\;jXE,ddlmZddlZddlmZddlmZddlmZddlm Z gZ d d d Z Gd d ejZ GddZGddejZGddejZGddeZGddeZddZddZddZy))partialN)nn)get_weights_path_from_url)ConvNormActivation_make_divisible)zIhttps://paddle-hapi.bj.bcebos.com/models/mobilenet_v3_small_x1.0.pdparams 34fe0e7c1f8b00b2b056ad6788d0590c)zIhttps://paddle-hapi.bj.bcebos.com/models/mobilenet_v3_large_x1.0.pdparams 118db5792b4e183b925d8e8e334db3df)zmobilenet_v3_small_x1.0zmobilenet_v3_large_x1.0c\eZdZdZej ej ffd ZdZdZ xZ S)SqueezeExcitationa This block implements the Squeeze-and-Excitation block from https://arxiv.org/abs/1709.01507 (see Fig. 1). Parameters ``activation``, and ``scale_activation`` correspond to ``delta`` and ``sigma`` in eq. 3. This code is based on the torchvision code with modifications. You can also see at https://github.com/pytorch/vision/blob/main/torchvision/ops/misc.py#L127 Args: input_channels (int): Number of channels in the input image. squeeze_channels (int): Number of squeeze channels. activation (Callable[..., paddle.nn.Layer], optional): ``delta`` activation. Default: ``paddle.nn.ReLU``. scale_activation (Callable[..., paddle.nn.Layer]): ``sigma`` activation. Default: ``paddle.nn.Sigmoid``. ct|tjd|_tj ||d|_tj ||d|_||_||_ y)Nr) super__init__rAdaptiveAvgPool2DavgpoolConv2Dfc1fc2 activationscale_activation)selfinput_channelssqueeze_channelsrr __class__s iG:\00. PROJECTS\API\Inventory\templateJSON\kerjaOCR\Lib\site-packages\paddle/vision/models/mobilenetv3.pyrzSqueezeExcitation.__init__4s` ++A. 99^-=qA99-~qA$, 0 2c|j|}|j|}|j|}|j|}|j |SN)rrrrrrinputscales r_scalezSqueezeExcitation._scaleBsI U#&$$U++rc.|j|}||zSr )r$r!s rforwardzSqueezeExcitation.forwardIs E"u}r) __name__ __module__ __qualname____doc__rReLUSigmoidrr$r& __classcell__rs@rrr&s' "77 3,rrc(eZdZ ddZeddZy)InvertedResidualConfigc t|j|||_||_|j|||_|j|||_||_|d|_||_ y|dk(rtj|_||_ y|dk(rtj|_||_ ytd|)N)r#relu hardswishz*The activation function is not supported: ) adjust_channels in_channelskernelexpanded_channels out_channelsuse_seactivation_layerrr+ Hardswish RuntimeErrorstride) rr5r6r7r8r9rr=r#s rrzInvertedResidualConfig.__init__Os // 5/I !%!5!5 U"6" !00U0K  $(D ! 6 !$&GGD ! ; &$&LLD !  `_. Args: config (list[InvertedResidualConfig]): MobileNetV3 depthwise blocks config. last_channel (int): The number of channels on the penultimate layer. scale (float, optional): Scale of channels in each layer. Default: 1.0. num_classes (int, optional): Output dim of last fc layer. If num_classes <=0, last fc layer will not be defined. Default: 1000. with_pool (bool, optional): Use pool before the last fc layer or not. Default: True. cXt|||_||_||_||_||_|dj|_|dj|_ |jdz|_ ttjdd}td|jddd d tj| |_tj"|jDcgc][}t%|j|j&|j(|j*|j,|j.|j0| ]c}|_t|j|jd d dd |tj |_|rtj6d |_|dkDrtj"tj:|j|jtjtj<d tj:|j||_yycc}w)NrgMbP?gGz?)epsilonmomentumrr)r5r8rGr=rHrKr:rI)r5r7r8rUr=r9r:rIrJg?)p) rrconfigr# last_channel num_classes with_poolr5firstconv_in_channelslastconv_in_channelslastconv_out_channelsrr BatchNorm2Drr;conv SequentialrEr7r8r6r=r9r:blockslastconvrrLinearDropout classifier) rrdrer#rfrgrIcfgrs rrzMobileNetV3.__init__s   (&"%+AY%:%:"$*2J$:$:!%)%>%>%B"R^^UTJ &33\\!   mm ;; 'C! #&)&;&;!$!1!1 # ::::%(%9%9) '   +1133!\\    //2DL ? mm $44d6G6GH  S! $++[9 DO ; sA H'c|j|}|j|}|j|}|jr|j |}|j dkDr't j|d}|j|}|S)Nrr) rlrnrorgrrfrWflattenrr)rrYs rr&zMobileNetV3.forwardso IIaL KKN MM!  >> QA   a q!$A"ArrAiT)r'r(r)r*rr&r-r.s@rr\r\s LP>@ rr\c$eZdZdZdfd ZxZS)MobileNetV3SmallaMobileNetV3 Small architecture model from `"Searching for MobileNetV3" `_. Args: scale (float, optional): Scale of channels in each layer. Default: 1.0. num_classes (int, optional): Output dim of last fc layer. If num_classes <= 0, last fc layer will not be defined. Default: 1000. with_pool (bool, optional): Use pool before the last fc layer or not. Default: True. