4 ms·
ML on AMD is currently impractical, the mainstream tools have hard deps on CUDA and the AMD projects are so far half-baked.
by lumost 6y ago
ML on AMD is currently impractical, the mainstream tools have hard deps on CUDA and the AMD projects are so far half-baked.
- alfalfasprout 6y agoI used to think so but recent testing with ROCM has shown the tables are starting to turn.