CrucibleIR.Training.Run (CrucibleIR v0.3.0)

Represents a training execution.

A TrainingRun tracks the execution of a training configuration, including status, metrics, artifacts, and timing information.

Fields

:id - Run identifier (required)
:config - Training configuration (required)
:status - Current run status
:current_epoch - Current training epoch
:metrics_history - Metrics over time
:best_metrics - Best achieved metrics
:checkpoint_uris - Saved checkpoint paths
:final_model_version - Resulting model version
:started_at - Start timestamp
:completed_at - Completion timestamp
:error_message - Error if failed
:options - Additional options

Examples

iex> config = %CrucibleIR.Training.Config{
...>   id: :train_config,
...>   model_ref: %CrucibleIR.ModelRef{id: :gpt2},
...>   dataset_ref: %CrucibleIR.DatasetRef{name: :wikitext}
...> }
iex> run = %CrucibleIR.Training.Run{
...>   id: :run_001,
...>   config: config,
...>   status: :running
...> }
iex> run.status
:running

Summary

Types

status()

t()

Types

status()

@type status() :: :pending | :running | :completed | :failed | :cancelled | atom()

t()

@type t() :: %CrucibleIR.Training.Run{
  best_metrics: map() | nil,
  checkpoint_uris: [String.t()] | nil,
  completed_at: DateTime.t() | nil,
  config: CrucibleIR.Training.Config.t(),
  current_epoch: pos_integer() | nil,
  error_message: String.t() | nil,
  final_model_version: atom() | nil,
  id: atom(),
  metrics_history: [map()] | nil,
  options: map() | nil,
  started_at: DateTime.t() | nil,
  status: status()
}