<ЦiSSKrSSKrSSKrSSKrSSKJrJrJr SSKrSSK r SSK J r SSK JrJrJr SSKJr SSKJr SSKJr SSKJr SS KJr S S KJr S S KJrJr /S Qr \RB"\"5r#S\$S\%4Sjr&S\%S\$4Sjr'S\RPS\\%4Sjr)S\RPS\\$4Sjr*S\RPS\\4Sjr+0\RXSS\)4_\RZSS\)4_\R\SS\)4_\R^SS\+4_\RHSS\+4_\R`SS\+4_\RbSS\+4_\RdSS\+4_\RfS S\+4_\RhS S\+4_\RjS!S\+4_\RlS!S\+4_\RJS"S\+4_\RnS"S\+4_\RpS"S\+4_\RrS#S$\+4_\RtS%S&\*4_\RvS%S&\*4\RxS'S&\*4\RzS'S&\*4\R|S(S)\+4\R~S*S+\*4\RS*S+\*4\RfS S,\+4\RS S,\+4\RS S,\+40 ErCS-rDSDS.jrESES/jrFSFS0jrGS1rHS2rISGS3jrJSGS4jrKSHS5jrLSIS6jrMSGS7jrNSJS8jrOSKS9jrPS:rQSLS;jrRS<rSS=rTSMS>jrUSMS?jrVSNS@jrWSArXSBrYSNSCjrZg)ON)AnyListOptional) struct_pb2)HistogramProtoSummarySummaryMetadata) TensorProto)TensorShapeProto) layout_pb2)PrCurvePluginData)TextPluginData)make_np)_prepare_videoconvert_to_HWC) half_to_int int_to_halfhparamsscalar histogram_raw histogrammake_histogramimage image_boxes draw_boxes make_imagevideo make_videoaudiocustom_scalarstext tensor_proto pr_curve_rawpr_curve compute_curvemeshfreturncd[R"SU5n[R"SU5S$)a-Casts a half-precision float value into an integer. Converts a half precision floating point value, such as `torch.half` or `torch.bfloat16`, into an integer value which can be written into the half_val field of a TensorProto for storage. To undo the effects of this conversion, use int_to_half(). r(irstructpackunpack)r(bufs ^/var/www/html/ai-image-ml/venv/lib/python3.13/site-packages/torch/utils/tensorboard/summary.pyrr6s* ++c1 C ==c "1 %%r+cd[R"SU5n[R"SU5S$)zCasts an integer value to a half-precision float. Converts an integer value obtained from half_to_int back into a floating point value. r+r(rr,)r+r0s r1rrCs* ++c1 C ==c "1 %%r2tc|UR5R5Vs/sHn[U5PM sn$s snfN)flattentolistr)r4xs r1_tensor_to_half_valr:Ms/$%IIK$6$6$8 9$8qKN$8 99 9s9cf[R"U5R5R5$r6)torch view_as_realr7r8r4s r1_tensor_to_complex_valr?Ps%   a ( ( * 1 1 33r2c>UR5R5$r6)r7r8r>s r1_tensor_to_listrASs 99;   r2DT_HALFhalf_val DT_BFLOAT16DT_FLOAT float_val DT_DOUBLE double_valDT_INT8int_valDT_UINT8DT_INT16DT_INT32DT_INT64 int64_val DT_COMPLEX32 scomplex_val DT_COMPLEX64DT_BOOLbool_val DT_COMPLEX128 dcomplex_val uint32_valc[U[R5(dUR5OUnUR[R :XaS$S$)Nr) isinstancenpndarraynumpydtypeuint8)tensor converteds r1_calc_scale_factorrbus9&0&D&D &I288+144r2c SSKJn Jn U R5n U R U5n XX$4upnnU R X4U U4UU4X4X4/UUS9 U(azUnU R U5unnnnUU- UU- nn[R"SU-5nU RU UU- SU-- 4U U-U4/US9 U RU U-UU- U- 4UUU S9 U$)Nr) ImageDraw ImageFont)widthfillg?)rg)rgfont) PILrdre load_defaultDrawlinegetbboxr[ceil rectangler")rxminyminxmaxymax display_strcolor color_text thicknessrdreridrawleftrighttopbottom text_bottom_left_top_right_bottom text_width text_heightmargins r1_draw_single_boxrzs#)  ! ! #D >>% D"&d!9T#vII tVnufo|d[Q   '+||K'@$tVW"(5.'D.K  +, {[01v:=> "K0    F]K+5> ?   