_write_results_to_json() — pytorch Function Reference

Architecture documentation for the _write_results_to_json() function in cachebench.py from the pytorch codebase.

Function python DynamoBenchmarks Microbenchmarking called by 1

Entity Profile

DynamoBenchmarks→ Microbenchmarking→ _write_results_to_json() — pytorch Function Reference

Dependency Diagram

graph TD
  4c5d5750_9c4f_4bb3_15ed_612570273bef["_write_results_to_json()"]
  e68c7f5d_5584_dd88_8af3_2bcda2115580["main()"]
  e68c7f5d_5584_dd88_8af3_2bcda2115580 -->|calls| 4c5d5750_9c4f_4bb3_15ed_612570273bef
  style 4c5d5750_9c4f_4bb3_15ed_612570273bef fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

benchmarks/dynamo/cachebench.py lines 146–184

def _write_results_to_json(
    cmd_args: argparse.Namespace,
    results: list[RunResult],
) -> None:
    if len(results) == 0:
        # do not write empty results
        return

    records = []
    for result in results:
        for metric_name, value in [
            ("Cold compile time (s)", result.cold_compile_s),
            ("Warm compile time (s)", result.warm_compile_s),
            ("Speedup (%)", [result.speedup_pct]),
        ]:
            records.append(
                {
                    "benchmark": {
                        "name": "TorchCache Benchmark",
                        "mode": result.mode,
                        "extra_info": {
                            "is_dynamic": result.dynamic,
                            "device": result.device,
                        },
                    },
                    "model": {
                        "name": result.model,
                        "backend": "inductor",
                        "origins": [result.benchmark],
                    },
                    "metric": {
                        "name": metric_name,
                        "type": "OSS model",
                        "benchmark_values": value,
                    },
                }
            )
    with open(cmd_args.output, "w") as f:
        json.dump(records, f)