Re: whisper.cpp backage structure based on llama.cpp hwcaps setup
Hi Petter,
On 2025-02-15 12:10, Petter Reinholdtsen wrote:
> [Christian Kastner]
>> These two issues together made my initial approach pretty
>> unwieldy. I've circled back, and will attempt another direction that
>> I'll share tomorrow.
>
> Great to hear. I look forward to testing it. :)
See the other mail I just sent. My conclusion is: whisper.cpp doesn't
need a complicated build, it's ggml that needs it, and having both
public and private libraries in the same package makes this quite
complicated.
It's much simpler to have ggml separately and to build llama.cpp and
whisper.cpp against it, and also much less work, which is why I'm
focusing on getting ggml done asap.
Best,
Christian
Reply to: