run_benchmark() — pytorch Function Reference

Architecture documentation for the run_benchmark() function in benchmark.py from the pytorch codebase.

Function python CoreTensor Dispatcher calls 2 called by 1

Entity Profile

CoreTensor→ Dispatcher→ run_benchmark() — pytorch Function Reference

Dependency Diagram

graph TD
  5a01be28_6429_ebf7_fc47_02232b445a1f["run_benchmark()"]
  8a4f57e1_634a_07c6_3a9a_c44f7ba2dbe1["sweep()"]
  8a4f57e1_634a_07c6_3a9a_c44f7ba2dbe1 -->|calls| 5a01be28_6429_ebf7_fc47_02232b445a1f
  320a5772_c0de_3469_ef0c_57110d2440fd["benchmark_process_group()"]
  5a01be28_6429_ebf7_fc47_02232b445a1f -->|calls| 320a5772_c0de_3469_ef0c_57110d2440fd
  9cb370ea_5749_29d9_5dca_4a2b481fcfac["allgather_object()"]
  5a01be28_6429_ebf7_fc47_02232b445a1f -->|calls| 9cb370ea_5749_29d9_5dca_4a2b481fcfac
  style 5a01be28_6429_ebf7_fc47_02232b445a1f fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

benchmarks/distributed/ddp/benchmark.py lines 86–97

def run_benchmark(benchmark, ranks, opts):
    group = dist.new_group(ranks=ranks, backend=benchmark.distributed_backend)
    measurements = []
    if dist.get_rank() in set(ranks):
        if not opts:
            opts = {}
        measurements = benchmark_process_group(group, benchmark, **opts)
    dist.destroy_process_group(group)
    dist.barrier()

    # Aggregate measurements for better estimation of percentiles
    return list(itertools.chain(*allgather_object(measurements)))