6bipSrSSKJs Jr SSKJr SSKJr SSK J r Sr \ "S/S9"S S 55r \ R\R5 \ "S /S9"S S \ 55r\ "S/S9"SS\ 55r\R#\R$S5 \R#\R$S5 \R#\R$S5 \R#\R$S5 \R#\R$S5 "SS\R&5r"SS\ 5r"SS5r\R\ 4\R.\4\R$\4\\ 4/rSrSrSrS!SjrS rg)"z9Keras Input Tensor used to track functional API Topology.N)object_identity) structure) keras_exportzkeras.__internal__.KerasTensor)v1c\rSrSrSrSSjr\S5r\S5r\ S5r \ SSj5r S r S r S r\S 5rS rSrSSjr\S5rSrSrSr\S5rSr\S5rSr\S5r\ S5r\ S5rSrg) KerasTensor!a+A representation of a TF-Keras in/output during Functional API construction. `KerasTensor`s are tensor-like objects that represent the symbolic inputs and outputs of TF-Keras layers during Functional model construction. They are comprised of the `tf.TypeSpec` of the (Composite)Tensor that will be consumed/produced in the corresponding location of the Functional model. KerasTensors are intended as a private API, so users should never need to directly instantiate `KerasTensor`s. **Building Functional Models with KerasTensors** `tf.keras.Input` produces `KerasTensor`s that represent the symbolic inputs to your model. Passing a `KerasTensor` to a `tf.keras.Layer` `__call__` lets the layer know that you are building a Functional model. The layer __call__ will infer the output signature and return `KerasTensor`s with `tf.TypeSpec`s corresponding to the symbolic outputs of that layer call. These output `KerasTensor`s will have all of the internal KerasHistory metadata attached to them that TF-Keras needs to construct a Functional Model. Currently, layers infer the output signature by: * creating a scratch `FuncGraph` * making placeholders in the scratch graph that match the input typespecs * Calling `layer.call` on these placeholders * extracting the signatures of the outputs before clearing the scratch graph (Note: names assigned to KerasTensors by this process are not guaranteed to be unique, and are subject to implementation details). `tf.nest` methods are used to insure all of the inputs/output data structures get maintained, with elements swapped between KerasTensors and placeholders. In rare cases (such as when directly manipulating shapes using Keras layers), the layer may be able to partially infer the value of the output in addition to just inferring the signature. When this happens, the returned KerasTensor will also contain the inferred value information. Follow-on layers can use this information. during their own output signature inference. E.g. if one layer produces a symbolic `KerasTensor` that the next layer uses as the shape of its outputs, partially knowing the value helps infer the output shape. **Automatically converting TF APIs to layers**: If you passing a `KerasTensor` to a TF API that supports dispatching, TF-Keras will automatically turn that API call into a lambda layer in the Functional model, and return KerasTensors representing the symbolic outputs. Most TF APIs that take only tensors as input and produce output tensors will support dispatching. Calling a `tf.function` does not support dispatching, so you cannot pass `KerasTensor`s as inputs to a `tf.function`. Higher-order APIs that take methods which produce tensors (e.g. `tf.while`, `tf.map_fn`, `tf.cond`) also do not currently support dispatching. So, you cannot directly pass KerasTensors as inputs to these APIs either. If you want to use these APIs inside of a Functional model, you must put them inside of a custom layer. Args: type_spec: The `tf.TypeSpec` for the symbolic input created by `tf.keras.Input`, or symbolically inferred for the output during a symbolic layer `__call__`. inferred_value: (Optional) a non-symbolic static value, possibly partially specified, that could be symbolically inferred for the outputs during a symbolic layer `__call__`. This will generally only happen when grabbing and manipulating `tf.int32` shapes