Re: plan of deep learning team for next stable release?
On 12/1/20 1:01 PM, Steffen Möller wrote:
> The way to go is via our non-free autobuilders, not?
That would probably be the way to go, however I have slight doubts that
modifying an existing buildd would be accepted as a solution.
> Since many DDs are using the CUDA drivers already, I think we have
> already decided to trust them.
> It should be somehow possible for me to come up with a buildd machine
> should this help the discussion.
That sounds great!
My gut thinks that the hardware is the key issue here. We already have
Nvidia drivers and CUDA in the archive, but those cost volunteer time,
not project money.