tinker_cookbook.stores.TrainingRunStore

class tinker_cookbook.stores.TrainingRunStore(**)

Typed read/write access to one training run's data.

url(path)

Return a human-readable URI for a path within this run.

Parameters:

path (str)

read_config()

Read config.json (cached after first read).

read_metrics()

Read all metrics (incremental — only new data from disk).

read_new_metrics()

Read only metrics added since last call.

metric_keys()

All metric keys seen so far (excluding 'step').

read_rollouts(iteration, base_name)

Read rollout summaries for an iteration as raw dicts.

Parameters:

iteration (int) – Training iteration number.
base_name (str) – Prefix for the JSONL file (e.g. "train", "eval_gsm8k"). Matches the naming used by rollout_summaries_jsonl_path() in RL training.

read_single_rollout(iteration, group_idx, traj_idx, base_name)

Find one rollout by group and trajectory index, or None.

Parameters:

iteration (int)
group_idx (int)
traj_idx (int)
base_name (str)

read_checkpoints()

Read checkpoints.jsonl.

read_checkpoint_records()

Read checkpoints.jsonl as CheckpointRecord objects.

read_timing()

Read all timing records (incremental — only new data from disk).

read_logtree(iteration, base_name)

Read a logtree JSON file for an iteration, or None if missing.

Parameters:

iteration (int)
base_name (str)

list_logtrees(iteration)

List logtree base names for an iteration (e.g. ["train", "eval_gsm8k"]).

Parameters:

iteration (int)

list_iterations()

List all iteration directories with metadata about their contents.

write_config(config)

Write config.json (overwrites if exists, updates cache).

Parameters:

config (dict[str, Any])

write_metrics(metrics, step)

Append one metrics record to metrics.jsonl.

Parameters:

metrics (dict[str, Any])
step (int | None)

write_timing_spans(step, spans)

Append one timing record to timing_spans.jsonl.

Parameters:

step (int)
spans (list[dict[str, Any]])

write_checkpoint(record)

Append one checkpoint record to checkpoints.jsonl.

Parameters:

record (dict[str, Any])

write_rollouts(iteration, records, base_name)

Write rollout summaries for an iteration (overwrites).

Parameters:

iteration (int) – Training iteration number.
records (list[dict[str, Any]]) – List of trajectory dicts to write.
base_name (str) – Prefix for the JSONL file (e.g. "train", "eval_gsm8k"). Must match the base_name used in read_rollouts().