Porting artifical intelligence (AI) code to FreeBSD

PacketMan

PacketMan

Silly question I know, and might even reveal just how much I still don't know, but here goes. Is there any effort to port some of the AI code over to FreeBSD? Are there any AI developers that are friendly and see the value in using FreeBSD OS to run AI code? Or am I off the rails completely, maybe AI code has its own OS, or is its own OS and doesn't use any 'common base' OS.

I would think if we are going to use AI in a heavy duty way, then it needs a heavy duty OS to keep it 'alive ;) and well'.

...or perhaps I should go back fishing? :p
 
Ask yourself this, how would an OS benefit from AI?

There may be some ports that use AI, some specific scientific research software. But I really don't see the point of adding AI to an OS.
 
So AI is not an overlay on top of some sort of OS? I always assumed that since the job of an OS is to essentially convert 'human instructions' into 'cpu instructions' that even AI (as advanced as it is) would still need an OS. So I don't see it as a case of the OS benefiting from 'serving' AI, but rather a case of AI needing an OS to run; thus my asking if some day we could expect to see AI machine with FreeBSD for its soul.
 
They're just applications with a specific purpose. We have about 25000+ applications in the ports tree, most of them haven't been written specifically for FreeBSD. If there's some open source AI code it's probably possible to port it.
 
Trihexagonal said:
Watson is made up of ten racks of IBM POWER 750 servers running Linux, and has 15Terabytes of RAM; 2,880 3.55GHz POWER7 processor cores and operates at 80 Teraflops.
Click to expand...
That is ancient news: This was the incarnation of Watson in 2011, which played the "jeopardy" game show. That hardware is by now quite obsolete; I'm not even sure new systems would be orderable any longer. Note that they are using Power7 processor chips; today's shipping system are on Power8, and Power9 just started going out.

Today's "Watson" brand name has only a spiritual connection to the system that played "jeopardy"; the code base used in the various "Watson" software products has little to do with the television game show.
 
The original request is a good idea. AI integration with the Operating System is the best way to get the following.
  • Text to speech
  • Speech Recognition
  • Image generation
  • Language Translation
  • Developer Assistance
  • and many other things

Here are my notes on using OpenAI on FreeBSD. Note this method currently uses the CPU and is therefore slower.

FreeBSD Install

Additional Links
https://github.com/ggerganov/whisper.cpp/issues/56

Install
cd /usr/local/src
git clone https://github.com/ggerganov/whisper.cpp.git
cd whisper.cpp/
pkg install sdl2 ffmpeg
cmake -D WHISPER_SUPPORT_SDL2=ON .
make
make install
Install the project...
-- Install configuration: "Release"
-- Installing: /usr/local/lib/libwhisper.so.1.5.4
-- Installing: /usr/local/lib/libwhisper.so.1
-- Set non-toolchain portion of runtime path of "/usr/local/lib/libwhisper.so.1.5.4" to "/usr/local/lib"
-- Installing: /usr/local/lib/libwhisper.so
-- Installing: /usr/local/include/ggml.h
-- Installing: /usr/local/include/whisper.h

cd /usr/local/src/whisper.cpp/examples
make
make install

Base EN
bash ./models/download-ggml-model.sh base.en
cd /usr/local/src/whisper.cpp/bin/
./main -m /usr/local/src/whisper.cpp/models/ggml-base.en.bin -f ../samples/jfk.wav

Medium EN
bash ./models/download-ggml-model.sh medium.en
cd /usr/local/src/whisper.cpp/bin/
./main -m /usr/local/src/whisper.cpp/models/ggml-medium.en.bin -f ../samples/jfk.wav
 
You do A.I. in a certain programming language , using certain libraries.
Then you check which operating systems support that programming language and has that library.
 
I dont know if admins will get mad because we are reviving this post. But it is a interesant subject and I was doing some diggin those days. Here:
github.com

GitHub - ggerganov/llama.cpp: LLM inference in C/C++

LLM inference in C/C++. Contribute to ggerganov/llama.cpp development by creating an account on GitHub.
github.com github.com

This a bare implementation of meta llama on C++ that has FreeBSD instructions. I have compiled it myself today, and got it running with CPU acelleration only. There is instructions for running GPU acelerated, but you will need a newer GPU (I think from 2012 up maybe?) my GTX550 dosent support DRM drivers and vulkan, so is a no for me since FreeBSD dosent have support for CUDA yet (I know some posts with people running CUDA with linuxmulator but havent tried myself).

I found something but havent played with yet but will send it here:
github.com

GitHub - mtnwrw/fyusenet: FyuseNet is an OpenGL(ES) based library that allows to run neural network inference on GPUs that support OpenGL or OpenGL/ES, which is the case for most desktop and mobile GPUs on the market.

FyuseNet is an OpenGL(ES) based library that allows to run neural network inference on GPUs that support OpenGL or OpenGL/ES, which is the case for most desktop and mobile GPUs on the market. - mtn...
github.com github.com

This seens a way to run neural networks with OpenGL and I find a cool ideia since we use older cards with this "new" technology.
And a extra:
github.com

GitHub - Significant-Gravitas/AutoGPT: AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters.

AutoGPT is the vision of accessible AI for everyone, to use and to build on. Our mission is to provide the tools, so that you can focus on what matters. - Significant-Gravitas/AutoGPT
github.com github.com

I havent runned this also yet, but it seens neat. It can prompt itself, search into web and support local models. I really think that can be a automation tool, or even a daemon in some sort to do things in an OS.

Grab thoose things together and it may be possible to have a low resource local model in a BSD doing automations or any task that you want.
 
You must log in or register to reply here.
Back
Top