it shouldn't be this hard
After messing around so long it’s a little shocking that the solution was so simple, to get accelerated tensorflow on a supported AMD GPU:
rocm-dkms
)because it has been up-streamed and most of the interfaces are in kernel 5.8 already.rccl
from source it’s a waste of time whatever was wrong for that bloke is fixedAnd that’s it.
there are no comments yet