Lr2c  ^^SSKnSSKJnJnJnJnJnJn SSKJ n J n J n J n J n SSKJnJnJn [%T[&5(d [(R+S5 [-S5e[%U[&5(d [(R+S5 [-S5eU=(d 0n[%U[&5(d [-S5eUR/5HNumnTT;d2[%U[05(a[3UU4S jU55(aM>[-S TS TS 35e /nU"5nTR/5GH=umnUcM [%U[4[645(aUUR8TlTU;a;[<R>"UTVs/sHn[<R@"US 9PM snS9nOSnURCU"TURAS5US95 M[%U[D5(aUUR8Tl#TU;a;[<R>"UTVs/sHn[<R@"US9PM snS9nOSnURCU"TURAS5US95 GM;[%U[H5(aUUR8Tl%TU;a;[<R>"UTVs/sHn[<R@"US9PM snS9nOSnURCU"TURAS5US95 GM[%UURL5(aJ[OU5SnUUR8TlURCU"TURAS5S95 GM5[QS5e U"UU S9n[S[RRT"U URW5S9S9n[Y[XR@"UUS9/S9nUR[5Vs/sH nU"U"US9S9PM nnU"UUS9nU"UU S9n[S[RRT"U URW5S9S9n[Y[XR@"U US9/S9nU"U RAS 5S!9nU"UU S"9n[S[RRT"U URW5S9S9n[Y[XR@"U US9/S9nUUU4$s snfs snfs snfs snf)#aOutput three `Summary` protocol buffers needed by hparams plugin. `Experiment` keeps the metadata of an experiment, such as the name of the hyperparameters and the name of the metrics. `SessionStartInfo` keeps key-value pairs of the hyperparameters `SessionEndInfo` describes status of the experiment e.g. STATUS_SUCCESS Args: hparam_dict: A dictionary that contains names of the hyperparameters and their values. metric_dict: A dictionary that contains names of the metrics and their values. hparam_domain_discrete: (Optional[Dict[str, List[Any]]]) A dictionary that contains names of the hyperparameters and all discrete values they can hold Returns: The `Summary` protobufs for Experiment, SessionStartInfo and SessionEndInfo rN)DataType Experiment HParamInfo MetricInfo MetricNameStatus)EXPERIMENT_TAGPLUGIN_DATA_VERSION PLUGIN_NAMESESSION_END_INFO_TAGSESSION_START_INFO_TAG)HParamsPluginDataSessionEndInfoSessionStartInfoz>parameter: hparam_dict should be a dictionary, nothing logged.z>parameter: metric_dict should be a dictionary, nothing logged.zIparameter: hparam_domain_discrete should be a dictionary, nothing logged.c3T># UHn[U[TT55v M g7fr6)rZtype).0d hparam_dictks r1 hparams..s$FAqz!T+a.%9::As%(z"parameter: hparam_domain_discrete[z/] should be a list of same type as hparam_dict[z].) number_value)valuesDATA_TYPE_FLOAT64)namerdomain_discrete) string_valueDATA_TYPE_STRING) bool_valueDATA_TYPE_BOOL)rrz=value should be one of int, float, str, bool, or torch.Tensor)session_start_infoversion plugin_namecontent plugin_data)tagmetadatavaluer)r) hparam_infos metric_infos) experimentrSTATUS_SUCCESS)status)session_end_infor).r<#tensorboard.plugins.hparams.api_pb2rrrrrr$tensorboard.plugins.hparams.metadatarrrrr+tensorboard.plugins.hparams.plugin_data_pb2rrrrZdictloggerwarning TypeErroritemslistallintfloatrrr ListValueValueappendstrrboolrTensorr ValueErrorr PluginDataSerializeToStringrkeys)r metric_dicthparam_domain_discreter<rrrrrrrrrrrrrrvhpsssirrrsmdrmtsexpseis` ` r1rrs( k4 ( (WX L   k4 ( (WX L  49r ,d 3 3 W  ',,.1 [ a&&FAFFF4QC7fghfiikl  / C  C!!#1 9  a#u & &*+CKKN '**BLBVBV"8!:!:A#((a8!:C#' JJ!(;<$3   a  *+CKKN '**","6"6"8!:!:A#((a8!:##' JJ!(:;$3   a  ()CKKN %**","6"6"8!:!:A#((A6!:##' JJ!(89$3   a & & 1 A*+CKKN ' JJzqx~~>Q/RS T  K  U$\ 3@STG #..#W-F-F-H  C +ACPQ RC7B7G7G7I J7I!::!, -7IC J #C 8C38KLG #..#W-F-F-H  C >CHI JC  -= > ?C>QRG #..#W-F-F-H  C +?