• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1# Copyright 2023 Huawei Technologies Co., Ltd
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7# http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14# ============================================================================
15
16import numpy as np
17import pytest
18
19import mindspore as ms
20import mindspore.nn as nn
21from mindspore import Tensor, ops
22
23from tests.st.utils import test_utils
24
25
26class Net(nn.Cell):
27    def __init__(self):
28        super(Net, self).__init__()
29        self.conv3d = nn.Conv3d(in_channels=3, out_channels=32, kernel_size=(4, 3, 3), dtype=ms.float16)
30
31    def construct(self, x):
32        out = self.conv3d(x)
33        return out
34
35
36@pytest.mark.level1
37@pytest.mark.platform_x86_cpu
38@pytest.mark.platform_arm_cpu
39@pytest.mark.platform_arm_ascend_training
40@pytest.mark.platform_x86_ascend_training
41@pytest.mark.env_onecard
42@pytest.mark.parametrize('mode', [ms.GRAPH_MODE, ms.PYNATIVE_MODE])
43def test_conv3d_para_customed_dtype(mode):
44    """
45    Feature: Conv3d
46    Description: Verify the result of Conv3d specifying customed para dtype.
47    Expectation: success
48    """
49    ms.set_context(mode=mode)
50    net = Net()
51    x = Tensor(np.ones([16, 3, 10, 32, 32]), ms.float16)
52    output = net(x)
53    expect_output_shape = (16, 32, 10, 32, 32)
54    assert np.allclose(expect_output_shape, output.shape)
55
56
57@pytest.mark.level0
58@pytest.mark.platform_arm_ascend_training
59@pytest.mark.platform_x86_ascend_training
60@pytest.mark.platform_arm_ascend910b_training
61@pytest.mark.env_onecard
62@test_utils.run_test_with_On
63def test_conv3d_input_5d():
64    """
65    Feature: Conv3d 5d input
66    Description: Verify the result of Conv3d 5d input.
67    Expectation: success
68    """
69    ms.set_context(mode=ms.GRAPH_MODE, ascend_config={"precision_mode": "force_fp16"})
70    class Network(nn.Cell):
71        def __init__(self):
72            super().__init__()
73            self.relu = ops.ReLU()
74            self.conv1 = nn.Conv3d(1, 1, kernel_size=5, pad_mode="same", padding=0, has_bias=False, weight_init="One")
75            self.reducemin = ops.ReduceMin(keep_dims=True)
76            self.reducesum = ops.ReduceSum(keep_dims=True)
77            self.add = ops.Add()
78            self.square = ops.Square()
79            self.abs = ops.Abs()
80            self.concat = ops.Concat()
81            self.batchnorm = nn.BatchNorm3d(5)
82
83        def construct(self, data1, data2):
84            batchnorm3d_01 = self.batchnorm(data1)
85            batchnorm3d_02 = self.batchnorm(data1)
86            reducesum_01 = self.reducesum(batchnorm3d_02, 1)
87            add_01 = self.add(reducesum_01, data2)
88            reducemin_01 = self.reducemin(add_01, 1)
89            relu_01 = self.relu(batchnorm3d_01)
90            abs_01 = self.abs(relu_01)
91            square_01 = self.square(abs_01)
92            reducemin_02 = self.reducemin(square_01, 1)
93            concat_01 = self.concat((reducemin_02, reducemin_01))
94            conv_01 = self.conv1(concat_01)
95            relu_03 = self.relu(conv_01)
96            output = relu_03
97            return  output
98
99    data1 = Tensor(np.ones([1, 5, 5, 5, 4]).astype(np.float32))
100    data2 = Tensor(np.ones([1, 5, 5, 4]).astype(np.float32))
101
102    ms.set_context(device_target="CPU")
103    cpu_mode = Network()
104    cpu_out = cpu_mode(data1, data2).asnumpy()
105
106    ms.set_context(device_target="Ascend")
107    npu_mode = Network()
108    npu_out = npu_mode(data1, data2).asnumpy()
109
110    assert np.allclose(cpu_out, npu_out, 0.001, 0.001)
111