make extract_delegate_segments=True by default

backends/apple/mps/test/test_mps_utils.py

@@ -247,7 +247,9 @@ def lower_module_and_test_output(
             )
 
             executorch_program = delegated_program.to_executorch(
-                config=ExecutorchBackendConfig(extract_constant_segment=False)
+                config=ExecutorchBackendConfig(
+                    extract_delegate_segments=False, extract_constant_segment=False
+                )
             )
         else:
             delegated_program = to_backend(
@@ -264,7 +266,9 @@ def lower_module_and_test_output(
                     _skip_dim_order=True,  # TODO(T182928844): Delegate dim order op to backend.
                 ),
             ).to_executorch(
-                config=ExecutorchBackendConfig(extract_constant_segment=False)
+                config=ExecutorchBackendConfig(
+                    extract_delegate_segments=False, extract_constant_segment=False
+                )
             )
 
         if bundled_program:

backends/arm/test/arm_tosa_reference.py

@@ -202,7 +202,9 @@ def tosa_run_test(op, profile=TosaProfile.MI):  # noqa: C901
 
     model_edge = model_edge.to_backend(ArmPartitioner(compile_spec))
     exec_prog = model_edge.to_executorch(
-        config=ExecutorchBackendConfig(extract_constant_segment=False)
+        config=ExecutorchBackendConfig(
+            extract_delegate_segments=False, extract_constant_segment=False
+        )
     )
 
     # Save ground truth results to file

backends/qualcomm/tests/utils.py

@@ -187,6 +187,7 @@ def lower_module_and_test_output(
         )
         exec_prog = delegated_program.to_executorch(
             exir.ExecutorchBackendConfig(
+                extract_delegate_segments=False,
                 # For shared buffer, user must pass the memory address
                 # which is allocated by RPC memory to executor runner.
                 # Therefore, won't want to pre-allocate
@@ -195,7 +196,7 @@ def lower_module_and_test_output(
                     memory_planning_algo="greedy",
                     alloc_graph_input=not self.shared_buffer,
                     alloc_graph_output=not self.shared_buffer,
-                )
+                ),
             )
         )
 

examples/apple/mps/scripts/mps_example.py

@@ -182,7 +182,9 @@ def get_model_config(args):
         logging.info(f"Lowered graph:\n{edge.exported_program().graph}")
 
         executorch_program = edge.to_executorch(
-            config=ExecutorchBackendConfig(extract_constant_segment=False)
+            config=ExecutorchBackendConfig(
+                extract_delegate_segments=False, extract_constant_segment=False
+            )
         )
     else:
         lowered_module = to_backend(
@@ -192,7 +194,11 @@ def get_model_config(args):
             lowered_module,
             example_inputs,
             edge_compile_config=exir.EdgeCompileConfig(_check_ir_validity=False),
-        ).to_executorch(config=ExecutorchBackendConfig(extract_constant_segment=False))
+        ).to_executorch(
+            config=ExecutorchBackendConfig(
+                extract_delegate_segments=False, extract_constant_segment=False
+            )
+        )
 
     model_name = f"{args.model_name}_mps"
 

examples/arm/aot_arm_compiler.py

@@ -211,7 +211,9 @@ def forward(self, x):
         logging.debug(f"Lowered graph:\n{edge.exported_program().graph}")
 
     exec_prog = edge.to_executorch(
-        config=ExecutorchBackendConfig(extract_constant_segment=False)
+        config=ExecutorchBackendConfig(
+            extract_delegate_segments=False, extract_constant_segment=False
+        )
     )
 
     model_name = f"{args.model_name}" + (

examples/qualcomm/scripts/export_example.py

@@ -96,7 +96,9 @@
         )
 
     executorch_program = delegated_program.to_executorch(
-        config=ExecutorchBackendConfig(extract_constant_segment=False)
+        config=ExecutorchBackendConfig(
+            extract_delegate_segments=False, extract_constant_segment=False
+        )
     )
 
     if args.generate_etrecord:

examples/xnnpack/aot_compiler.py

@@ -103,7 +103,9 @@
     logging.info(f"Lowered graph:\n{edge.exported_program().graph}")
 
     exec_prog = edge.to_executorch(
-        config=ExecutorchBackendConfig(extract_constant_segment=False)
+        config=ExecutorchBackendConfig(
+            extract_delegate_segments=False, extract_constant_segment=False
+        )
     )
 
     if args.etrecord is not None:

examples/xnnpack/quantization/example.py

@@ -191,7 +191,9 @@ def main() -> None:
 
     start = time.perf_counter()
     prog = edge_m.to_executorch(
-        config=ExecutorchBackendConfig(extract_constant_segment=False)
+        config=ExecutorchBackendConfig(
+            extract_delegate_segments=False, extract_constant_segment=False
+        )
     )
     save_pte_program(prog, f"{args.model_name}_quantized")
     end = time.perf_counter()

exir/capture/_config.py

@@ -55,7 +55,7 @@ class ExecutorchBackendConfig:
     # Whether to move delegate data blobs from the Program into separate
     # segments, rather than encoding those blobs in the flatbuffer data.
     # This makes it possible to free those blobs at runtime.
-    extract_delegate_segments: bool = False
+    extract_delegate_segments: bool = True
 
     # Whether to extract constants from the Program into separate segments,
     # rather than encoding those constants in the flatbuffer data.

exir/program/test/test_program.py

@@ -293,19 +293,15 @@ def test_edge_to_backend_replaces_subgraph(self):
         # two delegate blobs for forward and foo
         self.assertEqual(
             len(
-                delegate_manager.to_executorch(
-                    ExecutorchBackendConfig(extract_delegate_segments=True)
-                )
+                delegate_manager.to_executorch(ExecutorchBackendConfig())
                 ._emitter_output.program.execution_plan[0]
                 .delegates
             ),
             1,
         )
         self.assertEqual(
             len(
-                delegate_manager.to_executorch(
-                    ExecutorchBackendConfig(extract_delegate_segments=True)
-                )
+                delegate_manager.to_executorch(ExecutorchBackendConfig())
                 ._emitter_output.program.execution_plan[1]
                 .delegates
             ),
@@ -349,15 +345,23 @@ def test_edge_to_backend_selective(self):
         # one delegate blob for forward
         self.assertEqual(
             len(
-                delegate_manager.to_executorch(ExecutorchBackendConfig())
+                delegate_manager.to_executorch(
+                    ExecutorchBackendConfig(
+                        extract_delegate_segments=False,
+                    )
+                )
                 ._emitter_output.program.execution_plan[0]  # foo
                 .delegates
             ),
             0,
         )
         self.assertEqual(
             len(
-                delegate_manager.to_executorch(ExecutorchBackendConfig())
+                delegate_manager.to_executorch(
+                    ExecutorchBackendConfig(
+                        extract_delegate_segments=False,
+                    )
+                )
                 ._emitter_output.program.execution_plan[1]  # forward
                 .delegates
             ),

exir/tests/test_memory_planning.py

@@ -495,7 +495,7 @@ def test_multiple_pools(
                 memory_planning_pass=CustomPoolMemoryPlanningPass(
                     memory_planning_algo=algo,
                     alignment=1,
-                )
+                ),
             )
         )
         graph_module = edge_program.exported_program().graph_module