From 1390187475b2567893df7ff78d001e144c332a57 Mon Sep 17 00:00:00 2001 From: loyal812 Date: Tue, 26 Mar 2024 14:44:26 -0400 Subject: [PATCH] feat: chatting class --- requirements.txt | 4 +- src/chatting/ChattingClass.py | 66 ++++++++++++++++++ src/finetune/FineTuningClass.py | 18 +++-- src/finetune/finetune.py | 4 +- src/finetune/initital_eval.py | 4 +- src/finetune/jsonl_generation.py | 7 +- src/finetune/run-model.py | 2 +- src/finetune/train_generation.py | 6 +- ..._Swing_Analysis_Tool_for_Self-Training.pdf | Bin .../CulturesConnect.me_E.pdf | Bin 141490 -> 0 bytes .../Dating module_2ndTask.pdf | Bin 214672 -> 0 bytes .../regression_test003/eval_questions.txt | 0 .../generated_data/eval_questions.txt | 11 +++ .../generated_data/finetuning_events.jsonl | 12 ++++ .../generated_data/model.txt | 1 + .../generated_data/train_questions.txt | 10 +++ .../regression_test003/train_questions.txt | 3 - .../regression_test004/eval_questions.txt | 40 ----------- .../finetuning_events.jsonl | 16 ----- .../regression/regression_test004/model.txt | 2 - .../regression_test004/train_questions.txt | 11 --- src/utils/finetune_engine.py | 34 --------- 22 files changed, 125 insertions(+), 126 deletions(-) create mode 100644 src/chatting/ChattingClass.py rename src/test/regression/{regression_test004 => regression_test003}/AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf (100%) delete mode 100644 src/test/regression/regression_test003/CulturesConnect.me_E.pdf delete mode 100644 src/test/regression/regression_test003/Dating module_2ndTask.pdf delete mode 100644 src/test/regression/regression_test003/eval_questions.txt create mode 100644 src/test/regression/regression_test003/generated_data/eval_questions.txt create mode 100644 src/test/regression/regression_test003/generated_data/finetuning_events.jsonl create mode 100644 src/test/regression/regression_test003/generated_data/model.txt create mode 100644 src/test/regression/regression_test003/generated_data/train_questions.txt delete mode 100644 src/test/regression/regression_test003/train_questions.txt delete mode 100644 src/test/regression/regression_test004/eval_questions.txt delete mode 100644 src/test/regression/regression_test004/finetuning_events.jsonl delete mode 100644 src/test/regression/regression_test004/model.txt delete mode 100644 src/test/regression/regression_test004/train_questions.txt delete mode 100644 src/utils/finetune_engine.py diff --git a/requirements.txt b/requirements.txt index 564e59b..a25864b 100644 --- a/requirements.txt +++ b/requirements.txt @@ -17,4 +17,6 @@ llama_index==0.9.15 pypdf spacy ragas -datasets \ No newline at end of file +datasets +docx2txt +cryptography \ No newline at end of file diff --git a/src/chatting/ChattingClass.py b/src/chatting/ChattingClass.py new file mode 100644 index 0000000..717ad5e --- /dev/null +++ b/src/chatting/ChattingClass.py @@ -0,0 +1,66 @@ +import openai +import os +import argparse +from dotenv import load_dotenv +from llama_index import ServiceContext, SimpleDirectoryReader, VectorStoreIndex +from llama_index.llms import OpenAI + +class ChattingClass: + def __init__(self, model_id, data_path, api_key="", temperature=0.3): + self.model_id = model_id + self.data_path = data_path + self.temperature = temperature + self.set_api_key(api_key) + self.set_document(data_path) + + def set_api_key(self, api_key): + if api_key: + self.api_key = api_key + else: + load_dotenv() + self.api_key = os.getenv("OPENAI_API_KEY") + + if self.api_key is not None: + os.environ["OPENAI_API_KEY"] = self.api_key + openai.api_key = self.api_key + return True + else: + # Handle the absence of the environment variable + # You might want to log an error, raise an exception, or provide a default value + # For example, setting a default value + os.environ["OPENAI_API_KEY"] = "your_default_api_key" + openai.api_key = "openai_api_key" + return False + + def set_document(self, data_path): + self.documents = SimpleDirectoryReader( + data_path + ).load_data() + + def ask_question(self, question): + ft_context = ServiceContext.from_defaults( + llm=OpenAI(model=self.model_id, temperature=self.temperature), + context_window=2048 + ) + + index = VectorStoreIndex.from_documents(self.documents, service_context=ft_context) + query_engine = index.as_query_engine(service_context=ft_context) + + response = query_engine.query(question) + return response + +# Using argparse to get the question input from the user +parser = argparse.ArgumentParser(description='Chatting Engine') +# parser.add_argument('api_key', type=str, help='Openai API Key') +parser.add_argument('data_path', type=str, help='Data Path') +parser.add_argument('model_id', type=str, help='model') +parser.add_argument('question', type=str, help='question') +# parser.add_argument('temperature', type=str, help='temperature') +args = parser.parse_args() + +# model_id = 'ft:gpt-3.5-turbo-0613:personal::8XaasBXv' +# data_path = "./src/test/regression/regression_test003" +chatbot = ChattingClass(model_id=args.model_id, data_path=args.data_path) + +response = chatbot.ask_question(args.question) +print(response) \ No newline at end of file diff --git a/src/finetune/FineTuningClass.py b/src/finetune/FineTuningClass.py index 16ccbb3..fe6dd70 100644 --- a/src/finetune/FineTuningClass.py +++ b/src/finetune/FineTuningClass.py @@ -26,6 +26,7 @@ def __init__(self, data_path, api_key='', model='gpt-3.5-turbo', temperature=0.3 self.retry_delay = 60 self.set_api_key(api_key) self.set_document(data_path) + self.generate_subfolder(data_path) def set_api_key(self, api_key): if api_key: @@ -52,6 +53,10 @@ def set_document(self, data_path): data_path ).load_data() + def generate_subfolder(self, data_path): + subfolder_name = "generated_data" + subfolder_path = os.path.join(data_path, subfolder_name) + os.makedirs(subfolder_path, exist_ok=True) def train_generation(self): for attempt in range(1, self.max_retries + 1): @@ -84,8 +89,8 @@ def generate_and_save_questions(documents, output_file): for question in questions: f.write(question + "\n") - generate_and_save_questions(self.documents[:half_point], f'{self.data_path}/train_questions.txt') - generate_and_save_questions(self.documents[half_point:], f'{self.data_path}/eval_questions.txt') + generate_and_save_questions(self.documents[:half_point], f'{self.data_path}/generated_data/train_questions.txt') + generate_and_save_questions(self.documents[half_point:], f'{self.data_path}/generated_data/eval_questions.txt') break except Exception as e: @@ -140,7 +145,7 @@ def jsonl_generation(self): ) questions = [] - with open(f'{self.data_path}/train_questions.txt', "r") as f: + with open(f'{self.data_path}/generated_data/train_questions.txt', "r") as f: for line in f: questions.append(line.strip()) @@ -155,12 +160,11 @@ def jsonl_generation(self): # Handle the exception here, you might want to log the error or take appropriate action print(f"An error occurred: {e}") finally: - if 'finetuning_handler' in locals() and 'data_path' in locals(): - finetuning_handler.save_finetuning_events(f'{self.data_path}/finetuning_events.jsonl') + finetuning_handler.save_finetuning_events(f'{self.data_path}/generated_data/finetuning_events.jsonl') def finetune(self): - file_upload = openai.files.create(file=open(f'{self.data_path}/finetuning_events.jsonl', "rb"), purpose="fine-tune") + file_upload = openai.files.create(file=open(f'{self.data_path}/generated_data/finetuning_events.jsonl', "rb"), purpose="fine-tune") print("Uploaded file id", file_upload.id) while True: @@ -182,7 +186,7 @@ def finetune(self): print("Fine-tuned model info", job_handle) print("Model id", job_handle.fine_tuned_model) - with open(f'{self.data_path}/model.txt', "w") as f: + with open(f'{self.data_path}/generated_data/model.txt', "w") as f: f.write(job_handle.fine_tuned_model + "\n") break time.sleep(3) diff --git a/src/finetune/finetune.py b/src/finetune/finetune.py index f9b6e7a..abd16ae 100644 --- a/src/finetune/finetune.py +++ b/src/finetune/finetune.py @@ -18,9 +18,9 @@ os.environ["OPENAI_API_KEY"] = "your_default_api_key" openai.api_key = "openai_api_key" -data_path = "./src/test/regression/regression_test004" +data_path = "./src/test/regression/regression_test003" -file_upload = openai.files.create(file=open(f'{data_path}/finetuning_events.jsonl', "rb"), purpose="fine-tune") +file_upload = openai.files.create(file=open(f'{data_path}/generated_data/finetuning_events.jsonl', "rb"), purpose="fine-tune") print("Uploaded file id", file_upload.id) while True: diff --git a/src/finetune/initital_eval.py b/src/finetune/initital_eval.py index ba87e59..a17f418 100644 --- a/src/finetune/initital_eval.py +++ b/src/finetune/initital_eval.py @@ -15,14 +15,14 @@ # For example, setting a default value os.environ["OPENAI_API_KEY"] = "your_default_api_key" -data_path = "./src/test/regression/regression_test004" +data_path = "./src/test/regression/regression_test003" documents = SimpleDirectoryReader( data_path ).load_data() questions = [] -with open(f'{data_path}/eval_questions.txt', "r") as f: +with open(f'{data_path}/generated_data/eval_questions.txt', "r") as f: for line in f: questions.append(line.strip()) diff --git a/src/finetune/jsonl_generation.py b/src/finetune/jsonl_generation.py index f546e6c..5cba647 100644 --- a/src/finetune/jsonl_generation.py +++ b/src/finetune/jsonl_generation.py @@ -16,7 +16,7 @@ # For example, setting a default value os.environ["OPENAI_API_KEY"] = "your_default_api_key" -data_path = "./src/test/regression/regression_test004" +data_path = "./src/test/regression/regression_test003" documents = SimpleDirectoryReader( data_path @@ -33,7 +33,7 @@ ) questions = [] -with open(f'{data_path}/train_questions.txt', "r") as f: +with open(f'{data_path}/generated_data/train_questions.txt', "r") as f: for line in f: questions.append(line.strip()) @@ -50,5 +50,4 @@ # Handle the exception here, you might want to log the error or take appropriate action print(f"An error occurred: {e}") finally: - if 'finetuning_handler' in locals() and 'data_path' in locals(): - finetuning_handler.save_finetuning_events(f'{data_path}/finetuning_events.jsonl') + finetuning_handler.save_finetuning_events(f'{data_path}/generated_data/finetuning_events.jsonl') diff --git a/src/finetune/run-model.py b/src/finetune/run-model.py index 33fd49d..a938b6c 100644 --- a/src/finetune/run-model.py +++ b/src/finetune/run-model.py @@ -16,7 +16,7 @@ os.environ["OPENAI_API_KEY"] = "your_default_api_key" openai.api_key = "openai_api_key" -data_path = "./src/test/regression/regression_test004" +data_path = "./src/test/regression/regression_test003" model_id = "ft:gpt-3.5-turbo-0613:personal::8XCvxg1X" diff --git a/src/finetune/train_generation.py b/src/finetune/train_generation.py index b66f444..907d64f 100644 --- a/src/finetune/train_generation.py +++ b/src/finetune/train_generation.py @@ -15,7 +15,7 @@ else: os.environ["OPENAI_API_KEY"] = "your_default_api_key" -data_path = "./src/test/regression/regression_test004" +data_path = "./src/test/regression/regression_test003" max_retries = 5 retry_delay = 60 # in seconds @@ -54,8 +54,8 @@ def generate_and_save_questions(documents, output_file): for question in questions: f.write(question + "\n") - generate_and_save_questions(documents[:half_point], f'{data_path}/train_questions.txt') - generate_and_save_questions(documents[half_point:], f'{data_path}/eval_questions.txt') + generate_and_save_questions(documents[:half_point], f'{data_path}/generated_data/train_questions.txt') + generate_and_save_questions(documents[half_point:], f'{data_path}/generated_data/eval_questions.txt') break except Exception as e: diff --git a/src/test/regression/regression_test004/AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf b/src/test/regression/regression_test003/AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf similarity index 100% rename from src/test/regression/regression_test004/AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf rename to src/test/regression/regression_test003/AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf diff --git a/src/test/regression/regression_test003/CulturesConnect.me_E.pdf b/src/test/regression/regression_test003/CulturesConnect.me_E.pdf deleted file mode 100644 index e5c9fa6fad1fdaaa940f99ffbbfb927036f72481..0000000000000000000000000000000000000000 GIT binary patch literal 0 HcmV?d00001 literal 141490 zcmb@tV|Zmvw>BDdY<6sQY}>Yz727ttgN|+6wr$(Cla8JK($D+s{l4ey=R5n)nb)<} zHP@_BqpI$@M~zxF#~_s#7NudJWrHK#-J6<&gJvV3C$KfJfaB(-lQy?8)^{-XFg7A! zppzzGU}j+Epr?}~;Gh#BU|?cnVBnyW{VY-bQ^H8kL?=$5MZnC?M8Lqp4B+L3Gqy4M zqln;tTyTsG|87Ls*wEIWbthJcBUPTAegm`+{M*o00^i-F*C{GX$D zbaF7(w}yi@+cYpR(nHkKGtko;1&e`Y)oY{+Hc|zTIcE|eX<`aCS_dZ*q4Qrz_7`td zuXaNYuD=3{4TsqKvHu%|8djP=uIR^vkzPL5bJmaRL(}g-pg=OTuta~b{kw&~QFe6F zcW`oZFgAgsha)8wkrVwl?r=*(#63Bs=V89?*s{DGp|2M_zMBM6n6BX-cjfc%Rq`hauVqPi(lFq7v? z2@2^Yfxln?d5}S~|Ic3k!SU~o$rwB78|gdg6a3+ZBAvXxsqrTb{<{AQU!l)3D_hfl z^NQo&b%ktgoQ!Rp90?fyRawZv)=tpYO^g0hgq4AjfQ^F@Kquy4>ug8B^e>e%pXAd2 zA0nSD{8aOItBS^sw$2WQ#(&e&)ReoIc6>sX*+H0_ML&#r`Pxt4n`4{yB_4`WbaU%l`~d@z4Dy zqyAF+DIo7)Yp7)Gq(%2>vvkVFZccO()}J^D0qE4_3@nTd|J0VSrYB(d7e6JeKb8N% zK#}+J`eg2(_6g|!#n(Ro{~Jt(KVtuPiiny!I64tChp=-L06KJ&j+eYUOSWc{C< zX^&sCSmQ_hscTG&61s9*`9%P6jiY(HK4K%k|4FchwKA9*O5(@M#cS32;XG=(&YfU* z2iNV@_<0jCseNpyu&`oWEeYI!dJT!R^lVIZsFCm}WGq9=aLn$nUX;4GMa}d-GXHwM8@w2y_TIctxjJ%ZU37iMhxDmU21E{;7h z(xUf+;SNXrQi(oP8t1J#swn&XvTpLhK9Q@xdq@!C=mGs?4Fk(Bu0g$zs%BV9tdMUh z;KnWzW=bNoATXSubNy=4W!d-}zn@mycQ?0h=lR+yn3tE=Sl~xd9q<?Ln1EszG_d%?xWL$8}NG^pXVc=j>bLMym7`Zt(T& ztMs289f;#*sHLAeO;~MbYb)jklgg}P-%#~{h+$`D|L0ou*V^)bH<7=+1p_-9%fBr_ zJfvQ2Ep(q1@g0qau3nUFjFI~1CEjCfd*DSA;tjnNI!GJg_-~*oaOZa+up)?}UW|MC0|?#u zD*OlrR$W?MUR^?6p-zKN(RPF<-kmFncKqv)l&$RSAfp0HIjYXKH6W>euq{Kh&8>K&G(QwdxLMC6If><7|zMN6ZteR1^6fuTdVPPH#2Z`S{*Cm(0%1oKaE)@IwW-yjYwOls=h^p-A_=z$9SO3I z2j2$*cC@8#V)F*+Vz~w>771(`>ou?f=>n7fcM#e_vY96n(xL*mn(5L6X!__k&Rw)a zf2RFxZ}hFy+jC2dp-J4jHY)I>GD+2uU|A8^4+_`i5Yg@U>*ruC=tn=UAOnPg0&4CC zljQ=eF6V_Lu=xD$U?Gx3;xTxKB34+lq=D8PfDG#N0r3(7ok3DCSvvdD5<*}pqCz;) zkHUPyOZfuE2A5O>R^&>gFrpa)>nkpvL(OCa|8G8qBo)EAmKOwvA!~-Y#Zg!z7QH%| znpC7;XMQwM*=}j6XVHs`QQ%tIq&K0UWlI}`s!#1VMmLgg>S=HWrv;Bcb5Uvq&?!^Y z+YYA&=2vYx26?xgh-npEa1OytsYl^rj@^WlQ0rD;~_uueD!tu9bHBQUbW zudhY$xIlKyuc@op?^?Q8GL3OX#${QHOk)q@kUFf$1gEuGF3$~A4=gJwQI5Q$a3O-u z&(0=xiQ}E3ug#l`lB(kG5%QkPqxrfjl7FDkk5srMQpQ4M-cVs2b70u&u&MzRdD+lh zXLcaPk_1Ed)q{RrFj8fikP{v?lqF5qrA}g$2+9tThoqh>WH<(2;^{G(WE7t>q&K6n z9wAGM!}x~N07N8mF)+}|UTnlko>#`17mEhML*`D#WyWB?Wm>`@+5uUmnyRLzG~4?n zlyHP|TRoXZD3>AIp||7I**{p*P#a;^9(1#dRiCu2A8w=yG!%30@@q)sLRY*^i8>mQ|a6ao;kVF5dAmG67B5|1|s|-1%KU3 zad1g-lpyi`huy~t9jNa_`80yE$OVv?Dp(}}2S_k(Wzeo>-=#PA;$eDXekzrmVv;Hb z0}eS66;_K^!U&-XH$>Q9v-P~fOf4+hcQ#GUu?h3Le_>zf>xNF@K#488qn`fS`ADuy zxxQrH{>{RxsDik}sJRt9rBzQ4DSTC}&1Q%AJDYDFCj#~^q8eyUDbZ%9u0)pDEHrxy zb&jh~{35{3gS62RZYT7%5$5TpS0U?`NEMYFv+s6Ol!P4*=u zFMnD8ser3{qwFw5-*tvLhNkf)zO+P!WA;@XGi+x>uyL7h`oId;!4y`rT*{B9M7a2&2OA~zc-*tkuA3zYRFZ&s$?qXHX$@M%;!Paa^JN0($)Eb*Ih zlr|FnHnC%vUwHPchHcfX(roqaJ1+&2b9*pu!#8KV9&X|r;%&`?l(jCtd*?ElcZpCA zd-v#9$$4$H;_|OqU^{ql@w%sK9W(Q+_E2`nwGN_)MDteD67!B9Rpi6Ulx1V_&zNpC z`4(0(kHxeA@LW+Rm*TXrsK}Y_ueeASqbFl^-;pcUxm8Dn4SG|zteNYyx|tEbJ_w!g z9{VPPh3xPyZ@~@uExySyF4L`_U%>R$)dU`ZAp9m*wP2Jx_B*@@inb;kF#w?s*_@%npn%Q3wl~jtsgXjCtq~E*0xswT zB%-MZ^J}cdE*yq1DB3Hu7oIpKdVMrGlM)!-36NY*xu69T=u-fP*v2Jnt_72+^E)*3 zu_7>Zk)w;zUMyFQ9g5SHxDR1@EFP8)VPjQQcB<@8@^9CnVacG1Y&cXz*D+bp84{X# z5!soZl4mh8uSP;|B>=kP$|jly7pTl_$cLV8_hChP+QL9ysJ)|$#Y&T}DOqr((oOmv zM8JvLh9VdNXX1o$fl*1-!G%EF)TZM7lMR2jwm>rht8ohVGl&rR5lFVUMK z1!~xYP>7P>qk)K6g_ZS0kdT03hKUBi)YSOVk&p&@iL$?kf*B#nFAk^(ix`vN2t0aE zx$||r@*Sl-Pw+fWt{?H@A0hyi;1l@8Bn+ec0_6sn>F*#huPc#|z;}j@NXmf!ra8v$ z#Mbd?mHG;}7?|WS=XT1^;}$Kd`hL6$@zF%GoUV+1!^sw)p)fpzR4Jm)NlDR-w&Mcy zYd`ra`8RGI%8C^bBwO@b2ZOFpU1+JvFENw-s> z>fQPSMC%ds95TZKr@^@`=MrCoZzS|{aa0*~`=KDU>I_oE;Y+Y~(8k_<(FoBnvr|G! zz@GS|oXE!VosogAaez{DAXNL%uD+nEpl@Et0Xo3jp@O?dJK^_PdUPO=q5dm3HW15x%pVDtMTe2Ab1+x_r z@WVzBbQ;tQC-ujNA8#*3GOSb|Wi`ANER&w;G6cLITQ`FpbUM^g_v?;gGc+$KTc6tY z+AlInv>|6z}c~)h-(l0837QziV9`ddVhQ%855X|T@ zB6HJD^|-4rxZx%_O4Do8V8@J3G_9B$5uIUpVpBO?dK4JO;e>8z zaEBm=SSFB6A?`r~2F5k5>XzkSS#hv~HvDM1@%mWxxoSX{q0YJCgOs{4chBuR*~w(T zj}71sWb^YZuNB2W!S(<`G z8D$mH6*4YBB$q@X$xIHGXb{O2St=x=-?Cq;iHq>dbV>|>l<9M=fKuq_2@DEvSOvH+6&gVHI5+BZ3htASzzQG-nbtC|ag1A_?z ze}k%isNL1w$zALLrU9w`sjw}itONq_WASE5deTOTcq#38jF{x4e92VF7fCWnbn!*0 z&G_`hUWpU&7O9#7hkRKHKtgJEZgHxz8f92(n5ImJXh!^?JX>`&u;xn2w0xb$pSr_$(y!6oYIvg>ZEbYAY>~+8P*!DPXFSxFa3>H_WpM@fVPePhrApaYuG zqIl1QBNb&8Zk1S-$Wm>U6qN)OXT=C58r2K=428T>j$-?T&;{DEj_i@?baSYBMNKwq zS%Aid6rfnAWV5KVIHwG$Jf>nM|5`n+cw{k-43 zp@vz4zJwK%C4muOoNB#l@{*MD19Pl={XCOnV7$7{0&1jlE|o6Z_}tAE{z z7a0W;O9THoF%~=)F3vG-NYP3O zzG(c~`-!DQ{`BX(>JR1{;uEw}C|8)FIh@(ixy_=W{H&?y@onQf8`kA^+gs&AvBM~> zX)SfF*4p_x7n2BE=S|HPxK;Fq3d^>`hxCg$E^gQPOOng*t!J%92^WQjH9$_=s)MqF z%f9sa!}>-vRpQtb?uud4QNPZe&gEvO_u*^gZXApDvlD_Jz z_#ItU)1|+jfX{(9f9&{2LRf~8gfWJZ!B=q1YfDp2Qu%ng&F!R-qLY$GB1S<*g(pmO)7jMB@BIfQOX7~2 zT8YMQzp2@&xLgEp3~1br`Wfi!XD_*z7~JINzFN&m~f=2p)su3sQ~?|1~|8< zvuyiSHd9r!wy-uvH=<2>PT;cdL=Bj1=iW+L=kO#w92H(Fteh#+SG#h4b$tZfwz%xI z4&9gWO!6ezE^nE2nS0*u(hOE4uBkSqdD3{&y_r0(1@e}49&#(tbwb-JHc|o z>$IG?JihQ;c{T+t!o$HrAZ+1l@G-h%dHcQy^!hI$%;C{SUb+Ri$3Lp?+K-?dkgLeu zXR$_TL;$c{v9G+*pRAnxSgvm#=ePJ|4!MiC=~5KaO4C@=Y_hJCPkd`G2lmJIlfvp+ ze8oQ$t|hM)=zKZfSGc~ivOTB%_Bg5XV}rJp?P_paW4U7cy${%J8whNFgub%mqvyTz zf_-OiNdAC&EV--uIBKNh)y1PBrKRyc5(sbK0GO*5vZ|^&F|~f1h(MV@C4)30GCfL- zk0#ll1BpwahL4a%NlEaXC=$j6%Qj{MxVKm z`+67wmJZvq#2XjOcH6>G1Cq)n*S&L_oEw=8r7IUicgnyLLbnZ#qBf04ZVtt7QQxw1 zlRq3RHpmUo^vDARs>Jv{6(}VYmLMgS6{C?Y3Ve+E0T%%|L;Me0$Ui5n|LYc#iIwI5 zy@O1M*`o4Sz!>6VCCWnzSsj>dUhSXX6-^0 zY@KpR66ssZL29#pN2aTm%xIt^=hQ97mSbko5jO;NPk!vIYPEIbaiLAZ7NUXUnw{h(^+hx_k+t3nNuqWl zFUs@9eEuCRrKtGlfql1>F6l+T9}Ly?kX(Cw>^J@i0g1?oNqdr{j*!W zHX86wAky1IBdB}fJBHu&Yk0ZcPU3KV`yZ2bO^3zHiBw6W1}W{4jqJlx5g6qJ%aPstTAx%H%og{1L(ujkj`o(5(N4d@DI;U{J1M&pPRC{U zjL@B@P2kmq-2@Qg@{S*qSd6axQ&apiW-jfUle!%c5jrdIF7|W_umnuS_joXT0&4-_ zJ9P>mM)VgT4H$R+Nx`mr?Vi&zs~#APMnv#%^=;rhbc64uffRboXf@qrvfSuR#AGWN4XcU9NY~#-J!f@)%_`xpQ z_8_q9^+2(k{V3m8K~Dt^qSouyRnG=qQYSd`SoVWtH|POvUi8DRHu~{2rT`=Jy-AY- zxU!A_I8uYa$E5>o*Og5j5r|FG3>2q{e|IM!kZ3zU5WRbi#t9VAK~@Z$9mXI4*X#5~ zR!qwdAtMhQ`f^nbRhp5z$xqfTHZt?m2I0->dE<7s+cJLQ0cH2smc)c_LVDL@O3Rws zS(%Mk(L0$xo0Z6yx5o?dPC`ud8REu6AWjOr&*<&n8(SU`o z#e{bN%5aF1`U>HRT@gySsl{ooVpT7ygkIxc9cm@G$9PY;puA6LrFv}$oxTaDex{&L zER>M%3)9xhF?Uj{PvGovTOB~}jq(C%a4*mT_@2e=AGF~N!XnohcAId=5MBwXB4Q<| z{6kTSb-&tFxe?5QcyAGHe@jI4xTCLZugLky=zhNy0)krgEQWnSS}+LTkh$+;ZuiNrZTmzsU}Q7v`e^ zxhr|Z7!*_G;U`E>J9#m2MRbxQ;eKEn18Tv78zc`BG)v>p&~RLD#5LRq2?&qi8aDK5 z+%7s)%@$(RUfVo1UK1gg_OTj=GXKH095@{Q{AJ$oEV5Gd5wp^yni*!4@Ifn1W)@jq@5VvobT@c>msuUawrXd@<-wO9mN<^v803fu;DJJ2gLFvf-DGfN zvO#_d$qvj|(6e{XmIjfOOhT;8oZ1M-wdIy8Mn`7p^-dZy?U8ymOoy+Buh?Dj-n^LK zW_rYp=q+@1JV%rwjv5NRdgYubI$k(!dY~yx2aWo$D0J_jF|Y!lFlyn-7|d1}0|k91 zNU92RbkqDjRvq>-@`A<=`uYG32A**)tV&m>rKTQdBbA&kJ;8fMckbl}iWPffT19fI zS{^T2C&aaz6#EbHp`NuVE1k2MIO<q;g7wX zmd5Dg4`yZO33m}tf^*ao?JO+K>~+KLx9!ac)?}a1|3W02iz=`tqrqw!nVv}&k-%PD zv(o#TLqA|JvT$ybGW_fT(jzFQ=h)ccTG(V2*7G4hRG68#KDoKe z!u6!e=HJ2(3zf5&`2E!AUY<7<3;y$|S+$$2~OT)+SuyxS!HH zkngHGS1>lJ5SX_=7dDcpV1?Ue&QvvJ#4_Tpo65~I>cG)&4w|}h!LD+P+VayUC0bKM zsGK$)Nb}Lr_n-JpTv%6z+O*(+h58$`y-AkXt(zlmdO~Ell4L3*fEKlh{~0ooU8hIi z{(i*5X+&Hcn{t!0?UYH=<6X2NTdges4_j-~0`Jw7DP`rObH&PKF?0bu{Q791RMNnC z>{yjp_1=qzNnW1EBn&FLT-bTSv+b}sy zbe?Tc3LWe~yiQGz%cvuPoOlu^npwufALmz?Y!!8hjcs?^4UK=#)9@mG337iZg=dIb z!S}JUUtJiS66@)#>*-Gl*x)bW(bH@1~O$z-q2sHoFdVzv5x2Y!m+G3j*Fx!9RKO&cjIVRDfNrswJok8g0+-Q%jE zYeKG-b%|8zLzHB(9?zvBdW;ChQr(s3cHDP=G{tH}CQzy1auKZSa#Ytl0bx3^>0shw zW0P7lvMHkRf_T6zoLs|rc=eNyVV&gPw(DqG`V3A>oG$-th=5yJj33SiPhCn08=JY{ z5kjKLg=&|(esp=cZye75(WD8*BN~CScFLpEwk& zU0>XF2owf&%Dw>kc6b~BuLMPv5~P=wyXhzRm6bhlP)B!UD_l(HuZN^8z)Jv*!ZMOg zI~IGQBshDoT#wr)H}*Rw;Vun*hYQ`=tKI>paFr$kjdZVDVD8dN=j9y=2VenjX5EWV zYJ7qqQJDrAu~B>2Y((*Qx7vGr^d3zNRaj<{P`fr2o7L$Y5r&r1RU)hbiJbyvBUm{S ziiX-u$R+!I+tHyX$V4ckiSH&oF)5N{Mb(Nt3_Km&l|!x0`d#O9Xm12LMb~IWU0k^f zjEJC~jx8OCP?)E1b5;@-$D3S)@# zO(f?-aj`86jc~j2$Ed`@Q$wx~=B%Ur3zH^Np(4g!M$uMyNQrUp(Q?cf-DrTT-qNl; z6q?`p+z!O#tSYj~Pm#78u5taB(?;q%(?6VswZ_gV(Ij%I6J3nQS8D^>_-8z10;?^= zl^Rc2oT<(dyyDB zVjLx-9o=p+a~w9cm2jNKDCw89&tVi}62gqU-OQ<3b3BD1=BOq8XS8W1B?4=Nx17j# zT(WVh`k7%lS{R&H6a{uyOcanX2Jx#NBVO0$WIEbjLXggoKT0(t}+iSK{86lD&pWJ@*dwHv0#0JHS@c417B07=KoH|^%+MpOf_rc zX^6@U^kwH;TiA|-H(UbD2aC@P9?NDBJt{<-W$m6nPWU7yhdQ~3tCgWch(j0*-QeZ7<5AK;mB z9lR+jI1yv*rPS`uA@}UlVs2^fmPRRy9={OeP<23A=)dcNOE}#&ZNquRfx(; zcswY#_t>-qN1=<88ui-5(+o^VSiqj-1AVR3OA8nYHW%XH!~*GsMR*R>G`-z8VfqenLDkfiICG1J%v5}Ru>@*D4a z8@iG^V{c84p~nQqfkC#ELF{w>7Pc$WP$4TxKN-rBxkVNFdWHtK#r@=Eze-wUBpRQT z9|0AoH*%mUjg8XVW3c{H=!f8}0F7NATN_%vycdL>9J_kHeVg`U&0A=^L1>|Ibtn!X zCV=henT+H9oLN^)0PjR=v`M{B&vda>{RjGC^v1oJoqWkK#7mV16Qbl10ppEA>Yt{R`iHU_2k+-0$+ZLDR5T8@#rg9rwDWgvZ5A z9V%*B(hRDp)^QJC)Qf>E!fQ>^PLXtL>ngKLJh(0)(ywiMtBSv8na^TT&7W>A=MVC7 z9TsCc->)7oz3(D*f+k0_4W27$v^G6!9E z0a-j{VyOK#G{}n^A3i?1KO*HhJ{zUoJ~sy|v8lD1c101`php-?KOsCCW{(DXhEKBeu^CsI@eb<`MF4&q>@-j=SWvj*AUcAXc`g}LGBtjSa zf!WQbwk{bdne(dZ@O`y@w~pC~mwoL9CXNOCWvA>}jS%p$k;@rzxBdl{w%n$((*plB z7HUhsQ^eO(PmlBo9_yJ2rjnkrQ7wZ7qa$@SgLW$3R3Oopp4zLv)09jktuoPud#hGe z`f^y>MtYg{wLUipgN%b8Xg|YsW z)6h=0O zzS^XeZq-(0zv6^SJ*?F1mu%EqzZ_x}FF1m%e1G1@cZA-+BC6Bi2}=4^#JVs1l51e* z{r)SUJR&Q(Gz({rCzh{{DQ+X-aaRr)FmyV$w>ReOR3xy;AWq%U7P>z=Ojicy1uLre z4WR^{KSmxYW@N2=6bNk^3=I@dsd@0sU))HGCw7!&5W^{LIhkzTgBdr>%%u2QU+0&_ z(+Szljq@eVaX>R;ttiI|yM`^#G|W>R2@F2pRgfyX`nBdfD7LEsEDD#dWWb{0BM43H zreLupQ9`LGSk|Z{^xj+fyjw`Sw}4@4k}+=?SFg5tC9{_mk&f+-#=A$5(u@Jh#)@QH z#v+NTQP<84iNt+ac|Z4215==PLP*)`JFlqz0DABlteXUEd%xS3N=I>N3rVyRla#l& z6cdwF!R)NG4Ok9n72%HT zgG_^Ee#--CG_j9TUoRUFM;MRcwz)+VASR34vh`#JC4OmS8gw8ts0&QRCD^W49j}L_ zobT&75`JShR=u;QZD=QLcjs+#KSrsfG3rne+4WUQj)IjvWx$^*_FL;~aA`8C#2C%O z`}lmr_I#_^px3Fnp>DSyT3oFoiJ#ai%*!Xb2z4Hk5|Ge z@39Ux+MX}u3j~_(JKP892Y0zA(}D=vh;FM?~Lm5NUV$i+?!M{1XfOzaD8abNokyw%aNoej+lB zA2HyH8*FH!54tcd3os&Bd#$+}jXy8~s)xZP&Oi{3;PX1JEaL?k8#6iwnA4qNme@E? zvsaJ18k@(xz{(1$(LJ|rrOSwrdteNYiD|}fO$fcZT%A-<^ff=Kx>z7v^Wj`+Ds0;0 zERSl^y=S<*SOH^x-P>ziwPeaHnjm}Laj-qRaTDmZys~|$ zu8Ib?4$p^Q7XkK@?40-hfwd_EuzGEfCv%!<;b!kwgpn7$nmiafM`wmI5`F_ z*iNze4YgO^NGBds!x(&Syvr4k%tu10ezx{kDRtUUwYI3ZFESSGTJK5R(HOtN@5mJm@0Lp|@+Mo;#{RmWq?>4yfZg7{HuXG2EZY$eTG7XLu%-tSvtkU@v0_X)vaTYA zu~O^?z&U|sHuVt8uKpqr=>bt1ZK3|YWpp^)GVq>64*vz>130p+@5|ODmt7pd_s!u! zyKMb!vZd#mmv=19O7Gj@WqouXN5q-N5jfLaUKbE<5G;*g1}lM(s(!Ll?s8SOtj`bG zR_rC(T}|AdI|#+oJcrIqmsF`KAnIu_w7F7IM+ZN&i7V>VaBORfVdeU`e!