o V hA@sddlZddlmZddlmZddlmZmZmZm Z ddl Z ddl m Z ddl m mZddlmmZddl mZddlmZddlmZd d lmZmZmZd d lmZd d lmZm Z gd Z!Gddde j"Z#Gddde j$Z%Gddde j&Z'Gddde j"Z(de j"dede)ddfddZ*de+de e+e+e+e+fde+deede)dede(fd d!Z,d"ed#d$d%Z-Gd&d'd'eZ.Gd(d)d)eZ/Gd*d+d+eZ0Gd,d-d-eZ1ee d.e.j2fd/dd0d1dee.de)dede(fd2d3Z3ee d.e/j2fd/dd0d1dee/de)dede(fd4d5Z4ee d.e0j2fd/dd0d1dee0de)dede(fd6d7Z5ee d.e1j2fd/dd0d1dee1de)dede(fd8d9Z6dS):N) OrderedDict)partial)AnyListOptionalTuple)Tensor)ImageClassification)_log_api_usage_once)register_modelWeights WeightsEnum)_IMAGENET_CATEGORIES)_ovewrite_named_paramhandle_legacy_interface) DenseNetDenseNet121_WeightsDenseNet161_WeightsDenseNet169_WeightsDenseNet201_Weights densenet121 densenet161 densenet169 densenet201c seZdZ ddedededededdf fd d Zd eedefd d Z deedefddZ e j j deedefddZe j jdeedefddZe j jdedefddZdedefddZZS) _DenseLayerFnum_input_features growth_ratebn_size drop_ratememory_efficientreturnNcstt||_tjdd|_tj|||dddd|_t|||_ tjdd|_ tj|||ddddd|_ t ||_ ||_dS)NTinplacer F kernel_sizestridebiasr&r'paddingr()super__init__nn BatchNorm2dnorm1ReLUrelu1Conv2dconv1norm2relu2conv2floatr r!)selfrrrr r! __class__O/var/www/vscode/kcb/lib/python3.10/site-packages/torchvision/models/densenet.pyr- s    z_DenseLayer.__init__inputscCs&t|d}||||}|SNr )torchcatr4r2r0)r9r>concated_featuresbottleneck_outputr<r<r= bn_function/s z_DenseLayer.bn_functioninputcCs|D]}|jr dSqdS)NTF) requires_grad)r9rEtensorr<r<r=any_requires_grad5s z_DenseLayer.any_requires_gradcs$fdd}tj|g|RddiS)Ncs |SN)rD)r>r9r<r=closure=s z7_DenseLayer.call_checkpoint_bottleneck..closure use_reentrantF)cp checkpoint)r9rErKr<rJr=call_checkpoint_bottleneck;s z&_DenseLayer.call_checkpoint_bottleneckcCdSrIr<r9rEr<r<r=forwardBz_DenseLayer.forwardcCrPrIr<rQr<r<r=rRFrScCst|tr |g}n|}|jr"||r"tjrtd||}n| |}| | | |}|j dkrAtj||j |jd}|S)Nz%Memory Efficient not supported in JITr)ptraining) isinstancerr!rHr@jit is_scripting ExceptionrOrDr7r6r5r FdropoutrU)r9rE prev_featuresrC new_featuresr<r<r=rRLs     F)__name__ __module__ __qualname__intr8boolr-rrrDrHr@rWunusedrO_overload_methodrR __classcell__r<r<r:r=rs0rcsTeZdZdZ ddedededededed d ffd d Zd ed efddZ Z S) _DenseBlockr F num_layersrrrr r!r"Nc sJtt|D]}t|||||||d}|d|d|q dS)N)rrr r!z denselayer%dr )r,r-ranger add_module) r9rhrrrr r!ilayerr:r<r=r-cs  z_DenseBlock.__init__ init_featurescCs6|g}|D] \}}||}||qt|dSr?)itemsappendr@rA)r9rmfeaturesnamerlr]r<r<r=rRws   z_DenseBlock.forwardr^) r_r`ra_versionrbr8rcr-rrRrfr<r<r:r=rg`s& rgcs*eZdZdededdffdd ZZS) _Transitionrnum_output_featuresr"NcsNtt||_tjdd|_tj||dddd|_tj ddd|_ dS)NTr#r Fr%r )r&r') r,r-r.r/normr1relur3conv AvgPool2dpool)r9rrtr:r<r=r-s  z_Transition.