test_conv2d_op.py 2.8 KB
Newer Older
1 2
import unittest
import numpy as np
H
hedaoyuan 已提交
3
from op_test import OpTest
4 5


C
chengduoZH 已提交
6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35
def conv2d_forward_naive(input, filter, group, conv_param):
    in_n, in_c, in_h, in_w = input.shape
    out_c, f_c, f_h, f_w = filter.shape
    assert f_c * group == in_c
    assert np.mod(out_c, group) == 0
    sub_out_c = out_c / group

    stride, pad = conv_param['stride'], conv_param['pad']
    out_h = 1 + (in_h + 2 * pad - f_h) / stride
    out_w = 1 + (in_w + 2 * pad - f_w) / stride
    out = np.zeros((in_n, out_c, out_h, out_w))

    input_pad = np.pad(input, ((0, ), (0, ), (pad, ), (pad, )),
                       mode='constant',
                       constant_values=0)
    for i in range(out_h):
        for j in range(out_w):
            for g in range(group):
                input_pad_masked = input_pad[:, g * f_c:(
                    g + 1) * f_c, i * stride:i * stride + f_h, j * stride:j *
                                             stride + f_w]
                f_sub = filter[g * sub_out_c:(g + 1) * sub_out_c, :, :, :]
                for k in range(sub_out_c):
                    out[:, g * sub_out_c + k, i, j] = np.sum(input_pad_masked *
                                                             f_sub[k, :, :, :],
                                                             axis=(1, 2, 3))

    return out


H
hedaoyuan 已提交
36
class TestConv2dOp(OpTest):
37
    def setUp(self):
H
hedaoyuan 已提交
38
        self.init_groups()
H
hedaoyuan 已提交
39
        self.op_type = "conv2d"
C
chengduoZH 已提交
40 41 42 43 44 45 46 47 48 49
        input_size = [2, 3, 5, 5]  # NCHW
        assert np.mod(input_size[1], self.groups) == 0
        f_c = input_size[1] / self.groups
        filter_size = [6, f_c, 3, 3]
        conv2d_param = {'stride': 1, 'pad': 0}

        input = np.random.random(input_size).astype("float32")
        filter = np.random.random(filter_size).astype("float32")

        output = conv2d_forward_naive(input, filter, self.groups, conv2d_param)
50

H
hedaoyuan 已提交
51
        self.inputs = {'Input': input, 'Filter': filter}
H
hedaoyuan 已提交
52 53 54 55 56
        self.attrs = {
            'strides': [1, 1],
            'paddings': [0, 0],
            'groups': self.groups
        }
57 58
        self.outputs = {'Output': output}

H
hedaoyuan 已提交
59 60 61
    def test_check_output(self):
        self.check_output()

H
hedaoyuan 已提交
62
    def test_check_grad(self):
63 64
        self.check_grad(
            set(['Input', 'Filter']), 'Output', max_relative_error=0.05)
H
hedaoyuan 已提交
65

66
    def test_check_grad_no_filter(self):
67 68 69 70 71
        self.check_grad(
            ['Input'],
            'Output',
            max_relative_error=0.05,
            no_grad_set=set(['Filter']))
72 73

    def test_check_grad_no_input(self):
74 75 76 77 78
        self.check_grad(
            ['Filter'],
            'Output',
            max_relative_error=0.05,
            no_grad_set=set(['Input']))
79

H
hedaoyuan 已提交
80 81 82 83 84 85 86 87
    def init_groups(self):
        self.groups = 1


class TestWithGroup(TestConv2dOp):
    def init_groups(self):
        self.groups = 3

H
hedaoyuan 已提交
88

89 90
if __name__ == '__main__':
    unittest.main()