MLExperimentRun
One training of an ML experiment. Runs are personal and are never used by scripts, workflows or other models.
Fields
Unique run identifier
The run's number within its experiment (1, 2, 3, ...)
Where the run stands
Possible values (4)
TRAININGThe run is training
TRAINEDThe run trained; its model can be tried until it expires
FAILEDThe run failed; see its error
EXPIREDThe run's model was removed under the workspace's retention settings; its scores remain
The <ml-model> XML this run trained, as it was when the run started
How the run's model scored on held-out rows, against a baseline, the based-on model and earlier runs; null until it trains
Feature importance of the run's model
Per-round evaluation history of the run's training
Why the run failed; null unless it failed
How long the run took, in milliseconds; null while training
Rows the run read (trained plus held out); null until it trains
JSON Schema describing the run's model inputs (types, enum values), the same shape as MLModelConfig.inferenceSchema; null for a forecast run, a run that is not trained, or an expired run
Whether the run is pinned; a pinned run never expires
When the run started
When the run finished; null while training
When the run's model is removed under the workspace's current retention setting; null while training, pinned, failed or already expired
Referenced by (3)
Exact type SDL
"""
One training of an ML experiment. Runs are personal and are never used by scripts, workflows or other models.
"""
type MLExperimentRun {
"""Unique run identifier"""
id: MLExperimentRunID!
"""The run's number within its experiment (1, 2, 3, ...)"""
number: Int!
"""Where the run stands"""
status: MLExperimentRunStatus!
"""The `<ml-model>` XML this run trained, as it was when the run started"""
sourceXML: String!
"""
How the run's model scored on held-out rows, against a baseline, the based-on model and earlier runs; null until it trains
"""
evaluation: MLModelEvaluation
"""Feature importance of the run's model"""
featureImportance: [MLFeatureImportance!]!
"""Per-round evaluation history of the run's training"""
evaluationHistory: [MLEvaluationEntry!]!
"""Why the run failed; null unless it failed"""
error: String
"""How long the run took, in milliseconds; null while training"""
durationMs: Int
"""Rows the run read (trained plus held out); null until it trains"""
rowCount: Int
"""
JSON Schema describing the run's model inputs (types, enum values), the same shape as MLModelConfig.inferenceSchema; null for a forecast run, a run that is not trained, or an expired run
"""
inferenceSchema: String
"""Whether the run is pinned; a pinned run never expires"""
pinned: Boolean!
"""When the run started"""
createdAt: ISO8601!
"""When the run finished; null while training"""
finishedAt: ISO8601
"""
When the run's model is removed under the workspace's current retention setting; null while training, pinned, failed or already expired
"""
expiresAt: ISO8601
}