Shortcuts

Source code for torchvision.models.mobilenetv3

from functools import partial
from typing import Any, Callable, List, Optional, Sequence

import torch
from torch import nn, Tensor

from ..ops.misc import Conv2dNormActivation, SqueezeExcitation as SElayer
from ..transforms._presets import ImageClassification
from ..utils import _log_api_usage_once
from ._api import register_model, Weights, WeightsEnum
from ._meta import _IMAGENET_CATEGORIES
from ._utils import _make_divisible, _ovewrite_named_param, handle_legacy_interface


__all__ = [
    "MobileNetV3",
    "MobileNet_V3_Large_Weights",
    "MobileNet_V3_Small_Weights",
    "mobilenet_v3_large",
    "mobilenet_v3_small",
]


class InvertedResidualConfig:
    # Stores information listed at Tables 1 and 2 of the MobileNetV3 paper
    def __init__(
        self,
        input_channels: int,
        kernel: int,
        expanded_channels: int,
        out_channels: int,
        use_se: bool,
        activation: str,
        stride: int,
        dilation: int,
        width_mult: float,
    ):
        self.input_channels = self.adjust_channels(input_channels, width_mult)
        self.kernel = kernel
        self.expanded_channels = self.adjust_channels(expanded_channels, width_mult)
        self.out_channels = self.adjust_channels(out_channels, width_mult)
        self.use_se = use_se
        self.use_hs = activation == "HS"
        self.stride = stride
        self.dilation = dilation

    @staticmethod
    def adjust_channels(channels: int, width_mult: float):
        return _make_divisible(channels * width_mult, 8)


class InvertedResidual(nn.Module):
    # Implemented as described at section 5 of MobileNetV3 paper
    def __init__(
        self,
        cnf: InvertedResidualConfig,
        norm_layer: Callable[..., nn.Module],
        se_layer: Callable[..., nn.Module] = partial(SElayer, scale_activation=nn.Hardsigmoid),
    ):
        super().__init__()
        if not (1 <= cnf.stride <= 2):
            raise ValueError("illegal stride value")

        self.use_res_connect = cnf.stride == 1 and cnf.input_channels == cnf.out_channels

        layers: List[nn.Module] = []
        activation_layer = nn.Hardswish if cnf.use_hs else nn.ReLU

        # expand
        if cnf.expanded_channels != cnf.input_channels:
            layers.append(
                Conv2dNormActivation(
                    cnf.input_channels,
                    cnf.expanded_channels,
                    kernel_size=1,
                    norm_layer=norm_layer,
                    activation_layer=activation_layer,
                )
            )

        # depthwise
        stride = 1 if cnf.dilation > 1 else cnf.stride
        layers.append(
            Conv2dNormActivation(
                cnf.expanded_channels,
                cnf.expanded_channels,
                kernel_size=cnf.kernel,
                stride=stride,
                dilation=cnf.dilation,
                groups=cnf.expanded_channels,
                norm_layer=norm_layer,
                activation_layer=activation_layer,
            )
        )
        if cnf.use_se:
            squeeze_channels = _make_divisible(cnf.expanded_channels // 4, 8)
            layers.append(se_layer(cnf.expanded_channels, squeeze_channels))

        # project
        layers.append(
            Conv2dNormActivation(
                cnf.expanded_channels, cnf.out_channels, kernel_size=1, norm_layer=norm_layer, activation_layer=None
            )
        )

        self.block = nn.Sequential(*layers)
        self.out_channels = cnf.out_channels
        self._is_cn = cnf.stride > 1

    def forward(self, input: Tensor) -> Tensor:
        result = self.block(input)
        if self.use_res_connect:
            result += input
        return result


class MobileNetV3(nn.Module):
    def __init__(
        self,
        inverted_residual_setting: List[InvertedResidualConfig],
        last_channel: int,
        num_classes: int = 1000,
        block: Optional[Callable[..., nn.Module]] = None,
        norm_layer: Optional[Callable[..., nn.Module]] = None,
        dropout: float = 0.2,
        **kwargs: Any,
    ) -> None:
        """
        MobileNet V3 main class

