Training

TypeScript

API reference for enhanced model training methods.

`client.models.train.trigger(modelID, params)`

Trigger model training. Returns enhanced response with response.job.wait(). See Jobs for the full .wait() reference.

Signature

client.models.train.trigger(
  modelID: string,
  params: TrainTriggerParams,
  options?: RequestOptions,
): APIPromise<WithJob<TrainTriggerResponse>>

Parameters

Name	Type	Description
`modelID`	`string`	The model ID to train.
`params.parameters?`	`TrainTriggerParams.Parameters`	Training hyperparameters (body), for example `maxTrainSteps`.
`params.dryRun?`	`'true' \| 'api'`	Estimates the creative units cost (269 response) instead of training. `'api'` estimates at the API tier. The response then carries only the cost estimate (`creativeUnitsCost`) and no job, so don’t call `.wait()` on it.
`params.originalAssets?`	`boolean`	If `true`, returns the original assets without transformation.
`params.projectId?`	`string`	Project that owns the resource.
`params.trainingImagesCount?`	`number`	Simulated number of training images, for `dryRun`. Ignored if the model already has training images.

Returns

APIPromise<WithJob<TrainTriggerResponse>> — The response contains:

Property	Type	Description
`job`	`Job`	The training job, with `.wait()` method. Absent with `dryRun`.
`model`	`TrainTriggerResponse.Model`	The model data. Absent with `dryRun`.
`creativeUnitsCost?`	`number`	Creative units cost of the request (the estimate when `dryRun` is set).

TrainTriggerParams and TrainTriggerResponse are not exported from the package root. Import them from @scenario-labs/sdk/resources/models/train if you need to name them.

Example

import Scenario from '@scenario-labs/sdk';

const client = new Scenario({ apiKey: '...', apiSecret: '...' });

const response = await client.models.train.trigger('model_my-custom-model', {
  parameters: { maxTrainSteps: 1000 },
});

console.log(response.job.jobId);  // 'job_...'
console.log(response.job.status); // 'in-progress'

const completed = await response.job.wait({
  intervalMs: 15_000,   // poll every 15 seconds
  timeoutMs: 900_000,   // wait up to 15 minutes
});

console.log(completed.status); // 'success'