Arm backend: Support scalar_tensor op

backends/arm/_passes/fuse_constant_ops_pass.py

@@ -161,6 +161,7 @@ def f(node_name_pre_computed):
         exir_ops.edge.aten.arange.start_step,
         exir_ops.edge.aten.eye.default,
         exir_ops.edge.aten.linspace.default,
+        torch.ops.aten.scalar_tensor.default,
     ]
 
     def __init__(self, exported_program: ExportedProgram) -> None:

backends/arm/operator_support/tosa_supported_operators.py

@@ -205,6 +205,7 @@ def is_node_supported(
             exir_ops.edge.aten.amin.default,
             exir_ops.edge.aten.eye.default,
             exir_ops.edge.aten.linspace.default,
+            torch.ops.aten.scalar_tensor.default,
         ]
 
         return supported

backends/arm/test/models/test_conformer.py

@@ -35,7 +35,6 @@ class TestConformer(unittest.TestCase):
         "executorch_exir_dialects_edge__ops_aten_where_self": 4,
         "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": 6,
     }
 
@@ -92,7 +91,7 @@ def test_conformer_tosa_BI(self):
             )
         )
 
-    @conftest.expectedFailureOnFVP  # TODO(MLETORCH-635)
+    @unittest.expectedFailure  # TODO(MLETORCH-635)
     def test_conformer_u55_BI(self):
         tester = (
             ArmTester(
@@ -114,7 +113,7 @@ def test_conformer_u55_BI(self):
                 inputs=get_test_inputs(self.dim, self.lengths, self.num_examples),
             )
 
-    @conftest.expectedFailureOnFVP  # TODO(MLETORCH-635)
+    @unittest.expectedFailure  # TODO(MLETORCH-635)
     def test_conformer_u85_BI(self):
         tester = (
             ArmTester(

backends/arm/test/ops/test_scalar_tensor.py

@@ -0,0 +1,91 @@
+# 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 torch
+from executorch.backends.arm.test import common
+
+from executorch.backends.arm.test.tester.test_pipeline import (
+    EthosU55PipelineBI,
+    EthosU85PipelineBI,
+    TosaPipelineBI,
+    TosaPipelineMI,
+)
+
+float_test_data_suite = {
+    "scalar_tensor_float_1": (3.7, torch.float32, torch.rand((1, 2, 3, 4))),
+    "scalar_tensor_float_2": (66, torch.float32, torch.rand((1, 2, 3))),
+}
+
+int_test_data_suite = {
+    "scalar_tensor_int32": (
+        33,
+        torch.int32,
+        torch.randint(0, 10, (1, 2), dtype=torch.int32),
+    ),
+    "scalar_tensor_int8": (
+        8,
+        torch.int8,
+        torch.rand(1, 2, 3),
+    ),
+    "scalar_tensor_int16": (
+        16 * 16 * 16,
+        torch.int16,
+        torch.rand((1,)).unsqueeze(0),  # Rank 0 inputs not supported
+    ),
+}
+
+
+class ScalarTensor(torch.nn.Module):
+    aten_op = "torch.ops.aten.scalar_tensor.default"
+
+    def __init__(self, scalar, dtype=torch.float32):
+        super().__init__()
+        self.scalar = scalar
+        self.dtype = dtype
+
+    def forward(self, x: torch.Tensor):
+        return torch.scalar_tensor(self.scalar, dtype=self.dtype) + x
+
+
+@common.parametrize("test_data", int_test_data_suite | float_test_data_suite)
+def test_scalar_tensor_tosa_MI(test_data):  # Note TOSA MI supports all types
+    scalar, dtype, data = test_data
+    TosaPipelineMI(ScalarTensor(scalar, dtype), tuple(data), ScalarTensor.aten_op).run()
+
+
+@common.parametrize("test_data", int_test_data_suite | float_test_data_suite)
+def test_scalar_tensor_tosa_BI(test_data):
+    scalar, dtype, data = test_data
+    pipeline: TosaPipelineBI = TosaPipelineBI(
+        ScalarTensor(scalar, dtype), tuple(data), ScalarTensor.aten_op
+    )
+    pipeline.pop_stage("check.quant_nodes")
+    pipeline.run()
+
+
+@common.parametrize("test_data", float_test_data_suite)
+@common.XfailIfNoCorstone300
+def test_scalar_tensor_tosa_u55(test_data):
+    scalar, dtype, data = test_data
+    EthosU55PipelineBI(
+        ScalarTensor(scalar, dtype),
+        tuple(data),
+        ScalarTensor.aten_op,
+        symmetric_io_quantization=True,
+        run_on_fvp=True,
+    ).run()
+
+
+@common.parametrize("test_data", float_test_data_suite)
+@common.XfailIfNoCorstone320
+def test_scalar_tensor_tosa_u85(test_data):
+    scalar, dtype, data = test_data
+    EthosU85PipelineBI(
+        ScalarTensor(scalar, dtype),
+        tuple(data),
+        ScalarTensor.aten_op,
+        symmetric_io_quantization=True,
+        run_on_fvp=True,
+    ).run()