Arm: Move ReplaceScalarWithTensorArgPass to transforms (#8519)

The pass is general and can be used by multiple backends.
Use it in Arm backend and make small adjustments to make
it work.

Signed-off-by: Erik Lundell <[email protected]>

Author: Erik-Lundell

backends/arm/_passes/arm_pass_manager.py

@@ -77,6 +77,9 @@
 )
 from executorch.backends.arm.tosa_specification import TosaSpecification
 
+from executorch.backends.transforms.replace_scalar_with_tensor import (
+    ReplaceScalarWithTensorArgPass,
+)
 from executorch.backends.xnnpack._passes.remove_getitem_op import RemoveGetItemPass
 from executorch.exir import ExportedProgram
 from executorch.exir.pass_manager import PassManager
@@ -102,6 +105,7 @@ def _tosa_080_BI_pipeline(self, exported_program: ExportedProgram) -> GraphModul
         self.add_pass(ConvertMeanDimToAveragePoolPass())
         self.add_pass(ConvertFullLikeToFullPass())
 
+        self.add_pass(ReplaceScalarWithTensorArgPass())
         self.add_pass(AnnotateDecomposedMatmulPass())
         self.add_pass(QuantizeOperatorArguments())
         self.add_pass(FoldAndAnnotateQParamsPass())  # type: ignore[call-arg]
@@ -125,7 +129,7 @@ def _tosa_080_BI_pipeline(self, exported_program: ExportedProgram) -> GraphModul
         return self._transform(exported_program.graph_module)
 
     def _tosa_080_MI_pipeline(self, exported_program: ExportedProgram) -> GraphModule:
-
+        self.add_pass(ReplaceScalarWithTensorArgPass())
         self.add_pass(FuseQuantizedActivationPass())
         self.add_pass(RemoveGetItemPass())
         self.add_pass(ConvertSplitToSlicePass())
@@ -176,6 +180,7 @@ def transform_to_backend_pipeline(self, exported_program: ExportedProgram):
 
     def transform_for_annotation_pipeline(self, graph_module: GraphModule):
         self.add_pass(ScalarsToAttributePass())
+        self.add_pass(ReplaceScalarWithTensorArgPass())
         self.add_pass(DecomposeLayerNormPass())
         self.add_pass(DecomposeVarPass())
         self.add_pass(DecomposeMeanDimPass())

backends/arm/operator_support/tosa_supported_operators.py

@@ -113,6 +113,10 @@ def is_node_supported(self, submodules, node: fx.Node) -> bool:
             exir_ops.edge.aten.le.Tensor,
             exir_ops.edge.aten.lt.Tensor,
             exir_ops.edge.aten.mul.Tensor,
+            exir_ops.edge.aten.add.Scalar,
+            exir_ops.edge.aten.sub.Scalar,
+            exir_ops.edge.aten.mul.Scalar,
+            exir_ops.edge.aten.div.Scalar,
             exir_ops.edge.aten._native_batch_norm_legit_no_training.default,
             exir_ops.edge.aten.native_layer_norm.default,
             exir_ops.edge.aten.sigmoid.default,

backends/arm/test/models/test_conformer.py

@@ -32,15 +32,14 @@ class TestConformer(unittest.TestCase):
     ops_after_partitioner = {
         "executorch_exir_dialects_edge__ops_aten_arange_start_step": 1,
         "executorch_exir_dialects_edge__ops_aten_max_default": 1,
-        "executorch_exir_dialects_edge__ops_aten_mul_Scalar": 4,
         "executorch_exir_dialects_edge__ops_aten_eq_Scalar": 2,
         "executorch_exir_dialects_edge__ops_aten_where_self": 4,
         "executorch_exir_dialects_edge__ops_aten_logical_not_default": 4,
         "executorch_exir_dialects_edge__ops_aten_any_dim": 2,
         "torch.ops.aten._assert_scalar.default": 10,
         "torch.ops.aten._local_scalar_dense.default": 1,
         "torch.ops.aten.scalar_tensor.default": 2,
-        "torch.ops.higher_order.executorch_call_delegate": 5,
+        "torch.ops.higher_order.executorch_call_delegate": 4,
     }
 
     dim = 16

backends/arm/test/ops/test_scalars.py

@@ -1,3 +1,8 @@
+# Copyright 2024-2025 Arm Limited and/or its affiliates.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
 import unittest
 
 import torch
@@ -50,6 +55,22 @@ class Mul(torch.nn.Module):
         def forward(self, x, y):
             return x * y
 
