bitorch.runtime_mode.RuntimeMode

class bitorch.runtime_mode.RuntimeMode(value)[source]

Enum for BITorch modes:

  • DEFAULT: use the default implementation of all layers

  • CPU: use layer implementations for inference on CPU

  • GPU: use layer implementations for inference on GPU

  • INFERENCE_AUTO: use an automatic layer that uses the fastest implementation available (not recommended)

  • RAW: while in this mode, new layers are created as the default implementation BUT without wrapping, so they can not be switched to other layers later on (it does not influence already wrapped layers)

Methods

available_values

list_of_names

is_single_mode

is_combined_mode

from_string

mode_compatible

is_supported_by

Attributes

RAW

DEFAULT

CPU

GPU

INFERENCE_AUTO