PEFT (default value)
taskType	string	N	Training type GENERATION: generation (default value) CLASSIFICATION: classification
trainEpochs	string	N	Number of epochs to train the model 1-20 (default value: 4)
learningRate	string	N	Training rate: the level or degree to which model parameters are re-trained when tuning 1-1.0E-6 (default value: 1.0E-4)
trainingDataset	string	Y	Dataset file path to train on

Syntax

The following is an example of syntax.

curl --location --request POST 'https://clovastudio.apigw.ntruss.com/tuning/v2/tasks' \
--header 'Content-Type: multipart/form-data' \
--header 'X-NCP-APIGW-TIMESTAMP: <api_gw_time>' \
--header 'X-NCP-IAM-ACCESS-KEY: <iam_access_key>' \
--header 'X-NCP-APIGW-SIGNATURE-V2: <api_gw_sig>' \
--header 'X-NCP-CLOVASTUDIO-REQUEST-ID: <request_id>' \
--form 'name="test"' \
--form 'model="HCX-002"' \
--form 'method="LoRA"' \
--form 'taskType="GENERATION"' \
--form 'trainEpochs="4"' \
--form 'learningRate="1.0E-4"' \
--form 'trainingDataset=@"E:/Downloads/test.csv"'

Responses

Describes the responses format.

Header

The following describes the header.

Header	Required	Description
Content-Type	-	application/json

Body

The following describes the body.

Field	Type	Required	Description
result	object	-	Response result
result.id	string	-	Training ID
result.name	string	-	Training name
result.model	string	-	Tuning model name
result.method	string	-	Tuning method LoRA
result.taksType	string	-	Training type GENERATION: generation CLASSIFICATION: classification
result.trainEpochs	interger	-	Number of epochs to train the model
result.learningRate	double	-	Training rate: the level or degree to which model parameters are re-trained when tuning
result.status	string	-	Training progress status WAIT RUNNING FAILED SUCCEEDED
result.status.statusInfo	array	-	Training progress details
result.status.statusInfo.label	array	-	User data label, if training type is classification (null, if training type is GENERATION)
result.status.statusInfo.dataRows	int	-	Number of data
result.status.statusInfo.numOfTokens	int	-	Number of data tokens
result.status.statusInfo.currStep	int	-	Current number of training steps
result.status.statusInfo.totalTrainSteps	int	-	Total number of training steps
result.status.statusInfo.currEpoch	int	-	Current epochs
result.status.statusInfo.totalTrainEpochs	int	-	Total training epochs
result.status.statusInfo.estimatedTime	int	-	Estimated run time (average 1 epoch time * total training epochs)
result.status.statusInfo.trainLoss	double	-	Amount of training loss
result.status.statusInfo.sendWeightSuccess	boolean	-	Whether to save the training results
result.status.statusInfo.endDatetime	string	-	End date of the training
result.status.statusInfo.failureReason	string	-	Reason for training failure file.read: file recognition failed file.privacy: private information exists in the file file.harmful: harmful information exists in the file resource.timeout: response timeout due to GPU acquisition failure. Retry required result.store: failed to store training results db.timeout: response timeout due to DB connection failure. Retry required
result.status.createdClientType	string	-	Client type that requested the training API: API client WEB: web client
result.status.createdDate	string	-	Training generation date (ISO 8601 format)
result.status.updatedDate	string	-	Training editing date (ISO 8601 format)

Syntax

The following is an example of syntax.

{
    "status": {
        "code": "20000",
        "message": "OK"
},
"result": {
    "id": "czf9fbky",
    "name": "230821-130704",
    "model": "HCX-002",
    "method": "LoRA",
    "taskType": "GENERATION",
    "trainEpochs": 4,
    "learningRate": 1.0E-4,
    "status": "WAIT",
    "statusInfo": {
        "label": null,
        "dataRows": null,
        "numOfTokens": null,
        "currStep": null,
        "totalTrainSteps": null,
        "currEpoch": null,
        "totalTrainEpochs": null,
        "estimatedTime": null,
        "trainLoss": null,
        "sendWeightSuccess": null,
        "failureReason": null,
        "endDatetime": null
        },
        "createdClientType": "API",
        "createdDate": "2023-08-21T13:07:06+0900",
        "updatedDate": "2023-08-21T13:07:06+0900"
    }
}

Was this article helpful?

What's Next

Retrieving training

Table of contents

Requests
Responses