|  | 
|  | 1 | +# Copyright 2025 Arm Limited and/or its affiliates. | 
|  | 2 | +# | 
|  | 3 | +# This source code is licensed under the BSD-style license found in the | 
|  | 4 | +# LICENSE file in the root directory of this source tree. | 
|  | 5 | + | 
|  | 6 | +from typing import Tuple | 
|  | 7 | + | 
|  | 8 | +import torch | 
|  | 9 | +import torch.nn as nn | 
|  | 10 | + | 
|  | 11 | +from executorch.backends.arm.test import common | 
|  | 12 | +from executorch.backends.arm.test.tester.test_pipeline import ( | 
|  | 13 | +    EthosU55PipelineINT, | 
|  | 14 | +    EthosU85PipelineINT, | 
|  | 15 | +    TosaPipelineFP, | 
|  | 16 | +    TosaPipelineINT, | 
|  | 17 | +    VgfPipeline, | 
|  | 18 | +) | 
|  | 19 | + | 
|  | 20 | +test_data_suite = { | 
|  | 21 | +    # (test_name, test_data) | 
|  | 22 | +    "ones_two_tensors": lambda: ((torch.ones(1), torch.ones(1)), 0), | 
|  | 23 | +    "ones_and_rand_three_tensors": lambda: ( | 
|  | 24 | +        (torch.ones(1, 2), torch.randn(1, 2), torch.randn(1, 2)), | 
|  | 25 | +        1, | 
|  | 26 | +    ), | 
|  | 27 | +    "ones_and_rand_four_tensors": lambda: ( | 
|  | 28 | +        ( | 
|  | 29 | +            torch.ones(1, 2, 5), | 
|  | 30 | +            torch.randn(1, 2, 5), | 
|  | 31 | +            torch.randn(1, 2, 5), | 
|  | 32 | +            torch.randn(1, 2, 5), | 
|  | 33 | +        ), | 
|  | 34 | +        -1, | 
|  | 35 | +    ), | 
|  | 36 | +    "rand_two_tensors": lambda: ( | 
|  | 37 | +        (torch.randn(2, 2, 4), torch.randn(2, 2, 4)), | 
|  | 38 | +        2, | 
|  | 39 | +    ), | 
|  | 40 | +    "rand_two_tensors_dim_0": lambda: ( | 
|  | 41 | +        (torch.randn(1, 2, 4, 4), torch.randn(1, 2, 4, 4)), | 
|  | 42 | +    ), | 
|  | 43 | +    "rand_two_tensors_dim_2": lambda: ( | 
|  | 44 | +        (torch.randn(2, 2, 3, 5), torch.randn(2, 2, 3, 5)), | 
|  | 45 | +        2, | 
|  | 46 | +    ), | 
|  | 47 | +    "rand_large": lambda: ( | 
|  | 48 | +        ( | 
|  | 49 | +            10000 * torch.randn(2, 3, 1, 4), | 
|  | 50 | +            torch.randn(2, 3, 1, 4), | 
|  | 51 | +            torch.randn(2, 3, 1, 4), | 
|  | 52 | +        ), | 
|  | 53 | +        -3, | 
|  | 54 | +    ), | 
|  | 55 | +} | 
|  | 56 | + | 
|  | 57 | + | 
|  | 58 | +class Stack(nn.Module): | 
|  | 59 | +    aten_op = "torch.ops.aten.stack.default" | 
|  | 60 | +    exir_op = "executorch_exir_dialects_edge__ops_aten_cat_default" | 
|  | 61 | + | 
|  | 62 | +    def forward(self, n: tuple[torch.Tensor, ...], dim: int = 0): | 
|  | 63 | +        return torch.stack(n, dim) | 
|  | 64 | + | 
|  | 65 | + | 
|  | 66 | +input_t1 = Tuple[torch.Tensor] | 
|  | 67 | + | 
|  | 68 | + | 
|  | 69 | +@common.parametrize("test_module", test_data_suite) | 
|  | 70 | +def test_stack_tosa_FP(test_module: input_t1): | 
|  | 71 | +    test_data = test_module() | 
|  | 72 | +    pipeline = TosaPipelineFP[input_t1]( | 
|  | 73 | +        Stack(), | 
|  | 74 | +        test_data, | 
|  | 75 | +        aten_op=Stack.aten_op, | 
|  | 76 | +        exir_op=Stack.exir_op, | 
