|jyUUddlmZddlmZmZddlZddlmZddlmZm Z m Z ddl m Z ddl mZerHdd lmZmZdd lmZmZdd lmZdd lmZed ZedZGddeZGddeZgZdddddddd Zded<d?d Zd@d%ZGd&d'ej Z!Gd(d)ej Z"Gd*d+ej Z# dAdBd6Z$ dCdDd7Z% dCdDd8Z& dCdDd9Z' dCdDd:Z( dCdDd;Z) dCdDd<Z* dCdEd>Z+dS)F) annotations) TYPE_CHECKINGCallableN)nn)AdaptiveAvgPool2DLinear MaxPool2D)get_weights_path_from_url)ConvNormActivation)Literal TypedDict) NotRequiredUnpack)Tensor)Size2)shufflenet_v2_x0_25shufflenet_v2_x0_33shufflenet_v2_x0_5shufflenet_v2_x1_0shufflenet_v2_x1_5shufflenet_v2_x2_0shufflenet_v2_swish)reluswishc.eZdZUded<ded<ded<dS)_ShuffleNetOptionsz#NotRequired[_ActivationType | None]actNotRequired[bool] with_poolNotRequired[int] num_classesN__name__ __module__ __qualname____annotations__q/lsinfo/ai/hellotax_ai/data_center/backend/venv/lib/python3.11/site-packages/paddle/vision/models/shufflenetv2.pyrr.s60000$$$$%%%%%%r)rc$eZdZUded<ded<dS)_ShuffleNetSwishOptionsrr r!r"Nr#r(r)r*r,r,3s*$$$$%%%%%%r)r,)zEhttps://paddle-hapi.bj.bcebos.com/models/shufflenet_v2_x0_25.pdparams 1e509b4c140eeb096bb16e214796d03b)zEhttps://paddle-hapi.bj.bcebos.com/models/shufflenet_v2_x0_33.pdparams 3d7b3ab0eaa5c0927ff1026d31b729bd)zDhttps://paddle-hapi.bj.bcebos.com/models/shufflenet_v2_x0_5.pdparams 5e5cee182a7793c4e4c73949b1a71bd4)zDhttps://paddle-hapi.bj.bcebos.com/models/shufflenet_v2_x1_0.pdparams 122d42478b9e81eb49f8a9ede327b1a4)zDhttps://paddle-hapi.bj.bcebos.com/models/shufflenet_v2_x1_5.pdparams faced5827380d73531d0ee027c67826d)zDhttps://paddle-hapi.bj.bcebos.com/models/shufflenet_v2_x2_0.pdparams cd3dddcd8305e7bcd8ad14d1c69a5784)zEhttps://paddle-hapi.bj.bcebos.com/models/shufflenet_v2_swish.pdparams adde0aa3b023e5b0c94a68be1c394b84zdict[str, tuple[str, str]] model_urlsr_ActivationType | Nonereturnnn.Layer | Nonecv|dkr tjS|dkr tjS|dStd|)Nrrz*The activation function is not supported: )rSwishReLU RuntimeError)rs r*create_activation_layerr<ZsC g~~x w tMMMNNNr)xrgroupsintc|jdd\}}}}||z}tj||||||g}tj|gd}tj|||||g}|S)Nr)shape)rr rA)perm)rBpaddlereshape transpose)r=r> batch_size num_channelsheightwidthchannels_per_groups r*channel_shufflerNes./gacl+J fe%/  *f&8&%H   A 111A q\65 IJJJA Hr)c6eZdZejfdfd Zdd ZxZS)InvertedResidual in_channelsr? out_channelsstrideractivation_layerCallable[..., nn.Layer]r6Nonec tt|dz|dzdddd||_t|dz|dzd|d|dzd|_t|dz|dzdddd||_dS)Nr rCrrQrR kernel_sizerSpaddingr>rTrD)super__init__r _conv_pw_conv_dw _conv_linearselfrQrRrSrT __class__s r*r\zInvertedResidual.__init__ws *#q(%*-    +$)%*1$!    /$)%*-   r)inputsrc@tj||jddz|jddzgd\}}||}||}||}tj||gd}t|dS)NrCr )num_or_sectionsaxisrf)rFsplitrBr]r^r_concatrNrarcx1x2outs r*forwardzInvertedResidual.forwards #\!_16<?a3GH   B ]]2   ]]2     r " "mRH1---sA&&&r) rQr?rRr?rSrrTrUr6rVrcrr6rr$r%r&rr:r\rn __classcell__rbs@r*rPrPvs^ 57G " " " " " " " H ' ' ' ' ' ' ' 'r)rPc6eZdZejfdfd Zdd ZxZS)InvertedResidualDSrQr?rRrSrrTrUr6rVc tt||d|d|d|_t||dzdddd||_t||dzdddd||_t|dz|dzd|d|dzd|_t|dz|dzdddd||_dS)NrDrCrXr r)r[r\r _conv_dw_1_conv_linear_1 _conv_pw_2 _conv_dw_2_conv_linear_2r`s r*r\zInvertedResidualDS.__init__s' -#$!   1#%*-   -#%*-   -$)%*1$!   1$)%*-   r)rcrc$||}||}||}||}||}t j||gd}t|dS)NrCrgr )rwrxryrzr{rFrirNrjs r*rnzInvertedResidualDS.forwards __V $ $   $ $ __V $ $ __R    $ $mRH1---sA&&&r)rorprqrss@r*rurus^ 57G 7 7 7 7 7 7 7 r''''''''r)rucVeZdZUdZded<ded<ded< ddfd ZddZxZS) ShuffleNetV2aShuffleNetV2 model from `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: scale (float, optional): Scale of output channels. Default: True. act (str, optional): Activation function of neural network. Default: "relu". 