+cQ#Q^qc z9MTj|i}vma0IVCs1cvE4VqL<)z|xON$RY}m=1>jmIX^>;;sbpm0)E+~9=9jY43`$N zxLTqm`Fru>PMSmHAWLe~L!KqR=c#t`>xjBPp zNn));WoSsHq(rrFI;vco~PwMWc252EF-n_mN zRHG02959Ct&x8|MTVIJ4OWa0^+A&=H+3a1)^)=JnezN&yfdNfi37SSf2B+5J@K<%v zhv~3Z1%`b3LY!P`fX4c{A49)=ShE8M1;8yp`mqSIS|gg9{P$S%h`OB!(yR)h|ZcG{MyXW)C@G59S& zoPaO6?A(g9B2-Rmoj;$5S#M)oxKdX-Q#obLZy=#f?D(2GQv8!o{*RQne@LJDdvRrC z|Btx;PoEq!VG}@r8u-WyHkLB*WoG7tWYDNIqhG{VaScMgSu-~Di~Y{pBS>RdtoY4A zreic{MQW?U>?|A8Ejl%3J~DWsI}46DLY^{`RB-e|Cb zqGUiNya?9hvcRZY_w+bf1eN=nM|UI3!udmd4HLoD_1-jz5>4Re)bI>dU@!&Z4@!ADkx=N$UU0&SyNhw6+tHvL^QB$KnoD9APie2dQhe= zVG%_7A`u--(jQxZCQO1Xd^73k(5Ztjz;tW%$m0medKQQs8s~XjcjUPpcaU*p0{Z{; z+6KpC;T|33i>tS_c5)c64@faYkHTkIaHFq(%C~iWMiTop^l0ge+!gR%9mIQjdL|2)lzc0*GTpOuSh!cYx8DiX^F9Y;5cvaCiZ`s~_h=C~BfEupV+1 zHJj<2Uz~88jCwDXUi*W9+Bf`W!Uo${%y)s85j-ifcO?3*OyL9Qvd&SKH=uluQP&L> z*XxCj82j5m$Mu%9)Snz3*6SyGcrwgt7e~me*Pe{yX63HF&iEm{ZjsgIaala{;RQYI z=soNNCn(`!-89Go1U*D6Gwqp=Fhp-g-Q-{O+%2@DrV-Z7YZpr_#j8ZLd|Q;fhxm_l zDMqc^@*pQC6zlLGmLp#g%9762XS}WOtDbVy)Voy7Ls+8l{L0oML)#MgY?6$`bl*`djl*y9~_l_TOLKUMDEp zomISaI`)=A1j{<=e_^#6^$u7j!I=j~6@AoQOd=ul`7G~&+(^NYQD>F6d-t~7Mfw&E zFEE9UNp|m4C^)q(+X^5ykXdGQewfcc=i*$X37y_k;{AND9x_+O=b-z}%NPDmsx6QC z)D9LA5kcizVRN*V;?u{og~P&U9J5JPvF)-fl<%Fz$_uD948*{2+zEOB2JFrvSI<&z z+6$R__Y^GXNaKsZkI;~hkjc8UF` zkmKTnwS~p3)g>-g*C$AC&H9;8Z`BTJ?I!(r8900CZ7*^!Bv$6a+M0qUm)GmDVFSjb z2P$X_4GT0PTfXW5RTZCmRRvl5%bi=yT{M(h_Hwx=pri8;Z5~|hophf08`l?ilwH%s zRF28}p-vP&tXHY0oB3wquaSp|`)C`q=m7*~JI-&2IM^?54-n+DZ}B3lh@P{Z4YI@R zYMnZnmn$Sm$;m+MvCjqsvK3KrW9%MD1ImOE9 zJs$5TcU^K74+C^m2r0z9kd~cmzZE5)4Ft0NZeOvMw$<9S-yIX*Ka5;avqb|N`nl>n z5hOGlQgf1DM{}jh+!D1W z4XJ4*Z!!#43yDY2O#+w3bE-PzrL$?G9sUO1arR*AwRksPM!8&I#^AL0wb4>CxsXTX z5Bx`)I<`NL=d*m-)?Y509o6XE#vF=ZCLc1`rwv5BzJ6gOde|3Aqgi|$Wq^7)YRI(Ebb?7RXXtm!As0Ws*>7nSqPJp3?|FjLi{&=e7dRv;6Qa}|&17mqbvzCPK;K^wP5mgzv%SJ;ePxLbA2(8cO0v69T z$w!7z*{mgyg)d2EdzbMPm7O(IvQYeKMq&u0D`D6PY;6(Vx)p_46Ca}o zIZ3P@+kY9=<;4Qlmgp;$?@W@KN^&;>FpLo7qWE*c`X`TSZOfJji`doNd5(tz^ITGZ zCxFf1?LzQpjN_Lu!J3l$5I$W{i3+)qee&wAxy{Ic$xrrVZ*>{+p}`!_I`{u>;U1?G z21JE5`PrE>@i9o@4H1<~)?c8l`1RA+C#-VHZOyO9XwBXP_mbS+C=ga%GnV_3@E1Qi z1^jbfmvwv5O6gtqK(MWwwKc1Gm{>SDM7_k}Be51q^G&77Lw$qs2j~W}`azSd-#e9F zHEUdtW#QIa0HCx5CQ-W5)w4X1Oym0y0!6|G4bA1kJs&$*;E%5k);KQN9$n)5Jru*OdIox%g~`Jz`4H5u zF*%KV0o$vCnbL|(X?Guid6kxL7buwu@Eu>Pr*C4P%EjCTRLQ^m!WTOgJ6~T?Y>aoEmGYK?I%9ImT}Oi8SQ%wv(cnrh6Nla`K}^B` zuId(r^<}F^*>E%x7tFiMn%Q>Kt-TpQ?%TVI$2JSjjA?r zHxRGcth_@_I|Q~O9OaW16vIf%v;>D2IE-g?QtR(Ri5jtfZu2s!`J^?_ zNVRgaRS}3ATN^c91P4d6d3@wz0beycF5?3z1ScAE*9M>+mCoU+CyFmAX~sni?UK`CKqYh;S9e#)$V(@y>(Wx>8vh5svQ_8AU*E6SZDG1 z3qWY>_B>>hlTidDs-jW!BnP81H6*Gw(P`f7uvO!RU8OU?5%%Tt8i$Akr7rA-@fa5C znbfC!1JrDOzZ5CdCfO8k4EF`%8}VJdqknNnuKZlZYi&}R85OPV`*S$z*}MWuX)@kk zgKUk*a;rw7I55XYC{XR24m0oUU$cfzCJ^!tqY`P(1TH`y`|h23G&H#8N z_5GRanpwTNt9x}<-Boqhnsn&;^qGD$G7GG+t+2`$0F!n@3wM6{2R?bE=90lpNEy$o zG5A3yLHKpzY=Y@D!?V?rg@HMl(VY)t$q#~d_T`p}3oCL)N5|@JUacTqyLMknTy?Ut zb0)-2sg;zyQeN%{%`M~AemYz_7#OpaX)5@I+v$?tNTmlRot-zY$5=Vb*>BuL>2L~h zx9@#5Dw|#IOL*861IaGIiMJ)}+Hr!>GV0#+0)yZCId=H(D_sEniGRK2Rxb#P^6ph* zGY#1hx_?baM+;)F;AUfaJyE|dbg9{ z^|Tsd7j`5-NW?+$(1Z~Ivxa3KCp07VJ&41EWPb2yU*;lp5qPiAckK2$E^^w4o19f| z0poJ;@ddc4U0-5Jg2HiNydGpM&00m1sj7lxP>+?tvLH^GywiKHAu{3dGioe8ewjit@!Iq6d!R#AS&p3=u&|h% zzBn!V4TjmbYBN{NzeLh9vyav>>Km6~<-u@HgF`LR9A@Qfgj4-k9^VOAn4`J3vXbXh zOU!BaN-bLar;UW(Id#ch=Z5A_-$~9*suniSdgecvUY8M+{2+0<30#JL-7TuKfXdn3 z-a7d8QPuD8MUS|8qoH)LR=seZ-q;@ld1GS}rfQen9wf){Td3upoMNI&y2{r=W%jW0 zUTSL*d$7a@rOS>vFEDTK7blR$8gh2C59D;h|Ct>C8g4ICLWE=to3-xNYfJR?>&8m$ z`>a}Jy}%Tmh+3{O0o1z_@WmcBa*W>U`k-O-kTkS!DRigRO#=xKKNO9v9hIchN9f1jul0zb(t{rUo*Ui`-4%y%qNC zZM>I6%w8JiYe}lAhRR)=%-FFi+yl=p_}qVvRV& zjlqO$5a8KZ4zrJ_jtH7{tE>Z(bH?J*qRXdcQ&{0pW((Ss3Ann4fS*Oo80Z~3ex&j> z>qd_Q5ECSPk28wriy^LHm3XQCxO+_LwT`^oFBwBx?M7_ zP%$Ifn$ezfJahFLHP;K|yzZ5v4~z^(+cGt7@Oi!e)z_wQ?ex}j*P5+38?39$``pf6 zqvGX8mS;6#Y0I;iIO&70OOJ;uW*k#j)pz_3&=WN=L!Xl7Ra6EI@u5Ap?QMcap4KoR zmnx@BloGC%yM8?{zdK*4q0q9rt!ZSyZP4?eJA&3hg;+Ul8NWYR2P!Z?Je4kkvXRi< z4NroTR6K?4oMtUM2HUfnrS18NrZ>lO+pXX0HY&oi)LMm}=WIT{8XW%Fve>BVn06`< zVTmqT^xeprNxkU8Lj6@l{M|wGrqf$I+_E*FG?OZ4De@Z5#PuX#39x=cmX$|q-+rji z*^53mR8Y$eEFd-sOE|&2xKA`~h*+b5PLmS|y{JYiDWF&y71%ZUvqLi;$L|yu0KeX_ z^uE^pC#0tmmtMQ)UD&j*`;c4CTzeFrf%P|28a&o1pyB~epV5{+?Z7Q{UnLA?eBehC z*c5(WZ|VAn+@f#|PP^mXc5+-J2B$jz65w$a=-u}TEqY@2b0T2oJG{ZQaq;046Tjco zT_fbJ`yOfbE(S5X`y8Kbx9w(>&kh|HqHFi!^{iV;qFX*B(5o%waA}Oa<#1^E&d>H% z>|@e1u(r};bI2;T0Sfur2+N3Tr7x!nFT>^Yzj%bpBd_ycb+&d?wJ1askDuE#*AK8~Qc8aOvJOr9=(%r>e*E7Le8zu-qnfS)kG4Xmq4o+^%ggqVwp(E+>eg89##8@<;-{pfT(>eWg1Xi)Wj>4>??f%B&+x1H zbp^xc+97_=QlrYHEgg7EbmBR)#mKwzV1rZ5$D~gzR#lF7&vRXPPdqd6 zc#Ip1#x?y^O$L*guirH_&|cm-%daH?AtjC%kDI8}48=RjyKetly`LypeM3Q`Dpm#JG-?K$gHoU%F-H_k^_1!4{Ml+`fiI>o9x3V5_A(nd+ zAg0p3+?sK#X??b8gzHG=RPcH{?_I5HxA!Sum0Hw-R~#K=&(G@#eejOiGjXYnRc@o6 zZ}?g!{n^*4h~&;gbp=@NW}&7X?+F?SVsE3~svyb~Pf^(z1Mmk3LVX<@Jv@9`+SV`Q z^woim$_5xkt_>sKjQe|Eud?q=vzD&eE2TT|!|kc65x?6rmk$l=SA4IFwJ+aun+R*( zYmbR9WcNA9PkrB9wvsd4$bBB2CZ~=02;Aa$&N}ODUSmBq)rA>Bh`TL(H9@A>FlShx ziua)V<4)_Lk({?)owmd~S2%bq!0>C8qytYqE~&y=+rde@;WtC$legI+QeJ26oGS%3 zN*{Aq<%^q}e!ns|Un$cw;qd(Isp#3;C#6=|O>V)gkzwxi;oNWhDDsVNxH=mFbs+87 zV9l#sjl%e8k6hYXeCjt1#wx1#p{ z^`c{rn4nmhIO((Z1A&p)GkITM#M7EuYB68w?DI)!8qNozw#ld0$Ij`QFKoPvgHM`Y%K(1jvd%xo>zxF7wF zGgr6Pq>0Y)g*UcZ{-1@dmyyrg5YDYV#;}GBH>4%uMX#c7b=r@^p-!`e;DqfgkMu|n zCT+N^mb{s`&bv208!^RD*z^{ilaq=1iAU@ZAoafK(<;-CabaCOUtfL*Jv=i3+L@jU zVk*ehG$(uh)Rpq0bx|MI#%9%|8{xI6-=m_6PUsTvXROlKinuPpsy6Nz@9u;ImXLX? zj_Bl#8(&A&pN~q@I}P37pO_dn@*u3BAWKC%H5RRM=It{OR#F(~$vQSGxh{vZhYm1$ zp+<#2WHy{NRK@v#5c8B!xV6%ZfTxBXi8c9tpjrxksoYLYPK}9Q?VIxl^suCklMWEN zl!?4yU7CIoU8DRwUfW-#mL3S$mYM9vzCE}10m$B!@o`2~D5=es)LC>9dt}S1D>VIe z7>F^X26Ri0nRjRpTtE;9z|!kAw29jDlS#*or{T4(L~~gZrUjK}NV8E%;7lyhJ+jGDFLHPbkyZ(rKJ+!4Jk)dF)v=+^Jdq}Ll-qkgCO ztd|^|TLDy2W_S7A{UWsc;Sg^vk4qoQB8EP~fMXEY%q%4NwYap|_<4%a8#TWA8-d{m zNReSo`dc4y+rE5cwh+m1e4D|j;JjB3(Ci!MKbVoqKYbv;lvJ@1et1L^DkNz$7%mo5Yz`4bL)Z_W+R)j4_EOE5`)jWU#yiwU8xe1k%Vj`2vmKcAxcRqBbMami(w z*(6Y%&;Y83#|SoSe*+xYiq+Zj&_h0B2eb8(y(rZ2z(Qi_`QqfW-F9i0j`&8Qla#3T z^eqWs@`x4sSYS`gOrZTKmiO9bdn=Frze?FHHuJA%Q5`IsznpE1dgB zqeu7Vp=!++5c{u#6$cVra)`A&b;?F>`dXN(;UH7_*bc8&kPLoEZzVGnOf1dMwgXXf z_u1Gh!W@x$D!`~^SU-m0&K@^^fXyQ%h=*_D(}cXAW>el4sEkR-7wImvQ=?VvhpPtyX0~rlY z3@=23EJaw*^*EA@P-Hdn7f{4O*&=n@K6jwahn~#94lSSDeJA^gyVm9?^Qb#$HSXF; zaZb&$72X!}L}qg=8p!RHN~6HqcM*98`&lj*itu}brRndUUpc3UrV}4S+}i{B)0faB;p- z(WN=a`eZfZ;JmqVbli-4akb**SJq&}BXD{Da@mN#{GZD>I-fwJZz9qy2gAlVIMz7h zXY~V8L&32*UFtC^>Gwm$2G{%RZ#I)s6n?e)EFZXA#Vkdk#(CnIb10 zyy$z@_Kps7k)isiZ+=cIWQb&euH6ki(kaycZtJ#K#Z+w%BFj!G(EJXHofF~tmoZ|_ z`sYuCFCz8*GNa3c2p@W!7gwAAu?A=pf9Hg`iUrZs}o@WCZPo`L=`1QDL(K@5aY$*c-eetxX4+@z^E@-XiYL zuxIq;519?{&$CS64nDGj4CszRF`hKC;Yq)=x&4}E=R`tu2PbIZ^sg424^I z4iFmz*X6Xpcb6^j{g4JtFiY;JSQI!jyfAFC9~E*D%mNF7W<;;L`%Bl%BjjMyGl;+p zv9k!BZyEQS;lQVMnj!IF88+FVF#5FDKkB!k<1=GCB%laApcH*uREr*B0HS;n^T#}L zJHmt5lkJ51RoB7^q}+~Zoz|JMkOk%Fq0?i~PU9( zxbf`qnCi+k@TS1^BHg{_b-%>+s>&PS{`0BZBd3?(V;;^E%xfii4k=7R)rjSiQb)SX zPtDWhyZR!t)u}2Y(w{*W)lDjYy7+FC9ox9})g~q9LV(Mq7Xsjc?Q#*taBl2|2mifF z3#Z9(!==r2+Z1jSBf_dNDf&Jx6FMSM>O;m3gouY`6kk#g{H5M|?B=79+E72~BW%_JIs9H#lApfqknP)X zhNGFR`f$Wgb^M!jy+dIkeiC(zZ~{ldlfapSC=m6mqFuA#dle{lvlK_lp(?5S3^6nq zijg^*DJQ%N4+hzX>>r%ajaoT%dbnj#i`3C!7M23-;gF%|BkT4s@9Z{o|Lyb)rT4To zFZ>vul}UkyM&U-8RMA>uvs->JqdtX}=2sKUTY8^ZHuWB@DVU?pQ~ZR4u+9_NK$NZC z*W{W`*yNhT$Ne(16Cq`Z;b$86JpQYZve%Jg5mZH@Q5G~BG+W%cx}}WViRs4htHM%J zpmp2N6Mn}!IIAVTx<&;NV?&oWifS{eyiNzp=e3M7j6>r) zDw^Q`*$HPzLbnH}3Y`~uSw6PxOsL98Y3>^HuJ`DWlOlwjq%wJjc_*MfLlAsigrP=% z%H(e9Xo3->Y&UANTjIyAlbuiq5;h%!k40#Esu((HU`ACLZEK5JDp4dudyFPMAO$V_ znn4S{r9Uln*pfTtXW{&CBYzZsq8zsRB=w1Mv&$~Rxs>;Jtq62v*X|+K;f{*I`HCaO z30PvELlK9)OJ}9WA}NX9E3@z=W0PM_PMF-e34hECLhOF_z_fF1s1M`=tQnc=p-qLZ zh4ZO8r8P>;ZbzWt{R>s+bVJY?@{y|Ix|z1-mLCiADAke6kFgFCDkEQ3{~1GdgdR<6 z7(R*fXKC)aNvu8GR}B@rcA-lYU}85%jymiAxxcX#7Yy?)L(cd=V($bz##Uo&M-={h zTWyA?bszN>`d=dG?*wyQIq`3YvMd7}n z+VN$~G`3)F^f_$TFS+?6Zuqo;vh-Ii& z5~8WRcj&fdS{rX1WS{6p=u0{?M`)ae1kP6kU_zx~^9OBGO!**WS%$#HHX
V6lLVYBk9a5}{iKhwqV-p5bZu!0Gpg4K+X^pvOmb}^W zbRm3RdO1`F05I)hMx> z?GzU$#hhTsAOu^y0l)LFQKF^n_hl{==3T_Yty;p0o8G%o#Q!wvEWJesqdN_k5c$m8hP~%TyOV__BZBN%9Ue_CK_}(v zv97gULNAM>bRen*%E40JrS2(K6u}~NN6o4IBj%x4)0k9WvlJ=E0of&clI*h-505K| z4M@!oL7&-6N6lVi6M|@u$?RHSpAdCdtd$PFmfGl>mMJ=M^N#GeTpRLV)iN0o^fQf_ zr)vVGqCjz6u5Duic_v1nH!gRjk?wqFAX>99{=fVBOJj%yNf{r!vq20oG|(6B0lOuE zqVSphbq#cWbBMuz4F`N}S_WUU1iV9AJV7RPXJ2L~YZh8Rvx1qG5cC5vVr&}|4LYe& zO57#ero10*6p83)ynzn>B$HKIQY$TRG~=}y;O4}8M0*?4uAet=$EvG5cFtqfE>nv5 zydM(S(lCFAa-)`h18RQU4XkS#_s?^CyaApAagw&?<{^`f(`#p2-T($H3@B5i`S@hP zh7Q=OnvTZt0<~}O!r6D>!I(!~nL~?X25r|sI!DmFzLxfkgb^C}mnj7~HyLSwxi0p_ z+Y6hz(Zx2TGhwhrb_Q=c`u2C@Kf40hQZWp`9@(dtE{4?qpCRA>W)md*B=&GB^I}!} z)!_2$4g?VQX2m;0P9j6tknK2NX@^dOJ9#akk5*{uVB*zDYO~}%ls@;)2|ZtJI8bys zQPOx%lo5K*#RD6IX|h4QC)oWwHp5iJXwH2DO?Ah^<9}a%mJw$G8OKK~77<%wlRXeU zf#CKZ56ZffoP5d%uL9J?%>Vk;S5GpZ zPZO%3pUpwgkISM;`8?tekl2X1(JV0bzwgEvvIoQrh)}XsU&tw1uxlsSj24b`8Jw1vVl=7?V5;&7Mlb{tXwrn?Q ze3yc)?*eK?0jdaH$IBCXQg2tHaVT>F4*1~4-o~D_U6s%uy%~*(b`t!9Lfsy0>?OyQ zlc;Sg1AvwL?{RQ^J_Ah7b`G!x+9(gRI#6h+XOhyGAuNP-i(%e9h{Qx=28cx8lJ?`q zp|AAg9-UZ0ma6`*NlX|r2+Cv~*cE|~2uyU%WlxtKk~;c89n2eJC#>eF!dpLA0cf+` zuJ_vP%C&nPY*&|b|=>sjYGE>Iqsy<8y?@ou6U`A|csP*wbeTmxyZ+;OXj8gT1 zi@?3w*_IXAqD;~SKYxdBr2b)`7|A#Sz0p!|5um>~Eh23AXp8@8jP^)B!h!|ynf$X? z$_X`KQt$hR)N5w^MluwAZ~MLe=`^8&1x5ui9V4(9rrr*qN7z37Q5?K_p=zB9$lY^b zr4O}{5&>|N9v-^J0VKzdTCn=|#(&1EndF&CQzP!T9+5joGe-M)HV5V@=-dwchnIAf zkH<4}J7R1Q^nSF$51rfb&>plX8+}x=70l)UhZ1*y$7@@CWVuW5Ts4BwpZwFpZn2DN z1oOy_2saasxW9ICEk=;rPo;99>J82Y4jpX%Y^3Uxzu#|U8TT(paWQ>=U{6=fFYGDK zkjF#VoB=e2MkaU4NnaA@H9*<_xo0Qf67WnS$DBN;|B@+{h5`rR>~DX!r|8G^U)vx! zgY%sJGYAWTHOl*M39}BiuY9#D6mW|veX5*RDeVn%gm=~nNIzoqO3Nq=U>5(l;?9Qp zt-xJhhOHR+H>#1Y6OZij$U^T%faKV_^Ks&Ez>JEkDYb<097mjr`XYa|H^VZTV;xQz zq7%L6MvvADkeNP^ESXZm91x>Ay2AW2eb6Kc#I&(1AzmQXkeZTGb#!Z&X5GE^!O5F% zKtCf49s~Xil0D)~jbVc!r&f(w zQ{(5w(RORJp#zrIiz*`nU!O_%$b-Ya9Z{*H;=;8T185riiDM=zfK(+Za3X4#l#)B4 zstU~jr*(q%>~P0<%XJ*YqJEkl(~juKL`!;sH~gBbj)Z*9byv9Phoo3Q;eLaWq^mKG zsy37#f*#2S@l}sqA=QzD6a{{897}lx!}UIum-F**X8|h1Sv$r==iF02k?y4GfE}E8 z@;|P)iHISy_%;bo5yp(}0lG3UHTaJ^{dAR25a0E}?JZohZA6?g31?7l@gB!uq6l?vIN-f-6($;<k|1E$rlCvu`UB9NQy z$$k8}k6C2ffG7;;6*Yrg<0)tF2PoAe2QXMg)@-Vs!85Jt;U)5?~#TAL-1KNjn# zzlf%O?b(w-l?peWhMqo0X`qxO`2Zty*xj^WYdPa?Ma zRj;fUKv*24P^yosEAG9+UW7jPQRhqO^fHIO<%109Ncj+3Xh!|~u{_GIXybm!2;!&J z#CzkBYp*h!;w^joJTx^&HUldKCi*r(+`QRi0CP;32kC_rQ`$v%PtYEemNT7qzGWA# zSY6CFZB&Q54NQYH7esu;jh;+xx^_H{ii8B>y76(_$ix;Qr#`R1Af0JC;n8)$^HEhq zTMv|%+W-7~eM}}9GMOc&F#gKTIZ<)%43KzV6*ReGQoFXgeq)>$FhwW<5AJ#^+Kk#c z$=$PGLjn_xEu%Z38BIp{g{9ln{EJbV|QgArF`xo-MRrG+UwYELo)>-nueLb^@T{7U?#XVJ=D z_&Gxo@+eQC^@eF|$J2Bk(TK^*z*nKdssK{LBFNw-T-gUnwktm4Lq*f2o7=%&JzDg9 zUPCK+F$b#Q7s{3eBj$R(4DXGfD%EtRSk_SktI8VrqGa(icLW@~|3O;psPDLhSLi&8 z&50_wwuIx*c;Ov>L~qG(sksue8t}%3x ztqwp227H5Zo}Z$Bcym97!lgco6U$@WvC(o)z6M8y_1jT$yy?KzaWdIG7@`AQ)EATV zZrK_Tk#nHwZGBwlC2@Hww1f+azM>ehY8Ao66M7%B-HvuHpHvEqa<+5&)z$Kzym`YK z(raQ)Rqq|*>XJ(C^dZ*q{Z+3gVoh38WnhzbC!T`5WPF61{(sG~H0}L|X1(q|{J5Wu z*^OedJKRbZJmJrn8hRZ)X-hOS5?%B;-!;}ySJhv*XcsF}T)wn;4=H>8+R|)~uc6Sb z9RoX>*ZMf0$J!Uq&4V1Ws;)>I7hNq;B2hImK4@!()OX3*MJ8{r4}5}*EOGCNtM3dd z?y-l~-oK%z?ci*23W5TqafxD+ej_Ia@6D#=P@i+xn9wB3 zq(qj+*KhLly*Hf!srCFq`wmx(W583`S^nK)bU-}t7jtJveJ9`-H3jdj*PD;?YEi4# z*QZ<`DsjL2?Dk8&xW|#0%;>@QVI5e7pO9y>%SO&&IP)&NSBlBFJ(LPR=gH2*(Q$?R zW45k0@?aTnu-H}R=f0pq--?l#5nrUGKS;h{X)U63mYP@I1e<~kwx{Z;EuBt9t)4i9_dKO8 z=iq*xH;I3;HD|mJo5v(2s8bCpuVOPvlo~Y`JxoK2#?Ox0r!?gI3Bng}a}w+Jik+Ak#5$Ke4BDT1I-?tREPbtZT?$Y(3S37^$K&{+*d?saGPH%6N8Y5K)s?;jq;iIqJT5p& z>k+PBeq&E47OtiVv`^iS8f^WB-N$b4KmSF^e%4_8oFf6dn4k4`r!IQ$EHG{P9UTAk zI+69t+5|K-a8u{Nf|8L9=f(@V8X$VHHb!LV7d?tlae}j-GGm~O7MssG4!hKuVDO0# zoh#h~c1`0*%TU{FK;L#4ZdQN_Rl!DrPMM9Jq@4eB+7kPs zLdx?v@Pa`05SXZ>9%#FviEa4>m-;OsHgyF{OLcT8%6=Bx{5kq~DhHTj4P=a-WXngU zPZXuKOrj_XEbMCBO0LrI1hAzV$jWhsNoVJ{7Z}ZzS3gO0tjQPqBv4M#SOdOSkg7BW z?Vl$aNRyU~y=#+Ec;zGlLHrt>y~G9VvpA1T1@Rzn@MlOc6Ej#mrStKp!xMl7n8Y1e z+PU|t5(cPXa~O-BC1Rke`&zMUr`Q)58}>6$2eS%rO2O@Ao7sLUUN4c*d@Bj;fQGyU zZASJi?a+yr3jr$TOkY;jD2hwT7!c>QE{#YW6FV6taOYrL-nz<(7Gp^vP8AEM`!ZYG zlz^Lmx8o&9saZxWzM(}#o5Pf-`T2fpuHJL@=Duv_Fpq}`#Y)Knh{ArA~j;vTRxT`)tL^EwxzOOBHcQ5 zEx4h$xVSVyBm7yN-8qusQ}!!n=bDZ~F=Oy-PKYaNN#0cf=INA1&(|I-qW8zUyTsJ; zz50zteg>F#%6Zq*_Np)fN6pOy;34>gfwCb>uRbT)_m5<6ihabtBjX-7Dev<68G$A> z!a+J8^6A<85q-I#*xxk!GN?Ung??u44s$@u0z}DY_pO0uWK{5U51o{ktq}37T~t_A z)w5X!dF)Ziy($pyd->NW5&eoa-4E8k)Y;ft7SE@T-umX^O)DQ3ILeh)!SS@FF8+=I zaNg{|2MxupCPsr!PG}5!A zxYXU$I!iL&zeOY>>kil;;-R3%g!+klSNQ^0Vyb2ueLBNo%hxrP%zfNt7-yO! zQkDu3&na)k?U$vG$KEyqK2FcLkr!|N=tgTI6~R$+cQn&QR|KH+tccH}T2$chDv#0N zg2(Ii*(ee`nh??y;VJ;YWV3F&6K{cXI@opSRSLt{ChIoD)Qt~8jXUY|L7lvURu`RI z0nE@Z!u%DkI+6?eZ2260v~xV;qS|D8B>q;04%2gj#Z(odDp4CAnbTeTkI1c(HCtkMNidsi-aPn~eUm#m~81vaRLV&k*+0!S+xe{`6h=ETu`r#`Y9 z2*pRea1waf?()cgMrwJWzsr5lb5;Bcii8fPtL={Ma z8|yiy%u8GhYdfg9Z@NwBOk&KT%44p4=%SVqYO4oK86gOeJ(E_=3G%7p~3&F z%iE4zJe<#sU)hUIfYL|q>Pg-6^UDMs$yi?G#zH||U++z}5y1f5Z%=(Sd@`chboR6} zKiVjnD&N_aDaw78l+ut=4hu3UQZv-YD+Tx&A9&@BZ~+1C0>F9$wq?KOHXE z|89V&+13QXMy^>sBmEx7pK1>HK5hg02WOPLR0ka89gChQ@$?>z1(1hHU|M_qPobmxKX^#!GX%e6ic!`Gw)W6;joVDTlNz1IOTqW=2Ej%6vvFnL zIMVYCM^G9h{dza-_Y(}JZ3~n|-qu#Dwp6%d zubxsw93_?xZq2<5b2wt8Ko3eWsciFlFAmj{`m9$e<$&Trz%p4rSM7NOfTgrB@I}-I z3Y+h*ELZme1U?)y*%)%v!`*#n*LT-WX?Egk{!~o$BV)KcGbjl|UeI2s&H-{N)rtSj z7*I9@@J@_szu!jOp)D`Zo6~#`*I7ZSFey-mm9c zlCS(ky`_jJa>3n9gha<};@>?)=Qo|K14;zLJELUo$o1t{yZTDA)yer&6x>|K@Q&-C zerg-RCNM2C0xIZ0NnWeHg4$Y%OOKIULJE4QY<0}I=((W>+)3x-JJN|}kh}OhZZNe3 zW)~3A`P0H3Mj$KJVMr_ccQ6^0UAg^Aq6y+6&u-gVe*38DwQgg4BfyFfd*Hxkrs??D zgn&)f#r<&~!~jBzm>B3ej0NdLuU+WPSXas9ag&q)?Ce4n0N4uz0iY1ZHFR*{U|Z1D z_*(SmvwT-muNea(YuftC2@ zSm>YFLN^AKu}!botc)m6@a8ehUrpe%1C=xJ)vO&*1_mK}TT~yQ!?-E)NCO(+r9GGb zwTjvM`Ag3M+3*X*E<4UVm$i=pP>;guTLP;VD6o5~v&+oOiU-OEYzO{tR3v>lu&<3E z=^vTi0)^WHc(gdfehN-sa*=?(ox_dPAf_tzcfX7+eD{IJo%-!4}T`NX_S#Ctu z_%e}{Nqu9ZVMhK?bixbnTm8;yrBo~HLDcJv_p(WyPx|+VnE6X|-FGHqB^h-3$xRIt zD{;6Xu?HVOH|Q6VxGxLd#mU|{;dy8wo@swEoRm^7kD#a&y?_k~4Q0^8&^3l~zUR)||IE zQP)`_B}n#b)Wz*Li^bMs2}}~Z6lq{9aUn6jLDi&Tct1X0GPT2zK6)3*PUel{;Vt-J zh>emNNt8~EZ@3iCVlVOV=Go4@cGfhGt+<9KcdiobgtxrjttT=I{f&IQe24j#<`sXE za7qtq7JP_@57KA^>#ISVZ_M!UXAe!is^N49N#_0y;R&{D%wGIxkJo&ybIf~&Chf@{ z!O7nlN^ndAT=rveC74<>PFYmGxHsKz_*ldozJsT`8uu1Ue@J6iaHv*WIpr#C40sy<`ug=mPMt6?+}I) zd@SQ2@{OT~Y^TL%UFYe!4HxXYv;#C5(nK*!K5z&Jaet z(@ya2Se)pyU+_CplxvdPZ2l(?Z-^krt8ayyT=yj}D@`VC76UY8ArE4n<}?+!%X2^Nu{`? zl`<|wSk25j$QP2<+qTNQTF^~Z%1(j;R({m|D=e5i7Kxx>be~BWcmHD*X`)I^bw9R` zb&{ortIVgp>IzoNSdMBIrPmCVvbtsrsF{2CuQ)_L5|-y79VlrC-$BPomhrZUpVbT! zno8de;1lI|=RqPyWK3YXH7cIsG?y^NiQ5@O?Z79afwR(SuZv{qE1%WBo!=D^g?4*V zO4C9}IriPV%l>8%F4vWEUw&qLMO?`s*)OZ@VI(qrGCb)Nt-e8Iri7>DSQ-j6708{0?|#YLWryC@U!Vi}(g__=`d!7>x-PxgRl z#(L!aP$SMAIHi>*%%&5XzxkGr*;H=wzHz28)M6o{d5u=6@fYQ|b!L7g#j7XZ?yeU} z&`ds=og2Q{>PPTwe{>NQEo*Pd>Cf$Fi#$$Ht8o%Ae3CXuK;C3k@7EA;`~X`N@U@s? z*WDQJyweI9}Pji(<%(2GRB_SPS7IFMUwQ^qqErmP;EK zf-K4DVhnIXiz;Q}O~&>p;$?L|6u1b#*R?ZYfV>{SWxOJQze?HZBs9PwdM@~2<~P-N zi#66|c}p_UsS$Tpxv|3!ZOYiY+K-ZcY$vXj)&xg!hQV5V20}>>9g!?53bMs(#$}|M zqj)Ym5p1}q+x4b#ms*Y2%T0`l%GxfQ5;A&b;-2<{XWkmhw@X;58d|4>!~50c0a0oy z>AQ6udniYNN9P!=w3MNw)&A&QYsoUR%*~(!>r1MJUj2A)SR5G&uVlIBB}JzL(sJUbaXYF$3$QuudXrE4zl$<$s2TLpi3NM^|1Mg7 zkpIq^3sXtj+KQA~v*&!vI{+WBs(!z(TuR-_Y!h=}Eu_#FtpA%5yCN%0wBi*-HOQ=d zI4CGGMq%0q4*65g#mikmLUmyNMDcdgeqRR{^dhz}9`WYd9v9zF_myhl^3~2ffptaA6Rk zW_ct^A1gb_G!vlFeb???Wz!&XkJ0Z&f9A|GwF4!BU@|sFO3rFY8*pHpkaWK88P)&Zt{D5=}F1`dg(r}ybowv zQJy$W`TA^lFFDppO%cl#xSqwz+0Su4{6t-X;>{ApJ<|QgIx{>%Z0hYvg&}SrE_-;I zO%0ol{nBkfRyKUubPlegGxb@pnZC8cBt9)HB{SV;57iOA6Y*pcdZF+ zdVPjI3)d~bvJG!wpmo*EINQI(6y!jd{0b&cxG1x5>G*>&(Tak>z>jSg9?G@-0qo#c z)6le5j*j4#mQcqb$ygDHW%WI;wrDO{HtEFj^G zttwsezxWwLzD&0g;5&56 z<7is_Ncq_v&9bc*MQ=b>&VGYlI*PEx|JMpgiwq0L? ze2=3r<1I=RG(TirVyyVSY#4V?^^z!UqIP@8o!_^1pCxj;P3ErE+?RlB*msHY_W%SH zXNlyrs?BW-;4{XJdEzi`grwHi5(~P*SNi}B3FErQu^YXAj-$!75Q$&M@R-~*@L;zi zh5$uM73V`la?|cgjebVkYM0mPwju*fL)s%}o8Q|JG3}Aig%%sSYm&S5N7p@?xj(^v z$aalIH7Pu{sChI=*5w4P%^B0XQr79>#r$n=Tz&QWznNu0gjz_a9egelWtErb!W;p`e(RO=toLwRckY## zqe6SsL@d%desM+3=&Ijp{J5a$#|d|SWJA|Ih_acgTWrt?eh}%DG0;>y(h?~-(|^Fx z%Qr~Z0JZant`UxNa9Q5|LLIGtSAW4A(!Yj@?$u-|&rwYhpP7BZEVrl(U27aLU2kKKmN)yzmn`Uwne^!#tQy z&rdpCm5p1h;jAUg%_Ef6e2M*OANk zq(rF%e8Zyrs#MWeiVmwwhCZ6G#!c`l(ta;;nAdr_I)S~ySkwQj%I=#L>Gj(h*L=k{ z?l9jZH^r5>bT~lU8X>seZtcfD9PSpt_x%A$Z--2qqrV_))ut0JZev7=7FjLc-P%rn zK8zMNhZU@8ekmbSYx$EmEZgF6 z*!#jc!8JtZ?Cdfz6>Pk>D%)sKJO z1bM#|6TyLEaXlGTG8_MKAkBD0C2(ft{;GeEu;eQ3O+SL6Xf;y$6K@$ zJO&l+AIGKRaDLE=47R*7(SJREl~b)Vb>!)!(}weU{GD{~Rn9^0X$)rN7^9+bIXAJe zJsVl)*4sP%%xuD3@wp?EPPe+jyLI~`|JdjXFQhE~Hr zccfOZfAa9&>UO~9F?aUo`J>hX^oW<8pXU3I%;m1p<7!*hO8?SGB$JjuNG4b8=% z1J{_HYC6l@f=O+u(eiFt)-7Y$+Q};`ZvQrYkQ5MWLdJByl2`G*UGV3fuQ=UBJ-_e{ z>u1Q*hE}+{E#vxFM+XtJ&sLo`IBLv9FFzkTouT5#lb4#(V&Bo(4o%+Bq2R?n*BS|6 zZ+55@f>mJy`+m#%8+1__=w z;9B&|lv#3Sq$Di4%<^Wg$`#U7hu8Ir4zCG4bX^tHG6<8go;q^Ca-~nqQ$E}I-b(SF za_0_}TS$Xpsm)YhqE67faBA=&{^6li_SV>JUforOQ3fi&W$|v#!aU*{KbE%oN8Xm`>4bg zUa$b-xQB1Mz`yr3d56vzIFx&>IcC~IsF*9dylsZ|9Y{D&al+E@2EM_Mn3n9HrpkIc z*FaTq`HVGxV^7s?wu+9pQbmL``aXZ5l5t?&8;eg-+yVa?&aQ>q&UC6y&0C{}yY(%? zs+i^7Ib~`2HF4bcUtKjkh9vp)FHx=s$%pQg8%vw1kKrJ!eqUPoIpeWeZFLzzSs{EE zOZ7DL@SKCPGY58S&4_dTwb-v%XS0_sKS-1w$W93e8I@tt%XdTY&-Nh zMZRNbDX$K5y_-#MUkVDZ_QXWHS2cR6-*-bUl;gK@+Dcrb=lkbmYe+J>)Ft&3&)}rJ zs?(f0;c0|$^*atpCu1ho$z5)>HtF9Hjb9pjSp$8F{axfvWlHpT!@Iq)-iKOC;MGkl zH*v@Wz$+$S=x_P+eHVdze!YI5znz4D{BdZ$#&c(hzN^C4oL1&!jo8IcLTo0Zs99## zR@sQP;hG4i!=Pd`o9mX4~LfT*(rQv5)q**_{-gbHI)LZBU^-{bSjD7!k12gLFYTajR*e zc3+24K&+pYK9*<39dwGgCr)?h?S!wd4!?S9^A}IjNSEIt(bMYK_pLiwj60W-`1lW^ zo{?NiF~9_`b%ryAG$*J^Rc(V3LLZ+n@# zm(#NrQEHogPvmYTGpMmg8}(h&Fq)KSV3_?i7Y&`IJsi-zwG%?on*ZJ=Qs|-0I?YYh z!R>=w?(+T&+6X!g3FnRbL9E>LWv79xqnkdj{3|QsOIU;CXI#$n8ea;HsSzqOi3CVr z^E<@Qmy&W8@ubaAv-I7<&AODd{HE%XXyPMAUm5-xx2TlU99zjrkGHd~7y^Tb#HM&k z{m4hA_Q|ixd#_WE)FD|SIjXPa9_?ej{tp0?Ky1HbTB*gBK=9Mwq@!rm;8k&%5EsO~ zD>_O%lXtw6ea#{9H51~XcGw|bhzA>NvxuyA*g!{GagO`o zfe*rHZ}P7JT*&2}ZTt9-=|jJ%pP%NrLfA82fiheEE+_6Nzm$I6o(H&N)x?dpBO7Ws zfTT=;_(=@<{RbRJssjWEi#I;O^H^(ff?r+EBfkN#Sf01a9ekzt1nvR&EX7Wg-}#rR zC0pk*9+CWx+e+Jg>xj!xfv_C_haE5BFMSY3nOXrvL3L-j0?9SenlIAdD_emZX7YOreaQ@2|tR z?E*jVa?5s&QypK{P6oJ`jpNdY)z4PgE{l$^lWdj<(W5%4|SD*ZUlR)MaGY;TXNXnfkLP^&%!a92t|9=3N?U4n4 z(mw=P&uKxJ5PsB3@}QMI2%}5eY&(QhK{*9)RFiB$<){FZ_Kqdt0LOhI78SCYbv_q9 z&9L3oMEI1?28XehIGeKkpCjLN;k;|MKv>pjh_}Lg&)vI4DGZTqxc37Ha9vV11AFiw z{&q^%or%0kG~_tH6PB0$0)89KYf^V`1avDySYLn_(lzZ!AB54RGXMorPj6dD_I^AG z#<0WoPr_z!gyqlO0#5yAy)IZgV6p8kF}^6oev`Nl&3y&nWyUvMID!MO0AXVttE(9R zWrY7PLvm*Oec=gkPWZpE%#GxklUXsvQI!OMnc}JIGvN*PaUvwvAk z6Vm6j_dyt4S_+UW_4IZc$P$F#<3Ap=eJIK=LM~f(LSKrWuhtK(gb%jO0(agz(Z!$> zp8o0CX%@ekcwYHUmyOs>zGqpYz+6lYVTEJu4se`Uw?#}k+*7uIe-5!JTX#%6ES>Wg zFiSkcG4XG(`C?jl`$>;7gbe_AD7B#s@ILw>)YIGb{4DL^_;^8js3thl_O|{V z`bwD3RV^QCfXyFYSa3O9q&||m$d-81MU(fX6~adIu*^BF!opif>Yehdv`?+?gM6t7 z!USU(nl6cyd|Qlmv!8p_>)*{~4(Q?v%j&n9UIePT7QNjkJ;e;G*tqq38B>I0`T#u6`C8l}e`vS+*&xiX zfaK2fNpM#D;|kp1QWJ!6djKw-b!6w|Uvo#F*pF&eA1Qcu_Vb&uY^@U`{P*i%f#weND&b^^y>w<1)Vp zKOsHF73&8}PW^`qM{JFRhu59ld{PKKfYF=;!unV`X@Rg+0FrtFzX6E)Qb;_wysZ#T z8Scp;X87#4#EM&vQe+#&%3~o)5Vld`t&^2s`zKhftYN$HVkFyd8spTn&YMPP|D+$l zZOdPmQbZCJ4a##JVymVO2umyI&wGx~TO1y&rUZGatBdx7NB+B&_Sv~1>lsi>)9^;q z!hHc!2-|J(qOd^Nr!x2FlVBDdEWQy>M!XBc0#;vQS%gK%jTY-T{SsZH7jQ}tHkIF( z_pH**Vq|E^zYinkz30z`f4!f700xAG-mrya+9>0FNYG9zFaY45<*#ch`8`t(W!?_2 z2WLnP5SCxiImv=TNCvym@@G2Xt1ld1ZQws9*j{|5Pp^u{b=$g{Ov3IVf65b0BOx(~ z9l|muTKwfK5EdE%x9n?xOlE76w5{Ul$CfizN&>U(PfB4JX}Z;PT6u(j@5T+g6~MIs zc0w8~$xXtEWhZ0>6C@D!El6`V4Ht@ECEY$R9RFOIi~(T-N!VFBMS zTeeClOM7PRG8v7^mhw7V&XzOUzSaPkYx%RF@YN5FuOfD`r_Ctf3^6;Z$qxtCsdq238Pu-JYS#N`5>LjP)dh zao_EkAYB*+|G86E>}14$-`KzUX~U6!C#11`%sJtF1SH?y*~5?Xo4VaiDXg#64e1Lc;u&DILN-uvu|zuB~$<^^)5s09-B>^cwD& zd=N%L70m!$sXgijznL;2Y%D+r8H9Z-?evE)0bc*+T&~&&VW>16;H6H3vpWbY zIqTvpczt$VQ~uUfTdHY6Scxg89w(cho}FIC2QCIs!QPvqTsF8Tzt4>OJkgvlvBzFk zPgE!ddbf1BESe4*rEm!IzYEDZWe^q^3FjTR6%z{vLv0^~5m34@F#vmY-mDE_eiL5X zioUHVGS;?BP=v7HS+-jta=vdV#vH)rCMjF#5|$Uhdtxys*ZP|>k9-+@>V$R{%6*u> z`ZT2;NZKH*Ilu-^24RAaiKH?>R^c2ZKE%Ti7SM6#+I?FWjL4$~VcAA3+OcEq5W$P4 z?1*>0OVs@Z7isg)G$1T^g_z#%@k3#{99v=w`&ELl{5!>J6qZ|IZxQ1yLa6|^f*Ot9 zb@%8$!!sA0lc;*q0Fj-sWRVX*Rw(|yEajn$8oC4$1?PKjM!ax!ESJI|>_318P7wCJ zkZf6EO%cFB>avzMNji*Qgx&g8_axJTuv+}q z@VPbcw~V;OSA{SkpVZl+R{01Kb<>UNZ;1 zB6IK;*LRM0pZI3jtaxiFNj>>O4yI>3^)65UcD1Ho|MW|*b`2^QNy8Mv(nUgiHWvt+ z3*g90yA0q?I^mVRDa1n%_9xH$oW(CK%{<|~l*0+artlX}Ti(JYPb-{0yjfI0(JCxL zSk7lM+em*62pa?NN)lIDfrq`m_gY0k`!1D2x}qnb#BJD zp`KV-5w?dhg|Jt%rNqOi6~dY&@XJ8k4DerW4|TOeSeZzEH*CA1;_Va+VS=l11BtC9 zLLI`60IZWfbrrxV0~+5WS!Dv7moYp~7H66&Ec`DFj@- tS>sMD z?5}oGHUG%$?zV95Uj4IsPJKg$+*9cL!|~}EL0IMG2_LvZSmtDV`G6viE%_NtJfxUE zVhUldWPz}i0CyehiQtU@kF&`j>{ots>Cld*Q7Hhz_V$p{bQ^^I2Jk@Y9dS$*!U7)v z^pj0FUPx(KH{EZN)(q~Ka*iki=Ku!9rGol#HLs|2{!tqf6A0tqpH^*VWtyyzYvi&rT_?2G3PG;*d_g9u_}ZW zOqAtf;keo$wGJ?%pR}j&wj3#Bo(60JEV`X8g#%d7e2w1k&CLeF9JB?Xw9`FBg*9PQ z>lSW+{YfJt*wHiG#r*6jV+jLaacMVqFz{?{YUC0mDx zU_GQ)KAkz^_o*{pf;Ej8RR~it=MUsxUVhSVW&`YzLYOeY+eyCfm~tU(uA{n^&A_Ds zVR`t`Y%jgF5&sOv8L5NLxSzv65c0^nd9~-`40;1L)l1P3HpWfw_lB~9u#GmqQK#Li zP)9(X7`2tHF~1rHVG{vP%Dx!|a7hYbZ2+RA=F8kMDHg)U0X&zOMNMZ_Aj zN-k?tF6S}~!bZuw`z#y9Q&I?P0FW%T^1N?Ou@KgU|KoGV^@6eLCzAUBc1WMGNm|RE z3B9N9jP7S5Cc~vASR7OeppH@FlRXLuXmZL$uryEHcfSTrhPPf^w%fMm3N+p|J$oOQ zJRMzx+)OA>SW7_Js5p3Lxa2S-RCC~wYWwo5BZufa+cbU1VQN9F12urcW$p6;o z-W!?5n(-%+>?_*Snx6anr(eP!kiKytd9$XSry+jLX1I75O*`n=Gq`^ zpcKM}TmDM7N)$!1#mc@m6Wdm!xT~9sqd&jDyG#&yj2WHpND<8Ga4N-6fX#J5>crYSppv_sbRw8V-e7zNtY&CC8L+rJS?~P)+2II0KElw6tf5$&6 zZO}NcCWbJfB5MjHy|&P?sS(1s;n958BTjBTa7U2AE~{&eBwyGdVQKR0;QE97<|^S2 z4?tMJFSY~a7!gW|sNUSKovt+aQ<8%+e9(@I&#!Fxy$Jv3$SL_OX!YYvYy9uKvwAuV zhv^1xJ@e?v+3iC!u@=C$Xh|CiATubt>6}2qug^Kt`VpUIH&9VY>Dv@Jg9qvC{fU$c zVW*rbRdS|_06*D2d~w^M{c~)ow^{>SvBf-fd1?=d|0#bI#1uES_1Shw#wgwZ59~Dh z&x(T$sV{IGZ5Cq}eN-{$8%I&(5VnE#`?KxB)6XvN`MH!w!{}G&?0vjn3WcyAF8`NP zex=uye;uV?@{B!eInG()C^Qwg;i|*MsMMTCqVx00w>fUc4dc?^h19HGJj;9X^deCe zxzt8E5FOS}zNruvOpag>L)W6+@|x|z2g-X;GDq}wKUYef1paQW1eWhOuzTIi>UU(U z`$O-l!79&UL#DR>L@P!!7fo#tHWGCj7kWb>%oI-n!>ZG4&|v8~$pp_WaUSdK{r@;0 zc?)IQs2Y-QN@b$BgF1_G!@UPMj{V0UnY|nC{-U@F9n{P{rHo z!&E1Ai6S%dEMADosqpm^X+7H2eH0*U7`^6HcHY6y^^jbO>+XGDA9Q#LW!DO7-JH5) z(E)iDV<&6!^Y8+9IRyy&i&|>}UGp1xC?(tbx_dv)haH}Vu4>gqTbpL$&ZN{;JbMfB zERKIuOHBVv0m61tYt1LF55E`2ai$N){&#fu>}CeWGS$-E`zbyQ@ho+X0Ju=ots_+B zR3((>=+hYA?Ink+xhJ1-bG0%B2s=-$HA;@QExEF`q#=(w8fM=AM0f9J8K{`SIkTtn zVV0aM@gQ`yGKPIbXnslS3FagLWq!%vu{Gmald9 zejU?xN{w*LBGcK@O(^IWm*boZWRVz=<0wF1CSmi9qd8AM%Oo-s8@B^+lOH^Jkz z2B4L~TSTat|6%x!b)N&thsJVmr4wVR9@qq^_K$pN6ki7%(xz>+ulc5 zTL>VGqEYG?sxC51WD%nKkb}uNeS<_aL2NCOS~2Vuj>tU{BDgYGiCwa_`+A$e;#HINF{&V9hX$uk2-}9EvfC&( z!vEu5zi&RdtnaW$8Jnx$r}KzR`r^psb4+H>z^vq`5*Q0H#-_0N6d%GKCsIpX8uAE{ z_Z?Ofw}@^`ugQS>T8v>hl|NNNSa%$iji$PjEHUc zRmOOkT)>bBcDk^R{6az@PUIj#MG;KG_YO;dXH)`ci#7k|huU)ynX^fCQxK^JUuVkP z^S#C#h*at6dIm>S)T*?GjdYi~Jz=2HCb5tt42fVLtcN^N^22Wp5O4H>Mmr#&ugQ&Z z{m)`?zQjZrXf^m~dm(Sb9*3#MBZ$e9IXeUEx9Ystlit`B^!QXRG&q>GB927f!e%#A zA#%BL9Kd{|wIM$Vgk2!j<7RhoS?6UTjIn4(vUHi3A{dBaMyiZ|v)wP_;8gcao0Y^mAsDZjfk-m8SjN1v=#vuQLw=cX)V zF#)Lwl+Bh?Ne05edRVn=8WXqBGll5-CE-eG2%QFW(`zx^XQ)wsGKNHY6Y$F+tUort zWB;6MmT+QH&Ul3&3Y8=(u3#d`S=N<>EZQSAfo{gI(3d0wkyy_tefeBsFMQG31KY{E z*!c+_(J_S$)X#B(k%C!