[Benchmarking-Py] Adding the argument --detailed_conversion_summary to print the detailed TRTConverterV2 summary

Author: DEKHTIARJonathan

tftrt/benchmarking-python/benchmark_args.py

@@ -201,6 +201,13 @@ def __init__(self):
             help="Whether to build TensorRT engines during runtime."
         )
 
+        self._add_bool_argument(
+            name="detailed_conversion_summary",
+            default=False,
+            required=False,
+            help="Whether to use TF-TRT detailled conversion summary."
+        )
+
         self._parser.add_argument(
             "--max_workspace_size",
             type=int,

tftrt/benchmarking-python/benchmark_runner.py

@@ -378,7 +378,10 @@ def engine_build_input_fn(num_batches, model_phase):
                     line_length = max(160, os.get_terminal_size().columns)
                 except OSError:
                     line_length = 160
-                converter.summary(line_length=line_length, detailed=True)
+                converter.summary(
+                    line_length=line_length,
+                    detailed=self._args.detailed_conversion_summary
+                )
             except AttributeError:
                 pass