benchmark_single_shape() — pytorch Function Reference

Architecture documentation for the benchmark_single_shape() function in utils.py from the pytorch codebase.

Function python DynamoBenchmarks Configuration calls 2 called by 8

Entity Profile

DynamoBenchmarks→ Configuration→ benchmark_single_shape() — pytorch Function Reference

Dependency Diagram

graph TD
  38cd02c9_bda5_f921_f4c9_f7772e1a880b["benchmark_single_shape()"]
  ac0bb274_6e2a_862f_4e64_142f3728825b["benchmark()"]
  ac0bb274_6e2a_862f_4e64_142f3728825b -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  02c5a438_2afc_6211_922d_df8faf0ed669["benchmark()"]
  02c5a438_2afc_6211_922d_df8faf0ed669 -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  a45d8f6b_d010_10b3_6034_08ec9eff9865["benchmark()"]
  a45d8f6b_d010_10b3_6034_08ec9eff9865 -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  ab5a067d_4dce_7aad_050e_58d0942293ee["benchmark()"]
  ab5a067d_4dce_7aad_050e_58d0942293ee -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  f0518c93_32fe_49c0_0cae_0cae28698405["benchmark()"]
  f0518c93_32fe_49c0_0cae_0cae28698405 -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  3f9404a0_b5ca_d039_2c4e_873e66d4608d["benchmark()"]
  3f9404a0_b5ca_d039_2c4e_873e66d4608d -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  e5564ea1_2d08_4990_c09b_7d9b97e2b066["benchmark()"]
  e5564ea1_2d08_4990_c09b_7d9b97e2b066 -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  9ad13df4_e4bb_8645_a1ce_35d8e720f64d["benchmark()"]
  9ad13df4_e4bb_8645_a1ce_35d8e720f64d -->|calls| 38cd02c9_bda5_f921_f4c9_f7772e1a880b
  3d4e92c5_a13f_fad7_1514_52cd7fc93a06["benchmark_single_shape_for_backend()"]
  38cd02c9_bda5_f921_f4c9_f7772e1a880b -->|calls| 3d4e92c5_a13f_fad7_1514_52cd7fc93a06
  bf14c1d3_0a90_a5c6_58ff_3e6211ae65ce["check_accuracy()"]
  38cd02c9_bda5_f921_f4c9_f7772e1a880b -->|calls| bf14c1d3_0a90_a5c6_58ff_3e6211ae65ce
  style 38cd02c9_bda5_f921_f4c9_f7772e1a880b fill:#6366f1,stroke:#818cf8,color:#fff

Relationship Graph

Source Code

benchmarks/dynamo/genai_layers/utils.py lines 170–194

    def benchmark_single_shape(
        self, args, kwargs=None, should_check_accuracy=True, setting: str = ""
    ):
        for backend in self.available_backends:
            self.benchmark_single_shape_for_backend(backend, args, kwargs, setting)
        if (
            "compiled" in self.available_backends
            and self.script_args.custom_compile_options
        ):
            torch._dynamo.reset()  # cause recompile
            with torch._inductor.config.patch(self.script_args.custom_compile_options):
                status = self.benchmark_single_shape_for_backend(
                    self.script_args.custom_compile_name,
                    args,
                    kwargs,
                    setting,
                    fn=self.compiled,
                )
            if not status:
                self.script_args.custom_compile_options = (
                    None  # once fail, don't run again
                )

        if should_check_accuracy:
            self.check_accuracy(args, kwargs)