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Small architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import MobileNetV3Small >>> # Build model >>> model = MobileNetV3Small(scale=1.0) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] ctddddddd|tddddddd|tddd dddd |tdd d d ddd|td d dd ddd |td d dd ddd |td d ddddd |tdd ddddd |tdd dd ddd|td d dd ddd |td d dd ddd |g }td|zd}t| |||||y)NrbTr2rHFXr`(r3x0i i@ir?rer#rgrfr0r rrrr#rfrgrdrers rrzMobileNetV3Small.__init__0s/ "2q"b$5 I "2q"b%E J "2q"b%E J "2q"b$ Q N "2q#r4a O "2q#r4a O "2q#r4a O "2q#r4a O "2q#r4a O "2q#r4a O "2q#r4a O  'te|Q7   %#  rrvr'r(r)r*rr-r.s@rrxrxs8  rrxc$eZdZdZdfd ZxZS)MobileNetV3LargeaMobileNetV3 Large architecture model from `"Searching for MobileNetV3" `_. Args: scale (float, optional): Scale of channels in each layer. Default: 1.0. num_classes (int, optional): Output dim of last fc layer. If num_classes <= 0, last fc layer will not be defined. Default: 1000. with_pool (bool, optional): Use pool before the last fc layer or not. Default: True. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Large architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import MobileNetV3Large >>> # Build model >>> model = MobileNetV3Large(scale=1.0) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] cNtddddddd|tddddddd|tddd dddd|tdd d d d dd|td d d d d dd|td d d d d dd|td dddddd|tddddddd|tddddddd|tddddddd|tddddd dd|tddddd dd|tdd ddd dd|tdd ddd dd|tdd ddd dd|g}td|zd}t| |||||y)NrzrbFr2r@r|rr{r~rTrrPr3ipiiir?rrrs rrzMobileNetV3Large.__init__es "2q"b%E J "2q"b%E J "2q"b%E J "2q"b$5 I "2q#r4E J "2q#r4E J "AsB{Au  #AsB{Au  #AsB{Au  #AsB{Au  #AsC{Au  #QS$ Q  #QS$ Q  #QS$ Q  #QS$ Q ?" F'te|Q7   %#  rrvrr.s@rrrHs8+ + rrc |dk(rtdd|i|}n tdd|i|}|r_|d|}|tvs J|dtt|dt|d}t j |}|j ||S)Nmobilenet_v3_larger#_xzJ model do not have a pretrained model now, you should set pretrained=FalserrrC)rrx model_urlsrrWloadset_dict)arch pretrainedr#kwargsmodel weight_pathparams r _mobilenet_v3rs ## 7u77 7u77r%! J  _V] ^ _ / t Q D!1!!4  K( u Lrc $t d||d|}|S)aMobileNetV3 Small architecture model from `"Searching for MobileNetV3" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. scale (float, optional): Scale of channels in each layer. Default: 1.0. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`MobileNetV3Small `. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Small architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import mobilenet_v3_small >>> # Build model >>> model = mobilenet_v3_small() >>> # Build model and load imagenet pretrained weight >>> # model = mobilenet_v3_small(pretrained=True) >>> # Build mobilenet v3 small model with scale=0.5 >>> model = mobilenet_v3_small(scale=0.5) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] r#r)mobilenet_v3_smallrrr#rrs rrr+B  $)j DJ E Lrc $t d||d|}|S)aMobileNetV3 Large architecture model from `"Searching for MobileNetV3" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. scale (float, optional): Scale of channels in each layer. Default: 1.0. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`MobileNetV3Large `. Returns: :ref:`api_paddle_nn_Layer`. An instance of MobileNetV3 Large architecture model. Examples: .. code-block:: python >>> import paddle >>> from paddle.vision.models import mobilenet_v3_large >>> # Build model >>> model = mobilenet_v3_large() >>> # Build model and load imagenet pretrained weight >>> # model = mobilenet_v3_large(pretrained=True) >>> # Build mobilenet v3 large model with scale=0.5 >>> model = mobilenet_v3_large(scale=0.5) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) [1, 1000] r)rrrs rrrrr)FrA) functoolsrrWrpaddle.utils.downloadropsr_utilsr __all__rLayerrr0rEr\rxrrrrrCrrrs ;$#     %%P!4!4HBrxxBJY"((Yx2 {2 jH {H V&$N$r