|  | 77 | +        use_to_edge_transform_and_lower=False, | 
|  | 78 | +    ) | 
|  | 79 | +    pipeline.run() | 
|  | 80 | + | 
|  | 81 | + | 
|  | 82 | +@common.parametrize("test_module", test_data_suite) | 
|  | 83 | +def test_stack_tosa_INT(test_module: input_t1): | 
|  | 84 | +    test_data = test_module() | 
|  | 85 | +    pipeline = TosaPipelineINT[input_t1]( | 
|  | 86 | +        Stack(), | 
|  | 87 | +        test_data, | 
|  | 88 | +        aten_op=Stack.aten_op, | 
|  | 89 | +        exir_op=Stack.exir_op, | 
|  | 90 | +        use_to_edge_transform_and_lower=False, | 
|  | 91 | +    ) | 
|  | 92 | +    pipeline.run() | 
|  | 93 | + | 
|  | 94 | + | 
|  | 95 | +@common.XfailIfNoCorstone300 | 
|  | 96 | +@common.parametrize("test_module", test_data_suite) | 
|  | 97 | +def test_stack_u55_INT(test_module: input_t1): | 
|  | 98 | +    test_data = test_module() | 
|  | 99 | +    pipeline = EthosU55PipelineINT[input_t1]( | 
|  | 100 | +        Stack(), | 
|  | 101 | +        test_data, | 
|  | 102 | +        aten_ops=Stack.aten_op, | 
|  | 103 | +        exir_ops=Stack.exir_op, | 
|  | 104 | +        use_to_edge_transform_and_lower=False, | 
|  | 105 | +    ) | 
|  | 106 | +    pipeline.run() | 
|  | 107 | + | 
|  | 108 | + | 
|  | 109 | +@common.XfailIfNoCorstone320 | 
|  | 110 | +@common.parametrize("test_module", test_data_suite) | 
|  | 111 | +def test_stack_u85_INT(test_module: input_t1): | 
|  | 112 | +    test_data = test_module() | 
|  | 113 | +    pipeline = EthosU85PipelineINT[input_t1]( | 
|  | 114 | +        Stack(), | 
|  | 115 | +        test_data, | 
|  | 116 | +        aten_ops=Stack.aten_op, | 
|  | 117 | +        exir_ops=Stack.exir_op, | 
|  | 118 | +        use_to_edge_transform_and_lower=False, | 
|  | 119 | +    ) | 
|  | 120 | +    pipeline.run() | 
|  | 121 | + | 
|  | 122 | + | 
|  | 123 | +@common.SkipIfNoModelConverter | 
|  | 124 | +@common.parametrize("test_module", test_data_suite) | 
|  | 125 | +def test_stack_vgf_FP(test_module: input_t1): | 
|  | 126 | +    test_data = test_module() | 
|  | 127 | +    pipeline = VgfPipeline[input_t1]( | 
|  | 128 | +        Stack(), | 
|  | 129 | +        test_data, | 
|  | 130 | +        aten_op=Stack.aten_op, | 
|  | 131 | +        exir_op=Stack.exir_op, | 
|  | 132 | +        tosa_version="TOSA-1.0+FP", | 
|  | 133 | +        use_to_edge_transform_and_lower=False, | 
|  | 134 | +    ) | 
|  | 135 | +    pipeline.run() | 
|  | 136 | + | 
|  | 137 | + | 
|  | 138 | +@common.SkipIfNoModelConverter | 
|  | 139 | +@common.parametrize("test_module", test_data_suite) | 
|  | 140 | +def test_stack_vgf_INT(test_module: input_t1): | 
|  | 141 | +    test_data = test_module() | 
|  | 142 | +    pipeline = VgfPipeline[input_t1]( | 
|  | 143 | +        Stack(), | 
|  | 144 | +        test_data, | 
|  | 145 | +        aten_op=Stack.aten_op, | 
|  | 146 | +        exir_op=Stack.exir_op, | 
|  | 147 | +        tosa_version="TOSA-1.0+INT", | 
|  | 148 | +        use_to_edge_transform_and_lower=False, | 
|  | 149 | +    ) | 
|  | 150 | +    pipeline.run() | 
0 commit comments