6biK!SrSSKJs Jr SSKJr SSKJr SSjr Sr "SS\R5r "S S \ 5r "S S \ 5r"S S\R5r"SS\R5rg)z6Contains testing utilities related to mixed precision.N) regularizers) base_layercL^^^[RUU4Sj5mU4SjnU$)adReturns a function that asserts it's gradient has a certain value. This serves as a hook to assert intermediate gradients have a certain value. This returns an identity function. The identity's gradient function is also the identity function, except it asserts that the gradient equals `expected_gradient` and has dtype `expected_dtype`. Args: expected_gradient: The gradient function asserts that the gradient is this value. expected_dtype: The gradient function asserts the gradient has this dtype. Returns: An identity function whose gradient function asserts the gradient has a certain value. cJ>^[R"T5mUUU4SjnTU4$)z8Function that asserts it's gradient has a certain value.c>T(a'URT:XdSTSUR35e[R"TURSS9n[R"T/5 [RR R X5nSSS5 [R"W/5 [R"U5nSSS5 U$!,(df  NF=f!,(df  U$=f)z@Gradient function that asserts the gradient has a certain value.zdx.dtype should be z but is: expected_gradient)dtypenameN)r tfconvert_to_tensorcontrol_dependenciescompatv1 assert_equalidentity)dxexpected_tensor assert_opexpected_dtyperxs `/home/james-whalen/.local/lib/python3.13/site-packages/tf_keras/src/mixed_precision/test_util.pygradScreate_identity_with_grad_check_fn.._identity_with_grad_check..grad.sHH.M((8 "((LM. 22!8KO ((!-IILL55bJ .(()5[[_6I .-55Is'*C0C" C" C1r r)rrrrs` r_identity_with_grad_checkEcreate_identity_with_grad_check_fn.._identity_with_grad_check)s" KKN &$wc>T"U5$N)rrs ridentity_with_grad_checkDcreate_identity_with_grad_check_fn..identity_with_grad_checkEs(++rr custom_gradient)rrr!rs`` @r"create_identity_with_grad_check_fnr%s*$6, $#rcH^^[RU4Sj5mU4SjnU$)aOReturns a function that optionally has NaN gradients. This serves as a hook to introduce NaN gradients to a model. This returns an identity function. The identity's gradient function will check if the boolean tensor `have_nan_gradients` is True. If so, the gradient will be NaN. Otherwise, the gradient will also be the identity. Args: have_nan_gradients: A scalar boolean tensor. If True, gradients will be NaN. Otherwise, the gradient function is the identity function. Returns: An identity function whose gradient function will return NaNs, if `have_nan_gradients` is True. cB>[R"U5nU4SjnX4$)z@Function whose gradient is NaN iff `have_nan_gradients` is True.cF>^[R"TU4SjU4Sj5$)Nc >T[S5-$)NNaN)floatrsrkcreate_identity_with_nan_gradients_fn.._identity_with_nan_gradients..grad..csBu,=rc>T$rr r,srr-r.csrr)r cond)rhave_nan_gradientss`rrYcreate_identity_with_nan_gradients_fn.._identity_with_nan_gradients..gradas77"$=z rr)rrr1s r_identity_with_nan_gradientsKcreate_identity_with_nan_gradients_fn.._identity_with_nan_gradients\s  KKN  wrc>T"U5$rr )rr3s ridentity_with_nan_gradientsJcreate_identity_with_nan_gradients_fn..identity_with_nan_gradientsjs+A..rr#)r1r6r3s` @r%create_identity_with_nan_gradients_fnr8Ks*"  / '&rc6^\rSrSrSrSU4SjjrSrSrU=r$)AssertTypeLayerpz5A layer which asserts it's inputs are a certain type.c >U(a [R"U5ROSUl[TU]"S0UD6 g)Nr )r as_dtyper _assert_typesuper__init__)self assert_typekwargs __class__s rr@AssertTypeLayer.