BatchedArray{T, NI, N, AT} <: AbstractBatchedArray{T, NI, N}

A concrete type for batched arrays. T is the element type, NI is the inner sample's dimension, N is the total dimension and AT is the array type that actually holds the value.

BatchedUniformScaling{T, N, ST <: AbstractArray{T, N}} <: AbstractBatchedArray{T, 0, N}

Scale a batch of arrays with a batch of scalars.


The shape of batch can be multidimentional, which means member BatchedScale.scalars can be a matrix or high dimentional array, the shape of this member is the shape of batch. dims defines the dimmension of each sample in the batch. It can be multidimentional as well.

batch_size(batched_array) -> Tuple

Returns a tuple of size of each batch dimension of the batched array.

inner_size(batched_array) -> Tuple

Returns a tuple of size of each inner dimension of the batched array.

merged_size(batched_array) -> Tuple

Returns the size of this batched array after merging all its batched dimension together.

AbstractBatchedArray{T, NI, N}

Abstract type batched array. A batched array use its last N - NI dimension as batch dimension, it is a batch of array with dimension NI.

BatchedAdjoint{T, N, S <: AbstractBatchedMatrix} <: AbstractBatchedMatrix{T, N}

Batched ajoint. Transpose a batch of matrix.

BatchedTranspose{T, N, S} <: AbstractBatchedMatrix{T, N}

Batched transpose. Transpose a batch of matrix.

batched_gemm(A, B)
batched_gemm(tA, tB, A, B)
batched_gemm(tA, tB, alpha, A, B)

Batched version of BLAS.gemm.

batched_gemm!(transA, transB, alpha, A, B, beta, C)

Batched version of BLAS.gemm!.

batched_tr!(B::AbstractVector{T}, A::AbstractArray{T, 3})

Perform batched matrix trace.