`E|FJ6-XEpK%25ZpRxg+)f7Om#1z?Z-!~^d%B4Zd7CgA)7 z3FDzGWKkNa2?MC|Zc8){5cQTZ>e5yYAZrZYHe|)lH~g?Tl?F6H;n~nY$?QiL31Ogb zhDA=#+h5272*%|lnmGhvHWd;`7*A(nqecl6NX$)@_g*pxGbqJL)(9uI&srM15SxW9 zp{Hwlr6=cKG7^R7^Ss7M<~p)OVu@cs&85{GFZS?I`Pcm<6K6X<-`zwrb~T$d20{y# zQRSQkNXd|*uL^0N*W@r#TLYex$mr>Jr2I49K;yX2xLF;AERGozxV|-N$JIE9zp-aZ z1&lk!$;@#}33c%_ zlNh7ny7?ca%lIc^2!rp~>&hS%)_4D0uzJf0t0Eo#m|+WkA(he}MC=QpTR+}2XdV|q zZrN8{N}*XK8qopO0@&v+|ErkJrbnc@$fTThPTeX>IceD%atJ%YURCrJTuAVn2dyE(FI)kQhHPnV}NihhCFt3Un4+)+OSrXFZ8yAT@jnbqj zl|SiZ2AcE^3t{-_%KY4zGS><;@H6)d@qhp^9LP(pBg&cbI9HC9Fk59*-HTNdeud zh!UA)Y50nQoUje-h4A=I^FovM^=veFHhpG!+DO|HIli)|Ff8wNBi7@y=pq=rO2a}H zdTNUwr_JGRPXa_Fu>6i?U7%_tp1d~KV ze@*KF62CBs-W;}AOTR@%ZJQP8UBpNu)I>Y!sUsL8Pbl|_nv|Y4&#^SbF`K$5aQC-= zu~XB=A0J%y8=)0_k^-9!7&*y88fsnw%j_ z$mKfPa}b6$kC3NvKl%^`Omgm--1?Z%PtNrV4QbHNw{rlqB`mX-4I-A)BwW_>y=%K5 zrpj5+osof_*24SXG0O>sjVGWQ5V-U6yX;5aOKK?cX2&>6P64SKT7_tHs3=%|O8h^&JRn&%LK_1Jo z8NZFq^EmtI+gVRL9@0u}ACd$Kp-K?OZ?Qmqm0QRJLKnIX8>V28OfrjmSZyu&-13_vS6`$tB==|EDY7O#8E;yJtlCwJ&(&Bf<--CymZ*{ z40Cvvdg4icXzU|N*v5PU2fFh;zDm<_?9anR4h0CyVulBvT7)46Fz!WWaD+aBona=8 z|Ecoa&m?vxggp3)NvfT@I4ya|PD{%?BbF11!v}tn_5}9Fbbo34QcMn`%I&;8Mnwi; z{b?7K&`$*sxB2m*bR~>M49oPXg~LTKN?>=z3%$blH~V9(>B=lZ((eMXn9L;Z(C^^$ zZOMEJ*VfuY&Y_(`VgSGae-{V~xXt8PNf*sU9TMbYhd%TntQI?Y_{~$~7*mZAzu$?4 z!60XA5zLh{vnt2Q-d)>eX13SnaaZu)r!!3rZM06y0Mbd0dl2DBx=AW{GDVL=9Bmn^_^5N z`rIb*FZ5zg5GK50x_W34hQ7-~J~+=#J=e|`Ohpo?bo}%=D{kTOq|%_Bc_bs`qXFEpD|jL41e4x6m4eJQAonFvXO zqVPHY&xn~E@|j{;*s@F(KF7rK=Qw~Ppa1xW=8mX57-eCKkyz^jcH%C0SFJGg)#_NemfI)P(i;jOQ3KmkTQLW|I#g$ho4^xIhmf zST^*4jSDzVCL<^!*zz>Y#iCTKcQ-qBGl*r9OwDN9i8#G86EC%&XlM@E$t^)7Sd1aL z0F{>`Z!_XnY^)=_&5OyUIZx3F9Y?~2s((3P`ubBA&%)EKvCuZ>YibB~rdt4Osbo+z zRI}q(mV#Jj(veCJ@|L`05&PAdhU$<3$y)@RcN*btFqc86HTPDvzaP7B`|3%(8;5ud zz{Cu)x);G=1#)K!!`&r(O>r5vfwgt?J;gUELxwcM7uJV*2ZCW__0>T+XdG&jC4n~G z@{XOpDk;bD72r}XrT=`n9=-UD3Nyc56fvwZW63D5$TPBI*qDY?_N4a?1S6={s6qS5 zoArzkT4xgQj?B+|KIc;AoHwQ7SN?9>bjB48qjZK5M-Rfzd%&jxpd>?%`)P>fRCez` zFq{J)x73lmS*r+}(42(%zs$`aDt;%qQX+i$RrvXoe%b#>iR*V1mtk#LiYI+tY7rTbBVO)|*43_y_{1$weNNsr7 z9NyN6AZXIA1v-%SpZ!$!5#-*D&l|FuMPg7)V=43%4bGJ-fvXq`%PaE>!|Jx*J#Ri7zq&Z0JoiXTo_3evfgfsI(yrg>}nyu`Yd zr)PJsQK+Dd!kP4}-Xj`Hd4wcsTRZv)Ml-WsK;sQKARjla6z#FvKrR=|QWI^c4o(Y> zJt6yOQ2~y^G{nOo*PVaEwhPtReM_C2P%!CHQ-hL&h0=PvCg5kws+Yt7!aC8CP~l%5 zU^n7)pLA45E6aIex!?^`u9q~&=30OOnmr2D&x1}F-wg`#p`P+2<4~>f+xvavx9GwR z4=S&W3hPd@lb&K;z;yi%zRv)`fRa(Pz+cCRn(?xG5Y6$*L26TfLGPKt-_-&eusO4eZ-I*2C6M=F&(y_(l8ZLJ%-YsVfWSAwp3g? z((+i4(I5odMHx6o;I!f(3kVxULq5gmxT*a_bIkIP6+VA49V66RaT{5eOFInDgW*F% zl=~23tNPGko5@VBF#l3CJvuu{(?=D$N<5j3Q&+w zyJ$+S=}U9e@?9fq%o_A@nn_IvwJ@p!n|G0l zqBOu3V+MHw1ME17>>^f;K`WB#l1s;bj>YMXMti6h21)xTSz~@cW!Oqd!j_13zh21f z6H6JR-AfUVqcT(C`wO|c)TGoD(?CpjsXb&1<)8%EYRDLW3M}Y8>PDq1{We5X@O1IB zmY$;$Rb!@GB=?iYB#n8Ecpm6~S>YkmY~Kzt#MppAh%xEYXz{HDLtq6QQ#-WMgg!az37?ScoD z`xr%KDP|k&NvJAxoaIAzlbb(&!%lv;9PThrq*yYM9fZLRl=nzT0&Vu>S!5g2m&tz& z7Se6-a#!p?;<88K2{Ko=hA6y?8V6yV{w;1Y&AcWh$rR-bmC%KtEi{yND{lUV0&%dP zanT~b{lJn=rDLLLI%zPebPByE!)>%rU=*wQ|P)TQ;k4Z=cBun@^wCIjh@VLV3K zNX}DuKr{I*C>X%5p!uK>G2 zoZ&Tl2!rLi;XC^yZ-TIlj~EFgJ|}}I`UhjHTss9*n#-G1|4woo=LzzEuGpAPjMx># zM=&ZT@XzV)9%`evprz{5W>J#y5HNy9>uyhtWG{lSnal*zpZa9UF$TkQY(`0)sbC7L z!f{oCb!x0Led-YiQ>nqpDvx^*u7Yf!Vks;@*-G>bz4mDE$lb~`k5aXybT18oFocA* z1F7=;o`Jg4ptF>xO38WgSS~MQrm*!~#jP+&c2k&cYi3@LK-gI2#@uDZ&V;ES>!=`D z0QJl-n>nwe0mdL0#8%SVBEL;E`I3|m3ckotAXKYe)xtgaLa7q05T!HvZ{gpDSQfeK zCXXn*TCQ7GMH}%v9e<^h%XSofz5T@75G?$eK@bM_bLnkU7{%>dJju&!C`)-Cg9C8Jl!QtM%}R!mHY*zGB0#7|y6fQK~0UGnu~LUJvmy21o=#;-g7*SP*)*QM%4{sFqZoSN73kBk&|nl~X>$OA(m%-k61B~s#( zuZR9IeL$NJ#e>|XI2?x(2xlT=Y23s|#}h{r7E&2_0Aa+Hr}q_{EJya>X+0%Ixi9sIU7!tr7uc%$Zw zWi5?jcn!;am1ZMy=w}=lrM_*w?zFq`uQ&eK?VHzEb<9;larr#TC>1X=3(GSqX2pjz zu%t2)(|h9zQ$+4#%oO(*oc4s4Z`(&s&1k$Q4j2q!a6SyXtwHEnpJPZ8LQG~Z49r5b zR1pj&Bd|WKwVcbX_N#xCZa3&fUV9(DWG;3f2eDatnVnrYvwb0|UMAnLQ{0!R#yYa@ zEbVME*=!{?T+E8^c^G8dL84JeHb!9c`ub#NtJ> zckv<_{Y<_T_rL1_2zy<~AjJKL=wv5vyaBSnSS2oJppQCja(*q{EC4Sm*wsQ!%GQd% z*&|pz8bdgYY;$L-vEZb$P4(V1CF|Uw@>xxsU)?ZDov&(Us<|i@njj(Bm*Ful20*o;BI!-2v&>E zam+Mg4`lzs4g3DJW7obzNB1us^G*HYereTK<<~C420BJU|I=(W7iA0h1%Lkxe>Q`> z$RiMjF>!5_Psjc~anOq8Bid9ZRJ?o%Mo3~GYO^Y^SsNw|@uTy?JLfzBizb3fdA~>d zdKF@cU?0-*>RU?NO2w7cdel`$8YO#II{lu#s5ITyy8^Zc8432drLH`#c3< zuoHiD-gaA!ChhVR^)mJlvh0h3d+OaZ2J%Y|s5LCZPW@b$VW)(s$X;_JbT_UNzyU@GRs=jp*ej*betUGrN+!(0PYU99 z(;zwgh19HdY#TG&9e;gA7EU8#9j%T!; zAZt!oD~2_sui1XPXiMfeosh{Qx6#<1#Y<99-3ZmyUqLKL4tZ*#W56E?Q6;^0yG0r}NB*b-wLhoT2Fu?};yBp4=!kGPpnQP`*;?@q7hS-XWA)0 z$26fkSL|Dp;4UU`ug(V@O2^+kTBiw)@+&&%Bn#18Ag?ev7=vwM z6?CLLZj9<~)!U%74;ht*+T&RWgFQ%^o8A^E`Wn&S2VD`~r`VCAvDjBJJb)F9Q3*4?rW0PPqcSf*qT zLl|^NXl=MX&JNj9i@p@EVxSNn>upqdL&nQp?hO9}b+Af?>)5GnY);2yUL}c*Xw20K zw~?6XyHGQy`x*8EuBy99Md#;&n;9hSO8W?xEn-rw z!-lEuzX&gu^$_`UZn{R#fR~DB2Oy3fun#3C)FGu4elu)B`#){yMu10MzsyX^n*|G6ku+rsP_-f5K ztfDG*!-!(T^k;?Ovd0^!yEH24f@CHbD+03~ z6Ltx+QO(imtyh9+b*DwRK3J?>qW0rjb7~W@J+D6}(-0$dS8wfyj5U>s$iiP3yo0v8 zZ|ZuPT>#UHp+8>Mbk$%Hl-1Fwxkfc6UN3lj$Fvzob9(79J7$9Wfb(Q*ztzhq(9pWD zQM8I};!qbafiU;+?y_8=*uGI~bW5g$(GqCkzD+p|+a2uEEhn~H+dE}&v^wY}Yx!Kn z%i!9PknJJ`l)qXn?o zZIi+cbAU-guujxx+mvMFJOEt!XY3_(AIJNayv$ClK1)b>gLU_B&p}uLuR$<~%%bUL zD)*v1h0R8<8(vPs1#rx5n~aM(M+zuqU1)PU>ZWa!5hB7_cx*Cct-b$Zc#ZFv79+}eU~8`YGAw@{QQ^gQF{Y)y}()L9IvX7e0<8CHzK zdGcdjenCwWjMZH;TFg~Q$%fs2{{EBFp%fr&ju$~#l*SslvfO@uiGa=T`XsZ_cmcSp zg<@RL+b%RMD#QAe;Y0?M<{0Ep@GbIeqZ0F^%LrH6Y_RSkqyjQkm~OKJMb3VYIM(R0 z7eN?|S6_cPZ5&DR14Zh3rH98E-+y;CdrunHl2?eY$6l-(bO_-sAN-HA9?jl+L(ygnm^wTz7xht10cpEOW1pu9~s@uy{HZ^EJ^z z>@%-|Fqojy%zVzTFOrQUT+{O!Tj+uAPtG zDdqKW?*(o${4^c*>J)T6p~V$8px!dUW;clh!lrl^gxynV=4@x zNC06cy$ixrCQd%*;}yc!nT}lFA9|wq*EO5e!O*shMO7vCp--#y*`78VC#2O#5;{tY zt-G920hc_Wy!rKrplmF_8o5azWEntwC2Rx>N%ArXQ|`((=i6?PaY=TcCH6N@^gcmz zhihz!gGZ>oSZS^>Z_^?@7Jw24p2e9_D%X^u#x^mxHmA;>P~L_*UoXd;jhZ0`5ptfo z(7RrSU{F|}_c!SozIzY-6CUZlR!peMV>B5B z-S>p<_SYeTw*g}ExSlCS0+hmqu&=xf!R{-*`fpO_^U$4O(t4&ON=1z(8C`OH1F|*^ z@`ibmCDuqG#bnei9e;|cBVLAUar)FFx;s;c45kA7p=^<#g$rT7dl`fcQ);AAh9%Ms za5nET2}?=il-}=-!EH&(g{n<{?6@3wYc&I z<0dYI9q=v$JE-XUqg057)JM>rv3i(i1n|8#f$C4l>Xe^-jozqo-emzerQ4^8Y98q8 z8dkx4W{>FZMIGYjEE#CMFAQU#r&38y>PYN~S@%B-)N#?ELR&Vy>H4E6(H7JDh?W&6l{ zD*IMXi{F+$7M9c<|d~fN0O~abMr4C#V_^&6vBhBpS%phlrmB(SKWI^ z%Bqj1V7~W|0Mr+Uhdt1m`xvon)eP`7)X0rl(|9R$&?T4Lw2!#EdBCZpcu}oyo4dD# z#lthjFCIWLX7c^vUItH511L~}8fX;v$aqx4(5aCIv`hV1 z@-MGYS_SG$o9#)p@JM#+;G5d%tbcSbg)0DWD_kxFa0~N5a+H@rSZtu~Gx7_BtdcTm zMaio91~NbE$)VDs!5$e8obcNK1J;YL)UslSMS4I=x+;5vKydp*jjxF;xNs7}msV#Bu3kiC%cc0AgI-4YTnYQyLS(@_HMDxp?e- zpmygC(go2H4Y{b)O;2WD8&@T&$5Ry!-ow$VP6jk=CZ~w5p2&~{XPNIKpZY5tys2cR zWS3?qPB^$&s{Xy6x;k3{4m-bog7=~4c5=`$FN3hf`n{o!5VE?hA(pYqKbgqhWuS{! zxSZQ*0+v>-GN5IBS19;GnqGTEN;>iqk@=8MM2N zg5?pN3jh-E%TQMJG6=ioY{ga|5aN6y#Zc$qo$zbS1)vfGxZ6YHffx3X=^p!*CBsaY zby(rUo{>^tw-U)tb{@$uhJSRu`k%q>Kw`3E|0il+M}4W+J?@m$dieG+K*?I|O4Gdy z!jhcgCbWX?NMAS8)TN@FMXgbK_5x52e9D^asR5~orByWz32_7&~V4Y53fvI$;QHotRi)s4~Ki8)lllWU>=Tx)2e@y?IDGb_7lFZY7VVRUT-CiNNgocGz7y-}+}$yM{Jeif2atrGejE|?_bLc0DQlZnewk;D30bMf zbY)Yz{Z&q-#XN0Vsnnj(i(a#{*GWN!WCKpB@9?IW;K}5j?$T9+lSPLO6`qmDb=RSk znrf)ZGaeG3blAKhZm8XmPN#B|IVWF=yvq^NmtF;7^61sa8h3W#NTUW-%O(9!#o%E; z0H9K3hk*Y)I3VTlw5o(5l~|)~Nq%eH@Tk_htH(iD4Qr~}X=?*lRs)qEaXimt(oQyX z=c1TT_tx*DCAJEO2`g&*lOmXBGAyf-EK0JH#*NR((<1$VgzC5QaRw}gjw>|hSM|Lx zHNFvQUkt}otI39_%TqD&bGWDLIVWe<76#!8=%DzJvaVk!I0p4me8TcD(FR3AakfNQ z8PyD(D6U-@tF`L;TmRPYXXYr-F0X>H&-BKswBWePguDi#Wh=S*mH|fU8?LQMqrTew zhS~|bs`K~(OtmUvjJ{mwm(uTD^SIh-dv*E~R>O)3Pze-|PNU2H#yo`og4F1}&CcSw z${p`>o*}>2UIk&&=r@zb?m!H9C-kdWtR|BM%J&_Vc_Qp;t$QmqVRVmds% z3N}b#j%FC{1tn~bY09EJ(X%hu63f?hvuROt&W{vx0slg6&oV^ijz%7Wu)nxHulIB9rr8H5byV-P)C6nq`f1fjoGMH7->_n3f$VzwFM~ zyd}+O*ii>!gVmN{n_al=lr9PSIM0NXE-6PtStuVr4Zg;Q zunJzL9FEAr*;WXuAM%#ch{~Cy7ROM5!fHnYRO43t2qch z!qcqMDHp+#TuT7<(|fSd-FdU;h8~%O{O~r9XMYzzqMgp%1ng=)%JE(IwKA~iz7OAPyJuTE&R|7b$*|x70HpIH=taoWbyxl?MsJabYy2~d^XR4N^ z&F`}3MBrQxXP@$Ib1dIds{FcDe3(jwt?PKIwLUdLm_`+&KP9()W!s7AjSNc&vYaN~ zHK{JcbUMA%BkA62r=XEArRSRag?ULk$ zF$=Br1?seMhx8abYL(ogX1p-TX{iaq(yPls>Qj2u)%8J}j@w-Bi>X?#yXqSbq9VaA z*@Tqbb3fE@pC6&wwjBJggE1kjke6wKeTdcV)N!&DnQ4jgQklav43HioCx6?_Dz{a( ze7%7VQ%f23bkWM+R>RV*KB)-7qBY#hWmXAWhOe?nW^~gH_}*?xclHh2q;EJrAnBir zzl2gA;Yu2T8rpgFg|aQU_Wb%@rV1Nu4eO$Mi&H60vUGrDHMZ{_q6M6e^J{z5Eu~CZ zR-XWsyjIV zYvIoHG6=ITSXEO_J6E1+O2kjyW%%tx4~gKSyd7%I2ywSsMIvi-y-7JG+7T?>%4xXh zrg7Xqo*6W)y2{+&(mM|;22QFPYpvyOGbY0a_UAcQq4`tSaFC>Rxvl_8F7#5%?V*)+ z<6f$#jd4f2y$r%^-f{1!-qGBt2V`}wS9@Oyj|h1)RF)NHH(q0iR9JIO8-tbJ+%Mi> z$;~h^Eu*MKT1gtXma7@4P+8EoOaZqXycS^DJMAz_;_-ikA^|p5<^|X&u0eWe+ zEsO_i+!XqEybQu@e)rwfZn)=Cw1p_C?{Rg}gkJZM3}n%0oZ6AG)m4yAe&=MHpGqNR z#4L`bTvv^i+bq$fKFcFxR#{69X=JA)huk|pDi}tG!zsrDozDWQus*qS_PbZ~%w49+ zx`&e+fs(_=WlRL~OYkxXv!&lwjpZNg(uW-+{GMtErf!b1a8~X*}W5_fmTau}pSzU2AZ&o(B)JJ))w_SmabnYqXxx$W72c zr-H~e5REcL`GtI_8~>Z%Fd;0fmm!!f;h*W5%J=cEh11BYu7+tyXQ5|gP(xSuwpzcx zO@rk=JG=IZ-*?8+u0JUX#obyvv7gPYp@Iiq%d52oA-C7=Lf8S>gZ@nyrRlf|%Ppkm z^^8V1e)nPnHH>?%ZFQv!t_C0}1>JRP?^mTVUWQ-^)@`C{Hj?G8eP~O>Q7eQw>vN9? zp`NS}x!f$As1%@D6DJQwFFeUIlD)}EqJ_LR`7AGJZ6HObHTvVuk2NJ3r-)S9%R zE7F7Wm~A^yF>qS`rfIHcI38km&uNpVT#@pq*5P!tRv>(uhuX=CG^wMPL72_6RV|fl zv_c;Oh&V3ObQ<322_dx9XOgqNwg%H++1ZT~H(nZn4=D=e5|=}arFuNYbup39SMD31 z@1?;bcGF$z_k6OA-^(yJr|u4tMBV%=^W2ah9go*D9Qf(f<)+ZI%BdAJbz}s7xv zgnegSCaZjH&P3fYMHEs4fl|AyA>%lqsa68#RlmZj=yoc_dx!f zgNiq8(`H^X31Dw_r8QU&UQ1TsS1@vpbc()ro`D4?*7?X1?kBmA-V=iw0*2=?H^z9v-NvH;M&ovM{l;1_dnKal4 zMmGByP$A98O>-5seKtNlaU&rQgnReZ(Zln(>FlGGUb6~s*&2B2b%3H&5RD~M4IHpJ zI@;4%Vdv>Sh;*EjH*`awzgzF8N z@YIWDB`CxYI@1i}2CR}ECvw>^MTzUS5VQrk-z^wo`K!9~k;ZmCBEfbxk-flFDlF(I zbM>{bnTY?hiEYj5_wFF^w0&o00tDk?ybHo?K?VzyzgEc#S&E2fdN(4Yb-^n6h`E2nr+m0ozi8&Q6m%!mQ*v3PMrDwL=jItSUf=0!&k^Zw z&S>blgjzYr=?JEj{4OTO*bI6afLs6Ll#Z44Ke|K^+S?__Vtpx&h~D-x2(v|msd!>H zQ|Sg#36KhIr8LQ4DNI#fYc7fcNIvkBOMxv4Dc_`Lpy?FK^NhjNuLcyQf}T0n?@vR9 zD<*@!=E!qbiqQls+&#Ogv*~&!$ECgCa@%mGZhC*n)(f$U_f*t5CK>wfmzeSH&#cA~ zQaLY!u*Q~VWmK~|t8Ds+N|@Ak|L(Y8tQ1re42BoZ;?_`%MRg>PYp3x27nu50lA=`V zV({;sWWaFE)7u(ZJ_0cXh1>96d8`@N1tK$=mblz?GgKFU+-SlM?w!K6A`D8FHlyZl zII?5O#6ItuIASX3We}Fn8jhg${S4)9Y$V~?(!H3<8=L)b0oou^sG9w@c*C zS2~JP`3$E4bqw&+|DWE*RCZarb$9n{r`YEVkuXr9L`x%m{PSXVu(*4tuxgyA$jfYD z&#kDenua7vc`i+7ZOQn_P%*4P8^my8hUO5*)3PeS^$lzVqlh4_Jie7xrnCB&60%(8 zf~(ypsPH%*Ubx1Ktk=mt46lQ)O>Rw3yFG?roGHnxAk20&te%KfIjPA|5~hQ7$SL4} zkz)7W=uy?I!slPFAOW{YT4ePdyk#lvL+zN?=Vs(2} z;l&c}aqkpX%X1LMJ@YCE8|Qxip%|=3RC85VG2|l228v-T;vgO-Whf0}*-H&Q1frP9 z{v90tX(L$=)&-P&$wQ>StfX%ARo%fW1+ePR<%X9butOe1FttQ39wPf*K|N_* zl+v<&9SPG#dgPSJNHMHO9K>a$3?g3ib49yZ2D7C!i>oWK3`L%8R@geMxS3#KlopVzCp0NtO z*h#{4tR6YlGg1u85C?G-DMN|Rgi`t1CHVT)o0?D({c!l7W<6Ht(I#J2$r~avoa5XT zlYH0dOwrX4wgut+vIMJ(Qm>=Mfc4w?a8>LtZ-TJ>j&?m(zsohIHiRfy zdVM`|`qxM~Ohp_-;_!ibng>n^>G2_~5;dVCY6^gOUv$5gS8y$QnV>9ah3 zOLZxEM~|3F87hY%h=UkF>UryM?!HW6uOuZuYFTgdV{os)dZf-`80gkGem-p-3Un88 zL7#*MXspEwx{=q8KSJqaUpfA`PByiy3t^I&f{^1g_}d;;r{`WT~8xf}t&uVUS^ zh_q$s@OQ}@ozMu4b7rhpog@IShr-uK(Sn3dO7bdH2e@zBQk{d%xGHwvdm!wT-qRB& zN4oArTyC$f6zb*mN_;?M4Wx3iFK*^)mS1yd>X_CBQra7kjLtGWx=sgf7ufmdFeG4MbgBwJa&+C0X?mUr( z3ZgCQAnqpP7j3s@>{Ix*91(<3Dv@;+$2F}g<8g{DfF%;sJ~TEzi@$ouIm=;Eb5zDV zaXKl9+$gH>^&+$;>HUQ!dD~S&f z$8a7gLvwD`qusbngb+qLKmz@O#rS>3!_*90BteKlF+I78PU~YctU~6x1-*+2*A6+1 zkd=wIksOysvaCN1@TvWar2uI$rRn!Vwf*tNPTVyvZBx%_>FB!OB`T9wlX4g<2&H;TCinpse z<)^+ce~A8mJ7<3>*5%)icGL=e1-G!Jxk`O{8#{4bQ`(hP-GW5eJz zp16GV33|tQ?454`ts4MS?3cfken6$pUUD9poDRj8!6sMJ88WWAaoE%(om36&lqVpH0!vXF z73v~?K?-4uDKu^|7Uw3chp9J_U{>LaBgpDzIueQDWn6w;RYvkbMx1^N36QrAA1kEj zu>0d@z5=k9U|)XOJ0MK)oF2JVG+GpB?(nr~u#64hxJ<6gmV1GO z2v(7XHR35&o}J%W57Q1J$%OC6WIN6x4byV8^FaJS+H2x!FOAEavLRN#1MG)S(`~EB z;5`6l68FQYUV&idtcs0OO-O3Vsny8{EgWpHC?2DYA07Yqb{yBz_3A*NrZko-|hw$$4B>U`V-#NOhr=H(ER)B|)di6|R+1%&Na zZd_)y(>AfbbTDBAiCI^4{62$4aSYA-iH-;IC8ZMTF|R>3vOWPYil`i0c!D=Tn3@M=RkOyM z`3FL#+x5sNW3s`bSc5ctJt=X`X>vX2pG~?5+eW8te`B!j$bO){Ba#iyQSD3FJGZ_W z4(gryL@{!8IWC*Cm7zXIjjFEgOjuBIw>LnTYSyklRDLj#2j3@zOuyD6o==Sy#U!NB zyGV)aRE5kmB?)!2FeSg7M;IEul>I<0A`%XU&&M670U08Az3L;Pw3xeGUHj*8-b8)9 z0Ktx^pN*^J(zx$OY@|xmb@VseXjvfT+6kexM!Yu&Z;`4+V`+&5Jz1FP3HyP%P9mE$ z=mvmeIi31A4O4@TrWc~Vyieg?bK>snDX)X3(_$33%Z2!qv66pS%n2NJZm7$ zQ9)K4Pe%KZd2cmhq{b2m2jisUuu0KNLeFBEw8)iwD{_~38=1A!;ypJvTNjX~Qhub* z%a!bD2-6N(t;i;G*|$a>sv*#LgrRH2t{Jbw3DxRtdTohCb+HXLB+bemeBqdILO!H+ zuP@9fcB$1gyPfbpp=M-QU_qQ8Egwh0@pPmZ=IJv@cTV6#m`u?$dPEZzXSg)ZAdK8< zcFp*WRI7fY*cen7teS?5R1ALbLS7txl05YdB~A{1xaAh`ie|S-m**ea+&QfgMw0>7 zlCHzP^)Q6FbIlE85v90KmuTvFKpeUBkK8?GjYuv?s8nQCS)AxgMCJx1sjPr4oyJ)F z>=lcu%cAFNxa0Oenw?%irWb+Z`1NDbmLb2{AuRi4fI)=XmMzJn2$tYBn@URtTpQ~W z&14VAqoeMDOUo`H(pu6*SRIR6qd%MG+N21PKELr9@gJC4V3w-5}B((s6z7cYWXY-p1_i?0a{3^B3HAV|I6T z-tO$oXVk=^@FT-m9<6e{R9N)CZ>3Xfw}Y!H41oI|co)0Vk3bVs)Q$PL1@>mpcV9ZC z>QOHAcUDROvAiJeD@|7tR1~c%sD>DKP?9ms2K15Z6WkfYjJQ6eae&^Im<(73SklicM zW*>r+5j%B8+VCVJ7cC6qK}nBIg*|H++Z9`O6LAOT2Zap(1cVCN}F&$)*6226-#v$+TD4R#**fybD#uVya<`^Vn5bK$sD(h)q}8Jq@-2 zcx9Q6+35q=%{!Yr9&^z6;cVJCn6K4Mh6R?$}CwAvrrkWIG(2mI%SNmf2e$ ztl=J2^%ac1H@r+xVV8Y=;L{NSIH5L>legCsn1w=UzfK7qbdmViJ{WuLb@l(ZvN0dqOgG7qGmbC?2g=RPKiG^g_y2yJJ|dedOGTmhY*|(n zwpJBE(&I^c$DPf3T>$)XG+wtbMCH^ETvB1M*e6NnA^;6kmQ@cjz{hHlzJGOYrJj>} z>p9@|>~F(RPP#nVoz2=+#IG06VG(A-XzY>(W5B2BysRaj zEwU-gY5=ko@t8>GC!JdhV+32gT!qOT>cp8EV^9wTWDWCjZGUyAr z`aX>oygMD#-!!*WSg60N9?5gmWwi+QXG>7u;{fi4?H z6@m(Dq3%@SaRZVen#2-JpgpZDg7oHl$t@KYlP#1_+cfoFQ(cGeB{vYk5mdnh0RI;e>17Db%yb#dgFCn*{foZ!ai7mMjIf#RdP*(89PAJu^+qJ zSfj}iPtsFihCtp`Cn4UhO3lteY&ODJ8h#~CAGI)4SjDKXQDvJ7i~8t1%8byuE8wky zS=|eV;B>ps1=l8_#~+ZB%n9snm-UPG{O*q+Oz6PR#a;8~V3W+wON|bzz)W+16kgN@ zLxp*^7{=jL78O?HU&Htjqw$yA*sIS9GC2vS+q15GD+OpAgh6zuolPE}=YM~Y1woj~ zW3yHj-uF=_7pkhn<9d|rx z?m|v{E^092W@-S3x3GIX9Nj94Frlk3crKE3j)qjoL?`#aX5wGg^G{Z!PCi3XVKGQ) z_ulkPP+@-kgQ70NRd5>b=)zVNh6xHpNq7+ni1P05rV2CO0L%qCe8)8p7FEDztrNKG z>W;3{Sc%E`My11tgui{|#Mfjdl4$ga#*yFkS=1za>XWtVgey`13+{l_cECC3cWP^w zzV86fr}eI?ux)^;KwqygpCqfo27$Zo>*zZDCNVimI7Aoi{??Z>J7FMNfF&E16a4n# zuxC$}NBea4b$+LcpTlwRZ}{f&Ebc4!AZ-_i6;#xdURN;k%#Nu9@&=DOn>wU029OC5Sfc->tM2638Phu-NE zSfYgPkTM}!5(XXe)*vIjx8jm#6b$E)#BHxXSZgR$=!ua9PBU1uU3kRTvn{KUJ8s>m=^+sfdxB(bL>JN+p7a zE{BCx@TUiv0`g_>pi}01PH)tYg!MO0%qrK(CGeihD$MAq*DOq=r?Dgn-EUxvdmamH zdR2ICCa86o+vBN{7zKKSd!N{eG|;^$axuD$M=y@}o;9>s9nI%k~m(W}Dp zIlWK65#_KpPQP5hu&-TLVSI!BK=dxFBv};(j`RN3;Z+ijiTAyW@|h`DyK2R!0qP(% z)wQ?mU0Hufo3O@raL8Sr4fMjr`Lb2D2I!1`d!cI=9i?P z!gyd*CD41UiX}+J-D0hiH`8OhTB*ZX|1;h77 zQcz($$emo+9p3cC5+wZmaKJe4PaR(+Q3QvB&P+r+y@fW=vx%KKuT~XyILWIp zq|7*+=amXZlAupHzJb{N%|jC;KXmYCU?|!G#}`TPQQc3~SSM(&9ELW~Y#4%3JxnSx z9!~!zc@?(L5!>W@NkfHQ_i@JMC!RXsNfLTStU->-hX8GI_CeBU=$5NzJ`};|YZ&in zbPWqIlDrB-LMn}9pJvx34Hd@o(?U*Q9Qg#ED4oWh3SgYK^FbAt#U`eyYJY-`FW;dJ zv^a`j`1HDCY|`hVpC{?7u**mtGoeY-h9nUc#_|1_i`&~OL{Zv}Jryv2qMR;RECWOh z+G-@2oCPnCwltDx!Z%wG{!C(kDl4aU3gB1ag#fGA4n39;ro z>jfo1g@h(%x@vb--CBss827xl7erCa5&fTK3%uwvq#Hovq_4uTP`oO#IvJN#RG3i| z#d{5ATR(+qqCodbSqAF1g9~dUb}af+wL8Id6FbqySsOttT2fm_a(V5PPe~bs9cOt{9IEaR)dORi(B*0!QwK zBzYC)ht+W_M{hyW#Ku!A1Z)MAGbJJ&rV1O4JrkQj-Oh1gjouZRmmJF737W$3U`$`a zkc;wt$<)*zNA4yhc@^ft@;B;{wEQC}6wXeF<_ClK3kEsU9YQs$qUWQ9We5Jtg+=;W zWL|11cPE&774K!LttT1K+@!C)U6uiyDk~c85XOo&kY%&p;I= zbQF#KNz<`0ik6DZi+Tb!u>hU4c;oDx98}oqq_4t)kUHkL8Rf0hty_!gd17C=mWH{b z!ESRAL1F%D+d!{iBT+>O9Yy0fq}R-DEYoh0dHGVgyCOpCfaqoNR&r2bW0JlKdw|sO zw<4Wp>$xey`K_^%?+y*~TU-RX7`I0irJC5YaTd~RB{x)~HU}6b`xc9y* z7k+`OU0U_$sGMR#i#N>C$w7rV9~&sF!mheuz(}CdV@_+Il1CdyRf6&ZgQ)KUiOUPZ zjlK23j?oH+^g6>86(->GSx&vTtU~MX4BjxuCkqueB+09=i%1=xlAWUGKHjA7ta@+t zuo@z%L1A7-jUk%u7_umx!JZq4irwOh3KOs*BHVPva$lFm8|D-QsYvtX4ijygGvd`! z{+M!CD#@#`vq&A^QkbQ{YVZAi8+|HnE*n_MPn4DWke1690Tfz+tisT>3B=-b*%cKw z1U@SW(F0=0gNR|zKO#s45kq!gc}&6El+;z&Nh}+zpV~zEmhE#6(IWgy@6u8ixkDi9 zNk%}SsmLk}JxCG&ZIIR#4JL9tLC6swD6|ZSsiZIfG0PBRoWsYY#uL1kle!8!j@5Hc z<+;Ml8=OXx2gnYf+`x3;tZh|+;o!T5Bvi)lq-SC=w zD`~5+9auf9M||anbp(pqsw>Gu13=r~LTu|oF6I!*hb=14@%#MUB^3rfLnGXAj*Bfr zbG%U=)aT6greRD&Y;Js#R$*JQdj3HD{0obl>;oy*LXwB4U?ZK(Ep!cAD*eKdMWq(M z&s{F5Fz^THd&kd0ui%YRV;=4(G2VdKobLxANmzxUd3_)N-ChB8)ae4CP&-K$(gE7; z;qiRPEwl$+EH5F8N+W)k=Uq}^?H&983r5BvMvY$7r@{ilj7z8uKD{aFs<8D~JQiCS z@Yv~58}zBNA*rgcHCR0(99)-yI(qp>ohl6Q)>z5oX~Qk_HM&$TLKcC{F8%b4#{eaeWHswjG3+ktQHk~Sr|E<{`ZQSId{})CVOYAcdfA98oONAW)PRU9~ z4x=V&ggDGx+74k6hS3na(RN8yh2iCids=ytih}wn0RhXFR$+WA*8+%#LnQtsp$M|5 z@N=bpJ$h<5ia7{%P%Bt ziYm@+dO~zD9t&6S1v6D7uBd#(?eSZ;R2a}Gfifda+&c?DqeX#b`7qQlg7DR28;2HO z7-$bK#cE6SSpp&KCvhh#P1b{*8V3^##*CwJzRa3 zK+mITA&5BgDr_IG>6qkdpK%IJ)F4B@;fl&9j+mO{2(|@L0Oka5TxAhQSzlcyYS~+j zI5|4C2I+GQz4aSW#$G_{8e!LhkV$<3;XMi}jH@VOpaOXSgo8B5&{SMe8N==Hfom%4 zMX2k5I*BDPLe1ydOkE~?Y78~eyRAxV(O(`7`Q$0|r6^<7(7K+hK1=-o-Gowd9+;`M zG!IiaEk^`Eo@ZXuAVbJW9Y5vum(Dc}W^@Kk2ufs$V7ed7h=o?3bG3)-?h^5ybEQ0w^hbC21abN4!{{!K|E8=`_gDQ@Els z7UN8;iGG00M+=!#>k02C2j;XL099sKcqKx^XIZTEiOcTI>{0~Ue$$Jg* zo}6M)bbQz+fV(QZM|Ng(NQE_WXze=+ zb9NuCZ-l7TWBgtYfz)mRX%=$ASRDXgRl>B7~Uv9R~Ly;jx;rw3Kc>tMpQ~)t|vYU>}5BY=IB~6N^ql(G|UUv)K zQ(+O1YY-nJ9N<7WNpuk|;*Ih~ks<0LAsW3?H3Loy<&tbsxp|B(6alAp!(x$>bF^qZG(pSTw zr7tVY8SGp82vBB-BZt*!P~v&KLYEbIn8s}RMqDX&AVFI#Dy$o-sC>oiZi9;|EDwJ+ zp5Qhz!Y7DZM)+g18*h*c6lB5wg|5*Qbtc@_)j~bKSdVj{FjF;$mi|d$&LH`Kj8I4b z_IUMa0{A}9_5h*iz*C&pIIHy2>-~`Npukq z6Dt0$APZk`+CHBHTl%UvwDkFeIr|FI>?w$k(9i1A6v)riJw2+}XkMXx+Ri$^fzFFh zX;ERNP(_7n1={bT3agA@bU5Jgtipf)`WlI|-7k1Of^P@{&Wrvnt+?fCb+qxbS z-MR{4p7<^v(OJW}IE^Z{Om-~T)VNY?xV;QoR2Vdp;t{A6H`C?cE~+p94h3e1*=#g)p!D)hjkhwj0I00e#;^DSV~c^6gK%lyfh12C$*%k8MR=o%nC3-8rt z&euT|99nvDpIlHbz7d|;{6gmt=4n3Y35lD`?=uA!axq&@NFW=m$rQ(;BQPP@2P!p% z*V_#jRTv=3gy#(uZA{>UC1(*`gRk%gc`R9}uyW3<{3CJBK&bxhA?e5;ASDZ5thDHrgf;0qphQ8cR218s_+Nkhw8T91K@__wClN13jF|Z7=T=(q= z7gg9W?p$Q%H4*KD%uQK*U3d+K;0^Msx0dQqb%y zq|~=a@1I(Td&}kKt%+R&US#|h;Wg-lH^@*=a-yf5S$PTM)O>(uPr{b2iTXV8;g~u= zRG1Qm5-ZmHAjl@tUXv67A)BhKgp}e%=;d-#h0W&9#ZXQQg=0$b%z{gW*8npYnQwAZ zVNW`<^4r8YO9N;&8Wz7G$}Ah(mT^W?{>ooK%=Y*WtXIB1K3Dy94?qiTj-{2Pe5n-af z<#kdEpu%P=vA~L?=mIef1Zn#cf^b`JutNu+Dy)W^DvawK%&~uL2R;vV5ncjcL|>EQ zQbGXBhBgUX_6j(&@-rn^`-9(Xb(r0MIl@G7^Swp~0Q0gq4pm|yFXs%%eXu*NO@)E% z^v67&D+nsAp_?k~n&K1x+i}Hs-eVFlorpmTWu$}v7U?q*TlVrfv+^=%v-9}P#-ca< z!z;EEVWQrEEc>hYg6T#65g;oU9gAwKHWl_QLHm z)iqda;Z#{p3d;89Q}Y09*~{b1%F94@@v}Sx`*58NV(2`=MAhbX5{+ZSi|Z_^#=}*5 z%U<3ZbmyxD+EmzZh!8A;W$xg`NY-~#g+=ma;%%TYbT40;1ZgwgC~JF?5jo^!GaA{4 zAeQ3t$#9X#F)~SVpYo&awC`~~;fnq?^e|~}K&Z!lWp`%bWe^(La$AjehT_8bV)#Q{0Lb?D>>{dZ6{ho)Ip<#8QJWZnJfz=);L68n zIwlo&RfXl^aK|*vZh?NFC14TwP>XQtm7G*qR%aHzrvz*KsFk76?@v91nc4!lVQ}&T zs-A*ArndYeNR%6C6C+3{MKt<~&nc85mOAfwSei} z=02aEg@rA?^Ao@owQI?0A_C~{weGurH9af!+$neGqm^F?k zl;mO7uSJ+C&b{`mhwGU=Se1nU+W8~clwE~}3$$IM3LD0kffz>tKtLN8SH#WqA4X#~ zZ!;NHH<(J!=RDpDlRxBi8V(xd24#65+dK|d(ja=NoLP7&$E*XKMpq}U-d%Rdg9tO_ z7sl16gVj6 zy!INf9GvQD3;o}mZK8SGVjg@V-ozKZ0B087UxqhcK%G}G-4b@m8yHh{n%PA(E-gD! z@V>t?3kNApo0I+0@(E4RsKNk%d$Ax7ySU}JNrH~Mslqsq+YD&!h@7U9y?Bqe!UR!I zu!OlMbcjff(NTwhhxpmN{iN3MwqW^tQoQ{NH=6Y1++q9?ri!gm|B5QHfmvdWItw({ zZ!@C%bXhyO0s#_9?T9JJPj$fkQ15k9g|Sg^{23RuKg>m+XZ2E3W(@!iD+Vk{$W=W; zQu2sl%tvq4M`jsA{uHbr7iY~g5x)j^z{ra*RqRY-(J2})sJ*oDX$Hv3Hd1Q$X*Wo$ zh$+a=71je|t7@a(Ox9R#VcpF|pJMeQl_q8hfc>XHgVY5q_iUtL)Io1mW*N8qC|E(x z7WFZw)#tp*9%C1+h%i&{^Uhdl*!I+#>MYP+US9#Ze>B#p!q`3nUN7~{70)>y5RjZc z8Yp|84!58=Gu_*y(!_kn=RU&c1Zb4H93ZCQI5?R}2J7+Tp`toZinacnR-f=H%Sp$M z5qjT_*~2oPV@^;ibr$-7d{!DzN|n{A!uD{5AU1rPAK}=6*VSA%RT%p$-ATZXz{qSI zpX2x&VI)Uq1HaGbrNP>S zzyhrYGbTPQ<~pZN6}A+(`Tc`=<=XMO`ov8Y#)gghP~ee~VU0Pf<&z&{(P2*H$hdud0qb#ROOfDN?=$F(W@m1XF zNh?@GF6Qc4cB896k98^xSYAGuQ}rz`gCIr@M~b!BibM7JUMW>BMNG`uHR zXceo6v7Gcgf1s=u*#Y}=(Chj#ttzY*R|rbS7Hp-wef&aQRAF=I>#G9-iF zjnt~b0=YrPaWDm`c#*~pxqW>ptA8GOJVkpgx}~DRRo_vO0}C;I zPmhpfy{-e)`lj$Ae(HfZC}Nnh*^gHiAAy$4!s!sEikT(y9kx%!DV)_JLdOPM!>7mx zo$gvy7&psp!dHYB%2$`$RSRm}h`m!-O~<}y_;*lHVN4JQ>LtzU*BV3Tu&wDQx!jM6 zDm4pc?$$tlGYsQID)eT>kxQQETApkq=^Vd~)#d}M<&5d!Ruo~XDzQ3vi~S}y8E3VK zj){A0-}g|w+X_Yi+IH7Z!0!W+p8>{#&KjI!Jy6a4+p@iRWl!>Zq99&6`&Xry-!ZatFW0YzBY;6 zBkArX!IDnM{2<8~WQ5E7Q7UVi(!@Ja#Mm?%B=3^w8h#d2ZH-VJNrznK)y#%|RgqO1 z}m8e^_laez$&BXny}kEb%RaXTa~`L!jlRu$HaBLtJ;3-Z!6fKtVkp0@Gyf%P-G z*->d;%(qb(D+AVxSU@w=b^e@1Y3^Dv?*T+YTxFUlj^w;IDYw@h? zorUg^^y7o9SC&6FKT$Ti7?~dwl9LMiM{p(MyaW72aK>bhK&-Se0fuG4 znaDb#AuZNHrk$K;9?%=-?b*#1f(8i%Ibb>I75Uh~cdcQ>36kN=&V5}IkjHqG&^^-Q z*+}7`uk6sdR-pjU%Zrf^N|7w+Z!rxPo>8LSYz-e&(&h9mdht*Uuq;*-WU=2FA8;h3 zE?*+t4++7F_~iCLD_|JUF|2$-!K&*5@ju0Jlp`6buw!B>Y*&1}skl2iz6;##QlQGr67`(Xi)wJPc(smccfd@&xX{5g7&uM~RTOm4C# ztvt<{zqTv`yx5^Q6PM4w+%5Z5GE!kj#8gL$zF1bUl9{GPJ(^6fDu zi}4uWK(kL;G?7sliW5o_9mv>leach{>>J4Ri@v1F#~ak9&nA=!<@#Mn5Cw^=m!M(5 z2$8l~@}Q?T#g!}1-Lh{;au9VlK0#%_chpxzLvEyG)S7X(@4l6Gu6 zFU`S|oQ)Wd$?wZW3G6S&K)RbOP^a`!zmCBdJh8;w<2kQmWNIkOTxX=cWpo_9^Da1M zwqs^yW`@KVdmJ;y%*@OXGcz+Yd(6zt%*+ro#E#eR5B_`4?%uN>Zk?HKovv2*qf$v- z-IA)ZLl7pQ1CP?hQLx>faUV+f%}^GwQ=7?cJ<=~&c1l!2Sfd8c!LUnl`0~e~J0Aw* z#re>*HD97f=h}o<+?u4*0~X@GEd#C$GY^88&(M5nG4CQ>MzOW8xZuiakHRofdM%~C zyK)_6aFUduG)iL#hMrJuT5%&-md&9;C7!(GfUU|Rk`)`Sc-8!p>p^~v7wew;h?ft} z{*#CP$PbRt@6KM1Ulf9rY~p4-vHSWsR`hnuKg~sD;#r!DSW1z&DcFZFV#eB`VgK`p zPCi3GvU2>Km^?f97Y4FdjzTmX@TPvt7o}yA-KNvm% z=dUl(+!0&kg2bplGBI(4(cFcr;y-l%Jvy4fz|sc{U*Teny**gVVu;a-fs` zqX{D~CzwFPS(>-ydar>TBabB9ozIB*5_}I;cL9At4(<2qciGQiC1h;I>46pmmdUTn zfs}mxoOvQRHxD(RNF-ypif=C|(yNlmNS`9t!3~4P_ImI7Svgptr7YL{3XFD)P?dsb zQ31hi-%uWgBmp9FWZg04R_nQpixJ@LtoWXefY4JM69)_ZLq?s-F=#2JY~!PHo06r&Cf%{fbjzB{v;RD8M= z$jX=Cha>xA6_4)nV7e~SDDUyRf@8c)9v7{YswXA(uf?iLn{K}5isl5JQQ@}f!SEE^ zj%;HPw7`MDhSZDD#jz02Rp01k&YS2_;y1<0j87{XF=XJk>LCxPz5I+ zs|{EOgw{a8#^mcrJb z@Uc6wSOWt=Y`o{hVAxx8aJmcQxaAlzMcCRG;0BLe`&?%Pt7ecmb9~;<-^t|xeo-~) z2qi4VhUyYZQlDy>Qm&O)$dz^CBn30?CElhxhu%vio5e^Gr_d{AP?BO9r*6k%IuGpi zo!<4SCkC3P7R*zCDXDX%YQx^Ff5mVp;+%(_K~;<|ZAopEpXUsf77uJfCQYGiKen6- z9=Ah8fXZM_B$p2d5EZ?bwQ3xo`~0QQ|4E*=Pm||ZFPe=nT)keW{ImG!W{Vb0GEr2h zpCD_fS7k+bBoFWq7h>UQaL`&1!AD`|he>Cc11qFqZDhN)Z0s>jq%X!s`4_t%Cz#2F zxK|~0@JB3?hTLB(S(yyJZ(f+w1wWQ!FoJ1PKLZ}7&;^VnrNdGysn)<73Y=!GXH^{@v-;|Q-nvJf+Y_~_+E@b!>9|?TO$>-#SLeMC6Wr{qtkkdb*c$tQhUp|ZP0qs-#k&;DV&A#t$z!1#szvyY5_JOR; z@nYVz)DH8cK;_o+8(X1gbIot&DTdR>N*X7PeHEaOZr5X>4FItrK~~g`#vWs!m5E`< z%X*}#J0JZfz{f1*GmbR+Gw!niT9Z35AS${b5?S>mniTt<`@CR-vHOQfN>Q|Dvy`~) zT*kC)d_~-|s)^2UdTJxA*mD8PWH#sU_1X*vm1DWKQGu-$hM4Jd$fB~($}O2(|3O4K zI{MS0y|oD;u+-Ht@-$!HRO`U!2{klu?Uk|U0=ngtIOfq$x9GcwR|-WLNzvhw7p{UI zd3}43_|ehwD>{GV0qHXfYkI3CrSSM$zSxcp+@4(5UBQATA(G-U!sl(|kgW9|65q5& zLL;qukU#Ij9C>=1hhtxT%KuC`Hyy|{q%g*`D==!FC$3C-Vnn+du5s)B!091ov+DV0 zW+NX+#x(T-0KswxqX7fHj;)1*#r9h;SSP&uXmLlLQ-~+o(d;l`rco zarH4jiVF;6W5*cwA-u}$ilFQauQ|p1PPqrSmAS|lAdzPY z*odKs4NyCH8x^DZS|%<;cZI76+QYd^BCPD~tIua%B5T3^CaR!