3jXrSSKJr SSKJrJrJrJrJrJr SSK r SSK J r SSK J r J r Jr SSKJr SSKJrJrJrJr SSKJr SS KJr SS KJr \\ R8\\ R8\ R844rS \ R<S \S \ R84Sjr"SS\ R<5r "SS\ 5r!g))Enum)AnyCallableDictOptionalTupleUnionN)nn) Bernoulli DistributionRelaxedBernoulli)RandomGeneratorBase)_adapted_rsampling_adapted_sampling_transform_output_shapeoverride_parameters)is_autocast_enabled)Boxes) KeypointsselfinputreturncF[S[U5RS35e)zYDefine the computation performed at every call. Should be overridden by all subclasses. z nn.Module [z3] is missing the required "apply_tranform" function)NotImplementedErrortype__name__rrs R/home/wildlama/miniconda3/lib/python3.13/site-packages/kornia/augmentation/base.py_apply_transform_unimplementedr+s$  DJ,?,?+@@st uuc ^\rSrSr%SrSrS)S\S\S\S\SS 4 U4S jjjr\ r \ S \ R4\S 'S \S\SS4U4SjjrS\4SjrS\ RS\ R4SjrS SS.S\ RS\\ RS\S\ R4SjjrS\ RSS 4SjrS\ RS\\S 4S\ R4SjrS\\S 4S\\\ R44SjrS\ R8S\ R:SS 4S jrS\\S 4S\S\S\S\ R4 S!jrS*S"\\\\ R4S#\\\\4S\S\\\\ R4\\\444S$jjr S\\S 4S\\\ R44S%jr!S\ RS"\\\ R4S#\\\4S\ R4S&jr"S+S\ RS"\\\\ R4S\S\ R4S'jjr#S(r$U=r%$),_BasicAugmentationBase3a_BasicAugmentationBase base class for customized augmentation implementations. Plain augmentation base class without the functionality of transformation matrix calculations. By default, the random computations will be happened on CPU with ``torch.get_default_dtype()``. To change this behaviour, please use ``set_rng_device_and_dtype``. For automatically generating the corresponding ``__repr__`` with full customized parameters, you may need to implement ``_param_generator`` by inheriting ``RandomGeneratorBase`` for generating random parameters and put all static parameters inside ``self.flags``. You may take the advantage of ``PlainUniformGenerator`` to generate simple uniform parameters with less boilerplate code. Args: p: probability for applying an augmentation. This param controls the augmentation probabilities element-wise. p_batch: probability for applying an augmentation to a batch. This param controls the augmentation probabilities batch-wise. same_on_batch: apply the same transformation across the batch. keepdim: whether to keep the output shape the same as input ``True`` or broadcast it to the batch form ``False``. Fpp_batch same_on_batchkeepdimrNcv>[TU]5 XlX lX0lX@l0UlU U US;a[UR5UlUS;a[UR5Ul SUl 0Ul UR[R"S5[R"55 g)N>?cpu)super__init__r$r%r&r'_paramsr _p_gen _p_batch_gen_param_generatorflagsset_rng_device_and_dtypetorchdeviceget_default_dtype)rr$r%r&r' __class__s rr-_BasicAugmentationBase.__init__Ms  * 02   J #DFF+DK * $ )$,, 7D ?C%'  %%ell5&95;R;R;TUr .apply_transformargskwargsc>[RRR"U0UD6up4 nUR X45 [ TU]"U0UD6$)z9Set the device and dtype for the random number generator.)r4_C_nn _parse_tor3r,to)rr:r;r5dtype_r7s rr@_BasicAugmentationBase.tofsI#hhll44dEfEq! %%f4wz4*6**r cSURSURSUR3n[UR[ 5(aUR<SU3nUR R5HJup#[U[5(a%USUSURR53- nM?USUSU3- nML URRSUS3$)Nzp=z , p_batch=z, same_on_batch=z, =()) r$r%r& isinstancer1rr2itemsrnamelowerr7r)rtxtkvs r__repr___BasicAugmentationBase.