o ={cy@s|dZddlmmZddlmZddlmZddl m Z ddl m Z ddl mZddl mZdd lmZdd lmZd Zed Zed ZedZedZe Z             d5ddZedd       d6ddZedd       d6d d!Z d7d$d%Zd8d&d'Zd8d(d)Z d8d*d+Z!ed,d8d-d.Z"ed/d9d1d2Z#ej$j%d3ej&ej'd4e"_ej#je#_dS):aGNASNet-A models for Keras. NASNet refers to Neural Architecture Search Network, a family of models that were designed automatically by learning the model architectures directly on the dataset of interest. Here we consider NASNet-A, the highest performance model that was found for the CIFAR-10 dataset, and then extended to ImageNet 2012 dataset, obtaining state of the art performance on CIFAR-10 and ImageNet 2012. Only the NASNet-A models, and their respective weights, which are suited for ImageNet 2012 are provided. The below table describes the performance on ImageNet 2012: --------------------------------------------------------------------------- Architecture | Top-1 Acc | Top-5 Acc | Multiply-Adds | Params (M) ---------------------|-----------|-----------|----------------|------------ NASNet-A (4 @ 1056) | 74.0 % | 91.6 % | 564 M | 5.3 NASNet-A (6 @ 4032) | 82.7 % | 96.2 % | 23.8 B | 88.9 Reference: - [Learning Transferable Architectures for Scalable Image Recognition]( https://arxiv.org/abs/1707.07012) (CVPR 2018) N)backend)imagenet_utils)training)VersionAwareLayers) data_utils) layer_utils) tf_logging) keras_exportzDhttps://storage.googleapis.com/tensorflow/keras-applications/nasnet/zNASNet-mobile.h5zNASNet-mobile-no-top.h5zNASNet-large.h5zNASNet-large-no-top.h5`Timagenetsoftmaxc  Cs|dvstjj|std|dkr|r| dkrtdt|tr4d|vr4|dkr4tdt|d| dur:d } tj || d t ||d }t d krZt d t d d} nd} |durgtj|d}nt |sttj||d}n|}|d|ddkrtd|t dkrdnd}|d}tj|ddddddd|}tj|ddd d!|}d}t||||dd"d#\}}t||||d$d#\}}t|D]}t|||d%|d#\}}qt||||d&|d#\}}|s|n|}t|D]}t||||d%||dd#\}}qt||||dd&d|d#\}}|s |n|}t|D]}t||||dd%d||dd#\}}q&td'|}|rat|}t| |tj| | d(d)|}n| d*krmt|}n | d+krxt|}|durt|}n|}tj||d,d-}|dkr| d.kr|rt j!d/t"d0d1d2}n t j!d3t#d0d4d2}|$|n/| d kr|rt j!d5t%d0d6d2}n t j!d7t&d0d8d2}|$|ntd9|dur|$|| rt | |S):aInstantiates a NASNet model. Reference: - [Learning Transferable Architectures for Scalable Image Recognition]( https://arxiv.org/abs/1707.07012) (CVPR 2018) For image classification use cases, see [this page for detailed examples]( https://keras.io/api/applications/#usage-examples-for-image-classification-models). For transfer learning use cases, make sure to read the [guide to transfer learning & fine-tuning]( https://keras.io/guides/transfer_learning/). Note: each Keras Application expects a specific kind of input preprocessing. For NasNet, call `tf.keras.applications.nasnet.preprocess_input` on your inputs before passing them to the model. `nasnet.preprocess_input` will scale input pixels between -1 and 1. Args: input_shape: Optional shape tuple, the input shape is by default `(331, 331, 3)` for NASNetLarge and `(224, 224, 3)` for NASNetMobile. It should have exactly 3 input channels, and width and height should be no smaller than 32. E.g. `(224, 224, 3)` would be one valid value. penultimate_filters: Number of filters in the penultimate layer. NASNet models use the notation `NASNet (N @ P)`, where: - N is the number of blocks - P is the number of penultimate filters num_blocks: Number of repeated blocks of the NASNet model. NASNet models use the notation `NASNet (N @ P)`, where: - N is the number of blocks - P is the number of penultimate filters stem_block_filters: Number of filters in the initial stem block skip_reduction: Whether to skip the reduction step at the tail end of the network. filter_multiplier: Controls the width of the network. - If `filter_multiplier` < 1.0, proportionally decreases the number of filters in each layer. - If `filter_multiplier` > 1.0, proportionally increases the number of filters in each layer. - If `filter_multiplier` = 1, default number of filters from the paper are used at each layer. include_top: Whether to include the fully-connected layer at the top of the network. weights: `None` (random initialization) or `imagenet` (ImageNet weights) input_tensor: