fftkit 0.1.0

One FFT API over many backends: scipy, numpy, Intel MKL, CuPy, PyTorch, TensorFlow, pyFFTW, Accelerate

fftkit

One FFT API over many backends.

import fftkit

fftkit.get_available_backends()      # what this machine can actually use
fft = fftkit.get_fft_func("mkl")     # or scipy, numpy, cupy, torch, ...
spectrum = fft(signal)

Swapping FFT libraries usually means rewriting call sites, because each library spells the same transform differently. fftkit puts a single callable in front of eight of them, reports which ones are installed, and measures which one is fastest for your array sizes.

Extracted from modalpy.

Status

Under construction — the packaging scaffold is in place; the library, tests, and benchmarks are landing next. See CHANGELOG.md.

Install

pip install fftkit               # numpy + scipy only
pip install "fftkit[mkl]"        # + Intel MKL
pip install "fftkit[gpu]"        # + CuPy / PyTorch
pip install "fftkit[bench]"      # + benchmark plotting deps

License

MIT — see LICENSE.