floatngupstream

Yes, but this just makes a frictionless runtime for inference even more important (which is something that does not exist in a comparable form for AMD).

There is an enormous investment beside the training side. Once you have your model, you still need to run it. This is where Triton, TensorRT, and handcrafted CUDA kernels as plugins come in. There is no equivalent on ROCm for this (MIGraphX is not close).