def model(self, num_classes, model_type):
block1 = MobileNetBlockConfig(4, 8, 3, num_repeats=2, stride=2)
block2 = MobileNetBlockConfig(8, 16, 3, num_repeats=1, stride=2)
blocks = [block1, block2]
model = model_type(num_classes, blocks, [1, 2, 3, 4])
yield model
del model
gc.collect()
def model(self, model_type):
block1 = MobileNetBlockConfig(4,
8, (3, 3, 3),
num_repeats=2,
stride=(1, 2, 2))
block2 = MobileNetBlockConfig(8,
16, (3, 3, 3),
num_repeats=1,
stride=(1, 2, 2))
blocks = [block1, block2]
model = model_type(blocks, [1, 2, 3, 4],
fpn_kwargs={"stride": (1, 2, 2)})
yield model
del model
gc.collect()
def model(self, model_type):
block = MobileNetBlockConfig(1, 1, kernel_size=3)
model = model_type.from_identical_blocks(block,
in_channels=1,
levels=[1, 2, 3])
yield model
del model
gc.collect()
def test_from_config(self, model_type):
config = MobileNetBlockConfig(
input_filters=4,
output_filters=4,
kernel_size=3,
stride=1,
dilation=1,
drop_connect_rate=0.0,
squeeze_excite_ratio=2,
expand_ratio=4,
use_skipconn=True,
)
model = model_type.from_config(config)
assert isinstance(model, model_type)
def test_construct(self, model_type, num_classes):
block1 = MobileNetBlockConfig(4, 8, 3, num_repeats=2)
block2 = MobileNetBlockConfig(8, 16, 3, num_repeats=1)
blocks = [block1, block2]
model_type(num_classes, blocks, [1, 2])
gc.collect()
class _EfficientNet(nn.Module):
DEFAULT_BLOCKS = [
MobileNetBlockConfig(
kernel_size=3,
num_repeats=1,
input_filters=32,
output_filters=16,
expand_ratio=1,
use_skipconn=True,
stride=1,
squeeze_excite_ratio=4,
global_se=True,
drop_connect_rate=0.2,
bn_momentum=0.01,
bn_epsilon=1e-3,
),
MobileNetBlockConfig(
kernel_size=3,
num_repeats=2,
input_filters=16,
output_filters=24,
expand_ratio=6,
use_skipconn=True,
stride=2,
squeeze_excite_ratio=4,
global_se=True,
drop_connect_rate=0.2,
bn_momentum=0.01,
bn_epsilon=1e-3,
),
MobileNetBlockConfig(
kernel_size=5,
num_repeats=2,
input_filters=24,
output_filters=40,
expand_ratio=6,
use_skipconn=True,
stride=2,
squeeze_excite_ratio=4,
global_se=True,
drop_connect_rate=0.2,
bn_momentum=0.01,
bn_epsilon=1e-3,
),
MobileNetBlockConfig(
kernel_size=3,
num_repeats=3,
input_filters=40,
output_filters=80,
expand_ratio=6,
use_skipconn=True,
stride=2,
squeeze_excite_ratio=4,
global_se=True,
drop_connect_rate=0.2,
bn_momentum=0.01,
bn_epsilon=1e-3,
),
MobileNetBlockConfig(
kernel_size=5,
num_repeats=3,
input_filters=80,
output_filters=112,
expand_ratio=6,
use_skipconn=True,
stride=1,
squeeze_excite_ratio=4,
global_se=True,
drop_connect_rate=0.2,
bn_momentum=0.01,
bn_epsilon=1e-3,
),
MobileNetBlockConfig(
kernel_size=5,
num_repeats=4,
input_filters=112,
output_filters=192,
expand_ratio=6,
use_skipconn=True,
stride=2,
squeeze_excite_ratio=4,
global_se=True,
drop_connect_rate=0.2,
bn_momentum=0.01,
bn_epsilon=1e-3,
),
MobileNetBlockConfig(
kernel_size=3,
num_repeats=1,
input_filters=192,
output_filters=320,
expand_ratio=6,
use_skipconn=True,
stride=1,
squeeze_excite_ratio=4,
global_se=True,
drop_connect_rate=0.2,
bn_momentum=0.01,
bn_epsilon=1e-3,
),
]
def __init__(
self,
block_configs: List[MobileNetBlockConfig],
width_coeff: float = 1.0,
depth_coeff: float = 1.0,
width_divisor: float = 8.0,
min_width: Optional[int] = None,
stem: Optional[nn.Module] = None,
head: Optional[nn.Module] = None,
):
