我有一个像这样的numpy数组，
array([[ 9.10812400e-01, 7.55480960e+00, 5.85824600e+00,
-1.63298620e-02, -4.18150200e-02, -7.58900800e-02],
[ 7.93121340e-01, 7.59404000e+00, 5.81901550e+00,
-1.63298620e-02, -4.18150200e-02, -7.58900800e-02],
[ 7.73513800e-01, 7.35867300e+00, 6.03475950e+00,
9.80464400e-02, 1.60981410e-03, 1.03224980e-02],
[ 7.73513800e-01, 7.41751100e+00, 6.01515200e+00,
9.80464400e-02, 1.60981410e-03, 1.03224980e-02],
[ 8.32351700e-01, 7.14292900e+00, 6.36819460e+00,
1.98594750e-01, 1.33242460e-01, 1.14474740e-01],
[ 9.89257800e-01, 7.00563050e+00, 6.32896400e+00,
1.98594750e-01, 1.33242460e-01, 1.14474740e-01],
[ 1.69534300e+00, 7.61364750e+00, 7.87841800e+00,
-7.01231800e-02, 7.94170350e-02, 1.17905500e-01],
[ 5.77377300e-01, 7.75094600e+00, 6.56431600e+00,
-1.82861970e-01, 2.53925120e-02, 1.51244830e-01],
[ 1.16578670e+00, 7.47634900e+00, 7.42730700e+00,
-2.52029870e-01, 8.80286700e-03, 2.21323560e-01],
[ 1.49920650e+00, 7.26060500e+00, 7.34884640e+00,
-2.52029870e-01, 8.80286700e-03, 2.21323560e-01],
[ 1.81301880e+00, 7.08409100e+00, 7.40769960e+00,
-2.52029870e-01, 8.80286700e-03, 2.21323560e-01],
[ 1.83262630e+00, 6.98602300e+00, 7.32923900e+00,
-2.77232860e-01, 5.01918570e-02, 1.12273335e-01],
[ 1.75418090e+00, 7.00563050e+00, 7.46653750e+00,
-2.77232860e-01, 5.01918570e-02, 1.12273335e-01],
[ 1.06658936e-01, 8.16281100e+00, 5.56404100e+00,
-4.09735530e-01, -1.65083600e-01, 4.19206840e-02],
[ 4.79309080e-01, 8.35894800e+00, 5.62287900e+00,
-2.90299700e-01, -2.74592130e-01, 8.11367500e-02],
[ 2.63565060e-01, 8.24127200e+00, 5.36790470e+00,
-2.90299700e-01, -2.74592130e-01, 8.11367500e-02]])
这个数组的形状是（16，6）
现在，我希望使用自动编码器嵌入此数组，并且输出形状可能会或可能不会像输入数组的形状（可以动态更改）。
如何使用keras做到这一点？