Documentation

Welcome to SigOpt’s developer documentation. If you have a question you can’t answer, feel free to contact us!

Checkpoint Create

https://api.sigopt.com/v1/experiments/EXPERIMENT_ID/training_runs/TRAINING_RUN_ID/checkpoints

Only relevant for a Training Monitor experiment. Creates a new guest Checkpoint for a given Training Run. Note that metadata for checkpoints can have no more than 4 entries.

Request Method: POST

Parameters

NameTypeRequired?Description
training_runstringYThe id of the Training Run to which this Checkpoint is associated.
valuesarray<Metric Evaluation>YAn array of Metric Evaluation objects. For Training Monitor experiments, this is required even in single metric experiments.
metadataMetadataNOptional user-provided object. See Using Metadata for more information.

Response

Training Run object.

Example Request

training_run = conn.experiments(EXPERIMENT_ID).training_runs(TRAINING_RUN_ID).checkpoints().create(
  training_run="19147",
  values=[
    dict(
      name="Validation Accuracy",
      value=0.6543
      )
    ]
  )
Response
{
  "created": 1554842812,
  "id": "108468",
  "object": "checkpoint",
  "should_stop": false,
  "stopping_reasons": {
    "Look Back 2 Steps": false
  },
  "training_run": "19147",
  "values": [
    {
      "name": "Validation Accuracy",
      "object": "metric_evaluation",
      "value": 0.6543,
      "value_stddev": null
    }
  ]
}