Update: build.sh; WIP

[rezaarezvan]

This should make the build.sh script work. I've tested it on my local machine (i7 2020 MBP) and works. I will test it on my Ubuntu machine soon.

Using gmake was the fix for macOS according to, ggml-org/llama.cpp#1570. I'm not sure if this will break for other platforms, in that case, we can add a simple detection depending on what platform the user is on.

[ghost]

Hey, Thanks a lot!

I myself tank into this issue on my intel mac.

Can you see if the cmake commands work for you?

Asking cause gmake is just make on ubuntu and so this script might not work

[ghost]

No need for the previous line, you can remove that

[rezaarezvan]

Hey, Thanks a lot!

I myself tank into this issue on my intel mac.

Can you see if the cmake commands work for you?

Asking cause gmake is just make on ubuntu and so this script might not work

gmake will work on Ubuntu as well. The problem is specifically make on macOS, with gmake this works. So I think the gmake solution is fine. But I made the change so we use CMake now instead.

[rezaarezvan]

Hey, Thanks a lot!
I myself tank into this issue on my intel mac.
Can you see if the cmake commands work for you?
Asking cause gmake is just make on ubuntu and so this script might not work

gmake will work on Ubuntu as well. The problem is specifically make on macOS, with gmake this works. So I think the gmake solution is fine. But I made the change so we use CMake now instead.

I've tested on my Ubuntu machine - the gmake solution works as well so it's up to either you (@cto-junior) or @yacineMTB to decide which approach we want to use :).

[ghost]

Here you will also need to take care of where CUDA flag was set or not. See the whisper command for reference

[rezaarezvan]

Here you will also need to take care of where CUDA flag was set or not. See the whisper command for reference

CUDA flags? What do you mean?