Update base for Update on "[ExecuTorch] Add broadcasting support to optimized op_div"

Summary:
Similar to broadcast support in op_mul

Test Plan:
Tests added

Reviewers:

Subscribers:

Tasks:

Tags:

cc larryliu0820 manuelcandales

Differential Revision: [D69491815](https://our.internmc.facebook.com/intern/diff/D69491815)

[ghstack-poisoned]

Author: kimishpatel

backends/apple/coreml/TARGETS

@@ -76,6 +76,7 @@ runtime.cxx_python_extension(
     base_module = "",
     visibility = [
         "//executorch/examples/apple/coreml/...",
+        "@EXECUTORCH_CLIENTS",
     ],
     external_deps = [
         "pybind11",

backends/arm/_passes/TARGETS

@@ -7,6 +7,7 @@ python_library(
     deps = [
         "//executorch/backends/arm:tosa_quant_utils",
         "//executorch/backends/arm:tosa_utils",
+        "//executorch/backends/transforms:replace_scalar_with_tensor",
         "//executorch/backends/xnnpack/_passes:xnnpack_passes",
         "//executorch/exir:lib",
     ],

backends/arm/operator_support/TARGETS

@@ -5,8 +5,9 @@ python_library(
     srcs = glob(["*.py"]),
     typing = True,
     deps = [
+        "//executorch/backends/arm/_passes:passes",
+        "//executorch/backends/arm:tosa_specification",
         "//executorch/backends/xnnpack/_passes:xnnpack_passes",
         "//executorch/exir:lib",
-        "//executorch/backends/arm:tosa_specification"
     ],
 )

backends/cadence/aot/TARGETS

@@ -256,6 +256,7 @@ python_library(
         "//executorch/backends/cadence/aot:pass_utils",
         "//executorch/backends/cadence/aot:remove_ops",
         "//executorch/backends/cadence/aot:utils",
+        "//executorch/backends/transforms:replace_scalar_with_tensor",
         "//executorch/exir:pass_base",
         "//executorch/exir/dialects:lib",
         "//executorch/exir/dialects/edge:lib",

backends/cadence/aot/pass_utils.py

@@ -7,7 +7,7 @@
 # pyre-strict
 
 from dataclasses import dataclass
-from typing import Callable, List, Optional, Set, Union
+from typing import Callable, List, Optional, Set, Type, Union
 
 import torch
 from executorch.backends.cadence.aot.utils import get_edge_overload_packet
@@ -32,33 +32,33 @@ class CadencePassAttribute:
 
 
 # A dictionary that maps an ExportPass to its attributes.
-ALL_CADENCE_PASSES: dict[ExportPass, CadencePassAttribute] = {}
+ALL_CADENCE_PASSES: dict[Type[ExportPass], CadencePassAttribute] = {}
 
 
-def get_cadence_pass_attribute(p: ExportPass) -> CadencePassAttribute:
+def get_cadence_pass_attribute(p: Type[ExportPass]) -> CadencePassAttribute:
     return ALL_CADENCE_PASSES[p]
 
 
 # A decorator that registers a pass.
 def register_cadence_pass(
     pass_attribute: CadencePassAttribute,
-) -> Callable[[ExportPass], ExportPass]:
-    def wrapper(cls: ExportPass) -> ExportPass:
+) -> Callable[[Type[ExportPass]], Type[ExportPass]]:
+    def wrapper(cls: Type[ExportPass]) -> Type[ExportPass]:
         ALL_CADENCE_PASSES[cls] = pass_attribute
         return cls
 
     return wrapper
 
 
-def get_all_available_cadence_passes() -> Set[ExportPass]:
+def get_all_available_cadence_passes() -> Set[Type[ExportPass]]:
     return set(ALL_CADENCE_PASSES.keys())
 
 
 # Create a new filter to filter out relevant passes from all passes.
 def create_cadence_pass_filter(
     opt_level: int, debug: bool = False
-) -> Callable[[ExportPass], bool]:
-    def _filter(p: ExportPass) -> bool:
+) -> Callable[[Type[ExportPass]], bool]:
+    def _filter(p: Type[ExportPass]) -> bool:
         pass_attribute = get_cadence_pass_attribute(p)
         return (
             pass_attribute.opt_level is not None

backends/cadence/aot/passes.py

@@ -6,7 +6,7 @@
 
 # pyre-strict
 
-from typing import Any, List, Optional, Type
+from typing import Any, cast, List, Optional, Type
 
 import torch
 import torch.fx
@@ -95,9 +95,9 @@ def get_cadence_passes(
     passes = get_passes_in_default_order()
     pass_filter = create_cadence_pass_filter(opt_level)
     filtered_passes = [
-        # pyre-fixme[20]: Call `torch.fx.passes.infra.pass_base.PassBase.__call__` expects argument `graph_module`.
         filtered_pass()
         # pyre-fixme[6]: In call `filter.__new__` ... got `List[Type[typing.Callable[[GraphModule], Optional[PassResult]]]]`.
         for filtered_pass in list(filter(pass_filter, passes))
     ]
-    return filtered_passes
+    # The type checker can't infer the proper type of the list comprehension.
+    return cast(List[Optional[PassResult]], filtered_passes)

backends/cadence/aot/replace_ops.py

@@ -1719,9 +1719,9 @@ def call_operator(self, op, args, kwargs, meta):
         )
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))(
-    ReplaceScalarWithTensorArgPass()
-)
+register_cadence_pass(CadencePassAttribute(opt_level=0))(ReplaceScalarWithTensorArgPass)
+
+
 @register_cadence_pass(CadencePassAttribute(opt_level=0))
 class ReplaceScalarTensorWithFullPass(ExportPass):
     """

backends/transforms/targets.bzl

@@ -201,6 +201,20 @@ def define_common_targets():
         ],
     )
 
+    runtime.python_library(
+        name = "replace_scalar_with_tensor",
+        srcs = [
+            "replace_scalar_with_tensor.py",
+        ],
+        visibility = [
+            "//executorch/backends/...",
+        ],
+        deps = [
+            "//caffe2:torch",
+            "//executorch/exir:pass_base",
+        ],
+    )
+
     runtime.python_test(
         name = "test_duplicate_dynamic_quant_chain",
         srcs = [

backends/vulkan/runtime/graph/ops/impl/Convolution.cpp

@@ -475,7 +475,12 @@ void add_conv1d_node(
     const ValueRef out,
     const bool clamp_out) {
   ValueRef arg_weight = prepack_standard(
-      graph, weight, graph.storage_type_of(out), utils::kChannelsPacked);
+      graph,
+      weight,
+      graph.storage_type_of(out),
+      utils::kChannelsPacked,
+      /* passthrough = */ false,
+      utils::kOptimizedAxisMap);
   ValueRef arg_bias = prepack_biases(
       graph,
       bias,

build/Utils.cmake

@@ -357,7 +357,7 @@ function(add_torch_to_cmake_prefix_path)
   endif()
   execute_process(
     COMMAND "${PYTHON_EXECUTABLE}" -c
-            "import torch as _; print(_.__path__[0], end='')"
+            "import importlib.util; print(importlib.util.find_spec('torch').submodule_search_locations[0])"
     OUTPUT_VARIABLE _tmp_torch_path
     ERROR_VARIABLE _tmp_torch_path_error
     RESULT_VARIABLE _tmp_torch_path_result COMMAND_ECHO STDERR