directly as tensors. Statically inferring values in this way and storing them in the KerasTensor allows follow-on layers to infer output signatures more effectively. (e.g. when using a symbolic shape tensor to later construct a tensor with that shape). name: (optional) string name for this KerasTensor. Names automatically generated by symbolic layer `__call__`s are not guaranteed to be unique, and are subject to implementation details. Nc[U[R5(d [S5eXlX lX0l[U[R5(d[US5(d"[S[U5RS35e[UR[R5(dB[S[U5RS[UR5RS35egg) zConstructs a KerasTensor.z6KerasTensors must be constructed with a `tf.TypeSpec`.shapezAKerasTensor only supports TypeSpecs that have a shape field; got z, which does not have a shape.zRKerasTensor requires that wrapped TypeSpec's shape is a TensorShape; got TypeSpec z(, whose shape field has unexpected type .N) isinstancetfTypeSpec ValueError _type_spec_inferred_value_namerNoneTensorSpechasattrtype __qualname__r TensorShape TypeErrordtype)self type_specinferred_valuenames Z/home/james-whalen/.local/lib/python3.13/site-packages/tf_keras/src/engine/keras_tensor.py__init__KerasTensor.__init__ts)R[[11H $- )Y%=%=>>9g.. ""&y/">">!?@33 ioor~~>>115i1M1M0N>IOO,99:!=??cUR$)zEReturns the `tf.TypeSpec` symbolically inferred for TF-Keras output. )rrs r rKerasTensor.type_specs r#c.URR$)zEReturns the `TensorShape` symbolically inferred for TF-Keras output. )rr r%s r r KerasTensor.shapes $$$r#c<[U[R5(a[USS5n[R"U5nSnUR [R :XaURRbqURRS:aW[R"US9RnUR(a&UR5n[U5[:aSnOSn[X4US9$[USS5n[R"U5nU"X2S9$)zCConvert a traced (composite)tensor to a representative KerasTensor.rNr )rrr)rrTensorgetattrtype_spec_from_valuerint32r rankonesdimsas_listlen_MAX_TENSOR_RANKr )clstensorrrrs r from_tensorKerasTensor.from_tensors fbii ( (6640D//7I!N288+OO((4OO((1,2"$v!6!*-==)-%)Nt  6640D//7Iy, ,r#cU"XS9$)N)rrr7rrs r from_type_specKerasTensor.from_type_specs Y22r#cpURb[R"[RRR UR[R S95nURRRS:XaUSnU$Sn[RRX RSS9$)z5Convert this KerasTensor to a placeholder in a graph.)r rrc~[RRRURUR 5$N)rcompatr placeholderrr ) components r component_to_placeholder=KerasTensor._to_placeholder..component_to_placeholders%99<<++IOOY__M Mr#T)expand_composites) rrr rCrrDr0rr1nest map_structure)rrrFs r _to_placeholderKerasTensor._to_placeholders    + XX ((..bhh)N ~~##((A-"0!2! ! Nww$$ $nn%  r#cUR$rBr+r%s r get_shapeKerasTensor.get_shapes zzr#c[S5e)NaeKeras symbolic inputs/outputs do not implement `__len__`. You may be trying to pass TF-Keras symbolic inputs/outputs to a TF API that does not register dispatching, preventing TF-Keras from automatically converting the API call to a lambda layer in the Functional Model. This error will also get raised if you try asserting a symbolic input/output directly.rr%s r __len__KerasTensor.__len__s E  r#c[S5e)Na Keras symbolic inputs/outputs do not implement `op`. You may be trying to pass TF-Keras symbolic inputs/outputs to a TF API that does not register dispatching, preventing TF-Keras from automatically converting the API call to a lambda layer in the Functional Model.rQr%s r opKerasTensor.ops '  r#c [SUS35e)Nz%Tensors are unhashable (this tensor: z(). Instead, use tensor.ref() as the key.rQr%s r __hash__KerasTensor.__hash__s 3D6:4 4  r#dc [SUS35e)NzYou are passing a, an intermediate TF-Keras symbolic input/output, to a TF API that does not allow registering custom dispatchers, such as `tf.cond`, `tf.function`, gradient tapes, or `tf.map_fn`. TF-Keras Functional model construction only supports TF API calls that *do* support dispatching, such as `tf.math.add` or `tf.reshape`. Other APIs cannot be called directly on symbolic Kerasinputs/outputs. You can work around this limitation by putting the operation in a custom TF-Keras layer `call` and calling that layer on this symbolic input/output.rQ)rrs r __array__KerasTensor.