mm_op

→TensorImplPtr

int mm_op(const int & a, const int & b)

source edit

Python wrappers

pylucid.ops.mmfunction

Strict 2-D matrix multiply: $C = A \cdot B$ .

Composite — rank-checked alias of matmul_op that rejects inputs of rank $\ne 2$ up front so the error message names mm rather than the planner. Gradient flows through MatmulBackward.

Math

C_{ij} = \sum_{k} A_{ik} \, B_{kj}

See Also

matmul_op — generalised version with broadcasting and 1-D vector promotion; bmm_op — strict batched variant.

Parameters

aTensorImplPtr

Left operand of shape (M, K).

bTensorImplPtr

Right operand of shape (K, N).

Returns

TensorImplPtr

Tensor of shape (M, N).

Raises

Failure

If either input is null or not exactly 2-D.