Spaces:
Sleeping
Sleeping
""" Create Conv2d Factory Method | |
Hacked together by / Copyright 2020 Ross Wightman | |
""" | |
from .mixed_conv2d import MixedConv2d | |
from .cond_conv2d import CondConv2d | |
from .conv2d_same import create_conv2d_pad | |
def create_conv2d(in_channels, out_channels, kernel_size, **kwargs): | |
""" Select a 2d convolution implementation based on arguments | |
Creates and returns one of torch.nn.Conv2d, Conv2dSame, MixedConv2d, or CondConv2d. | |
Used extensively by EfficientNet, MobileNetv3 and related networks. | |
""" | |
if isinstance(kernel_size, list): | |
assert 'num_experts' not in kwargs # MixNet + CondConv combo not supported currently | |
assert 'groups' not in kwargs # MixedConv groups are defined by kernel list | |
# We're going to use only lists for defining the MixedConv2d kernel groups, | |
# ints, tuples, other iterables will continue to pass to normal conv and specify h, w. | |
m = MixedConv2d(in_channels, out_channels, kernel_size, **kwargs) | |
else: | |
depthwise = kwargs.pop('depthwise', False) | |
# for DW out_channels must be multiple of in_channels as must have out_channels % groups == 0 | |
groups = in_channels if depthwise else kwargs.pop('groups', 1) | |
if 'num_experts' in kwargs and kwargs['num_experts'] > 0: | |
m = CondConv2d(in_channels, out_channels, kernel_size, groups=groups, **kwargs) | |
else: | |
m = create_conv2d_pad(in_channels, out_channels, kernel_size, groups=groups, **kwargs) | |
return m | |