mirror of
https://github.com/ggml-org/whisper.cpp.git
synced 2026-10-04 05:21:27 +02:00
Shrivas Shankar
050b8567a0
Drops the hardcoded f32 GLU kernels in favor of a single template. We now load/store in the native tensor type (half or float) to save memory bandwidth, but keep the actual ALU compute in float to avoid exploding math in geglu/swiglu. Also opened up the dispatch gate to allow f16 inputs.