Georgi Gerganov

Ok, you can track the state of `ggml` becoming thread-safe in this issue: https://github.com/ggerganov/llama.cpp/issues/3960 Sorry for the inconvenience - you might want to stick with whisper.cpp v1.5.1 for now

Does this change have any positive effect for `whisper-v3` or it's still repeating stuff?

Did you run some tests?

Wow! This looks like a very important work. Would love to give this a try at some point Any reason to prefer `-ac 500` over `-ac 512`? Round numbers are...

Use `-bs 1` to get the old speed. The quality with more beams in general should be better, but it's possible that you don't observe much of a difference

Do not use large-v3. I haven't checked what is the status recently, but last time I did this model did not work well even with the original OpenAI repo. Try...

Hm, I can't reproduce on my Mac Studio: ``` ./bin/main -m ../models/ggml-large-v2.bin -f ../samples/gb0.wav --no-gpu whisper_init_from_file_with_params_no_state: loading model from '../models/ggml-large-v2.bin' whisper_model_load: loading model whisper_model_load: n_vocab = 51865 whisper_model_load: n_audio_ctx =...

Ok, to me it seems better to aim for the more general solution and for now not merge this change.

Not sure why it fails - I have very basic understanding of the CoreML stuff. Probably somebody with more expertise can help out

You can use the `whisper_pcm_to_mel()` + `whisper_lang_auto_detect()` API. You will get the probs for all languages in the `lang_probs` array: https://github.com/ggerganov/whisper.cpp/blob/59a3d0cb576db605f76f82f07350647837e15c7a/whisper.h#L244-L255