Optional Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. pooling: Optional pooling mode for feature extraction when `include_top` is `False`. - `None` means that the output of the model will be the 4D tensor output of the last convolutional block. - `avg` means that global average pooling will be applied to the output of the last convolutional block, and thus the output of the model will be a 2D tensor. - `max` means that global max pooling will be applied. classes: Optional number of classes to classify images into, only to be specified if `include_top` is True, and if no `weights` argument is specified. default_size: Specifies the default image size of the model classifier_activation: A `str` or callable. The activation function to use on the "top" layer. Ignored unless `include_top=True`. Set `classifier_activation=None` to return the logits of the "top" layer. When loading pretrained weights, `classifier_activation` can only be `None` or `"softmax"`. Returns: A `keras.Model` instance. >rNzThe `weights` argument should be either `None` (random initialization), `imagenet` (pre-training on ImageNet), or the path to the weights file to be loaded.rrzWIf using `weights` as `"imagenet"` with `include_top` as true, `classes` should be 1000NzWhen specifying the input shape of a NASNet and loading `ImageNet` weights, the input_shape argument must be static (no None entries). Got: `input_shape=z`.K ) default_sizeZmin_size data_formatZrequire_flattenweightsZ channels_lastaThe NASNet family of models is only available for the input data format "channels_last" (width, height, channels). However your settings specify the default data format "channels_first" (channels, width, height). You should set `image_data_format="channels_last"` in your Keras config located at ~/.keras/keras.json. The model being returned right now will expect inputs to follow the "channels_last" data format.channels_first)shape)Ztensorrr rzwFor NASNet-A models, the `penultimate_filters` must be a multiple of 24 * (`filter_multiplier` ** 2). Current value: %drr r validFZ stem_conv1 he_normal)stridespaddinguse_biasnamekernel_initializerZڊ?MbP?Zstem_bn1axisZmomentumepsilonr#Zstem_1block_idZstem_2z%dz reduce_%dreluZ predictions)Z activationr#ZavgmaxNASNetr#znasnet_mobile.h5ZmodelsZ 020fb642bf7360b370c678b08e0adf61)Z cache_subdirZ file_hashznasnet_mobile_no_top.h5Z 1ed92395b5b598bdda52abe5c0dbfd63znasnet_large.h5Z 11577c9a518f0070763c2b964a382f17znasnet_large_no_top.h5Z d81d89dc07e6e56530c4e77faddd61b5zDImageNet weights can only be loaded with NASNetLarge or NASNetMobile)'tfioZgfileexists ValueError isinstancetuplestrrZobtain_input_shaperimage_data_formatloggingZwarningZset_image_data_formatlayersZInputZis_keras_tensorConv2DBatchNormalization_reduction_a_cellrange_normal_a_cell ActivationZGlobalAveragePooling2DZvalidate_activationZDenseZGlobalMaxPooling2DrZget_source_inputsrZModelrZget_fileNASNET_MOBILE_WEIGHT_PATH NASNET_MOBILE_WEIGHT_PATH_NO_TOPZ load_weightsNASNET_LARGE_WEIGHT_PATHNASNET_LARGE_WEIGHT_PATH_NO_TOP) input_shapepenultimate_filters num_blocksstem_block_filtersskip_reductionfilter_multiplier include_topr input_tensorpoolingclassesrclassifier_activationZold_data_formatZ img_input channel_dimfiltersxpiZp0ZinputsZmodelZ weights_pathrU9lib/python3.10/site-packages/keras/applications/nasnet.pyr.@s6[                              r.z&keras.applications.nasnet.NASNetMobilezkeras.applications.NASNetMobilecC"t|ddddd|||||d|d S)a Instantiates a Mobile NASNet model in ImageNet mode. Reference: - [Learning Transferable Architectures for Scalable Image Recognition]( https://arxiv.org/abs/1707.07012) (CVPR 2018) Optionally loads weights pre-trained on ImageNet. Note that the data format convention used by the model is the one specified in your Keras config at `~/.keras/keras.json`. Note: each Keras Application expects a specific kind of input preprocessing. For NASNet, call `tf.keras.applications.nasnet.preprocess_input` on your inputs before passing them to the model. Args: input_shape: Optional shape tuple, only to be specified if `include_top` is False (otherwise the