zlacker

To be fair to AMD, they've been trying to solve ML workload portability at more fundamental levels with the acquisition of Nod.ai and de-facto incorporation of Google's IREE compiler project + MLIR.