#NO PC S=G,,D KsR4 +R9 R> !Sc[U5R5nURS:Xd"SURSURS35e[ U5nU(aX[ U/SS9nU(a [ U/SS9n[ R"S S 9n[ US 9n[[R"UUUS 9/S 9$[[R"XS9/S 9$)aOutput a `Summary` protocol buffer containing a single scalar value. The generated Summary has a Tensor.proto containing the input Tensor. Args: name: A name for the generated node. Will also serve as the series name in TensorBoard. tensor: A real numeric Tensor containing a single value. collections: Optional list of graph collections keys. The new summary op is added to these collections. Defaults to `[GraphKeys.SUMMARIES]`. new_style: Whether to use new style (tensor field) or old style (simple_value field). New style could lead to faster data loading. Returns: A scalar `Tensor` of type `string`. Which contains a `Summary` protobuf. Raises: ValueError: If tensor has the wrong shape or type. rzBTensor should contain one element (0 dimensions). Was given size: z and z dimensions.rE)rFr^rG)rHr^scalarsrrrr`rr)r simple_value) rsqueezendimsizerr r rrr) rr` collections new_styledouble_precisionrr#rrs r1rrbs"V_ $ $ &F qx KFKK=X]^d^i^i]jjvwx6]F"fXZH &6(+NL%00YG +6 '   gmmJKLLr2cUR5UR-S:a [S5eUR[;aa[URup#n[ S 0SUS[ URVs/sHn[ R"US9PM snS9X4"U50D6nO[USUR35e[R"SS 9n[US 9n[[R"XUS 9/S 9$s snf)aOutputs a `Summary` protocol buffer containing the full tensor. The generated Summary has a Tensor.proto containing the input Tensor. Args: tag: A name for the generated node. Will also serve as the series name in TensorBoard. tensor: Tensor to be converted to protobuf Returns: A tensor protobuf in a `Summary` protobuf. Raises: ValueError: If tensor is too big to be converted to protobuf, or tensor data type is not supported lzAtensor is bigger than protocol buffer's hard limit of 2GB in sizer^ tensor_shaperdimz has unsupported tensor dtype r`rrrrr`r) numelitemsizerr^_TENSOR_TYPE_MAPr r shapeDimr rrr) rr`r^ field_name conversion_fnr9r#rrs r1r#r#s||~'G4 O  ||''+;FLL+I(="  0?E||L|!)--15|L!M&1  C5 >v||nMNN!,,BK k 2C '--ClST UUMs4C6c X[UUUUUUUS9n[[R"XS9/S9$)aOutput a `Summary` protocol buffer with a histogram. The generated [`Summary`](https://www.tensorflow.org/code/tensorflow/core/framework/summary.proto) has one summary value containing a histogram for `values`. Args: name: A name for the generated node. Will also serve as a series name in TensorBoard. min: A float or int min value max: A float or int max value num: Int number of values sum: Float or int sum of all values sum_squares: Float or int sum of squares for all values bucket_limits: A numeric `Tensor` with upper value per bucket bucket_counts: A numeric `Tensor` with number of values per bucket Returns: A scalar `Tensor` of type `string`. The serialized `Summary` protocol buffer. minmaxnumsum sum_squares bucket_limitbucketrhistor)rrr) rrrrrr bucket_limits bucket_countshists r1rrs<*     " D '--D=> ??r2c[U5n[UR[5X#5n[ [R "XS9/S9$)a0Output a `Summary` protocol buffer with a histogram. The generated [`Summary`](https://www.tensorflow.org/code/tensorflow/core/framework/summary.proto) has one summary value containing a histogram for `values`. This op reports an `InvalidArgument` error if any value is not finite. Args: name: A name for