Full threaddeserialized·When they get an equivalent to CUDA and the nvidia container runtime we can talkView on HN