Add 16A8W support and test for tanh operation

Pull Request resolved: #13797

Add 16A8W quantization support and test for the tanh operation in ExecutorTorch ARM backend.

This follows the pattern established for linear, mul, and sigmoid operations, extending int16 support to tanh operations.

Changes:
- Add INT16 dtype validation support in op_tanh.py
- Add test_tanh_tensor_16a8w_tosa_INT test function
- Enable test_tanh.py in test targets configuration

The 16A8W configuration uses 16-bit activations with 8-bit weights, enabling higher precision for activations while maintaining weight efficiency.
ghstack-source-id: 308986671
@exported-using-ghexport

Differential Revision: [D80510815](https://our.internmc.facebook.com/intern/diff/D80510815/)

Author: Ninja91

backends/arm/test/ops/test_tanh.py

@@ -6,16 +6,23 @@
 
 from typing import Tuple
 
+import pytest
 import torch
+from executorch.backends.arm.quantizer.arm_quantizer import (
+    get_symmetric_a16w8_quantization_config,
+    TOSAQuantizer,
+)
 
-from executorch.backends.arm.test import common
+from executorch.backends.arm.test import common, conftest
 from executorch.backends.arm.test.tester.test_pipeline import (
     EthosU55PipelineINT,
     EthosU85PipelineINT,
     TosaPipelineFP,
     TosaPipelineINT,
     VgfPipeline,
 )
+from executorch.backends.arm.tosa.specification import TosaSpecification
+from executorch.backends.xnnpack.test.tester import Quantize
 
 aten_op = "torch.ops.aten.tanh.default"
 input_t1 = Tuple[torch.Tensor]  # Input x
@@ -105,3 +112,105 @@ def test_tanh_vgf_INT(test_data: Tuple):
         tosa_version="TOSA-1.0+INT",
     )
     pipeline.run()
+
+
+def get_symmetric_a16w8_tanh_quantizer(per_channel_quantization=False):
+    tosa_version = conftest.get_option("tosa_version")
+    tosa_profiles = {
+        "1.0": TosaSpecification.create_from_string("TOSA-1.0+INT+int16"),
+    }
+
+    quantizer = TOSAQuantizer(tosa_profiles[tosa_version])
+    quantizer.set_global(
+        get_symmetric_a16w8_quantization_config(is_per_channel=per_channel_quantization)
+    )
+
+    return Quantize(
+        quantizer,
+        get_symmetric_a16w8_quantization_config(
+            is_per_channel=per_channel_quantization
+        ),
+    )
+
+
+@common.parametrize("test_data", test_data_suite)
+@pytest.mark.xfail(
+    reason="missing int16 tanh ops support; fails at TOSA reference model with Unsupported operation type or rank. See: https://github.com/pytorch/executorch/issues/13975"
+)
+def test_tanh_16a8w_tosa_INT(test_data: torch.Tensor):
+    """Test tanh operation with 16A8W quantization (16-bit activations, 8-bit weights)"""
+    per_channel_quantization = False
+
+    pipeline = TosaPipelineINT[input_t1](
+        Tanh(),
+        (test_data(),),
+        aten_op,
+        exir_op=[],
+        per_channel_quantization=per_channel_quantization,
+        use_to_edge_transform_and_lower=True,
+        tosa_extensions=["int16"],
+    )
+
+    pipeline.change_args(
+        "quantize",
+        get_symmetric_a16w8_tanh_quantizer(
+            per_channel_quantization=per_channel_quantization
+        ),
+    )
+    pipeline.run()
+
+
+@common.parametrize("test_data", test_data_suite)
+@common.XfailIfNoCorstone300
+@pytest.mark.xfail(
+    reason="Vela compilation fails with 'Invalid arguments' for int16 tanh operations"
+)
+def test_tanh_16a8w_u55_INT16(test_data: torch.Tensor):
+    """Test tanh operation with 16A8W quantization on U55 (16-bit activations, 8-bit weights)"""
+    per_channel_quantization = False
+
+    pipeline = EthosU55PipelineINT[input_t1](
+        Tanh(),
+        (test_data(),),
+        aten_op,
+        exir_ops=[],
+        per_channel_quantization=per_channel_quantization,
+        use_to_edge_transform_and_lower=True,
+        run_on_fvp=True,
+    )
+
+    pipeline.change_args(
+        "quantize",
+        get_symmetric_a16w8_tanh_quantizer(
+            per_channel_quantization=per_channel_quantization
+        ),
+    )
+    pipeline.run()
+
+
+@common.parametrize("test_data", test_data_suite)
+@common.XfailIfNoCorstone320
+@pytest.mark.xfail(
+    reason="Vela compilation fails with 'Invalid arguments' for int16 tanh operations"
+)
+def test_tanh_16a8w_u85_INT16(test_data: torch.Tensor):
+    """Test tanh operation with 16A8W quantization on U85 (16-bit activations, 8-bit weights)"""
+    per_channel_quantization = False
+
+    pipeline = EthosU85PipelineINT[input_t1](
+        Tanh(),
+        (test_data(),),
+        aten_op,
+        exir_ops=[],
+        per_channel_quantization=per_channel_quantization,
+        use_to_edge_transform_and_lower=True,
+        run_on_fvp=True,
+    )
+
+    pipeline.change_args(
+        "quantize",
+        get_symmetric_a16w8_tanh_quantizer(
+            per_channel_quantization=per_channel_quantization
+        ),
+    )
+    pipeline.run()