Przemysław Pawełczyk

Can you do one more test? Building using `cmake` with `--config Release`: ``` cmake --fresh -B build -DWHISPER_NO_AVX2=ON -DWHISPER_NO_FMA=ON cmake --build build --config Release ```

Thanks for testing. It seems that either one of this options happen: - Debug configuration is much slower than Release configuration (and 20 mins is simply not enough), - Debug...

Thank you for the report. Can you provide what's your current OS and compiler? Were they the same one for the older commit? EDIT: Sorry, I missed that you confirmed...

You may want to try also with `--beam-size 2`, as that's what seems to be the default in the older commit. It was changed in b6c5f49b78b214b7b4aa7392a8ba489c78b7382a. As Georgi commented in...

> It looks like the former default was beam-size=-1 ? I was referring to changes in `whisper_full_default_params`, where `beam_search.beam_size` changed from 2 to 5, but you're right that `whisper_params.beam_size` previously...

Could you do `git bisect` between good (v1.4.3) and bad (v1.5.5) to try to locate the main commit responsible for performance drop in your environment? Less than 10 steps (experiments)...

@ggerganov, do you have any ideas what else @Linux13524 could try or tweak in pursuit of restoring whisper.cpp performance from 1.4.x in 1.5.x?

Just a side comment and follow-up to my earlier comment: > I'm wondering if WHISPER_NO_AVX512 shouldn't be introduced in Makefile, to make it easier to disable AVX-512 (setting 3 variables...

Thank you for your interest in metastore, all credits go to original author, David. From metadata perspective I think there is limited value in storing `atime` or `ctime`. `ctime` is...

Sorry, I will be momentarily afk for rest of the day, so cannot help much. I would need to look into llama.cpp, but `madvise()` typically can be replaced with more...