@f$@s|)+Ih!69Psp ze$SUN?tLW>XZ|k9)-++Fq!v{4MyyR%g@f}WF+F&!BDVSOxvei>rDX&BUyq<_E<4fq zjr%b_3%b^`A~Mj(bTXg9yrbYEa%4NPnV6f+m>jLfs*Y2h52ggdO08i`Nr4Q$#8lhy z3n}|RPg~us_S#0D-%Tl()?PE_@YD!02F zT90Qegej{pjBlMsUsZMBz<$no3YCUK!?qvHAx)oK)0oIFMD~IXC$TP;H@Uf}6|Bma zw!M{Y|8{QUXuZ(uj8TaoFl+v8RC--MWfgXtSfcG=M@v-EPJ68Y-+=PkeA$gxa&vy6 zl&n0W*P=3VVIr9h^*@han=I5pSQ z_?W6#O~j+y!AP;?V8FLHMGA|;Ayw0bd@CXnxRw3#{sf9|-q#w>>m(fN5wF&+jk z$-#LMj)5Ip(cg)O0a9FHn-NztM}%=|&tt|io_jKIxdUC+rpsref<9o6GS2G!v*tG$ zaou2Duvt1ti>k>{eZZ=V?VoWRq=-XP2)2$NjQHh)+Ww-Ajd+kWs^KU{oESnLj-7iB zUx1YZN(4It0dj_(UC(o5=F35)Zn*6lSAayvf8jRL6;QHs}3V zkv`brGkuf%@DHh9WkC^rzn4a5>}$*Dq=9OFBY-C^6tgl};Ug9a+@X0vJ_p$I9C=3UL7Cr$MC@pz%M zyD(j)VCyZrUwQaxWxMbj1)Lb{1G4xm-2hlT1ayl#c0X^<)9lO;Fu^b~jR>Iw7se7eeVAw-^V`sY!vUO}H%m@e{GK?<|C z<~0D%m-$xkex#fiO-1Z9XA-B)XZSS*SwnUF(V#0;!>JSW>DzxGEZ#E+qopBQ-GLgJ z7~Kjf=a5b_36iye;ySL0#UC^#>YLSuc?p~8w8B-$AAz2ToB!zs!6N5qS>;}Wti|jv ze7U3@MvWhvVd2i>9XBWte5Eu9gwF@Lf@YN?_2q#Y0gRR%n^$`H!|vIve=%xXss;a7 z5V%jwy_ZcjucrISLJN$@-v;WFKaaAX@J~tTB0b_Y`fK7WEE$yFcGu)~q$6S~b+R{Xb(nm1i}_v<%Yjp(1(Ew$^^3T8#$wC<(Y!4ZoZ(G ziI*7i6Zn@xzD4N39Qn{nExT=jTt2NVeQa@{Zxg=brkvPmhhM*-A%y7IOo|$O(*-@5-BFr_Db2Veon?^*O9ql!^ajmlE;nbLG1Eb%-(^7z70ssCIKymP!=T%|6n5YxkwWIKK^ zeqPM`B*9EGE_RF2yR>-4gqk+ACdlMlw-AM*H}qRhWrWFJNkJDZ#lGdQ49iWh+Ru{g ztypVruktZjTFS$08(YmiUMUBqc+1!Y}U=t|Ws2#ITL2KM|YwxD(L@G)^h`cZOonX=G9$gA{#9bw< zmPOtEr~i_iBYAclC_%*bC^Paf=Au~*<04;;+0{0 zIm#ziO`{Mp7mU+yQLZCIJi*L%;ptqjP(e)w?Wyh0^G{ zEsKMlbG}soAEcReQNY^D_vxg@Po9@;bQsQLv0fej+S!>$bvEvmpsr~w)vnQ zKhF{b&dB}bqNlAmfooq5y^ZwZ54P)I)H@m8s4r9{iWtZarL}}rv5xOoLL$LW>lv52 zeTBU?l@z!9O^urdli#0@g>D#bp45|)q@G^5(M#H+J`_?aJwf17huSQnP^)$Lt z-ptQ;w&ojK5}10E4dn((vc~7?Mgl__AS7w1r!j7fakx*Yw@^(K3VF?01PSo)P|8uf zC9~-jE<=oJK1+mfS333T4=CQ&m68 z5c@KfKUUFAmDLgoU@oaV7a+8V!p@Mpb6{2c*|W){Z3xgqy!9;wV}-V>DgG*-=)d+J z7k1wxI(qPy|2g6aPJk2s+M}lga4(_4mUBLix>&+DK1o!QxMiCmv*JQ<{S#fs>lqo> zF#1BHqO;A4*%VaLx02?DU}EQI5J_A27vd@O(W|aiZjuGk=_7_2LR(>snv+C< zhr00|2;i)Nflz3^Q}*SmWxa4&kxXXAeU_TzSnzr;`xx-Gx@^nfwARRk-P(w+U$mN&6}4~o;16h1J_l$j!y5dm z!>ZWfIBQPtF;O7$;vD9}cz)0R7Hhbj3-WK?Q7pF42;R+_cjvPkyzzl`sApWV^o8K< zQYYnFh3!GZP5tuj0C&Q#$k;XBF|I2%FZWn`7XxiyZ?fULn&TFpMMbx%z*!W!AyC%E zH(uUyGgq(l(m{wQv26efqdBKoi)bi7LiW-t`@KHkxsz!;Y~Z%bY)b8v2Bap6CRM8| z*4ZdFpCD4_b2zHWt<|_3>1w#B{66Da5G6Og#Zf(&4mAgQJ1w@i1lnkm4YQlre@S#~ zu>2^R%B55+%#}av<r*oIswn1#uhBCX5@yv2j;xBK86WesT5h z7d)J3wr8Pg@_$+?BK-m6{iX0bnJ7kq>D??N0>w;R8gEV#_yuq3k^2t(4;-@mscEW# zp8g*2QXKDvB5z(2=5~vbbfs(?L+&ZLxZ8<)mBev^Es~K>fXcL1MI8mCwDJ86aY+px zpS!@-rZ0^>LrvUty~p)r?n`mtD#LTGbnx0cw5a`nBC*EoH+-YtU_WGa^wvkelkzrs z3{wC}A1?+W#6t^Y3p*tR_YQwSXtJ2kh*Y#%cAy?zw(_v_@u*?x?sXYfuw0>!4o{rH zVyDglS-65^Kyv#NtFY%RA;JZ`U(CMl#L=e98W^@nW~CRrBcAYlrGK>AZZsKRb(6tz z*Mp^^=3U{D_O4?>2$KMjD?VTwZljIR9-r1!ACQR7R$}|RGz4!0_gkLu9ryW%0#arv zZP=8cw2zjqdn&gTvxFV1+x_zwWw=M>brX3T+*N+m#b~W(90=r|o_x8gfKSbSwVzAX ze|Nc-46TcqAk_pXpDEwZoLsF)NL{sLt$`=PWp z-5x|P|GbbFes0h1G6Drf>ZQY#rTdq)2ex#CSpyTlZ}f#D69VOaPGK(rv$@W$2)sO; zWLKM4B7(14PqB0y%{@(tUAAHZF$`9QX>_3;Z+DLuD2_KbPiL0I6jn( z4dY$o*PHx!f)#y_ix9qx-`;mKiC;N3}(rY>$ee%$^6`p)6lzV@?xIuw9XSrO7RCy~63 z?#6jg5Wws-Y^jb=;~tdjh+s2K#%V0bE`q(ml^=sb^jQh>%)xGdgXx9%Vyb9FeXDM( zGHz^8i%tur-5RK?M)z1+a=S zm$v6XYt;x?X$}Vl>NgDmMX=47`>GaN%YpU)%krrF#z4g7$m3_U!c}b6+@S;RAL1@w z{w}JUt|0`8vh=zO1}Yop3!#$ZK5*;XJtu{hFb>l+5(JRquel_Z1=g7^6cJDL%s_U$ z*^<8zs_J94z*7Z(My&3C&%vB95ZYykIT`p9f|hPOOTtHm8~X_HixKiPG}U|`QvVF< z*FpLqJU{Q?k2;p|ySKp72!vv>M-#X&1H#RW1%C+;kwa)DV3BUhW57vm@2oC4po!v& zWi|DDY<&9)24>6w&wu1#grZA4Z=uO*AS|W+)O<(xc~d!*&on!J7Hx_udLDuinVFPo zF{cRT>N_U6!WR$I&9=&92a7do56B=X!58(Y7RDx5}m! zX>LONDq{VrPxOP^%6^!UA()|LZltjXfjK_IO+c`}4fBP13-sxh4V02npCqVJIo$88 zuT}jQwK1!=$mCn*vbN3b)tLJ>jyb1}wuqRDM*t~*ZYC-&F4Fz*&J2EJVbc$Yt&GB2 zzOR>7{aT5STmt11XrnpIvbCoNH7 z+qkv}+@P72bhQ6))b~YrGp&yILvZas1fo!aI|SMf2STQCB1&-%DB=fS^cW=7%;JCR zo%NQ*)SCaZ+EvH*^|5qD^bsfM+6#0V;#1q+J2YT`{@MPnHAqQujq&YOQaOAATtx=w zDp52kF_5UveXy_b@!cD^*5Hk>-q5}XQRo#&=LM!n!V5{a(I?A)L>+J&NO^>j^4Dr1 zBHnvr$SE=(FX+(@#;!|GEhtI7YBeIUhgU68iTilfO1#gYxp4o!Rl&b#gm0jFJQ=6S z2+ol6sb;`0>BOtk_xJqcEeQ;o?t!I5{KaG8e^;$n+-*JYmZ{S5IY&Q6xE+|e_G4kc z_5}yaxiqw=ztZdnb{0NMw;%B-W3Iqa2~;)s_L81n?iw~lWJByQy8PXH-9FWi^xMhA zC;sHQa$>svE9mo7H(!s@^p${9mIo1@<1bYBthVqjO~(3nQwbenSD58aSLsZ)*PzN1{}DmLjalpU)b@n2F?hCW}Mo);@n*gsH_sx>0jQFe0nA+(P;4K#P@g2D9Y1VX z%sk`<>O&Xx;+-`(VM4!AB<6AXL5j3$uf`t+v1UpoW$Xs}V>v1cD*JoWbm~y(x@rZv z-y&_J&22tkH9;7B#RkYO=I&+rnViu#0X@Pe8FxZk4z7V{y+>9BRTx>jJ*zw9 zpLj(|lnn>>86P-7b6%QToGm}K+M|`Gy-G%llpRtc^l*9Bq=O&G)d#Km4p>ouEnSEh@*%!Uoh3?xulQ4>i2f_4N1s(;t&kZzjOa3_)ZwnlJF{1| zJd|Kd&itWl5MEeYnpl%{G-9?x#%4V6{#1aj_mB9mk$P(J*esS}637k*LY3J2Yz< zuFy;0=I5~>qd~J;lkGXua5p*)hnUWLmzGs@{D+n9w#_fRt?V+Z^`}R=O4Z}PIX(oD zlEW{_^n`zaAJE@d68H8J8*Ho%)8=d|5?K%j@-1HehJTEfNP0)K%=PDZ zX9sKyus2f0{x(Jjtg1?J-;_IPK!6Jho>#-+2?Z2r&RpQ?u)h=C^S%^s$n zq7rZSE5GpvSjWneIxMxPef>?98(SJN&&1nZ%iiN47})?vym7J+^*$qkC-oDZ2iAse zEaVh5KB>LBjc}Ylobv~vl#oi6Y=8IjMM8`a4r}F4iICmz75Lr9Aj{q%2`-`>(}%V zkN!grcC0Ptm#ZJ+4|*^vTE7)y;<2Olaa<s~ABP{-?kn-I}PHXfnwDaPvC~3lgJXN_3%XD#EgMJQHi)y0{@)PPb!(7h->d zK6r59=m%xnO+uX5%jh0cD4*W2JwImG!5Gq2GhvMX=Ka&V^g*(@(_(OJz>K-aYEKBk^pf zDfaaV*EWcAKqV6p7N~*C;iiU$%=W;-{HzF*fH{5JI>#_kVf$>PKeN3av8uGvzw^R^rCM}nqe zKb5!NzW`3|CJv=@{ZLi&^tH(-gNv1Pm^vhdqHGDS71}WeN$lEOIjZojB2a%rz>TrJ zODq}!(QNvgtF8&fF4`sxOMR<6i6fGba=!>g0F+S$? zAPu+gZcln3g|(>rgV$jDbL|)fC?d(*Q#Mg?i7cnk>i2Ng8B>(rwGb0V7h{2yr~j4o zh9{&{ak|jZtZ~-Yzt6UKTJV~a2&1dhwiLI_y9CTf_Y@Yn$++uGBt(s#l+-^NEe^M? zkY9>HlKqK{f7O-Fs$rfpVxFoW!$5sBs95O|(5?qqbU6J-Uw?A`5Do>)L$FE-md=&1 zi9(H-mc;IT636~|f%SDYkJp2m+Wvvp@^_!>@z42$!e?cHCFQ^1M#ncgN6DJ>6X*s8 zwfd3AR+s6lZo{r1BNHE^o)bwM335~aC@p&Wfeg5_bJ0k-rHY7&ru^LeQJ|i0?SyH| z1(XLzT=8gX|#E>E%Kv6jC|^qOD6e0@_WXHr?8E2 zl6iiOty?Pm`;pu#?{)n9i!KL$je=oTqG9ObXEpTP#<+Yl%_+^6H}}0)t(zL}etF%) zvtJ?R5j~$@L|Q*!3nth>lzT_&z1llbx5P=t8OFG#ca8|S89sWpJ5 z(jAGwIbBaey71x=ONT8HCWQvYtd*&q~^T4L`8YSm5-$R?c`MED8F`BK6q_ z?SCJA%y^c_fMp=ICbRS)arr~^F}H*_*vh^@g9-@dZ+EC4YBW#oN6T|w1YMeJ#(hMN zx95(r4ZDxv9e_$%g%ZV4`-Jlb-wVV*+Y>ayDV$2AWtwTu)GIhscmHBO*af})!PE*u zBGU=mM6#+nW^Py9b8^VOBDmn4dhDe2f-XSmAumL zuBE}LlLsi4{8;i0nfsjXRpk;e#N1>xU#;{MU6sD%UF|eGv9l%~%*>08#mwVKS5|i` zmXz#0G@&lw{QUeh=FY28$pLMK-)Y;>ijtagS`ab(Tdsh!m{r&9Gwk%=A~7b(oS1?s z{a)KYg(GHUMOPo=IoQ9NW2`0a)hjtI_5;R!8F8{{jq?>H%_@wU9<{Q9Wy<$S(M(W; z5@EAb8AsftsQ-Xb>bQT5+xqCVILuG=Q%4i5bQ+EagXoi)GN0yCnLs_rKfA;;kchaQ zk7~MYSL}Ze%m#S6SZ%hr9M%VQPv9+A2$P;}FjK-bjwX=ly7)_{{lWmAaX34YNlbp* z_ZdpWmCC~a<{61Wv8&l)JylIQtQ~-WrFxI#JCuM73 zfyBl_!b0*-LO_67#N5eI-q=CZ*4ob2#@NP*gojzw*2>mF$xh$Un1qe}^Gro%33Dqa zV+Uq2QDre>Lt7(bW=UgnQ!^(LPEK~_AI3JOPG%&WT)aFi%t|u)j+P{>y#LCj&%&cZ^^m!U98!NAn(0|~-f2@4$0KNbvq$B`f-~h;ea9RN1;|hQ*{0Y+- zU78xso?6$AH^u<@|4iHHdBNbm^>sR{8($f*AzG6v>P zV&LZI=l|#b9|A){K=^|E1s?;Sf{lihhK-+tjg9TU%>UVc`-k}dwf%o?!2eG#@UaPi z1ju~!08nAUX(9U|0Av6#R4{N<@Q(qod;k~#90Cjs3;^*@k^doR7zjuxa9A)nc!Ylr z|HA|SALc(SFlZRqPv#c@I2brMBrFsZ0uq+mKRLm`Ay6TqpwUQK(S;TC?fv8OU@*w) zdNJA94Q8>)MHH0`9h~y(cdyx^Xg`I|5ZXvEbM;H?k@4s;8{)>0-zkqLU{|otJ4S@J5eK1sTRKQmN=ooG|}wr;S4br(Fp&T`94|6YYL zZcMmZX1U@3FEH`oYM=3nkoeoy?Dn_|{sm3uB_&A9MiVI1J;d3%pcLU_vp?7*LnzZS z7noeWDRWUN=?4Rq`;UD8MaYalt|}ePc9acEcd{PObyeS*uK2%KsqbNu^RGMB0*DCx z{&VL4c6VfsY<25JQ6WxEwP28a-X&BRqu5?Iag>KmozJHiAwPRD%o1J|HY|%4-xkSi z>|Nm~-Ll%FIC;CFOskR-sgQNeU9WJtek^FOWu#C;wYW3qja-EWK;gUTs#<(uZam1m z2LK=@N^m@mQl#rn0APnxw5uT3LGN6m|6gqXThRaCjKQ#h`QEuCjq2I(cF(sSZtG%s z{XTE&$*yZ7B&9F;z+4~;i?({xCc1iS?%FU=a;}RVdeh?qaxfwUV;uDQ}wt%^Yr-i`TPcBg8emxf?TlPT}s zQ_jcCKf1>|q$0=D4-9-1yh-UBwK4L&h zp}=1%p~^DHP3ktdFBrEg!Y?Fr)FhpLt}8tBQ0m#G)j!dx9d*bQ zt3#V$y|kN)%7bb{Jve=5Vs||iP4PA_c`TqrJckNPFOlt_q;93PHkIjYQY9~e-Qq1S zvc~Fn0!V27kGRQ_y4lpdgS+hG>uELbC_`2(k8$Y|6P2u2DP5+H+jCX>UU<(a~N9;(fj%zCExHV_&0iqiJ=OiP9EYr2f>9<~Xs` zpq!US#9sCfCE-sgc3Wvyo^YHE6u?%4Lw)h>rzvdOXF<0%XdDXsS58=&F+Sv zEMM;$Uyq4!q$o6{eCs?SChIsklp5t2JqB`|2@f0EOy|^{Sso>^I!^NtaDK<;qvKv% zP*QVfW+tr*n+eCHZIuA2X<8~JCbaK$O-ucGP5b(5>lct9)7pl7%!yP@$hu>mM}s(3 zWoT!^lIs{)5Mb+Fnf`>x3uQjdF=0r{mAIH-T?DEc&VqYL?M^06eH+mz$uwcHwM?2j z>U^iJvW}bN<;LUo+*Y6J^5#_!4H4rAP9qNb!D#e(K**kE0FIZ)Z~1TfXkpYsF;Dx?IXdiAW=moYcHRRz-otKi30)Fl0miy* z909B`>lB-opx}HI%@|6&l!vSeg9*FjB0-b+aDdMTAO&B+MYm{7iZX4xCy7?si6pOT znSI$!Gm2rV(`%}ez*%fx)_H+e{Eb0l4d^s|yR3Fl_V&u*)E3Dy)1y>eg|j3@IZ|nB zo^x1TGph7*@9Ei+d0-dhw7J>z455iqxMI}~6O>=T2f`vs+LzJU64{yxmMe6#I&2a7 zovE@xS+|~~={8$QC_n`s5UA)_f;)QCOzC~h&RRr3K}w2QOpW*?+98eN4%EeAQ;u4k zbR3Vctp;;xlkz>aQ`2ud=E+FQz2!3x_f*9>T*oKg->;0b)@jt!ZO5x%hq$(@7-JgD zk2H%6)PD9kD$deZhdgBM=VmayXmx*#5%o zvpn*4B#5U|c7)T>d#SM&yY8LimSQ$)c^AVcf0Sn(EoaNO_AqSAG5PyNT+rhd^dxSb zaPplB&r+PTU;q5mTxcEEA0IkmeXywlE|_X?s8*6gRhd+QO6I1bylIQV0?QV9mJ^LC z=YN@TF->oa z{S9+7=?ZwxQ(5##a>d`vsw>ydjZ3ovn0zZ2*6xKH^qlxr%Zb&0V%R8s?6-l`oU-fe zI#0*=_VXI3nT8rc8VU)kw2WC?{Gbw9Cu3T7d8c(R*Qln)0#P81NSC(mPhC~K@y|6k zpTjQ#$w==e=@&8{D0nj3gndep8x+^|Ckx}}hn%G%cKtjho>bLKOU)ZA%VU4Z@4kxd z7;VeYzGob#wRuv4;uOj}>|Hb%hd*XNAFKq741-GrBFRL7Ak)zDvTaAH>&rGly?asS zzCqggT}tcv1e@#pN5NAr*GXp-20loir z6Of{8HHgr$iWNW4dgo&jpR8qt99z?mRy25k^}Bl4Knq$au60ec2En6MDtz8>m4psg zDSi%zn3!HUvmJ zuK;9P<9~!koR*?DNfMBwml?e-ZL-=2Ju&(8+fvQq@xV!dT>DfyuYc3tR_i1?*YesB zGX@vITOfeCk{=t)Zl;Zs18e(rex6w;Us2AS_rB1SQDq-h`ea&@&cXAl(2B4rR#2J^ z3YNdw#vW2_hLtNv@w9Rq*>iBS7x-s+acYFqPW^UpkzFU+@(v8(oifv_1ySSGE-zTX zNpbryhCooxtf@~>SnO%-^R^deRS_{p*Jx;#G}dEC?_WhO>z6v##7Y1Wpe?6bDxz8sbJc2 zn8iSFh$*9NF&n;)O=z~+-hRawsZ}km+GuXv^<US+p^!g-5yo zx!0;6&&gDR2j+(qgEVYzGL?Ru4i=_i^4*lte)m{9l1Ij}E_CcSz(RqM4z7Hvcwt@j z?Ocn&(Bvyq$_QFN_~9AcWA3Uz>JMwx_Y$NtXrG>i@0FrZJj?zvhs(1Z_A5prZIJucc!#)lvh93 zWTPuK&T>`XU=g!Me0*p6W?lAeiKnjlv~+%Mxv{&>dzL#7Osb4>c>KnkcS%0m3IKp4 z@EPgi=)Q&Trn8x|Uh1~k29Om)iAiY+LJd9HG7Z#7_3ACPr=_lQX>g>SP2I6{TOzBd z(X>rs9D7UYnr}kRTqljj%GB`E&~7p}S(PE!izrfY9Fq3ZrQ`WM`!<)pldQMM*|7AY zdeHryml8~>Dnuy{u47uq(Xx_3h(N$yloa>GtWBA5ROSY>9 zjOx;I+i`oa<Ai%VVzl$SN^(~>oXWIwWlQoMv~59`;7^0v z2W`Aj%F1SrR)sYP_RcA7SRSLMK8|}`?C3<*ub=kud{?1r)dxU2H*UTJPeDGsUWy_G z?Rzdk2FA%y)@65$tp)%l(C@Rm_{uT4p9~BnAlS4hhof#Tv3ajT?B`M^kibXItD(EI z^ad${4%;)I=y{wCYMP~;JqQ6{yDNaB*-2wf){0;98jYv4yxZ2Bl~ps-imemH1(w$< zPW}PdQ3ZY>a2ghT}9 zH*anU0rTJM5{*Ipeqg&Z|Lw&7Uwd%Y(1FEq#}msb*=#a6itYI9!-gW)>>+OQ3KgXW z&&>`vYCtcdGkj7}NR-%_xb|=qd4JO!d%#U&(kk9;rHs}zVdP+I7xHVIT8Hb!>P-8z zvP}k0mI|%}|0)#>Gi{#%%R?PnRrqdV$-&Um!rG&wRZ~B-ezqJ^>aN{PA}1Yn&Cv7Q zx_3sfW;=|E1TDzA>FmO#_lP}mac^F0>Np*AI-^#`Gw3i{*gg%1=r=6e$yc`&xKGhC zFXK`$XJOx=^H5inOXNQ}%HozuK*u0K9%ADvd%2!u&G)_Xh!Nyc`MJW*d`?MAVotxa}1N$b5#mNS2! zr;NIj)9E$sEU9XWEDLd?!4^)lysj*sUXP`le(q(#{;p!^&)Q7mHyDY>DoQTv zxYSO6xUvd>P161Y(mHM)Hd@s*v{aQ-#z6awM7;HKZ*$@jMGuwKxe6RmNPm0ryvEe& zQ)63H6x+iJakARK6+}Sj`L`nw>gdk-Qwt70?VPt6co5I(5yuF*|`$_6*OlcAM#3G|$-m?L?&j z{#>DC5kmZcinE$*zBBm!!YeUe(j(>ML91 zkb4ofhO)5~ub>JwsG7?D{_E-;ak9`Mn@r0MJWJuhK1OENX1tnLP2wHbkjp7I4cC3Q zqS-0-VLSRGMwC{{?I=ppnfJ&gkj{KaieoX|Q|T+IEXQ6RP7y|BysGMnK-V}?n`>E4 zmH|Qn^8nD-$MM<@Y_=B3?i*a$Y zEpEL?m9qf1w@gBcV`*W2pI>pDFzBbEiUo)fv$6x1_*F{z-HDah!soQkZ+3L2qOW2| zH9g{?K4H<5=dqy)g?%JWzl_?`*$RLx_1U}qo4}qIABl4zu!xbQ5s-lLk>oCyn8wYy zV*ogH0D!C|vqETaA^y{uE3(Z}Qj@M}RSYR8*~|wYhEug8%OY&+b`}42-THMD!0$|0 zsTPfvo^0y!a+B2teR$@}FeRj&Sm;-WeRL^y`uGe$WpIBQ|?Fl&8qIT;n zse$6#T(wfJSoRP5PZ?o5+rkm*k$qf#z~S2g2YLECcPs?P_W!@Jm)<3_q^|Qed~|Ck}FsCo;5Rjtywekn+2z=j5RKQ z+wm`I&|JG_SUx8dhmRptnnJ%Ts}z6d+17~5`UEW^p64_AxrO`67O`D=?~AM#Z<<4e z;+<0!eu!5Ya#0wuA&AfjgVaiis$$TM1FG+>jHMN(kPR3DKvPpAl zs7tFc0j(xob^XMVV_GRrh9jDg1PWz4uBvt$-k=R=2rR4cVg9?{=B$ zmy>B3O4%F3*k1$F&rk&Ie6gGxbko{q02;Pj%Xk0VvSwck(RWUW?G@pCU^h^tiwYEy zwZzv^4Is~CAThDU*;GnajfR!umU=B;1ii}Tu0q*mb1A)E2#a-ZNftmvb6hZ7VvxD8 zmeP5enHBW((-K#yaH{?KS+FldMRiK~Xst*D2@Y7Rh?Y^o^HSc^{w6|!wyjt-^$JU>4fRyD z(w38juqohmx}j4Y%4UmGSf$L#YidP%n`%6E3@w%%1`yf7sT}Td=L$52lJu^^mQlMJUrQ6>J9_2gL_o=2JS+x~?<1Kt8-A=|S;p$U?z;EXG9-jCAXKG6 zzW& z=Esh)6iymCG|+((Fp3XuKAL#sb0v=p*KDCC6P@vRBYJYqppxI%0^C?3p{E4DPHmX~G zA)8_w;$)$8>7sU;DA^2sBgzt8*1{6XZY?XO7+X40;#!kps;{9i)F;ZT*lM3!oNI1t zYd+WcRfAvV`l)|!Yw^~>*VHdM8tM(Eq1XPn+q0ypX$4)apX9feWxR^7g_mM! ztjQ~ZcHXCqjN&~4NdU)+1Mk>;+O15W@bM9fRdfEDv@LU!_-Cwy$O!Va;(b(k*$Xeu*Ch_8ZR$9n35$xvn zAQKkMkr4A*4y&62I4!Pw*Gx7Yp@aHa2j zmWQoNCeG1K(KGQ)0rsRGg1%d)OwETkS0_GPQL{GJM2e5)f994VA_51wq&p)Vxk0av zl3f;$m0jv}6BK1K(JFt^9ryqL?LQ-byWu-Donh*=M%OGGfM!r56@P2jbLX(Glinyi z7!!spR1F|EY&UwQTFNkj{2;eUuT4`B)hB)NaCgo&d%B5VIPqRz!+c>$A-83|XSA78 z(ow#Hi(2J6N-I`xvyB*hhWutV_m%oHrDErE9bY{?T01A^M%x#T(j8o|C z)}m`sjWD&xI8j#h^wGHh4i%q;oAO=##ftku{u6jtmAd7hHiP1}u_^!L;pp zvzpS6Uxdx9WSJ)F)t)<%xDHio%Etm%uZ98-6~11Swad0Y)UlE>{2ZHueGY7XOo}CO zBCPT(;%}(X(oFjAjLzcC?Gop-gLxmYuREVKS+kjV!T1t_*42q|4b_Q#dns0z9?K7v zTzSC$X*Xk{FPye({`daiqUrNEz4_O6<0_*0`kdIM8=-uUB!D9e5?am(D|0=YOdv`` zKiHQkO|7;*)#k_K)t-c0LHOP^l zZZb+zOq1MYEIr}hyHVQ&cd-FLu0aZz#E1@5)E6RvBQAuiSD=#J^GD`N$``)<865y; zT`n5;u&V_{{!0HOwaNuU!yQ^!aA)qy-d6o1z$;=JSJnFW3pfF{R3wkt{I)J~S zs;}^)6;Fahdk0T>$AqevSRcIsm)t+6mlXgvVi~23Ow?%My5w$r31fj3yTMHh3TV8e z4bX|>g0DX3Lk0%ADp4yY{hYGsvat-9^XoR}G)apPO`)qnQ~feKdNMXmpMylWkvBP8 zKN9z<-Ag~@>zCI{M0YF-MfPnKWt1p$QAeoLNT5hxqNG+P#K07%3~Z0R$G4w6`s-Xm_zy}` zt;fA}v-z2!>FzIw;sXR8U-G6IwuC9F6}sGTV72Z_b!dOZ5b*B<7}HH-3EWDVaogWU zzh=Mq7+Mz3N#jPx0H^^KJvX1~|QJ#==M@nbanUm&fb0EEF%*lY#;)*B+w9^zN z$ZdMbiax1Tq`8nf!+p_phLrJMjj44RPTn(>}sP#vw^Ql?!__DXiFd_ z`KA+AFUO^do5Ovn=U=qy5tt-d(QivLXGZF$2FC7Yn)i_gI5uj{2_oFv)mAIE3q<10 zNA?Hpn!_wepwQGlQJocIU?k+TA(s;BlHk!aQgo&bnS#IA_23OXK>Ih_?>NGH;suJA znw3_^K{0Lv2p2m;sh?q{YZm&Lp$v7I_>#0Iw;|OOLWy#)NU&CBeBG6y644gtH1VMl z2?>er?HB~2ZuHd3`Q0_0Y6aj(ifcLfbq9FGq!GJa=J#y54jW=?ACCt^!vRxF-LQUq z{ev&YY*%a@6G|>`_HRjT`+57RRLzWV*l}EGID6zW?32@4rq{n{vRW)Y4vP1u?o8)U zCKAsoebLT$EV+XAWE}@1r+(IaR&?7%Mv@@~jJ(@4?B~7SRNP9J^N4Vqf4ZIt<;$Ey z8RR)g^3GULv+S{*%Z66_o7nLm-Qqt=fBCtLVOB@WaW^=bntobNYU|ej6f;tF_lcLF zJ&ps?OBw#*e};#cr)-U0i!W*YjMNeRS)=qcV7rxj;O5on!p|3bn5Tu5$r2J)Ij;7L zFM|{(9h@(_#6aC;^V452U%vH^HuKfPbnd0tYX*72492!Ak(M7e8G@!=hJ$H7iR6^M$8Nv|Ms-jP1-h4tVnf zwh_eurdcXS&ivnv-g0N{dat*1iBNg!JTI+$F&BfNVaP4RIiu}W$1gCBEloWhx<6T9 zY^koG9H?A;pNO;DLmL@fi5X0I`yUP;`)|SCB zjoAY#leIoov6vKJQfKiR$0b#_2C-*WGhY>ha0i7Wh8@Vg8eG$bm;m^ZOeK2SCK8i& zCqUhT$po*bUJ*Ufq$P?H^P3!H^2}1^tflY4&T#hg(#r&o{>uBwmwbIkm~0!5j>r~Z zLUV|aIqP-)kLQL&c5&726OUl-lU41a@-MsqMW6oi4nV7N&^EY~x5&9tbk+-^qqybw5EUH`Y zY5a#~rJZ5FL%Ck1UuJgd@s$LZ$PZ9u8s8exSN>4mqQ!D*P)-4cn%WrW3DV|$7T&p@jn1Fm z4;P))9&TtWG%K>Lm=aZ)e19>1DVg*0xysw(NK%I<-`ZL3kvBmk`5ks~mBj8U!)caz z*c!X%*66!8hxpJiTG0Z+02UCGk-9{sHSjpSSOpEV(~2bDfUVv$>}~%$Dbq73tmOOL zoXx}{+xzcbha$fU*G@0$T4o14?vo!$ehO(DY3_2CEb*!Ft!LfR0)47E^FaSmRhqGl zBiSw+nlNQMu+)$oVAgjhUcU35S|qAL60PC5+&oY&Z^&tqubg z%Sq2yYIp(RCb`&(AEMAKK0HLD2ukGkg@F${I3zZdxwdN=&N_aXX?0a=*g8IZ9d@yTo`b}eC6v5JQHmlP3^v6H~bQ4WVG91~)U{Bl&&ziDEE-@Wr3?-nskSpyH+@h=Ti%D@lloq+%^w zfn>=LCDR?_;y)F({j*eA6#p zz*BW@Ica-(jgh2c&@#|{I3w%ND9K6>e2P>WkcZu&%uuz zUULoYvnIxggCA_R&k{zElH~lfZt>oI0bB-wqmEEzb5JuwWofHRT6vtz0Vl(LQMATZ z6qghv&3_abn;ckmhg$37nlDY17Xmw%pYlhpuju_KN;F7U`N?=M#V~?+6?)YZGB>x= zRoQ;mr`^pEGk<}{5_%`La@&j^3gn^XA9r_npu7^G&UN*d|G*z`X};EaKOnoG70}^p zxaVyH()}cxO}npF+rYp8_;&;=D*&5E7lgPW-ZAEzTb<+e(3J0`J}QXtj>X*bugczc z17~YVq9yfxbBuLayH82F`m#$g$xi6KfQ+<8^Jh2ma|<)#i$EKvEK>|B6GS7W(Zv2i z1=1S3qn@&j+7?kGggs!q*3je{f9 zk!%Q8MK%cT%zS4~u}m^P6Wg)qtTgIwBb+rhc6^*uw|duD%>U^OmsJ(4ti_nIBU&V_``-% z^^C(2Q7o|kv)7E~B>QnzjpT52&#=+%Py61o6O74;&5!Lt?ludL@8=P-ap-fDB;k4bH|7(_e? ztC_df>JL3n`zG2J`R0mi_E|$${C~+U;O^7;3RSNe%SX*p?%32Sno!y z5JOEfd{FWt&F0w3i_=Oyrh+5LGY{zL{_O%XF;4r`*X`Dl-@ORERKg)5 zLY;=rMQT<)a&WJ&CY&cLDTSTad(F)qu91WBS4BrF%zYw zCm79DOKnb-X!8ZX8lVSeOJ6%8!kH$Tza%jo?cEs6dbhS00f?0XTMD_<9XRwsda>i^}!MHDW$3mNmL+_R;OCxxUMspr4iAPehW0+?mY6xWl#-SNrA$=%FmBB(MJ z5sVCH4Xu!MJkt}t505d*2HYvMgwp*u+q4??cJtZrz*=3PC!$pN);Y}M|KTCM|FqF* zC*399uE%un(wudEi+>*7V|~g&#@gcCTnh&*`c$S^Q?UZ-$flTOtJ&^a)?qC5`lmA+ zYFei8jmd#9eT2}^x{DcI@&pdQJUn81?zjNHdZP6xFK@c!l8&eipD`{D5|$KxoW0^# zr&34&Wt!z>npB(mS+^VC>G834n%V5KOO((z_m#p-DXYPr=CtW*kNVtjj1J(_+!Eh{ z=kElg^82QgosFzm>ooH+ZYNiIwBl&FdT?Mtc8BYT_v(jQ-4|_p)bUUF66E-jR2EOy zSeNGg!fDv1$r6<^ZIJFEJ@W@%T1(5Y=aQd?^CntEGET!o*ZLml;sgU5_RoLVam)7l z{z0~23jJNWK8{EMBW2qgKKv!Q7aoq)M5>P+xB`Vw7H7vDADFEq>EHK}>~qCBkQQ&v zCyDshF6|kV@Nc&n?_=RQH)7pev{K`%b2ORD%>>xhua_r|C3iSR zu$sbx1QD)$OoPEwLKMMn%t&F~Jq@vvEU!3D;uIT{2;%;JU#<-dH~UnY{k>*G;!tJY zM?a%^i&vB=(2(;HuDPxpDeUghr?vE^ugDE-;guEQBajw{@RWG3hE z9?Culh`U57D>{PmQ(2u1ObM*uFJHZq!jV3-nwhIOS)4swiENss{oIaYvZgkos#kpl z#d{eX7Zi{tnY>K6K~>s~4YyBS{e~-o=BgwGI~;d>DKk2-N_LOCf-u{^1kRlKysX_U z+dK>~b=%K-z@Ya3SYrP(lK=MS`hj~W?`7rN(U{caV_q(7kXPYHaYrd5Hi#1~FT<4t z?5BPkyv5GjP!^FRgB} zRw;C=P&KqazD@F8@_T2pnwJT@OX!)KTjaMyJa7yTlz^K-ynL+EaUDiSW_YDcqnW`t z#Zfqpov-dXe77rV()weE0W7~k;|eM%3~bb|G@_%LUakg>Yd3jYLnB|5s{-l|TZ4K! zM0a;XSwOJ8n3d?zF8scX8Aj*f;Ge%q|IgU(IkH5khmc{9$=`wCk~f;s9b@p74CuY0 zPZbF-O87CRunHuq;H$4J6 z4|xYps(W^XKHu^T4eZD$3G_-hD+5^{gH9LGpQ(nS($4b>R28hcrP+>{r}#xoXO4p6l(h3;py?Tx}w2yNkH>7@27KX zZyr`J-O4sdGcrkNCIbhUr#WU<1cV=28Ya4F z8Uk@Y_$=DiV_l*3kevT%NR)x6;3=K+YE!!TG;gl_u_jBYPEP7zgjZDHd(fIwu5qlaMBA}Q1+47;v?I!-Ll4L>hiSbPPR2YOsl#- zd914VnlX}G#!GbN+72n|bu}4e*n7he^b`f8AjqZY;J}ZJZ?26MpPByHcfV~l8Cf6n zF_c;6WLV~GTI(YG$!04>)%_j_c$hcUT_wKy4HOVbM#zvc297qw!gxerJ%g6oZ=IpH z>pMDVxd-c=aY%$)ykEt9BJLS;$j#Ex zpe8Ms6rdhO(g$s7Wa>HSxEAUPGI@$rcF9U#H6&D67ZKd5?NAMmegdqSMFU?ZmyviU zr(MX#_Cs+lX?>pQ?v3Ei4mj|04s5e?a9B|zO-H+WUAmSV3=qM$+El%cfJQh0mMp5No7dCvCCv~Fm<6zOsgKNN9@rhF|_D&ljfNI6e%yG1c#S|v*e4Lv*oVC#z z*zC1*KnLQDGHmu;UxgItez^GTFL!jv_wY7{?ah&B3qHw$cM4} z)Fl=e;a0(FtH+~0m;?hB%4uw>ag-0Y_hd;v9wPw?J47fo?bfGa_>V)4SMF>MiXZv0YBPYi_?6^IzASZqu$L{ z37In`ae&Pgkv&vdyy-yvFRSu-`vJk@2EcL&N|NFy*0$1$9VKi&o;u9hxZDi4X zU*TQQO?{C*5%E#9&>uKc2nLCwR80EZP!VI{laVk>Stoywj2I=VhEl>Wkr_ zjG(vxo6gIgtC>^Vr$&&w$@Ud0rJIOa$})U&a^DV7H{Ib(=9C&Ju?mrPCJVGTuE0mx z{1(NCm!&eQm0TI`(Mx)6e+3pau-dE@dnSDBf|}<~O;GY`G(6W;yib*GUxl;z#MC#C zB*;V;YJ;fa6&cArpAvxHofyv%6vWEPPiRnjs;aSQisi~0Q+YuBo9N1>@%z)WwJ8B= z?ErB5P%vH%7@Ds13(e9bn9HQ>uI}v}TUd~p(lk0t-OmGlTDvMf+&&&m5Z4XGs{9JP zvC)ukUR2`k#2|d@zmk!%5qsarY;SJWow*=1BvaPB-tw|%DNTPKz0p}P-HL2EQYr2( z#X2P@w~>aSWTRcpt$6KZDEfAjvP(K7-@6y*_BaCvndT?vS*(+L!=k^rf+hI2J$bTk zfLKbH$~>#WXH1vDbnA)GQspDtySb{IdoQ{nBm{`6>5I3?ZjW;2I3F^;dCBFkC)N^U z?cG7ExdP#o=COuL_F?D3h^ZgLnGoHtkRv^(R#UDr_W)o$Byct934`zNHJaYebPkVy zE6eF_8@RW}mMj^_GL}NF^!$Kx*W$F2D?#p$raC2swzFL_4J)}6jPTo5hkwbB-)I2x zduXC0=Bjf&A9kEff8j0~j&s}Ww$TN-IU3R4?m;NwShb=!#92Z{LG${`g!s z(VaEYJn(iXuEeu??Gvh&XVg^>6YF}ZM`aY)WV7SLo7Sk_KAf!fwKT~#!Z@OB(KY%N z3|AJ~Y&5$g>k|N&0)c)$G?{eK%`d6Q=B`PF}xq&CT)1N zt)AP17c!o_HnrW5Xj3`W^3ejuPLlaZg`VgG}r$QXn80R`ia%x}wt@0Vu;;ucf_dNZh-iLu|O z-UWW>Q{>Y_z|8j*eQ-53{dK}rk&g&|o0}hqS!3|i>IGLo5T^z!TyncIudIA^grGib z^2$1M5m+-3AFJf3P)JGPvGJ4-jVrK7yHa&EO!(|F1hF92u@*V&N3}v) z48Rn?-2mldbpm*=nY_(+Y@^g-`%8Fo%j zfqt%*@QI!M+-jEE4ce;g1Rdbq`$vG?_lzWVJ*PR{+lKfcV%%2@K^p|goeqsjRVvUf z%0s5}M#7)zmX`b{GESpY>vITcX7ZRhT&~K-PpLmo6<-_cDRwh9MNn39$k4KIOTQcP zPSVPlo+7-c<&nIg0-EWZBt;zkd^9aNy1+uC&9w&Kv{DqX=YNnYsKv{UPq~B=jLn*U z!r(`r{LdK(?1+EiEhuojDXOsSx1*(x`MBaB5@?mk)gKa4;Y}jnyJO(IqB0MT4Hb2h zfD=?Qsbknw=Xu;N>#NF+X{=fkd>#j2? z4jT*Bjz;KPgd2GnWUsjtp{XJ3!I90tg82xQDn!;(_UiPsw%#M4QQimsiZdQwPb873 zEmb33rbhmUy|}Gx#oBdxy7Ak8g~7l;+^Y?b=v)+BDVnvkoSQbw;eN`#i~S^$>JpF! z^uPkmVyH)|B~^LHn}WrqF>0W6F&_%l0A<1BXAv<=6(dXdRZp_EJ4iZgf=Vm_ATY9q zLhXAJWnBS=8*PYuydXhK!nvEq;FgC?Ksz;cmR{-rh-u^ixmcTDx4+|^#Ab&`Q%z0X z4M?n;5t}|8QrEi3Gi_25{hbD7U7G_MaZdFd!JQ{4O4Ypg1q^QTb&5zo8tZ5>jajsv z+;1jy1-|&BCV`7b7*WTEX2?!PdVznE8v+x0c%RBXzbgC8^>m` zuo_-lj5(xO{O*7^9oRMP*)5wCWoPI;CEH_~J^xqOcWnGZrgxncD+wXt9O6%HGdLGD zK(Z$-xP7AW3F2FgK+|eF=neHi(FX@7#w?LJmm^7A#P+&G;+MwWU#vA#j;Hb!FNU`T z9BtVpI`fiiY$azJ!LDF%)ue+YI5$Ic&ZW9xEB+HhWaEDYaVDM;6cDgwhKXMB>{2OB zvX?lWb@$clZSFF8Maa`c3Kc(7bfMW-8w@JZgt_Et2YBvovII`DLy1c15%}oaC-nIIZK}c9};Z9N~?b8BcvZ>)l z8$6I%RwFOxL25zQsy;01is2?u7hat?8~TbE=yNS1?xD>L8P=GqA*hwQj2^WZC>o7* z%KkD-5+(X11~0483d<^=f`%|NV^z>Y#ZL6)nV08|fuh^HrOyOdUoc!aHi${N#$jaO zn~xXi$sX8Q9^}1eQN$sk=gIsS1BM#Olq9$c2?pt;c5z3>5|om8Qp#F1v@!v5jpm>V8{bSF&V`s<`ALUXAL-0fK?$K=5 z+PXy3xRgU~XSZxtHS7AgntniYy60{MEa`{xBU?qOR+E+!Szns7srM^-Y@B*570T8K zsyE0EHa4x__Z2b$qn{Lf)we`wlV^T(>NRdqUMD-M?tu*ngU(GEfqJ^HyO)+v%tf|} z492x>!%rV=B+n*IfFgg&@OFYiL_9jHLf$lT4l2*Mxi>?;F{vfbWa$Wc`;EDuj%{Nm zPHG2MJXGY^Y>Ss5W}~Xim!#kZ4&Gx{bF-PpC|*D|i(@&Z@P`(Fp0AlujZx;oxB~nX zINzQcY+g`=kiYi3at%(s`zy!%_>hyyUpv^u+g-$OwxrXpE>z-U+hd|hjrb6eTeXS8 zo+-yp775N|V0iG45-}*tH!O`Ivn5B&9bI%I3pPzr`j7qm0bR@35qkYBb+*?Wrz+Vz zh`>Z~kQ`jPAbo(bsDhScz4{xG9A7<)-3eAnwRc8JmlXVo2*SVil#kOD6RXhg3GMu# zg+iFBuml<&^6K$PBY6+1{4WE|{6h?{FQX-$b%QFaJZ2eqMQ;5jFA=njn#1*7n>VXUO*Sw6Pk4@<^>n0!Twn%JSG zKAm=a_I9|_>6=e?MJ|3&*N$sdoRfD zIhAFYJ-2Z^+>>D~-E-Q>Fk5RKT>I4fAR{{*-gt{aL+Ib9{0&=W>6eMEKtHvRqA-nS zH&`c92af&rYOiQBF|oUMO2`VAO?SGxU$_uZWVSg}bNOCA{N8T&bcpB4qZ2>W>ppjq z9!SATtQC|E2kubE}P|>*c5wqrv-o3{QGwWd}{o zl}3k7y_`=d)|X9=K7NCC41I9BSwIxrTx}97D8$DT5r)2(3V7L!InS$mZXB1oIN$!p z=ex(DyhPEl5fFW6fRbu+u#YJyGW9p6XzLHa$iSAv=#MFm;ta9t^k8iEbw$9!Q`+8_ zX=}xlD!a?Pfm%rHzUE#`7#74Xe;EzvQTVP^{L8iRuWSVix4$7?XU2vaWni_nX3;_A zV`Cv!C7uUP_mzV0q_ACoJD+4SS^7+DT@*M&ifnf1=XI9v<qYp&HJYu6-d%DX zmfWKvehiEcp8OUU^D|V|$8OS=nSatzxozF|7~I!Pm*fMSCfOJ<%eCKQZ7?fk?Ws~~ z4>2c$4w5#lqkIHGzK6a{-(vNpAh66jYO{yC`yaCJE&I3QO&%{vKzl&FiDY5;2>o`oVX= z(}WP`POLDaWIUUimKk_dgFUa+{vU?>RnYi0bq@YDy>9Q?rrFGfi<>Png%#nY%LmPG z-@LeJ@ZvYyu664gDvgdn5RpF_uE&1Ymvp9RIx=MvgC~(szd|<@53j+vK1fVzn>Y92 z_!{-%7vP4p{45Er(`L$-E`R69efATN!1au9PsmBvMQ(W|fgAZGYBqU}in8u>Q5vZT z)ga@tZq>)B4$9r_ilxNw1{&L4dmfH?mH>PoHidX%vmuLu)UPIt1-cdlI_D`Fe^FFu z9dG+a5wA7MXiN$kfX+)zq7FTTozb~Gb0od#T8VkQYe7f!U1b+tWfulUv0DshUUNiR zz&yQ3-pzslaDGvDkI_y;)pI~unTe@Y9`(z+w?5Ho(dh)_hj$Ug6xF$w{;qGDMBKfA z48>1H-P9x~QPv+>^f}ItKUS1~5#D*Z3?GZple{ESA>!m89@daJAX#7~Lb~T-=^`Xv zCao>X%$5K{L=ijF6+S+KUJfJaHdSx>18w{Z(2Zg%2)4f>WTkp?l`@Z$CJ)Cx-#Qgb zFK*!Hpx07{VU`Wwi5w@5#k(o?C^ioxq%+i_j}W6?cCW$+YBltd3Gx$VRfwi>{(&@w zk=D_oGA?SJzwANEBpWr)f$?6vb?%C}|EgY`9*%$<_tV0vvS$L~9(Gqp#=l;lk6%;X zGSWV69nZ9mhZ~!qy+W-CZ{f~qpa69yr6r`xnk-bJ$)&&Yhp#RVrZJ5{;D7m)O?kIc z@6ywsO%@!hf^N6B3}ims%&d-wh-O!gS?l3}WJQ8nt?kA(KoYNqR@rf}%im3y3J5Z@ zS-9A7s4FL*f#cVAa##Swe&Yrh8KiB3D>X9x>S{63m4Ewbv1t+?vHA_gbFa);4I-1y z>seMcF_}A+IDTXpy7gNP>1}!S4UJr~*V|t{CFXPU$mGosaSG?OGwobvK8WG(nR(};m$ zPkO=a;c;)GVKsFYfO>cDR2Z+(X+S@p#XM-oO4e1ZD95~K`L&{Jz=PG92^~jH6%3SK zEr^m{o~Ayr8J)!i03HNRFFshiHB?W>7D~d<$am42y|bF0?MLL!jJ18PFZ#F1$-SLL zsyvt&kklvBm9-_OLx{5YE=y1a_4MapyNe~ zC*B)qpcj@TUo$?Jf}D=k+!z^KK^ocf0I;O#LX9V975*lUE)%hb`*Sj{KP~GF$s0V8OGdRa8q9ASA5A=~GiYBn~{=&Zb9y*_%?x<)v^xMyP` zbA-e*WExfv-@p%Xeo)k5CoLY`OcYzLX`j!Mets)}6(@mRB& zV?Nxny|%ICC^@*`raUeF{neK;Q&asbbuXSNhrV7{6P2ic^J#hUTc%A!h)zkU*seSC zW$wt=B0({gj~*X2Ip;eY+h?bb58?bH9XEjQM7_W^tP>6V>0J?rf__@?%`TP{OHr%$ zqrLLB8sT`Wg^iBpOyp;k=!klMD8ui%TIcLo651B{i)gWSLczgTrN06!J(O?L^Lkw= z*3Ao7NnmrQ6_sf(M6ji@IsiNOiUYCi2TStMd-J6H`zAvBVTtF@+0{nH@faw9-`t|bHgmO^V)58&EMg-|bk71*Fq7|aGvR8(unM58LtYy|WzyTCutPY`Gqz*1`0@1YucF*S2MKEPMz9zpWZNcxEgfu{Y<;C|`7M;k7 z2Eg~L^NzJ8WF=%>>{u|EdgY1}TQ;TW2+3GUkp@ef($`IBf$Y}6huYC=WTm$xv($Ga zvc_l3Vrs(|pzKXh)bN z>L`fonbQ__`+(~8`)LEOX1|z{)VE!ZxL%cAo-fd}l^DSrJ z9{n}=r=6i6{5AADM19V5^4l`LX=2f#3XaV*`w{u|I3pFSAEHo|H-g#Xnq4 zs+j6SJr2Jl-mUT@EB^b}uY7E@q2&@79v7o#;9Who=dm&xz_aJ&;}_~uaf%%f+#V6h zs%%&XENRnE{AmpQd@tyobomxyw$UGV>2~y1y!kLOrblYNb-=s?wz|2x@Hf zlD*QtU{oq%tPf@}sQ4>f z63Y&s>-U&O604;;bRR(ScbDzsmbbC&Ct*M@>q*Iqq$_x$JnU(J~3C-B`2WX{SVikLuy>EtZPRm(Sck za}{*Q#M*D})bw9BnaGhZA~y!8QTDWUzSOn@)a|e75y-QwPiSItD$NaRfWSWidYoMOacSgsu=*H zfAb9Pl2*pnW>#2JIO*FIIjzP6jjVPbvPO0vdDte^qxyvtgtsFn7@Fzx+L;PNnD~2a zKeqhufBwFlqnG!wPICL@x|IA);+^c}l2Y4J1@+O<*Z}KCd*+T{-o@{MP|FSr*SG{n zDTSab7S<2seen?9{IcCryu@Ad4f;Kvf#^=GE_#j1RyuUA(&Ux-bTF&h;V_Qswb7?x zR~zgSu$jj7KKQ)W2h42Z?h)d^!EW=J=AmXL;Jh6W#J;I_N%)Ay zUxNfcbJQ;IIf~Z#@bRm$g|)LfM=glW{8s+x*lHbfb#;<#jQirPjJd4chUjl-?gCsO zfg}SQ&@h|L`EzL`@Q5c7cgJ@$hyUTbI=o-t95SNhyyTs1N|_<$5^^~sDXQj6LGQEb zHp=&iY-gPF#DvVF2YmEHRzAODk$I@5w-57N5;O&e(8@kt(qbyDX2)4uSvgE0eLRr< z=@*aL9zLCbq4qP1SZ7!4;MT5w;Q*?hX`IPn7<~Sf$>iUhVwoI z%hw4@MAv|k1EKD!aqa1(*sYtJp3Ehl%Dno^*&TD{JM#!pJpYO&(lGkgk6@P5!0bhb zsMzuL&o&oY!q_^-0tS{y(tG~K8?;}o2CJE0*c4!FUAIrid>IA z>|C#Wo@)`$dxo^Hj!7^*Xc7yFaJwNmiWqACre{E!JfUaOnL(12_73_D8P2rak$0%L z-U`NAHYGw&`CY(rgupP|Qbf!oFUUsDdsE+cAsYf9NmtahcPtWP$?8?