BVh DdZddlmZddlmZddlmZddlmZddlmZddl m Z ddl m Z dd l m Z dd l m Z dd l mZdd lmZddZdZGddej&ZGddeZGddeZGddej.ZGddej.Zy )z6Contains testing utilities related to mixed precision.)dtypes)ops)tensor_conversion) regularizers) base_layer) array_ops) check_opscondcustom_gradient)math_ops)nestNcHtjfdfd}|S)a|Returns 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. cFtjfd}|fS)z8Function that asserts it's gradient has a certain value.cr&|jk(sJdd|jtj|jd}tjg5t j ||}dddtjg5tj|}ddd|S#1swY?xYw#1swY|SxYw)z@Gradient function that asserts the gradient has a certain value.zdx.dtype should be z but is: expected_gradient)dtypenameN) rr"convert_to_tensor_v2_with_dispatchrcontrol_dependenciesr assert_equalridentity)dxexpected_tensor assert_opexpected_dtyperxs a/home/dcms/DCMS/lib/python3.12/site-packages/tensorflow/python/keras/mixed_precision/test_util.pygradzScreate_identity_with_grad_check_fn.._identity_with_grad_check..grad3s xx>) M2@"(( K M))LL 2882Eo  # #QC (@**2? @  # #YK 0$    #$ i @@$ is"B7C7CC rr)rr rrs` r_identity_with_grad_checkzEcreate_identity_with_grad_check_fn.._identity_with_grad_check/s% 1A d7Nc|SN)rr"s ridentity_with_grad_checkzDcreate_identity_with_grad_check_fn..identity_with_grad_checkFs $Q ''r#r )rrr'r"s`` @r"create_identity_with_grad_check_fnr(s("""#,( !!r#cDtjfdfd}|S)adReturns 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. c@tj|}fd}||fS)z@Function whose gradient is NaN iff `have_nan_gradients` is True.c@tjfdfdS)Nc tdzS)NNaN)floatrsrzkcreate_identity_with_nan_gradients_fn.._identity_with_nan_gradients..grad..bs"uU|#r#cSr%r&r/srr0zkcreate_identity_with_nan_gradients_fn.._identity_with_nan_gradients..grad..cs"r#r )rhave_nan_gradientss`rr zYcreate_identity_with_nan_gradients_fn.._identity_with_nan_gradients..grad_s YY  # r#r!)rr r2s r_identity_with_nan_gradientszKcreate_identity_with_nan_gradients_fn.._identity_with_nan_gradients[s% 1A d7Nr#c|Sr%r&)rr3s ridentity_with_nan_gradientszJcreate_identity_with_nan_gradients_fn..identity_with_nan_gradientshs ' **r#r )r2r5r3s` @r%create_identity_with_nan_gradients_fnr6Ks( "" # + $$r#c*eZdZdZdfd ZdZxZS)AssertTypeLayerz5A layer which asserts it's inputs are a certain type.c ||rtj|jnd|_t t |di|y)Nr&)ras_dtyper _assert_typesuperr8__init__)self assert_typekwargs __class__s rr=zAssertTypeLayer.__init__ps4>I5::"  /4)3F3r#c|jrktj|}|D]P}|jj|jk(r'Jd|jj d|jyy)zEAsserts `inputs` are of the correct type. Should be called in call().zInput tensor has type z" which does not match assert type N)r;rflattenr base_dtyper)r>inputsinputs_flattenedinps rassert_input_typesz"AssertTypeLayer.assert_input_typesusk f-!1#yy##t'8'88 1 YY^^T.. 0 181r#r%)__name__ __module__ __qualname____doc__r=rH __classcell__rAs@rr8r8ms=4 1r#r8cHeZdZdZ dfd ZdZdZdZfdZxZ S) MultiplyLayerz8A layer which multiplies its input by a scalar variable.c L||_t|tr$tj|t |_||_t|tr$tj|t |_||_||_tt|.dd|j i|y)aInitializes the MultiplyLayer. Args: regularizer: The weight regularizer on the scalar variable. activity_regularizer: The activity regularizer. use_operator: If True, add using the * operator. If False, add using tf.multiply. var_name: The name of the variable. It can be useful to pass a name other than 'v', to test having the attribute name (self.v) being different from the variable name. **kwargs: Passed to AssertTypeLayer constructor. )custom_objectsactivity_regularizerNr&) _regularizer isinstancedictr deserializeglobals_activity_regularizer _use_operator _var_namer<rPr=)r> regularizerrS use_operatorvar_namer@rAs rr=zMultiplyLayer.__init__s$$D+t$&22;BI)Md!5D&-#/#;#; wy$:d &DDN -'C!77C;ACr#cn|j|jdd|j|_d|_y)Nr&ones) initializerr\T) add_weightr[rTvbuilt)r>_s rbuildzMultiplyLayer.builds2 __ DrEs rcallzMultiplyLayer.calls%F# >>&$&& ))r#cP|jr||zStj||Sr%)rZrmultiply)r>rys rrhzMultiplyLayer._multiplys'  Ul   q! $$r#ctt| }tj|j |d<tj|j |d<|j|d<|j|d<|j|d<|S)Nr\rSr]r^r?) r<rP get_configr serializerTrYrZr[r;)r>configrAs rrozMultiplyLayer.get_configs =$ 2 4F(2243D3DEF=%1%;%; ""&$F !"!//F>F: --F= Mr#)NNFrc) rIrJrKrLr=rfrjrhrorMrNs@rrPrPs4@ $(! C@ *% r#rPceZdZdZdZdZy)MultiplyLayerWithoutAutoCastz:Same as MultiplyLayer, but does not use AutoCastVariables.c|j}|dvrd}|jddd|d|j|_d|_y) N)float16bfloat16float32rcr&r`F)rarexperimental_autocastr\T)rrbrTrcrd)r>rers rrfz"MultiplyLayerWithoutAutoCast.buildsN JJE ''e __ R#%% 'DF DJr#c|j||jjtjtj fvsJ|j |tj|j|jSr%) rHrcrrrwfloat64rhrcastris rrjz!MultiplyLayerWithoutAutoCast.callsWF# 66<>&(-- "E FFr#N)rIrJrKrLrfrjr&r#rrsrssB Gr#rsceZdZdZdZy)IdentityRegularizercj|jtjk(sJtj|Sr%)rrrwrrr>rs r__call__zIdentityRegularizer.__call__s) 77fnn $$ $   a  r#ciSr%r&r>s rrozIdentityRegularizer.get_config Ir#NrIrJrKrror&r#rr}r}s !r#r}ceZdZdZdZy)ReduceSumRegularizerc,tj|Sr%)r reduce_sumrs rrzReduceSumRegularizer.__call__s   q !!r#ciSr%r&rs rrozReduceSumRegularizer.get_configrr#Nrr&r#rrrs "r#rr%)rLtensorflow.python.frameworkrrrtensorflow.python.kerasrtensorflow.python.keras.enginertensorflow.python.opsrr r r rtensorflow.python.utilrr(r6Layerr8rPrs Regularizerr}rr&r#rrs=.+905++&1*'*"Z%D1j&&1$:O:zG=G*,22<33r#