remove "Bundled" prefix for bundled program schema (#397)

Summary:
Pull Request resolved: #397

Now we already have a specific namespace for bundledprogram, this diff removes the "Bundled" prefix for part of bundled program schema's variables.
Rest variables will be updated in the following diffs.

Differential Revision: D49337170

fbshipit-source-id: 57e87ffdc195aa691b8bb4cbd8bb69c1758d2e98

Author: Gasoonjia

bundled_program/core.py

@@ -8,46 +8,34 @@
 import typing
 from typing import Dict, List, Type
 
+import executorch.bundled_program.schema as bp_schema
+import executorch.exir.schema as core_schema
+
 import torch
 import torch.fx
 from executorch.bundled_program.config import (
     BundledConfig,
     ConfigExecutionPlanTest,
     ConfigValue,
 )
-from executorch.bundled_program.schema import (
-    BundledBool,
-    BundledDouble,
-    BundledExecutionPlanTest,
-    BundledInt,
-    BundledIOSet,
-    BundledProgram,
-    BundledTensor,
-    BundledValue,
-)
+
 from executorch.bundled_program.version import BUNDLED_PROGRAM_SCHEMA_VERSION
 from executorch.exir._serialize import _serialize_pte_binary
-from executorch.exir.schema import (
-    Bool,
-    Double,
-    ExecutionPlan,
-    Int,
-    KernelTypes,
-    Program,
-    Tensor,
-)
+
 from executorch.exir.tensor import get_scalar_type, scalar_type_enum, TensorSpec
 
 # pyre-ignore
-supported_program_type_table: Dict[Type[KernelTypes], ConfigValue] = {
-    Tensor: torch.Tensor,
-    Int: int,
-    Double: float,
-    Bool: bool,
+supported_program_type_table: Dict[Type[core_schema.KernelTypes], ConfigValue] = {
+    core_schema.Tensor: torch.Tensor,
+    core_schema.Int: int,
+    core_schema.Double: float,
+    core_schema.Bool: bool,
 }
 
 
-def emit_bundled_tensor(spec: TensorSpec, bundled_values: List[BundledValue]) -> None:
+def emit_bundled_tensor(
+    spec: TensorSpec, bundled_values: List[bp_schema.Value]
+) -> None:
     # QuantizedSchema in tensor has deprecated and may not be used anymore.
     # So here we don't emit it.
 
@@ -64,8 +52,8 @@ def emit_bundled_tensor(spec: TensorSpec, bundled_values: List[BundledValue]) ->
         tensor_data: bytes = bytes(spec_array)
 
     bundled_values.append(
-        BundledValue(
-            val=BundledTensor(
+        bp_schema.Value(
+            val=bp_schema.Tensor(
                 scalar_type=scalar_type_enum(spec.dtype),
                 sizes=spec.shape,
                 data=tensor_data,
@@ -75,55 +63,67 @@ def emit_bundled_tensor(spec: TensorSpec, bundled_values: List[BundledValue]) ->
     )
 
 
-def emit_prim(val: ConfigValue, bundled_values: List[BundledValue]):
+def emit_prim(val: ConfigValue, bundled_values: List[bp_schema.Value]):
     if type(val) == int:
-        bundled_values.append(BundledValue(val=BundledInt(int_val=val)))
+        bundled_values.append(bp_schema.Value(val=bp_schema.Int(int_val=val)))
     elif type(val) == bool:
-        bundled_values.append(BundledValue(val=BundledBool(bool_val=val)))
+        bundled_values.append(bp_schema.Value(val=bp_schema.Bool(bool_val=val)))
     elif type(val) == float:
-        bundled_values.append(BundledValue(val=BundledDouble(double_val=val)))
+        bundled_values.append(bp_schema.Value(val=bp_schema.Double(double_val=val)))
     else:
         assert 0, "Unsupported primitive type received."
 
