Binary

Binary#

add#

def add(x: 'Tensor', y: 'Tensor | float | int') -> 'Tensor':

Add x and y element-wise, with broadcasting.

sub#

def sub(x: 'Tensor', y: 'Tensor | float | int') -> 'Tensor':

Subtract y from x element-wise, with broadcasting.

mul#

def mul(x: 'Tensor', y: 'Tensor | float | int') -> 'Tensor':

Multiply x and y element-wise, with broadcasting.

div#

def div(x: 'Tensor', y: 'Tensor | float | int') -> 'Tensor':

Divide x by y element-wise, with broadcasting.

matmul#

def matmul(x: 'Tensor', y: 'Tensor') -> 'Tensor':

Matrix multiplication of x and y.

Supports batched inputs with arbitrary-rank batch prefixes. 1-D inputs are automatically promoted: a vector of shape (N,) becomes (1, N) or (N, 1) for the left/right operand respectively, and the added dimension is squeezed from the result.

Parameters

  • x – Tensor of shape (*, M, K) or (K,).

  • y – Tensor of shape (*, K, N) or (K,).

Returns

Tensor of shape (*, M, N) (or scalar for vector × vector).

mod#

def mod(x: 'Tensor', y: 'Tensor | float | int') -> 'Tensor':

Compute the element-wise remainder x % y, with broadcasting.

pow#

def pow(x: 'Tensor', y: 'Tensor | float | int') -> 'Tensor':

Compute the element-wise power x ** y, with broadcasting.

outer#

def outer(x: 'Tensor', y: 'Tensor') -> 'Tensor':

Compute the outer product of vectors x and y.

For 1-D inputs of shapes (M,) and (N,), returns an (M, N) matrix. Under vmap, operates on the non-batched trailing dimensions.

Parameters

  • x – Vector of shape (M,).

  • y – Vector of shape (N,).

Returns

Tensor of shape (M, N).

Contents