the generated node. Will also serve as a series name in TensorBoard. values: A real numeric `Tensor`. Any shape. Values to use to build the histogram. Returns: A scalar `Tensor` of type `string`. The serialized `Summary` protocol buffer. rr)rrastyperrr)rrbinsmax_binsrs r1rrs9"V_F &--. ?D '--D=> ??r2c $URS:Xa [S5eURS5n[R"XS9up4[ U5nUbXR:aXR-nXV-nUS:wa[R "USXg- //SSS9nURSU5RSS9n[R"URS -4UR5nUSSU2USS&USUS'Un[R"[R"US55n [R"U SU SS - /S S 9up[U 5n [U 5S -n A U S:aX:S - U O[R"S/USU /5nXJU S -nURS:XdURS:Xa [S 5eURU5n [!UR#5UR%5[ U5UR5U UR'5UR'5S 9$)zDConvert values into a histogram proto using logic from histogram.cc.rzThe input has no element.)rNconstant) pad_widthmodeconstant_values)axisrr{)sidez1The histogram is empty, please file a bug report.r)rrreshaper[rlenpadremptyr^cumsumgreater searchsortedr concatenatedotrrrr8) rrrcountslimitsnum_bins subsamplingsubsampling_remainder new_limits cum_countsstartendsum_sqs r1rrs {{a455 ^^B F\\&4NF6{H 3* ( 6 A %VV{BCD ! F K044"4=XXv{{Q0&,,?  "[1 3B 22::fa01JaB!1C-D7SJE JE c(Q,C$)19qy3"..1#vds|AT2U C!G $F {{a6;;!+LMM ZZ F  JJL JJL K JJL]]_}} r2c.[U5n[X5n[U5nUR[R 5nX-R SS5R[R5n[XS9n[[R"XS9/S9$)aOutput a `Summary` protocol buffer with images. The summary has up to `max_images` summary values containing images. The images are built from `tensor` which must be 3-D with shape `[height, width, channels]` and where `channels` can be: * 1: `tensor` is interpreted as Grayscale. * 3: `tensor` is interpreted as RGB. * 4: `tensor` is interpreted as RGBA. The `name` in the outputted Summary.Value protobufs is generated based on the name, with a suffix depending on the max_outputs setting: * If `max_outputs` is 1, the summary value tag is '*name*/image'. * If `max_outputs` is greater than 1, the summary value tags are generated sequentially as '*name*/image/0', '*name*/image/1', etc. Args: tag: A name for the generated node. Will also serve as a series name in TensorBoard. tensor: A 3-D `uint8` or `float32` `Tensor` of shape `[height, width, channels]` where `channels` is 1, 3, or 4. 'tensor' can either have values in [0, 1] (float32) or [0, 255] (uint8). The image() function will scale the image values to [0, 255] by applying a scale factor of either 1 (uint8) or 255 (float32). Out-of-range values will be clipped. Returns: A scalar `Tensor` of type `string`. The serialized `Summary` protocol buffer. rrY)rescalerrr) rrrbrr[float32clipr_rrr)rr`r dataformats scale_factorrs r1rr sx6V_F F 0F%f-L ]]2:: &F# ) )!S 1 8 8 BF v /E '--C=> ??r2cD[U5n[X5n[U5nUR[R5[ U5-n[ URSS5R[R5UUUS9n[[R"XS9/S9$)z/Output a `Summary` protocol buffer with images.rrY)rroislabelsr r) rrrr[r!rbrr"r_rr)r tensor_image tensor_boxesrr#r'rs r1rrEs<(L!, ??r2c URSn[U5nUH)n[UXS4XS4XS4XS4UcSOX%SS9nM+ U$)NrrrhRed)rurv)rranger) disp_imageboxesr' num_boxeslist_gtr+s r1rrVsj AIIG %  Q$K Q$K Q$K Q$K &FI  r2cSSKJn URupVn[XQ-5n[Xa-5n UR U5n Ub [ XUS9n UR Rn U RX4U 5n SSK n U R5n U RU SS9 U R5nU R5 [R"UUUUS9$)z=Convert a numpy representation of an image to Image protobuf.r)ImageN)r'PNG)formatheightrf colorspaceencoded_image_string)rjr3rr fromarrayr ResamplingLANCZOSresizeioBytesIOsavegetvaluecloser)r`rr&r'r3r7rfchannel scaled_height scaled_widthr ANTIALIASr>output image_strings r1rrgs#\\F7()Mu'L OOF #E 5v6  ((I LL,6 BE ZZ\F JJveJ$??