exp_inplace_op

→TensorImplPtr

int exp_inplace_op(const int & a)

source edit

In-place natural exponential: writes $e^x$ back into a.

Math

$a_i \leftarrow \exp(a_i)$

See Also

exp_op — out-of-place counterpart.

Parameters

aTensorImplPtr

Tensor to exponentiate in place. Must be non-null. AMP may promote integer inputs to float before dispatch; the promoted dtype is written back to a.

Returns

TensorImplPtr

The same pointer as a.

Loading…

Lucid3.0