__array__!s tf %- -  r#cg)NTr<r%s r is_tensor_likeKerasTensor.is_tensor_like0sr#cR[U[R5(d[R"U5nURR U5(d[ SURSUS35eURR U5n[URU5Ulg)zFUpdates the shape of this KerasTensor. Mimics `tf.Tensor.set_shape()`.z$Keras symbolic input/output's shape z' is not compatible with supplied shape r N) rrrr is_compatible_withr merge_withtype_spec_with_shaper)rr s r set_shapeKerasTensor.set_shape4s%00NN5)Ezz,,U336tzzlC227;  JJ))%0E24??EJDOr#c$SnSnSn[US5(a'URRnSUR<S3nURbSUR3nURbSUR S3nSUR <U<U<U<S 3$) N_keras_historyz!, description="created by layer 'z'"z, inferred_value=z, name=''zKerasTensor(type_spec=))rrilayerrrrr)rsymbolic_descriptioninferred_value_string name_stringrls r __str__KerasTensor.__str__Bs! " 4) * *''--E $     +&78L8L7M$N ! 99 $TZZL2K NN !    r#cSnSn[UR[R5(a'SURSUR R 3nOSUR3n[US5(a&URRnSUR S3nURbSUR3nS U<U<U<S 3$) Nrhzshape=z dtype=z type_spec=riz (created by layer 'z')z inferred_value=z) rrr TensorSpecr rrrrirlr)rrmrntype_spec_stringrls r __repr__KerasTensor.__repr__Ws! " dnnbmm 4 4!' |74::??:KL !+DNN+;<  4) * *''--E%9%**R#H    +&6t7K7K6L$M !  !   r#cnURn[US5(d"[S[U5RS35e[ UR [R5(dB[S[U5RS[UR 5RS35eUR $)zDReturns the `dtype` symbolically inferred for this TF-Keras output. rzKerasTensor wraps TypeSpec z, which does not have a dtype.zLKerasTensor requires that wrapped TypeSpec's dtype is a DType; got TypeSpec z(, whose dtype field has unexpected type r ) rrAttributeErrorrrrrrDTyper)rrs r rKerasTensor.dtypejs OO y'** -d9o.J.J-KL// )//28844''+I'C'C&DE3 (556a9  r#c.[R"U5$)ayReturns a hashable reference object to this KerasTensor. The primary use case for this API is to put KerasTensors in a set/dictionary. We can't put tensors in a set/dictionary as `tensor.__hash__()` is not available and tensor equality (`==`) is supposed to produce a tensor representing if the two inputs are equal. See the documentation of `tf.Tensor.ref()` for more info. )r Referencer%s r refKerasTensor.ref~s((..r#cb[US5(aURupnURU$g)aFind the corresponding `Node` that produce this keras_tensor. During functional model construction, TF-Keras will attach `KerasHistory` to keras tensor to track the connectivity between calls of layers. Return None if there isn't any KerasHistory attached to this tensor. riN)rri inbound_nodes)rrl node_index_s r nodeKerasTensor.nodes6 4) * *#'#6#6 Eq&&z2 2r#cSnURRb/URRVs/sHo"RPM nnUc [ S5eU(d [ S5eUSc [ S5e[ XS5$s snf)Nz0Cannot iterate over a Tensor with unknown shape.zCannot iterate over a scalar.rz:Cannot iterate over a Tensor with unknown first dimension.)r ndimsr3valuer_KerasTensorIterator)rr dims r __iter__KerasTensor.__iter__s ::   '*.**//:/3YY/E: =NO O;< < 8 L $D(33;sBcUR$)zEReturns the (non-unique, optional) name of this symbolic Keras value.)rr%s r rKerasTensor.nameszzr#c[RRHnURX5 M [ US5(aURUS5 gg)z%Register overloads for all operators.experimental_refN)rr-OVERLOADABLE_OPERATORS_overload_operatorr)r7 tensor_classoperators r _overload_all_operators#KerasTensor._overload_all_operatorssL 88H  " "< :9 rKrNrRrUrX__array_priority__r\r_rerprvrr~rrrrr__static_attributes__r<r#r r r !s)Ob4 %% 2-2-h33  D          K * && /   4 EE,,r#r z$keras.__internal__.SparseKerasTensorc\rSrSrSrSrSrg)SparseKerasTensorizA specialized KerasTensor representation for `tf.sparse.SparseTensor`s. Specifically, it specializes the conversion to a placeholder in order to maintain dense shape information. cURn[RRR UR UR S9$)N)rr )rrrCrsparse_placeholderrr )rspecs r rK!SparseKerasTensor._to_placeholders:~~yy||..**DJJ/  r#r<N)rrrrrrKrr<r#r rrs   r#rz$keras.__internal__.RaggedKerasTensorc<^\rSrSrSrU4Sjr\S5rSrU=r $)RaggedKerasTensoriaA specialized KerasTensor representation for `tf.RaggedTensor`s. Specifically, it: 1. Specializes the conversion to a placeholder in order to maintain shape information for non-ragged dimensions. 