__repr__ls466(*T\\N2B4CUCUBVW d++-@ A A**-Ru5CJJ$$&DA!T""A3a /00A3as|# ' ..))*!C522r rcU$Nrs r__unpack_input__'_BasicAugmentationBase.__unpack_input__ws r Tshape match_channelrWrXc[e)zStandardize input tensors.r)rrrWrXs rtransform_tensor'_BasicAugmentationBase.transform_tensorz "!r c[e)z9Check if the input torch.Tensor is formatted as expected.rZrs rvalidate_tensor&_BasicAugmentationBase.validate_tensors!!r output output_shapec>UR(a [X5$U$)zStandardize output tensors.)r'r)rrarbs rtransform_output_tensor._BasicAugmentationBase.transform_output_tensors@D &v<XRXXr batch_shapecVURbURXR5$0$rR)r1r&)rrfs rgenerate_parameters*_BasicAugmentationBase.generate_parameterss*  ,((6H6HI I r r5rAcnXlX lURbURRX5 gg)zChange the random generation device and dtype. Note: The generated random numbers are not reproducible across different devices and dtypes. N)r5rAr1r3)rr5rAs rr3/_BasicAugmentationBase.set_rng_device_and_dtypes3   ,  ! ! : :6 I -r cUS:Xa[R"S5S-nOlUS:Xa[R"S5nOO[UR[45(a[ SURU5nO[ SURU5nUR5S:XaUS:Xa[R"US5S-nOwUS:Xa[R"US5nOW[UR[45(a[ US4URU5nO[ US4URU5nXV-nOURUS5n[UR5S:XaUS$U$)Nr)rm).r) r4zerosrHr0r rrsumr/repeatlenrW)rrfr$r%r& batch_prob elem_probs r__batch_prob_generator__/_BasicAugmentationBase.__batch_prob_generator__s= a<Q!+J \QJ )),<+> ? ?+D$2C2C]SJ*41B1BMRJ >> q Av!KK A7!; a!KK A7 DKK*:)<==. A/@$++}] -{1~.?m\ #/J#**;q>:J z A %f% %r paramsr2c SU;aUSOSnUc UROUnUc UROUnU(a[XSS9nXlOXl[XSS9n[X#SS9nX4$)N save_kwargsFT)in_place)r.r2r)rrwr2r;rys r#_process_kwargs_to_params_and_flags:_BasicAugmentationBase._process_kwargs_to_params_and_flagssn0=/Ff]+E !'V#m  ($GF!L!L(%HF#EEB}r cURXRURUR5nUS:nUR [ R "[UR5R55/USSQ755nUc0nX$S'[ R"U[ RS9nURSU05 U$)N?rmrs)rAforward_input_shape) rur$r%r&rhr4Sizeintrpitemtensorlongupdate)rrfrsto_applyr. input_sizes rforward_parameters)_BasicAugmentationBase.forward_parameterss22; VZVhVhi #**5::s8<<>;N;N;P7Q6dT_`a`bTc6d+ef ?G * \\+UZZ@ -z:;r c&URXU5$rR)r9)rrrwr2s r apply_func!_BasicAugmentationBase.apply_funcs##E599r c URU5nURnURU5nURnUcURU5nSU;a [R "S/US-5US'UR "X R40UD6up'URXBU5nUR(aURX5$U$)aPerform forward operations. Args: input: the input torch.Tensor. params: the corresponding parameters for an operation. If None, a new parameter suite will be generated. **kwargs: key-value pairs to override the parameters and flags. Note: By default, all the overwriting parameters in kwargs will not be recorded as in ``self._params``. If you wish it to be recorded, you may pass ``save_kwargs=True`` additionally. rsTr) rTrWr[rr4rr{r2rr'rd) rrrwr; in_tensor input_shaperfr2ras rforward_BasicAugmentationBase.forwards"))%0 oo )))4 oo >,,[9F v %#(<<Q0G#HF< @@^W]^ E:DHLLt++F@\V\\r ) r0r/r1r.r5rAr2r'r$r%r&)r~r*FF)NNrR)&r __module__ __qualname____firstlineno____doc__ONNX_EXPORTABLEfloatboolr-rr9rr4Tensor__annotations__rr@strrOrTrr[r_rrrdrrhr5rAr3rur{rrr__static_attributes__ __classcell__)r7s@rr"r"3s.O# V VV V  V  VV.4ROXc5<</0Q++s+/G+ 3# 3ellu||)-" "||" % "  "  ""U\\"d"YellY%PSUXPX/Y^c^j^jYuS#X4U\\HYCZ Ju|| JEKK JTX J!38_! ! !  !  !J59*.c5<</01S#X'  tC%&S#X6 7 , eCHo $sELLGXBY : :d3 ;L6M:VZ[^`c[cVd:iniuiu:PT]\\]+3Dell9J4K+L]_b] ]]r r"c \rSrSrSrSS\R S\\\R 4S\\\ 4S\ \R S\R 4 S jjr SS\R S\\\R 4S\\\ 4S\ \R S\R 4 S jjr SS\R S\\\R 4S\\\ 4S\ \R S \ S\R 4 S jjr SS\R S\\\R 4S\\\ 4S\ \R S \ S\R 4 S jjrSS\S\\\R 4S\\\ 4S\ \R S \ S\4 SjjrSS\S\\\R 4S\\\ 4S\ \R S \ S\4 SjjrSS\R S\\\R 4S\\\ 4S\ \R S \ S\R 4 SjjrSS\R S\\\R 4S\\\ 4S\ \R S\R 4 SjjrSS\R S\\\R 4S\\\ 4S\ \R S\R 4 SjjrSS\S\\\R 4S\\\ 4S\ \R S\4 SjjrSS\S\\\R 4S\\\ 4S\ \R S\4 SjjrSS\S\\\R 4S\\\ 4S\ \R S\4 SjjrSS\S\\\R 4S\\\ 4S\ \R S\4 SjjrSS\R S\\\R 4S\\\ 4S\ \R S\R 4 SjjrSS\R S\\\R 4S\\\ 4S\ \R S\R 4 SjjrSS\R S\\\R 4S\ \\\ 4S\R 4SjjrSrg)_AugmentationBaseia|_AugmentationBase base class for customized augmentation implementations. Advanced augmentation base class with the functionality of transformation matrix calculations. Args: p: probability for applying an augmentation. This param controls the augmentation probabilities element-wise for a batch. p_batch: probability for applying an augmentation to a batch. This param controls the augmentation probabilities batch-wise. same_on_batch: apply the same transformation across the batch. keepdim: whether to keep the output shape the same as input ``True`` or broadcast it to the batch form ``False``. Nrrwr2 transformrc[erRrZrrrwr2rs rr9!_AugmentationBase.apply_transformr]r cU$rRrSrs rapply_non_transform%_AugmentationBase.apply_non_transform!s  r r;c UR"Uc UROUU40UD6up#USnUS:nURnURU5n UR U 5 UR 5(aUR XX4S9n OUR5(dURXX4S9n OURXX4S9n UR XUUUcUOXGS9n [5(a6U RUR5n U RUR5n U RU4U 5n UR(a [X5OU n [5(aU RUR5n U $Nrsr~r)r{r.rWr[r_allr9anyrrrrA index_putr'r) rrrwr2rr;rsr ori_shaperraapplieds rtransform_inputs"_AugmentationBase.transform_inputs,sd@@"NDLL AG L) #KK ))%0  Y' <<>>)))U)XF--i-\F--i-\F**#'0'8)i>Q +G#$$U[[1!,,u{{3%%xk7;F?C||(;QW  [[-F r c UR"Uc UROUU40UD6up#USnUS:nURnUSn URXSS9n UR U 5 UR 5(aUR XX4S9n OdUR5(dURXX4S9n O>URXX4S9n UR XUUUcUOXGS9n U RU4U 5n UR(a [XU S9n U $U n U $)Nrsr~rFrVr)reference_shape) r{r.rWr[r_rapply_transform_maskrapply_non_transform_maskrr'r) rrrwr2rr;rsrrrWrrars rtransform_masks!_AugmentationBase.transform_masksVs;@@"NDLL AG L) #KK ,-))%E)R  Y' <<>>..y%.]F229e2aF229e2aF//#'0'8)i>Q 0G %%xk7;FVZVbVb(ER io r c |[U[5(d[S[U5S35eUR"Uc UR OUU40UD6up#USnUS:nUR 5R5(aURXX4S9nU$UR5(dURXX4S9nU$URXX4S9nURXUUUcUOXGS9n [5(a6URUR5nU RUR5n URU4U 5nU$)NzOnly `Boxes` is supported. Got .rsr~r)rHr RuntimeErrorrr{r.rrapply_transform_boxrapply_non_transform_boxrrAr rrrwr2rr;rsrrars rtransform_boxes!