def __init__(self, block, drop_connect_rate):
super().__init__()
self._bn_mom = block._bn_mom
self._bn_eps = block._bn_eps
self.id_skip = block.id_skip # skip connection and drop connect
self.drop_connect_rate = drop_connect_rate
self.input_filters = block._block_args.input_filters
self.output_filters = block._block_args.output_filters
self.stride = block._block_args.stride
# Expansion phase
self._expand_conv = block._expand_conv
self._bn0 = block._bn0
# Depthwise convolution phase
self._depthwise_conv = block._depthwise_conv
self._bn1 = block._bn1
# Squeeze and Excitation layer, if desired
self._se_reduce = block._se_reduce
self._se_expand = block._se_expand
# Output phase
self._project_conv = block._project_conv
self._bn2 = block._bn2
self._swish = Swish()
def set_swish(self, memory_efficient=True):
"""Sets swish function as memory efficient (for training) or standard (for export)"""
self._swish = MemoryEfficientSwish() if memory_efficient else Swish()
for block in self._blocks:
block.set_swish(memory_efficient)
def set_swish(self, memory_efficient: bool = True) -> NoReturn:
"""Sets swish function as memory efficient (training) or standard."""
self._swish = MemoryEfficientSwish() if memory_efficient else Swish()
for block in self._blocks:
block.set_swish(memory_efficient)
import torch
