Tensor Operations

medium · tensors, numpy, shapes

Implement fundamental tensor operations that are the building blocks of deep learning.

Functions to implement

Create a tensor filled with zeros.

Reshape a tensor to new dimensions.

Transpose a 2D matrix (swap rows and columns).

Matrix multiplication of two 2D matrices.

Add two tensors element-wise.

Sum along a specific axis.

create_zeros((2, 3))  # [[0,0,0], [0,0,0]]
transpose([[1, 2], [3, 4]])  # [[1, 3], [2, 4]]
matmul([[1, 2]], [[3], [4]])  # [[11]]

from typing import List, Tuple, Union

Tensor = Union[List, float]

def create_zeros(shape: Tuple[int, ...]) -> Tensor:
    """TODO: Implement this function."""
    raise NotImplementedError()

def flatten(tensor: Tensor) -> List[float]:
    """TODO: Implement this function."""
    raise NotImplementedError()

def reshape(tensor: List, new_shape: Tuple[int, ...]) -> Tensor:
    """TODO: Implement this function."""
    raise NotImplementedError()

def build(shape, flat_iter):
    """TODO: Implement this function."""
    raise NotImplementedError()

def transpose(matrix: List[List[float]]) -> List[List[float]]:
    """TODO: Implement this function."""
    raise NotImplementedError()

def matmul(A: List[List[float]], B: List[List[float]]) -> List[List[float]]:
    """TODO: Implement this function."""
    raise NotImplementedError()

def elementwise_add(A: Tensor, B: Tensor) -> Tensor:
    """TODO: Implement this function."""
    raise NotImplementedError()

def sum_along_axis(tensor: List[List[float]], axis: int) -> List[float]:
    """TODO: Implement this function."""
    raise NotImplementedError()