}V$b4rY4!|_ z*yw9+>VeYKeqS9xpXxy%tjfCLnXiwDkMuDw_nQ1hCn5Txe4iz0Wu(RR^657+j5mY;BbK{zfE@4E2K(tcfpdm?|wmXV;_ zogw8%_EB*0Pf>m##&mN6WkN3Kxo3qb2HNPO%8BZ=#L`N=eaI%0l~dGbsMI_a7ytms zzH)jng^WiJ6@YG(&Gqq_8b454d%|JTI~`mZySe@bs%NW}8d=At%iSH1Axhnexz=v$ zN35xPnGGT=x3oL7mo~f9bSq|!9J)OHYDPrf3FtM-qfFK~Ni?!xM@K1G^(GhANLePP zNJgWE(!pYWCkkexM_;M=>6JXVnVIbs91Y(muWvrrZp$w-mb^gz{6?JH*X0*NN$i|e zKBl7qcIG>JCIGI-Se)wLr2XpHcy~3|HIEfPuVb8f`|gJAuB98%Qx8?D@W^BgeeNXB zP`iYqzp)b4Uvh!fnkp1zSz%-e)GD3xcfc}X4R0ReGQOPP=p`z~^JGhHz4$vWMn$cD z75vzM%l` z(3yZotjq|+uI|Vh#9EsC< zBZliV@u%4wRr|}kvfAl{M?KNG%0K`7UEmT`FnWaV^k%Sc6`ofuG5Egn?ESFuP?_?9 zLE@*v{Px-tN}}G5gPU}wzKqWX%@&dWoblx^=tJTScIw_;wQzVh;<~#jUV7^Btgz-+ zr;w8G@Qg*H527@psM2CVFfR(DR^e+{6#eMN7R%5Y%V&}tKb3f2mC*bg9^Mv{2ftNt zCI$x#T0)8kXk@IEZD>vE6ha-6u#Bp%u~+QTFp3gw8@V?%yAA3SZlGv8McMu4PkaHm zdCh>AAn#S z-pJU<70#P#UF_6vnM>nPMn9I|7_0C1BTFh?`1KtQqlz79oF6LQdD4kXSsG=c!bz3o zS7qQAp^b#R!G1* zjeiDU$j0e~zttt19U-^6X|)6TM!m5M7N}dWqoM7$#<5^rYx$t6N=0{925?{T1>nX$ zssXvqbKg%W@Sob(e;$Zt+XiS0E|)d<0;KAMV-^c#WYW9#Vnk}ChB$ezpz*e}6tEmT z;=HLPpc=tr!DerUOh~OOy}B_lAe>^zLO)TMdK7O-<4|?wLo;*0vP-nb{hzmLKisK(rSLbIFbA6#pBoET3X{0S=Hw~P_boOHEB+Gl=91sGa zL*c9Oxu7S(k*Fyw&WZPG9j>Cdf*DOSm{^?e>C`qRTz#J`&QCHrStEIsAD^s_h|paD%X(z8M)%LHTWsTu$bpX@{TvO^v@R&s}~<|9b&3zKyvqkNZ%cX8_a3eHiIzWVd-%2W zB7{n*`m{>Cu{CFw1!rlG*t$ZgYUz;@Q9U__X`Rq-gDEf;Lz07$f|5k;99w zy+UjG__MsVEvZk2@IE8>kO=@FBgNY%osUh*KNx-cbNYlSX;&?idEu_eUSu}NNNhJQ zATGo(KQlFKsl+;Z)j|m0Y4>;992vgZ=1oj%i_8A_$>T@*s~$BE0vOC#K5jefUiQ5B zST=3eEZt1-20VIMhSCOx0oc;JusG->eo=*v^lYR-V9tXQu(_=NLsb)Bft{(4C zth285itsZYqIab0p9KECwJIi0)=*|t6XrJO#gGSMujJJZn3r=)=FRht%@Cgo&rl5T zVDJ(bPD@I>rX!0B=GErC7o$1pmAX6I<2i2f;5bRmv|&NYB%*U)zp8r5Fr=Y!Tm7i3 z>4=Ddv(X27j+xw~$4BF_WV%hJmR4)aIcre;WzSuH9Z#j#vN9u^bo5JmvgKX^tsFIA z#7y}dYec=kXaTlkw?xDkbSf^8OR|$!12#@D_Cv=`NL&db50L_Q=BJ(JG~wq`(alEo zSdkaG$uI!9=F8a_nMg z6`*&*V!il-#wXwSKGam(wDNfGwEXz^@U@*rG>49$ggrMD$JNd6f|$-=gM~sX=`5b{ zc(CTu^GZ~@PblyrxI;J6&kPy={mU!Ja!FlIEw`bk`rH{38>dd|@=Nu`O*S#5F zkG3|7uE4A{)@PLJsf6LmpS_ayOk3HW9()1tv<F$Xm)u;O zoVC2=sUc2zm4?o#U!jw+iK2z0Y+z>$YI{ZAl7kJ!(yGFCt4Rzs ziIp8<5o>W%G&5h5ZV?>xJBQnrd4_{WbHN4O+g$gdkZD#@+G=7DwDGHI&Tl>08A(`R)vE5naEwS*Ha6TZHk`rtWvT-2$2g;rM0`ws z=ELsGaj^zWnmxWgo0GaOc6)F=nn9Q5MCoIUDsr>RIjnswjdWxJl3wN<(19-QY>3Nl zk*b)Jw6@LhRf+dg2_^aW7&wk&Hf(yI#Q{?U#jOe|p1%`P?4e?M9V_#8*Bn?3k$t=? z%#iF}e5aPa$~dS&%a#1_1;E|?!$0)3M{^t%F11!0JCC0&p!Pa-Vi_d+bswyd@d$bP zt$@J%19K9iA~BERautNLr8_NfGV6@}{n}9{GSX!u(t$zp=jfdvXNS910y4 z=eY>_qlmo{q`fj(mx*s0{5^D1-|NHuKJ#b3p^Z|H-JyqKp6?TMzlKI+J>qMCX3TEq z#{br8ZL?Q)4sChU4iOoJcW1?X{&Q!>>u|67W&ZzfC7Z8izyzY^rSpHjHj2rUG@ZdU z6X)v*=2*Zxc$^KqIPq;$cQ^4=aSN3 zmpD&j@znAm#{`T{t~&eJ*!IAQrKt}{fowGA3aCB8{`MB->L-;HXdOd0n(C~Iak3`dJLO*ojGRJun$Tif?sLW8^oEo7Yy;}7+ux_}*n z=5Ql=UAg-tp`k!hwKA?5c{bvbG>eWI&MZQI2HYHGTq0%BqcpQ9#;~#1YhoSVv}&Tt zO-jGcs)k!6-s*4nu4ztctIp1i*r1g9$mRW%6FD6_?Re8Ry?E22quYdg% zN2ORpD6ElR{;GXWjL(R=8Rdx5SNLE=#{xxPz-O*!u!t}w?tc2vqKjai#vy8ot4XKt zxZM!`3c4EkzWJ}4i=Bc<&amblOLXDO!+IYfQ>UY5ZvVPQ7(v3ZBjL(B{N<6co>2KE zkQ++u-UDHtgR&kKJ<`S-#0vu51dYy{~wsf;x!G&vk)5h{h}bef~ZB{#o35ZS<$ zCIf-DX46nBZ=ckMjq_9^b>lBh`=i-DFs3@>lNzU#$DJ^hC2$m?QJ^KP#9aBPc>2J@ zc|W%3PkdgHMF8@MFV`I)mT$J9rDJr_WmNVRM@aI3$^I`-IHDMNUc5&A669 zD>4quTjk-PwY14zy`5K2LGMHZ#nS7X_Fv&}J{py@JdKol42Mx5U&+APzURMg+l+4^ zOygeTxoWB+-=I*hf2?_2!0*W(9+_p=%u4vk29Q=|&xtx!*6h<3Xo@__Fr~L?Hoeg7 zW!g4B(DGxW)iVJd##m@plte_T7fRK3aQuRPG$DPYqRSk+M&rXob1MdH3@6#OSRaP$ z)w5Ssy)cr8>!#`Zbafq7EntpR53cA9m1JJma7z&7lq&;F8awV@sKCcPdH zM{(f&5f2y|;jy7=;4>Et(lqt$reij}?gzUVhnZOwb5iZ}u%xwO$HbQnO75I;*mO1O z?N@?}7yLTKbw(5Q?{5`Us(p+U4Ek6;3Lx)~A@U>DiIXi8-K%yS&Aa#b>}W`6*RE{CE)ydV+?~hrlG+;tcK29@+Ls_-&pzew-EC%&SbGB zcY}dN!!WBQN5Lhg%Iy=iW3ODzFXnC2U?p72uLp88o~+C{297(7F?gv*cm}vpi59jH zPvbBJcJ|Y7#`?ZI{10GN|4PtWAzj-e#s9H^|8*UkM`u4t7j=EPFW}fk4{*#)c-rQh0x{~Xb zEfwB~uy+ux_So-Tm*ep%Z1QZ`qv?GLc*;&<{4ChEWsY-49Cw6?(|+EvPM4-hxg=(0 z2KBAqbJ&BeAsHI&E{wxUDSOl#!Rriy!By;p8gx`v{r3{Xc}C{YG;70V9f;AGC_24k zdVIOof?-&XdnDaD_wR2bHsfaH(M1CDUVSnPfN~k25tC~S>XN;Ra7b(ONwxMR!9_1< zOcfH3VJ}*qc8C%!&F*siE)QE>T=FWM_FS%0=bZ0lM0Q3*B3bghh2Pr0yGVKX;kGGi zFGZOuKigf#^7)rqApR_VA&sd=+3U6saDea44V`fS;O<#!tJzJGuCSoXWC8G9K6PO0 zMTeP*O9zkIE?#`jwVpFKQ-Q9w>C5lv6Kla14Iy>Eyw+mQ_9qo8vp}ADx*ipdgeH;) ze*`r*vHN4HFU53u&d(6>U^2+ zs2pCdoBZkN$xt$c?1pa#?Znfjze=E@inS={Zje#Q^Yx$)Hdy59qlE!fB z(9dsW?8mXdGU_}cn}Xkm;0n%*euI2Yos(GBuKqjO#LFN=Lki{Mub$=W)+hO}UrhBE zfD&WyyF<-TYdzOFH5Ts%^fA+Qg|)j0RUoudYvdnWIGV574Wt1VKD@lL%sM9oJ#!<) ztJBx~uFQA?kktykC()Sz_AGPFjh180DMjZ>bWVUP5-G)>d&}MIP9z*t@jvU@9BzFV z&#g=@Jf2Vp**0d`_j}|@Du${R^)3Eb_4Kuf@uOCAQI;z=2kf4yD~nf!aueP+{tpA~ zPUID5<%+iukH#?Y#b4o8*AtQq-aTv^1bT)rx$~5G`nFTrd`CV?bHfnR9Hrknrnano zIex>)uIlC(1}D^#>(luWrb zUUaA^eQ4}mIeY;e=nm(dRUC^(+8DW1jkK{9^>zjxOBTAg#+9i&p+Tl|?*%|+kx9SI z$oN&i0=>##=Z4{v*NkXOPF#xy-T-(LC`Na!G}hTprJ0rWc$Q0MD{YLD0rU4BB;ZS~ zdrd3>YA#nksWQHlOH1iF`Z5Or`Gq4el|@#uE*;Y{&7|M*td&q$*l*F-pY3c#MY-U= zL{Pia&KTrj{Rm7i@k?}l1wH?5->|hbSQTcBmb2d2HfOD4LVxakPWr=$kI45T1@hY_Hel}cZ&C+w3( z^MVkn3%mS@j(5j=;*q+Zr)ATPT)C0?P7dD^OV9$n7dK6*+*fm|^HR3iiK=EL$no_! z)(vM4e2dn*4}bkkUvS^J&kH697r_LmFQWg}wx6e8I`-|`PYIrYLEmpBbm78!o+s5s zpM2d2`l%Z@;3Y zxA>sXNAQwnYhh5;{4P-@9Sp1mTKUQ#DP`02S>@B-OZ?hTi`b~ULNG{j$-kwCA8>GC)Rxdc09vW{oYm$!h zL`@B6Xg!S>1B2Kt+m|wt*-+$f$0e?&aNfs9$6^8s))7KbH$+#}0^{^R>zuCu1=M{m z6`4zmDBPMn&=;VJxqVa#Gg|w^Ik~o8556~ZE2>c*tg zUZ|AK@q7#rNCN#;;VBNZYw|lZEC0Nw!W@bk*L^#|gF$S_xY%u{+3D!N?pHqf`(oZU zaVhJpBO40}9x)lR8Mn^$mZ2QW1sY{f(Y@T^J+3St_8$6G*ubOHX(^jtOxvRJd7BqB zuA~RY>8w=4ze@5q8qNWj9>2IA@vTuAB(I8Uxx3CEmYJju1hwb}jDf~usi0LS8O~oa zqy2|r%j=Ka_|so*Ra~Gpzvh_A#8hum?;6QJ1YE&rWa3$CCk{A396p6zfj8cUT{rpP zBDBO-Y#Q$aSLqM`uRFR=R#ksykM;idQ;z<;pYe{O-0<(AAF}w< ztS9yXTO0<5=;N_td)^SYzi~jiWqYw{Fi$L^x|DyzRB#OZaPP@rXyVEY z*>e!4q2p=|M_U|Iqs29YIJw8JiF~Phwt)}C_RWY0J@OFk=+;Hdg>Okd9*3~!nxAMo zPPf2+9+0gt*HVD~fIFRUZt|J)_hc_Qu@a{9_OKj=`=(8Jtv`w}b)MQ#eVDb+|MWp6!a8W<1KsipJ)Lx+_)whdmh&!~0c+svJ`ini$)xKQq6qsi5z_3!JEmdd; zD1J+a`0_Vm0<=t78|f~k%9P{uZZc@>ErodL2_L$z108QUh0N~IRRjlDY zzoo489^P)KhbicPlh8>oa;06)Z_zYd#2K-&EmEZeaVWO@@j4-L#v)Hh9E|zYNRQ<8 zNlK$O;Pkz()pHvvu;|b|=Mwoec;Q9HvZ(dpkDBa=21NMp8oAtt^owYA-s=dqwLNA-M3a@W>>e7^JiIF3BU(e-F02YTM zv(boJxE+Dp(zS?L>F}=7NjqHbM9GcE|1nR656?i^&wJ^b3EEu;@$cf#Ixzh2C;Vo1 z5M{hMIu|k?rtkGk;Nv}Wuy1)Dr zt)$k_*{j#>g`}fHsgW~AFQ&V<4r=NiUF=GyO%=MIxMtjo&fn9?=q;~{9&-G|pVM8* zH4%*MaliKCXn!A;$tOxwk-g~3A2kS;hM3>A6ky(eW$Ec)pV~Og+|C)@Vo~A1w79wijgLc)_Yk;2wkP!Vn^m6`RNqBpmtiS5 za`dUzU;~lmVW;VKP4`uqfMHPRM|%jF($rqn=)S|M)H4+o7b;w6F;oj9YJG3Jey`38 zj4&aAqEH*@2A_$V{Tn`5?Zca;!Bp@x`ZMJ`?U zZgfUzWGa}qy3JEoA^(FJuLq;IKQo_rQS`iiCED^+oSz!B{Uv{@I;@_5QKl~9Xc?BF zxUo*B@(Ei$-b2g;gx`y#3XSF=r`-0BiBET@`V@OhkW6D283f`ali@qH9=%8rq&Bjj zY)XM|6x6fys?^k{DeVg%j>=$6z?=U+MlRSjm(0@osIBK$bjXw{eM$TUd@zPD(dNut z%8;KKjw(1;GITH1kj#noEh=j+F9V1mzuDPRdqbpp{7tUXJDk&PX+)OKMQz+9)+1Hk z5W#;+(es+&OZIin{xY3xkCc|SAr#@~F2W!)v-TaO#sSMsYX)eS&~LlK&)fZD9`^1@ zy%(QK<=k$3u_dOhaYUyNeV&<+4^$(sHOIxFsX(=zqv(8jgcRuCiGzA7Y~{~iO+qL56Y+nvb^B@@bL zhWERnt9pAsTPJzDn6>p$Qk)d8-9qE6kK>sObLUrnNy47?S4f4Mz3@czmX zj5xDLBzh!j**JawO1$MtFa8((jP9`xU+g5qernTmm>5GkcV{yG{+`0L_oU}N_}iP- z8OOJnCh%UOpA;|&$BLO-og|{x|7#RmGh0oT)T#dnB3!le!-Nc>!o8_sG&gEHMA-Em z#^~yj(n$%8rE}!vUb2-N_t!-?QQT*uVQ@EpZ=j@P7SLK>gNN)>Z;#p&6|di$V}rY& z(@Ra{Ydriw!Rd8|6%I%<&@&oDq1Mzzr^Bik?=kaqEc040qDh6_x0=B@)AMJZr^ocU zAETpI4vt#I86?8rxdkC|1^$LFb^JWTv=DSXs5AY%d8el--`r@nBhGdJ8PmShq*R9W z<*H&=g1P3W*0C5q;52%OO!%-TRTE>~!-2v8RI%69 z|654Rl(!u8Fn=CD-uv)k({`$oFO?iOhFoKLr1BkH>3+87VAIMHtu>m-{Y&I$;!tgU z(pzpDch1nFVtRaore)pxJb!%Ji5mH(ZcSMA@P7a|4yFDBun*l_+4>N8k@lgY^~6nh z(6Le)K0s{${j;NmPZ`sH{{Hb?0Y1CZM{$=vpiYAZh2WSqx9;B)uSp@yyLMt3l1VCvC{R)|nVWK&_ThDs4hi@prlAo#vAC(sE03h%R##?=a-3#+&xvx*Ym-FvxKUiQD19@S+fZ6Aqjk8KK)1_YxT~E_uP p7p(` z?O$pg;WLZ0u4MC*LLq65>-S%8)7DAg9HZBKDIO+1+O4o+DOk#l#A(EP^}ED4l-dmE zGCYiuz;U$$2hp)G>)KCD<+RF`Hm5pgQzm=rrn3`#zM`37!EkLh zvrR-+wdYtrR~NbWbGhzf)xV7|)sFI)a?u~DCJ!dv#W-^rnBo_Jo=Bh3r41F8D3rhw zzlvXjV5~vo1jn4LA~Z_=Ma?ZuHT&&lbAgbYdTpzcDEovWe^d3`-N+4a$^O`*6@9kA zUlTDrdop!{X15~GRd_JzD;>P!ggB#h$SGP4Skvm7W|PaCt$&|{)Lz()Y{$1ZFPC(i znEag7ef#I$V{V8flxx{uGeH~?T3(x_b6P5#2;`g}YMzK#G0BNhu)Wuyls2vNXVu3P ztZl_ZVN7W=tZ#MmdAp1aYpz*HT($DrtXf47&2o)Ylx+DCt?8eg&nP}VF-fLTzrW{) z_~aVl==9B2UH7h^VT*AlEG^!7NP=nQUQF$%PTNR%Z)B2DwV~m7tNtNU z#JZzOE}r~B`Mt%hn%2c!+Q}dD&gSKXqS>%*0&Yi@qYmzoRSn@1X{duA`TV=+!Ph5h8H@?b*F?|y~GkL{g| zV4WBRSWFbNEfVFBxky2@QmdGd|CTNcalwYB%@q+cq-NjkEOUk@RLsu+dl5EC3v}r@ zW^uIBm+F)WtmTP96-u+B4=^KRL6z}~b6f{y##U8s0$C1gwO;q(m9%kfwLVwmUGpq4 zALPEQP#J&KhXQVAeE|Sm-8O$xT>{r7Rgd!}Ps*m=WP_EVaP5T3xrDV^{7=8T`%fT+ zU?2Ca=>?mK^bMibeq$}Bl1h$a*4Ht1OWMN`niC^L_i|G|MxB#T64@h{t;iEtv&vVYf&My3x(LQi7I+&m&THId*+r{ z$oO=o*ftHsnh}G}G$K+7Dury_?9> z9}6Jt=YApiex31J78WC-7#@|?02yweuCk*K!#hqkjbQyB{;I1!WffIM9*VLN*%`Uq zT(}-K6-Dln@0u*}g3*Wj3M{s5LP>N;?>hZ01xysTj(xge#dxwwTp(;{VB;VeY zsWUeyIM@%emm?Q~5%$4n2K`Fzg|4{~xe{V2c}3$L4SBvjCuMIS5=lFLc2jbdwW)MW z?rgX3IaVD@(|n4HLjeI3EjD!iu8j;4u$i2?^pUrt*8obTyfh*mF7Wz_h~>_>Vdx>| zH5(#)?u0*|2F~MBFBQx*7cOU}&|Y$~cwVr#K?sB?EL9}(qxmNM@Zd=22{ z7K@t$+Ely5QuyAHUe@vByb`p6klO#(#ts%Lm!}^RbVf?dQq;Jo7>8}#6EB?%`Z*}L zK$$%yKNK~asBuV#wC2%W(s+&;HI6{pH-_)U8SlAjZ+p@%Q~w!y@{aH8gQm=Uq6&6s zZFZY0c2qOi+2T-;2$co9fF|0)tRX0(tFm;}%zImXRY5{gwvtnX`f>F>+h?m4p7nH- z?_ZV0^j83XVY!G9H*e%PL7ki>-jdUU5wd4^B4Qj zNTM(`sQMA^v%QnFM8|Mzold;fOdP~Z3-V*JE_?_Z)U5Soc1h!rXXPmL{r4F$s}Zgi zetyCtz}xe#O7H`lM!1CgZ%3wu*J+6sCabeJkx=@T$*WVMwoSXcN|#AhEfDBKbLd*Z z-{DrxvZnhAq#4da(Zz2yvZAVKH zJ;IK)NqJB~7_Eoer@>GyFwiPCRthiuLoW7}r0YJ%7M6W{Jph9VVroR}`u_W7 zkx_6wy1Oh||1CgD@=let=BJj4P=1EUb;Gg%2oE^Fqloykr^gc{B&Y!*Yd;i2#?D5D zb=q!84XV!E);gL<>{tAl@Hj+?uZqm2A8|52Z>3`VQ8UOya+LYo4H(fA_r_l%OkT(p zZ;>--M9ftBlrOE8IMLoFI2&3~{faVfTfbCy49|0n2G@5(U7gX@o?hy(%AilHYi$8A z*s+VNF>Q=Kt;{UUfQ10^si`TEh3#xs$+x)-)*r!rD2m0V~){2F;Va5LWGAb7G?2E^jWAYq#ESZe&^Mxw;IErT2qj zcg88B;40T{GO(8e3*;W#XpbF0naJk?Pza-fVd<->faP^v3CPh|mFW+k!hmiKMAr{Q z`9U_(P{gplBmiu}!fWyDrKu(kIgT}|RIT-6k~EzX?ycr*9Zt6vEj80$LTr--jf$tx z+3O-Ri)OT)y6Wp))9w<8DxQz+mh*3GKz=rWicSF1f`NdWZ3EH+FjQg%jaCJE+%3F03SP2|!wwINpM2KShI5BNy5(=`k!K zDW%&8=`5NgWu3x0S3Q@>U+WCU8Q{1l#vQz)8GLhf;-D28&W3#Uct@vKm%6=Qe%<2C z->UxDrt>B5%PtXLuA0wjg=pvEWiE_TYWrA3g-Vp_+!|`JCa+*B6lcsYthqAivWK=w zs_(@cl`U^cE7th(26rc=^oifHC*pI}rD!~L@=K2i98n`dsp}7;yrtxtQfCBej??^E zcqF>Duq?7Wv4>f|RYl@!dN0()R@!Uj~}|3I%y2tXdaPchlmJ}3Po@P4XJ z`D>FqmAsm^R769GxyQ=UHchz2Os%VnOJUO~ms5PMk;lmgRTiZcvi&rALl=j3aq_t~ z(Jy6H3PvdS7DD7?M(9*c1ZV3#d_Ee`v(sc>{_*MV$j_lM)d#!R)(sxIh zpJN#8>d7djcfqG@q8HG~q=IgXYcP5ro(4O1GK;u0&F?Q$ZNM7)XcXxh>hx7AIhrW# z`wac)@nqbk!%Vv-#H?#&7tK9@1wG|#Mp zRNn*X5~Uc+ckc%e?Vdq{6AH(-9g9@I-Ln2CcS9mB*ZS_l7T82#jB`8Jh$Whw=cSL1 z5lw&7`CZap8CoUZx#Mp`zr?GNy6h(EPOSK!*lsYl5YOF|P;I z^MtnYEhS|FM>Xt=KTwq`K75B}EQvE+?RNNh!BAAK-HHOU)E{ z&q@+dJSxg~4&I%6c5ISrYb%Ur(r+4H{Sn|6BOxIz{Djt4x592HpGzB@5%^c_Q{pbc zAZgiFy_0mcu$qQh-`-h&X3$#fC?6M@d*txd97qd~^*(E7a4&>JkdzsVS4`Jj{ZuNxHHO+ZGK2Q4lmbnVBl-uem0t3z8N#*Q1UTu z*ckMtz5BTOp{%yOhWH%*iR5a#k#9<)_^xyUvY4Eb{mad}SC(Q_RD3{#(HN?k{+b~H zqA%}aK8gF-5UG@;sRQ5r(S%9y`Y5+7IE|~c>|LC8FQJlsu~XrZCKWT>6$d) zPZ!C7smLfgiXK~P8cykGr0&c-Nl{&o$86Q3CW7rIqR??l@&D|E$?v3EKBfh^G^QC< zEY^E0xElg$IS;bK+}~X&HdRQz&y!b-c$I~v0oLLYxItF+!TvqUY8i?0AQ0GM=%oC18_UUbenS2bnleTP6Wk^7%ydRQ1;V1E|b@r zW8%B4tj+=}AjiGuxlaNj0_!;nmvY$*jVsU1yXxoOuQb%h6sq~>tX>({%RGjbNnF>t zdE|$*0q+Zj+?X2Ng*I={aes~MkasweTXjKBLVnygMe82WJQJv>7+?v_9|HW+*dVg+ zi5qc(wRTXZ0Ky8v)Vm%~CMz)LqBYI8?lffk{7^6d_C2*&F+Z8Hyu$YOLmc(@XO7)h zbk8TjV)3VE3TtjeaQO60bv9rV_ zE#*Ql^=fMbn6dC6ZYe7yC86o|9RbzI8ooQwLxEe z+gBt5eW32*26C0f7y30-&8-v2AEm9;M~0tvJYfe)AztnvyvO(Rk<5HC)SDBFGe@@N z#o<@v_NjMLi^2TA*!?%_MKDIfi_<8pZva~c0P^BR`#XA~;{v=z7IOSS3LM56+@R*x zGNfsNT%S+_`y|ha`moi(j})Iy<(6@ci)EJaeyNHKPyhRZLenWA=-|e?)vQ#TD5L#i zm1hlOXG~Y0N8^-WmMFEb!*VgRKzxdR2urZDWIaf!dqHRg#1zAk9cykGnp^n;Q?mvV z-)9ma@zG_HyKS-Oi-yHHM+nA2c=IN#DvIw;!EkzBo`9QW#M=r0$p7nIS{rSqCXy*d zqu0g%%l=TbKWv>N+AVRZaB)sFJeS8c*!GDT^;i}DKY-kZE!$I01*$-7G1tgpl;KoG zD!S6SuQaW^;wVQyO;is$CR(=R%WCLppEBI4-7@$M40)(Ef%WdRWlo8~Y+EG*Uo(T3 z58ajoX&z6s(-4KBWZb?=oTWeEPeshAoe@J8{}Pe_H`<3~SSl<{c^y5~GWXx-_$S$* zv|7fTml*aY6?$i_plKI9lZN8=T^{X$T54aVc^vS!pYAEXEpfTfjcOgu`9lhtlb-M~ zt5}})8=_`uv40YHGyC&(pRdbJH(&xmp7c^s#H=stj0;w~y<8v%8U5o!JLGmRn`hbM z1;CJ(xYt;LKnCDeqy&YJt?N{1iAICI?e_Y1(XqC3s%0uuC*6FIVb?jj)WceXX8;Nj zS>ZOpO-{$hQ#lS}W@jAFjjHciSqSIzCff1_-e~?t9LN{+@MUiKXoZc~z;k7z%BPp? z89t*XDKfrPfj6q403u3wUY2%UbSQTenz zjGppYF2{G`X{AjjM6&zdL@t~B9rt)C61DO^92*i{X+ZEhLr5JGtr;TA)kQ9cvID=Y z_*Ax+&xwPdA6EFXdtViLI95OI(26lkr$l=+X&u$(RDnHHgN0uRGfLHf<)c;lfa^av zY^n4-`Jn0E8~_sbV2YMTvCX#|UC^vOt$>gB0y_@merDj8{Xc~+L1}w7)#9iV9Xa=D ze27Tr^mvRXihUpC^6sK*>5Ly2n|e}Csw>{8Kd>LiD>Mu5P@Zcactj-h+n*rho%}4+ z(g`5TG}UHPbx?u>(Xq32{8F>@j_Ybd-t4w=$~C^n_GznW2C@s0azxASk-ZW3H z=;;Ch6#+9g36IuJRiF*ShvX&Xa<#uCh%_XgkV;&t0TZEa3=ywmSNkXH{I3?Iy7;(T z(oQYVb8~3IrE0%MYP)u|d@|}|6;D z@2>w%vU*N8;O0khBarxb<-(drm+YV!t=d6*;N)nhC|!@T2I~3ogLvnW0)l(Q#tWj; zLT%A0uJWK%ZG`wn;zuP%S+v%}p%CWZ7q;CP^7Mv5FvH|TQT9Ld7URBfQR%F|(8%bhPjhHHjWFk`A%{yvzT zpm8cBpQVwdQ+p6_5KLr*ZL7%#Z}YQu+q}K8 zwf7BhKZpj-#22?G6qMYoZWx&$_ObiYN*`!gEPR5Z6#hav-m_jSNM$B`2ONQxc%82u z{0*?Wdi{7Q{$Y_a^*t>nHP9z!z9*3LI3&&V%2%sla*}3gn|?tnjo{rN7kCyd zs-P0ornas7i3TspMbH~ISr^cGzVJBztmbFF3kovMJC@6>CWNjFc9BDmTs4RNR68d_ z-a)A`no_5eH;85s;{DO(G3KMu;uPX`V?q8SktfOL>0&i8mZXVG3k5I!cby&>u*L;v zNJm`ee8=+kZ`MtMXBH8l&gz;7`pJ2QP1+I+$q=pj^iQq7`cTnAb!Ie)81xA4+q&oq zr!~*R4AHFV09VE^T?d&%rG zJ~=!m3pR6! zpVipRYHk~0QY^4Wdjybm@E>uUT^4x`(AdAKfN|1rRobMe%jLPJCHIcoRcLpvprX%= zPL5vR=5|ma1IwAH{tV&kQ9k-LF;#$f4vC2cjeR9LII`xeF3@Z173LHn6(27--bKD4LZ?5h&%vtLLSj%Ud=+ zw0*C3eDz4*Eg-1soycPVmUQo|IL~wVHj>}B3e0x!Y`2G zdT5&M>ZW$^_XFqYQ4=(K?IFowv8)gsBauoD+ z02Jmo=<~x8$@mSWm1MEbHoZJAbK*=Gqc|%lYd8?2-(J1bYAkXmgNZl6HDYK9F$(iY z7aW(9?T@<~dYCH5y7zwMt|EHg_ePND5p@p-je4y33z9o_*F)rHaa{LFPA%uA%G0gY z;WMQpfqt;kvE6>g-L1g;R*L|@vs%0c5i$SMvtG>i9x^2YamacfN~9cRFgjO+)KA9x z#U8E`NV8{&+Vbx(Jn!7yFaeCi&w2 ztqJiP>!}U;FR$D^+)jwWr;IPZ0I-Y;xesBwi4&p6u|Rgs=9OPcgeI9L56YG_Q%0WA zrRM+%$$vWe5B-lJIodFk=zZuN<^bfkuEbYWQ^E4(w;0jCz`qFrTQeku>dtIwq;6g= z>c>N^2_e=fJ)PCs(X}nd(Qh49(Ux0pbD_Fi`kF!G3-lm?;WhSEH|dlPo0 zIeU%#QHYJPnNcr#Nk=Q1f)*4s0z7A5cJ530{D1t^Zysqu)n&0bp(ioTJ-2n6@rl00 zXez2ZLh-!N6pZ$f@pN%j+y7(iJK*7Jx`$VqAR!1sv|#n#i4uvrdhbNpu&myeAf6yP zv3ifPR`08mAc)?H608=z6D6MayYiIxeg5tD4eqXcXU?2CQ|_EH=LoH{7iWgqPOLhm z>r$EGJ=d|5!ytf2Ov3Q#DNp%vYE# zcPv%fvg3QKUp({8jRCq!i=!{CeaWhn`ea3*8$BjK?@;CUD5K(#M21FrQ?>rC(b9@> z!}AD0h5P9j2oSR&98*za-nG<378R$!lvqocBwV!}ClXcUxUZhBl?WX88Tyqo3I+vJ zj-)DyvE}ddIJqy9_;CcUk0ybYJ<+?wZ=;{M!Rr(P0zjefX;G?e^6HBTQ}$+oJuTsA zI5W9#Gy8VZh$LhqQzSbF*n!;fB~d4aY;FYwkIl5W@={j#1L7-1?~!c}g_!3)fSW@- zh4nkJz-#(iVp3!xNd=gDr{~V3vNEnSgbc8rwF(fc`rh`b+>VnS_d;sT+XWtzl4J$d zDOA{%#jK6h{!Xw{SH8gj4arFsKDMIi)3h|&%t!7ikkQ2UpE5Mzu0Iu*#iGfc;vwo(xIK?o4d|%J9ItT07RR4^I%t6y~h#M1vC=`B0#tGwaNvrd3-E?rKI>V z(8f19L~!ek8NY@Mm*FU2bqQOmn)ooqi9Eg_WSbnMYh7xo%T7m@dJR0##ojzL&Uzs& zL);E{Oe10U&;UV$mzkVmR8CNc{W-IC23|Pj?Tv zg*zB9arTw-;H>oDcr;@UOv7Go+U)vcLZaXlvKuX7g?AZDrLSIr+<0*;S|$vBd$;w0FtZ+;kddi?&_n$OyUA})aRIQ@wD|_ zc(%4(rUPgLGz6r>S(x$mxn^onqq+80ZPaEwbV zoQ%G)yW2m+)b5m()w1L3*~1ZIYfnuir9)!uo;E3#Ve#Ya9VY59ll)SL_a^gu zUTG=MbfV`*hI1PY| zXk8V)4te>tqiHPW*{>~R{IA^UMQ*W8Z|Rv)oIkJ!|{W+nP4}? z>JZ;Apa-Nk%OgXUit&FJllsBqyh5{W%@a4gmg?6yzX2AiGQAnrQk>)b?3T^Tp|`W4 zs8c2>???C3G4_@PJkh`^^=o>2f&|0quaUzyR$k|na*7OJE?GUzLv3&)bN54DydAXb z!(t)kytzL}j3gK!MzJ6K-?s6c88=b}+tcUb?z?YEV<39F?EJeUy=w0a0|WK!??bM* z*{Y~!7{nVN1_%q_v%h1;2<;2mDO9OG!c!vZaAQGrSXESAFjv&R$Q%eb_V z#s!#~*>7Bw%fpSJ+x!*RNL?o!BZN_IF~B69+37N#rm*~AxMA2zZp&ujoPbz1a=FxT z`ax2~f^HK7wh9QM9$>x;%mJ?D5rCeHf+bMB56n}^{__$ntKXKjicI>7@%KjN;Monn z2hrY4N&jeN4qg`vc_Hd=LV-C>Kg@7^%<;z?Oz1$+q`9EDRj3yRX|}3F`zzCE_h(b? zDIFJQbb0bvG(2I~?420tcFlaU+37)}!cAj=b(9j|e3{(<9xn0Egm{T{TA6wzaL|Lq z*aEH_BFz88ZhzyLYR!ZLWA}!IP7LUPVjab2x0k+vKrVursKU_@&a`d4V4cKleEhl6B&~SOhlVF~d_wmP6_7+JVedg|HUK910vLD)#o>Tc7IaWMRgREI zptHQ3%p9I}j2xNB`1Bn_fa@ix?THkQ5^&_+5E%DKxd2H@*5fYTSm|pUJ3LUH89n-v z_TC)V^~di|d>y4+-E~SktOdsZUI(Tys_+h;+>{Y_duBkWLz6ks)zkKQLw5hfM*VYf zpPs6zf>kwUa-jc|t>y0KAhBLzyI(z|Vkw=NexEuf zl7?(tz5MKlx(wVfl`eS{Q}fXodBUl!+0G^ct;_gI+xln5`*TeS&W0vu5W$~K#zv|= zC+3!df3r>Z&h9fM3OBP!&>-xkO#uv_!a?$Yh2F13Yv(iV zE--ok3;ZMUnWfz@l?T*euJ$0dQ?YJogMmxRc&%(MR{^` zJ_BJ3$?pPLWd(xarsha9YvnKbel#Z53+xf+@!`pz62|uC_hp-T<~;l22#&xDs-QaKAszdo2&O!!Y*Ux}+OnVy-2gM*`7vctI?T*XZ9Q zXQ$?(dVglC}lzD)(IM@B;pz;j?rPP*>T48Yy^}q~g zQowzZ_t!PvsMybz>)^}P4mYqZ!-$y=Mo5(VvX9m5yRAk+zb;-e#n&sipUK{$`%q;d zP15?=JpSMBeTCJB^}dc&Ssj*a3binb?Av~yh3XzA#%Uh-j#1K;lhMuOtu%k<-P@Ru zA` z)_yz0(gsZjv6BT{V|tzL#sqCk-;cNdsn#OqIZx-x{(Nh!L^^V%zW(qz|E26Qj|_xV zYj`yPXL?b9Qy`NZW!(Zb(p;a4QuslN$JA3nMEZVhf_#e7!swDIEy+zHu1b>O69D(d z^1jWvHR_bhJQHlKeZ3xK0qYSW;k~k+djM#k+829yDox+KJtE@v=9M|{Wt2?^%YXfX zB~*I`l>aNe&e_r8ndJetAraN;>GVp^ZXXVVKym_`sb>x4!J4L1$LVYEH6c+IIpBq4 zI@ndsCgLsZsHGRAFM2q{c|0^DnBfzPra43^mu3KVC{L$7iL^P2nQO-*QR9h2LW4Ns zI*e2%By-tCja-)ar*NjCQp(a<3T7CgJ}r(G(i}Su zZD}MCTwfN17Y^Xb&cz*F<`8a9{SmIsHCiwh?wIKmdOG*m{EG{)$Y5y{w1bE~89Z&c zfNOe%mAvX`SwA623BY~62)f;)AJ}6|X(Um@Gc9_V_p@AruIUx&_l}|Oiw020Pk}Hy z&@Bb}Yf`-V6>W$PiG16fs+?u9+{NZA;XfYbJH(!yHm0+cKblxo4VxcuaZex0Q#0PL zzBCqxJpF=ZFFaxs8uHv%*_E$h+Rn>*p-?3)iQAV~Qm@#mE7`JNw$2t#UOsbAtG8sR$0 zKk=SOVzO*9{z3SJVIdxtr19x;~ekd}aW$gb-KJt)T)+fTk*IaA!EX3_;+-L4w>I=m?%>D=e)LmS9Tbu^u#YgVc_^%kD zeC8Nh9Eo+$lu(7K=>4xgPOX~9#LwNZYT?n01lrg_HE9^~jVt|1%^7QA)@{0=%Qjse z4oWFGInS-SHF~*(#m0;ZgRsLTuIXD>4CS{D%WtKE@KMgc5CtwbUr2p8F!ba`E^rE2 zH=59dKfIuI)@4{a8#6@@>B1TX;=!%_uw(QqWo7SWi2yu88OwD-f^JvpuB1+=9*r-4 zU3yuBTU!E|I*ZzTYm1@O`UN4otxq`Ov0)kI%gW?%JVWhrJ%Sv57!~>%2*j;^Cw%&^ zJ&$=?)lbdgbMb|#$oL0;Ei&I;pZmT&s+L_Cr1;2iRgOx%E#54wpM?jgJ%EkKe`Z4wPhpmdaDucN@kH$EM0!9lrJqUaL%uH=YBt4!^IRNk~BUcdB5K1@JoOXw!ZQ18^#+DvF}pRr z)m))9|F_*7Tm+f<_AJW6V+RSX2;|=7a&M9S|h%GhFloqRNryHRXj`iy&}mQRQ&y`BcK~MUV=c02!VIC@C9bt+`aJQ z#7cuRE>vlUO$$hciAR`VolqK>!GZg-Q4Y|4coi8H>pqX`ymLqMEdS z+;xE*=Xogyg)V}=aJ|)j|3V*+Vt+JlukJ$xDRmh&EqVB9*y+x3;yy>3Q@Wh#5%b8s z?%@Nt9K;vtx&B(;MSw9?`Gr;qJI~vwEXc*Z%BWzhKZ7{oXk|BuGC0nLD>I)x;%r6V zzLNjRBX2vMpavF-uRRUf0lxjQW~~6bmPZc)(%r5^ZH%4vYM^QK??^Rg_o|GI(JZEw zCf%!Bx)X~%tWB!K?{Q>n30q^;PFOpzwhEZ^z&B4bTe1=B3tLX1N808$Z#5AUecAtAATqx`=Ok;|v+R;oYm)rwV~!=-@c>ZOv+L?X z*bNLV9ZM&-qP#gQ9lHKB$Vj8zdkum4-Da|VL5W4??V zQ};Zb^qhJ2eB8iVp)Hdxu+L8C%vZMQp=I+Aa3>`C5b&vXXwA}|NM}s&u5=%RqQi9k z3*^G#IyZJ9A*^aTvrB>kpVnl#u|K!tNdf1!NfQ{i4S>~IBa_-D3t47z*V%Ih_Mu@O z{1-K8r{cTcNW%SJhtClkzf}R~9%{|_$qWx3 zKxUb+;HkTH3$75OOnxa4SnT$|;%OQ>7F*u6**a5Isge{Os=avVOLDP?3+RDd^k0nY z`aj?uBFI=uD9OhP4VB)S-D#jAlF!qNp2~9_b7T~Gd*Hvh>7UUu8_&jsY+ppg)6Z-F*s+EVXNBkgy&1)&1mXRX^ zJtdZSJ$Q?bAOkZS+q^W9OCs4L;9Y7V=IBinozb$IDmFY<)^%;U3Mt|?qRpHBnzN@S&CuLkeaNS�|DOZjPWI9{gSyR5(=B zG}7wo9_>?)=uXVpRawfdG32H(%kyD^Wkp`uEVJd>UgauH6;in@uz_cK2E^^+t?urI zE4gz8*(mVNaX%S%EmY4-w=c<*OL4o+}MB5@gB$abaBfzPw2>m;cU|{ z?4wN2Qy(%K_%eROrc|%14gX@Ln~kz_CQHhclA&s0c{NAtWx6F4OMwT&9^g4eeppaQ zpnLiRt@CXu_!DRwupdsxxS0r?9-3C#*7T++pxf3A^bgLsr<{Q3nJ&~!h)j8`d zSR<YFs1AI)xyosFivw4yUCdwt1sqKemYA{E zmQ2TjOaV8MFR$ypPB#L(IP{BJ#iXC|rcoj%I`z$I3y=Dz_*aVs^#F(OK7{XNemha- z?a;uCZz&kQ*4g7;8T97ehyHdD+2t`aEgw-Qh+Zk%v1-&D5RA4{<}~JF2p?QWlu468 zCMQ1YDrYOr*0(^s)Oq)a(w98piFs2dWS6@}o!n6VTFN6;;(Z*&THp;Zbep%u#Jb^$ zl4rx~;K*mq4@WQ(P&LERm>Hd{xv-7w5L4lBp))uLPw^872 zbSq)`b3dd?bXAL(J7IRatRl#Y0!I`@NJt#Rq_r(8KtM=>5j`B_z{FYpjsuUQFFVG4 zD%G)nW+>EpUwQa4I&AHF+w((kY>Qh>EN6=Ho@vGxUdhS4^+Ao-%gjxbi=B3z!)mZT zGrlBtl|WL7R-ZFF^<|eM3TGmC?0rAt3J^~MfEEZuoxzMQ4|KxXTw`XXEQW4>xy1wX z?|vgQAcv?1j*vYFz)3fiR{6TIpB!GbTF4{SK_=v_fTD}q5?-L(#Ww#0(cVzXNw$BE zvgsPve?D_{D`29Jcb@GXT0Kh69JS-OHj=F-W*G}{9bg7V*e{734*q_2s3ss}T!0q? zHfvmdUs4l(1dHtTQ=MX(_n)HL72N9;e+xV&?~|+hpi?iAM&z-TVd&Su)D-(f%!fz) zZEuDBR*T3fsuXaz%4XN=I+5POif&&u=scOvy@X62KdgLhwc&i`GrrKKmHygtM2>WL zBu--_D`=t36a0cWJo}9)gw)(RA#3OXQg`v&tdQ@SPbMN`+@z1Gpdr!d*^O+CYvTM$ zWYWRngqikgp3@S7H0R@aSs!ZgzqfzGEqz<;uvufgK4CvNxUnBYX$HcJmhoHB@ipgZ z*FG~5BAdPy{@{W@8(!`^6QSbH;mht{FCD=3J zo7iaTrB;MrRQxA&_A=;Zh3U*~^}sLt_A{!d<8jBxn-STu6Fn8m%fxK;5CaJpi1r(Bmh z(c5NAQZsLBP;l~GiF;BvtoSvFDQ$=IBPdI&LtnA$54G&I3?T?#ke3`&V=ZmOh@KuF zyShYreOu>fvN;i+tB&bLMs8Xj?^K)4n#&u(_lc2PggbhYB=kg|o;e_CNJFZXS18io zr>l}hRh)qPw7$ifZOk&idne2_UlRaY4*(QJuhrR{Q{-&+tI#}Y%A<;EuaKkIs(urH zK9SRe(sjw`uHV!^1R+VcVxC!)r)j?-P|H0}Y+-VaM~LeW&l4@f^MON>g_6l zn6mJd*FW4&Q4(Q#TX>PBu}>GS8(&c@7*=?%x(DBAnG z3ol&7pWS>r{0n>Ff+PEC^XnYs;z0lVbb;e{yRScuE{=@9Pv=4V^lZCjfrx=0Y9t=L zMBHmFds2!iekyH$8#=$DyNbJsmEbC!s*>Vp@evDooHOL}KwLg>UtWkMZrT)xX61?MNAz!L{qQ;7L<~3MGO4*pRD31oE?=K-z$E8^GoG(>UkMS*%LAOGJ_i~i=Sw8(NRc;j# z|IQ;fQI$^A8!zQT^BlKu_*alBz(#9&j!TPVHL8hjL=0{J6kQ78UrEZ^BV>E=pj0ze z)XT?~CNf$|PGJYRZ7#c<(VQrpv%tBKa*uy)h>s_Kp!Z-3E!0QJy~GQti-mFNO))Zc zmfo#!(($LwrtxMytchXnGFK1!tZnjRdqyw3NfE_lR>rCH}wTV@PbQPMYvbpY>%|W4_sDNJ>R_s%oof-6B)_$B0Zi_ zFZAzZajsm^1{T>`YmB#PnGBrvRQsliKl1ptshk7C&3g#~@ma#0kBhX=Y$1|GT^cyD zflgxCOZD5T9L9PrvW-F+O;OL6#OlFj_a~0T)4#%g{=8ery$>T;-cpXcHw9nbr5<0b zh@dYzF2`H)&SYRSgcxmYZyElW!`$TwGGQ(?k$S>oN-6U(*2~OOPG&}-Bm_iYBXWBn zI15icy}^QFCfvHN$|#-tCIL%#2G@vt@8n;m0so_xAM#dKKg zc!&>cnN2+HYv4a(AG%{e_oyFjS{iTLgYIEF=V&sh>8m%Gp;%_OgnAngo8yUk7_-Lz z&SI3_WE?UVruk)bPW!q%e(45zxStP*ym5^NQD4ljnD~$^-mOiu@g8v0kT>FC>v?QRkzwd6&w=^q{GQs6D!5=giF4;`Qlsa66?P13f?X>t07tAXhFH#$ zO)c$=6*Kc3N^xfl;Nrv?yw~VD5437G&XvpGcq`KobSkC>&WwCebhq?E>q#KhF@85B zJZmgH3-u%_BevhjmIrwjg|Y7&J^@jo{@hC4mG@8&?9wxvzbg7YPH}5PQc?v9*glwN z_z=ouX&S+sa1psD99i#71VtgSFn>_<&8CD3~h${M22Z&+K(Vwxg{O z?hBTV1z*hStcc?liN0?S*dK2mLb&O5FD;{a#mp~~R z@|$!6q!74`HK^yVXs&~}hz+&=eM8Zwsl@6*gBAA6QK$0i!=@;)#kClMjNvzU_*Lg2 zbV@GoC&o8!M31Gjx7#4TpsY99uFBXQfa8c@Rje`o#w+)tS*VJ{Yb!REZx-qrUo=4R z`yUUfe927%L)x*n+`X{JYqh2{DqMFDHS=4ckHKz#qI%!gPl`NPZpUf5CYmH46Q>>- zF&>C!lRFS@!cVT>QE3QD?22J|!=~5KozXYwj7JMlNK!iXqa%RHTGq2I^7G&k;k$$E z;+P^HCmu0o#)e1_7cz0}9VMjY6n|(%OsegqxENT{t}2Pi0npJ7J~iZ$G}21yP;Pi& zF|ii$D9k8Loc`L;E|6{>s_Tk;36%vr=^s>CMQ7oL6!S#&^}jGQCOOTsb(Mt2NyD@} z(2o>+^OEeD$2W>JBN-dnz!Mcw>!)Kzc$rH~#!BvI(5+Q^5a>56 z>>u*epWC0Dz7H=RVDvd6cEVr%Mj*cPe#Lv~Q$9b1LSrGdCy^JxlIR)lIh z6~A;5Adt?5TL{8;{v*>kbK_Fnp2{LA9bU4Z3KV~VNUpn#-)UD?v-o=b}5@8>uhHBFPetzkq(Aqhtl z4Yc#rF}(;~%q@^a!$G=Fe~}R|CsX5g$SdY4T{^Z^YmtpZM(sS_ZHb#5_l)+x!`H!H zJ>goY*J3C{<}HWDrM|=g>*k1E2my_`oj16sabstG# z@vOWfh=?%Yx+&KQUm22HoMxpz6&{0`TcwHO)f?o?vXbT(U4yDUUQd0n_yHs)e!VX| zqrHiv|6pWbd$FK&*;^B;wM4-^c zp^e@@VWgdmUDCh3^Bd}u#roKpkw%QQmOxgR0pP6i&yDllXg3-YsO}&js~3z;e6OHS zzQD-;kJ=riuPYDM(FGW-S%T znNdV#z8pj}7*_R-qMy-;ExMH*xtf%n-Z-5}72Ttwzhb7Jy#DflmQ8c=EJM8dQibX2 z*264aMXTI$4K^Qb`>|>#%sA>*8^>9nG%!!