__init__)r_r`rarbr-rfr<r<r:r=rss"rscspeZdZdZ       dd ed eeeeefd ed ed edededdffdd Zde de fddZ Z S)raKDensenet-BC model class, based on `"Densely Connected Convolutional Networks" `_. Args: growth_rate (int) - how many filters to add each layer (`k` in paper) block_config (list of 4 ints) - how many layers in each pooling block num_init_features (int) - the number of filters to learn in the first convolution layer bn_size (int) - multiplicative factor for number of bottle neck layers (i.e. bn_size * k features in the bottleneck layer) drop_rate (float) - dropout rate after each dense layer num_classes (int) - number of classification classes memory_efficient (bool) - If True, uses checkpointing. Much more memory efficient, but slower. Default: *False*. See `"paper" `_.  @rFr block_confignum_init_featuresrr num_classesr!r"Nc stt|ttdtjd|dddddfdt|fdtjd d fd tj ddd d fg|_ |}t |D]>\} } t | |||||d} |j d| d | || |}| t|d krwt||dd} |j d| d | |d}q9|j dt|t|||_|D]5} t| tjrtj| jqt| tjrtj| jd tj| jdqt| tjrtj| jdqdS)Nconv0r)r Fr*norm0relu0Tr#pool0r )r&r'r+)rhrrrr r!z denseblock%d)rrtz transition%dnorm5r)r,r-r r. Sequentialrr3r/r1 MaxPool2drp enumeratergrjlenrsLinear classifiermodulesrVinitkaiming_normal_weight constant_r()r9rrrrr rr! num_featuresrkrhblocktransmr:r<r=r-sP        zDenseNet.__init__xcCs>||}tj|dd}t|d}t|d}||}|S)NTr#)r r r )rprZrvadaptive_avg_pool2dr@flattenr)r9rrpoutr<r<r=rRs    zDenseNet.forward)rzr{rrrrF) r_r`ra__doc__rbrr8rcr-rrRrfr<r<r:r=rs6 @ num_params_metrics_ops _file_sizeurl transformsrN r_r`rarrr _COMMON_META IMAGENET1K_V1DEFAULTr<r<r<r=r $ rc @r)rz`_. Args: weights (:class:`~torchvision.models.DenseNet121_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.DenseNet121_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.densenet.DenseNet`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.DenseNet121_Weights :members: rzr{r)rverifyrrrrr<r<r=r[ rcKr)a{Densenet-161 model from `Densely Connected Convolutional Networks `_. Args: weights (:class:`~torchvision.models.DenseNet161_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.DenseNet161_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.densenet.DenseNet`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.DenseNet161_Weights :members: 0)r|r}$r~`)rrrrr<r<r=rurrcKr)a{Densenet-169 model from `Densely Connected Convolutional Networks `_. Args: weights (:class:`~torchvision.models.DenseNet169_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.DenseNet169_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.densenet.DenseNet`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.DenseNet169_Weights :members: rz)r|r}rzrzr)rrrrr<r<r=rrrcKr)a{Densenet-201 model from `Densely Connected Convolutional Networks `_. Args: weights (:class:`~torchvision.models.DenseNet201_Weights`, optional): The pretrained weights to use. See :class:`~torchvision.models.DenseNet201_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.densenet.DenseNet`` base class. Please refer to the `source code `_ for more details about this class. .. autoclass:: torchvision.models.DenseNet201_Weights :members: rz)r|r}rrzr)rrrrr<r<r=rrr)7r collectionsr functoolsrtypingrrrrr@torch.nnr.torch.nn.functional functionalrZtorch.utils.checkpointutilsrNrMrtransforms._presetsr r _apir rr_metar_utilsrr__all__Moduler ModuleDictrgrrsrrcrrbrrrrrrrrrrrr<r<r<r=sp        A U ***.