# Copyright (c) 2020 Huawei Technologies Co., Ltd
# All rights reserved.
#
# Licensed under the BSD 3-Clause License  (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# https://opensource.org/licenses/BSD-3-Clause
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import torch
import torch_npu

from torch_npu.testing.testcase import TestCase, run_tests
from torch_npu.testing.common_utils import create_common_tensor
from torch_npu.contrib.module import DCNv2

class TestDeformConv(TestCase):
    
    def test_npu_deform_conv_1(self):
        x = torch.randn(2, 2, 3, 3)
        model = DCNv2(2, 2, 3, 2, 1)

        x = x.npu()
        x.requires_grad = True
        model = model.npu()

        output = model(x)
        l = output.sum()
        l.backward()

        expedt_cpu_output = torch.tensor([[[[ 0.0359, -0.0297],
                                            [ 0.2135,  0.0879]],

                                            [[ 0.2250,  0.0972],
                                            [-0.1099,  0.0224]]],


                                            [[[-0.2693,  0.4263],
                                            [-0.2629, -0.2155]],

                                            [[ 0.0654, -0.4343],
                                            [-0.0067,  0.1704]]]], dtype=torch.float32)

        expedt_cpu_xgrad = torch.tensor([[[[ 0.1239, -0.1093,  0.1239],
                                            [-0.0060,  0.2868, -0.0060],
                                            [ 0.1239, -0.1093,  0.1239]],

                                            [[-0.0404, -0.0229, -0.0404],
                                            [ 0.0341, -0.3033,  0.0341],
                                            [-0.0404, -0.0229, -0.0404]]],


                                            [[[ 0.1239, -0.1093,  0.1239],
                                            [-0.0060,  0.2868, -0.0060],
                                            [ 0.1239, -0.1093,  0.1239]],

                                            [[-0.0404, -0.0229, -0.0404],
                                            [ 0.0341, -0.3033,  0.0341],
                                            [-0.0404, -0.0229, -0.0404]]]], dtype=torch.float32)
        self.assertRtolEqual(expedt_cpu_output, output.detach().cpu())
        self.assertRtolEqual(expedt_cpu_xgrad, x.grad.cpu())

    def test_npu_deform_conv_2(self):
        x = torch.randn(2, 2, 5, 5)
        model = DCNv2(2, 2, 3, 2, 1)

        x = x.npu()
        x.requires_grad = True
        model = model.npu()

        output = model(x)
        l = output.sum()
        l.backward()

        expedt_cpu_output = torch.tensor([[[[ 0.0568, -0.0061,  0.0660],
                                            [ 0.0097, -0.1420,  0.1683],
                                            [ 0.1852, -0.3518, -0.0962]],

                                            [[ 0.1061, -0.1260, -0.0641],
                                            [-0.1111, -0.4993,  0.2399],
                                            [ 0.2185,  0.1643,  0.0179]]],


                                            [[[-0.4639, -0.0570, -0.1230],
                                            [-0.1023,  0.3752,  0.5703],
                                            [-0.0807, -0.3313, -0.1658]],

                                            [[-0.1327, -0.2725, -0.0910],
                                            [-0.0751,  0.1237,  0.4084],
                                            [ 0.1052, -0.2177, -0.0489]]]], dtype=torch.float32)

        expedt_cpu_xgrad = torch.tensor([[[[ 0.1238,  0.0521,  0.1238,  0.0521,  0.1238],
                                        [-0.0504,  0.2393, -0.0504,  0.2393, -0.0504],
                                        [ 0.1238,  0.0521,  0.1238,  0.0521,  0.1238],
                                        [-0.0504,  0.2393, -0.0504,  0.2393, -0.0504],
                                        [ 0.1238,  0.0521,  0.1238,  0.0521,  0.1238]],

                                        [[-0.1459, -0.0728, -0.1459, -0.0728, -0.1459],
                                        [-0.0267, -0.3579, -0.0267, -0.3579, -0.0267],
                                        [-0.1459, -0.0728, -0.1459, -0.0728, -0.1459],
                                        [-0.0267, -0.3579, -0.0267, -0.3579, -0.0267],
                                        [-0.1459, -0.0728, -0.1459, -0.0728, -0.1459]]],


                                        [[[ 0.1238,  0.0521,  0.1238,  0.0521,  0.1238],
                                        [-0.0504,  0.2393, -0.0504,  0.2393, -0.0504],
                                        [ 0.1238,  0.0521,  0.1238,  0.0521,  0.1238],
                                        [-0.0504,  0.2393, -0.0504,  0.2393, -0.0504],
                                        [ 0.1238,  0.0521,  0.1238,  0.0521,  0.1238]],

                                        [[-0.1459, -0.0728, -0.1459, -0.0728, -0.1459],
                                        [-0.0267, -0.3579, -0.0267, -0.3579, -0.0267],
                                        [-0.1459, -0.0728, -0.1459, -0.0728, -0.1459],
                                        [-0.0267, -0.3579, -0.0267, -0.3579, -0.0267],
                                        [-0.1459, -0.0728, -0.1459, -0.0728, -0.1459]]]], 
                                        dtype=torch.float32)
        self.assertRtolEqual(expedt_cpu_output, output.detach().cpu())
        self.assertRtolEqual(expedt_cpu_xgrad, x.grad.cpu())

if __name__ == "__main__":
    run_tests()