__init__ss3-8BKK $ ) )d  "6"rcUR(az[RRU5nUHTnURR UR:XaM)SURR <SUR<35e gg)zEAsserts `inputs` are of the correct type. Should be called in call().zInput tensor has type z" which does not match assert type N)r>r nestflattenr base_dtyper )rAinputsinputs_flattenedinps rassert_input_types"AssertTypeLayer.assert_input_typesysn   !wwv6 'yy++t/@/@@yy~~t'8'8:@( r)r>r) __name__ __module__ __qualname____firstlineno____doc__r@rM__static_attributes__ __classcell__rDs@rr:r:ps?#  rr:cV^\rSrSrSrS U4SjjrSrSrSrU4Sjr Sr U=r $) MultiplyLayerz8A layer which multiplies its input by a scalar variable.c J>Xl[U[5(a"[R"U[ 5S9UlX l[U[5(a"[R"U[ 5S9UlX0lX@l[TU]("SSUR 0UD6 g)aInitializes the MultiplyLayer. Args: regularizer: The weight regularizer on the scalar variable. activity_regularizer: The activity regularizer. use_operator: If True, add using the * operator. If False, add using tf.multiply. var_name: The name of the variable. It can be useful to pass a name other than 'v', to test having the attribute name (self.v) being different from the variable name. **kwargs: Passed to AssertTypeLayer constructor. )custom_objectsactivity_regularizerNr ) _regularizer isinstancedictr deserializeglobals_activity_regularizer _use_operator _var_namer?r@)rA regularizerr\ use_operatorvar_namerCrDs rr@MultiplyLayer.__init__s(( k4 ( ( , 8 8GI!D &:" *D 1 1)5)A)A$WY*D &*!  !%!;!; ?E rchURURSSURS9UlSUlg)Nr ones) initializerreT) add_weightrdr]vbuilt)rA_s rbuildMultiplyLayer.builds6 NN )) !   rcZURU5 URXR5$r)rM _multiplyrmrArJs rcallMultiplyLayer.calls# '~~fff--rcXUR(aX-$[R"X5$r)rcr multiply)rArys rrsMultiplyLayer._multiplys!   5L;;q$ $rc >[TU]5n[R"UR5US'[R"UR 5US'UR US'URUS'URUS'U$)Nrer\rfrgrB) r? get_configr serializer]rbrcrdr>)rAconfigrDs rr|MultiplyLayer.get_configs#% , 6 6t7H7H I})5)?)?  & &* %&"&!3!3~!^^z $ 1 1} r)rbr]rcrdrnrm)NNFrm) rOrPrQrRrSr@rprursr|rTrUrVs@rrXrXs3B! # J.%  rrXc$\rSrSrSrSrSrSrg)MultiplyLayerWithoutAutoCastz:Same as MultiplyLayer, but does not use AutoCastVariables.c URnUS;aSnURSSSUSURS9UlSUlg) N)float16bfloat16float32rmr rjF)rkr experimental_autocastreT)r rlr]rmrn)rAror s rrp"MultiplyLayerWithoutAutoCast.buildsO  + +E  "')) !  rcURU5 URR[R[R 4;deUR U[R"URUR55$r)rMrmr r rfloat64rscastrts rru!MultiplyLayerWithoutAutoCast.callsW 'vv|| BJJ7777~~fbggdfffll&CDDr)rnrmN)rOrPrQrRrSrprurTr rrrrsD Errc \rSrSrSrSrSrg)IdentityRegularizercnUR[R:Xde[R"U5$r)r r rrrArs r__call__IdentityRegularizer.__call__s&ww"**$$${{1~rc0$rr rAs rr|IdentityRegularizer.get_config rr NrOrPrQrRrr|rTr rrrrs rrc \rSrSrSrSrSrg)ReduceSumRegularizerc.[R"U5$r)r reduce_sumrs rrReduceSumRegularizer.__call__s}}Qrc0$rr rs rr|ReduceSumRegularizer.get_configrrr Nrr rrrrs  rrr)rStensorflow.compat.v2rv2r tf_keras.srcrtf_keras.src.enginerr%r8Layerr:rXr Regularizerrrr rrrsu=!!%*1$h"'Jj&&,DODNE=E.,22<33r