mirror of
https://github.com/ggerganov/whisper.cpp.git
synced 2025-05-29 13:34:13 +00:00
It's helpful to use expm1f(x), because expf(x)-1 will result in overflow for 25% of single-precision floating point numbers.
It's helpful to use expm1f(x), because expf(x)-1 will result in overflow for 25% of single-precision floating point numbers.