super().__init__()
self.__block_configs = tuple(deepcopy(block_configs))
self.__width_coeff = float(width_coeff)
self.__depth_coeff = float(depth_coeff)
output_filters = []
for config in self.__block_configs:
# update config according to scale coefficients
config.input_filters = self.round_filters(config.input_filters,
width_coeff,
width_divisor, min_width)
config.output_filters = self.round_filters(config.output_filters,
width_coeff,
width_divisor,
min_width)
config.num_repeats = self.round_repeats(depth_coeff,
config.num_repeats)
output_filters.append(config.output_filters)
self.__input_filters = self.__block_configs[0].input_filters
self.__output_filters = tuple(output_filters)
# Conv stem (default stem used if none given)
if stem is not None:
self.stem = stem
else:
in_channels = 3
first_block = next(iter(self.__block_configs))
output_filters = first_block.input_filters
bn_momentum = first_block.bn_momentum
bn_epsilon = first_block.bn_epsilon
self.stem = nn.Sequential(
DynamicSamePad(
self.Conv(in_channels,
output_filters,
kernel_size=3,
stride=2,
bias=False,
padding=1)),
self.BatchNorm(output_filters,
momentum=bn_momentum,
eps=bn_epsilon),
)
# MobileNetV3 convolution blocks
blocks = []
for config in self.__block_configs:
conv_block = self.__class__._get_blocks(config)
blocks.append(conv_block)
self.blocks = nn.ModuleList(blocks)
# Head
self.head = head
@torch.jit.unused
@property
def input_filters(self) -> int:
r"""Number of input filters for the first level of backbone. When using a custom stem, use this
property to determine the number of filters in the stem's output.
"""
return self.__input_filters
@torch.jit.unused
@property
def output_filters(self) -> Tuple[int, ...]:
r"""Number of filters in each level of the BiFPN. When using a custom head, use this
property to determine the number of filters in the head's input.
"""
return self.__output_filters
@torch.jit.unused
@property
def block_configs(self) -> Tuple[MobileNetBlockConfig, ...]:
r"""Number of filters in each level of the BiFPN. When using a custom head, use this
property to determine the number of filters in the head's input.
"""
return self.__block_configs
@torch.jit.unused
@property
def width_coeff(self) -> float:
r"""Width coefficient for scaling"""
return self.__width_coeff
@torch.jit.unused
@property
def depth_coeff(self) -> float:
r"""Depth coefficient for scaling"""
return self.__depth_coeff
def extract_features(self,
inputs: Tensor,
return_all: bool = False) -> List[Tensor]:
r"""Runs the EfficientNet stem and body to extract features, returning a list of
tensors representing features extracted from each block.
Args:
inputs (:class:`torch.Tensor`):
Model inputs
return_all (bool):
By default, only features extracted from blocks with non-unit stride will
be returned. If ``return_all=True``, return features extracted from every
block group in the model.
"""
outputs: List[Tensor] = []
x = self.stem(inputs)
prev_x = x
for block in self.blocks:
x = block(prev_x)
if return_all or prev_x.shape[-1] > x.shape[-1]:
outputs.append(x)
prev_x = x
return outputs
def forward(self,
inputs: Tensor,
use_all_features: bool = False) -> List[Tensor]:
r"""Runs the entire EfficientNet model, including stem, body, and head.
If no head was supplied, the output of :func:`extract_features` will be returned.