        Args:
            inverted_residual_setting (List[InvertedResidualConfig]): Network structure
            last_channel (int): The number of channels on the penultimate layer
            num_classes (int): Number of classes
            block (Optional[Callable[..., nn.Module]]): Module specifying inverted residual building block for mobilenet
            norm_layer (Optional[Callable[..., nn.Module]]): Module specifying the normalization layer to use
            dropout (float): The droupout probability
        """
        super().__init__()
        _log_api_usage_once(self)

        if not inverted_residual_setting:
            raise ValueError("The inverted_residual_setting should not be empty")
        elif not (
            isinstance(inverted_residual_setting, Sequence)
            and all([isinstance(s, InvertedResidualConfig) for s in inverted_residual_setting])
        ):
            raise TypeError("The inverted_residual_setting should be List[InvertedResidualConfig]")

        if block is None:
            block = InvertedResidual

        if norm_layer is None:
            norm_layer = partial(nn.BatchNorm2d, eps=0.001, momentum=0.01)

        layers: List[nn.Module] = []

        # building first layer
        firstconv_output_channels = inverted_residual_setting[0].input_channels
        layers.append(
            Conv2dNormActivation(
                3,
                firstconv_output_channels,
                kernel_size=3,
                stride=2,
                norm_layer=norm_layer,
                activation_layer=nn.Hardswish,
            )
        )

        # building inverted residual blocks
        for cnf in inverted_residual_setting:
            layers.append(block(cnf, norm_layer))

        # building last several layers
        lastconv_input_channels = inverted_residual_setting[-1].out_channels
        lastconv_output_channels = 6 * lastconv_input_channels
        layers.append(
            Conv2dNormActivation(
                lastconv_input_channels,
                lastconv_output_channels,
                kernel_size=1,
                norm_layer=norm_layer,
                activation_layer=nn.Hardswish,
            )
        )

        self.features = nn.Sequential(*layers)
        self.avgpool = nn.AdaptiveAvgPool2d(1)
        self.classifier = nn.Sequential(
            nn.Linear(lastconv_output_channels, last_channel),
            nn.Hardswish(inplace=True),
            nn.Dropout(p=dropout, inplace=True),
            nn.Linear(last_channel, num_classes),
        )

        for m in self.modules():
            if isinstance(m, nn.Conv2d):
                nn.init.kaiming_normal_(m.weight, mode="fan_out")
                if m.bias is not None:
                    nn.init.zeros_(m.bias)
            elif isinstance(m, (nn.BatchNorm2d, nn.GroupNorm)):
                nn.init.ones_(m.weight)
                nn.init.zeros_(m.bias)
            elif isinstance(m, nn.Linear):
                nn.init.normal_(m.weight, 0, 0.01)
                nn.init.zeros_(m.bias)

    def _forward_impl(self, x: Tensor) -> Tensor:
        x = self.features(x)

        x = self.avgpool(x)
        x = torch.flatten(x, 1)

        x = self.classifier(x)

        return x

    def forward(self, x: Tensor) -> Tensor:
        return self._forward_impl(x)


def _mobilenet_v3_conf(
    arch: str, width_mult: float = 1.0, reduced_tail: bool = False, dilated: bool = False, **kwargs: Any
):
    reduce_divider = 2 if reduced_tail else 1
    dilation = 2 if dilated else 1

    bneck_conf = partial(InvertedResidualConfig, width_mult=width_mult)
    adjust_channels = partial(InvertedResidualConfig.adjust_channels, width_mult=width_mult)