$L LLN ==)  r2c2[U5n[U5n[U5nUR[R 5nX-R SS5R[R5n[X5n[[R"XS9/S9$)NrrYr r) rrrbrr[r!r"r_rrr)rr`fpsr$rs r1rrsv V_F F #F%f-L ]]2:: &F# ) )!S 1 8 8 BF v #E '--C=> ??r2cSSKnSSKJn SSKnUR upVpxUR [U5US9n URSSS 9Rn U RU SSS 9 [U S 5n U R5n SSS5 [R"U 5 [&R("XgUW S9$![a [S5 gf=f![a [SS5 gf=f![a6 U RU SSS 9 N![a U RU SS 9 Nf=ff=f!,(df  N=f![ a ["R%S5 Nf=f)Nrzadd_video needs package moviepy)editorz6moviepy is installed, but can't import moviepy.editor.z2Some packages could be missing [imageio, requests])rJz.gifF)suffixdelete)verboser)rO progress_bar)rOrbz5The temporary file used by moviepy cannot be deleted.r6)moviepy ImportErrorprintrLtempfilerImageSequenceCliprNamedTemporaryFiler write_gifropenreadosremoveOSErrorrrrr3) r`rJrRmpyrU_thwcr"filenamer( tensor_strings r1rrsj),,KB1  f3 7D**&*GLLH4 xt< h   P ( ==am G  /0  D @    4 4 NN8UN G 4 NN8UN 3 44   PNOPsjB7CC.1D1 E7CCC+*C+. D.9D  D*&D.)D**D.1 D?E$#E$c@[U5nUR5n[U5R5S:a[ S5 UR SS5nUR S:XdS5eU[R"[R5R-RS5nSSK nSSK nUR5nURUS5nURS5 UR!S5 UR#U5 UR%UR&5 SSS5 UR)5nUR+5 [,R."USUR0SUS S 9n [-[,R2"X S 9/S 9$!,(df  No=f) Nrz4warning: audio amplitude out of range, auto clipped.rz%input tensor should be 1 dimensional.zwaver?rY setnchannels setsampwidth setframerate writeframesdatarArBrAudiorr) rr`rgarrayr>rofio wave_write audio_stringr s r1r r s? FOE MMOE 5z~~! DE 2q! ::?CCC? RXXbhh'++ + 3 3E :E  **,C 3 "" ,uzz*  <<>LIIK MMkk"o)  E '--C=> ??  s AF Fc /nUR5Hup#/nUR5HupVUSnUSS:Xaa[U5S:Xde[R"[RR USUSUSS9/S9n[R "XXS9n O([R "US 9n [R "XZS 9n URU 5 M UR[R"X$S 95 M [R"US 9n[R"S S9n [U S9n [SUR5/[5S9n [[R "SXS9/S9$)NrrMarginr+rh)rlowerupper)series)titlerr)r multiline)rchart)categoryr!rr DT_STRINGr^ string_valrcustom_scalars__config__rr)rr r MarginChartContentSeriesChartMultilineChartContentrCategoryLayoutr rr rr rr)layout categoriesrrcharts chart_namechart_metadatatagsmgccrmlccrrr`s r1r!r!sdJ *+'') &J!!$Da H,4yA~%~!44"55<<"&q'aQ=#((zG!77DA"((zJ MM% +4 *--ADE%(   3F!,,9IJK k 2C ,,./%'F  MM8 V  r2c [R"S[SS9R5S9n[US9n[ SUR SS9/[ [ R"S S 9/S 9S 9n[[R"US -X4S9/S9$)Nr"r)rrrrutf_8)encodingrrrrz /text_summaryrr) r rrrr encoder rrr)rr"rrr`s r1r"r"s!,,N1$=$O$O$QK k 2C KKK12%+;+?+?Q+G*HIF }}!6TU r2c US:aSn[R"XX4XV45n [SUS9R5n [R "SU S9n [ U S9n [ SU RS5R5[[R"U RSS 9[R"U RS S 9/S 9S 9n [[R"X U S 9/S9$)Nrrnum_thresholds pr_curvesrrrErrrrr^rFrrr)r[stackr rr rr r r8r rrrr)rtpfptnfn precisionrecallrweightsrtpr_curve_plugin_datarrr`s r1r$r$ s 88RRY7 8D,.",,)=K k 2C ,,r"))+% $$$**Q-8 $$$**Q-8  F '--CfMN OOr2c [US5n[XX4S9n[SUS9R5n[R "SUS9n[ US9n[ SURS 5R5[[R"URSS 9[R"URS S 9/S 9S 9n [[R"XU S9/S9$)Nr)rrrrrrrrErrrrrrr)rr&r rr rr r r8r rrrr) rr' predictionsrrrtrrrr`s r1r%r%#s-N N D-.",,)=K k 2C ,,r"))+% $$$**Q-8 $$$**Q-8  F '--CfMN