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 ShuffleNetV2 model. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import ShuffleNetV2 >>> shufflenet_v2_swish = ShuffleNetV2(scale=1.0, act="swish") >>> x = paddle.rand([1, 3, 224, 224]) >>> out = shufflenet_v2_swish(x) >>> print(out.shape) paddle.Size([1, 1000]) floatscaler?r"boolr ?rTrr5r6rVc t||_||_||_gd}t |}|dkrgd}nY|dkrgd}nN|dkrgd}nC|dkrgd }n8|d krgd }n-|d krgd }n"t dt|zdztd|dddd||_ tddd|_ g|_ t|D]\}} t| D]} | dkra|t!||dz||dzd|t|dzdzt| dzz} n`|t#||dz||dzd|t|dzdzt| dzz} |j | t|d|dddd||_|rt)d|_|dkr*|d|_t/|d||_dSdS)N)rArA?)r0`Q?)rr @r?)rrrrr)rrtir?)rri`ir@)rriiizThis scale size:[z] is not implemented!rDrCr )rQrRrYrSrZrT)rYrSrZr)rQrRrSrT_)sublayernamer)r[r\rr"r r<NotImplementedErrorstrr _conv1r _max_pool _block_list enumeraterange add_sublayerrurPappend _last_convr _pool2d_avg_out_cr_fc) rarrr"r stage_repeatsrTstage_out_channelsstage_id num_repeatiblockrbs r*r\zShuffleNetV2.__init__ s  &"! 2377 D==!:!:!:   d]]!;!;!;   c\\!!>!>   c\\!>!>!>   c\\!>!>!>  %#c%jj03JJ )+A.-     #qAFFF$-m$<$< / / Hj:&& / /66 --!3(:8a<(H);HqL)I#$-= """ !A..4s1q5zzA.EE!--!1(:8a<(H);HqL)I#$-= """ !A..4s1q5zzA.E ''....+ /.-*2.+B/-      4033D  ??,R0DK04kBBDHHH ?r)rcrcT||}||}|jD] }||}||}|jr||}|jdkr,tj|dd}| |}|S)NrrCr) start_axis stop_axis) rrrrr rr"rFflattenr)rarcr=invs r*rnzShuffleNetV2.forward`s KK   NN1  #  CAAA OOA   > $  ##A  a  qQ"===A Ar))rrrT) rrrr5r"r?r rr6rVrp)r$r%r&__doc__r'r\rnrrrss@r*r~r~s4LLLOOO&, RCRCRCRCRCRCRCh        r)r~Frarch_ShuffleNetArch pretrainedrrrkwargsUnpack[_ShuffleNetOptions]c tdd|i|}|rq|tvs J|dtt|dt|d}tj|}|||S)NrzJ model do not have a pretrained model now, you should set pretrained=FalserrCr()r~r4r rFloadset_dict)rrrrmodel weight_pathparams r*_shufflenet_v2rps  / /u / / /E z!!! _ _ _"!!0 t Q D!1!!4   K(( u Lr)c "t dd|d|S)aaShuffleNetV2 with 0.25x output channels, as described in `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`ShuffleNetV2 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of ShuffleNetV2 with 0.25x output channels. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import shufflenet_v2_x0_25 >>> # build model >>> model = shufflenet_v2_x0_25() >>> # build model and load imagenet pretrained weight >>> # model = shufflenet_v2_x0_25(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) paddle.Size([1, 1000]) rrrr)rrrrs r*rr1@  %)j  DJ  r)c "t dd|d|S)aaShuffleNetV2 with 0.33x output channels, as described in `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`ShuffleNetV2 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of ShuffleNetV2 with 0.33x output channels. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import shufflenet_v2_x0_33 >>> # build model >>> model = shufflenet_v2_x0_33() >>> # build model and load imagenet pretrained weight >>> # model = shufflenet_v2_x0_33(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) paddle.Size([1, 1000]) rrr)rrrs r*rrrr)c "t dd|d|S)a\ShuffleNetV2 with 0.5x output channels, as described in `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`ShuffleNetV2 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of ShuffleNetV2 with 0.5x output channels. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import shufflenet_v2_x0_5 >>> # build model >>> model = shufflenet_v2_x0_5() >>> # build model and load imagenet pretrained weight >>> # model = shufflenet_v2_x0_5(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) paddle.Size([1, 1000]) rrr)rrrs r*rr1@  $'J  BH  r)c "t dd|d|S)a\ShuffleNetV2 with 1.0x output channels, as described in `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`ShuffleNetV2 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of ShuffleNetV2 with 1.0x output channels. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import shufflenet_v2_x1_0 >>> # build model >>> model = shufflenet_v2_x1_0() >>> # build model and load imagenet pretrained weight >>> # model = shufflenet_v2_x1_0(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) paddle.Size([1, 1000]) rrr)rrrs r*rrrr)c "t dd|d|S)a\ShuffleNetV2 with 1.5x output channels, as described in `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`ShuffleNetV2 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of ShuffleNetV2 with 1.5x output channels. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import shufflenet_v2_x1_5 >>> # build model >>> model = shufflenet_v2_x1_5() >>> # build model and load imagenet pretrained weight >>> # model = shufflenet_v2_x1_5(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) paddle.Size([1, 1000]) rrr)rrrs r*rrrr)c "t dd|d|S)a\ShuffleNetV2 with 2.0x output channels, as described in `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`ShuffleNetV2 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of ShuffleNetV2 with 2.0x output channels. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import shufflenet_v2_x2_0 >>> # build model >>> model = shufflenet_v2_x2_0() >>> # build model and load imagenet pretrained weight >>> # model = shufflenet_v2_x2_0(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) paddle.Size([1, 1000]) rrr)rrrs r*rr=rr)Unpack[_ShuffleNetSwishOptions]c $t ddd|d|S)aiShuffleNetV2 with swish activation function, as described in `"ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design" `_. Args: pretrained (bool, optional): Whether to load pre-trained weights. If True, returns a model pre-trained on ImageNet. Default: False. **kwargs (optional): Additional keyword arguments. For details, please refer to :ref:`ShuffleNetV2 `. Returns: :ref:`api_paddle_nn_Layer`. An instance of ShuffleNetV2 with swish activation function. Examples: .. code-block:: pycon >>> import paddle >>> from paddle.vision.models import shufflenet_v2_swish >>> # build model >>> model = shufflenet_v2_swish() >>> # build model and load imagenet pretrained weight >>> # model = shufflenet_v2_swish(pretrained=True) >>> x = paddle.rand([1, 3, 224, 224]) >>> out = model(x) >>> print(out.shape) paddle.Size([1, 1000]) rrr)rrr)rrrs r*rrbs7@          r))rr5r6r7)r=rr>r?r6r)Fr) rrrrrrrrr6r~)F)rrrrr6r~)rrrrr6r~), __future__rtypingrrrFr paddle.nnrrr paddle.utils.downloadr opsr r rtyping_extensionsrrrpaddle._typingrr_ActivationTyperr,__all__r4r'r<rNLayerrPrur~rrrrrrrrr(r)r*rsm#""""""******** ::::::::::;;;;;;$$$$$$&))))))))55555555$$$$$$ Oo.O&&&&&Y&&& &&&&&)&&& 3** @OOOO    "/'/'/'/'/'rx/'/'/'dB'B'B'B'B'B'B'B'J@@@@@28@@@J*"""""L"""""L"""""L"""""L"""""L"""""L&&&&&&&r)