input shape has to be `(224, 224, 3)` for NASNetMobile It should have exactly 3 inputs channels, and width and height should be no smaller than 32. E.g. `(224, 224, 3)` would be one valid value. include_top: Whether to include the fully-connected layer at the top of the network. weights: `None` (random initialization) or `imagenet` (ImageNet weights). For loading `imagenet` weights, `input_shape` should be (224, 224, 3) input_tensor: Optional Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. pooling: Optional pooling mode for feature extraction when `include_top` is `False`. - `None` means that the output of the model will be the 4D tensor output of the last convolutional layer. - `avg` means that global average pooling will be applied to the output of the last convolutional layer, and thus the output of the model will be a 2D tensor. - `max` means that global max pooling will be applied. classes: Optional number of classes to classify images into, only to be specified if `include_top` is True, and if no `weights` argument is specified. classifier_activation: A `str` or callable. The activation function to use on the "top" layer. Ignored unless `include_top=True`. Set `classifier_activation=None` to return the logits of the "top" layer. When loading pretrained weights, `classifier_activation` can only be `None` or `"softmax"`. Returns: A Keras model instance. Raises: ValueError: In case of invalid argument for `weights`, or invalid input shape. RuntimeError: If attempting to run this model with a backend that does not support separable convolutions. i rFr r0 rFrGrHrIrJrKrrLrMrNrrOr.rErKrrLrMrNrOrUrUrV NASNetMobiledGr\z%keras.applications.nasnet.NASNetLargezkeras.applications.NASNetLargecCrW)a Instantiates a NASNet model in ImageNet mode. Reference: - [Learning Transferable Architectures for Scalable Image Recognition]( https://arxiv.org/abs/1707.07012) (CVPR 2018) Optionally loads weights pre-trained on ImageNet. Note that the data format convention used by the model is the one specified in your Keras config at `~/.keras/keras.json`. Note: each Keras Application expects a specific kind of input preprocessing. For NASNet, call `tf.keras.applications.nasnet.preprocess_input` on your inputs before passing them to the model. Args: input_shape: Optional shape tuple, only to be specified if `include_top` is False (otherwise the input shape has to be `(331, 331, 3)` for NASNetLarge. It should have exactly 3 inputs channels, and width and height should be no smaller than 32. E.g. `(224, 224, 3)` would be one valid value. include_top: Whether to include the fully-connected layer at the top of the network. weights: `None` (random initialization) or `imagenet` (ImageNet weights). For loading `imagenet` weights, `input_shape` should be (331, 331, 3) input_tensor: Optional Keras tensor (i.e. output of `layers.Input()`) to use as image input for the model. pooling: Optional pooling mode for feature extraction when `include_top` is `False`. - `None` means that the output of the model will be the 4D tensor output of the last convolutional layer. - `avg` means that global average pooling will be applied to the output of the last convolutional layer, and thus the output of the model will be a 2D tensor. - `max` means that global max pooling will be applied. classes: Optional number of classes to classify images into, only to be specified if `include_top` is True, and if no `weights` argument is specified. classifier_activation: A `str` or callable. The activation function to use on the "top" layer. Ignored unless `include_top=True`. Set `classifier_activation=None` to return the logits of the "top" layer. When loading pretrained weights, `classifier_activation` can only be `None` or `"softmax"`. Returns: A Keras model instance. Raises: ValueError: in case of invalid argument for `weights`, or invalid input shape. RuntimeError: If attempting to run this model with a backend that does not support separable convolutions. r r r Tr rrYrZr[rUrUrV NASNetLarger]r^rrrc Cstdkrdnd}td|mtd|}|dkr1tjt||d|d|}d }nd }tj|||d ||d