flip_op

→TensorImplPtr

int flip_op(const int & a, int dims)

Reverse a tensor along one or more axes.

Equivalent to NumPy's np.flip: along each listed axis the index ordering is inverted. Unlike roll_op there is no wrap-around.

aTensorImplPtr

Input tensor.

dimsvector<int>

Axes to flip.

TensorImplPtr

Tensor of the same shape and dtype as a with the requested axes reversed.

Backward applies flip with the same dims to the incoming gradient — flipping is self-inverse.