whisper.cpp

b27726da - whisper : add note that ggml_mul_mat_pad does not work with CUDA