v0.7

Added

• eye
• allclose
• ndarray.conj
• ndarray.copy
• ndarray.astype
• ndarray.tolist
• linalg
  • norm
  • matrix_rank for real matrices
  • svd for real matrices

Improved

• ndarray.real

v0.6

Added

• ndarray.dtype
• ndarray.real
• support math functions

Improved

• einsum for three or more operands
• zeros_like and ones_like inherit original dtype

v0.5

• Roughly implemented __getitem__ and __setitem__ (not including fancy indexing)

v0.4

• Improved broadcast
• Fixed reshape

v0.3

Added

• ones
• ones_like

Improved

• Improved einsum:
  • trace
  • inner product
  • transpose

v0.2

• Roughly implemented einsum
  • vector, matrix -> vector
  • matrix, matrix -> matrix
  • higher order tensor, higher order tensor -> higher order tensor

But, some basic functionalities includingtrace, inner product are not implemented.

v0.1

Implemented mynumpy.ndarray's:

• __str__
• __repr__
• __eq__
• __ne__
• __add__
• __sub__
• __mul__
• __matmul__ (limited)
• __truediv__
• __len__
• ndim
• shape
• size
• T
• flatten
• reshape

and mynumpy's:

• zeros
• zeros_like

Note that the broadcasting is very restricted, i.e., only a tensor + a scaler is supported.