Hi, I was running the imagenet in PyTorch example.
Do I need the add code or give the option for using tensor core? or Pytorch use the tensor core as default mode?
Thanks
Hi, I was running the imagenet in PyTorch example.
Do I need the add code or give the option for using tensor core? or Pytorch use the tensor core as default mode?
Thanks
Besides what @Dwight_Foster mentioned, on Ampere GPUs the tensor cores will be used through TF32.