d d|}tj |ddd|d|}td|}tj||d|d d d d|}tj |ddd|d|}Wd|S1swY|S)aFAdds 2 blocks of [relu-separable conv-batchnorm]. Args: ip: Input tensor filters: Number of output filters per layer kernel_size: Kernel size of separable convolutions strides: Strided convolution for downsampling block_id: String block_id Returns: A Keras tensor rrrZseparable_conv_block_r,rZseparable_conv_1_pad_r!r#rsameZseparable_conv_1_Fr)r r#r!r"r$r%r&Zseparable_conv_1_bn_r'Zseparable_conv_2_)r#r!r"r$Zseparable_conv_2_bn_N) rr8 name_scoper:r@ ZeroPadding2Dr correct_padZSeparableConv2Dr<)iprQ kernel_sizer r+rPrRZconv_padrUrUrV_separable_conv_blocksl   ((rgc Csztdkrdnd}tdkrdnd}t|}|dur"t|}td |dur0|}n||||krtd|{tjd d |d |}tjd d dd|d|}tj|dd ddd|dd|}tjdd|} tj dd| } tjd d dd|d| } tj|dd ddd|dd| } tj || g|d}tj |ddd|d |}Wdn1swYn^|||krtd!|.td |}tj|d d dd"|ddd#|}tj |ddd|d |}Wdn)1swYWd|SWd|SWd|SWd|SWd|S1s6wY|S)$aAdjusts the input `previous path` to match the shape of the `input`. Used in situations where the output number of filters needs to be changed. Args: p: Input tensor which needs to be modified ip: Input tensor whose shape needs to be matched filters: Number of output filters to be matched block_id: String block_id Returns: Adjusted Keras tensor rrrr NZ adjust_blockZadjust_reduction_block_r,Zadjust_relu_1_r/r_rrZadjust_avg_pool_1_r r!r#raFZadjust_conv_1_r)r!r"r#r$)rrrj)r!)rrrk)ZcroppingZadjust_avg_pool_2_Zadjust_conv_2_)r(r%r&Z adjust_bn_r'Zadjust_projection_block_Zadjust_conv_projection_r r!r#r"r$) rr8Z int_shaperbr:r@AveragePooling2Dr;rcZ Cropping2D concatenater<) rSrerQr+rPZimg_dimZip_shapeZp_shapeZp1Zp2rUrUrV _adjust_blockPs    ,  C C C C CCroc Cstdkrdnd}td|Ot||||}td|}tj|dddd|d d d |}tj|d d d|d|}td)t||dd|d}t||d|d}tj ||gd|d}Wdn1srwYtd*t||dd|d} t||dd|d} tj | | gd|d} Wdn1swYtd"tj dddd|d |} tj | |gd!|d} Wdn1swYtd"0tj dddd#|d |} tj dddd$|d |}tj | |gd%|d}Wdn 1swYtd&t||d'|d}tj ||gd(|d}Wdn 1sBwYtj ||| | ||g|d)|d*}Wd||fS1sfwY||fS)+zAdds a Normal cell for NASNet-A (Fig. 4 in the paper). Args: ip: Input tensor `x` p: Input tensor `p` filters: Number of output filters block_id: String block_id Returns: A Keras tensor rrrZnormal_A_block_r,r_raZnormal_conv_1_Frrlr%r&Z normal_bn_1_r'block_1rrZ normal_left1_)rfr+Znormal_right1_r*Z normal_add_1_r/Nblock_2Z normal_left2_rZnormal_right2_Z normal_add_2_block_3Z normal_left3_riZ normal_add_3_block_4Z normal_left4_Znormal_right4_Z normal_add_4_block_5Z normal_left5_Z normal_add_5_Znormal_concat_r(r#) rr8rbror:r@r;r<rgaddrmrn)rerSrQr+rPhx1_1x1_2x1x2_1x2_2x2x3Zx4_1Zx4_2x4x5rRrUrUrVr?s          LLr?c Cstdkrdnd}td|mt||||}td|}tj|dddd|d d d |}tj|d d d|d|}tjt |dd|d|}td,t ||ddd|d}t ||ddd|d}tj ||gd|d} Wdn1swYtd.tj dddd |d!|} t ||ddd"|d} tj | | gd#|d} Wdn1swYtd$.tjdddd%|d!|} t ||ddd&|d}tj | |gd'|d}Wdn1swYtd(tjdddd)|d!| }t | |g}Wdn 1s&wYtd*-t | |dd)|d+}tj dddd,|d!|}tj ||gd-|d}Wdn 1sawYtj| |||g|d.|d/}||fWdS1swYdS)0zAdds a Reduction cell for NASNet-A (Fig. 4 in the paper). Args: ip: Input tensor `x` p: Input tensor `p` filters: Number of output filters block_id: String block_id Returns: A Keras tensor rrrZreduction_A_block_r,r_raZreduction_conv_1_Frrlr%r&Zreduction_bn_1_r'rZreduction_pad_1_r`rprqrZreduction_left1_)r r+)rZreduction_right1_Zreduction_add_1_r/NrsrrZreduction_left2_riZreduction_right2_Zreduction_add_2_rtZreduction_left3_Zreduction_right3_Zreduction_add3_ruZreduction_left4_rvr*Zreduction_right5_Zreduction_add4_Zreduction_concat_rw)rr8rbror:r@r;r<rcrrdrgrxZ MaxPooling2Drmrn)rerSrQr+rPryZh3rzr{r|r}r~rZx3_1Zx3_2rrZx5_1Zx5_2rrRrUrUrVr= s         &r=z*keras.applications.nasnet.preprocess_inputcCstj||ddS)Nr1)rmode)rpreprocess_input)rRrrUrUrVr}srz,keras.applications.nasnet.decode_predictionsrrcCstj||dS)N)top)rdecode_predictions)ZpredsrrUrUrVrsr)rreterror) Nr r r Tr TrNNrNr)NTrNNrr)rr_N)N)rr)(__doc__Ztensorflow.compat.v2compatZv2r1ZkerasrZkeras.applicationsrZ keras.enginerZ keras.layersrZ keras.utilsrrZtensorflow.python.platformrr9Z tensorflow.python.util.tf_exportr ZBASE_WEIGHTS_PATHrArBrCrDr:r.r\r^rgror?r=rrZPREPROCESS_INPUT_DOCformatZPREPROCESS_INPUT_RET_DOC_TFZPREPROCESS_INPUT_ERROR_DOCrUrUrUrVs         & U V < \ ]t