trinity.perf.tensorboard_metrics module

Contents

trinity.perf.tensorboard_metrics module#

Helpers for TensorBoard metric parsing and aggregation.

class trinity.perf.tensorboard_metrics.TensorBoardScalarReader(log_dir: str)[source]#

Bases: object

Read scalar metrics from TensorBoard event files.

__init__(log_dir: str)[source]#

trinity.perf.tensorboard_metrics.extract_raw_metrics_for_step(metric_map: dict[str, dict[int, float]], step: int) → dict[str, float][source]#: Extract all scalar metrics that were logged for one step.

trinity.perf.tensorboard_metrics.collect_step_metrics(metric_map: dict[str, dict[int, float]]) → list[dict[str, Any]][source]#: Build per-step metrics from TensorBoard scalars.