1、numpy.ndarray -> mxnet.ndarray.NDArray
直接调用mx.nd.array(ndarray_in_numpy)函数
2、mxnet.ndarray.NDArray -> numpy.ndarray
就是对mxnet.ndarray.NDArray对象,直接调用函数asnumpy()
3、numpy增加矩阵维度
numpy.expand_dims(a, axis=0)
4、数组堆叠函数:
c = numpy.stack(a, axis = 0)
c = numpy.stack((a, b), axis = 0) (注:a, b shape必须相等)
c = numpy.vstack((a, b))
c = numpy.hstack((a, b))
5、numpy.repeat()
a = np.array([[1],[1],[2]])
a.shape = (3, 1)
b = numpy.repeat(a, 3, axis = -1)
b.shape = (3, 3)
6、numpy数据创建:
# 创建 0 数组
np.zeros((2, 3), dtype="float32")
# 创建 1 数组
np.ones((2, 3), dtype="int8")
# 创建数组为10的数组
np.ones((2, 3), dtype="int8") * 10
np.full((2, 3), 10, dtype="int8")