    if arch == "mobilenet_v3_large":
        inverted_residual_setting = [
            bneck_conf(16, 3, 16, 16, False, "RE", 1, 1),
            bneck_conf(16, 3, 64, 24, False, "RE", 2, 1),  # C1
            bneck_conf(24, 3, 72, 24, False, "RE", 1, 1),
            bneck_conf(24, 5, 72, 40, True, "RE", 2, 1),  # C2
            bneck_conf(40, 5, 120, 40, True, "RE", 1, 1),
            bneck_conf(40, 5, 120, 40, True, "RE", 1, 1),
            bneck_conf(40, 3, 240, 80, False, "HS", 2, 1),  # C3
            bneck_conf(80, 3, 200, 80, False, "HS", 1, 1),
            bneck_conf(80, 3, 184, 80, False, "HS", 1, 1),
            bneck_conf(80, 3, 184, 80, False, "HS", 1, 1),
            bneck_conf(80, 3, 480, 112, True, "HS", 1, 1),
            bneck_conf(112, 3, 672, 112, True, "HS", 1, 1),
            bneck_conf(112, 5, 672, 160 // reduce_divider, True, "HS", 2, dilation),  # C4
            bneck_conf(160 // reduce_divider, 5, 960 // reduce_divider, 160 // reduce_divider, True, "HS", 1, dilation),
            bneck_conf(160 // reduce_divider, 5, 960 // reduce_divider, 160 // reduce_divider, True, "HS", 1, dilation),
        ]
        last_channel = adjust_channels(1280 // reduce_divider)  # C5
    elif arch == "mobilenet_v3_small":
        inverted_residual_setting = [
            bneck_conf(16, 3, 16, 16, True, "RE", 2, 1),  # C1
            bneck_conf(16, 3, 72, 24, False, "RE", 2, 1),  # C2
            bneck_conf(24, 3, 88, 24, False, "RE", 1, 1),
            bneck_conf(24, 5, 96, 40, True, "HS", 2, 1),  # C3
            bneck_conf(40, 5, 240, 40, True, "HS", 1, 1),
            bneck_conf(40, 5, 240, 40, True, "HS", 1, 1),
            bneck_conf(40, 5, 120, 48, True, "HS", 1, 1),
            bneck_conf(48, 5, 144, 48, True, "HS", 1, 1),
            bneck_conf(48, 5, 288, 96 // reduce_divider, True, "HS", 2, dilation),  # C4
            bneck_conf(96 // reduce_divider, 5, 576 // reduce_divider, 96 // reduce_divider, True, "HS", 1, dilation),
            bneck_conf(96 // reduce_divider, 5, 576 // reduce_divider, 96 // reduce_divider, True, "HS", 1, dilation),
        ]
        last_channel = adjust_channels(1024 // reduce_divider)  # C5
    else:
        raise ValueError(f"Unsupported model type {arch}")

    return inverted_residual_setting, last_channel


def _mobilenet_v3(
    inverted_residual_setting: List[InvertedResidualConfig],
    last_channel: int,
    weights: Optional[WeightsEnum],
    progress: bool,
    **kwargs: Any,
) -> MobileNetV3:
    if weights is not None:
        _ovewrite_named_param(kwargs, "num_classes", len(weights.meta["categories"]))

    model = MobileNetV3(inverted_residual_setting, last_channel, **kwargs)

    if weights is not None:
        model.load_state_dict(weights.get_state_dict(progress=progress))

    return model


_COMMON_META = {
    "min_size": (1, 1),
    "categories": _IMAGENET_CATEGORIES,
}