Otherwise, the output of the given head will be returned.
.. note::
The returned output will always be a list of tensors. If a custom head is given
and it returns a single tensor, that tensor will be wrapped in a list before
being returned.
Args:
inputs (:class:`torch.Tensor`):
Model inputs
return_all (bool):
By default, only features extracted from blocks with non-unit stride will
be returned. If ``return_all=True``, return features extracted from every
block group in the model.
"""
output = self.extract_features(inputs, use_all_features)
if self.head is not None:
output = self.head(output)
if not isinstance(output, list):
output = [
output,
]
return output
def round_filters(self, filters, width_coeff, width_divisor, min_width):
if not width_coeff:
return filters
filters *= width_coeff
min_width = min_width or width_divisor
new_filters = max(
min_width,
int(filters + width_divisor / 2) // width_divisor * width_divisor)
# prevent rounding by more than 10%
if new_filters < 0.9 * filters:
new_filters += width_divisor
return int(new_filters)
def round_repeats(self, depth_coeff, num_repeats):
if not depth_coeff:
return num_repeats
return int(math.ceil(depth_coeff * num_repeats))
@classmethod
def from_predefined(cls,
compound_coeff: int,
block_overrides: Dict[str, Any] = {},
**kwargs) -> "_EfficientNet":
r"""Creates an EfficientNet model using one of the parameterizations defined in the
`EfficientNet paper`_.
Args:
compound_coeff (int):
Compound scaling parameter :math:`\phi`. For example, to construct EfficientNet-B0, set
``compound_coeff=0``.
block_overrides (dict):
Overrides to be applied to each :class:`combustion.nn.MobileNetBlockConfig`.
**kwargs:
Additional parameters/overrides for model constructor.
.. _EfficientNet paper:
https://arxiv.org/abs/1905.11946
"""
# from paper
alpha = 1.2
beta = 1.1
width_divisor = 8.0
depth_coeff = alpha**compound_coeff
width_coeff = beta**compound_coeff
# apply config overrides at each block
block_configs = deepcopy(cls.DEFAULT_BLOCKS)
for k, v in block_overrides.items():
for config in block_configs:
setattr(config, str(k), v)
final_kwargs = {
"block_configs": block_configs,
"width_coeff": width_coeff,
"depth_coeff": depth_coeff,
"width_divisor": width_divisor,
}
final_kwargs.update(kwargs)
model = cls(**final_kwargs)
model.compound_coeff = compound_coeff
return model
def input_size(self,
compound_coeff: Optional[int] = None) -> Tuple[int, ...]:
r"""Returns the expected input size for a given compound coefficient.
If a model was not created with :func:`from_predefined` a compound coefficient
must be provided manually.
Scaled resolution is computed as
.. math::
r = 512 + 128 * \phi
Args:
* compound_coeff (int):
Compound coefficient :math:`\phi` to compute image size for.
Returns:
Tuple of ints giving the expected input spatial resolution
"""
if compound_coeff is not None:
x = int(compound_coeff)
elif hasattr(self, "compound_coeff"):
x = int(self.compound_coeff)
else:
raise ValueError(
"compound_coeff must be provided when model was not created via from_predefined()"
)
return self.__class__.Tuple(512 + 128 * x)
def config(self, num_repeats):
return MobileNetBlockConfig(input_filters=4,
output_filters=4,
kernel_size=3,
num_repeats=num_repeats)
def test_construct(self, model_type):
block1 = MobileNetBlockConfig(4, 8, 3, num_repeats=2)
block2 = MobileNetBlockConfig(8, 16, 3, num_repeats=1)
blocks = [block1, block2]
m = model_type(blocks, 1.0, 1.0)
del m
def model(self, model_type):
block1 = MobileNetBlockConfig(4, 8, 3, num_repeats=2, stride=2)
block2 = MobileNetBlockConfig(8, 16, 3, num_repeats=1, stride=2)
blocks = [block1, block2]
return model_type(blocks, 1.0, 1.0)