[Arm tester] Run delegate nodes using tosa_reference_model

Instead of executing just one delegate,
we execute the graph and dispatch the
delegate nodes to the tosa_reference_model.
This makes the tester more flexible and
enables running tests with multiple
delegates, multiple outputs etc.

Signed-off-by: Erik Lundell <[email protected]>
Change-Id: I5fb0192da2d187f29c025e0c51238a9cb7d7ec90

Author: Erik-Lundell

backends/arm/_passes/tag_io_quant_pass.py

@@ -1,4 +1,4 @@
-# Copyright 2023-2024 Arm Limited and/or its affiliates.
+# Copyright 2023-2025 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.
@@ -202,6 +202,20 @@ def is_tosa(compile_spec: List[CompileSpec]) -> bool:
     return False
 
 
+def is_quantize_io(compile_specs: List[CompileSpec]) -> bool:
+    for spec in compile_specs:
+        if spec.key == "quantize_io" and spec.value.decode() == "True":
+            return True
+    return False
+
+
+def get_tosa_version(compile_spec: List[CompileSpec]) -> TosaSpecification:
+    for spec in compile_spec:
+        if spec.key == "tosa_version":
+            return TosaSpecification.create_from_string(spec.value.decode())
+    raise RuntimeError("Could not find TOSA version in CompileSpec")
+
+
 def get_intermediate_path(compile_spec: List[CompileSpec]) -> Optional[str]:
     for spec in compile_spec:
         if spec.key == "debug_artifact_path":

backends/arm/arm_backend.py

@@ -1,4 +1,4 @@
-# Copyright 2023-2024 Arm Limited and/or its affiliates.
+# Copyright 2023-2025 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.
@@ -10,8 +10,10 @@
 from typing import Callable, final, List, Optional, Tuple
 
 import torch
-from executorch.backends.arm.arm_backend import ArmBackend  # usort: skip
-from executorch.backends.arm._passes.tag_io_quant_pass import TagIOQuantPass
+from executorch.backends.arm.arm_backend import (
+    ArmBackend,
+    is_quantize_io,
+)  # usort: skip
 from executorch.backends.arm.operator_support.tosa_supported_operators import (
     TOSASupportedOperators,
 )
@@ -23,7 +25,7 @@
     PartitionResult,
 )
 from executorch.exir.backend.utils import tag_constant_data
-from executorch.exir.passes import PassManager
+from executorch.exir.dialects._ops import ops as exir_ops
 from torch.export.exported_program import ExportedProgram
 from torch.fx.passes.infra.partitioner import CapabilityBasedPartitioner
 
@@ -35,6 +37,22 @@
     logger.setLevel(logging.INFO)
 
 
+def is_quant_node(node: torch.fx.node.Node) -> bool:
+    return node.target in {
+        exir_ops.edge.quantized_decomposed.quantize_per_channel.default,
+        exir_ops.edge.quantized_decomposed.quantize_per_tensor.default,
+        exir_ops.edge.quantized_decomposed.quantize_per_tensor.tensor,
+    }
+
+
+def is_dequant_node(node: torch.fx.node.Node) -> bool:
+    return node.target in {
+        exir_ops.edge.quantized_decomposed.dequantize_per_channel.default,
+        exir_ops.edge.quantized_decomposed.dequantize_per_tensor.default,
+        exir_ops.edge.quantized_decomposed.dequantize_per_tensor.tensor,
+    }
+
+
 @final
 class ArmPartitioner(Partitioner):
     def __init__(self, compile_spec: List[CompileSpec]) -> None:
@@ -43,6 +61,7 @@ def __init__(self, compile_spec: List[CompileSpec]) -> None:
     def partition(self, exported_program: ExportedProgram) -> PartitionResult:
         # Run the CapabilityBasedPartitioner to return the largest possible
         # subgraphs containing the nodes with the tags
+
         logger.info("ArmPartitioner::partition")
         partition_tags = {}
 
@@ -52,28 +71,42 @@ def partition(self, exported_program: ExportedProgram) -> PartitionResult:
 
         logger.info(f"Partitioning for {tosa_spec}")
 
-        for spec in self.delegation_spec.compile_specs:
-            if spec.key == "quantize_io" and spec.value.decode() == "True":
-                # Exclude IO quantization from the partition
-                passes = PassManager(
-                    passes=[
-                        TagIOQuantPass(),
-                    ]
-                )
-                passes(exported_program.graph_module)
-
         capability_partitioner = CapabilityBasedPartitioner(
             exported_program.graph_module,
             TOSASupportedOperators(tosa_spec),
             allows_single_node_partition=True,
         )
         partition_list = capability_partitioner.propose_partitions()
         for partition in partition_list:
+            tag = f"tag{partition.id}"
+
+            def is_partitioned(node: torch.fx.Node, tag=tag) -> bool:
+                return (
+                    "delegation_tag" in node.meta and node.meta["delegation_tag"] == tag
+                )
+
             for node in partition.nodes:
-                tag = f"tag{partition.id}"
                 node.meta["delegation_tag"] = tag
                 partition_tags[tag] = self.delegation_spec
 
