Re: Installing openAI's GPT-2 Ada AI Language Model

On Fri, Apr 21, 2023 at 9:41 AM Amar Takhar <verm@darkbeer.org> wrote:
>
> I have a tutorial I wrote on how to setup PyTorch on FreeBSD with Stable
> Diffusion.
>
>   https://github.com/verm/freebsd-stable-diffusion

Thanks for the pointer but it assumes I have a (nvidia) GPU available.
  Since this is a VM on bhyve there is no currently working GPU
passthrough for it.  (I do use a nVidia Geforce GT 1030 with
nvidia-modeselect but it is not passable to the VM and due to hardware
resources don't have a physical machine I can assign to the task).

I have gotten as far as the part in the blog as installing pyTorch and
keep getting "no compatible versions found" out of pip and when I
search for any suitable ones it seems to not be there.   Thus I have
to revert to building pytorch from scratch and it blows up because the
host/guest CPU is not Intel.

>
> There are some issues with getting PyTorch working on FreeBSD and it's kind of a
> moving target I had it working once but it quickly broke this was years ago
> using the Linuxlator it's been rock solid for years.
>
> Another thing to note is the miniconda-installer script is hardcoded to an old
> version of Python you'll have to edit line 56 in the script and change it to 3.9
> or 3.10.  Not sure what to do about this yet I haven't had a chance to email the
> maintainer.
>
> I'd recommend going this route.
>
>
> Amar.
>


-- 
Aryeh M. Friedman, Lead Developer, http://www.PetiteCloud.org