~Q*dfTd zQ!BCaJWqzcuP&pS{SH1)3%{U2bm&Slg6mzagUEipQOw$h7I2lG6kee@MR;M4Jl~~G z`45sohQfBqXXU=!SdW6m13Cin_itW7l)oEV5KH6L>zaepXHMB`@*C&TbJ12sl2CD+ z-2@Rmm$zqt^m1Fx)sN`8c+Fx1BiWp_G)d)!zWARw68{{{!VOK5%-Zlk-i*uCK8f2< zA8cU~O`fN@;mcVn71qo2Rkv6^^)&W~7^$sl*U4=y7c_xfrv2XgXmAK2{T zRzFi>&M7f~wAAO73-+BJWJ;#;CA+IIgPyI4(L}co@~t9GOX5(L5=@@8%8`kmV}^0y z`PVqyiT03G0jv4x&WGkQrOHVf5yx=J&85&p(kQW(?&LN0B0{`VlgaG#zau5-I`$3x zyjoqI3V|Sm1&CZFji8rx@@$^8eHzg^u^D>aYpY!#Tm^vaq@&4pOJiqtE`B@Isk~Tr zd5J1zOvVm(FiPIH$i8Z5)TQd{)>1)vbjM)m%nab{cMR_+Y#sEz`%kjGL#TFaVVzG( zx7d57?W2BQ(uZR5znW9^y5KpzjhvdcPTOtJN@Q)dj`n%$vDo5mFg5UE*ZN`R8IKvz zU3OBi4Bf@*FPmxsx3#oNMz<-#UU`h_Y1`#o3QsM#MXEM4zC5)!VDl9clTMAv3rd1P z4rKYiwn5i9Ppay9k!`jn?D+=QwQ6u(9BYh_`)jCY=tk)YueV)FcMmCv2#jFBuZ6Kq(xixbSfF9HoTG& zOQa>sDfL4}EKqVGBi3Kly2K+kbql13h;Ogx7)$k@YkwCSJ)z@Y$&K*SEcqyJ`W=3< z?WI6$QpL?b0rz_oD3(I*w25Mq&zUGy+MEbs{N^8Gl`k)#uqo^k^10iZk3b6z-c7h1 z+q6zsa$1(i7T%*MF>M5Ff|)-3(H{@HQaOM~3zZauxuIU}tH$D;KjWp#mYgGJ(nUrD z+4l9HJ)Wk;o#6C;Ok`g;`SuOt>%PXdFz$UDKsBY4@A8;gY4F$sYexc5)EpN z(gNI=%}nJX}_@FT;~M{(=?SN532I!fyFW;?mV9eQbD;D_&JI?H||t zK%DBS(p3Mxvk4;AG#u8toat3Vl|+zXp<4%W-9|IJshTP0ycV&oS^-hLLi~aFEhi-_ zyZt1tr=B9a>B(qjk)y25QMTHb7q3YC+1XpvhE^RQPaT z3s3=oK*SPzL`!Fd#_m5+Qy-fr+lB(m)l3o?E=RtNs)*pp3+bIfhO)076MqKIeKPiL ze!O9vefT(Va?%qf4_`Gs9HRAAZUawnMdQV^vTbWuX zjtLinn(Z^0-%(1w&tE&|VV-i0@xBL+rcIxiX_&KNEe(pnOZGv%ATKRsN5Fb`OiR)< zkV|!Cj3G|4Y~7?W2CxOa79jGS!-lAc|F}EHnOJ|}0N`>*bir!U#lc;uVwX6W162W0 zZOPN9JAU5uc~#U2-CbTEZG{D%TV@S0sA8Wd6k%EV z@QW)uFBB&ue>dfQ!>N6gxVWcY!2QC3`$NOC+s~8P%kL5#IB9Q-n~O*26#b+x@2vT~ z1AUZNBP#pVpRzD}T-SGjfTB~gZ->v-%yNN!57)h3WkMOtI-Zm&g&dJAJI@aF9R@vk zAMh7WfZCW&y3h4ovoxyHqHrRCozyE{+w-lfGOXjv;dW-3>ATN*x$l|_I*hcU1Tn~( zy6t(znivA@$BCBfqpH0+?5wDkCb;SAa4?#}rZPv7Be%z9Xq+n~4$|xHM?+j@{!>NRs*u zrAM}eR)c8TZFJ^eTM!W}4zc%FtxqK~3!vrg#NTI(O}0+6wiMgQU4)I0>e)O4T~968 ztNqE2x*pcnOCQy-tn^8ZyzFj~x!R91on|-JNi+P4G40z2c-rY6V)hxgBL&Wi0Q0OwjCqL4Xe;-xE<&YSqmQp5S z+tRV`dHdSOoB_||P)k$ahpJJ@j2p$SudnHfIoe z)m%K8u*x!=#9;DedtFahr~1)Ye3~xX;8W9ysrcq@lJ+CRT{nd|yzT_(l%>})=QEt? zuh*lGT9Pw(|9G`W>s#Sv2q{B!t9bb7z^Pm{8vST0sC_ij!FytPMQNcWM=bYz=yptY zIhl>5@u5AJx#E}kQ6Jy;xoT-BmI232WoppfKSwjKRYf0VNU_{*p_`q(Ids(@>^?E zaRE36$;%tB085Nh61NUb zhi{D<4y8jq#wum;Cgam*pcEC1KCXlb(GGpfXLqH1pe)7l+z zW6g&RDd_I%vVEJU>Z1n`o+!P`2&NjFcI}|+MVG#T7ynsBjlVzmJtiOx}S zVTysag^^MCoS4gapw~f*HGfbVM-8@A>lLwqy7IUgpd|6+g}!*%xs}pY9oahLD+k?y z*05~#fJRVtGsPl;`ab-HvE4+KZ{^#nJ$Q+7lf1JO8k%gxWmu%?YBpo0(=^RZE|PE- zTFGAn_W<1<{Qv^V__AZiE4~`dxfF}~6^UGi87D3TnTVIhjVF}1jibO*TNf--1((ZoO5{51lEX<>EM^4Gx=%tsPo{&(+T_`4 zM!gcgrL8MGB->;x8XSg(g9IV;e>98QfjMeJJBeuq@DI3V?L5_`y*BxVe?a}NVT+2S z(IX$N2_Iy7+L-~ze%LYdLUJ06oWa1PiA?M11SQk?W4({My}sdPv_eKs zE^c`0{O~$=Hvv@Y%fi95lhzmNJ}#>pdNvwpSCc*B*h#EyON@2S6+DUYn1ArgaRfo; zSPM(BU~`ev6px1+vS8XEhFGR;!DSqDE&dC{T^!mEC_H`Qb!HQ4O*=c~e>OP+Mqrj7N|4I2 zaD|P}t0{B)Ai`&VknPu1?{{LpOum?4LUx^`8T6{u7BS#F-r1m+EVC1TeC6U{oL{$s zD!4S=GKn$2btdpQUAyL6l#!-MDYpwEP=T>i!tv&P-MxfI<|q!SBgA5#-QcIf4t(_IxVlO7TEzGZ(tB7R9BPt|v9fX6``47zRWWcj-hGyYu6Ue)r!L))< zGZO=!aH6AW6KRan8E^MIDmae}0P^2F$om1fz4X=yUos8ZT|WK`1bw4lnqJ#7?>>q{ zR5Lz{%I6T1dig}0tO`pnc26TS<{7LMVrwz9SIFFG=kCmtsX0<8)-S63iV!Q6EdpaoO-E}_ zOFtspd){N2awVI-d~;yx(o?dRf2v=q_nPKx*I1Cf=oqW%7;gohEYp%3N*GJvq@^W2 z8Oc_aN_d#E^fo%Ee!P^_7UZJ^=wix{u(xdx12{NCu1b48=bVCLSC+Qaa4IcBNerU4 zQ|H<9isKnaZwU(0Ol<-Li;YiM5ZW87=n2{PiEDu?+armqcO1^cESF-_j1qjRW@;a_ zt7m(2+0_?Vu~1MstL9H&(o{LzIQGB>G*o?um!g_#%8TgLzK@oh;%PBpt(ACZLu(Yv z2M}@6b-#cD&z=|ie6{m^4>)=sJth|IX)9!Tl`+D*ANO%Ps=){18XN(v`n(B+zy)hvxCX%f=cLa zMx#s}iy=e#ghLHEJq%_fyu-x-!3wN}+C{0Vc`JD(qpd$0PH4tFkz%Qh9M9YKEw)8J z#-Z0@&E_h9Qd!t(wd)tNVyF@r3l?aLV3+~Fpa}$%`&hq9z-`Ny`tuJes0q0S-4teH)9{6FH;mARFG``4vNSer z&A!Jeg@)^UYE7;YNzkYZ5{AYIo<&n93MhDL2UUtXvd*8t%H;O@-fwUmE23Cb7rYF= z((JodG-@VUCiWfxFP}p?|G^_W?^b*~F-sWJkJpD%8*6&e2?k20LW`=^VR?^|S3N4~ zr_DwZAg;Bi=sTl3VG+DzDul#BN{wSnn$vU`X}K?_Qkhm6sK`zbExvO$2M{~-fimyHhhzknB z&KF*#*y|?t%KFSlqKJH=r4;bFu#)^oFW!vFGw=o9V`ij~kYpRzDJ^ApaXr(;6d$~A z78bzKyon+(ou3!v_zI*HPF# z?6KeY!Yr|CLT>0I#s{+hpc{X27iDRvt2IHcMyFOmTFCAZUns;}?tt*@c5<2Tuq;@n zN6Y8g&>CRCbJ};h2x!cC3W|mcu;8vNH{;YbXx_AGnOtA%>qj>FKuscVLvI2gP?9^HLc1 zLjv6SMm~dACh}`9ZQ*{Cq8DOs5kI?$lzcUnjum?SzHGS^swu~#2f6lfdJ?bu?#?|& zl1CS0tU4zc`lX$ZcJsRDwV}(0g&+MID_VQn#5mkg7?0D$dJ1r^nNAJsPM3-lLq2() zomO0A&Td{tD%W-g!>6?@HCB0bY-uqMM!p|AweJ>Cu&0JHa5LjK-Mso<$JI@Fr#4qO zbVGY^EZY=pMBFmjDb#qnWFO_}rliI!mxEAw=U6HtHqfnaqzYG?QetW#oW??m1LOKW zfNlm{qxoM`(eWXBw3n7}9p3W(0#Kdc~g>h|q1;p`1|EpIH~Zd2G)M!~((W%4_O z)N|*Eth{J^XAA0IF1Be*APT8?>*e{>@D`cx>`jFvVmuG+!>il9x~&9NGxb0HOuF(M z@HLzRYX`htrEKS~zMIHe!V&C#G!1pB@VhdIScy5D*YSOolJ*>(SiNLY<){l zSm{P-|FX5wW9qVxL`FFU%UdK-$T#eF6)x>=<6q< zYES(s1*iJUMw?iq2yA<%I(24?3fm>796ri`)$q8~@GL=h|6qrH*`!lFnv1!aY&w}u zzR(8#HTZ{p$Pz{Tl-%vV?A=6(ajj@|jK(#EJAH7e+D^hvcn)xTni*IAljWT@!rK=Y z_q+$$DWgoJ?gYiJY(4kjT4ik~GW=FLWn))flJw;x-{+C2lUC)?_Sb_@Ri?tyu#82V z9?`#hkFoyhPTxVbTQr9G+f1O*M;L6j=yT}J@$DC@i)(@3!csTF{o}pO2-#egP^ZDW zarB%5{ES4nqSil4dZ*fvAxbOm%`iz9Q92-s4!LWQ8%60J7Lcp@VuV z$|x53=7&pXSs9WWX`mZy#%*i(4vijm-Dy2Y)1GAWz;F|TLceTjCB7To4Jm|XgscZk zHb7k|g@~tZ-BXi5d|HDGN{$;H0%8qjF)yn;SD$C(OxAW1sfouUyT_Za80Eg zz=m=OOMJ&M%Q&Y$zu>2Hkqm%I~CEqzj9ObN$3NHbQf)g(R-u8L`{@=+vLv$h{ zauf;6Pi=%92!5#*XvwY)=k%Ieq<8-GLiN4RtW9Yx zV5Ajrjpv(xP2Yz)TD!TsLMSE#m+nvRu^y_J|XgR~UB1VS9);OqcEqenQv9o@tck_;9ACQxzU`}J)e z2KrxF-0dY9WUe338>ni~zi@Je(u;BPaar;5^3jWkar0a7T8mnXankei@(b|r3h?j= za`Ewr^NEQI3e*3Y7^LulZxXK7uf(-v<^QY?_)n6-*4^D%oQKED%ZuAffZNH{hKEl~ zOpJ$@pNF5H3&_Fc=I!Wif#7mx%wP ze>Tj)SylD#svO<8e>Ica%E^HTVd2cf$IZ**;K1|0l>EQT{ZECQ|D#?1s0?@jR3`90 z)&x8Q=>I1*{%a}3|60u2>hD6%9 zqno>hqZL$1R?5QzW-ShQle4mb3R`mVi2x+G782y+vak{r;}YW)77?=&1$vs-QiA8d zmHjuZ|AZ^+WaV*P@n3Ld<%I=#Wn^Tz1Z0HdxrAioNq*U zrL>{0uvhdluU^3%Vea1a8XlH#n42xsm7brTpPQeX?>}JuS>nF{`)dmRHYJs&SO@Z-$K z1*2zmu&`3srf0Qqbr2S0W0TR{n61!RZAtSqi);WcClo_|03XMU^eK6G;ZwGR9}J%23-+EAB&-n|~C*E1RplX?NS z@UVu`D?34L;SxN5-uyS<>zu#t{u9P8EdG0i5&&-QFn2igpVj@UO9t-#&%6Id;(E0b zhg&$>NHQR}tf8+gJmBsOQcwpA7@Qv%>aI{{SEwV@!`1EoLh%0$)qJo6QJezNi&Un8 zT>}vXmYBfOS|46qV>lEZp44QZgfcY^MmB5I66DYz$Q}(YfqF@5+YMNN+`zaVgI+*@RS-~Z_i0);^J)=;AfDIcjmY>rY`NI&n~W?H4A;uJ8!(+>Oo~$sEe-IwC8o!h`;dWv%l}jN zd<16mySlgtbSo%{eYxSCwY@|!_mr9CkN0a8GlRQWK0kt1BsXbU_-qaI0WFFpHrHf zR}74yfTGlt%w%BOS;5fA&JGxH7;)+85^ZDZYHnfZWMb%I>Evc)YUE~U>}2HP3hZVX zyBQj~I+@zpn7Ns`7@8PZm^iw)nV6cJT39+68M-+E#hihJnW29S<8l{o$22qf1kS^)&4wVjRknZkI=@dkoJv{r&?9R^Y zAMo`DWdOO{&w0nI4vmt86blC{50qwWb94gA!46}G*&2U<3JS8xnb}x4TEh6iXH?jv ztUfxL*|SM~G;%bPFf*|=HDgnNaS01U9USe=j6OkK7fO>{lvJ;=f3=+y@MuFOkYD5M zv3!Xi6UWT=FAkuN77gK`pr4EB#1=$hq|{WncrVc|6;n+K4GZ>b(T^6F5JE@cet!%n zUEj2N_i#0Q=WK2vVr$XXbn8`ZYfk`NU`CvPd8T24L3zOnaqTAetEd>?(F~w{{zAS`a z$>VRo?Wuoml!{}8u^w!a&etw(k}f=AdEqLjEBCBp;n`Y0N#rZ+XH|q7MLTv}F@js7 z0i<1mIRr*qgjZWzLpv0n5Tke0QhukJ-sU%}e@q{!EE+lKg5$nzg*JUf5mus)3*o(v z``6_!3WMi_!BPH#yIy;w<)gedl}C~2hgew-xe7g79?m2$Gqt-JDwf6j^RSO>ZA_v`N$OFrba4T*H~k235W>bE_PvZs_}>kP)HY+R3b- zc|lDzi4`gbbGpVI#X7Q%h5L3p-t8Xu*M@`dsFmq@WP-Bg}+;@k9RFED%=v*Ab$ zk>q@7Di9QYp}N9}V8C~X;)Fc!H)LId!1ckYRA4Z}F}>lSM22Etk z$-L~(p&4T=#$ohl%vn}Kb;RLE&Io1|Umg=G#wQ9`5zR8fkqfe_fcpry*eREc5Zq~Z zO2Uh+{|x_^t}@9=jERX$}9B9hxme&l1lufY&X7ke#cI=F8u5h! z9)fZVFF3df6kig@1@w0@cj1@}aNM$^CWj~MBzq5eFys9QG&5YT%$YO&i091V%;Jpr zlgtBcG2*BbZSylPX*%i%F@E6Yn!R0j{ik~I`s#WzU6Es|%mAJafpz~g23L+&41-9K z5E1{AcJwuAqz~{8esmOJzo@pbXRt4^w2@|!?gKD$`D5N((Dq zhry}uQwragm5J4hz6*Sp{`xXXei=Zi6O?CjHyQOAFFQg=mGJBT^Z(8PbiGde+NYjVMko z`}$K}-=rtcCpPyJc1b6j%Fz#W4tjUlW>)ij)O_ZCjlsG^jVp%N{Mz4%vxtjyGItIK zS-eDyD3%hI63ZKwvWUo>e7(oq7^45iHV>WffH;_FFv5k{g)l0aU2a1jD!(sJmVzap zQb<^sCa)tOJoI&layT)0I(3VOwW-aacr?of(ToUFVgSpot~d+*gPc%VrKu z_r$EduUYMHd8QKFNyfzw!P8FGa|_ejhI(dR!!Lf?Rbwt3R_tX)rW|wV3AVW3km^*i z404bDocTGcFv!|bHdfYuoY0cu6?QLm>wIgAR)$6y+=7Nq@|vWR1cg+GWR+J-z%Ju^ zx(&Z-x_?HJ0FOYMbEVyS)1hO-!CR*YN4=wk{hR6GW%+&e9qwJ5?ab}g<$JWap0DO* zX3nmLf_FlR9_fFsw)VDD4%~*$(b;s$KbHIJkD9ZapIRJdjPHEf7(QD$Dm~4)&A-0e z%{zU4`TI)!R``tT@8IX89fIUBLtD135(SdE$$76bXbN_(XVT1S|M% zM70+}XxzvnFUwHLpmwNQB=)R0*hFNF92aPdUxv+QD}p`pHFGiN)`RGq^<5BdX&jS)VX0>v(nyQ3NrOi%8$#7}H z_8m_7toZ1nm!S4ct8T(oN>-MwYFmkC`vyubtlb_x@FAiUv1uYVL@sQE0 zvN6#gEaEhLv*Yq|=-+nH+5Xg^d#uuG&1Cn(^G7aaZmwI}{7=-y*7s0GTeWMAY1`S= zR`c6n(q>X2fqUnfBhERs#)I>y4)T*@wE1PtPs)0&h)s01qmwPj9*aK45AUlLT@3bG z%Wu1HaS@HtYNqRFr%0wm>Ra*35@ylJ-U>P1e_mVP=3i`F{;U1BK>8Qvbq2qhEmO{KVO+41zHB`?lHQ|oYzcb%ioA}1ndp5OOtw)g&8 zgjj@)e9dIx74THBKmL6^bnaDpu1&5h)A+gile*4q_h~q3k{_F%rA2Ml(djX-)BO3L zex@NN`3#$k2o1e*?z@efcj8Gp8k?{GSr5+&mDM?p?g~$CE;>9&twp@!ndcF(9d4fX z5_q^=7Wn3ZxLg0% zJDroC^DsLxn`b-II_=i)X|b zzqP-0uIhsG`110wBICJQq1hOLT%6BbtI40e<$N}E(D%UZVO&IpZ@D32nEuOWGqlA+ zcj;NN_|P2VmJ8aKP47Nvcro-3N#Go>$TI-}CMWw<($(o+`0%^hsl9Rdk^~-i?MD-~ z%(>0SnSX681cUpKhfk02qy2~jP%|6T|AJNU(G$jfLbgx;a`JF;K7IA+V-C*$1AtYX zj2+!Ro3Y7$GO{pZ)37oHXqbnSn@z^d%EHnS#>>UcCT8X6pk!t*Zu{x8t&N!t_#~US z?MGXC)z3yIW-tzp{~vb#FSh33disX{8$};xAiCm8E(WRPzddQ8n!wzY|03Xm?AIghX5FK^f<2PZnu9@uJ&lLC zg~dvW+p2}R#lO4zIt(!+ih%z)fAOY~2skspe*K!0lard7S~KKb>e$#=adB}?&8xmMoT~T{```o;|akBc~_4W0x6s%UeTOptFW#r^uW*Cr(B2lm+Nk@@MdItms78Dk4 zZ*P-{cG*}+>YTk zHGADd(1U`csi|o!cD6Lx2&JsHwzdoN3#27_57D7^tJO(e0xh26Xws%;7K0B@KV$xl zjBon)J0#?&R|mv!?Kw#3k@dm|TCu5_cJ!Pd>dGJSc-U6&5hu8~rs2W&N2_+MuC{%c z{@&v8Z)8SL;f@YhM#eRup%RM}-{)=Jdztja7$bEB!TU?ufDlRr@JC5$?B?&m&S;j% zr-ho6*4xN7!q+4sP59bBKh9Seb#`=gm^v_E%pT5H7fFO-)vdT6SkXtT{ZMbYzg(0Y z!>P6yeYpR(I!%)=s_<6j?LN25iJq{xbZo5RNh4v3*2*meJTZho(wTo^VU+Krq-rkdy zmfPL_84*cLWmrQ)!}Av}>u>gdgr|~na&aBZlp6NnquO_`+sk)}%tra3x{sEdlG@UEZ1oG}Qg*U@ z9xm)NdF_moSoKKIV$b~$P*hY@SShjPvxM&xb!=p(F<9S1;olHoVPRolL?Ie`h!ANQJut?eG2! zvKiD^D!&=w7Krl9I2HQ0G+ARcshZkI84Pt;4Spp@htgK&x zirY?xR0NSx&Gq@tY`H-eB)VsDad8e?kK^;2C|V+m&QI5~jaglXVelFsw36X3j{tWAX`rvp5Ee2eZtMUi`)Z@+NtN7DJq zw>hk5n47*B)yt>8`ew7-sP_H+^_7o=wDdV9na9n=p6;x70+>qVK6fz<>!9TIeeVC- zw%uQzuKO@Jle~T#I@YNvh1PPtoz@wQp2cgok^0(NfvDZ(XyG+z84i=4&(I{8tJYZ` z1`!dz?Mm~kzQ<8PlK#z_Kl;PX0UMJdV@E?kEI9bK$A{Y~zmA;)xv}R4UVo=YMIO^L zGh4y0gp5u5m4wfJsoqX^@kg2(9t&ARtPQRwgc-bm6tCj;+mno|EU znnID=!)ih|He>BJZy~*6^-}HW9~zH`K3zRMsVm--@;o_AO3X-K`{U^HSdz8D}8-^>Z{|X6Ze0|de!E`+~7QIS2IoFQ@^PeH5RX-i`@N+PmG^} zlm^mo?C+9zBsZOp2sF?mZ12AhxR^;rw+pbwWm5mz_{w%AUEA?^$?bf5cm&@#c5o`~ zhZ5OKvE+lPZ>nI(^`Q_P8<2_vqNDl6J|=MieRp6#_rBz&uegH0gA3 zNE2MnZYO7_i$!2YMVz!flKC9C7xi+D3X1Dya#=)PCuI+{{T@gVf0z8))EG`x+4J`G ztgZ=5!rCGx@`#HOQg6C714jDYW|W9R2syuFaSA$?kI2K7Ne$}TFBLWn(Qkpa!K@W4 z6JgAOSodjBD!#n{8|e?ZyOmXZ*y-@rR2UK%6-!Mn<$q6F{;h-)z)F&-WDxuW7Jdy6 z50fCHqthLyU$A~vh9wzz^W}U74}ROw(^K$IP*gjWVaf;D?=}utx zWeR!l=jWqL7b~0m-JfQ~k+HO-AAY;Z&n)ngOfcEu@$Ou)36f(gpz9}V?Ln zKqVb1(V0fW4E^G`x7sE`%C_2+__PE9sX4jv5H|1$9Jfa+NMz+GfBag*cpX6h!SiOX z=t}3E-qv7OC^qOE=zCv1->L6vgWih>v-hW+=$pAaBN-lVp{AhXu6#ObynU?2Lg4y1hU^bY@(q;XlwJ2Y7 zOeT&_@uvVPR1O)83bP&5IJ^)_R=ZwOmpNRdBTnNkBqmfEYHT#;{i%Lo9)tSN6iy_< zp7qyBqCT^FE|?!madD~M^M)|@H;a}d@HRoawBAM#autFlB65Ebhk{ULK1|f517@_( z9nW(_v{zW;Wul49S{t8{-4S{@s4|AVZzdH*hR==`>*8MYOWRz)rT$5?Z8bvTJrszA z#6TFR3AMBYnmt|IPeK(1^9lx+@}yh-A$&YPjg%i_f|HEd%!?B4@By>7oONO-oe!Ji zbUm0{n}-(3?f&MkgG>*6w5bea(+TC9sND7lf+m3X2qptoeuc`g7WcpQGTdN2OhMj= ztj6waZRvU()ppSuCw7{@?8H_SY|o$8uByOeC3^j#tLEkv4$`F24(aQ3VZ&d~@Qb*F zq{lHyA{;AmLds9tY?c*LI0FY-&w7Y2{dMYWx-*BGhTDvcjKGAbfe$S}S7w`=UtYe) zQi6^6T~BKm2z`TSH99irg-rgqCsttA`B~DCWWj)Olh#QyMoq$|TKKhdZ!7dv|L4aP z*&w7GMd!VVK&hP3r$zDjaQ{?3MQ6g%;>zOymX4#R6x|BL>z?^mbm}ZMUy>n>Ob^0K zjGH{%U0xrDv9WswzZ21-v2LDIyq4uG^*4?LPtn-Zt*M16TRv!kI~psOiRQDYLw zFEfe>1`f5Vsy(mm`S+7KOe{%vCXasTBRNgnfkAu0s$ex=Re`T%Y6yQ^R;0ok9ClWm7eBrJFl9J# z13;JN0ma{gS?QG1aW94Di-~tdIima1KX#huj5T2Jsx^qxM6WEz@}$uvkP;}r?cAbYPgrqr2+HVVTtPP?shty zizyKk7w6-2KAihe));2p;Cf;wAps8!*xkIbngLI%y!8fp&;pRcC6mouh3_+WbsQN} zQ)=!u@Ttc^DE#xw-En`-wvY3iO&P-(I9>VkydNZk0r!{}oc=B5o80<#aop!HYeLBT z@!o}s7OcDm_rD)*aT%T=R5$WSWk1tva<0wX!2*Yj5JhCqyV2Fzu#P#i(6{ou-Iv0s zJ-CyP00XO$?le2C{1J_9^%EIgYd;zS79nv`TpH76I3?P8IRH)(#60@d<`MTPm=D{1 zU{U*{)MlF3;vnT1^~chfGxk24ud_AClZptLV$vvk)o*NKVhd19dc%D!*$6bi*R2zk z2SeP;)AMT|2iu|8`e$70Pg@r_GV1Dhv7KbSs;a8C&1XGea3&W8iHnILU3Br5St6i4 z7#@FQQToi!wF+qG$z0{;5H8Z&r~v>iI}>xM7*R?2F9w*(pK3*&SkRR{_c)l`yP^#% z9cCj$@72$Q5m#m09p;#GyxcU|{?lgpLuteGH`U;edu7XwPI+ei(IRqePNkx|SOA^~ z7foscSX2-miwwQY^f>0I6(8V)r$@#=nk9n z?rw;F67LF_h<=ACY|ZRmH;xa#d3?Cij;FrGEm!J5;howUYypnm=ngMZ$QWqr4#(fY zdDn>{WZ-!wY3?bZ)4>Wi?nz$e4|&sqp%Mc>B)19xt7JGHhYI{mhuI+tZPeS2uwYSwD2WY(FfTHmARzF=U0b6}1c7#PIAFbd~6O;;=JP;EUATWqGid z87}E-*bhV+pTE-@iKNa)3xQ%4Hb~kf+O^P6Ydbp@S=<2FJ5CiVOI!oNCY6S5a?4d|E}j?IMSs!j2_3j!w*Dk9Wt^u@xgu+FO%(5N2&`%dBfGIekp4gE}Anu zbK5&5upQTcSxD+lwvME+DiIM95_$&Y@}fo!?>Q+y6>_Ar3jB0T^h^`pxxk{9l0P$dLQ52)NOVNk-NRs->k@^#7F$E3mP! zVzbM8u^we3$wkP>$xrV7t^zH0{+?dNCYzGad7g^SXXl6y4!Vn}v5 zq8>1@-9BS}I-LvsoxD-z;jV=ePaN=m}(@_=?MEhR-uM>mL% zOu-84%D505*;5Ni7J0nmY8!6rn#ZJ)x;S3GjpZ=;@L?a&A&<+$dBD%XiuMMZ@xAkK z>gys?&HTaYQDF~&PO@I{w*j%NZpnUdzEsTIeEUh#aa|9bj8n8odeSc|d&%+&@tdjp z#CUjkL_|am`g+NImWmX!G6Y=oo0tCrS*H2u=i>HIit4K(x~B7J2XIJ;Ckx=`yg(&N zpX@&Y>_iLbmH;_v<=`P|)_#}{pRxo(4N#T{c|T{Tg40)@u;*B3Y- zF=h&TxdYhi4Me>T3y;&a_Jgvf$aFor@W{wxAmyBGbdTB;OiWDJfJFiJn`#D%%{q`v zPr>|Rfc`MOn7#xQ$$EB#-%sX;R`oWJNLW{t9|6ToVAZ>c;8~&fGeZG&)@k50t>)NQ zoynIQ9yXdOq?2h7lFPn)wVM|;ki<6XnkC}X>HwiU1pC70;o4#cd)52)laJ4%n7-wB zt|VBaw|hldpiO1vzjKk}(>HZjONF?I`$;iT6a2w^@v4hU@tPy0X=1M<5E5+V=qTI-PG|L4p-nS2Q#- z5*ZO8M8oF3f`jB;4-~gm&-391RoKeP3OKzRph&PfBZ!6P@x|1uz${4lJpu8_U+bG+R6++)F;J0= z&z?HByB0$6;O^?G=^;WAfC|kAkSkFb83~E^NK#`!9Csj*E(M>Q&ZD7X~fvr z*eQ7HSjuQ+1yexXSgbO8m!9v$PEDOJm%=G`jiRLuVX}#)JaeJIl~GVYqvgD%gsQ;u z!DJcXu|Nb8tCX)E6Nd5e!~Lz^NnKYkIq&CR*AJu`RZL}7j22kq9_SCrktP-Pyo$fPk7X$p z)k%L98TX}e6E$5#yTRTJ)IYTRx20|mk~!bEa(3<_hi#@Jz+sSat_td~^+Uo@I^e}1 z8qENQs1&>%PU9JU*aN6s9h}l2TU#@*Ak5BxoozTAH|&sgkoVddlSp(n|DDmvS5#1- zg-gMJ05wmTb4!A?U7Fncih^Qv;Tn)lAg}d(TKnZYMBg7!0wmu~%t34{h)w-^5ct_M zIdd|BU^ip4o)t@!fmx95!`5`jS!Ovf|3ia~2`9G-)F|tpa?MIg1;7rX$yGL1C-n-0E)|~)C6SRk<{0N(3n>#X=%g}@4j8tnD)m~Qz=r_v3~?O7AYQ3o$q^% z$r&lUcn1r$2c#}bMj9}u{i!#{_6-kWixrrE-5F3jIcekY@#ag6=o? z;<0!|?*#8JCe!`o`<_T>I3A0=<-n6BUM5ELT~sl<0gUl?x6@YT6%`d=2fY8G9sxfW zaHH+f76Ac|UIUL+R|fYgZNq4BC(|Rc)p242mFQJSRf^!Z2T&?<^>uV6UknvhT7fWh zVKE>}2!=@=5Dci0>!HuHNOpq@}MB2e^)4mgwOTHbJHYBC@9U$KB89?cMcnX69UCP zS2{nR7(``UHPHaTYq`o!fEU2oJL~`_!EyJ(CLd7vfKrzgH;etLZ|2o--eeY%Z<`<# z94c_7oNj+i|40DgI5+5TJlw4DTbosQO9l)Q+MEG;j4H3K80pV%d8Je{!Mq;M2VMac zTC;dC8oAfSI1s8&f$&lHVLFUz1`lc5?uqu)VACn+Rv5x{y8^WoDFoq@VKScstxa4> z$uzn`jx(?3Qd0vz=F*jT^AA1WJHUN_oSC;_`Ulu~o65*}$;%50n2o`7prA9p4B?GB{zO&em1~MvuB(3LFbM!h+n?8s4+tM|G>*lSZa5T4Fdc)UM+( zxJ-Ip*35w8`XokHs99x-Oc`M)PwN@R4HE^%%ZZfl*kr|BUksIi^Pvo_quv%}3krgm zZ5=wQLXo75)Jeg!ctX{4w#v+dh0X2V@w~mQmz3Y7&d2FPR0j4o2+8@ko-{0iY>zQn z;rk0!Q%19&mr5z=--K)PavCUw@NT&jj&vG!e_86~vhzc~aeG{Te8u>?h6w`)e>ebh z%!(i)InvS;qd{}RXm{MiN^4{TCrnx~J94gyl zke3s4h!?HbkB;QaTjMjku{U!4DHKV1Bi6=FC$2bxzIR3tcoM}S?c8<;u*Y#2|lSNY!63lWh;}nIh%pC%iVTtUYGpAscr4yM*!2My$^7iuR(tj5N8}$g06bEU#MC}LO?@CYF zB+-{j1hmWMK?k6+2-H~Q0R>U*Sx!#?Pwb1@9|XY46CwY;ODZ$Hr5+?LME0;O==hTc)Q2oZB%7jBiY(2LJi+2d@`{ zI<_s{u-rmt1V0YUvYecp9@w3Joek**d7BYoz|ZVCG#8_ETqH>7-5#TQLc%Bi}rEr4^#euBW;ok1$t85?X9HYe=NeFTS;|L+qVN zCD@QfhHd-K57E!UK$mu+$HkU~ng@&#GGYi4uaNO=edfo?eHu&X4-WuTf1Uzz2nD2Z zMZoA+ASD(OwCKBN>;0t;tEr8~e7mu$a`oH864BH3Qk^<9G_;>tVwq^xuzZS3OkQY1 z=u|DlE*U~q2av-oP>=8lIUl>vC*2feeK9cj!cF15f^9@QwV(l-o`WoP>?h{?`xR2s zMaD!nH!EOw_#)*{UlXl^6CX z;`yxkA{G;c&1Mp}L^=Q;Y15NO5B9?Ht!aao0|H+gnEP9vSEuWK%(4l4dg+!y`l6!G z?sJ}C*jKw>+RyFk8@TK?B*Y$p@}{NavChFB-?@T*Bhp=_l;c2|WjP zi$nTPUK#{3y;L6C;|gXgJSlc_P0i#2(LPk2c_#!j@3U@vKATm3z#RXKiKE*o>AV5S zm=gV#)X<~Sx|M8jE&ztes)jlqAEhGvd4zXATG5$DX@SPE6fU`-_g1hKi?fIDaUmqgF`0xmzR z!(2bIO1G0x!};ZdLk4Bl{7t8hmBOPV06BT2d%sr}7b(#WN#`QECxRFofg~5%>&5*g z*ptkf6^1=u330w$*}sJbgk^sM6dm1Yygm_U;xk9n3mjQsVb+5c9qFJ99c^B*d?rDa z)Cx!vw5nMdHnNc(DKX6)xc1CH(D*9>M~BfD)uc3=#!&3xgVW9RQnc^We#F z?s;MC@1N+yD}>M?G3WM)%K*VI>~khE8YzD~Ml)DQKn33HjoN|zF2Au&F2TGSdhx8< zDC+v-5B1UzJESif241z<{pKnvSo-1D^_@LGfsx>db{z9g04M01(7F>q+(0MBlALq&q z+JOJ0S^hPTi?_Uij~?UWM#a+mXvq4+h>Hq1LGz;lDdcQ8dI#X!k`Blby-YZ9;7Cpt zQ+XMz3OpDI`O?2BS68mve=WrhM=K<;R0XBV3PiyV7IjFLUt`4yaez29!w|q45Lbqj?)&f#}sF;|ib}v|N-RGoPc$kJLWxl@%803bS9zP(!g-=T|HD zr6hj{T%JDbzV!FNdKsfjiR?Gtf_Ju`7hbB}CBL3i6!#T8AL0sp54*CkKCMs20(*(tl%BKT>J?UlKv0S4LFDIar=Nt z7#98jJa66j_BXyDgBI<7?x779Uxj|_$@6t899ZCb7ce?9f*^L@eZbNmYup~k!iRRb={BjuYB=w+=8klm!G=}IcRH8>V>FD{c1MaU z+TUrpgFYH?Q-!_drn+-+eqc?ByJ)==r^nc#NCZnw1g6X9s@$e+5E714$S4;RvVhi` z62{qHu2WZr`Rw!w@$~>I>l7RhqE#?T-la`pH3UDz&&l@olq(6cm0H*^!d4`n<7i?P z-=#z7*7>Li9i>iGF#&iN7C#|)=KjIBp#axQo~Ck`NVT4!Idzq+uA5Z= z72d&n@|yQDMEz(o+t z&+|Hxm$iNK$B1x}=c*2*)c_@#2ZVZkP!5}YiOlo#4Suw^%%Z0?r&b3z1TmQYPvPP^4c{GX=%SPxEf z&Ku3>@b|J0Xr+pO?Ovzb9Hp@|Lvk9(-L^1ib*uw0b|C4oS{8jS7o~blc+x^f9q`ye zPYTdL0@D*VM(YQ|R~6(`XuT|oMGsv(zwUH3t$5!aZD%0M%F4=_hoW3+(671a1D{}9 z7>xBDiI2R34!ui+DH1R)_rmBkbB6t^v(592;0^9@hK2FU>{NdEfP>UB+ggP5dnrd8QO;r-?VDVkT{0BF zfdN^%C%aULFr{Xw^Fv?<6UI5KrIgYe0jEHW9i$F-aN>3&*R4Ys%0s3%C{p53 za446~8^ik>g99Z`w4(o)@AF|D-a{#IJwrGX4hsaavw9NDG4~Vthm@IiTOBAh4b5+W zx+}}$4nZq})D<<;?r#mCoN_QNUgE3bsQ6|>d6Nt+ID3tK?#>Wd#HJ6Sroh>x_5d4L zudQ{Wq*&?wx#h&F+o&Cg2L`EEi$oBUkbK>cR2<0?Sq*Exm|pdHTy8E7IKBZ-VBv>} ze!?p7G8{0{M1T~bi^4~WhiA@7;e#!wsOCFX<%s3b`TOLt<#{3=AxP|~3I35d&6X9u zJ>0>ao{9F8Fn zDhl6NAz2eYStz0tW3}}pYV5E7UR)eb6recO{h+PJPLT3^y#FVUXlQ866DfE(XL2c> zd`9rnw294L9Y&Hv$(y^}?@_Gg@JE1Yz}YL@EP$G|c>S*ZOnANR}lq31{!#P7OXHaM$8Rn=e;rn%ga1pooN zQh2^@{TE7?S%GLd1*w9!T`g9oihIEq@D7MSxq7}JDTnnykA56uuYE$4`@R_eZJ>3> zj^q9ma*kgV4FO?Ma9$5l0Jr@nGYM%-B&Sk`=TWDf0?(*HdMtzkdK4o~mJ^xm2OIQ^ z_RD#}H{OK=&L0OfEy)=vijr#=g^$U)u-6SnkQ`C6oVW1{S?v}bMBMbz8#7@f_8N|YQt;Epy$luzFN&cH}p9a&6W-eahopq&#{!l zL#yFE^y^=<^c=CH~VvA9nJyM(6#3FTX% z@IGP!6W9F#>#N_@n6$^XY-j5NHDDmqM1>gK6fG8V{%q3`2>1w+nq0_FajDWIM~)t4 zG`6-o&d4M~BT4-ZnzC4bpbhutZS0&Ws1_-hY43afOX)H7axgFt$MMge=c$m~zf;XH8e@}XF%Yi{^ zA*tLj02DBOAOmb$Z!dwD8cpLl0Z&o7xcO=$yqFETefPO|nGnnQZ0E8YFP#yuntPO> zxw#oh1+4^dB08C%#EM99!DRL+cf5XMOI*(pUsV2Sk@*T^unz9sm|l0 zQ*&eL!|rKZ;}a8eb1GS={Ep@6ig3T~9bOw^zqEFL0OVPcDLJa*tJhbZ9{8y_AFm?^e&VE>a`H#yYpMMA~d;w7f6jxdAEe zV+ZWKJz5|H(Rsw8sHpfNur}FT{#!9E)p(gF0j*`V>H;-$K>e4si_Z;cZvuSyLjEz` zsgL^M>G$>i719!>d7F7*g_MrfpFZf%!)@C)^pA1dDdswJ_s|eXa`QjnI%u+kJnwJOMKI(ZW+yR%X)m#6}t1IvOoAs`AI!7@iLVZgu^ zKtBMb1|BLi^HA%o-qZbmKj4nc(p_3??8dDa`4I5Yumr!D$V4jTSbOM19BwQ2Y6yf7b@u8xmJhZZRa9|1DfDDMn@Yb=1vcVxt03QH^Puc)n0TX!yAVz>G+`y`x z1<~@4^Hm@sjfRX|S@Z}%K@d8*!nv$3l2O*ruOPnz`e=vpd~IzF81LO(U40{B>5>Ro zL_|a>i0coh>z#0LaKJ`GL?;tO#f(i(mO-0&)vg6ZMJOE=)n>jHM+^mPT&IsIDm;Sy zw3-ItAgocz&awg|!-YU7i{@RTM8sSI0b^vmghiup|0rzm`v|W zwG(m}1yJnpa6@o^hg!-}BHa(wE29fP?x7pnztwA_ypzD(2^$&XOpLx1zOVH`g;57%)=}G&$_xK;yp$txS7#%yN~(Q?e%t>tWAeiPY7n{kt&Xg zVk8YxY3SN5GKxIhd-a~RB`J>As&u{ z!=0Y-85t2@DGp<|^BDu-3k~z>zS!F+H>ChO@fVOWKFpNNgRRI^2%P7iVu4Vk=jXPf zqQ3yD)M{~eqWuEAfl}QD&OJ~7R=>yU?MzIEVrkwrx}4~@dR~L`j;B}71#cdhRkf?H zT-0Gt5v4kt<>Z^!0QiH5w%I^D2r;+Ke-FE}1l<52bVgm?6B__$sat6r7?jK_l3k4W zl0XSuA$l_Hzyx~zhXB}+;tways8-f>dzun#6dtB@)m+A81AepqAK8y&Q^c{+&3n2JuS2!Y=j2SWc5zYb~sS+gt z1Q5&#XSO=ZM59q9SgT$WD_m|Fbt_pU-#W0yxx9-e5p<0GDF#3 z28Rgl3AI{|Q!*hsih(N!SNG%6oI@~IuR<$@tDzyJb4zD5()5#q0Qjk_V3RJ+k2we~P0NU+`eCt1Wt_@)6E z?tp3JbUuQNCCDx?X@j;o+IHM}CI0ov$6%M1@z19tWDObGy8|g(5-9w#F{jIAuviF? zs&)}y;H$K!+7TldYY+i(j&A#hC;%=BL2_#K#vea^)FxwfE(Ko{k~YCXl1@Isb{mEu zT_e7DVITcc?|od!(02w63D;1>q#b*{+fVpn+=wx9BTbmP zqOXzTC3lop;}|hWOhC91+^tzF5d|5ZaXQ`=5XB=liY|wIQ_c%|y;;5ORX-2dttQWi z!gtY%yC;K>T&g35>ghTS$VON}n1(?vTp4h|Wj{@-ZlFdD!=_;h4~(Kx%bo8 z(iK?Ls_yoP4U*~>_YwL#GQLW+AZyZ){Tf5F&D#sC>5yBb1Q0gXt}8o)=7UClm3t4H zB?2a#!fSf^zCcKh*El!tXI6fods~+^1gW0xZ%;l=u|tq?)SQS6HT`$Itaw7hE|%_u z@CRsN6DFXZ1bTWv@ey#6^#0IfAmsfD<7c(|3QerfYJ1d6P6AxE$NkCpEFDZewUPz!lIE?R1b8cucxQS z_-ZIzYoj7S9}U#gO~Kv>VvT~g3BcZeZL{PG!aUBhL6E(f(ll|Di4w5vdcHjXn0(EM zv0w(SX#hRE;|GuB*fZP!;vMuRh!LyX;-|V=o8^T&0gI9JCv!daPMXXZOVYqsE}7kz zY|!=Fo$c=2Q|jKm3WQd?>GhjkL7Rg6PQTeT!NpN<09|f$A3l8eUS7_|-*tE@SOooF z)V+07m2dkmD4hz@&89)5k&aD=ASo%50!oJnlA8uWDFu{H5d|frK@bs?5|r+iP64U8 zw(sx!&ipfH&6=5Y)|z*H7t0qmn`b}IecxAn;tFp`B%Kn|TDh){lHk2OOc<`Ko>Gvp z(7hEBa;fMPdlaaUy64by1Fk6^rS(SELtTW=J7GW%_L08_3W4q9)OD%X1^9)Lo{o!< zZLqg#E3PJj%xb>8Sa-dNx&3BumG*dzFo=AuAmSJA>uw? zu8AjF_v1*p*73pu1yR=<7&vInNlu5!J!8!}^xrV-{8dKpCF}>J2 z=pJic;^KsoceIWrL*S|iINp6U;Su$)y$tzYESQaN-O~mTHrNdCP)totJIQ4J0CNHz zs3r%m&dF2)H#2vPyA?khNCw%dpV`m;UaNMyDzh|YiH+mfMOtK zjF(wR^U4xpyM%?2UBC}P5{JnBom zZt4k*^|=LzJMYRBsiFya_NgM<^kTs;Tw4C1OXl9DuCqciHjhc&Ie~fux z!X)rR|GPSH#}2v%T7-}Ry9&1oT>(MYtS%h&kp$F9tq~GlIGA_+(rTUyPpTS9OmMR_ zi-?PtQ*vMglkTVAOWPWN&hIQ6Ytbl$W`Wm-cWcf((39^5$x{MlSGR~9MM!9FMeI|| z+=TABq0?t@nz&5X|1y-H4!ISGZarpkb`%&UR07sI4lo2CkV!Talty%vQ4{Fq%$N6o zaI%#>1d3i&Xc$f!i$1Z_2fA0bLPfDal{Tw&pY6?*{!QsvSM4+`RsIU(CnTxO9`mY~ zc|9;%-JF4Q=fkwQx%on%O@0RBg@EO)7?Jx;`wAe9bm>QBAI*t8B7wVHa7vHM_L*-w!kcW5r;;EX4^j+Ploinm&>$+Bi zNSj`%!qha2pJ_Kp3x*}X$qNZ3)>)U|ZQAGbL0c|lFyn-Z2Ev2VLu3d3N!Q|o@1gf) zMP+4|Ae3X!J%}}>sl{G z??cGxwJh=CEGQ@l_%rGNlH|Wo8qLqm+1dU7$>rgHmaYFU{?$ujYWmgI##2$LGBorc zpZP3fD|>r;eIX_P>$C5a1T5#~>*LwSV{1{;K7RzvThC(W8vea%IU6^=k=bv5X4jH@ z@o6~e1>@+lSDONa&_}^YRYjY! z|GZ^}xQ$K4q^y2kUd1z}Ux$ll#@pLk7S&QL(kIze7Xd!rXR3ECgUNaSV}7xm5*KGX z{-{81+sZ1R^Df8t+UNXhKKHsNZ#((Q?M<{c`49Fv2n`H8ePc|KW&8E~mR9$}p;*5c zGDb%Mt}R|}$@WuGR~%jibLqt7(08B8tU5m@cTtiQh%FC+3=ggxF2~ku%J0zq*DBKi zy&p*c=St?I&&9!1vYYQdKa=s9d1)kc8F&Oh9&3w$W+%LfoF0l$sjy7nexBrf0%j5Ad_Y-pOz7@Kh=C<=`XvV zGQ^#LH$?Zv)EIFnPvdLB)skZPpQmyh)!v?8{lLH4d4ryL>%llnXgGFyI#(!#$fB&j z@X%ZB*qE-5QSVpMG8m-|hTPPzT<-D7+JN#xhGWrD%lT=s@mSep$qyj={k{m72}+(C z=Xz%nGpTf!I6kgB=E?U$tDznovIL3i*Zt>(jcYZC{@*w(E?S`a{ogm(^hWgR+AGHT#Lg(kjAM(70%5Ju$6{#=8wep?N;0XKscpD7~ zJ>445o^9ZgmFtuEL$=FvEIqv`^i|s?h*=`Rv^vtE;YyGh;{?5`{vVBqqP1lU~e0NL2^z<&ubvjR@nCILYM_bG;Kl^(bessWnyJO=V z4>QJzi`91ZM+|f1u+vP$jatojV_DU!zJD5>EOm8@Bp(efr8J*mUAyhEJP~_)!fE*6 z)X@`LXi^1L<9hRVuqAS%o}#PNHH`d~dHwXv#d3BG|E=SLnxoYlvdck~XXkjcmtXlB zMAB8uK1ua%sId%r<^R$myuD*7jy^c;@6a3nyk<(?hQk(lgy42|MD!siLz!+@;Mw2c zi1*EBn6DX`M@0vAG?Ljgf$Fu>6I|TG)DnlG!+>PgPZ~RDEW2LUuu<(FWJ0EMR3zv$ z%=v@V_{E4tMHT%QPVLGixhKKV!M1lyU`IPj(nW`zTKPDCF=bPY6mwmT$M-v^e*fl% z%r&DwwZEw@!u(V>2L)#8Z~3$4dtS%UU%gfn!>rmUCbqun6d>qZc#X;KpraO&pnIkq zcd8^_?6&MaNK@)c<&STuC>Y=3y=8Lp_1@0A@xS`M`a|7LriO!&mo+#S&(CNDh%Xn( zTXn2f%kbWQ7;LTX?yC%73^)Pp5-zM#$;H2XOkQrmrGz3?-fVWb_|+2_*+_HxeK-7e zX>wZq?o|V;1%+z5EVrgW-#se(3LfTLv4|meL4N+PW|xk4OKI9hzsZx;xaqVuwVb(1 zSpO|44#RK9RwHEDFF|9d>TGm5^z{tA8&h6$+GZuK-Xd?kZAL;r`45J=&}(VC z8U@e2?gY3#Uu9YtEShAx*=2W+aqE#K^1GNE&c;*sM>hMK(x12XsG5UM)0O=`x@tox8NNP0F)n`T zwuZGtr`nbdadG(P$f&exI`lzUYe=bXHm1jlpUZ_n%oh-b=LuJydz%$}<|ANbL*4J6 zv2wbcEwB2NQ1(OIg-X@)xXN`dG9PAo_naU$H>Ztz)92p&vyCKnKhwxP1_#$J#M?PJ zm!<-@&S5j%p-6KY-KyI@tgX31EwK3U!NXAC&|=Vu@BP{9d%wK5FM`Kw{P0=qIhR!V zo%E0&S@aCfJa_{E`yLXiX+nF-fskp8RqbsIihugH@I+GY;hO(Jy%Y2O?uuu6t~Z?g zF9*qud|6EnE4gZz;(oBozNY?*j;Z$q$;w*s{kAtWjBF_Ii-*hB7B@1N5-NeBvv`9q za6)!qU97R;u-WXPOkltOQ8{wAUEk;zT^9Tz1f+W9`{D~x)x**HmxFm~12_&h#toni zLpn!hgg7~LKwK+p@o+OVyk$+H9(BwA3iJ6ep`X#vhp*W(>YP3im)3^3Jl#j4`$L&N zDcd)mFXfUI`Z(OH(??|VGPJn0TymOnF*4)~qER?cY1w|Ra^*SSmYkDM3TTLbenLc@ zNQ=llTdKX73NAbN`6-O@n*Upq^cm9qm@fj)2tOlM>Wi~4zcKne=zJe(HQ}BtzY{n$ zEcDCiar2Gy1j&2zMeK81jpepP7jbWj9UkxKH!KIxdT|ok?Ux>}>-HF(DaXan5_u;E#95)aGzPAm1V-{buCr>HzYs=Qwjf*l)yuv~^7X9YqKX=AJM#dnS z^!;~sPxO*mr{JxfJ)LshF2+nM#>~OI=K>EkZpvFH&$5Nmi|$U<|N3=S-rj!a&Gw?P z5VKYs&B@=~ui3=c?tXLW=mh?{(Nv_@Kpc!$PyQ5q4p!`yrg59UTZh@ z6^q9BB<&pWA2QsK+E=_xh`vu0oqi!S+7ymI+x@=}C2wKnzJ{Y(O6HBbAa~{3!8!Kk zgBvmoK@hqStY15p76@V{#J@6@jnA~p4RmNXxz+V#Ck(%Nx1~WDE$MsPPduu-)JQ(@ zLM{389v>fFh~MLR&Z*~be1DcK|2liz;{SJjZREaAj6Rnh;&n6}^%4vhL9;Q{i7w+h zY25t$#Sa@~Y2Im82ds|QOa~cESFa_l{yd};7F}s4`_|3<6JH_S^dd(=;l`Tp=15}O z^h0{zserYq&ry|xZ`#UR81D}T>3q4y1Y65zbM?`o?1}&WSk$JJ5>BWF&{4Seif?^X zEV1cB^)bkJ;driW#~}M|J8C-Iq#HyhZ0&__kU2MvTL_;|re2R1h{N>zeNPQX@e)Ve zi|ZRpe_6gkHqQ-3!A|j&zZYAVOY7>sP6w=QYXDbE>(*3j|N45HT!DtW<;Qh(h#hc~ zDrd{td21xSY`H&cPy|$z z8`qr#Cbz zfm$c&oS9C{Z(A40lmK@O$Xc{6HcpUtlvM6AFew)NIlM4_9e5FjpE3Y6hXw8IW{gGN zx{Hd_=qJKCw74=s2?XZ@%=6MdJN}+C7+x~6XC|4C@9M}h?*CYDmpglK7gf&3xc9&j z9-;ijZh7h3Q?|sU5lZ!gmrRUO8w4^5&p-)NEuxRWFQASLmsn4-XyS~S+T9blSVuwr z!*(id=NPTz+9&sD+meu-l^$mxH#z(41;Y#HpY07QPSbnG0~Y8t&n-n5o;GrZ8;m(r z^*a$rAJJ^gI+8p3Pl~e%Ja=Ba@91f4u-tOk)TM#4>!s-42VPRnP7cjF*?sk#&&g>z zvVjS6E@O8_q7z>D?58h>M#n{6IlEYRdGoN#pY1Dr^dm8yO4&2#gf;4<03q>uctBiP zGA8*iikr6kT1bCZH(Xp6KZcyLxqK$GDlCiSjmfC(uQ%)`t!hj}-ns~jh7)V^Qs>CD zv)$3>!$^nc{d#_~32R-zj)PZybH;b>xEgw2 z2@aOvndlBB(no|?IRy($gaAeT;;==Sg_)-mn@@eMJ3N`&}%g~y64e^VN z%!`D%mkE{<6nJ-r%a9%MBy7>af)GJ!5*UH(u#^`nnFB~+%swuQ%hh286xJT(2R zqbjn3IA0rClzUQ6npf?*Jg(M>(A?xRJK?{$(b_Ea^W#~}Bb2_uNnj!LPN?1mNSsfF zEAi7s2%a4!b1|K~Ga9V)5^pK$X@cYTaD0p4)WsNq7bfghf6uK%*cH-spO#j)**0>L z4|x+**W;8v{&>kKwOM!Yr>Uz;4S0AdursWzou|(+v|stvG!0P~>g6WkT}VxcdK{`Z zovQ|qHK{mdzin|paJ z!6-|^N9-(*_LfrIn&a7gd$P#SzVBiSY*oQ2{-Se~IYC4HueaGx@#!zFlP^NIt;U*t zEGhzVvbcXQz4+5p^TD76vXNR^#4l&AEPJhWC(t-BV`QPFfJQ8Ta;bW&SZspt7>qyH9mj*06p$ zj_Rlt1&}FNx3!&jk>}R>KU#gXH(wM)cRaoo)2Z&fvYe`*@>cajg;wt0J|#w7{-|7Te`}FSLH;WF`NN6F_sT4&_|I9Gy!h@g zxHU~({*G`JKYbp@P=NO!2Y^!+;Sc;mz&e8hYg67V(VEOI`l&k3PTiV{fzlKa3@#lJ&=67mMx#|8M@1_d1yu z{WtaSd&@92sL@xzjcr!*uOBxuSq(*WH<{d?W-`-o*XVed5Ace3f_Ml zXgT=?s0^8vymAVp*#C?FW>-3ZG6aI0Bf#{7k6_M6N8Icdm}It+XF0ekh}d@nh3PBE zE(*NR@6x$60)2Qe5*vViJ0NtF+`2g+)~-wTn2n%*3ak>KKm7jg11uXfigU35JZLZ^ zG;IyVnFh8;i3R#vDnxuXCq@C@W>ty(`POs~AZSpmE;(ceih;&R&bUM$M{Ku~?OB1q zGO*Y(L37mld9>023OclRz(6AcKNMkS&LH7#HS*pXj7UMkT{c+bax`Rd5*v;(&C>;%opUnqso z)C}lpD?=w3LI7%g44~%W`luH`wD2%2YFsn3Lk76C(trdL38(}uaAg=Iw!xZ6;Ul;P zagLOHfv#^D%BODE}{9sHpyV|M1IlnO#Zyn7;IX# z6IoT{z?jh({UO}y7GD6ALJ1HsVCZmPxyAuNdkj1Z{vIew%o}}o07;s7y!H;Tdbpu< zi&WQ55=c7dlipdjWMKm8`rCk@G7xe;j+)6t3-(8NP0?e;Bkqu8z`qoQu7gCVC z>$gaG@(3h*3E_YvTX_Fa0c*w)z@-2vg_Y-F11-4&>+8WB0evaOE)K%q0Up%OUqHwhv3LNlrYHm#2@q|LzlejI zW>|KImI+{GPcHgAfczwQcg?;{ng&G657@VBLx(ro`8ZZUK980g62 zD6N3}wGXM8!o36Qqh1tn;3g2s-}7#_DbWYPEa@B|)*St?KleNafD3Q~ylh>i z+Gx~4UK(l#yd{oOAft?a3@l1Sfl(gg)Gmn7F&fyN~+;-RG~J357+o za~m+;U~tdWY8kzaK2*ZvW}XN%zXmKgp6#9sQ~4!};ELeu&Zh`+V&J>joMi6EVKKqQ!bVsDo%>&)&6m8xLEpM< z)W&glrt~3j)@(T6pCHQ`Hl-eAw9|CL7N#u#?wLCn5i>ttVW}Z0fTK${^@w-@B%qR@ zZ#X$vqh_a0`?J@G+M*4 z1TU4pU)^we0lOaUbjy=iC72-V{|B3hALw}ZVV_Ru*`WV2=@oL)YQSAbVPhj% zvcRT;yB%y}j=wLXA$xE@UFpvadTfgTwAtmNA^VfNi=J4>(J?t%r4APS_Ly)Ufd zK25+H0T-Cs`KQuopEuyObOw=#TS5D6iz^wR*KEI~^lfZUdN}ruKZ%2{%S>-&K_Y5@ zeKfg`9OQjBoTw(|VQ+*^{d{MB2z$*)=|9IFE~YGTE(kZa6JUS0;von?>UFj(e<2C9 zf*F%ip@h~jJb4hz7sB!ZRWn2zAmjLf-a)|pW4F#R*p9*?mw^{^Ww*)9q#c~ByYmc_3z)`3SoO)mZ&In}Ny6W7(g(ExxID9UkCxp)tSqMHKa!Y>6utk+_ zg4(CreOA#{SO$HLeR{ynn1sv)_}vs7OS8kUPrsW3*O6dXP4Plf*m?Zp2HGnF&V8`* zAnhw7htUSG64j$#VI^Ke|5Je!>)*E&h?UX*2b+a_2OLsUuibtC3qe2PKY{Kt@C{EO zki)hUFH!vWx_`g=pEvm5^EEvDz1ws%5GdiiMzdvzaAqC9#!~W%!TAHE)Pq3S#7g+5 z+u+|p&1Z&@hZ7vcws=<<5){Mp znP8bs1C8CDgkTg{4HT>}*=ifbe7DN+2e91>30WF2Hc>lnBkvX3&EXEtply92bWT?~ zJTqjmQ+f+J_Y;svwHUI%FD6)z%ZuVtRB*vrp-KJk7E2tN0Q&+I?}d0Q{x%Sly)lGH z;Vz=T{CfoO<=;2ZPofQmKTqX<{OG@5``E2r>m;i89ilI}gmPKVzD;uiup>943P2L&G7ujQ)RY44s0vd;PX>R-)=^Yw7NgvMki(H zRP75fHt%0iTGEh|0n*!Gb$s*XKS|EQDe%+bsV(R2K}yjX3_H4v;0Q_g?F_C8jyBEm z36ObG@oXT4ZRB*gU;j7=Xog{0(n3SM%94w-A~YD|9YG&kt6P)U=%^YH|Wb%7^7 z6=z_xbs7aR7&x$Az{`B*iYqKm8{V5TWd;vze+E*$?^3v-=X&RGTAxv&b4HN-sa?Nz z^MI#RTO=tJNkt)+fdCHrLv}6!TvmqX#Mp+fb#sB~B@}FS z_?bi=(!X%+=XN+)t(J>m9}V=%lA?br2bAC?`h{>>g3zr^fL>g#e zVdYB{oN|R=M5WxJWsSfd1-m4?eUj|HPcL+Ry+uf!BasBH(UM)4G|+D9Rz%>gr9C8+ z(2_LV81O+H#Nu%#V1QxdL|mjW{G_+|$G)RI@E-~8Et1?7PDSM~&|#9X{?`?w<0F^~ zOU;Ip!xp2{_UhtD1+)2uF{g6v7!>l7yF`11#Ci6mRx9%F&x*i+q!0YuUR}d#{m$Y5 zS*z+(hV%|(81|tTfl3EtRhyJbWj6gv_qVMM>a~A zALpK(9NtE2&AcF~uv(ggxep)4v0Z>!Ouh2>Q|BM<4C9qXIVTS8Zp1ixM0VRc`w8SAe84q?Q3&};VtG^Gw2YSXEq5k< zc0l;D+*t&UZ|@UG!MnlZ8*XdoS8wyHUJ<`EUl zrP)56EzRo+*>0zJ2ogZZ=v*Oh$cXSa=4Pn|(l-tv1wttnOfuTgM((0JTt44Yc)UQ` zfwKi!4td@I_;^b7x(}H;fM8$-azyr81A_;N0+!9d$prCiw>`l&A8_}=DUYfp=Ho=J zM**Sw14QxHA0RRYd~X2Oq!esY9jV{!3=ehx0y|ILYdP~#hcyCXc&_BpHaOY+WkGyH zT1CR_fg+Dt8Z;&FIjqS;@Q+Dpv1$#4GHWl^r9kMz#o1QyJTFa%!6NvwCfGsw7y2^q z^{4Ti1u$Vq9cvZu{+y$4hL>qT#fO$4!0>|xi^Bk45XL}$q#6pJkm+G=H*7`eH{oF& z$bJ)WcfbzoxQx5L5SGe`G!Qt)CC09X&vpvoyyJXlxqS^%sF(!Cl&9kt@Z0L!D#Fbd z=l(0S4-RSZEkcI*SVuf@8#`YFt-phUc{hAn@8SYP5(x;7H=p*PBpVezmZAOG_HEHZ zDoB^~=gX{{UI_!it}5mZTzKK(vaBK(bpDX35J~NWC;QvbWkV{yM1tRX!mP(e`+>rt z2g-Dq8osU(HnAv~6jxOhBhhSvZZ9&VfAssHtDPZa>d1S1mnFHX$|G^--oZeGIg1b2 z6?2xW-^qV|sj*XE;$%O#{6>kcmYmOZcl-@AQ+EbCrGlPJOr)n)C>X?k{go47Y*3~7 zapzqoD>uI0JIATQux+DHNbZ--N_**vCN}!@7?{@zA7I>XJ@<=t4Ww3{^ux~b zrG8Z$I^<*5%X{B)hP;9RtK@Ez6z&knDIo0QlpW$tC>sIofz`8ns^LX$?(39pn6fxTUEnwiN z%Jn&TQVpoV8bd(CH{`tmM0}`YGXr+JE{^+xp{7(bX#s1u96a$!?__!nJwgJwR}|1wKgCxeqS5=Md6M@v|br^g8d;xv$+qvWJ>0bmSa;B7eg6XJW2e zh_sEPDhi=^scqbZn*}2Yn;6>#@rnS+newHDJxrw5Zh;C8!y5%=_1+~exJ6WU_x{)~ zj?vMw5v@kc;CHmRe#MtGU^O+xG{X}j<;h)CaYBSgDdg&7=cu0Ee1afLXUpP=ja14O zt#J$Q$Z^ze7crQnoT?_!Kw?f2a%SImdK0<56t5>&XmI1_Exz|s^A-!~1J|+pj$ed2 zV?3!j8eS@LRLj1lD|){(-XtX4uD9kRdDjR|u-}(6r_&6vIZCt9RE8a2b&>R49!C3H zeAgZ9KE3WtpYvugA#Qg-#C`p}4#vmm>aAG&)lzfeXEFmw|DrY^@LR}XwUUrd;%haN zw2y$o#~@$za;g$68``T%1b52YuV^T-dy0@jqs8}4sUSYmJCeS(GmGqfFuo4#YTBe% zR^F-^l-*IfC#$99X4%@&DBT@Ua&+MEyt19OhQ?jS1sDT7>b^}w6RkZtlGAmA$}A%N zh-XJo6#Gjdg+UzZ^swckkWi2i8|JO>PU!V;V|_r*b8}q4*$jU9yO7p2+#G}~Y~@;9 zAJUg72vZ|;7w`JsscrOiyZcGKw>K}cZM?bErVwZPtEF@WiBo|#`SN4VX}sS#bj(^I0axRCW_r-Z2={jpzeF2TRy znV!#Di6OocqCS&($88Zyl1)VaSRraPx!boyi1b_N$?>(k6~p19aOY0skroYjs808Y zPp?by*xDUo>AmX1z-pC<5ojIN!KH>E#u2I%*`o{fwl5?uG`v1gT&aQS0gz-Jeuu^|m6DBmwPq*c#VjF_=%eNdtO z`)H66M0R0T*f?M85Pt-eL+P5p;Fd=yfqVNtG&MYVmG4_4#LREi=}L%)6iVj!f%Zkz zSk?QJ=Vr@AP_QGG8cfP@f)dz;#TNp)-2fPAOefm24)h>>ajHof0>fIn-|Hx9q@YFy zBAp)L(Z6{W>%5alea6xvyKJ3DjP3NHB#4VeU+Assd3Z>|^u*HT-=aKd<8X8>tsX5fKNtJ+ znW*8@;CHBAx<@Rx z9qa)N@8oVwev1%CSMseOx;*wiE}g6lAKcK-=uVg46TN^_I$4;Ebe1`_OczrxXj6!gYG=(JVi z*%EXH#$vr**9RM2paK9#Z5pu;u75iXd2@#^ao<5cB z!ahrL`a++~X?5f2IQuswY1o^(w0nl~_0k+m-7GTw_A@bb%2L7}Y@_fX9xLqA_-f&X z-cY5MeoMu_94AQnpjBzyas%sNtahC4JJow%^@gKLju>ImdJ`)%L&^F;{9o4Ue-ie8 zwvCz~NlwrD8W)X8BSw};TZ$aBsP>({(Fpdsh|yG*KbT%WyG>K$#a16}TdMLAH>z&; zh0FG*0+xqq%*?)ubos4m>4hGW4cf{I;tE_she1T68K_IRKXHhtstO4^yrG0|1RtyQ z(hyW5#4c=VOwFzf705V=sQG0Aj@+LZe3TSXH)!@i1YE z#gkshMtPrC5j{(;N1CIagnjij!5Q`-bhIGCgX zf}YUiwob;+jeO4l)^)TFmp`(8K-!9+62Z^?X)d+RU$hY?O%(iqpY64qaE)>SMGX{_8~QVHm~M7^Q|kCIue8uDZm zX_EdT7TO@N2mJc3%JUaQph5t2xOxCuB8p(d-h@}*Vpno8=vIG}({Zq!GjxK0J~6%4 zxljVvW=-?AEdqpE@qVZ8pylI^+a{ikwl4*NQ0v;9A)ggusR0Di!c0k&9z70-+AGc9 zpo0eO=Ls1>8z=(V8Huq!zXqc=6>1CRm;j*cdW8`4IrxSYo`oy`y=9lflHFB%ec99VW!cTE-~(4NVmtz14L#@4g#MV&rx zEcx7n8ri*kG2o{cy+Q90AHnpV@Z5%_42upc=B^VPE-Nls6pCh$+p<{(x36>T*?SDE z-0u{iR)g}$L)a25T@{XFdCQ{tU!K757cO-%027Y#+8m-yp}L9-HYKOPY{hR^i27no z0m@X!BgCQvA)^L&rz3@-Z2-{h!cof4R-y8(qGkmpCEX)buf|93-vi+^(RCiL4WUj1 zSJeb*#GW2(VCRSC`W(SfRm%Q%4C#5J9^5H=9dqv)y zRY01MO;_-sDP48oQPaVgj7q!WN7j@#7|<;HRgYXbG%%#BY2Yg|p6BsB7v_L zHRSQN-Y<38ha<%e%LZYCfVBvqngBHO~()M5KWu?z{m+7cp*lEEr#($oYzfBz0VJ> zcAt{7!MS0ascgJz{MO{^BySowR&Gwt_q0nXE78dtSx7lDmysEeLxBWBjLOx@@P=RPthE z)(m@%wqmUF$RhXid-76BUctrcW7b3GgSYyODt#F*XTZ^t}(QZ z5G>Kl!`v*mia4MMd;BJD;VEi{+)=^P$E#O4pw<*Xyg=MgwYe{VYVg;#L;E4RQAQyO za;r2>Urq?co@YYMSP-hJ2=bDTmU%xAV!cs7-5 zo4O=FNE<5B*8H8vkt_Q;x;fNV8}NjvG6@1GF1JP-7)bNXyD3)@jcgeIx$}JE1%W3~ z+X^&2*Uv@eU5mm{2~*jpqKT&u$wLz_N&`--OD-|}CB_!AGH?{;rc+^Q4A%UKnavxz zdZMOqY$bAuU{d@Q#d zh+>S4=a~ry8$D04=9@Ee=Lmj?4b4?b$L34tyRh#R8#xP9RJp_EPR)O3 zq@^?T9N;43L=D(%el?Iodu*pZwU{qs2(x)tOA^<{YVDyXmhfW4R51o)63 zURCOYeXSTcVe<(Ba6=bJWcL#$XU4#8^Fiz<*tghFQRT;K;AbK95tFQeAOvY7z>o9tx%{bD5AT zlRo2D3dnV|VUTRNX0EL2+3-Ha_rX|x-`Kn2#xGvd_n&iP&GJ#}Zx}}K@ne&#iTM;E zq+PWhJNK0t9vt&&R?^=fSlQ&H8h3FLAJjK(IB&bvETLUL?s#9oP~j^6i~iXOVqsD_ zgS-v8>L)GT8F}Lddij(3^n=F8NAZ-pYLy%niXjs8PCc4>PfJpbeXq(#3I<*E(T=ys zd@vZ|b8u%rB|dZEp?Hu)LY>9MoOl^}5tt#d8GsW4O|VV*b}S_Q1e2k7>BYcJ0UX(@ z%%K|xThlE7k3(!`Q*MT0#FtM#{#^_SjbI4=L*P!3%>(rR+)7z#%OA)SpwykkWAe3v znO-|?lAju2uHGWr13wUYIp7Df1Db7+aq`p#qm`T#nvw0zpjubq!Oj^bW|ljhzBoS& zBaxG^>m+^ka3QRr2y2$b0z|lRWhdYoYXd`x&dbCGbORT1`(bL48J}<4wtV4}91Sef zP)RTlVENV2rjWauHg5G=pgSCQS0LrHADX*?cHiFxSvUpWz4&`F=H6V`75zN!t43HWsrIQ; ztNP3HM3K4C!pz3DwT1PTCp792&gv_cWh!*%99g^d(xVTI zom5L>uIVPWTniUlkL2C&Fz3WNRzA6bBY9&tb>HWgAL+SsBWV0u-zK+;I;*55{dl*Q zWOpm~h2->D&+C{hj2hfnx=Xh|;$EKMc}x^5D8xOf#qsw=X}$#$eRA^nd!_!Q1out__mOqC6jnZX)IGPwZ|869<`zUNd$hm1v!eTuqbsrM=E&4wKn%CyA;1d*}YgSE@V}I&T~+$)J$-u>t@EzuAtS# zu|H@_o65p=|1F_Mb5HEe?Sb4VtuA&wFHQdg5^@y+8$?@cD?q8x_1K4}rb(&BiY3`; z#ZJHHT?{xWO>}~4zdrB^V49T~IN3b_$mYT%5`jxbd=o(4CV&N?h#8>@&3|=|AQS=P ztaw0fjkUuK)hN_4;D@mB);#nomgK+(z%x&(Z=C#y?;fk=1*jM|N%jX=^rB5Z;GGC8 zp0e91;o5w?)(UC&!KEYT7x;H*Lg0I#D8YgDTeK!Y77UP- z{~L~kv6j#bfA~hRGs^@##s2^ngvLBSz=cK-QFk1F(Nd5w`vCh9q|q)DwR&D5XwVuP zpeTz9I}I>T0_N93Y{Ys946TL{ZCI5sG9!;baO<|Glphs!Kzdtq*THEAG4 zRDxf^O4LQzM>H9Ug_aDgaC(BSd*&h-HbVTJry5O-AHt16L*xHn{intZd%qHN@^I(h zl7;6p!qxPb&k+KvoaRxjEhJfpLYC+BPYalE^7H|%+ z`jCH56uzLJ?*HNu|IZ)&kJtY9)Bis$htW6R;8AL8YK{RD0Q_gd)@p&#fVM7fhJ+_( zNfH27aAMtrMb=g83()M^WW(X%bZ|D~SZKdIG{rAYm*t`DK#mm?1_eaF&g3)k(PwKo zEsO0+#INk&W88%p42-iY%XBP!jQ!e z<15F@V=z^F4sbr4k3#dFK%)=@8$Gmx5W{f1CbRDh3Nis9`%v*I-+ZVs0W6kc9hs>H zN>#VeSM&@EHQ)!pEQvO|ZJP{DAnHVmrmL5tDN3(-dV6BD*KBL#A~*wJ*h*^u44^H? z&}=sWP;K`YV&lrr>~dfTKOa4i8GJOQuK$V|UY9V|bLUAe^NIGU>>raO zRbWCsL*OdbDi;EIL-}VK!8xE2L(oPNWCy^WIGp?9zA&0Wr%xx9Qg{sB%$3n7*h7-r zp4gjiNeXsqrN9a3s*hnz!6kn}W=mr{cKjLeTbOg=G_5I;i6SNHkIZO9;CdPgCqP%^ zDlfsbP$1)Hq@o|V?s$UvCM7ZU$4i+s!glIr4EFHgH739>1;`YV#n@`0axmcXM_nsh zG>RsWPSVqE|B=NPWJL-pATbX3CNlPK*?CvnYQV$mHdHIG;uZWiEhPx_8L?YnQ^~C^ z_F2?QUVfuRqG>BP!dHIe#Gn2h>4;zd>8cPpcQh&*RS<>}bspKyd%{2G)=jw#%T-Aa zXl!JvYeKa6*q1IsmgDxuSBMw4k87(P>1wfAByQs>UV*&fdoouaU(JM~QAx$i6dHWY zZN7TN^%R=Un*vP>b96p5(`T+@RWZc*rBl!hO^_Nf1WwM)6z;2&0ZC+7xw>sbz$i)J zJKy?Y&<@L&wh*^3=1QE`*(d72#K$gB9rQksyRUBbS0e^GErsRakJd~_LV$$^-3eOZ z+*D~yQ}V|YI>&2EfVUfRk2#9S!_XQo5dz@%EjT-mRYLu&Wk-ikT1{>j!levY9~K5w zGH)1)q4kh^r;1`)b3QCI$IgQW-{y$!md03`raIo>Qz|>m_P&go&`IInA*1>kuCEkK z&eMgL?KhZuuJP4>ODSpD|NV_x;0l(|lW#fJU(17x76L|u6JUv>nToKpe?-vu;P3nrA=%sHMAN?(+ln)>5{<#82GLYyWe=Ph6TJ7bNLJJfTtOTTIhL2 zy&FOf<1NW;l)B5B|E>bxL{N}`0`mT6V8=W6(=G#c7m!zrwyXyv@GyJOB`>oXM5EfX zG~#e34*+!0?v13mT)-B1-F9RQ4v3r?9J`yO9J{2Nn0ezsIOhNk(( z31(;W?QDS?R?NfT*NJhQh*y#!Qn#VUEi%W<9O61bXTTG%{tiP;o7{261+4^U0 z-BS2QlE}knPuD(RA7%_GqlVhkusO_Bx{b&;!mx=5RRrZ`NU!sn4N|!BIlWL)pz)c} zJ0A9YFO@_lndv8Jr5wN1j+yZMfWm=t<=ZM|ry!qP)L$FxDX zKG*=OC1Dx*!(qoM*G+eU!n6Hc>D*4mxp33{J$T@f2cwC913*SZb_vRGV-#XL2?5;Y z0nEuO1b$bZzk!271YK<65Unk?I$Kn9uOj3!JD9r*kd^N z>XlqhV5%;znmwKoA;rhVzC@x_OH$28}A;vXfB3^8y#RjLF=pa&}n*Hh31q1R6>) zz9Mo433d1td8!h~7Q{aC-n;AiFPU*w`Hl$RsPf^b+NspBzxN{;$D5171-woO!}kex zN(7A@O#frjQh^{)sX%C^cn#W2w!0Fziu8#hqj^H#<^;$F82G)%r07UWu4bTy*vXZJ zYqC}QkrdX}iDvZK>^cwHjThWVFeLY6(n=L1*=Ho((wVfF#)HDb@jH^OT*H@tzNGHC z-t%Hzgi21 z@Zx#;&TtnZxI3hRlnsG9#DPI6`em%ap5=9Py)A*>s!HU7b6p}ac{Q$QeIEahvs-dU zROx=-ont!kSQp4A^A~aP#eO)%->TUy)^@z!Qj-)Hh5y^1%aK^%Wp}OeE($}6GFR$N zpZO1-Z}~|mB#C1vn7oQX_H`ce@*}6=a{6Sw`96FRnIt~pIQVIx2cZZZH9=fGO zS{hVT8U&RtMbe;@lm23}j*{r3 z4|C9^Wamn{j`uXGFftH6@5mSp8;yrTdcs2|f=J2=XQ%OD_pg^21&FKKa`k2s+8`R3 zX+2l-PhGJ$@CcAMUf|t#0PF)8K1K0oHw9Z+3&u9ogLufO(^LxNN~}9<`Nlk1HIjfj z$Rf5hf*!piVsZSQt)<@UiB>BEq400^YzLJ08rBHH5s45vP<`mLqRhC@%jJ)4gXT+U z@~6lWdf0b0yUzjXwX1XMoR4P8Clt$H+?&l=ag-KOz`HD7m})vpwOm|OmAFLIi?<|g zkK{|(QI$ImO<D>6~D9I0S!Z`&Ib?b^7xVS=Xj5P}$$ykIt3@Cau)cu1PhISPXnnYb1a3&@YxF$vlUf=P3$&by4F#}ZB+)}z8KgLR z`4{(TSI38RS?JUA7W&eJ*Vq_KIGuid|6o`np@w z@*`;aAkp%T9|WPg!dUoz2*94+OKW2MMA|1!VA3CCwS@_VsQLy4^UUF358BfssQF*P z0Z)H8mI$XdP3+Z(6e&~ak|@B=!aa$~)=O5i{3D-E55EQ0;&*$J90L)5UpF~FY~8(34okZvxy@pKEBCnbze!ErHw)gd=;KDqQ1w3+GAeV(LeOmBd}ac zYVb?0Dj^I~1#R$*ohL5|{uI!<#LWnLw-5O$GSJr--L z?d30>mj>n-1nzR}+Xy6;XlJQs^u@Ac=mot9eE+<)ePAWAFPqPi(-S(OOhQ5BfFtZY zqkeTLt?#c4t{(XmlB`uv4yQA(0M5H*+UlIZ&9B>LW<{5Hsjt8-sLNP-lRuS{`} zbNkHq6payiq#glF4U^*yNA4n_c07=olxms}L=jhyKxd>O4k7LundqV!5a^^m0J}Y9 zMKiIHvA2JG?UAD~HVc~MkjsbDkgY`1e z2K}j|Nd%%zCcaEqC_;#}N)6vnDywvA~@Q~*`PPglLI>|$eSN&Af5PI@J$c^%hopAu58X|uit zs#cm!Q8T^(o~OVe1_~9=(8%FFsYX z^0z6A4Wl}MvT60)gg?=D?P(DvzTiZ_1_@0z=PY9J7x%|SUXhQ3 z8*_vNxH0O&eO<5y!ojkJ&d~@)L#84I-MfsKD=-#x@l~vX|3Pp5s4$dXF)aAU+mP?- z98JXgK*hvg;*ljuI|kElf2&Bu1ca3yOaP$_{{&V?f*{M$peM|RU}BJS4SqU3!Z2It z35B}87N4RcnK+FnUyJwFt65m9 zl*Zab>HLbU0@zMJ_JIvDFf$J4W0ia=+d!49Do>q-29~bLTd8o9XFUSEQM&kD8Ead$o=LuA zT%4pmJBGC^{E8luGa)LOsI3(lvV@S`TP2l-)##EZeUm*8ipYRumg>nHw}JJ{pu#5=E5jYX&4M3g3@Tc%;%-@?L|Fpmcq2 zf1gZ|$gtApz(@8RLyX02*FiG_q8IW3rAX4K6pZ&QXm%<=dnI4?0ah4Q3-o3EwQ3An zu+xyItEPTbzqB^=bHQ%R&uAI=*vfEBu2Xi(2S2EuX?lV1P}a{ zs@cGmKEs3>0*&b^*;L)yRazm^$GL(THE;seS<%Jn*F&tG-vQWYyN(@QzKS?)Z|q*1nN?z!Vm zpW#}S{>p}z+t-6!V@TC(k3nnH2-X6p?)Top?A>L*^~p+5K$No{C3Cz-bjM z&cLR}7s76we9#`*fjU1vkv0nO@eDb>Rw9))-uG>J%iFObSK|qKu^W*7jTjMUIfRi_ zbg?oOS@^mW@|L?)GQui)@2*QdhcvT^13|;t!B{fivo5iu^S@a4r@TY22dE|iYca1w z8K0Sg!Or{bCXb45nS|)hSQEp^8Bha65h4&x-#35Coj~zBKb(+$08q!cF0~Trt%x9< z#{ysAsIvK|eo6Wg)+;dPewHiMy$@dB01I)r*#JJ+ve(B1Jv4+P^q;U`*+0C5U?|u~ zLSw>$-V?0Qtf>KxiFs=a@Qw}J0zKz0QfYb}fz0HG3_l2jg@#5@$vEP{8?PteanyZU zP_=G@Y&nDVN~x+67FjZvBF-C#Utrhj{7NB9J-9d*e+;N8jGcafNx?BMdhw8{R=iBV z8%AmVF|r)5d&PGNq|V9yUTWdbg!}M=TlM6ntXvx%?=kKKp_dCVm@M+*^xHLA{NNyy zy)A(?(6gRwT2s<-xE(`l9Zb@RwbWQ3GLSKfPkuXAfHwen1*)cGCZ#5}acu?zM`ZJBXG;Yq)tDv@ zDr#p^CGZputl?|b(6vt7Q6rO7y7=?&nbe)2kx$K|JAr43IaQ0NT z{yK)_J{6IYQ()oQ(D5BTV2VM+qDRLdKm8sobHsCfHeUixA#75W8cvWaiooWU?`Pn8 zpR$XI8@C9ld=CgcmsB%SSFN+S z32R#wog;!2phw0IscND2!k8Owg-Px(a`*h}(lTm+CS!HbA`&|CA~F1K#wt z{%w?EHT{5f_%Hd@VkN&MF$9ys#ZC#4ed)cIr~FR8Vi-2-4=>KW#Fbu~ zD?Cp)93e3eTJpCtXQ_wF4aoi(fJhThD)V^ z_Pms=`2wGjSTM;R7D3;YX-8pLW29w_7xXoLNtWg{gT3+OpyTa1GS}~ej_pCGfBwBD zO3gN4b*%aN`dN`uvxea5Q3szzX@!>lgc;+BuiM09_fRp>^BBL=?xw){r@u@+DY8}r z`=)}r8se<3FEziX2VTWmJw`>9d5Lw={hAF@QWk_HcwA**1&#XzgNs5|x$GZM2`ili z%&o-xVuU5lK!L|7t`>#m7hv}q6qC>|2T~wnI-#)j@+XSNYo(xV(;aNHqsJ&a(y~Zj z+kOW-SjHKh0Af<6(0l)$sORuRq z?nX>t1L(&Gh&61_(%NB~LGr1R+X(lBwte~oWBEsFe7kLdBrw~NZBAxDquk!RAW}GnvMhWoWdRD~&C5B_Xwq69zS$#POn*K5x0_tIi)JXw# zr!sof0nTg|oCR$l^oo*74?*dUt9t7P8RSl5f3_H4xDWRd3nQ}sp%qivxY_H%XpbjN z8&UC;DSgt2(Dvtu!z6$9y#8k(YLR`7I1p`%Pkzr&xVJT~pZ8;-_(uMgmN2D|evB(Q zISSDrcO{F9lZ7-Yldpdx@lu#1&5_m2id(RL=r|%w`$AO=h(DTWzJlAxq?ctgo3zqe zy8I1_#Fau`Ipj~gJbO&nCMtTU6l`;WHS&7qn(+ZigQp>pXi3}UtMXHg-pbVtS#|YV zpVLMH**{F_*VVR z0LsiE*h>+eLjEYC9Jes!_+TCP`uIMr45BA8jn`oy~z30$K%M|c4 z{#GDxncfL0S0cbm&3{hu4@dL*B4#scMu)=r{LS-Gu`)YrS;N;d!L=j@GYCUz|5oG> z$DtA7=Ym&-ydMU98m*sevmldladGFYi})fByyn;ubNaWA0hgk6_MXgtDpr11^3SQ; zXqKI1W`~vR6QQd8)Sttq-yb_-{mv9F{%(q&zkC~PDwgyHXY|$QU30ZC%Eg3=?HbX< zmC}Nf>cafP!DBcIzPld1NUp*It2H%kKD z-w}D62(H{R+i;2Xv1es$%qqaao>k1NvKLRXKtW%$|O~pY)yGH!n6oDvjLE~~INvXcm#c>h^+ICRsn10{wx ziIW`RWt;B1Q_zi%5Nk0QYu2WmZy9ef{=Ooh6YJW%yMByHq>I^-Ud9eyHs1>h;|@L&Mu|G;Jk!z4JbqXiYhfsq5xvs{Dt z^8fyXP$dwg`+v9V3r<1iAWb;q8nxAjLN=QRXhS3whNYm1MKskG^tvTBRh>(1GxY!b zXa8?VcPuYh(HCyeghco>*ufACA_d?cgOK}h5};GaGoU;^Q&X0fy|fpDer?rui}xl5 z$XBF46TG2Oq7;TC)Ra9ldl!1`KrV5oiVmdJM$u=AA7ODX1sR-5Y4GQwHs4|L>0R0W z7P5>5f8jDan9RI!`y#ME_@B@mJ_FSJK%XdkFAUiJKxYSnwGJ@qm+*~X6k=(;q5c1r zEMA&$vOW9)*ad;aDv}*A4T5t`<(>aNwHCM+5C{Ogf8F-vD-Hx#VIc#I0XU$H=it|p zj0 zY+m>#5EPgl{{Ks0h;7f`Umd~3axKlWUcho0lPv;iXt(l+HQ8`5hCpDEV*E~_5Vvf$ z!IetkXI$w`I9rQS>>|4;l7SRGc5)!+ZvFwaO9^f}mP{mhQr=Cs*1o--1J4xk&G-Lk z1#g>Fo9%(!gC2ip`I9@hVd{dekeM0E??>U09C6LR8QHN)kx$V+7lqQ z;FXYxcn7>3;Ck?Ib04|#?LpU)2kZo}B>D@CG>q)i=d1ipvp0U_*3kf5J~oW8Z1z#Z z$>7mHgTy8ybr3;=`1bQ_YbDLA4IWeJxWEg+n8+NG^2z{j^8GtNg)m%(EneV40hJ{z zH%WT$KK!N8OXH_t`eELKn!m_u&!a(aW7rK4#f3NI0KLON&fWGScQD`+zah77SM?i2 zkC=;Q-j6b%lh@P`ibFHsCtc1o)*pn>Nyuo^yby-oiIZ|(qYzL=7!!FbDrLtR2H?U1 zZ`I)%O&t@?5$NLC=`qaX>tlf+tfv>Cy9EmE!tcV1>?(jJW+ zjG+ji686J5C>jja@7#$6CqBY}WVP7dsj&}3FPey>Z2&7Ia0jphO7Hrxj6VaRCj|6b ztNJ-W719c{N49(bt?m+*2Q9Vt!yE8UNp|-En+c}fXRw-pD|7;>pZ2YLK*-Ga21XFl ze#=_7Gtd}!7-L6_Ur8)_z8}1CB*BjZkYBn=7A>uzrG zKLCBdhZn)1pveMP+9_mz1*-Omr30RM&J!bfkWVm z;~s0;n+X$Db%FhdH29)G&ZOwF#MI$O!c5{41>AeEQs?g_0x2S)Mipip1+aVt^HtQ$ z^19isxlzl!1=u}PU3aim`=m4pjc|4q%gu92pfr51a?~J43o=r(0iaC}M&WalmG@r_ zh2g1z!WDMui7(T6TFD)toJus9Q?u1zX;G{+i!x+a=^$Q;S(mFhtrfpj$ByC(%b@8x zjL5-H?I$krF2Vg>R=4H==;hy8$>jaV@tp+c<5S;%kUxJY5DYV+i24tejIu{w&-AWq z?52iqzoEH)neb8Cl6JoQgI|GN1}x7Jy3XdwSFw(=>4^l_%y6c}5MRoKB)IXW@EjwXuqEloQuA+1D0y~SLVlc5YOKQ$V%Xsg>m%4XpEkltix>322=xhj=&9q zr@C7eB!S`M+PvU-_K-RB@78)yE5yj{3I+^>({*E1n25=#ZA{K4z_>_N*rp94&|tPP z{r>*FdTmC=g|>k8ML!o{K@c;*?iMEB!{;EjgpY@rb`<zJWoj6Aiy1bUoMv-85^1hzT+U1vwpLX8P4t*viy2>qYTmWD&Y-R1+*MIU ze8jc-_4{L3QnwU=`4`oDpjqgU4rTeRT}PLeqHAuf^+<)|1Y$LUoyNlYFMNE(($-vu^xO()FNa+-nf1QrOnErQcdh1Nr_7y%>vQLOScYr z5o-ov71kMU9kGPNDZmgHR?T@Q2lB!H3At5$HAg1jS8hfbjxaXmzLh<%oUCaP4y%bB0 z!`Ya}?u5kc7K;%__WIPp!{S`5$wIb50@&Wt(~q~>7EDp*n7RMPQs^P7&Sc7vF%G)- zE!t`$nw?#;i?h--o}Nsl%0 zPaGOwW=ibHNtX_9g9c^A0K}Y_VEN#X_3vVD4qh55|1J}e=_Fls)@!88VAQ}~&2%Jz zuM)>wq;$I5{Z`-K{x*o&sRrY+_4i#~1KGI8D`wXrqKjw>a=hlXpOfN#Kx+XRnk)Pc z*9EYe9~nSAo!S#pS_oNu!-s=jK*qaog3BGr?|rCXNF1`{gbH{oAQ%3GmOK!KCrnhk z5XR1J1y;(?Xh4ZB@eK~;z)gxx<#-}E!q0NuCPS$VGpxf2aTg-0at$fwK^TC>uep)X zB$1gIQp%CE4HAxMKTKw1Sp!+@)3Dz}ZnYd*6W~P!etlr(N5h@^kZwVIHx)R9*`$25 zj*2kSNS>Ea>)rlpeG<|&QX@WJb3ru&X%9ThTvb~t~}FBjkXQ(M*dK)-aNmr4D| zA2KJEl`X-}gGi!{x?rBrIcBfvc#_^QG|ojD)}dvYBwhg7rc}As_OkZakRmvO;3UXXu(9ky)0sGC`xqK z(mL-P=C0SxwBIKp^p@}EY4gZPUl$mTXq%D`Oze28erq6R7_Q9yjJj$L3EUg+bz>f< zC?;3QF(Grt?v5WOrAQ38W*GLNs^vJFZa%%ul=!ilB|BCehkbEEnD#<=NW8}ZDQ|na z;~sO+&w9@Tfa&ThV{zcWM^Aytu6*DNE{QN5F%TZjf#1ft>g`o-w)@Z zoHiTM*0QcBf(`lpe&_th)3Bi`x(qb#22#iqgUL+E|nD$j8;$z1HWP$5^*!KTVQdHucb*@j(07CNR3r zlJFxEM_DU#odo5*c$Zjj?G8_MTD8lq3&GZGL+S9TXk5O|e?REoUds~3e?1xXP@#-y zfWYeLxFLslXuMaMv?D0*txn5oZ)P^{<38%P*^ z9$({B8}!QR`%$7<>i_Wr+>4!mc;llx^5oNJRSw7=%k#qHc9b`e6>)ZvFEGbS^UyY; z&Z5yxJ?ZlB?s$QAmH-YkV2{FEMrP^63|^76xaQ-q-vQ64CTL|u@88wMHv-9Scx|^E z8N);I>}VU2nnb_|Pi(Pu^QA?BA_G;ZAqSOx12!PQCjH_vW>N_}%bHRcS*Wg;TrCHV zHZKTt;jvh#w!=AwpgceXXp5#s(Gz}%Ra#=ppIIhi9mE$i#J5}E+@SR- zG;56J`cX%dYO@%T@A)tbffGNhwAAA9SFt#-*=(s|V7wC*cYQ4S5O|2p5!+-<1o&99 z7$iBbj9BaP2KSxR$we_&j~=(gEe2a;liE+EX1uhb`$Vc*KXV6HB1)K+uNqkh^H}8h z_*i1UsE=Eh1V5WkEQAkL2IR|kGx>0ce)}#6m)BY_<5T*W^zb~jVMvQj3P~GPsO4_A z(H;xjo5EL-s#N>suc4*v6(Ksrg8wtn17F6FGy+NkiRYXO*pOMd4=;4d8oY$6=+iOc znAjPavxABKpC4n#Sz+DYzP+Z15&gWF(3jgTxxib0Ve(9ct$}q60IZ&QZD3xzq-}-0 z;{8tAGb=XjuRQm*fdm2Ivg{;&Jkl$dsq=*yk7VM`qkm_`?x1Sah~x6#ABii-+82vM zD{=XUajSJBqR}Z=ynO(#s;M_&77~ za)}oiwQibu8Rv*i!lf@VD>nWtr%VMEMMNHwgCXn*J{AK*!9JGy@0!SXs`zu{7&Dad zBSgjq%}BzDda3P^=tqr0OLM{d-{wxLEfEXWGnLehLNix(cH@?5_51^(5;i zqt)(le9>(nN-@Ashe}h3TS{&K^qYPZ0&@fdojK7d7+i45!1st`jC%kSQb=fAy%spZ zGDA-oWYazSa?~-Oo4u`r)j{Ttu_c9(QI3}xRNJGG^x~Bf_ zYf^<1lBV_#46S-CANgDOmdLe6PPAy4oZjk{CIZrCLvm$iq7i8Ph5XEI#xkwo6ij_igUM)WA^+yP7??7r3P-n_kn8fRr& z|FuJ{s&a({;Mweimj%6~o?fj(*OD!!mjpjA`ZdXLnO8(Kllx;|@Y!wz>7Dwjc=C+|d_Z9?$Ca613AqJFwEb!JUbA{~--IaGCl57MLum zux@|AR}3kkMA7pA;G{Hz*wj%p6n!2!7(jK4iT5_`I_xln%#Y$ER|_@l*%ApNe< zOgkxIl8MIEA1qkZUXyTXWr%yTv-Ve@d1T_3=4506ZZoF}8f&86Zn(S$ZLS{heJ(KY zO+}?;!v6N`UETB}rn??j2&X$#^5dr&WJ&Y9)8=?b?2tLbD3Ka~ImEz{5Fob(?_k(( z>MfaP!(#)8Ex*!@+`bUK$9R zS%dl1v&9niQrX`@t6wAOE`D8{zlLpA-(OLT`0NS4G0)^-0Fr_K=MB1)9cvmKZ4xk6 z4M}{)lDi-82v#%MpOt)8)Avg5WsjTF&%8?oCx+4`>kQ??#Qn{2Npg?8=&*Lrdt9@L z12*3ts%aLCy-9=ngNIk)j(w<#&%}~_A!DJg0A7umLJ&V0{%Fk$hgGc=i{VWD_#V!^QQ zra_T#d_hfVab)+O|E{pd!wDc6zykW;#usujCoX+!@c8^pF#x;xot zxSvWKzim%b#J7_g-1qbe6l)cTjS6(2-{Z?&Xxe`G=TVixwY=(@MA`3FrEl%pJPqR% zZa{s?<6<5qptz*Y(7A8_VmqJxeG%^X?h}gE@5G?Qm>Jjh3N+;<^Ylw+Wgo=q_bQ*f zo11bp>8O}FVEQ&XUs9pafNz8lM^wd8>S3{No8Ls1Q-pn`#%Y!Gm&!Wj*-C0^yS#EM zJ=be(PF&@Wr88Ygg#Hp^PqI*t$v}$v2vPh44|2~EEN)KH<=4ws#6}-q%ei}nb9^j> zp5#*)(|6LO#HP&w7ZPhz-yMBl=NVUnwty!Zn)46Q*@-nw-J?n7)s5bx!Y+z=*koDm z6kqhe;~W38Lz;J@T)cA0WeuTJn<+zOf^JK}XTAfH!ewvbkHZnW=nBF-uw%9a+b5WT z!L0Qk>~#O=#eKy%oo|A>*lrn`)PQ0fGx%txF!!Nc&S0&*OH({2X)hLjeXHpj6ebloqN%Gw%K4gW5ck5U~EMf_*QHVhKB!s4;K$ zf>m2s$3IulaIxY)aqJlvV&Fv0KZtBpYEAmScX(4wOSL#q$N3loTWdYtxP}aT_k3He zF^8ZnYqo&ubn{o@*TGbWvd*V%76Cr@19cvIN^29~)LQ&pcM0T)3;N`JVmlyot(>r} z_0zD2(Y3bZujB~%fS{4|H-k36^J31JqzlBhS(WZ-txN7dq@g(dLiN}eH zNpfIyw-Pl~n7d}cOW8o-D@MIJr6@@g+kcJ>zT5->UF}#n3l@v~Uur|h+v@PuxkHf+*0Hw3>=8ueg$z7PrK_#e-_w{TdThR!LPX(Z2p^ z*&y?$M>spX;giHmfxYIVC`l~JYGy}g{0MG%NckCHq;7NE(9(5J?-Qb3A^O!?TxDMRW+I-7 z*KXv1Vou3Gj>+r|FLhM#SR)PTqu1DiQy%yl*w%Hv$tMbUCy%i#j4pr#xH9teM$@@4 z9YW8;$P1R8(&0UEh12zHK^^`2h4MMF>G2n(ILl6@-hSEQ&X@5$l(zRwo7AA!7D6W9 zxywpMCLcdtX!Q;7MU;fvb>-yMFr63dg(r7O6%Tb>&lR5P*JRVwnjc88wt?)sU%6WC z$`v!dXwVl4?vv;HGXnzj#xaL+k_ODi?B<`VELlF87`4)UOfUa@Ff8Ii%~aX1z20uO z{n3KjQ}f+vpBTQ%{s^#g3~i~rJTsctja&XL}bV4KeOO|ex;3-{Y#9bmASb2 zakk#C>ghk#+3$qQgG%ix%bKNM%~#nXBT^L@2GR&ia5_0w=*hsP!;e-!%fidvLD?d1(?B)Q?KrqSTi*LBGJ@hVSsP zE*MfT6xl+N0S9zjkrw||{XGP`oX;aL69nsPtR0s9<^+I8qas&Q+1W|)7uY}IvNK&W z?w?mz!l0%Eptg1fq9YjZ8w69Pgo5TV5YW+$Kk#M#Km4!DFb4JlTcBjb%sU{Q1A+N9 zsB;alOazB;yT2PJ@M*CKfS)beA1HW*KnIwb+6Y)J4nFqOokq7W0IUA|59g&RVuN3G z3Zel39g+~BuIo;kgX0VY;R#c!>LT~n!4@4Z7iiu=Ck`iODZteV$!t4<|3RK%z`k^X z)kJP0Z5=OQDxkq29`nXrA94$_Cj{_6Ax8l7#Xv`nF-`~Y6hML(rmwSDqqgFqj*j8| zKdrw1FYoXFJrEdY_W#TK`~R0Mc&ctcd}p|u&9K$)HiksIzaZH6h1)15Bbs_JMNV6o zm4~DK1_a=ods2ntcQloyJ)ku_5pjC9h!jCIyaN~w7>)MOWyF3>#$c@Qvsl5YW8v8< zbZM}Mc>`_m|M|n-4V}fYo~5uhyMGf>B5O2j@|c3N2$QCO!8TJuq`wimHhWAhUa;bJ zT>$=hT#QwgtF`VBEV72@A0am?IbCOytUp*A9Um zKBkLKw>2;~yUMI~5g@uCLWTiYsGPUI+rLy_GH~9ME>5ClKy82!xhkM#z{Nm;XKCm| zVM)`O6lX3H;YwCiQdIJ)BreTS-LMP6RZ6L5NZLUfl0Agmaku@6%*D&sO((;S93tJn z?gQJm!u#%*XT}Sst*x!UppVS4z4iJ|!Ngg_K4F{LL7MMSy_GU)XS6(4h|x? z7|7pwq>#h&np;7(-n=v;h)L=dtVDrC z=U~?`m3I_d-D03%Kr(HF?7dua1Fouf>w1Ny&|4~hs8;s@qNKAk8;r#+u4J!=w=s#7 zkhe%^;Nk0fepx(hHMj5uLhTx$T5TA8eyLF$umbi0QM`BLsc!KY7RkO}B{4;(ij&nk zexxOnV8=5`r`B?ETAA;RB$wN&nH3K5v2wERVmixpgpbwrqTmgD5`*aINg<1VfyyQw z%u1~2@J)(0vt$=QR9N1&g=#bQZ+<~FU37mnpu0=xm@v#TA$?`;f;$}FB4c5DZCm|w zsd*{2rq(f>o*XL}%jgn53~J$L#<^GmTk#F}{ieunQ~~*Lr$Yd7+ZJBh1wK+D zA|kFLB6=1I8p3NO_2re7n%31wh6g^=SF2H7F>vDttZDc`P31VY-F-hj@}Sn|I}l8w zu*PTz&$zDl-p5mUb849Sk>RZ-lX{N7NQ{Juc%`rNSEIQI>zjs8YKC|&)^gfj8Fjo9 zrB+s+G*D{`8)Umd`(crZ*3M1(N{ZT6@_D>T&Aij{h$%II@7JmE>=Xp-RwxobW{?HY zc<=>2`tws2Z7>1!113B%Fr(lbyaPK^Kz}D_F=kbu^oA3_v^Bu>RS3igAqdwR24ML; z2Re`OOkgR1fXtGPDh;QeY7AsPvXEO*BZ0BRdGzQ}32KcfzpE1?-!6#;Bt{!(Cpj`0 zm4@0(k?MxRjCHvPTRC9r&pGh5anc`#K|RpKc7tkH_Yk|2H$TOK_MFay zmQqS&WE5-dHs)x`B=aNTgyUoL3S^%-=r_}l=SJSB?e{vHG;!A@Kp6wbGkYH!bUDBoEm)<5W%elu44 zjEf%&-kG{IWQX(?xk@-FW387B|a zWAfF*`$j(Z3h0xFUGn9;%cUys8gx+Qjj|t2lcw;_K0D=kN40DcRV_^^^er@HCP_24 zShY2Gad*vBHkhZxuZc!7h0^PT3DZ(brIn~RaU4E$-zUS=OS!nYZ3Vnm zy%TDF3^>rY;4WFOAXXiUw&}%g>E(`pk3TBt56>>DCyuQDSQ9gw1 z&*0dHeM9#PnZ_~57Bd1fWxJht7^at!r4(V$3|z`^;7G*_Lpa%<@q4J%TA6TGr#k@! z6bGNQ>UN{RVkaFv7*~YBa}PG+zd_TcxIZKgT@NoA)-c}UL+s+aagKbnCF?zsPCoY@K^qB05F1K>k?QX^5N8Fh z_AxqwB3Qg7X|ShUz8%ZctyCp<*~nNxJ+|2B{Zoc{-hu!l_8&tZVzFyagnR5-Ep3{4 z>#>BmRqb$};yd{zIrkyU`uh4x?