enable_cudnn

dragon.cuda.enable_cudnn(
  enabled=True,
  deterministic=False,
  benchmark=False,
  allow_tf32=False,
  disabled_ops=None
)[source]

Enable backend to use the cuDNN library.

Parameters:
  • enabled (bool, optional, default=True) – Use cuDNN library or not.
  • deterministic (bool, optional, default=False) – Select deterministic algorithms instead of fastest.
  • benchmark (bool, optional, default=False) – Select fastest algorithms via benchmark or heuristics.
  • allow_tf32 (bool, optional, default=False) – Allow TF32 tensor core operation or not.
  • disabled_ops (Sequence[str], optional) – The operator types to disable using cuDNN.