+            if not is_quantize_io(self.delegation_spec.compile_specs):
+                continue
+
+            # De-tag outmost q-nodes upwards and dq-nodes downwards.
+            # De-tag if at least one input/ output is not part of partition.
+            for node in partition.nodes:
+                if is_quant_node(node):
+                    for input in node.all_input_nodes:
+                        if not is_partitioned(input):
+                            del node.meta["delegation_tag"]
+                            break
+
+                if is_dequant_node(node):
+                    for user in node.users:
+                        if not is_partitioned(user):
+                            del node.meta["delegation_tag"]
+                            break
+
         tag_constant_data(exported_program)
 
         return PartitionResult(

backends/arm/arm_partitioner.py

@@ -1,4 +1,4 @@
-# Copyright 2024 Arm Limited and/or its affiliates.
+# Copyright 2024-2025 Arm Limited and/or its affiliates.
 # All rights reserved.
 #
 # This source code is licensed under the BSD-style license found in the
@@ -50,7 +50,6 @@ def maybe_get_tosa_collate_path() -> str | None:
             tosa_test_base = os.path.join(tosa_test_base, "tosa-mi")
         else:
             tosa_test_base = os.path.join(tosa_test_base, "other")
-
         return os.path.join(tosa_test_base, test_class, test_name)
 
     return None
@@ -83,6 +82,7 @@ def get_tosa_compile_spec_unbuilt(
         .tosa_compile_spec(tosa_version)
         .set_permute_memory_format(permute_memory_to_nhwc)
         .dump_intermediate_artifacts_to(custom_path)
+        .set_quantize_io(True)
     )
 
     return compile_spec_builder

backends/arm/test/common.py

@@ -0,0 +1,57 @@
+# Copyright 2024-2025 Arm Limited and/or its affiliates.
+# All rights reserved.
+#
+# 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
+from executorch.backends.arm.test import common
+from executorch.backends.arm.test.tester.arm_tester import ArmTester
+
+
+class TestMultipleDelegates(unittest.TestCase):
+    class MultipleDelegatesModule(torch.nn.Module):
+        inputs = (torch.randn(10, 4, 5), torch.randn(10, 4, 5))
+
+        def get_inputs(self):
+            return self.inputs
+
+        def forward(self, x: torch.Tensor, y: torch.Tensor):
+            z = x + y
+            s = torch.sin(z)
+            return s * z
+
+    def test_tosa_MI(self):
+        module = self.MultipleDelegatesModule()
+        inputs = module.get_inputs()
+        (
+            ArmTester(
+                module,
+                example_inputs=inputs,
+                compile_spec=common.get_tosa_compile_spec("TOSA-0.80+MI"),
+            )
+            .export()
+            .to_edge_transform_and_lower()
+            .check_count({"torch.ops.higher_order.executorch_call_delegate": 2})
+            .to_executorch()
+            .run_method_and_compare_outputs(inputs=inputs)
+        )
+
+    def test_tosa_BI(self):
+        module = self.MultipleDelegatesModule()
+        inputs = module.get_inputs()
+        (
+            ArmTester(
+                module,
+                example_inputs=inputs,
+                compile_spec=common.get_tosa_compile_spec("TOSA-0.80+BI"),
+            )
+            .quantize()
+            .export()
+            .to_edge_transform_and_lower()
+            .check_count({"torch.ops.higher_order.executorch_call_delegate": 2})
+            .to_executorch()
+            .run_method_and_compare_outputs(inputs=inputs, qtol=1.0)
+        )

backends/arm/test/misc/test_multiple_delegates.py

@@ -0,0 +1,53 @@
+# Copyright 2024-2025 Arm Limited and/or its affiliates.
+# All rights reserved.
+#
+# 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
+from executorch.backends.arm.test import common
+from executorch.backends.arm.test.tester.arm_tester import ArmTester
+
+
+class TestMultipleOutputs(unittest.TestCase):
+    class MultipleOutputsModule(torch.nn.Module):
+        inputs = (torch.randn(10, 4, 5), torch.randn(10, 4, 5))
+
+        def get_inputs(self):
+            return self.inputs
+
+        def forward(self, x: torch.Tensor, y: torch.Tensor):
+            return (x * y, x.sum(dim=-1, keepdim=True))
+
+    def test_tosa_MI_pipeline(self):
+        module = self.MultipleOutputsModule()
+        inputs = module.get_inputs()
+        (
+            ArmTester(
+                module,
+                example_inputs=inputs,
+                compile_spec=common.get_tosa_compile_spec("TOSA-0.80+MI"),
+            )
+            .export()
+            .to_edge_transform_and_lower()
+            .to_executorch()
+            .run_method_and_compare_outputs(inputs=inputs)
+        )
+
+    def test_tosa_BI_pipeline(self):
+        module = self.MultipleOutputsModule()
+        inputs = module.get_inputs()
+        (
+            ArmTester(
+                module,
+                example_inputs=inputs,
+                compile_spec=common.get_tosa_compile_spec("TOSA-0.80+BI"),
+            )
+            .quantize()
+            .export()
+            .to_edge_transform_and_lower()
+            .to_executorch()
+            .run_method_and_compare_outputs(inputs=inputs, qtol=1.0)
+        )