Start TrainingSession

Starting from a ModelTemplate and a Darwin Dataset, spin up a new TrainingSession. If the TrainingSession completes successfully, it will generate a new TrainedModel.

post/model_templates/{model_template_id}/training_sessions

Request

  • Base URL
    https://darwin.v7labs.com/ai
  • URL
    https://darwin.v7labs.com/ai/model_templates/{model_template_id}/training_sessions
  • Auth
    API key in header Authorization

Path parameters

model_template_idstring required

The ID of the ModelTemplate to base the TrainingSession on

Request body

namestring

The name of the TrainingSession

dataset_identifierstring

A <team-slug>/<dataset-slug> string, identifying a Darwin Dataset

devicestring

The device the TrainingSession will be started on. It can either be "gpu" or "cpu", provided that the associated ModelTemplate supports it.

train_paramsstring json

A list of optional parameters, needed by the ModelTemplate Integration to successfully start a TrainingSession

classesstring

The list of annotation classes to be targeted

Response

200

namestring
dataset_identifierstring
devicestring

Example response

{
  "classes": [
    {
      "name": "African Grey Parrot",
      "type": "polygon"
    }
  ],
  "name": "Bird Segmentation Experiment",
  "dataset_identifier": "nest/bird-species",
  "device": "gpu",
  "train_params": {
    "num_epochs": 100,
    "val_percentage": 10
  }
}

Changes

No recorded changes to this operation across all 1 revision of this API.