+    class MulScalar(torch.nn.Module):
+        def forward(self, x, y):
+            return torch.ops.aten.mul.Scalar(x, y)
+
+    class DivScalar(torch.nn.Module):
+        def forward(self, x, y):
+            return torch.ops.aten.div.Scalar(x, y)
+
+    class AddScalar(torch.nn.Module):
+        def forward(self, x, y):
+            return torch.ops.aten.add.Scalar(x, y)
+
+    class SubScalar(torch.nn.Module):
+        def forward(self, x, y):
+            return torch.ops.aten.sub.Scalar(x, y)
+
     class AddInplace(torch.nn.Module):
         def forward(self, x, y):
             x += y
@@ -91,6 +112,10 @@ def forward(self, x):
         ("Sub_", SubInplace()),
         ("Mul_", MulInplace()),
         ("Div_", DivInplace()),
+        ("MulScalar", MulScalar()),
+        ("DivScalar", DivScalar()),
+        ("AddScalar", AddScalar()),
+        ("SubScalar", SubScalar()),
     ]
 
     const_ops = [("Add", AddConst())]
@@ -108,8 +133,8 @@ def forward(self, x):
                 scalar = dtype[1]
                 tensor_scalar_tests.append((test_name + "_ts", op[1], tensor, scalar))
 
-                # Don't add (scalar, tensor) test case for inplace ops.
-                if op[0][-1] == "_":
+                # Don't add (scalar, tensor) test case for inplace and .Scalar ops.
+                if op[0][-1] == "_" or op[0][-6:] == "Scalar":
                     continue
 
                 # sub(scalar, tensor) does not work in any case.

backends/arm/tosa_mapping.py

@@ -11,6 +11,8 @@
 # the standardised TOSA representation.
 #
 
+from typing import Sequence
+
 import serializer.tosa_serializer as ts  # type: ignore
 import torch
 
@@ -99,7 +101,7 @@ def __init__(self, argument) -> None:
         if isinstance(argument, torch.fx.Node):
             self.__process_node(argument)
             return
-        if isinstance(argument, list):
+        if isinstance(argument, Sequence):
             self.__process_list(argument)
             return
         if isinstance(argument, (int, float)):

backends/cadence/aot/replace_ops.py

@@ -1,5 +1,6 @@
 # Copyright (c) Meta Platforms, Inc. and affiliates.
 # All rights reserved.
+# Copyright 2025 Arm Limited and/or its affiliates.
 #
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
@@ -37,6 +38,9 @@
 )
 from executorch.backends.cadence.aot.remove_ops import RemoveNopSelectOpPass
 from executorch.backends.cadence.aot.utils import get_edge_overload_packet
+from executorch.backends.transforms.replace_scalar_with_tensor import (
+    ReplaceScalarWithTensorArgPass,
+)
 from executorch.exir.dialects._ops import ops as exir_ops
 from executorch.exir.dialects.edge._ops import EdgeOpOverload, EdgeOpOverloadPacket
 from executorch.exir.pass_base import ExportPass, NodeMetadata, PassResult, ProxyValue
@@ -1713,65 +1717,9 @@ def call_operator(self, op, args, kwargs, meta):
         )
 