 
-def get_program_input(program: Program, plan_idx: int, input_idx: int) -> KernelTypes:
+def get_program_input(
+    program: core_schema.Program, plan_idx: int, input_idx: int
+) -> core_schema.KernelTypes:
     return (
         program.execution_plan[plan_idx]
         .values[program.execution_plan[plan_idx].inputs[input_idx]]
         .val
     )
 
 
-def get_program_output(program: Program, plan_idx: int, output_idx: int) -> KernelTypes:
+def get_program_output(
+    program: core_schema.Program, plan_idx: int, output_idx: int
+) -> core_schema.KernelTypes:
     return (
         program.execution_plan[plan_idx]
         .values[program.execution_plan[plan_idx].outputs[output_idx]]
         .val
     )
 
 
-def get_input_dtype(program: Program, plan_idx: int, input_idx: int) -> torch.dtype:
+def get_input_dtype(
+    program: core_schema.Program, plan_idx: int, input_idx: int
+) -> torch.dtype:
     # pyre-fixme[16]: now assert all input and outputs is in tenor type. Support multuple datatypes in the future.
     return get_scalar_type(get_program_input(program, plan_idx, input_idx).scalar_type)
 
 
-def get_input_type(program: Program, plan_idx: int, input_idx: int) -> type:
-    type_lookup = {Int: int, Bool: bool, Double: float}
+def get_input_type(program: core_schema.Program, plan_idx: int, input_idx: int) -> type:
+    type_lookup = {
+        core_schema.Int: int,
+        core_schema.Bool: bool,
+        core_schema.Double: float,
+    }
     # pyre-fixme[6]: Incompatible parameter type [6]: In call `dict.__getitem__`, for 1st positional only parameter
-    # expected `Type[Union[Bool, Double, Int]]` but got `Type[Union[Bool, Double, Int, Tensor, BoolList, DoubleList,
+    # expected `Type[Union[core_schema.Bool, core_schema.Double, core_schema.Int]]` but got `Type[Union[core_schema.Bool, core_schema.Double, core_schema.Int, core_schema.Tensor, BoolList, DoubleList,
     # IntList, Null, OptionalTensorList, String, TensorList]]`.
     return type_lookup[type(get_program_input(program, plan_idx, input_idx))]
 
 
-def get_output_dtype(program: Program, plan_idx: int, output_idx: int) -> torch.dtype:
+def get_output_dtype(
+    program: core_schema.Program, plan_idx: int, output_idx: int
+) -> torch.dtype:
     return get_scalar_type(
         # pyre-ignore[16]: now assert all outputs is in tensor type.
         get_program_output(program, plan_idx, output_idx).scalar_type
     )
 
 
 def assert_valid_bundle(
-    program: Program,
+    program: core_schema.Program,
     bundled_config: BundledConfig,
 ) -> None:
     """Check if the program and BundledConfig matches each other.
@@ -145,7 +145,7 @@ def assert_valid_bundle(
         plan_test: ConfigExecutionPlanTest = bundled_config.execution_plan_tests[
             bp_plan_id
         ]
-        plan: ExecutionPlan = program.execution_plan[program_plan_id]
+        plan: core_schema.ExecutionPlan = program.execution_plan[program_plan_id]
 
         # User does not provide testcases for current plan, skip it
         if plan_test.method_name < plan.name:
@@ -167,7 +167,8 @@ def assert_valid_bundle(
         # Check if the type of Program's output is supported
         for index in range(len(plan.outputs)):
             assert (
-                type(get_program_output(program, program_plan_id, index)) == Tensor
+                type(get_program_output(program, program_plan_id, index))
+                == core_schema.Tensor
             ), "Only supports program with output in Tensor type."
 
         # Check if the I/O sets of each execution plan test match program's requirement.
@@ -239,10 +240,10 @@ def assert_valid_bundle(
 
 
 def create_bundled_program(
-    program: Program,
+    program: core_schema.Program,
     bundled_config: BundledConfig,
-) -> BundledProgram:
-    """Create BundledProgram by bundling the given program and bundled_config together.
+) -> bp_schema.BundledProgram:
+    """Create bp_schema.BundledProgram by bundling the given program and bundled_config together.
 
     Args:
         program: The program to be bundled.
@@ -251,16 +252,16 @@ def create_bundled_program(
 
     assert_valid_bundle(program, bundled_config)
 
-    execution_plan_tests: List[BundledExecutionPlanTest] = []
+    execution_plan_tests: List[bp_schema.BundledExecutionPlanTest] = []
 
     # Emit data and metadata of bundled tensor
     for plan_test in bundled_config.execution_plan_tests:
-        test_sets: List[BundledIOSet] = []
+        test_sets: List[bp_schema.BundledIOSet] = []
 
