Crate tensor_optim 

This library offers basic tensor types:

• RefTensor: Statically sized slices with a mutable reference for data.
  • Ownership: The tensor does not own it’s data, it has references for the lifetime 'a to each slice.
  • Allocation: This tensor has no dynamic allocation, it doesn’t have it’s own memory, just pointers.
• ArrTensor: Statically sized arrays build the foundation for shape and data memory.
  • Ownership: The entire tensor is owned by the struct.
  • Allocation: This tensor has no dynamic allocation, it lives on the stack.
• DynTensor: A dynamically allocated tensor with far more flexibility than the others.
  • Ownership: The entire tensor is owned by the struct.
  • Allocation: This tensor dynamically allocates everything, shape is boxed and data is wrapped in Arc.

Note: The crate is fully documented, no-std compatible, and well tested. It doesn’t even need alloc unless the alloc feature (off by default) is enabled.

Structs

ArrTensor

A tensor made up of statically sized arrays.

RefTensor

A tensor-like structure that owns no data and holds only slices.

Constants

MAX_STATIC_RANK

A small limit to ranks mostly in ArrTensor. Primarily for matrix multiplication.

Traits

ConstTensorOps

Operations only statically sized, non-allocating tensors can leverage.

TensorOps

Operations even the simplest tensors need to get indexing, shape, and other content.