Tensora

Tensora is a tensor algebra library for Python. You can create Tensor objects in a variety of sparse and dense formats. You can do calculations with these tensors by passing the them to the evaluate function along with an expression (e.g. y = evaluate('y(i) = A(i,j) * x(j)', A=A, x=x)). The expression is parsed, a kernel is generated, the C code is compiled on the fly, the binary is invoked, and the result is packaged into a output Tensor.

Tensora also comes with the tensora command line tool that can be used to generate the kernel code for external use.

Tensora is based on the Tensor Algebra Compiler (TACO).

Installation

The recommended means of installation is with pip from PyPI.

pip install tensora

Tensora definitely works on Mac and Linux and definitely does not work on Windows. It will probably work on any system where ffi.dlopen(None) works.

By default, Tensora uses its own code to generate the kernels. The tensora[taco] extra makes available the option to generate the kernels with the TACO binary, which is provided by the tensora-taco PyPI package.

By default, Tensora uses LLVM to compile the kernels. The LLVM is provided by the llvmlite PyPI dependency and does not need to be installed on the system. The tensora[cffi] extra makes available the option to compile the kernels with CFFI. To do that, a C compiler must be installed on the system in a way that it is available to CFFI API mode.

Hello world

Here is an example of multiplying a sparse matrix in CSR format with a dense vector:

from tensora import Tensor, evaluate

elements = {
    (1,0): 2.0,
    (0,1): -2.0,
    (1,2): 4.0,
}

A = Tensor.from_dok(elements, dimensions=(2,3), format='ds')
x = Tensor.from_lol([0, -1, 2])

y = evaluate('y(i) = A(i,j) * x(j)', 'd', A=A, x=x)

assert y == Tensor.from_lol([2,8])