2. Overloads the KerasTensor's operators with the RaggedTensor versions when they don't match the `tf.Tensor` versions 3. Exposes some of the instance method/attribute that are unique to the RaggedTensor API (such as ragged_rank). cj>URnURS:XdURRc[T U]5$URURSn[ RRRURU5n/nSnURH/nUbUb[USS5cSnOXV-nURU5 M1 [URSS5HnURUnUb[USS5cdXGS- nUbUS-n[ RRRURU/5n [ RR!X9SS9nM[ R""XaR5n [ RR%X:SS9nM U$)NrrTF)validate)r ragged_rankr r1superrKrrCrrDrr.appendrangerow_splits_dtype RaggedTensorfrom_row_splitsconstantfrom_uniform_row_length) r ragged_spec flat_shaperesultknown_num_splitsprod axis_sizeaxis num_splitssplitsrowlen __class__s r rK!RaggedKerasTensor._to_placeholdersnn  " "a ';+<+<+A+A+I7*, , &&{'>'>'@A ))+*;*;ZH$**I$Iw5=D+D  # #D )++111b9D#))$/I WY%F%N-Qh7 )!+aJ1100:,88U9Y0L0LM@@UA:" r#c.URR$rB)rrr%s r rRaggedKerasTensor.ragged_ranks~~)))r#r<) rrrrrrKrrr __classcell__rs@r rrs" %N**r#r __getitem____add____radd____mul____rmul__c<\rSrSrSrSrSrSrSrSr Sr S r g ) UserRegisteredSpeci/z7TypeSpec to represent user-registered symbolic objects.c(XlX lX lgrB)r _dtyper)rr rs r r!UserRegisteredSpec.__init__2s   r#c[erBNotImplementedErrorr%s r _component_specs#UserRegisteredSpec._component_specs7!!r#c[erBr)r componentss r _from_components#UserRegisteredSpec._from_components:rr#c[erBrr%s r _serializeUserRegisteredSpec._serialize=rr#c[erBr)rrs r _to_components!UserRegisteredSpec._to_components@rr#c[erBrr%s r value_typeUserRegisteredSpec.value_typeCrr#)rrr N) rrrrrr!rrrrrrr<r#r rr/s#A """""r#rcV^\rSrSrSrU4Sjr\S5r\SSj5rSr Sr U=r $) UserRegisteredTypeKerasTensoriMzAKerasTensor that represents legacy register_symbolic_tensor_type.c>UnX l[URUR5n[ USS5n[ TU]X45 g)Nr) _user_registered_symbolic_objectrr rr.rr!)ruser_registered_symbolic_objectxrrrs r r!&UserRegisteredTypeKerasTensor.__init__Ps= +01-&qww8 q&$' )r#cU"U5$rBr<)r7r8s r r9)UserRegisteredTypeKerasTensor.from_tensorXs 6{r#c[SU-5e)Nz?You cannot instantiate a KerasTensor directly from TypeSpec: %srr=s r r>,UserRegisteredTypeKerasTensor.from_type_spec\s! M   r#cUR$rBrr%s r rK-UserRegisteredTypeKerasTensor._to_placeholdercs444r#rrB) rrrrrr!rr9r>rKrrrs@r rrMs=K*  55r#rc*\rSrSrSrSrSrSrSrg)rigzHIterates over the leading dim of a KerasTensor. Performs 0 error checks.c*XlSUlX lg)Nr)_tensor_index_limit)rr8dim0s r r!_KerasTensorIterator.__init__ks   r#cU$rBr<r%s r r_KerasTensorIterator.__iter__ps r#cURUR:Xa[eURURnU=RS- slU$)Nr)rr StopIterationr)rrs r __next___KerasTensorIterator.__next__ss= ;;$++ % dkk* q  r#)rrrN) rrrrrr!rrrr<r#r rrgs r#rc2[RSX45 g)z>Register a specialized KerasTensor subclass for a Tensor type.rN)keras_tensor_classesinsert)r7keras_tensor_subclasss r $register_keras_tensor_specializationrsS$@Ar#cP[U[5(aUR5$U$)zFConstruct a graph placeholder to represent a KerasTensor when tracing.)rr rK)rs r keras_tensor_to_placeholderrs#![!!  ""r#cSn[Hup#[X5(dMUn O URU5n[USS5b[ UR 5UlU$)zCConvert a traced (composite)tensor to a representative KerasTensor.N _keras_mask)rrr9r.keras_tensor_from_tensorr )r8keras_tensor_cls tensor_typer7outs r r r sc0 f * *"  1  & &v .Cv}d+7263E3EF Jr#cSnURn[HupE[X45(dMUn O URXS9$)z3Convert a TypeSpec to a representative KerasTensor.Nr,)rr issubclassr>)rrr rr r7s r keras_tensor_from_type_specrsL%%J0 j . ."  1  * *9 * @@r#c[U[R5(a[R"U5nXlU$[U[R 5(aB[R "UUR URURUR5$[U[R5(a [R"XR 5$[US5(aURU5$[S5e)z@Returns a copy of TypeSpec `spec` with its shape set to `shape`. with_shapezjKeras requires TypeSpec to have a `with_shape` method that returns a copy of `self` with an updated shape.)rrrtr_shapeRaggedTensorSpecrrrflat_values_specSparseTensorSpecrrr)rr s r rdrds$ &&u%  D"-- . .""  JJ     ! !  ! !    D"-- . .""5**55 | $ $u%% C  r#rB)rtensorflow.compat.v2rCv2rtf_keras.src.utilsrtensorflow.python.data.utilr tensorflow.python.util.tf_exportrr6r rr-rrrrrrrr SparseTensorobjectrrrr rrdr<r#r rs@!!.29.26h,h,7h,V  ##BII.4<  = (4<6* 6*=6*t$$R__mD$$R__i@$$R__jA$$R__i@$$R__jA""<5K54:YY __'(__'( [ B " A r#