Target: Neural nets with auto-diff with BLAS as only dep. WIP! unfinished, just started, ignore!
See Benchmarks for matrix multiplication comparisation between Intel OpenBlas, M1 Accelerate and M1 with Tensorflow (GPU) or Pytorch (MPS)
Requirements: cmake
, ninja
, for non-Apple: openblas
.
- Linux:
pacman -S cmake ninja openblas
(orapt...
) and some c compiler. - Mac:
brew install cmake ninja
, command-line tools required.
Clone repo, create build
directory:
mkdir build
cd build
cmake -G Ninja ..
ninja
Run (in build folder):
./bench/bench_neurad
or
./test/test_neurad