Remove direct calls to serialize_to_flatbuffer

Summary:
Use `to_executorch().buffer` instead.

While I'm in here, remove all other references to "flatbuffer" in these files.

Reviewed By: larryliu0820

Differential Revision: D48366594

fbshipit-source-id: a68184cc8c2479047ebbc20396fcb44a03e40ba9

Author: dbort

backends/xnnpack/test/test_xnnpack_utils.py

@@ -31,10 +31,10 @@
 from executorch.bundled_program.serialize import (
     serialize_from_bundled_program_to_flatbuffer,
 )
+from executorch.exir import ExecutorchProgram, ExirExportedProgram
 from executorch.exir.backend.backend_api import to_backend, validation_disabled
 
 from executorch.exir.passes.spec_prop_pass import SpecPropPass
-from executorch.exir.serialize import serialize_to_flatbuffer
 from executorch.exir.tracer import _default_decomposition_table
 
 # pyre-ignore[21]: Could not find module `executorch.extension.pybindings.portable`.
@@ -158,16 +158,16 @@ def lower_module_and_test_output(
         quantized_dynamic: bool = False,
         # TODO: remove this after we migrate to use long term flow
         quantizer_api_test: bool = False,
-        dump_ff: bool = False,  # for debugging, dump the generated flatbuffer file
-    ) -> exir.ExirExportedProgram:
+        dump_bundled_program: bool = False,  # for debugging, dump the generated bundled program file
+    ) -> ExirExportedProgram:
         """
         Helper testing function that takes a torch.nn.Module and lowers it to XNNPACK with
         the given sample inputs. It then runs the lowered module and compares its
         outputs with the outputs of the eager module.
         """
 
         if quantizer_api_test:
-            assert isinstance(module, exir.ExirExportedProgram)
+            assert isinstance(module, ExirExportedProgram)
             edge_program = module
         else:
 
@@ -197,46 +197,42 @@ def forward(self, *args):
                     edge_program.exported_program, partitioner
                 )
 
-            program = delegated_program.to_executorch(
+            executorch_program: ExecutorchProgram = delegated_program.to_executorch(
                 get_xnnpack_executorch_backend_config([SpecPropPass()]),
-            ).program
+            )
         else:
             delegated_program = to_backend(
                 "XnnpackBackend", edge_program.exported_program, []
             )
 
-            exported_program = capture_graph_for_xnnpack(
+            exported_program: ExirExportedProgram = capture_graph_for_xnnpack(
                 delegated_program, sample_inputs
             )
-            program = exported_program.to_executorch(
+            executorch_program: ExecutorchProgram = exported_program.to_executorch(
                 get_xnnpack_executorch_backend_config(),
-            ).program
+            )
 
         # print("Graph Module with delegate:")
         # delegated_module.print_readable()
 
         # Assert the backend name is xnnpack
         self.assertEqual(
-            program.execution_plan[0].delegates[0].id,
+            executorch_program.program.execution_plan[0].delegates[0].id,
             XnnpackBackend.__name__,
         )
-        buffer = serialize_to_flatbuffer(program)
 
         ref_output = delegated_program(*sample_inputs)
-        if dump_ff:
-            filename = f"/tmp/xnnpack_test_{randint(1, 99999)}.pte"
-            print(f"Writing flatbuffer to {filename} ...")
-
+        if dump_bundled_program:
             save_bundled_program(
                 representative_inputs=sample_inputs,
-                program=program,
+                program=executorch_program.program,
                 ref_output=ref_output,
-                output_path=filename,
+                output_path=f"/tmp/xnnpack_test_{randint(1, 99999)}",
             )
 
         # Test the model with executor
-        # pyre-ignore
-        executorch_module = _load_for_executorch_from_buffer(buffer)
+        # pyre-ignore[16]: Module `executorch.extension.pybindings` has no attribute `portable`.
+        executorch_module = _load_for_executorch_from_buffer(executorch_program.buffer)
         # pyre-fixme[16]: Module `pytree` has no attribute `tree_flatten`.
         inputs_flattened, _ = tree_flatten(sample_inputs)
 
@@ -353,7 +349,7 @@ def _test_xnnpack_dqlinear(
         self,
         weight_qconfig,
         use_bias: bool,
-        dump_ff: bool = False,
+        dump_bundled_program: bool = False,
     ):
         assert weight_qconfig in [
             weight_observer_range_neg_127_to_127,
@@ -413,39 +409,39 @@ def forward(self, x):
         composite_model = CompositeModule()
         composite_model(*example_inputs)
 
-        exported_program = capture_graph_for_xnnpack(composite_model, example_inputs)
-        program = exported_program.to_executorch(
+        exported_program: ExirExportedProgram = capture_graph_for_xnnpack(
+            composite_model, example_inputs
+        )
+        executorch_program: ExecutorchProgram = exported_program.to_executorch(
             get_xnnpack_executorch_backend_config(),
-        ).program
+        )
 
         self.assertEqual(
-            program.execution_plan[0].delegates[0].id,
+            executorch_program.program.execution_plan[0].delegates[0].id,
             XnnpackBackend.__name__,
         )
 
         ref_output = captured_dqlinear(*example_inputs)
         ref_output = composite_model(*example_inputs)
         print("ref_output:", ref_output)
 
-        if dump_ff:
+        if dump_bundled_program:
             mm_str = "addmm" if use_bias else "mm"
             filename = f"/tmp/dqlinear_{mm_str}"
             if weight_qconfig == weight_observer_range_neg_127_to_127:
                 filename = f"{filename}_per_tensor"
             else:
                 filename = f"{filename}_per_channel"
-            print(f"Writing flatbuffer to {filename} ...")
 
             save_bundled_program(
                 representative_inputs=example_inputs,
-                program=program,
+                program=executorch_program.program,
                 ref_output=ref_output,
                 output_path=filename,
             )
 
-        buffer = serialize_to_flatbuffer(program)
         # pyre-ignore
-        executorch_module = _load_for_executorch_from_buffer(buffer)
+        executorch_module = _load_for_executorch_from_buffer(executorch_program.buffer)
         # pyre-fixme[16]: Module `pytree` has no attribute `tree_flatten`.
         inputs_flattened, _ = tree_flatten(example_inputs)
 

backends/xnnpack/test/tester/tester.py

@@ -28,7 +28,6 @@
 from executorch.exir.backend.backend_api import to_backend, validation_disabled
 from executorch.exir.backend.partitioner import Partitioner
 from executorch.exir.passes.spec_prop_pass import SpecPropPass
-from executorch.exir.serialize import serialize_to_flatbuffer
 
 # pyre-ignore[21]: Could not find module `executorch.pybindings.portable`.
 from executorch.extension.pybindings.portable import (  # @manual
@@ -253,7 +252,7 @@ def __init__(self):
         self.buffer = None
 
     def run(self, artifact: ExecutorchProgram, inputs=None) -> None:
-        self.buffer = serialize_to_flatbuffer(artifact.program)
+        self.buffer = artifact.buffer
 
     @property
     def artifact(self) -> bytes: