aiq.profiler.inference_optimization.workflow_runtimes#

Functions#

compute_workflow_runtime_metrics(...)

Computes the p90, p95, and p99 of workflow runtime for each example_number.

Module Contents#

compute_workflow_runtime_metrics(
all_steps: list[list[aiq.data_models.intermediate_step.IntermediateStep]],
) → aiq.profiler.inference_optimization.data_models.WorkflowRuntimeMetrics#

Computes the p90, p95, and p99 of workflow runtime for each example_number.

The ‘workflow runtime’ per example is:

max(event_timestamp) - min(event_timestamp)

for that example_number.

Parameters#

all_stepsIntermediateStep
Must contain at least two columns:

  • ‘example_number’

  • ‘event_timestamp’

Returns#

WorkflowRuntimeMetrics

A Pydantic model with ‘p90’, ‘p95’, and ‘p99’ attributes.