_AugmentationBase.transform_boxeszsO%''!@e QOP P@@"NDLL AG L) # ==?   --eU-XF  11%1\F 11%1\F..'0'8)i>Q /G #$$U[[1!,,u{{3%%xk7;F r c `[U[5(d[S[U5S35eUR"Uc UR OUU40UD6up#USnUS:nUR 5(aURXX4S9nU$UR5(dURXX4S9nU$URXX4S9nURXUUUcUOXGS9n [5(a6URUR5nU RUR5n URU4U 5nU$)Nz#Only `Keypoints` is supported. Got rrsr~r) rHrrrr{r.rapply_transform_keypointrapply_non_transform_keypointrrArrs rtransform_keypoints%_AugmentationBase.transform_keypointssF%++!DT%[MQRST T@@"NDLL AG L) # <<>>225%2]F 66ue6aF 66ue6aF33'0'8)i>Q 4G #$$U[[1!,,u{{3%%xk7;F r c |UR"Uc UROUU40UD6up#USnUS:nUR5(aURXX4S9nU$UR 5(dUR XX4S9nU$UR XX4S9nURXUUUcUOXGS9n UR U4U 5nU$r)r{r.rapply_transform_classrapply_non_transform_classrrs rtransform_classes#_AugmentationBase.transform_classess@@"NDLL AG L) # <<>>//u/ZF 33E53^F 33E53^F00'0'8)i>Q 1G %%xk7;F r c[e)zMProcess masks corresponding to the inputs that are no transformation applied.rZrs rr*_AugmentationBase.apply_non_transform_maskr]r c[e)z?Process masks corresponding to the inputs that are transformed.rZrs rr&_AugmentationBase.apply_transform_maskr]r cU$)zMProcess boxes corresponding to the inputs that are no transformation applied.rSrs rr)_AugmentationBase.apply_non_transform_box  r c[e)z?Process boxes corresponding to the inputs that are transformed.rZrs rr%_AugmentationBase.apply_transform_boxr]r cU$)zQProcess keypoints corresponding to the inputs that are no transformation applied.rSrs rr._AugmentationBase.apply_non_transform_keypointrr c[e)zCProcess keypoints corresponding to the inputs that are transformed.rZrs rr*_AugmentationBase.apply_transform_keypointr]r cU$)zRProcess class tags corresponding to the inputs that are no transformation applied.rSrs rr+_AugmentationBase.apply_non_transform_classrr c[e)zDProcess class tags corresponding to the inputs that are transformed.rZrs rr'_AugmentationBase.apply_transform_class%r]r rcHUc URnURXU5nU$rR)r2r)rrrwr2ras rr_AugmentationBase.apply_func/s) =JJE&&y%@ r rSrR)rrrrrr4rrrrrr9rrrrrrrrrrrrrrrrrrrSr rrrs (-1 "||"S%,,&'"CH~ " ELL) "  "-1 || S%,,&' CH~  ELL)    -1 (||(S%,,&'(CH~ ( ELL) (  ( (^-1 "||"S%,,&'"CH~ " ELL) "  " "R-1 ##S%,,&'#CH~ # ELL) #  # #T-1 !!S%,,&'!CH~ ! ELL) !  ! !P-1 ||S%,,&'CH~  ELL)    D-1 "||"S%,,&'"CH~ " ELL) "  "-1 "||"S%,,&'"CH~ " ELL) "  "-1 S%,,&'CH~  ELL)   -1 ""S%,,&'"CH~ " ELL) "  "-1 S%,,&'CH~  ELL)   -1 ""S%,,&'"CH~ " ELL) "  "-1 ||S%,,&'CH~  ELL)   -1 "||"S%,,&'"CH~ " ELL) "  "+/ << S%,,&' S#X'    r r)"enumrtypingrrrrrr r4r torch.distributionsr r r $kornia.augmentation.random_generatorrkornia.augmentation.utilsrrrrkornia.core.utilsrkornia.geometry.boxesrkornia.geometry.keypointsrrTensorWithTransformMatModulerr"rrSr rrs$>> IID 2'/u||U5<<3M-NNOvvCvELLvQ]RYYQ]hs.sr