Skip to content

Support 32-bit input in bBeamFormerCCGKernel

Bram Veenboer requested to merge ccglib-32bit into main

Passing --nr_input_bits=32 to bBeamFormerCCGKernel is now also supported. To make this work, BeamFormerCCGKernel needed to pass the correct precision and not use 16-bit precision as hardcoded default.

Merge request reports