6bi@SrSSKJs Jr SSKJr SSKJr SSK J r SSK J r SSK Jr SSK Jr SS KJr SS KJr S r\S -r\S -r\S-r\S-r\ "5rS!Sjr\"SS5S"Sj5r\"SS5S"Sj5rS#SjrS$SjrS$Sjr S$Sjr!\"S5S$Sj5r"\"S5S%Sj5r#\RHRKS\RL\RNS 9\"l\RFR\#lg)&aJNASNet-A models for TF-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.h5c 2 US;d9[RRRU5(d [ S5eUS:XaU(aU S:wa [ S5e[ U[ 5(a&SU;a US:Xa[ S[U5-S-5eU cS n [R"UU S [R"5UUS 9n[R"5S :wa/[R"S 5 [R"S 5 Sn OSn Uc[R!US9nO1[R""U5(d[R!XS9nOUnUSUS---S:wa[ SU-5e[R"5S:XaSOSnUS-n[R%USSSSSSS9"U5n[R'USSS S!9"U5nSn[)UUUUS--S"S#9unn[)UUUU-S$S#9unn[+U5Hn[-UUUS%U-S#9unnM [)UUUU-S&U-S#9unnU(dUOUn[+U5Hn[-UUUU-S%UU-S--S#9unnM [)UUUUS--S&SU--S#9unnU(dUOUn[+U5H$n[-UUUUS--S%SU-U-S--S#9unnM& [R/S'5"U5nU(aK[R15"U5n[R2"X5 [R5XS(S)9"U5nOAU S*:Xa[R15"U5nO U S+:Xa[R75"U5nUb[8R:"U5nOUn[<R>"UUS,S-9nUS:XaU S.:XaPU(a[@RB"S/[DS0S1S29nO[@RB"S3[FS0S4S29nURIU5 OuU S :XaPU(a[@RB"S5[JS0S6S29nO[@RB"S7[LS0S8S29nURIU5 O[ S95eUbURIU5 U (a[R"U 5 U$):a Instantiates 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 TF-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 TF-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. >NimagenetzThe `weights` argument should be either `None` (random initialization), `imagenet` (pre-training on ImageNet), or the path to the weights file to be loaded.r zWIf 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_sizemin_size data_formatrequire_flattenweights 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 TF-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)tensorrrzwFor NASNet-A models, the `penultimate_filters` must be a multiple of 24 * (`filter_multiplier` ** 2). Current value: %drrrvalidF stem_conv1 he_normal)stridespaddinguse_biasnamekernel_initializerZڊ?MbP?stem_bn1axismomentumepsilonr&stem_1block_idstem_2z%dz reduce_%drelu predictions) activationr&avgmaxNASNetr&znasnet_mobile.h5models 020fb642bf7360b370c678b08e0adf61) cache_subdir file_hashznasnet_mobile_no_top.h5 1ed92395b5b598bdda52abe5c0dbfd63znasnet_large.h5 11577c9a518f0070763c2b964a382f17znasnet_large_no_top.h5 d81d89dc07e6e56530c4e77faddd61b5zDImageNet weights can only be loaded with NASNetLarge or NASNetMobile)'tfiogfileexists ValueError isinstancetuplestrrobtain_input_shaperimage_data_formatloggingwarningset_image_data_formatlayersInputis_keras_tensorConv2DBatchNormalization_reduction_a_cellrange_normal_a_cell ActivationGlobalAveragePooling2Dvalidate_activationDenseGlobalMaxPooling2Drget_source_inputsrModelrget_fileNASNET_MOBILE_WEIGHT_PATH NASNET_MOBILE_WEIGHT_PATH_NO_TOP load_weightsNASNET_LARGE_WEIGHT_PATHNASNET_LARGE_WEIGHT_PATH_NO_TOP) input_shapepenultimate_filters num_blocksstem_block_filtersskip_reductionfilter_multiplier include_topr input_tensorpoolingclassesrclassifier_activationold_data_format img_input channel_dimfiltersxpip0inputsmodel weights_paths Z/home/james-whalen/.local/lib/python3.13/site-packages/tf_keras/src/applications/nasnet.pyr8r8@s#x ) )RUU[[-?-?-H-H <  *D 0  ;&& K  z ! 47:+6F GJN N   !33!--/# K  "o5 9  %%o6*LL{L3 &&|44 L LI$Ib$5q$89:a? D! "  0026FF!BK!R'G  &   A !! 64j "  A A  1g+Q./( DAq  1g**X DAq: aGdajA1  1g))K:4N EAr!aA :  ' 'Z!^a/0  1   #Q&&J/  EAr!aA :  '* *Q^a/!34  1 &!!$A  ) ) +A .**+@J LL M    e --/2A  ))+A.A..|< NN618 4E* 3 )22&-!)@    *22-4!)@     | , S )22%,!)@    *22,3!)@     | ,#    7#%%o6 Lz&keras.applications.nasnet.NASNetMobilezkeras.applications.NASNetMobilec,[USSSSSUUUUUSUS9 $)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 TF-Keras config at `~/.keras/keras.json`. Note: each TF-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 TF-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 TF-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 rFrr: rerfrgrhrirjrrkrlrmrrnr8rdrjrrkrlrmrns rz NASNetMobileres9P  !3 r{z%keras.applications.nasnet.NASNetLargezkeras.applications.NASNetLargec,[USSSSSUUUUUSUS9 $)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 TF-Keras config at `~/.keras/keras.json`. Note: each TF-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 TF-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 TF-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. `Trrr~rrs rz NASNetLargers9P  !3 r{c [R"5S:XaSOSn[R"SU35 [R S5"U5nUS:Xa4[R [ R"Xb5SU3S9"U5nS nOS n[RUUUS U3US S S9"U5n[RUSSSU3S9"U5n[R S5"U5n[RUUSU3S S S S9"U5n[RUSSSU3S9"U5nSSS5 U$!,(df  W$=f)a%Adds 2 blocks of [relu-separable conv-batchnorm]. Args: ip: Input tensor filters: Number of output filters per layer kernel_size: Kernel size of separable convolutions strides: Strided convolution for downsampling block_id: String block_id Returns: A TF-Keras tensor rrrseparable_conv_block_r3rseparable_conv_1_pad_r$r&r sameseparable_conv_1_Fr")r#r&r$r%r'r(r)separable_conv_1_bn_r+separable_conv_2_)r&r$r%r'separable_conv_2_bn_N) rrK name_scoperOrW ZeroPadding2Dr correct_padSeparableConv2DrS)iprr kernel_sizer#r1rqrsconv_pads rz_separable_conv_blockrs0026FF!BK   3H:> ?   f %b ) f $$&221B,XJ7%AHH  " "  $XJ/* #     % %'z2 &     f %a (  " "  $XJ/* #     % %'z2 &  E @P HQ @ ?P Hs C4D55 Ec [R"5S:XaSOSn[R"5S:XaSOSn[R"U5nUb[R"U5n[R"S5 UcUnGOWUXe:wGa,[R"SU35 [R S S U3S 9"U5n[R S S SSU3S9"U5n[RUS-S SSSU3SS9"U5n[RSS9"U5n [RSS9"U 5n [R S S SSU3S9"U 5n [RUS-S SSSU3SS9"U 5n [RX/US9n[RUSSSU3S 9"U5nSSS5 OXtU:wa~[R"S!U35 [R S 5"U5n[RUS S SS"U3SSS#9"U5n[RUSSSU3S 9"U5nSSS5 SSS5 U$!,(df  N=f!,(df  N)=f!,(df  U$=f)$auAdjusts 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 TF-Keras tensor rrrrN adjust_blockadjust_reduction_block_r3adjust_relu_1_r9rrrr adjust_avg_pool_1_r#r$r&rFadjust_conv_1_r")r$r%r&r')rrr)r$)rrr)croppingadjust_avg_pool_2_adjust_conv_2_)r,r(r) adjust_bn_r+adjust_projection_block_adjust_conv_projection_r#r$r&r%r') rrK int_shaperrOrWAveragePooling2DrRr Cropping2D concatenaterS) rtrrrr1rqimg_dimip_shapep_shapep1p2s rz _adjust_blockrSs0026FF!BK,,.2BBaG  $H}##A&   N + 9A W !2 2##&=hZ$HI%%f^H:3N%O,,"#-hZ8 -   ]]qL"")(4'2 #))2B)CAF&&0@&A"E,,"#-hZ8 -   ]]qL"")(4'2 #&&xk&B--$# %hZ0 .  