         # emit I/O sets for each execution plan test
         for i in range(len(plan_test.test_sets)):
-            inputs: List[BundledValue] = []
-            expected_outputs: List[BundledValue] = []
+            inputs: List[bp_schema.Value] = []
+            expected_outputs: List[bp_schema.Value] = []
 
             cur_plan_test_inputs = plan_test.test_sets[i].inputs
             cur_plan_test_expected_outputs = plan_test.test_sets[i].expected_outputs
@@ -285,19 +286,19 @@ def create_bundled_program(
                     expected_outputs,
                 )
             test_sets.append(
-                BundledIOSet(inputs=inputs, expected_outputs=expected_outputs)
+                bp_schema.BundledIOSet(inputs=inputs, expected_outputs=expected_outputs)
             )
 
         # emit the whole execution plan test
         execution_plan_tests.append(
-            BundledExecutionPlanTest(
+            bp_schema.BundledExecutionPlanTest(
                 method_name=plan_test.method_name, test_sets=test_sets
             )
         )
 
     program_bytes: bytes = _serialize_pte_binary(program)
 
-    return BundledProgram(
+    return bp_schema.BundledProgram(
         version=BUNDLED_PROGRAM_SCHEMA_VERSION,
         execution_plan_tests=execution_plan_tests,
         program=program_bytes,

bundled_program/schema.py

@@ -13,7 +13,7 @@
 
 
 @dataclass
-class BundledTensor:
+class Tensor:
     """All information we need to bundle for a tensor EValue input."""
 
     # The scalar type of Tensor
@@ -26,33 +26,33 @@ class BundledTensor:
 
 
 @dataclass
-class BundledInt:
+class Int:
     int_val: int
 
 
 @dataclass
-class BundledBool:
+class Bool:
     bool_val: bool
 
 
 @dataclass
-class BundledDouble:
+class Double:
     double_val: float
 
 
-BundledValueUnion = Union[
-    BundledTensor,
-    BundledInt,
-    BundledDouble,
-    BundledBool,
+ValueUnion = Union[
+    Tensor,
+    Int,
+    Double,
+    Bool,
 ]
 
 
 @dataclass
-class BundledValue:
+class Value:
     """Abstraction for BundledIOSet values"""
 
-    val: "BundledValueUnion"
+    val: "ValueUnion"
 
 
 @dataclass
@@ -61,12 +61,12 @@ class BundledIOSet:
 
     # All inputs required by Program for execution. Its length should be
     # equal to the length of program inputs.
-    inputs: List[BundledValue]
+    inputs: List[Value]
 
     # The expected outputs generated while running the model in eager mode
     # using the inputs provided. Its length should be equal to the length
     # of program outputs.
-    expected_outputs: List[BundledValue]
+    expected_outputs: List[Value]
 
 
 @dataclass

bundled_program/tests/test_bundle_data.py

@@ -9,40 +9,35 @@
 import unittest
 from typing import List
 
+import executorch.bundled_program.schema as bp_schema
+
 import torch
 from executorch.bundled_program.config import ConfigValue
 from executorch.bundled_program.core import create_bundled_program
-from executorch.bundled_program.schema import (
-    BundledBool,
-    BundledDouble,
-    BundledInt,
-    BundledTensor,
-    BundledValue,
-)
 from executorch.bundled_program.tests.common import get_common_program
 from executorch.exir._serialize import _serialize_pte_binary
 
 
 class TestBundle(unittest.TestCase):
     def assertIOsetDataEqual(
         self,
-        program_ioset_data: List[BundledValue],
+        program_ioset_data: List[bp_schema.Value],
         config_ioset_data: List[ConfigValue],
     ) -> None:
         self.assertEqual(len(program_ioset_data), len(config_ioset_data))
         for program_element, config_element in zip(
             program_ioset_data, config_ioset_data
         ):
-            if isinstance(program_element.val, BundledTensor):
+            if isinstance(program_element.val, bp_schema.Tensor):
                 # TODO: Update to check the bundled input share the same type with the config input after supporting multiple types.
                 self.assertTrue(isinstance(config_element, torch.Tensor))
                 self.assertEqual(program_element.val.sizes, list(config_element.size()))
                 # TODO(gasoonjia): Check the inner data.
-            elif type(program_element.val) == BundledInt:
+            elif type(program_element.val) == bp_schema.Int:
                 self.assertEqual(program_element.val.int_val, config_element)
-            elif type(program_element.val) == BundledDouble:
+            elif type(program_element.val) == bp_schema.Double:
                 self.assertEqual(program_element.val.double_val, config_element)
-            elif type(program_element.val) == BundledBool:
+            elif type(program_element.val) == bp_schema.Bool:
                 self.assertEqual(program_element.val.bool_val, config_element)
 
     def test_bundled_program(self) -> None: