6bi&SrSSKrSSKrSSKrSSKrSSKJs Jr SSK J r SSK J r SSKJr SSKJr SSKJr SSKJr SS KJr SS KJr SS KJr SS KJr SS KJr SSKJr SSKJr SSKJ r SSKJ!r" SSK#J$r$ SSK#J%r% SSK#J&r& SSK'J(r) SSK*J+r+ "SS\RX5r-Sr.Sr/Sr0Sr1Sr2Sr3Sr4S&S jr5S&S!jr6S"r7"S#S$\Rp5r9S%r:g)'zHA `Network` is way to compose layers: the topological form of a `Model`.N)backend) layout_map) base_layer)base_layer_utils)functional_utils) input_layer) input_spec)node)training)training_utils)serialization_lib) serialization) json_utils)network_serialization)utils) generic_utils) tf_inspect)tf_utils) tf_logging) doc_controlsc^\rSrSrSr\"\R"S\RR55r \ RRRS$U4Sjj5r\ RRRS5r\S5r\S5r\S5r\R*S 5r\S 5r\S 5rS r\S 5rS%U4SjjrS&U4SjjrSr\S5rSr\R@S'Sj5r!Sr"S(Sjr#S'Sjr$Sr%Sr&\'RPS5r)Sr*Sr+S&Sjr,Sr-Sr.Sr/S r0\S!5r1S)U4S"jjr2S#r3U=r4$)* Functional1a A `Functional` model is a `Model` defined as a directed graph of layers. Three types of `Model` exist: subclassed `Model`, `Functional` model, and `Sequential` (a special case of `Functional`). In general, more TF-Keras features are supported with `Functional` than with subclassed `Model`s, specifically: - Model cloning (`keras.models.clone`) - Serialization (`model.get_config()/from_config`, `model.to_json()` - Whole-model saving (`model.save()`) A `Functional` model can be instantiated by passing two arguments to `__init__`. The first argument is the `keras.Input` Tensors that represent the inputs to the model. The second argument specifies the output tensors that represent the outputs of this model. Both arguments can be a nested structure of tensors. Example: ``` inputs = {'x1': keras.Input(shape=(10,)), 'x2': keras.Input(shape=(1,))} t = keras.layers.Dense(1, activation='relu')(inputs['x1']) outputs = keras.layers.Add()([t, inputs['x2']) model = keras.Model(inputs, outputs) ``` A `Functional` model constructed using the Functional API can also include raw TensorFlow functions, with the exception of functions that create Variables or assign ops. Example: ```python inputs = keras.Input(shape=(10,)) x = keras.layers.Dense(1)(inputs) outputs = tf.nn.relu(x) model = keras.Model(inputs, outputs) ``` A new `Functional` model can also be created by using the intermediate tensors. This enables you to quickly extract sub-components of the model. Example: ```python inputs = keras.Input(shape=(None, None, 3)) processed = keras.layers.RandomCrop(width=32, height=32)(inputs) conv = keras.layers.Conv2D(filters=2, kernel_size=3)(processed) pooling = keras.layers.GlobalAveragePooling2D()(conv) feature = keras.layers.Dense(10)(pooling) full_model = keras.Model(inputs, feature) backbone = keras.Model(processed, conv) activations = keras.Model(conv, feature) ``` Note that the `backbone` and `activations` models are not created with `keras.Input` objects, but with the tensors that are originated from `keras.Input` objects. Under the hood, the layers and weights will be shared across these models, so that user can train the `full_model`, and use `backbone` or `activations` to do feature extraction. The inputs and outputs of the model can be nested structures of tensors as well, and the created models are standard `Functional` model that support all the existing API. Args: inputs: List of input tensors (must be created via `tf.keras.Input()` or originated from `tf.keras.Input()`). outputs: List of output tensors. name: String, optional. Name of the model. trainable: Boolean, optional. If the model's variables should be trainable. )_layer_call_argspecs_output_mask_cache_output_tensor_cache_output_shape_cachec >URSS5nU(ag[R"U05 [TU]X4S9 [ R RR5(aj[[ RRU5Vs/sHn[R"U5PM sn5(d[R"X5upURX5 gs snf)N skip_initFname trainable)poprvalidate_kwargssuper__init__tfcompatv1#executing_eagerly_outside_functionsallnestflattenris_input_keras_tensorclone_graph_nodes_init_graph_network) selfinputsoutputsr!r"kwargsrt __class__s X/home/james-whalen/.local/lib/python3.13/site-packages/tf_keras/src/engine/functional.pyr&Functional.__init__s JJ{E2  %%fb1 d8 99<< ; ; = = WW__V44%::1=4 #3"D"D#   1s C)c SUl[U[5(a1[[R R U55S:XaUSn[U[5(a1[[R R U55S:XaUSnXlX l[R R U5Ul [R R U5Ul [R RUR5(dSUl O[UR[[45(a)[SUR55(dSUl O][UR[5(a7[SURR!555(dSUl OSUl [R"R$R'5(dA[SUR55(a [(R*"UR5 UR-5 SUl[R R1SU5UlSUlSUR6lSUR6lSUl/Ul/Ul /Ul!/Ul"0Ul#0Ul$0Ul%URHKnURLunnnUR@ROU5 URDROXEU45 MM URH[nURLunnnUS:XdeUS:XdeUR>ROU5 URBROXEU45 M] [QURUR5upxpXpl)Xl*Xl+0Ul,URVH0n[ZR\"UR^5URXU'M2 URa5 /Ul1/Ul2/Ul3/Ul4UR>HnURbROURj5 URl(dM;URdROURj5 URhROURn5 URfROURp5 M URs5 URuUR5 [vRx"URz5 UR|b![~R"XR|5 gg) NTrc3`# UH$n[RRU5v M& g7fNr'r, is_nested.0r5s r7 1Functional._init_graph_network..s'H *=QBGG  a *=,.c3`# UH$n[RRU5v M& g7fr<r=r?s r7rArBs&? *FQBGG  a *FrCFc3D# UHn[US5(+v M g7f)_keras_historyN)hasattrr@tensors r7rArBs!DP&GF$4555Ls cUR$r<)shape)xs r70Functional._init_graph_network..sagg)A_is_graph_network isinstancelistlenr'r,r-_nested_inputs_nested_outputsr2r3r>_enable_dict_to_input_mappingtupleanydictvaluesr(r)r*rcreate_keras_history"_validate_graph_inputs_and_outputsbuilt map_structure_build_input_shape _compute_output_and_mask_jointly _call_specexpects_training_argexpects_mask_arg _autocast _input_layers_output_layers_input_coordinates_output_coordinatesrrrrFappend_map_graph_network_network_nodes_nodes_by_depth_self_tracked_trackablesrrgetfullargspeccall_set_output_names input_names_feed_input_names _feed_inputs_feed_input_shapesr!is_placeholder_batch_input_shapeinput_compute_tensor_usage_count_set_save_specrassert_no_legacy_layerslayers _layout_maplayout_map_lib_map_functional_model_variable) r1r2r3rLlayer node_index tensor_indexnodesnodes_by_depthr{_s r7r0Functional._init_graph_networksy "& fd # #BGGOOF,C(D(IAYF gt $ $RWW__W-E)F!)KajG$&ggoof- www/ ww  !4!45515D . ++dE] ; ;CH *.*=*=H E E 26D . ++T 2 23? *.*=*=*D*D*F? < < 26D .16D .yy||??AADHLL!55d6J6JK //1 "$''"7"7 v# 15-04,+/( "$#% #%$&!#% A        & &u -  $ $ + +U ,M NA    ? "?1$ $$    % %e ,  # # * *E|+L M,> KK, (v$-(.%$&!22E/9/H/H 0D % %e ,3  !#"$''E    # #EJJ /###&&--ejj9''..u/G/GH!!((5( ((* D//0((5    '  9 9&&  (rOcUR$)a&Retrieves the input tensor(s) of a layer. Only applicable if the layer has exactly one input, i.e. if it is connected to one incoming layer. Returns: Input tensor or list of input tensors. Raises: RuntimeError: If called in Eager mode. AttributeError: If no inbound nodes are found. )rTr1s r7rwFunctional.input3s"""rOcr[RR[RUR 5$)aRetrieves the input shape(s) of a layer. Only applicable if the layer has exactly one input, i.e. if it is connected to one incoming layer, or if all inputs have the same shape. Returns: Input shape, as an integer shape tuple (or list of shape tuples, one tuple per input tensor). Raises: AttributeError: if the layer has no defined input_shape. RuntimeError: if called in Eager mode. )r'r,r^r int_shaperwrs r7 input_shapeFunctional.input_shapeCs% ww$$W%6%6 CCrOc [US5(a UR$[UR[[ [ 45(a-[UR5[UR5:wag[UR[5(a`[URR55nUVs/sH/n[R"[URU5SUS9PM1 sn$URVs/sH@n[R"[U5SURRR S9PMB sn$s snfs snf)N_manual_input_specT)rKallow_last_axis_squeezer!)rGrrQrTrYrRrWrSr2sortedkeysr InputSpecshape_with_no_batch_sizerFrr!)r1namesr!rLs r7r Functional.input_specUs3 4- . .** * d))D$+> ? ?C   E  E  d))4 0 04..3356E"  "D $$243F3Ft3LM,0 "  "  %A $$215,0))//44 %  s ;6E AEcXlgr<)r)r1values r7r rvs"'rOcUR$)aIRetrieves the output tensor(s) of a layer. Only applicable if the layer has exactly one output, i.e. if it is connected to one incoming layer. Returns: Output tensor or list of output tensors. Raises: AttributeError: if the layer is connected to more than one incoming layers. RuntimeError: if called in Eager mode. )rUrs r7outputFunctional.outputzs###rOcr[RR[RUR 5$)acRetrieves the output shape(s) of a layer. Only applicable if the layer has one output, or if all outputs have the same shape. Returns: Output shape, as an integer shape tuple (or list of shape tuples, one tuple per output tensor). Raises: AttributeError: if the layer has no defined output shape. RuntimeError: if called in Eager mode. )r'r,r^rrrrs r7 output_shapeFunctional.output_shapes%ww$$W%6%6 DDrOcB/n[5n0nURH{nURnXR;aEURURS5nURSU3nUS-X4R'XR;aMEUR U5 UR U5 M} Xlg)zAssigns unique names to the Network's outputs. Output layers with multiple output tensors would otherwise lead to duplicate names in self.output_names. r:rN)setrfr!getaddri output_names)r1 uniquifiedr prefix_countrproposalexisting_counts r7rpFunctional._set_output_namess  u  ((EzzH*!-!1!1%**a!@#jj\>*:;+9A+= ZZ(*   X &   h ')'rOcSn[R"5n[UR5H+up4UR(a UUSU-'US- nXBSU-'M- U$![ a Nf=f)zBDictionary of layer dependencies to be included in the checkpoint.rzlayer_with_weights-%dr:zlayer-%d) collections OrderedDict enumerater{weights ValueError)r1weight_layer_index dependencies layer_indexrs r7_layer_checkpoint_dependencies)Functional._layer_checkpoint_dependenciess"..0 "+DKK"8 K == !/2DD'!+&6;k1 2##9$  sA A+*A+c `>URnUR[TU] "U40UD65 U$r<)rupdater%_trackable_children)r1 save_typer4rr6s r7rFunctional._trackable_childrens1:: EG7 LVLMrOc|>U(aURU5$URnX;aX1$[TU] U5$r<)rrr%_lookup_dependency)r1r!cached_dependencieslayer_dependenciesr6s r7rFunctional._lookup_dependencysD &**40 0"@@  %%+ +w)$//rOcURnUR5VVs0sHup4XC_M nnnUHnXe;dM URXVUS9 M gs snnf)z@Handles layer checkpoint dependencies that are added after init.)r! trackableN)ritems_handle_deferred_dependencies)r1r{layer_checkpoint_dependencieskv layer_to_namers r7#_handle_deferred_layer_dependencies.Functional._handle_deferred_layer_dependenciessb(,(K(K%*G*M*M*OP*O$!*O PE%22&-3QsAcgNTrs r7_should_compute_maskFunctional._should_compute_masksrOcbURXS9n[RRSU5$)N)maskc[USS5$)N _keras_mask)getattr)r5s r7rM)Functional.compute_mask..s ga5rO)_run_internal_graphr'r,r^)r1r2routput_tensorss r7 compute_maskFunctional.compute_masks311&1Dww$$ 5~  rOc"URXUS9$)a,Calls the model on new inputs. In this case `call` just reapplies all ops in the graph to the new inputs (e.g. build a new computational graph from the provided inputs). Args: inputs: A tensor or list of tensors. training: Boolean or boolean scalar tensor, indicating whether to run the `Network` in training mode or inference mode. mask: A mask or list of masks. A mask can be either a tensor or None (no mask). Returns: A tensor if there is a single output, or a list of tensors if there are more than one outputs. )r r)r)r1r2r rs r7roFunctional.calls&'''MMrOcF[R"USS9n[[RR U55[[RR UR 55:wa%[SUS[UR 5S35e[[R"USS95nX R;aURU$0n[UR [RR U55HupEURS-nXSU'M [URR55nURSS9 [U5S :Ga(UGHnURUn U GHln U R nX@R ;aM!/n U R"S n [RR U 5HWn U R$nUR RS UR&<S UR(<3-nU R+X?5 MY [RR-X5n [R"U SS9n UR/U 5n[R"USS9nUR0R3U 5n[5[RR U55HunnURS US U3-nXSU'M! GMo GM /n[7[UR855H>nUR:UunnnURS US U3-nUR+X65 M@ [RR-UR<U5nUURW'W$![a GNf=f) NF) to_tupleszInvalid `input_shape` argument z: the model expects z input tensors.T_0_0reverser:rr)rconvert_shapesrSr'r,r-rerrWrzipr!rRrlrsortr call_argsrFrrripack_sequence_ascompute_output_shape_inbound_nodesindexrrangerfrhrU)r1r cache_keylayers_to_output_shapesrrK shape_key depth_keysdepthrr layer_input_shapes layer_inputs layer_inputkhinput_layer_keylayer_output_shapesrj output_shapesirs r7rFunctional.compute_output_shapesm--kUK rww{+ , GGOOD.. /1  1+?%%(););%<$=>!!  '' tDI444// ::5#%    < LE  V+I16I .  $..3356 % z?Q #,,U3!D JJE 2 22!)+&#'>>!#4L')ww|'D (77*,((--MMOO;++113D (E*,)A)A$*& *2)@)@*d*&+0*D*D*+'+3*A*A+u+'"'!5!5!;!;D!AJ$-(;<%5%*JJ1ZL!1E$E =B : %G"$XM3t2234262J2J12M/z<!JJ1ZL,)HH $$%<%GH5GG44$$mM3@D $ $Y /U   s;N N N cU(d[URRnUR[:XaSn[R"[ R "U5US9UlgXlg)NModel) zero_based)r6__name__rrunique_object_namer to_snake_case_name)r1r!rcls_names r7_init_set_nameFunctional._init_set_namefsP~~..H~~+# 33++H5*DJJrOcT^URU5nUcS/[U5-nOURU5n[X5H upSX5lM 0mURn[UR U5H3upxUR XS9n[[U55n U/Xi-TU 'M5 URn [U R55n U RSS9 U Hn Xn U HnUR(aM[U4SjUR55(aM<UR!T5unnUR""U0UD6n[UR$[&R(R+U55HupU/Xi-TU 'M M M /nUR,HRn[[U55n U T;dS[U5-5eUR/TU R155 MT [&R(R3UR4U5$)aComputes output tensors for new inputs. # Note: - Can be run on non-Keras tensors. Args: inputs: Tensor or nested structure of Tensors. training: Boolean learning phase. mask: (Optional) Tensor or nested structure of Tensors. Returns: output_tensors N) ref_inputTrc3,># UH oT;v M g7fr<r)r@t_id tensor_dicts r7rA1Functional._run_internal_graph..sO;N4;.;NszCould not compute output )_flatten_to_reference_inputsrSrr_tensor_usage_countr2_conform_to_reference_inputstridrlrRrris_inputrXflat_input_ids map_argumentsrflat_output_idsr'r,r-r3rir#rrU)r1r2r rmasksinput_ttensor_usage_countrLyx_idrrrrr argsr4r3rrs @r7rFunctional._run_internal_graphss226: <FS[(E55d;E /MG"& 0 !55 V,DA000@Ar!u:D!"&8&> >K - --.--/0 %E")E==O4;N;NOOO#11+> f**d5f5 #(("''//'*B GD*+.@.F(FK%  $Ar!u:D;& L(Cc!f(L L&  ! !+d"3"7"7"9 : ww''(<(rrrFrr!rSwarningswarnformatKeyErrorr-)r1tensors ref_inputsref_input_namesinpns r7r'Functional._flatten_to_reference_inputssN  - - -*Wd2K2K,,J77$$Z00"112 *d++ #)):";>H#=Gc&&,,11Z # 7|c/22 FFLf$+LLNONq6NNOG !  0-<ZZR8F!|+ " 0B^^F> : :]]#5d#JKF FKKM*It/77 d//667 // >G}}($'='=gll12>N'OOCJJsD)P rOc0nURHn[RRR U5R 5unnUHgn[ U[R5(dM$XEnSRUVs/sHowRPM sn5nXQURS-U-'Mi M U$s snf)N.) r{r' __internal__trackingObjectGraphViewbreadth_first_traversalrQVariablejoinr!) r1resultr descendantsobject_paths_dict descendanttrackable_referencesr5 object_paths r7get_weight_pathsFunctional.get_weight_pathss[[E((88%%'  !* j"++66+<+H("%(()=>)=A)=>#K>H5::+k9: *! ?sB? c |[URVs1sHn[U5iM sn5[UR5:wa[SUR35eURHn[ US5(d(UR R n[SUSUS35eURRn[UR5S:d3UR(dM~URSR(aMUR R n[R"USURS URS U35 M [URVs/sH-n[R "URR5PM/ sn5nUR#S 5 [U5S:a"[R"S [%U535 UR&H;n[ US5(aMUR R n[S USU35e g s snfs snf)z4Validates the inputs and outputs of a Graph Network.zThe list of inputs passed to the model contains the same input multiple times. All inputs should only appear once.Received inputs=rFzInput tensors to a z8 model must come from `tf.keras.Input`. Received inputs=z# (missing previous layer metadata).r:rz model inputs must come from `tf.keras.Input` (thus holding past layer metadata). They cannot be the output of a previous non-Input layer. Here, a tensor specified as input to "z6" was not an Input tensor, it was generated by layer "zx". Note that input tensors are instantiated via `tensor = tf.keras.Input(shape)`. The tensor that caused the issue was: NzEFound incompatible static batch sizes among the inputs. Batch sizes: zOutput tensors of a z] model must be the output of a TensorFlow `Layer` (thus holding past layer metadata). Found: )rSr2r rrGr6rrFrrr loggingwarningr!rr get_static_batch_sizediscardrr3)r1rrLrrinput_batch_sizess r7r\-Functional._validate_graph_inputs_and_outputs1s# t{{+{!1{+ ,DKK0@ @#$(;;-1 A1.//>>22 )(4''(c)LN$$**E5''(1,$$$U-A-A!-D-M-M-M>>22j!! "& ,227**>=>?C A !<  $A44Q5E5E5K5KL$   !!$'  !A % OO((./@(A'BD  A1.//>>22 *8*5BBCFg,N s H4+4H9cJ^ ^[RRU5n[R"U5 0mUR R 5H'up4TRUVs0sHoUU_M sn5 M) U(d9[RRUVs/sHofRPM sn5n[U[TR55-5m U U4Sjn[R"U5nSn U(aU S- n U S:a [S5eURS5nU"U5nUcURU5 ML[!UR"R$UR"RR'U55n XR(;a>UTU'UR(R+U 5 UR URU5 U(aM[UR,5n /n UHtnXk;dM UR,RU5 U RU5 [.R0"UR25UR4U'U R+U5 Mv UR7U 5 UR95 gs snfs snf)aInserts Layers into the Network after Network creation. This is only valid for TF-Keras Graph Networks. Layers added via this function will be included in the `call` computation and `get_config` of this Network. They will not be added to the Network's outputs. Args: layers: Arbitrary nested structure of Layers. Layers must be reachable from one or more of the `keras.Input` Tensors that correspond to this Network's inputs. relevant_nodes: Nodes from the Layers that should be considered part of this Network. If `None`, all Nodes will be considered part of this Network. Raises: ValueError: If the layers depend on `Input`s not found in this Model. c>SnUR5H5up# nURUnUT;a[UTU5nM-UT;aM5 g US- $)z5Gets the minimum depth at which node can be computed.rNr:)iterate_inboundrmin)r min_depthrnode_idr inbound_node network_nodes node_to_depths r7_get_min_depth1Functional._insert_layers.._get_min_depthsgI(,(<(<(>$1$33G< =0 #I}\/J KI!6 )?q= rOrr:i'z6Layers could not be added due to missing dependencies.N)r'r,r-rrzrlrrrrrRrr;rr#ri_make_node_keyrr!rrkrrmrrnrorrrx)r1r{relevant_nodesrrr rrgunprocessed_nodesrnode_key layer_setdeferred_layersrerfs @@r7_insert_layersFunctional._insert_layersps<$(((0  00668LE  %!@%$+%!@ A9WW__39:6%%%6:NNT-2D2D2F-GGH  ! !IIn5  FA5y L%((+D"4(E}!((.% !:!:!@!@!FH222&+ d###''1$$U+2248) .556 E%--44U;&&u-3=3L3LJJ4))%0 e$ 00A ((*}"A ;s ( J #J c [R"5n[SUR55n[ UR R 55nURSS9 USSnUGH!nUR UGH n[RRUR5Vs1sHn[[U55iM nnURU5(dMg[RRUR5H"nU[[U55==S- ss'M$ [RRUR5H&nUR![[U555 M( GM GM$ URH"nU[[U55==S- ss'M$ Xlgs snf)aCompute the #. of tensor usages for all the output tensors of layers. The computed tensor usage count is saved as `self._tensor_usage_count`. This is later used for saving memory in eager computation by releasing no-longer-needed tensors as early as possible. c3J# UHn[[U55v M g7fr<)rr rHs r7rA9Functional._compute_tensor_usage_count..sJkFBvJks!#Trr:N)rCounterrr2rRrlrrr'r,r- keras_inputsrr issubsetr3rr) r1ravailable_tensorsrrr rI input_tensors output_tensors r7rx&Functional._compute_tensor_usage_countsk)002JdkkJJ$..3356 %^ E,,U3#%''//$2C2C"D!"D6 O"D!!))*;<<"$''//$2C2C"D*3r&z?;q@;#E*,)F )--c"]2C.DE*G4 llF s2f: /1 4 /#$6 !s-G cgr<rrs r7_assert_weights_created"Functional._assert_weights_createds rOc[URU/5up#[R"SURS9nU"U5 UR UR 5 URU5 URX25 g)NF) unconditionalr&) _map_subgraph_networkr2rAddLossr&extend inbound_nodesriro)r1 symbolic_loss new_nodes new_layersadd_loss_layers r7_graph_network_add_loss"Functional._graph_network_add_losssq 5 KK-!  $++}':':  }%556.) J2rOc[URU/5upE[R"X#URS9nU"U5 UR UR 5 URU5 URXT5 g)Nr%) rr2r AddMetricr&rrriro)r1r aggregationr!rradd_metric_layers r7_graph_network_add_metric$Functional._graph_network_add_metricsm 5dkkE7 K %// U[[  )778*+ J2rOc.[R"U5$r<)rNetworkSavedModelSaverrs r7_trackable_saved_model_saver'Functional._trackable_saved_model_savers$;;DAArOcJ>[USS5(aSn[TU] X5$)N_has_explicit_input_shapeTF)rr%_get_save_spec)r1 dynamic_batch inputs_onlyr6s r7rFunctional._get_save_specs* 44d ; ;"Mw%mAArO)rdr_r`rVrrrtrsrgrerPrrrrTrUrkrlrhrfrrrrmrr]rqr2rr3r) checkpointr<NN)T)TT)5r __module__ __qualname____firstlineno____doc__ frozenset itertoolschain training_libr_TF_MODULE_IGNORED_PROPERTIESr'rIrJ no_automatic_dependency_trackingr&r0propertyrwrr setterrrrprrrrrrrdo_not_doc_inheritablerorrrrrrdefaultrErUr\rorxr|rrrr__static_attributes__ __classcell__r6s@r7rr1sI^%.     < <  %!__>>2?24__>>H?HT # #DD"@(($$ EE '&2 0 ((N)N(`D :Nx&(P&P>$=~T+l6@ 33BBBBrOrc$US-[U5-$)Nz_ib-)r) layer_namers r7riris  Z 00rOc ^[U5unmUVs1sHFn[URRURRR U55iMH nn0n0n[ U5HnURUS5nURURS5n[Xx5nXvUR'XuU'URH&n URU S5n[US-U5XY'M( M UHXn U RSn X;dMSXk'STU 'SX[RS'UR[U RS55 MZ [R"[5n UR!5Hup7XR#U5 M [R"[5n UR!5HupXR#U5 M [U R%55nUR'SS9 /nUH+nXnUR'U4SjS9 UR)U5 M- [U R%55nUR'SS9 [+5nUHnUR[-U55 M /nUHnXHnURnU(dMUR.(aM+[0R2R5UR65H0n[-U5U;dM[9SUS URS U35e [0R2R5UR:5HnUR[-U55 M UR#UR5 M M UVs/sHoRPM nnUH9nUR=U5S:wdM[9S US UR=U5S 35e XLUU 4$s snfs snf)aValidates a network's topology and gather its layers and nodes. Args: inputs: List of input tensors. outputs: List of outputs tensors. Returns: A tuple `(nodes, nodes_by_depth, layers, layers_by_depth)`. - nodes: list of Node instances. - nodes_by_depth: dict mapping ints (depth) to lists of node instances. - layers: list of Layer instances. - layers_by_depth: dict mapping ints (depth) to lists of layer instances. Raises: ValueError: In case the network is not valid (e.g. disconnected graph). rr:r#Trc>TU$r<r)rL layer_indicess r7rM$_map_graph_network..cs M!,". The following previous layers were accessed without issue: z The name "z " is used z6 times in the model. All layer names should be unique.) _build_maprirr!rrreversed setdefaultrmax parent_nodesrFrr defaultdictrRrrirrrrr r r'r,r-rurr3count)r2r3nodes_in_decreasing_depthr re nodes_depths layers_depthsrprevious_depthnode_deprrrlayers_by_depthrrr{layers_for_depthcomputable_tensorsrLlayers_with_complete_input all_namesr!rs @r7rjrjs&0:'/B,}.-D tzz (A(A(G(G(MN- LM23''a0'**4::q9 E*$)djj!"T ))H)--h:N%(N%CL "*#40,,Q/  +)*M &)+M+ &:;L33A6 7   n[-=-=qA B !,,T2N#))+ $$T*,"--d3O%++- %%e,.o**,-JOODO!F*1 "<= &' n))+,JOODO!  r!u%"$")DJJEuT]]]):):;A!u$66(&&'S EJJ<@..H-IK<6A&**2a517*11%**=*$*00I0 ??4 A %TF*Y__T-B,CDHH  &/ AAIz1s A O3O8c[5n[5n/n0n[RRU5Hn[ UUUUU5 M X44$)aThis method topologically sorts nodes in order from inputs to outputs. It uses a depth-first search to topologically sort nodes that appear in the _keras_history connectivity metadata of `outputs`. Args: outputs: the output tensors whose _keras_history metadata should be walked. This may be an arbitrary nested structure. Returns: A tuple like (ordered_nodes, layer_to_first_traversal_index) ordered_nodes: list of nodes appearing in the keras history, topologically sorted from original inputs to the `outputs`. (If outputs have different sets of ancestors, the inputs to one output may appear after a different output). layer_to_first_traversal_index: A dict mapping layer to the traversal index in the DFS where it is seen. Note: if a layer is shared by several nodes, the dict will only store the index corresponding to the *first* time the layer seen. )rr'r,r-_build_map_helper)r3finished_nodesnodes_in_progressrrrs r7rrsW*UN "M''//'*    %   + % 33rOcURunnnURUnX;agX;a[SUSURS35eXT;a [ U5XE'UR U5 UR (d"URHn[UUUUU5 M UR U5 URU5 URU5 g)z"Recursive helper for `_build_map`.NzTensor z from layer "z" is part of a cycle.) rFrrr!rSrr rurremoveri) rIrrrrrrrr s r7rrs       +D   fX]5::,6K L   !"=1 $ ==''F !)  (tT"$$T*rOc,[RRR5(d[R "U5 [ X5up#pB[RRUR5Vs/sHoUPM sn5U4$s snf)zReturns the nodes and layers in the topology from `inputs` to `outputs`. Args: inputs: List of input tensors. outputs: List of output tensors. Returns: A tuple of List{Node] and List[Layer]. ) r'r(r)r*rr[rjr,r-rZ)r2r3rrr{rs r7rrsm 99<< ; ; = =--g6#5f#F Av 77??~/D/D/FG/FeE/FG H& PPGs= Bc[US5(d[U[5$UR(a>[U[5=(a' [URS[R 5$[U[5$)zCReturns True if the first layer node should not be saved or loaded.rmr)rGrQrrminput_layer_module InputLayer)rs r7_should_skip_first_nodersl 54 5 5%,, %% uj ) ..q13E3P3P %,,rOczUR5Vs/sHo"UR;dMUPM nnU(ax[RR UR5Vs/sH.n[ U5(aUR SSO UR PM0 sn5nURX45 U$s snfs snf)z?Adds layers that are not connected to the outputs to the model.r:N)rZr{r'r,r-rrro)modelcreated_layersrancillary_layersrjs r7connect_ancillary_layersrs*0022%5<<6O2 ,224  5E+511##AB'(()5    -> L  sB3B3$5B8c0^^^^^^^T=(d [R"5m0m0m[R"[5mU4SjmU4SjmUUUUU4SjnUUUU4SjnUSH nU"U5 M T(asUSHanTUSnUT;dMTR U5nU(dM-USnU"Xh5(aUR S5 OUTU'MXU(aM4Mc T(aMs/n /n [ R "USS S 9n [RRU 5HunUR5upnU T;deTU nT"Xm5n URU RnU R[RRU5U5 Mw [ R "US S S 9n[RRU5HunUR5upnU T;deTU nT"Xm5n URU RnU R[RRU5U5 Mw [RRX5n [RRUU 5n XT4$) a[Reconstructs graph from config object. Args: config: Dictionary returned from Network.get_config() custom_objects: Optional dictionary mapping names (strings) to custom classes or functions to be considered during deserialization. created_layers: Optional dictionary mapping names to Layer objects. Any layer not in this dictionary will be created and added to the dict. This function will add new nodes to all layers (excluding InputLayers), instead of re-using pre-existing nodes in the layers. Returns: Tuple of (input tensors, output tensors, dictionary of created layers) c>[U[R5(agTRURU4S5$)zBReturns node index in layer (might differ from config_node_index).rN)rQrrrr!)rconfig_node_indexnode_index_maps r7get_node_index/reconstruct_from_config..get_node_index8s7 e/:: ; ;!!5::/@"A4HHrOc|>^UU4Sjn[R"USS9n[RR X 5$)z0Deserializes TF-Keras Tensors passed to `call`..c&>[U[R5(apUR5nUSnUSnUSnTUnT"XB5nUc[eUR Un[ RRUR5U$U$)z7Deserializes a single TF-Keras Tensor passed to `call`.rr:r) rQr ListWrapperas_list IndexErrorrr'r,r-r3) r5rrrrnew_node_indexr r layer_maps r7_deserialize_keras_tensor^reconstruct_from_config.._deserialize_keras_tensors.._deserialize_keras_tensorAs!X1122IIKqT qT  t !*-!/!B!)%$++N;wwt||4\BBHrOTwrap)rconvert_inner_node_datar'r,r^)r4rrrs ` r7_deserialize_keras_tensors;reconstruct_from_config.._deserialize_keras_tensors>s2 011&tDww$$%>GGrOc>/n[RRU5GHlnUR5n[ U5S:Xa0nO*[ U5S:XaUSnT "UT5nO [ S5eUS[R:wamUSnUSnUSnTUnT"X5nUc gURUn UR[RRU R5U5 MUSS:Xde[ U5S:deUSn [U [5(aQ[ U 5S:XaBU S[R:Xa+UR[R "U S55 GM[URU 5 GMo [RR#X5nUb[%US 5(aUR&(d[(R*"U5nU"U40WD6n [RRU 5SR,R.n U TUR0TU4'TU==S- ss'g ![ a  gf=f) a$Deserialize a node. Args: layer: layer instance. node_data: Nested structure of `ListWrapper`. Returns: Whether the node was processed (i.e. the layer was called on the inputs specified by the node data) Raises: ValueError: In case of improperly formatted `node_data`. Fz"Improperly formatted model config.rr:rr##_preserve_input_structure_in_configT)r'r,r-rrSrr node_module_CONSTANT_VALUErrir3rQrW_COMPOSITE_TYPErdecoderrGrrunnest_if_single_tensorrFrr!)r node_datarx input_datar4inbound_layer_nameinbound_node_indexinbound_tensor_index inbound_layerrd const_valr output_indexrrrnode_count_by_layerrs r7 process_node-reconstruct_from_config..process_node\sX ''//)4J#++-J:!