.ff -> .pte

Summary: .pte is the file extension for executorch binaries. We should stop using .ff and encourage partners and clients to use .pte. Pytorch/pytorch has never enforced extensions like .pt or .ptl so I'm comfortable not strictly enforcing .pte

Reviewed By: dbort

Differential Revision: D47967450

fbshipit-source-id: 13c5fc2b51d49ae2e63996d5814108b7808ce167

Author: JacobSzwejbka

.github/workflows/pull.yml

@@ -37,4 +37,4 @@ jobs:
         # Export a test model
         python3 -m examples.export.export_example --model_name="linear"
         # Run test model
-        buck2 run //examples/executor_runner:executor_runner -- --model_path ./linear.ff
+        buck2 run //examples/executor_runner:executor_runner -- --model_path ./linear.pte

backends/xnnpack/test/test_xnnpack_utils.py

@@ -116,7 +116,7 @@ def save_bundled_program(representative_inputs, program, ref_output, output_path
     bundled_program_buffer = serialize_from_bundled_program_to_flatbuffer(
         bundled_program
     )
-    output_path_with_postfix = f"{output_path}_bundled.ff"
+    output_path_with_postfix = f"{output_path}_bundled.pte"
     print(f"saving bundled program to {output_path}...")
 
     with open(output_path_with_postfix, "wb") as file:
@@ -223,7 +223,7 @@ def forward(self, *args):
 
         ref_output = delegated_program(*sample_inputs)
         if dump_ff:
-            filename = f"/tmp/xnnpack_test_{randint(1, 99999)}.ff"
+            filename = f"/tmp/xnnpack_test_{randint(1, 99999)}.pte"
             print(f"Writing flatbuffer to {filename} ...")
 
             save_bundled_program(

codegen/tools/targets.bzl

@@ -143,7 +143,7 @@ def define_common_targets(is_fbcode = False):
             srcs = ["test/test_gen_oplist_real_model.py"],
             base_module = "",
             resources = {
-                "//executorch/test/models:exported_programs[ModuleLinear.ff]": "test/ModuleLinear.ff",
+                "//executorch/test/models:exported_programs[ModuleLinear.pte]": "test/ModuleLinear.pte",
             },
             visibility = [
                 "//executorch/...",

codegen/tools/test/test_gen_oplist.py

@@ -132,7 +132,7 @@ def test_get_custom_build_selector_with_both_allowlist_and_yaml(
     ) -> None:
         op_list = ["aten::add", "aten::mul"]
         filename = os.path.join(self.temp_dir.name, "selected_operators.yaml")
-        gen_oplist._dump_yaml(op_list, filename, "model.ff")
+        gen_oplist._dump_yaml(op_list, filename, "model.pte")
         self.assertTrue(os.path.isfile(filename))
         with open(filename) as f:
             es = yaml.safe_load(f)

codegen/tools/test/test_gen_oplist_real_model.py

@@ -15,7 +15,7 @@
 
 from libfb.py import parutil
 
-MODEL_PATH: Final[str] = parutil.get_file_path("ModuleLinear.ff", pkg=__package__)
+MODEL_PATH: Final[str] = parutil.get_file_path("ModuleLinear.pte", pkg=__package__)
 
 
 class TestGenOplistRealModel(unittest.TestCase):

docs/website/docs/tutorials/00_setting_up_executorch.md

@@ -51,10 +51,10 @@ pip install .
 
 Via python script:
 ```bash
-# Creates the file `add.ff`
+# Creates the file `add.pte`
 python3 -m examples.export.export_example --model_name="add"
 
-# Creates the delegated program `composite_model.ff`, other options are "whole" and "partition"
+# Creates the delegated program `composite_model.pte`, other options are "whole" and "partition"
 python3 -m examples.export.export_and_delegate --option "composite"
 ```
 
@@ -65,7 +65,7 @@ $ python3
 >>> from executorch.exir.tests.models import Mul
 >>> m = Mul()
 >>> print(exir.capture(m, m.get_random_inputs()).to_edge())
->>> open("add.ff", "wb").write(exir.capture(m, m.get_random_inputs()).to_edge().to_executorch().buffer)
+>>> open("add.pte", "wb").write(exir.capture(m, m.get_random_inputs()).to_edge().to_executorch().buffer)
 ```
 
 ## Runtime Setup
@@ -119,17 +119,17 @@ conda install -c conda-forge lld
 ### Step 3: Run a binary
 
 ```bash
-# add.ff is the program generated from export_example.py during AOT Setup Step 3
-/tmp/buck2 run //examples/executor_runner:executor_runner -- --model_path add.ff
+# add.pte is the program generated from export_example.py during AOT Setup Step 3
+/tmp/buck2 run //examples/executor_runner:executor_runner -- --model_path add.pte
 
 # To run a delegated model
-/tmp/buck2 run //examples/executor_runner:executor_runner -- --model_path composite_model.ff
+/tmp/buck2 run //examples/executor_runner:executor_runner -- --model_path composite_model.pte
 ```
 
 or execute the binary directly from the `--show-output` path shown when building.
 
 ```bash
-./buck-out/.../executor_runner --model_path add.ff
+./buck-out/.../executor_runner --model_path add.pte
 ```
 
 ## More examples

docs/website/docs/tutorials/backend_delegate.md

@@ -202,7 +202,7 @@ model_inputs = (torch.ones(1), )
 exec_prog = exir.capture(composite_model, model_inputs).to_edge().to_executorch()
 
 # Save the flatbuffer to a local file
-save_path = "delegate.fft"
+save_path = "delegate.pte"
 with open(save_path, "wb") as f:
     f.write(exec_prog.buffer)
 ```
@@ -257,7 +257,7 @@ exec_prog = to_backend(gm, AddMulPartitionerDemo).to_executorch(
 )
 
 # Save the flatbuffer to a local file
-save_path = "delegate.fft"
+save_path = "delegate.pte"
 with open(save_path, "wb") as f:
     f.write(exec_prog.buffer)
 ```

docs/website/docs/tutorials/exporting_to_executorch.md

@@ -196,7 +196,7 @@ executorch_program = edge_dialect.to_executorch(executorch_backend_config)
 buffer = executorch_program.buffer
 
 # Save it to a file and load it in the Executorch runtime
-with open("model.ff", "wb") as file:
+with open("model.pte", "wb") as file:
     file.write(buffer)
 
 # Or you can run the Executorch runtime in python to try it out

docs/website/docs/tutorials/profiling.md

@@ -56,7 +56,7 @@ Here is an example of what a ExecuTorch run + profile + post-procesing workflow
 This runs the sample program with profiling enabled
 ```bash
 cd executorch
-buck2 run -c executorch.prof_enabled=true examples/executor_runner:executor_runner  -- --model_path add.ff
+buck2 run -c executorch.prof_enabled=true examples/executor_runner:executor_runner  -- --model_path add.pte
 ```
 Run the post-processing CLI tool that calls into the same API's listed above and prints out the profiling results in a tabulated format in the terminal.
 

docs/website/docs/tutorials/selective_build.md

@@ -32,7 +32,7 @@ load("@fbsource//xplat/executorch/codegen:codegen.bzl", "et_operator_library")
 
 fb_native.export_file(
   name = "model_1",
-  src = "model_1.ff", # checked in model
+  src = "model_1.pte", # checked in model
 )
 
 et_operator_library(