[docs]class MobileNet_V3_Large_Weights(WeightsEnum): IMAGENET1K_V1 = Weights( url="https://download.pytorch.org/models/mobilenet_v3_large-8738ca79.pth", transforms=partial(ImageClassification, crop_size=224), meta={ **_COMMON_META, "num_params": 5483032, "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#mobilenetv3-large--small", "_metrics": { "ImageNet-1K": { "acc@1": 74.042, "acc@5": 91.340, } }, "_ops": 0.217, "_file_size": 21.114, "_docs": """These weights were trained from scratch by using a simple training recipe.""", }, ) IMAGENET1K_V2 = Weights( url="https://download.pytorch.org/models/mobilenet_v3_large-5c1a4163.pth", transforms=partial(ImageClassification, crop_size=224, resize_size=232), meta={ **_COMMON_META, "num_params": 5483032, "recipe": "https://github.com/pytorch/vision/issues/3995#new-recipe-with-reg-tuning", "_metrics": { "ImageNet-1K": { "acc@1": 75.274, "acc@5": 92.566, } }, "_ops": 0.217, "_file_size": 21.107, "_docs": """ These weights improve marginally upon the results of the original paper by using a modified version of TorchVision's `new training recipe <https://pytorch.org/blog/how-to-train-state-of-the-art-models-using-torchvision-latest-primitives/>`_. """, }, ) DEFAULT = IMAGENET1K_V2
[docs]class MobileNet_V3_Small_Weights(WeightsEnum): IMAGENET1K_V1 = Weights( url="https://download.pytorch.org/models/mobilenet_v3_small-047dcff4.pth", transforms=partial(ImageClassification, crop_size=224), meta={ **_COMMON_META, "num_params": 2542856, "recipe": "https://github.com/pytorch/vision/tree/main/references/classification#mobilenetv3-large--small", "_metrics": { "ImageNet-1K": { "acc@1": 67.668, "acc@5": 87.402, } }, "_ops": 0.057, "_file_size": 9.829, "_docs": """ These weights improve upon the results of the original paper by using a simple training recipe. """, }, ) DEFAULT = IMAGENET1K_V1
[docs]@register_model() @handle_legacy_interface(weights=("pretrained", MobileNet_V3_Large_Weights.IMAGENET1K_V1)) def mobilenet_v3_large( *, weights: Optional[MobileNet_V3_Large_Weights] = None, progress: bool = True, **kwargs: Any ) -> MobileNetV3: """ Constructs a large MobileNetV3 architecture from `Searching for MobileNetV3 <https://arxiv.org/abs/1905.02244>`__. Args: weights (:class:`~torchvision.models.MobileNet_V3_Large_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.MobileNet_V3_Large_Weights` below for more details, and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. **kwargs: parameters passed to the ``torchvision.models.resnet.MobileNetV3`` base class. Please refer to the `source code <https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py>`_ for more details about this class. .. autoclass:: torchvision.models.MobileNet_V3_Large_Weights :members: """ weights = MobileNet_V3_Large_Weights.verify(weights) inverted_residual_setting, last_channel = _mobilenet_v3_conf("mobilenet_v3_large", **kwargs) return _mobilenet_v3(inverted_residual_setting, last_channel, weights, progress, **kwargs)
[docs]@register_model() @handle_legacy_interface(weights=("pretrained", MobileNet_V3_Small_Weights.IMAGENET1K_V1)) def mobilenet_v3_small( *, weights: Optional[MobileNet_V3_Small_Weights] = None, progress: bool = True, **kwargs: Any ) -> MobileNetV3: """ Constructs a small MobileNetV3 architecture from `Searching for MobileNetV3 <https://arxiv.org/abs/1905.02244>`__. Args: weights (:class:`~torchvision.models.MobileNet_V3_Small_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.MobileNet_V3_Small_Weights` below for more details, and possible values. By default, no pre-trained weights are used. progress (bool, optional): If True, displays a progress bar of the download to stderr. Default is True. **kwargs: parameters passed to the ``torchvision.models.resnet.MobileNetV3`` base class. Please refer to the `source code <https://github.com/pytorch/vision/blob/main/torchvision/models/mobilenetv3.py>`_ for more details about this class. .. autoclass:: torchvision.models.MobileNet_V3_Small_Weights :members: """ weights = MobileNet_V3_Small_Weights.verify(weights) inverted_residual_setting, last_channel = _mobilenet_v3_conf("mobilenet_v3_small", **kwargs) return _mobilenet_v3(inverted_residual_setting, last_channel, weights, progress, **kwargs)
# The dictionary below is internal implementation detail and will be removed in v0.15 from ._utils import _ModelURLs model_urls = _ModelURLs( { "mobilenet_v3_large": MobileNet_V3_Large_Weights.IMAGENET1K_V1.url, "mobilenet_v3_small": MobileNet_V3_Small_Weights.IMAGENET1K_V1.url, } )

Docs

Access comprehensive developer documentation for PyTorch

View Docs

Tutorials

Get in-depth tutorials for beginners and advanced developers

View Tutorials

Resources

Find development resources and get your questions answered

View Resources