pullrequests Search Results · repo:ggerganov/whisper.cpp language:C
Filter by
616 results
(88 ms)616 results
inggerganov/whisper.cpp (press backspace or delete to remove)In log_mel_spectrogram_worker_thread() function, we allocate enough memory for fft_in and fft_out variable, and then we
can avoid allocating memory for even, odd, even_fft and odd_fft in fft(). In my testing, ...
- mkycoder
- 2
- Opened on Jun 18
- #2242
The target is https://github.com/ggerganov/whisper.cpp/pull/2239
Not sure if this is better, but it works.
Increasing the number of threads of the BLAS backend seem to results in better performance with ...
- slaren
- 1
- Opened on Jun 16
- #2240
target branch: https://github.com/ggerganov/whisper.cpp/tree/sync-blas
Using BLAS just in the Encoder since it involves large matrices
- ggerganov
- 10
- Opened on Jun 16
- #2239
TODO:
- [x] update CMake / Makefile builds
- [x] try to introduce ggml_backend_sched: https://github.com/ggerganov/whisper.cpp/pull/2239
- ggerganov
- 1
- Opened on Jun 16
- #2238
- ggerganov
- Opened on Jun 16
- #2237
- fix issues where whisper may be set to run on a device which is not the current one for all mel_calc calls.
- workaround for #2230 as opposed to #2235 : forbid mel_calc_cuda if device cc is below ...
- iboB
- 1
- Opened on Jun 12
- #2236