OOr2czSnUcSn[R"[R"XS- -55nUR[R5nSUS- 4n[R "UUUXc-S9up[R "UUUSU- U-S9up[R "USSS25SSS2n [R "U SSS25SSS2n U SU - n U SU - nU [R"XKU -5- nU [R"XKU-5- n[R"XXUU45$)NgHz>g?rr)rr-rr) r[int32floorrfloat64rrmaximumr)r'rrr_MINIMUM_COUNTbucket_indices float_labelshistogram_range tp_buckets_ fp_bucketsrrrrrrs r1r&r&>s<NXXbhh{q6H'IJKN==,L.1,-OLL & MJ LL |#w. MJ :dd# $TrT *B :dd# $TrT *B AB ABRZZR88I "**^"W5 5F 88RRY7 88r2c SSKnSSKJn UR"U5nUR UUUUUR UUS9n [ SURS5R5[[R"UR SS9[R"UR SS9[R"UR S S9/S 9S 9n[R"URX5UU S 9n U $) aCreate a tensor summary with summary metadata. Args: name: Uniquely identifiable name of the summary op. Could be replaced by combination of name and type to make it unique even outside of this summary. display_name: Will be used as the display name in TensorBoard. Defaults to `name`. description: A longform readable description of the summary data. Markdown is supported. tensor: Tensor to display in summary. content_type: Type of content inside the Tensor. components: Bitmask representing present parts (vertices, colors, etc.) that belong to the summary. json_config: A string, JSON-serialized dictionary of ThreeJS classes configuration. Returns: Tensor summary with metadata. rNr) json_configrErrrrhrrr)r<tensorboard.plugins.meshr as_tensorcreate_summary_metadatarr r r8r rrrget_instance_name) r display_name descriptionr`rk componentsrr<rtensor_metadatatensor_summarys r1_get_tensor_summaryr_s.1 __V $F66  7O..$++-% $$&,,q/: $$&,,q/: $$&,,q/:  F]]  & &t : N r2c:SnUb[R"USS9nU$)z5Parse and returns JSON string from python dictionary.z{}T) sort_keys)jsondumps) config_dictrs r1_get_json_configrs#Kjj= r2c |SSKJn SSKJn [ U5n /n XR 4X8R 4X(R4/n U V s/sH oScM U PM n n URU V V s/sHupU PM sn n 5nU H%upU R[UUUU U UU 55 M' [U S9$s sn fs sn n f)aOutput a merged `Summary` protocol buffer with a mesh/point cloud. Args: tag: A name for this summary operation. vertices: Tensor of shape `[dim_1, ..., dim_n, 3]` representing the 3D coordinates of vertices. faces: Tensor of shape `[dim_1, ..., dim_n, 3]` containing indices of vertices within each triangle. colors: Tensor of shape `[dim_1, ..., dim_n, 3]` containing colors for each vertex. display_name: If set, will be used as the display name in TensorBoard. Defaults to `name`. description: A longform readable description of the summary data. Markdown is supported. config_dict: Dictionary with ThreeJS classes names and configuration. Returns: Merged summary for mesh/point cloud representation. rr)MeshPluginDatar) rr(tensorboard.plugins.mesh.plugin_data_pb2rrVERTEXFACECOLORget_components_bitmaskrrr)rverticescolorsfacesrrrrrr summariestensorsr`rkrs r1r'r's,2G";/KI (() ##$ %%&G %,EG&ayvGGE004;rs  && & <F8JC 2 0   8 $ &5 &S &&3&5&:5<<:DI:4ell4tE{4 u|| S   JJJ(;< MMIz+>? NN]J0CD MMJ _=   KK*k?;   MMK?  LL; o> JJI7 KK*i9 KK*i9 KK*i9 KK*i9 II I7 KK*i9 LL:y/:  KK*k?;!" OOnn6LM#$ KK..2HI OOnn6LM LL>>3IJ JJJ8 8NO MMO^5KL KK*lO< LL:|_= NNZ?5<5 (Vzz(MV"VJ@B@,4n"@LKO@""4@(V@<"J  IMP4P69B8vOS3$r2