recompile_profiler_experiment() — pytorch Function Reference

Architecture documentation for the recompile_profiler_experiment() function in common.py from the pytorch codebase.

Function python DynamoBenchmarks Microbenchmarking calls 1

Entity Profile

DynamoBenchmarks→ Microbenchmarking→ recompile_profiler_experiment() — pytorch Function Reference

Dependency Diagram

graph TD
  8d5b4cea_0c1f_6c73_230e_f6e909faa7c7["recompile_profiler_experiment()"]
  3473d1a5_c1f5_fc97_006e_79a1d3081bef["write_outputs()"]
  8d5b4cea_0c1f_6c73_230e_f6e909faa7c7 -->|calls| 3473d1a5_c1f5_fc97_006e_79a1d3081bef
  style 8d5b4cea_0c1f_6c73_230e_f6e909faa7c7 fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

benchmarks/dynamo/common.py lines 852–863

def recompile_profiler_experiment(args, model_iter_fn, model, example_inputs):
    prof = torch._dynamo.utils.CompilerProfiler()
    opt_model_iter_fn = torch._dynamo.optimize(prof, nopython=args.nopython)(
        model_iter_fn
    )
    opt_model_iter_fn(model, example_inputs)
    write_outputs(
        output_filename, ["model", "profiler report"], [current_name, prof.report()]
    )
    met = prof.get_metrics()
    guard_failures = len(met["guard_failures"])
    return [guard_failures]