2 ms·
There is a lot that has been done. AMD's HIP and the open SYCL infrastructure can already run TensorFlow and can run a large amount of CUDA code. The reason the
by sudosysgen 6y ago
There is a lot that has been done. AMD's HIP and the open SYCL infrastructure can already run TensorFlow and can run a large amount of CUDA code. The reason the transition isn't being done is momentum and tradition, really.