KJIX !W ,##&>xj$IJ%%f-a0MM""28*="'2"--$# %hZ0 .  Ke ,F H}JIZKJe , +F Hs>-I2-D I7*I2!AI!>I2 I I2! I/ +I22 Jc [R"5S:XaSOSn[R"SU35 [XX#5n[R S5"U5n[R USSSSU3S S S 9"U5n[RUS S SU3S9"U5n[R"S5 [UUSSU3S9n[XSU3S9n[RXg/SU3S9nSSS5 [R"S5 [XSSU3S9n [XSSU3S9n [RX/SU3S9n SSS5 [R"S5 [RSSSSU3S 9"U5n [RX/S!U3S9n SSS5 [R"S"5 [RSSSS#U3S 9"U5n [RSSSS$U3S 9"U5n[RX/S%U3S9nSSS5 [R"S&5 [XRS'U3S9n[RUU/S(U3S9nSSS5 [RUWW W WW/US)U3S*9nSSS5 UU4$!,(df  GN=f!,(df  GN[=f!,(df  GN=f!,(df  N=f!,(df  N=f!,(df  WU4$=f)+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 TF-Keras tensor rrrnormal_A_block_r3rrnormal_conv_1_Fr"rr(r) normal_bn_1_r+block_1r normal_left1_)rr1normal_right1_r0 normal_add_1_r9Nblock_2 normal_left2_rnormal_right2_ normal_add_2_block_3 normal_left3_r normal_add_3_block_4 normal_left4_normal_right4_ normal_add_4_block_5 normal_left5_ normal_add_5_normal_concat_r,r&) rrKrrrOrWrRrSraddrr)rrtrrr1rqhx1_1x1_2x1x2_1x2_2x2x3x4_1x4_2x4x5rss rzrVrVsZ0026FF!BK   ohZ8 9 ! 3   f %b ) MM  !(,*      % %z* &     *("( 3 D )~hZ%@DTLxj/IJB+   *(F}XJ-GD)F~hZ-HDTLxj/IJB+   *(($XJ/ )  B RGM(*DEB+   ***$XJ/ +  D **%hZ0 +  D TLxj/IJB+   *&}XJ%?BRGM(*DEB +    BB #!(,  O :X b5Lq+ *+ *+ *+ *+ *C : 9X b5LsA?K65J+K 5J>K8J$K2AJ6 K''K&K J K J! K$ J3 .K6 K K K K K)c [R"5S:XaSOSn[R"SU35 [XX#5n[R S5"U5n[R USSSSU3S S S 9"U5n[RUS S SU3S9"U5n[R[R"US5SU3S9"U5n[R"S5 [UUSSSU3S9n[UUSSSU3S9n[RXx/SU3S9n SSS5 [R"S5 [RSSSS U3S!9"U5n [UUSSS"U3S9n [RX/S#U3S9n SSS5 [R"S$5 [RSSSS%U3S!9"U5n [UUSSS&U3S9n[RX/S'U3S9nSSS5 [R"S(5 [RSSSS)U3S!9"W 5n[RW U/5nSSS5 [R"S*5 [W USS)U3S+9n[RSSSS,U3S!9"U5n[RUU/S-U3S9nSSS5 [RW WWW/US.U3S/9nUU4sSSS5 $!,(df  GN=f!,(df  GNk=f!,(df  GN=f!,(df  N=f!,(df  N~=f!,(df  g=f)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 TF-Keras tensor rrrreduction_A_block_r3rrreduction_conv_1_Fr"rr(r)reduction_bn_1_r+rreduction_pad_1_rrrrreduction_left1_)r#r1)rreduction_right1_reduction_add_1_r9Nrrr reduction_left2_rreduction_right2_reduction_add_2_rreduction_left3_reduction_right3_reduction_add3_rreduction_left4_rr0reduction_right5_reduction_add4_reduction_concat_r)rrKrrrOrWrRrSrrrrr MaxPooling2Drr)rrtrrr1rqrh3rrrrrrx3_1x3_2rrx5_1x5_2rrss rzrTrT s0026FF!BK   0 ; < ! 3   f %b ) MM  $XJ/*      % %"8*- &    ! !"..q!4#H:."       *(+H:6 D),XJ7 DTL1A(/LMB+"   *&&'z2 '  D ),XJ7 DTL1A(/LMB+    ***'z2 +  D ),XJ7 DTL /KLB+    *(('z2 )  B RH%B+   *(GV0@ .KD&&( 3 '  D T4L /KLB+    R $XJ/   "uG = <0+ *"+ * + * + *+ *e = rs 0"! 4(2)*=9K.0BB#47P#P ,/@@"36N"N     #bJ ,.O   #SSl+-M   #SSn?C9 xY xZzqh:;< <==>=*>>EE 22  3 3F ,>>FFr{