Dsxir.Optimizer.BootstrapFewShot.Sampler (dsxir v0.2.0)

Copy Markdown

Persistent state for a Dsxir.Optimizer.BootstrapFewShot session.

Holds the data that Dsxir.OptimizerSession needs to checkpoint and resume a sequential BootstrapFewShot run: the round counter, the remaining examples for the current round, the accumulated demos pool per predictor, the predictor declarations needed to compatibility-check labeled demos, the labeled demos already selected during phase one, a deterministic shuffle seed, the resolved config, and the running error counter.

Serialized via :erlang.term_to_binary/2 with :deterministic so two checkpoints over the same logical state are byte-identical, and deserialized with :safe to refuse atom creation from untrusted blobs.

Summary

Types

indexed_example()
t()

Types

indexed_example()

@type indexed_example() :: {Dsxir.Example.t(), non_neg_integer()}

t()

@type t() :: %Dsxir.Optimizer.BootstrapFewShot.Sampler{
  all_examples_for_round: [indexed_example()],
  config: map(),
  demos_pool: %{required(atom()) => [Dsxir.Demo.t()]},
  error_count: non_neg_integer(),
  examples_remaining_in_round: [indexed_example()],
  labeled_demos_slotted: [Dsxir.Demo.t()],
  max_rounds: pos_integer(),
  predictor_decls: [term()],
  rng_seed: integer(),
  round: pos_integer()
}