PyTorch-DirectML: Train your machine learning models on any GPU (opens in new tab)

(devblogs.microsoft.com)

14 pointsmands4y ago1 comments

1 comments

1 comments · 1 top-level

The TensorFlow version using DirectML was 4x slower than CUDA on the same machine. Is this comparable?

j / k navigate · click thread line to collapse