PyTorch-DirectML: Train your machine learning models on any GPU (devblogs.microsoft.com) 14 points by mands 4y ago ↗ HN
[–] lostmsu 4y ago ↗ The TensorFlow version using DirectML was 4x slower than CUDA on the same machine. Is this comparable?
1 comment
[ 3.2 ms ] story [ 14.5 ms ] thread