Many Users have a strong focus on AI/DL for their science and we provide some really high performance systems for this in Hilbert.
You should always use the latest versions of tools like TensorFlow, PyTorch and their dependencies like CUDA and cuDNN to get the maximum performance.
Since we got some new DGX-A100 systems we have some contact to the research and development teams at Nvidia.
Available Software
Inhalt nach Stichwort
Es ist kein Inhalt mit den angegebenen Stichworten vorhanden
TensorCores
To get some high performance Networks you need to use the latest features of the hardware. At the moment these are TensorCores which can be used directly from TensorFlow or Pytorch.