 
-@register_cadence_pass(CadencePassAttribute(opt_level=0))
-class ReplaceScalarWithTensorArgPass(ExportPass):
-    """
-    For binary ops like add.Scalar, sub.Scalar mul.Scalar, and div.Scalar,
-    replace the scalar arg with Tensor arg.
-    """
-
-    scalar_to_tensor_ops: Dict[EdgeOpOverload, EdgeOpOverload] = {
-        exir_ops.edge.aten.add.Scalar: exir_ops.edge.aten.add.Tensor,
-        exir_ops.edge.aten.sub.Scalar: exir_ops.edge.aten.sub.Tensor,
-        exir_ops.edge.aten.mul.Scalar: exir_ops.edge.aten.mul.Tensor,
-        exir_ops.edge.aten.div.Scalar: exir_ops.edge.aten.div.Tensor,
-    }
-
-    def get_replacement(self, op, args, kwargs, meta):
-        return super().call_operator(
-            # Replace with .Tensor variant.
-            op=self.scalar_to_tensor_ops[op],
-            args=(
-                # Tensor arg.
-                args[0],
-                # Scalar arg - replace with aten.full tensor.
-                super().call_operator(
-                    exir_ops.edge.aten.full.default,
-                    args=(
-                        (1,),
-                        args[1],
-                    ),
-                    kwargs={"dtype": args[0].to_tensor().dtype},
-                    meta=meta,
-                ),
-                # Other args.
-                *args[2:],
-            ),
-            kwargs=kwargs,
-            meta=meta,
-        )
-
-    def call_operator(self, op, args, kwargs, meta):
-        if op not in self.scalar_to_tensor_ops:
-            return super().call_operator(op, args, kwargs, meta)
-
-        # There must be exactly 2 args (3 for add and sub containing alpha)
-        assert len(args) == 2 or len(args) == 3
-
-        # If there are two args, just replace the op.
-        if len(args) == 2:
-            return self.get_replacement(op, args, kwargs, meta)
-
-        # In case the op has three args, it must be scalar add/sub op.
-        if (
-            op not in {exir_ops.edge.aten.add.Scalar, exir_ops.edge.aten.sub.Scalar}
-            or "alpha" in kwargs
-        ):
-            return super().call_operator(op, args, kwargs, meta)
-
-        return self.get_replacement(op, args, kwargs, meta)
-
-
+@register_cadence_pass(CadencePassAttribute(opt_level=0))(
+    ReplaceScalarWithTensorArgPass()
+)
 @register_cadence_pass(CadencePassAttribute(opt_level=0))
 class ReplaceScalarTensorWithFullPass(ExportPass):
     """

backends/transforms/replace_scalar_with_tensor.py

@@ -0,0 +1,75 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+# All rights reserved.
+# Copyright 2025 Arm Limited and/or its affiliates.
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+
+from typing import Dict
+
+import torch
+from executorch.exir.dialects._ops import ops as exir_ops
+from executorch.exir.dialects.edge._ops import EdgeOpOverload
+from executorch.exir.pass_base import ExportPass
+
+
+class ReplaceScalarWithTensorArgPass(ExportPass):
+    """
+    For binary ops like add.Scalar, sub.Scalar mul.Scalar, and div.Scalar,
+    replace the scalar arg with Tensor arg.
+    """
+
+    scalar_to_tensor_ops: Dict[EdgeOpOverload, EdgeOpOverload] = {
+        exir_ops.edge.aten.add.Scalar: exir_ops.edge.aten.add.Tensor,
+        exir_ops.edge.aten.sub.Scalar: exir_ops.edge.aten.sub.Tensor,
+        exir_ops.edge.aten.mul.Scalar: exir_ops.edge.aten.mul.Tensor,
+        exir_ops.edge.aten.div.Scalar: exir_ops.edge.aten.div.Tensor,
+        torch.ops.aten.add.Scalar: torch.ops.aten.add.Tensor,
+        torch.ops.aten.sub.Scalar: torch.ops.aten.sub.Tensor,
+        torch.ops.aten.mul.Scalar: torch.ops.aten.mul.Tensor,
+        torch.ops.aten.div.Scalar: torch.ops.aten.div.Tensor,
+    }
+
+    def get_replacement(self, op, args, kwargs, meta):
+        return super().call_operator(
+            # Replace with .Tensor variant.
+            op=self.scalar_to_tensor_ops[op],
+            args=(
+                # Tensor arg.
+                args[0],
+                # Scalar arg - replace with aten.full tensor.
+                super().call_operator(
+                    exir_ops.edge.aten.full.default,
+                    args=(
+                        (1,),
+                        args[1],
+                    ),
+                    kwargs={"dtype": args[0].to_tensor().dtype},
+                    meta=meta,
+                ),
+                # Other args.
+                *args[2:],
+            ),
+            kwargs=kwargs,
+            meta=meta,
+        )
+
+    def call_operator(self, op, args, kwargs, meta):
+        if op not in self.scalar_to_tensor_ops:
+            return super().call_operator(op, args, kwargs, meta)
+
+        # There must be exactly 2 args (3 for add and sub containing alpha)
+        assert len(args) == 2 or len(args) == 3
+
+        # If there are two args, just replace the op.
+        if len(args) == 2:
+            return self.get_replacement(op, args, kwargs, meta)
+
+        # In case the op has three args, it must be scalar add/sub op.
+        if (
+            op not in {exir_ops.edge.aten.add.Scalar, exir_ops.edge.aten.sub.Scalar}
+            or "alpha" in kwargs
+        ):
+            return super().call_operator(op, args, kwargs, meta)
+
+        return self.get_replacement(op, args, kwargs, meta)