Member Junction
    Preparing search index...

    Input to TrainingEngine.trainModel. A standalone/manual train is the default; pass experimentSessionIterationId to hang the run off a generic Experiment Session Iteration (plan §4.4) during an agent-driven search.

    interface TrainModelInput {
        candidates?: readonly CandidateSpec[];
        experimentSessionIterationId?: string;
        pipelineId: string;
        sidecarVersion?: string;
        timeoutSeconds?: number;
    }
    Index

    Properties

    candidates?: readonly CandidateSpec[]

    The candidates of this search; absent means the pipeline's own algorithm as the single candidate. More candidates than the bound protocol's budget allows refuse the search before a run row is written (too-many-candidates).

    experimentSessionIterationId?: string

    Optional MJ: Experiment Session Iterations id this run belongs to. NULL / omitted for a one-off standalone train outside a session.

    pipelineId: string

    Id of the MJ: ML Training Pipelines row to train.

    sidecarVersion?: string

    Sidecar version string recorded in model lineage (provenance).

    timeoutSeconds?: number

    Seconds left in an enclosing budget (an experiment session). The trainer's timeout is the smaller of this and the protocol's compute budget; absent or null means no enclosing budget.