#ZA%#A!7OF !!EFF!} ; ;;%/]"%/]"'1!}$ ./A B %3!&"&- ,;;:nZZ   078   &! + &w"! !s I IIc>USnUT;aTUnOSSKJn U"UTS9nUTU'[[U55TU'USn[R "USS9nUHnT UR U5 M g) zDeserializes a layer, then call it on appropriate inputs. Args: layer_data: layer config dict. Raises: ValueError: In case of improperly formatted `layer_data` dict. r!r) deserialize)custom_objectsrTrN)tf_keras.src.layersrintrrrri) layer_datarrdeserialize_layerinbound_nodes_datarrrrrks r7 process_layer.reconstruct_from_config..process_layers '  '":.E M%jPE).N: &%()@)G%HE"(8%== T ,I e $ + +I 6 ,rOr{r!r input_layersTr output_layers)rrrrRr#rrr'r,r-rrrrir)r7rrrrrr layer_nodesrrxrrrrrlayer_output_tensorsrrrrrrks `` @@@@@r7reconstruct_from_configr s $@{'>'>'@N N$//5I HUunnn[URU5n XR;aM=XZnUR[,R."URUU/55 Mu [0R2R5UR@U5n[0R2R9U5(dU/n[,R:"U5nUUS 'U$!,(df  GN=f) a0Build the config, which consists of the node graph and serialized layers. Args: network: A Network object. serialize_layer_fn: Function used to serialize layers. config: A dict to append more config entries into. If None, start with a new dict for the config. Returns: Config dictionary. Fuse_legacy_configTr!r:rrr{Nrr)!r serialize_keras_objectrrr!r{rrrrirkSharedObjectSavingScoper  serializerirQrr rrSrergrrr'r,rrTr>rrfrhrU)networkserialize_layer_fnr7serialize_obj_fnset_layers_legacynode_conversion_mapr kept_nodesoriginal_node_indexr rl layer_configsfiltered_inbound_nodesr layer_config model_inputsrrrr model_outputss r7r=r= ss\rF(??w+U33(?? +?/?\\F6N1%88Qa )253G3G)H % %ejj2EFH1110:H-a *I M  . . 0^^E%' "-6u7K7K-L)#)%**6IJ555dmmm!%&(;!I+11)<.M%,,1B*.'-e4L#(::L ,B )   .#$$)x' 1,L 3w,,- .*1*D*DQ*G'z<!%**j9 11 1 ,6  %**nl!K L /77++ L 77  \ * *$~ 33LAL)>M 3w--. /*1*E*Ea*H'z<!%**j9 11 1 ,6  %**nl!K L 0GG,,M 77  ] + +& 44]CM+F? Mw 1 0s*AO!=O!BO!! O0cURRcgURR5nU(aSUS'U$)Nr)rKr(r)rLrKs r7rrgs4ww|| GGOO E a LrOc6^\rSrSrSrSU4SjjrSrSrU=r$) ModuleWrapperipzFWrapper for `tf.Module`s to support the Functional and Sequential API.c >[TU]"S0UD6 Uc'[US5(aSnO[US5(aSnUb[X5(d[USU35eXlX l[ X5n[R"U5nSUR;=(d URSLURl SUR;=(d URSLURl g)aInitializes the wrapper Layer for this module. Args: module: The `tf.Module` instance to be wrapped. method_name: (Optional) str. The name of the method to use as the forward pass of the module. If not set, becomes '__call__' if defined, or 'call'. Defaults to `None`. **kwargs: Additional keywrod arguments. See `tf.keras.layers.Layer`. Raises: ValueError: If `method` is not defined on `module`. N__call__roz is not defined on object r rr)r%r&rGr_module _method_namerrrnrr@rarbrc)r1module method_namer4methodmethod_arg_specr6s r7r&ModuleWrapper.__init__ss "6"  vz**( (($  gf&B&B },FvhOP P '-$33F; /.. . 1$$D0 , o** * Oo.C.C4.O (rOcSU;a"UR(dURS5 SU;a"UR(dURS5 [URUR 5"U0UD6$)Nr r)_expects_training_argr#_expects_mask_argrr!r")r1rr4s r7roModuleWrapper.calls[  (B(B JJz " V D$:$: JJv t||T%6%67HHHrO)r"r!r<) rrrrrr&rorrrs@r7rrpsP" HIIrOrc[R"UR5RSSn[R"[R5RSSnX:Xagg)Nr:TF)rrnr&rr)cls init_argsfunctional_init_argss r7r>r>sX))#,,7<rrOr7r<s O !! =*00A*38.14=HM,)'=.[B##[B|1xBv!4H++\Q"-*&n9bZz,IJ$$,I^rO