&Xwb$7S!H>AL9g1SZ%xcz%d6(-!M`@fzikCGI)6 z?35Tgq~!2kl!>u`xrQAvEv+B#@OYw}L!32dGJ*T@o*p9 zjH)^osc3s*+$GVNc*aQIi0FGOMjJE+$XmvBU8Ebs-t~}c9W337U2ITYa&?RShV!FF zF+u&yO*gV-?Z}-4t>4cQk>(GzQ@i36rObQ;uc?xrlwIjM%%X!M%~XZL`Vy+Qi!p1pzE$66{9&$pvy zW?G6^coC{Y;Hmm=Q7`w+89V7%I8gB;bO4NEt+dfY`*0|H z3Z-seU#?2rA5ofn6|l>7eaKWX*cDtmOy=ms-i5u>d4byjGGw1;c<@9kVlx-_GpgOtWRV)gQSO9j+kRA!>cu*337D2 z0;iy6s)zmzOAkoP9ZwF{BWDG!rakq>i%1LD>X6O0dTel3gWc(;U&eIU$jQCFSM%cX zu(ObzqAxM}(HifNs|-J)Ijm<$^+U(~oQ`_eq}eqo_uZ!U8r9O7hF@2E9-NcSMc2w1 zqeFgB$5~~Gxvc!~^ARzx z$fx51tZuDH(cybsO*BkTNUhWQez{f<56o@b_vo06T0iNa2~Zcvo>)dbDN6RiHkTov z7?kNXMbc~ZHx(~9dwrPrbL03@@0;6`9F;^gN>S!wn>ZD)Q2pF!9)00IU4F}sI!@+q zmW3Dl5WD!H-WvIUIyK=Ka7lh0SfsWSrQqIgCjAZaGD-y;sq1UxE|*>%m(q)X8=ngO zlLah1l}&ljCc?DP2KBSlw3Av3X1TLS=;}EG{J2@oJ^S(s-Dn| zsW3rz^3mE#X``wXQDxY7*@!ljf;FpLW9_Pf=Sg@n%c#UX3Ei*oukDlm~ zGh5r=xm^fC#CHC{{1-T@sPpoX`HYU?J8h4$CJ-Mpu#2lv;S~ngxh~`dBeSuIr1L|k z-qDALnLF?PW1~#*wX)$K{~9+ivEj8zl>bwhVZq^d&^|lf5q8I9gKNgz9> zWCWCJ@r+S^*b<=~&#aA+056bP%3+c(!1C6|@K1I9Fl3JoV?;7<2(U12&FEk~OxD!B$*C$s&dt+hOlpxo z_c_pa+lt3vh*SyDmz(1dK*-9!i)L54_t(Qs0Z2>O|Ew>LO{E^_p`)4=Qt{OtM#{zZ zgSIC3fI8)Clye^6p-PKch94Oq|C(`MD*z@S>=JXi-z*Tk8EV=@n`~ae(v|XGbt|6q zhZ5okdU|^+)P=vFqq@k2%nwPt*XJs+ZHM>Hh7dLbqqykEK<}{}d7PwiA-Gv&` zWd5GKU%dKT^vvl9Vd<0j>q^@~wsBZl%KI21lYV~0Z_J$M;Ee%Tw!5ih`2W=8~JueTpAl3Z}) zn%xkS(Dr=9nIer%GghI;WZDD#pdi>GfWK74CWg&jfK(qMyMn-Wgtrxx1NXr~XlugH z7#a7p!{kPkx*#rP(AsDxKO#>sem%-*^YlRU)mkh$Tj9h6F%)f0@XTtfBFw>Lk%1df zDdA$U8Txep=>_R>And-U^F>+5^$PJ0u?y{-pE z?MFd(bnRQ@&I6UYzI-E{*J`PtpfiJxe1Nz5)78_QlHqsQ#rJVu z%0YDD*}eKnbmwmsp=jn(Ll?$}{uZ)_r;GMBp2ED)VCfY^(60_tWK=s`08we1j!UE{ zgQ*fVn33NDYd8M`F(1O|5sY`n&j6FZ&0P7&kbmg%mDG-`Zx69+73aXm4uJY}_S-4f z3w*kWASvbVjrb}Qy`A)&DG~sXXRreTCi)1W-RJC^FdgR_BPAakHAA$Dbp9fY4^d&i zO`Bjh6~br-h#66uO-k=QOoXva6x&*p#SgF?*TgI4qNsYH<|G>mklmhkQ6PaD+9v6H z@x1j4YK2Pn(Y@R3i;P;q{2j0EdoZ&1O#b-1+f%57r=nQF?P1#~dGhn1i;QySyAKCa zAyweCFxhIG)&L##>twls>e?1Z8};2|$r9{^IhrXoQ)J-3r<1k3lnz1%@?ZYUyAamuWGyF5VdQ_Wd7^ZqMNqR=^-Qxe}2;8v2rQ zDk5IXW?Iv@5iAHl+^fMMUWPd4QVo;Nr(f4z1Ycz%j`AC`TD}p0Ve)IS-Es8%@|@-- zAt~-M)g2Y37C8efoLSk&uTj~eAQOb7NKHn%Ur}GesvGmO9K&r~xx0hx>F#NTL z*#!73ZXWP4Z+e$Ai8&Bz(0+Z<46j%9VD^=4Hjj`_qPM$>h&Ef6_ex&-*?3+KN!vN( z))VH{gD6pZ_ID-fJmZ}>RU}SO^;YfP!;l@xSkJRRn_8i>hcJFF51KX& zQPN(Lh@CrVnvs{c&e`(KyHV4rfsnM%voLSeyBa8!oN!`gO3XF7hKZrZI)mf(AI*Tw zYW~s)4XLgQ`GPAz-1_)Fdk+9lTn%v5u!c2@xLx=O(5`@`3%si6H2Q0uw4m z)M|4IwI@n4$AUvyGDPM0wNs>C|%Ma9XH*r(jwiBfKuK_ zBaPA^NGT|xk^+KsmxM?oh!O%y$9G&l&+~roynD~yGv7aZHsd(V&E3~^o#%0^wT`tc z1_Rj`IQ^eIJZC3v2qOCMx#jNiVosZPtdsiwt+~0?duuLqo;ku{0>9R}(Y(^nH#D%r zDO6AllW%8zGVhXo=knbTzw-DDRo$7jaas6Gag(7G6^RK}B0I>Ys=m!Y^%fu#M&nlQ`03Hw zBAzlERT&tP1w%}qwUPFAQNkP;$N}J+jk?|;T4rS++v(D=_X-qc#F5=Y)G_=jCvtAf z1*kHzIGANET?a-;zMU4uGD(pNwlOIp%{pA~Dc16|b@kJsZtc0OQ*q+IPoong` zA7GC?Yn*^Fav1V>RdE!IK@r)H-n-5mY(!s%VsB-{+zoHgw{pAx>qWtgLI;H_VNEgS zh7wAeyeqh>W?^cYnUB8bcFt?8>CrIP)|%5eC=&Lc8X34F! zjd85y6K@d5@m*fLJzn@*!Zv_h4OHaR$OZ{c{O8yRdh#e?AmJ3rqrM1AH z79#XWFz{}lDdGAvA&yRfg4#{1X$oe6ljgzHU;>Xd?}9?aJ0P~eqEkU>$2;r(-ah$>{(08iX;)` zE^6Gm{95S<{Q1T&(`6Ka!=x+rDHsii1C|CJ0!npVG*6b@9>$y#`Y6%f_5+Cy(RG|w zg?FMf!2(eSN&*8_6Kr2%uC3!lozci;-K*esFRiLfEjNDNor~##A--6AM!s=ZE$RVz z&)lAmKwRyOB~JgUM~er)!C{zhYicJ%%B4p(nG!lfRN=bIt-Wy_|7-`;XZO&hEG=sC z7#H2=I_BtF-b*X`BMk;+lbo_#cK&bK4-E_U3`$ERt_|KTupN|KahQ5dkH0K4>tt9) zBkC40L7B$6&i#JjHb?2Dv&vNWFX|g*V)^S1c{TF^T9pRcX}KFV4>52o&Hq8WkXtZL z-Cf9>@cl!s0Ddxk_xdWfZ8Eh!)d*|^a-3nk#&Gx%pvb-6kemO3Xj$fbCdQQ~|2Q#? z))}wY?P1@Fxs!rP$XmVEOG$giGW#7w7ds}W)iFPd2~ps3(~yzvf};kBW~6{Z-?VFk z3k+$X|FDzel0=jmQ=fu}n8BX32;7@7GU&@L>J+o1fh7bl{mc}YirR%kva;U|I*6}8 zIs|ybVi6DP3}Htx5>=Os{Ui~}ei3)4f0Egc0Q!wAIZXEg7U#{!hHnl!)*1)TdH`5D z2PYNm755+NU93r)y9+w?-$2eL1`Ywi1x)!RI}`?{XspDmqLjM7k?@cQ7uMR97rxEG z3;2T{a&=&$wDt#S;1K!UfVcoD)v8D?=~pm}3Hy{?gXe>k?imTJ$g~(@)hvLmA_?|s zb2q5tgoLnPKf)l$1#TQMpEYcs^Sg2_3_B3aLV%yS9li7dj{X&rFvROqw~c9zLULZbylxwDbk!b8>CU%>Fz2H<_(Vp*Zkkmj8g8I$u-sRb1iByn<(bRYzl_! z35`;V>A!278EZ<#wmR|JkeXzsPj7xCPkmPCNL$n5RQ|GNwfxf?`k+G*!h>lZv2B|B z40JRr3p#SLd!$Y7|&Ue+?J&{uoo_+ z<=v}Lgi4#_CafoXT(i4PA^b#ET=n3Z=mLrLpko--i|_eDjWl;AuU3YPg@hBPemm~D ze{c8c`AU6OejY_X$Fj3Jzo_KBx>&BCo_ld$BSl`RkPQbpx};xp)C^y!Pk7Sy;9-~> z|IEc!TFdxC1!q^xtEr`+Ux%Yd=l;A{aqI9&pa8RtLd(7MXx^-uwEVzP8d%ux1`0_D?afw9V~tvm$nto;CdKUkjK!PjHho2e#tP}L_$1E zz7K<zFe|d}ZY({R8d{Y!W96?IH5$nEdGk8WE++QLVw3_Y-Tz;!bKQ_iLZdmk#MRS5O4xh#H;<86W}5aBk3y+$pLx2 zXF%7$${OD`_|8g6Nl8sj&DnV!G!=*`WyrC_)(8E_cHK7EFQ&na6rkJ1(7wLDf`S5= z%!QodKj6r_RO?S?#$yB{0yZ!u1Z3`}ehnB`=0l?oMF6kqdz}!9E@YT~YgQ9$3R8vu zs9u8shp_-vfbtBP4+1Bwn-I_ef@*M#_h{P6BgaR^Oh5vF90Oz&6E+?sSx5pDb+F-q zGR^`5M;<(cQN55aKxB@kLmosdFC{c=|m>AfRF~fKsJ;E$8ciqi_sysDjLXPSX|;<99PA8YmW;8n zF-UoaV}NnE5UkihEEjnS+9dy^6+s+9K)of0 z%K%ULfA*iHOa*8m;H5)DL$ArDc*Co~JYk_OZz%D_EC`c>H~q$>Mr$tP1|iP(ACLyG zt)VeCNhKGF#VG@M7Q|Fksn4G60e9ZZkZ-*Up3xw&01g=ES`~z>2Wr0|mZK#G8s+ic zJ>eCi2lBMyC&zQ=S}H1rphjvnnMFRNwxy+|volv}sArfcwjwg34d>1)8#oe^=kzjZ zq3O}z@I7#ova&u{e7f)_7`gcb%3@Qj9-_q|Ynb9Dz(5XL0wxJ^1d&6D{E7?b_TR&+ zg8uvKe;wZc+zzSxz^;Qq#YRjIp!E{D%7g5*EAHU1fg zsfTbFj)oWlFj*VSm6;11oIsde(5uSf-40WCbo>FA=vlGMsRO7&j_d4A11Bva%)I(< zVmC$NAtyfQ4`$#>x#$vsL|nu{qy${LdGkgavyI2Vo}K}K_u;(o9<G|q*T|0u*W=fO(BM!$|WSWUScmC zj9`WP-ir&(uPt23h(;SCJ%Kk(0i`YgKEMnk5ys?^9J37rVvqV_evvHD;It}Z`Zz zVQi1lLK|Z5U!FH7;A|uH!Zb&^bx1|Q?M;G`ok7ksohVLCU47R^Z{Y^<`Uz-cK;D(F zdN(qu0F}0no2W1XY`Tn^fH|6PdU@qxs7hPGyFqu90Bik6mg{zf)^7(#U$7HrQgllm zNqHxwU+$}4@Za9!I1H7!;f)hQ1{c~3iD!F@^;ZT%wU5Jhjivj>hnjo zwe_Kw7^eem2aM|0ZXA>QH{0ne$Xe&-kmnOq;0{}KB`_%HvF3lrF91~xNWM4(R}m&1 z29|+u0#0fxN5t*!^Kt?8Cv-Ar;WVz_KbKChq?*b-LZS^$)$?E$-VxNC8am)I5(N02 z>Ns2#Fq~6&CgZizm+4yn5(IU$&EWJ`rw}VZ3!H>NUJ4+MXWw{(8Mu&~D;Pt$P$>`~ zzWoj49i8FO#eA>okL5IG$rNBtwnt1BS{gCkvk< z5V6}Bl4CVF)#$$uDg#9p^F^7ChIKe&~tJe~QrCv&zzZF~PANp>{Kx#kKc zew+6JJjTI>XHYCVhwiRQ<@8nW8|w1m8e1UY0YLjYlUGO+9z|JI7*+W4;MBc;#0*3J zA@5NhNVh=C^0Htk2`nBw?x|;wvPe^yNyNWHoYU#Q49I#9)-;8$V4c{=dDKw3q9T0{BOybp*)`z>r&Qq9pa`cw03eL!yTg)Lz+!=R;0z7 z(uCRf(kqE<%B@JMg7`I+YN1>T?B7zs;?eBWU1GE(e8+Iq?sqz9(%Ng%Zk9&B%2%Pl zZ9D?~BXbfm)GCNoMZPf+7)&y`RvC#67wV{5psabD%ND#-w?|`{5Vn8b{11M?MTkw{^1rU@t>; zh4$f{!bR@_(FjP|$H78`6=_m;vs~UHlo+##%Gaa9$d!!@N1*P3+RHc;7G@K%Qh;(0 zpjvpIF)7d5lod!Zetv#f^JITamW6EBexJ085B#1NAv%jh<;AMj7Q#!Mtzfg>9%yxA zmVmU}gv);6;RmoGy6)YRhXoTta8}HXMjZ`RT+etq!92q1Gf29~bkTgp7JP*XD@a(Z z{>dmHuJzB%Z;t4Q=f4BE53m8oKq&i$h?+1hk_!CC;N;FjYRLrB%?vC)a3jSM&jd01`Wge&H`G|d_4SdCEVdtw*YPO?-! zu*U6*L}4uq3K8VK=?QOM>x2II7*$qBLc=Wo%RThOz*-)YFrN%Fb+Xzlv@%&=vhbpw z(BZu+9No#mQBLFMFK=GtpQccF7(!^{5!!aEKQ8n73X@h1ySAbzi3GQIh6upJTzuxM=MTU)_=<}ksx4>zoXLKcv!+XE^>Kyh#UV}HZKe{e^ zDe)(!ib{aZ#ELe}j_6P~o)Tk7%g8Q?7w$9q#;^C4bI2p5ZfbO~mXSTm+75NbCOv|& zRx#IlChf97ctfwf2iiB7nG8*CiiMPGL9A+k1TedOOlqEYpr)A#XqSCQmUwMJ2wL)H z4hJVMEkiVt{=&{bW*iXk;Tl75oGGliJpIsod?&Epq~uJS)`kox>~MX0p+Km%_}w6I zWW3FrrU2)?ySm8OA}G2F^=H6O@{NA%Rl$yTn`gJ^g|KH>E4l_fRxC~)WupV>pO8Fc zHEVq^{YIo48_P76)uh4G_tg{EzGg}ZSnwy_clAh}xR&yRKJR)2Uv!xF+pHTZovdDV zM8Q^q_dS$x^~lNZdN1O(<7~F3Vdg31Q&1iS#g(9(&!4#cGqC;hDGkn|zrYa6x_77u z$0PevO2I;;iClqqpL=xsjCkLu*##$hF5kmR(-zqtxx?C`HZ6(8^*pP2uNA}M2M4%# zcD=68`sA9=nGv2500mwpC4;cj5p6NRU`nv`{E)(Py^6ty=W7cgAi zwDGZg#^=Bvc;tOd-mUCT2cya^Uw{}E(fqS9y$`!{P4Bx3%)BVjxUEDmKg3Lbo?W)s zSAyFLCF_t~fK}wpgHaCRl}RYzbP}2~_yHawXaaU(-Mq;3mOVH_owi^60Ld)|T!Y=E zig~E-{-KfzFJ-0{UFsQI_{&hgUWYyT_=?(M;@KQ5#+fUP7k{^_jlYk8_gPP z?O(l>jE%MtKLV@#;zSU+$(dn}UMkkNx#9QE{<9c)S+rq<-5WnBxHr_E9gKkIzYM27 zy=mpqQY6Q3sF&wo2`_4DbV6IH1KA5J!BBn|PL0H#l_+YRKX$=mFp>fR=n!9P?0io=~)h9EtZy zbHrR-w=&ynHd1q;hdRU-ElEPyAz|Jl?fZ8|>t-dLo5YA}#EVUOf>) zw3RQpo}<#aGIz# z0CTiquBJRWpZkco5p0~d1(TP9^yyt&eFNh)a8pV!r=>v9VnS@?N)(FqleBMn<*M&C z&D~q??BNRxF~Kb(V6tna>QCseY+ilHfNf|GV z-lofv5C7A~-B0cjVIobgQY+(?mGYq8ilqRBpM;f6u>A(`Xwx*Z9!Queez#-z!1>hza3)lL<)B}5>$a9oy-M44I{KQ~yR1FwAx0DdC6&@^v zY7v~0`6QQf{TA^w6bYxAyW&AS%2aIUclD|Ll~$33$=jiqygc2P_*3!0{fr^h_SLTr z8n;lm{y@zvaUzZ1x^I5AL!~bP4I8CvR7nefHX&g7^F{h<(IOCa$v;LTtM!JV3b=6q z^)ix$M6sQ_zCVzM0dR|rpa}}ir#|hIiXs~W_V=y(^BF283>+L*`J%6YyMRQ5_4Y}) z`w?UltaCSv;e30{0zIaWuNQk<^(FF%yZN=y8g7Z;ELxUiKEw0rJ0Bm1E&6=1``3%2 zz6~DM#fuI4WOr~wlRn%aW^*(B_`V-+m`S`5|3JdJsS*wWQdU-1?_Egryd zsN}aMR5v0)G&=2DxgJ~@OXI709PxDub_}<9ae|5_*_vYe zp3BGRCdZol>*oF&Po`bM`hW9_l#h)2(6H<1Pmd<)({&|dGUQg6th**&51~BPMAt6x zfz%+RDCQx_+(ptm^nEm7T_Orow-6w;58ya#B`b922NxF^%1~eRJ~Ydz7ch+{ubwh= zv894s(`Q(b3g?*;x&jtgSgwc}y{JvG<6xAN2@5yl4$1FNfG_3{;pl-O1eY_oDBouK z9ZG@wEVTa@MHf(NX5c&P|B&S$hXlPMls)0^5oqt4z%X*Xa}{xxm8#Y$RHHbvrEaUC z%@uaZXh7{!;8Jysl|bZreE18Uwk@4XxKut?mI{WeQ0% zX%E3lxemv%4X>SnYAZg}MQhd#KlGQQh-L{c`qQj}W6{c8H1Vsh7EJLDVy3VCogd`_ zrD1DN>>++jR>@`=x$jREB^EZM%AOamYhIv&L7~z%69p3&*?b=Ydp+acpbkg{4b7 zYM73Hx)@UoDpyE#Oaaa26-ht7qI%AsWN-0gemfGEL#xvNxn5sDgLM+t^O2YOY%6cT zAifq#^t2x|&(v{tpgUUn2qRh%F(^sIRZf!?RBvFFLSTHU`xXQ&fTh)!ZFQcyP;2Zk zup7()E9y->PFUb5NcbWL*)>t3^7`PN$m;&~$;yjASlUa-z{=QW$R!?B(9mtboG&$7 z@qN=X{T)%Umx)=)LjQ0L0HyR_QPey<5&P)LOX{CW$Fc7)hTY^RtjxT*U?4IXt#Fqj zQc;Afa?7Me?h9k_zTBi>v$fpH*<%LUuWtC*`k$_KyC#ZaL7`$I{iPU_fsAnF>@~+; zG>y<>7n>;}{I>B1#@Vn+o6Etu1QdCi{j1(Y{2K_>(#5U~GCMu?nIT zTx)j3eCz^GeU$f}dufA_tpq1XuHwjk6Q0Sv9|Mnwpx~+nA$?ivY{{4Ix^~LM|g_9M}4d49jJ1Jjcv4OLN=Cs|?@#@!=X= z!hL&b8f_*7<8*wVgm-X4@dZAPD}fA31B-ZxOE#rNOQUA~x@a~#keX-7OhG~fGF2IW z#R1&-dyn3pW7Aq;8(@!>N$DW0g!hhy0g_Q{JnnfbvH zb^NenY>c2-*tqYXeBN~V9q;?L<(a$YPq+3*DlqY!(^o`KFPeu2XWasKKpJ z3Kb0vf};gMwTty!=_VjT(P^wiw!$ZQ6?#7e^?6h~UtR_4ED`uWiB#z#zU9oo**FO*<60;TE$@@se^J zwIRQ(-+`H~jK-uP{s@sr^gm?d-%u<)@XCY`RHWS3TZhI||N z$K=+w#FY)Lns*d`0@#VkWDb&tP`_eH*m6l7-x+t{#oO5pP#g_CI{fiMKSLNIt>)mZ3v^?NISFbd~C@0J~v;~9fRSf#2fll zt6rv#ZZpMbsBuu#sdM&YsISF>^8vQT!K`gGtZcMw2INhQ-`rG&>gUD3|#;W z>V4-8ua~HGwO3%twLA_m6F^bFq1azE|KSI-QuRU?)SknLC&CkzN(o;c=_%;Ecb);z zL)v#R9aL0nBfY#2L!Ua7u4`Z3m#RNbfmAs%sZi-%DD__OBZ03cR$I@#Mz2i_AJ8sz zwg7Kb?f5x5C=3??ip?H<{D;ue1ppFva#z@arn@)5y+9I*XIe_zCEm;kib!so={E!^ zSXxYu#PLHELG$;1bWD@WipKSJ@Eq>ekbRCNnGU{^a1p~jWJWcwc$j!^MC*eK(OB&b ztGo@2j}@9zwNzhqujMnJDa06`_myHYtS6<(t?dzLGuCpS9bhCji?!B~D^O=WYT-PK zF~7FU>_O&lFM25^OY6)~v9XNY-|i)un4xQoaX_en2ae~Ziz{oHahLK@>$6#nyO#|{ zi*B{J4jyh5-!IKL_&Uo@`n_Xp;`)5M%vfST$(Hui^6zgFMPl7UcR$2m>_Y#L&;}t> zu6tI>3g^*d7oYyQ28=Lr6&?mk_bGpX7|Akg!@p6dcTK~Z3ETDH)+Ahiv&+?Bp1ROg(e!y@9$G-jZv#^X661zOMIEIy-G`eL zk($*d90Eeg;0>X`P9*d6W0v+M((O~pbZf*wXPGkY4_jlTn+=;}EUg&nuzfk-H=wK#ElPqXI$}Fj@GBYA`H#fmqJ(ts7&2kU}P!)__{nj_nV)CK}pt65JYwf zXoBdR%Ef<}5Uv6X1gz-a80Tt&`7phKncR3GxyeO<>8rZf#Oy8G*6?4l4NHx%MEPsA z1wX&Z<{b5rLi1u#^AzXz(;LZoF52(+8^pH~o4M1{rU2HeKCl{&)mXo|MUuPxo;or7 zL^VF<;F=$q0Gqg5a(r`HXvD^~oC{6&IsI;>w{0@*)UbBxHzV6a>Aee9{ht%xGJg5! ztnpfv(u2BYqFGYo4*-Psg$nr zsgsC)Z-%z-hT>_yoPRf)YhD>UubF~P81J_OYf2b!v3!ZwOGVh@ zSbO(hx!t8uv3QUBXJ*wPiJ=c)y6S9qZdKu=W=}C9U(6;~G5HZIZm=MsX4ZsXcJk%n zX8I%{3oOQ2Ge)D@sMok&p0(U%v5YyjGW<-J&icKD@Nc$L)a7Ekt7|D{VtSO%Rr(t~ z=rN59;Us3jkC~jDgfUQ%=ZZ>3d9=g`AXB!{i;6KIYhAZnH*vx$e@vE7U*BawFQ5LhufDOY}h>R>K zENp~OEYSZtf(Q{{NyI4^DC_;T(N#F}p47JJ;WO9gn=}Q~H5NKT^}up0oIH7skN9lY z%^5KRvBplx%*;H*<`NO+pj|GP`8;0tcm@*@$xx(1#tD6mi5bh13=KhO@`!{bF z`8c7rwiei0rwf!3tliz+K*{Kw1L1h8t*1vS=&t+Wr)uP5+G8GWsh2l^q{Kd6>4$s& z3{zax`FXg#y$u}YgoqI?LmQfriOEV{n1h4k=#Gu_9)tAeCuq4QC4HT*o+DDWW78H!<<4Mx&D_ig;Oo@FU5+82mVse2pNtL=?D=^ z{IBDR5UvOx8~QJn_5bcMe;=iS{(Bnyq8bx{RBp1Gg6o(gn$W>zf8qa%%yZz~%xJzS@PgwKb@^ z!NH$N^CXklvYCvArU5|pckkYv3jZJ@6-eb#PZwnKSQ}ZU;MDsAkAB9jUiA4Hm*#9@ z0tXIbv;>S(Aujt~^knz`68Vu#20aFC^`X~#fz!(QYf z*2a4etB%$%)sywfMC3_$fl}@wDnbTop8 z%VruVMq*f^qKw|n_dMlTTQh=7J}oINP30|enKPgwu?Q1!aH>Ed$T)KqIP>mryYVtb z9C1`m<8L%gD(E@*`StM?$n|0IQOP`mLNendO!UfpKE1H8FcZ@YcDEMmwl?OrPP-eC zRQoo>Up$WY2sns!HS#oaAT=Gar$jLc_4tl%cyV7agwX1@5tjuh79FbH`BO^Y; zanseR=*UHUuNhe2g#O+=JsmzFazv0*k18 ztb(@xLiIoH^7m}{pN~{~h1uQnrO|7ZCrJGs@;bcnY z)vV4=3e1CNR8`(D0qe8!!%N=zX!Eodd*{zYck}D%I{zB-uDpB{VQ&%eP1r5&A&4Gy zYqqhLlX$rkV38)& zA@pl29vg@vB;8Nw>l&C5Ek+l5aW!|eVtK;cS4msxyK5J_aaf&&6f|~Pk**ACf2rkXO2ERm`|Qi%wZfVdo?Ev|3KA5N*GW&`(_q;Ss(vKZ zicfFcU!8W-7Z5npcGUTvK{vY6H)X=U%43uPdI}aCJxp>HaJ`8jy!H@^wqKnfRU*ah zA)!z)HT?=*b7w5gK2&vJWl~IVdw&>mgrPAn1I^d*8cfm3-2X8Kru9b`C(4X=f#?9Y zqGf9nOUp%A6InGgenXl40qAF7v}a)$FqfX2djrx74q)B|bRq1P&|0PmxqK1wv`wfD zV|!C}4WhHn!NJmU#E5UOZ{zC^4`ZsFTigEbVN_=udpBn*T{(lS3w*Y^1*YRuW1MW7%e7}ohyur zjpa6Y^A*xz1Az#>{$M!~Y96LfaA3d;+Z8yf>KaOQM1s~v(RqlZV*to5_xacGVtWi6 zU@=J8O=T#S-2dto3=$X*obH2npy@`QDe04H2FM%ll<4Gvd{fhws^5`DITs{ z(Zy`8zY7(i7EXv9>H?S9v-Rp=1zq|plD>Bre|&z)!Ex?IXWctaAYfD}0116&Jp6Kt zrRAi)>j&o)8miw&x6V^x!4AWfO|E6~d(4RGYf(YmF5q9imVSPG?|4t+7FiW6KybCn z?ZV6lUQ_6ffK87iAfr5A!YPmhl1w27xp#r!xg%ozi4KMVTt|y8Km5oT^ir=da)5^{ zF2M!930@EmHrsnELR0AZsu#ZgID%S`&B1xj_BT!du`Y>Ai1K%iAO(9FK5w{Ovxv#a z9^1ir8NY`jzv=R2%vp&gnZPDY3xYGlr=GA2;Do$FMo!G~=mLIWgcQRzS;Cd1`x^>s zj(4!uT$v}wRZbAo)d^zv8@EdMAatxj z70vt|X7BNDh#ikQy{QWy>Pv)^s-Z!UTvl2kctn{xQ$mLe!s&LLGM+v20S)pq9$QLO=JS^$$WNx$i-|$yAXa;h~yI{gAkr>AM z_&1h&u?E1Ct*{)`B}0G9T^9%`SY$KeXI}q3*tE|tTI^jIYL7`iIcCx&?;4c3NOvm_ z{}LrL1L?|u=I`K@KFutO&#zCl zoa{F>GPA-J?9Gvol!tQOT?J*RODUNT4UFLfpG4i zB^7YtX5rU6yLb`^-(FY=O2ZbopttauK~i!uU1lU$i~`l&Avld*#ep%&;m_x0O==RPwKbKNf9>a1qKZiK9n2c2A(h!Zps_Weg$P~G#D=BXF zy?s{=v?(YBL9`eo5SNSOr7YhO^n8JGmxEXveu)6}5-@KW`*PRb7JvDP;J|!*1!mns zIL+TqjYDa$_2ifQmog(f51P^4VpQ4_D&ARtstyrR9Ka8aRVM%(1;h+&B1$P+e@nn==My@#q1L2(Sia zxNsfj{y!4%fjez}0KiBAI3Z!pXGc!+f0#OHy$CvINJ5E6Slmb1o}(yvOjTUr$#Ba` z5*{EZ{R-LN4M{Btv$E67%5pC{VRF)Yx_ZGUs+KpbTNS!7U)wxwdF#>)eC=`9 z()SYF?u>7;s3#ALX(?J;yT44MBuk#>B=%(Ohb!=`+K}*#N$Pb$iESvX0(=XLx1{H% zbqZ0@a_EmR1_uLB%nG>Z>0rtg*zx6P==nVB(U!;-6&Y*dS{Fv#3=WLGA3%8y#nuko z#TYa(=jn5NFjlwd(RpoEJtie31++-I;(9Fx^pjgTar|I5)ar3wNzU~WOo*b{qcPDi zk~a3nj3#@DjYc{^G^`%y>VG~aKS|X}XZGYLqF49MLY853*mq^ge!ss_)bRGEGC2-r zridLqsS2Ol-NlKy0%*Pj{0C-F^G&L4)1nW)rDiX83{IFA2Q*}9lTq?&RnwpvKD>|p zT&n+eKb!T<5zqbpj9b$y7K0mKe(F)#%A_~kvk=IzI}2zeDzLsm>@Q*f)g<1xCQMMR zBGSTt6Ypd%k#Z5tfnhv=tkbS9Z9yhOEdvy3*#O6##c0e09lNypLQudP09U?;H^M#| zi3WQ2;(c@?Vx$K`f^M&mk1(|A(=cBGnE(eC>UJtJGC^2u6Pc6|*E)c(n^Hjm zapJ3e>5qR^_~t`VfC$OqR0NzTj!q^g1=^E#z`@<+VDtalr2nt4Ee(@DF(D!0&!-Hb z#bY?gCjqDT1w7_9h+3}B1!abebZxiHwv|B|*EcQ+UucT{{2ZOp8ZIlFIZ-DJP$O}g z*kt~6aL@-e*y{FO!ZR2uwb@yUcer|PmN50+X}?2rjnklr%VaP3@Wt(F=Wf{Y67k#Q zUe9>peIR>^j=CZWF>biOq1sbsqbCse{yslhJM8Qyw@LKr0ote7vmy$)sHh8<4>nePm$_BmS7b9UOW=fJ z@hkYAL~G%45^Krc$;(iQ$_`N5xIjpYqye61gCH2di4;DZ#o%L_KP#QHkvRrelhf<= zy@QlZZ&a2SPTLJkrLWJv=WJcOdbOdy`tfq_8fXsWjE%p5I@9Zm5j7p18>o=d3W|zc zCo2}<#f~624Fp&SyaWNc+)KL247cA6>df`1SitV-+0V+h?LN5UeM8RT6c;b^DUrw6Y(g2+(Ih`(LJUmWHy z7V{UMK_C=p$Nr*)dh{150A&DwyV|v)%>Uh>TQ@`OA=2DHV8vS%HvGS5(BEdATqO!Y zb>If%S^n+>u$&?~YxrvauQ$VMBcRRS=Kt@T`S(5H=0qX^0fA^^H1x$CZOqPdIWOt& zmZ_?A^|&en;9xohHWEv(Pl4+L+&&X!gmy1^d|WWT2dRwV{=hF9~$AlC3@iL^X<~;ul z)kT(=pvC7`DZhV1MX#%|5Yqch_6$;zozAEh|7pFVp`~S;p{WAzsOQEAIyWt?&RTWn zdxxS+UjLr;VE?zD`h)hHF0it#&9F@`+%lvm`DqV#;j{4aCl8kaJ;o+M9gVvmookra3Y zlgZJFUU>|kX*w5+uJA2`a3G*iX=&J}40f^i-4=jH+*8%o)>ctLQU@s+ z+Je}mrKP#KHOUf1>HiB{Lyp9MfpGr~DEoiK@BSOOhF~lI1;Qam`!Cq{UoY`5un##A z{xy#A-|)MC19$M4e?97dVuKuxI90F@-a?)c!5Q1-5Nwz@f7v5}GeHV@LiVT=W#lI0 z_y52Dx7>G*Wp;Ao7QT(EwU@W2jfD&Ho13L0KEEJ6A3gGm2tS{Y(0^{d$@lMDB_(dn*huVnjTt3k)oc3Xi1b-voyPB7gqF zE%n!0{rXH$mIiV1Z zkL?oR?R}v7PjJ6dP^+kn*JUccZ5oE9y)rla!vR~SL3}s)i^AvQ)1}-!wAMTpS*O>W zvR0p)ws*|mS;ZGy^}oOQ>Sme2uH^dHt@onCZ;m4VvCSFj4$G>^`0kN!ck;qlL3sJx z{tW2zL-04cnBHnEH;ZR!2j^vh|1L zY_#XPucx~|&zv6lZb6eR4c%&YK%Y<#mZqmy_ zS*tder}LAoRs#!?ceC~rbN)E;L>yFN9t0c)9JR~D?;^Do4S|J_J&A_6T zF*$gtID|ti^QqfzT#k4815?$amqc`1?sWRWgG$fOK7|wV%-dy4i{))>C`=Lw=@M1C zuQC{`cu{Cg@q28RY;6&X%t9rqt3%8ueBE$9M#*X6s;d&HYbo$o$m>QhXR#el)oeRD z?%XDc3n;HHz!f`mq}Cew(C<|DY(T*|F}~b0c5%$fa`Tz*Zvsnv}G>da6( zcP%T%X+JZ3N{eJk&5r`(Ll|r#0;4UaqO$Rll^ZGaNl~7cf;IlGv3MNvdYqTC!--tP&%=~)C zs#G97^g&vCu%B{rph>b<=U~mvDc^UZY`Upl2B|-nB)Y_W|2**hQzvRkGufrSd9|^x zx(iG6yh$}^E@iMbGPcXp`t1q&DhpBHAyJ?6J95d2Wtn%B2UotF(ipgK`dM?DzBgO= zY3basm~F|zOnfyJTUgTM)x(-whdj}*e0vY>ykY%zV;P@9^}0a`ok7WUgIBtG-Sui= z)SN0WF4Wl1mR@j*+Lljxt9p6x{k`l|-4~mu;uY63)eR?p&X!L!&pkhXe)n7!tFomx z_QKd?WHKw0s#=cClw|&r;(LC5th8Q|!h|fuiB^xr<2*Fh zqiPSxDm~aQcyK9vC-WD`xi9y%`#_ZZvEOULV$vn7LvOv1=i1wsXOcYpJyFRyL#t2N zw>8ulqcOGcm*2kZK9gorlg3h|rb-?tDn^Cm$B!lrE-YuIc?t0DSBJhA{q{ityZexU ziyYeG@n%oY8%p_*dfsdU@n; z^(5R5>WIy|cb}E4C4A9jbKm`<0?7O(I_R*T3qmi}s8ZvoP+c60B8Wzu_?rz0l$T2U_K5f%;JGHdpGy9#A2X`9H zUalP9p{VSb`#$B^zd>C=S@-Pih0Ok+I-bRG_Zv60q@pG-*Eh^7ZJ1@DdS{fRydU}1 zka%J_oi-MI#;&`17<9?*MU96#=VWOAH>L6LW0#JL8?3)R5xSEc$Vx;LNz~cPRlfbM zTYN~O*0<$NA@$?NL*=gcdX4WK5r^5XyAp(#$SUXVU--Rr_or1cgbtIg7I1-aEOZ>y{|SQW-!ml8iYC2~`2UvHnS(N(rmR^00B zKD*++ewJpRb&%;4AzfK3qej@)fg|i4IZebbyn-_2y27%}F3TV35%Cz$gL!9TZgELo zbvuG|Nta+z;rp{kyV&MzN5a>&R7JC%yl7U8LI*J4RQBU;N^vcbm z=j>WytXkr%T9nnyVFx9mhoV|>c99hfsw~p1B-fZ?M^ctXmX`$Fe|V(hDx2qaS#y)| ztKOmZwEfXt_B+|gLG(zBM&hF7mzT6DSv?d?s|C3OPf?iswYs^Ic&mbkNuxWT{fg)x z^Dl?#sSHr61%CKal5kkB^H3}OffmQ7;pyl;FUKjdrv!?!P6C5Y0;-bbICcz9RRceD zNb}{xB^7eX@&`MHJB_Fa>KMXlBiV;nO@~$)mm;X1U^sU*+V?cv>wD~%*v5FsA|rpy z;&sOI_N0o1E&wn8{i9tL^%fhN0hiehZUg7AE+1ZVCGKd>_UIu}Yn4{DlsuN63O!N^ zi^4Dqo5?~P3mg2Dl;n~R#&2qk^T>S)e`F-&NqsO5d~Mv{Mtyn7QTS#1n_<)P&*2k} zicTy6Ln`@?%o^)C2+;_HjL`8o0YRZhPvLK z!=A-dEWQt25*}^s;T>z`esgWU9N_kGVBc)rUw+<_VVdqq%2zs`6~-NVjvbZmPn9DP zFA9k|dvs`d<-7e3jd}cqjLzJvPKIx_e|i~xag>7Sx}@l@nf-F9sE4NxLfbSyeaU|o z>oPwO6u%%L{^-Rd_I>$ND@%qcGBTR3TCBD!zj6$hMXd7o+ZH5wN>n2bCZ#+o-P--H zeo2>P&>He5TbdB|R&P~J$~~ptFJxX<*Fp`?<+q_`b2FIlk;FNeYJAPjA?QZ`wx%l= zj@BoPOeRv%BUDL(e?$*yWN=jR`BUk7h%Mxk)B&_*G*q@3R^D_|5Z!G7l@&+W&bgYKBU(lu8a$2AHp)bcO9^Z3P{S~elAF_G>i zk~)lryEI>(E33Y^&URFxByX5pzF?GOic@ptjquyVKkq3|-`~MHa4bDD&cvv@*3eHp zYeC?~HXO-VJL=a+W=v30IOE

V3_)UPB}yCbIl3p5r=CtMFqcT165<_0UgGnC41Y zM)0abm>#BCe{P%%6ee{GIo=RCUS}sILcPnwz0SqjJuN@^Dd>9D95qkVP3PvDdY(=N zlWeosu~LtY2MoKjqpBpx3Qa#q;PG6XLtEF=HefC`rbjmjQtaE_nDA^roqFZ}_UZo0 zt1VdpgZluhHO>X??QgT}TbZl;V$K1iEzUp{_cYoGKXk(hJ-U{Zu^XW`qPNQac7^;A~61eR+Fh1V0J+<)%`MW&bKVQCTC z7NT{Nl7sjpE-N;@HQ1G}BOaw^T#ZU(NK~g6<4VOA@s`NSZ8=oQOYy+SWUsg+7{_jt zKzq9-n17@d(f|@jpi2I+M;%0q)QipPz6y$Y6PSUh=?Ev2uOz@z4w+-rArl*-a&c` zy@nck6)Bvye_m1)2F}{@_Yb4pp-fQnQ=X_>oR?2t*U&T#%B|qCY20(c zs$&cxd}SNyRVI9;N{LA?^Jn7w+Q#=bht#+$OTxKABmhjAnY0B5XD?v&t8JvQcO$&tfb^YFkczz|0ir{9T$6@6px! z7RtaI=aX?QfHRwil}N`SWSA3_y(?#XAJNd03tkoei6hST zG-ip44d0a$@c8pOL`9eK$T}kf#?I7T-xcPN$KJw5Q}sN#Qp5`S9Peif9Y-1A9+kXcUTnY{P4*C zO@DR($!g$qC=neqcMPMZ{8Rt!waE{8Lvw{k-*sPX!#QVE%TB(3@chC>ygc-~=@aFT zovP@a%BOtGUF{bQPeTtMe|=W>^g4(Gu+{jIZ#^eyJ%9d}EaZV*f&hMk&U!)VC3an( z${dmM99m_L`1>doz**7`_(nb8n9>0Nc#Fa1C}C@$`nS<|u8x;QpVPUtvcH$UF)R`> ze5eX3f9KRn#I{wU@p7s##%8j$C+cY*{R4y*?}UxkLY9gCk5yU6XYPS##=oAhQP45) z3O;jrICA=FF~GYqKCHxEru8qGs$U)shLS~V#=p;-5g|jqJ6n<7_^Q(4)V7l|NHywbPzeIw;5Eq2~lP98?m2%kLF`eWUJcQ;wdp#S5$7Dq(4qFx*Er za^Au9+8er4=7=XL!m`n`xUjoj0^rYzui-??1&oasC{MiN3aqmbz&lC>TA^-S7n1O| zFDWlW5(OaMr7E|n2v6t@>zU`nR;Ff?z6Q0C2V_vp{w z5f^FiPyyf5*Vlx{VaulpB098s+^{oa6Yxr!5zTYfwC@+F*@)%%t{l$EIS|#{U+QG?17c}q4Ii6-1fvE-zAeW)y)twz{2ZR7fu-tML2(&_%|Ma= zK8!irv?Qwdlc9`3{>b|&QwW1#IpsjF?X$PxT_#t1CLxTbK4c%Ge-lQ{^hSyH@?}Re zIR9XXeVFmtMD@wY@n_>9nTP$4vcWN{3=;7_T&Nxvzbb%5((eZ|Rx3YK<`Bu`)lB2) zh7{HN8&V9$EmtpwoK?4TvjFm`6*wnj(n7V1#0Z9pJBNw`++Rqeij9}4I~}bjOQPK9 zr423rIOD%DXs>B%H;Tw23Wu6Y7(I^mjQ(6=zT(z6`8{dU95!JcMA$lO^w37wW!<6ZG~67D;2)vx{&Muivc%du;PCDp+spSj&;3~s5>+&HBS(4ulCK`S<( zJt1WSNV7V$=M41DQ&cFNqg6;7(z3fj`;F>3^zv;KqHmd-@6Ijg72_>QjGBiB+34_{ zzD8xn05U4BOZZjW;#ZC{eBM?T%C@kIVA9dmpmyYN&AP;DmvA=jW8$O|nC5XL)l~nd zxK})0dDB(vElEluS0Z>UY_?h2%CoJuk*`{-ArRiKFH{|noxVmZEDm{!Vpvqh03a|s z34P@6-7o#J>YF-|Av1T}SFlmgWs=-yPD0y8-_JFvWJCz*xDj?<2rAW+RoYzniX^dF3bO_v$YuN$x6^@~1%jXY?41(U}{iD}h- z*l_dpK{b6_aeDxdcn(O70lu~dvVwO$ceVZr`sNn<1%wMKiv$WGKmW~C ze*5q0^xxGAi^Vs>=&yyl^zY*F|@rb-r~Zgc*ij6>1)6@6a^F4reg%rYJ;duzMXcqN4(K5Jo5*X!Mty3CqG#lBi-yp zVJSU5%jxL?0@rH49>PYp%ziP~S5;N1q&0;N*sw3xns2aq`eM2*i}6A0#l+!v*R>E-&6TeP`*ZhTXI(R|lii+&p5TIsG%o<_dogaY6GzQR&|jy9hvx zZFLVo{6s)za5{(jtD7sH&IpTTM9+%ZNl4-&at8WKRQMhb+o5QE)yc8tVC@q{TJbql z@i=xgy@QUv&I_Nad`}^4AhJ9JVarwqtF!9FmrE|KODGFK#F89Mtg=Gp330v-SXB#=Wx<`ZAd$Q`i-b5dc{EsV+v}MZ2WZSg>A) zBXFQLJ`_BlY1YLZBbmN_xSop!t)2!INRJGPdMC=!nhyXh$%ijdf?G<)8xoVHa+J5s zTTm#^==emiK|CS>9-+cZ1Z_vHPU`P2ecq_01G|-zk$b`iyP&WO6t-b-0I^CuARti` zh7hW$`8Faum~p-MEL6H2;uMc44?ye&UGfAVM#va`n$~w9j1tbclIEeWRvwA(;m za@3c(c^f^g%jcwkf#4_wn?iO6ZBhQAOipH8P^j6ex@afs+LBSK4FZ2ee-2F1eKOv7n9kfB9FGm>Exl ze!i#~F+Pw5fCEf(M2W2tK&#D)|5(dt*B6mWHqIkaB*L!nd6O$hyd z0~7s0;8Uxqscb%|4eq|s0!8m|phJXDQTxGZvUGo3ruIXGL$!tZomG?DHpPJ7 ze++%t8hQ=0cCva%)DoPRSJWMxn}3C)_K35i>*04pV<7W@LSkKQZvE>60(m2FaPjfA z0q@C+LB(T5#rS73-#9bfK-shJEJG!U_sCOD8-{ONET~0oI&BO!XIqE9P~E0SdB2N| zUzn6c-fr@YKTaei!Y-|kh=L!x1drG|fb*D#0DgvAMohbHS)=Un2tJK`vqI*UNi=1H zk0<4yH5}+#%HDdkdE%Us@BAv7|7J>}T7X$I@ruRv7S&{{PW^-I_K*H>EOLRy0HdNfe zjET5Ba_k8W=*q*hSa-9)B2nV_0+0-H-cmj>u&cMl0-1)AI)Dw3Rp!d4!`xmpsFe=s z5h(X*gM}1bu~`V|7#Tx;RHve__|VN$3H@prcu05bue{jDY%kGKj+&fUKv>I%^@)f# zwtd2o%LAR<8~^^dt*@TBxL63}`7g8CTd$kf1*ud2l*9_g*~wu+d1fId zB1wo#FSJ3*WzaxM?muEs8wJSGSHCRL`jO=1A9kB9t5j+hN4I$kRW$+8%lWD;h6(fr zdZLvT?S4FIf62`h8wQYZ`Ux{*RefasSC&2d(GI2JDXEWO=-ERWnzoWp=>>y}`lLR4p#s^?}FhRyIZyV5Nq!?y2yivGIU))_B0jti3Af z402TXfjtWW%0TNjPa)mA@4>56sW%rpA(j5fT3Kmxh7bfTk(_6uAN8H5GEbxIZCRqV zoBuMkTFG5@f%`-?ttG!`JFC~kqeWYEC?KoY)ida_xe|VtRh?^U?rJ5wET5k!QB`?R z;qKxHmcR$sUTcD~(`754@%&colp~ZH5)!}4svokbTM=8)s25S1@|Hknyt?W=R&b}c6Lj;~>819&#_tqcoBdgcbOO_Xb z3k*t58`99S^FXBTp-uh360hkb_BiZKeOt|4O`V-${mQSibMz@diNwwywze8G9!Esz zTKp*e-D2gPcHKs$oN*p;a)hhia3{ThA^ff4AV2*dKE%2zDREaFB zC}{PTV-FokrD0+QZIR;TGKE)F@4~{m@xV&Q6J5el8cRZ&Iw1rZH4{gnq-{2ZtG5k? zOx5F}Blfa>qc4ACv{aWX_5@%ELZB-RIig3UG*0q!`Nz%XPFQyLCH10}(yA>1n9WTH z+=sxr+gTuX6IdHh>eR2VIiO(%8&s8CZCQK_t#&t~0LT6)vG-h!EPFLwfoZiQZu@fI zj@^NmZksoiafJ-uHzI$gh|eMJglFi(L<+72Lk4uBW?0Fkjh{-~OyGC&Gl@I_lR+4H z+t-Olgdh$XgB!p6v&-bP%Ec2SxiA}BU$av^Eh*u!`XF?$)-j{Y!{z3?P3iO9`xk49 z2b-647j74sE|oT~YbIY}K_AwkccCT~-GN#M@5w{WFLKkz%+K?-TeSk z`@_|9y+Lo2`b%5#?hfMDBI3k^kAT=AgaEO>r)NzoS<>p;!uNh@xhH5rI6%FPigU0vOS)Os<9RQ`R} zwV4&fV93(CcVgXRV=7S8*uJx;KY-gx?=ju`;A4~=8;udzEDW}3g1YIm*KfC=EEN1Q zQjUBv`G8*!M|`62%b9B@a22G(V?+y%B{sVJdF+|r$wR;Akere(j1StGzE z;=|QPf{Pu)sl*4xSbWJ6c3O(H^J!7nlbaz$d?94V<$sJ%quWdTjL2FghzAGpWTNLz zHLKj7UMR#%78O7oV>_fbPNpvv%2YPovK5Hk=yv`HN*$YcEL?Ju3L8CxEyhW$YPa8K zRSzmM*Ti4KkrwPOm9*Pm)3*KofauiSH%o*f#zA0&PU*jdY}7iwD4R{UjEs#xP0H`K-HrY!wUaGu1Z)31@@7vR0`k zND>)mNC=KyjqqabQ2Z*62jyy@)`!j}CSP&PS%7_c-zNuV$TED) z+cO{r6~xcn?ok+mJg%kR!2e)ap>(76#LU2b;DZigDcNOe=6;N3*C#ZDwUKl~75^q?7#@-3>6t#7#CLINAcnmAM2ny=H9n(CN znnR^&NVmR0Y1~j4-pulX`Nl=jjYcwLp!K-Bk6x!b*)t*ayF)VEZRik2ZWvKN8N-sB z=CAuO4EF|50VYASPYs z^|E)TAs!LB$!_!KE)I~nBr=#b=7Q3+_-%A|aQ#*NWX*dKxL=Tsf2Tgb7~*iAK(eyb z*;0kRb;N0|&*l(;VQF+9ukFQbss(w>)Do`*L;di+0kj}SU}@2iN{u6I{o_@*&G{El z=wCJ6-elk?K&XTq#!9@AJMc%eNoC+U4lh6yo@P{F=SU;Wk07gww!e>pT%r_;N=RR8 zfd06Ja&t1#KWAwz=cNh)ye!Ul{vo160rC=f z-?R9{Ipi{{!4mqKD>Kn&wX3u2& zm%BAUEXY6YM!~NsU{B&x`d+15RgkNH1Q*{3mh=SQ$BUcikL+x8B4J4biZ2ME9l?#& zFB+G>KvXTBg@(AwDP|R+e@f=hMjcqsD}$*W*U|m}Zz+K*@qkb9_f3IM<6L~ttZV%Z zAu}lOGpZNN^;-D4x!$e1@xlY%`c>IvlCYnlNm4_5>M~n4g3|c|Qad)RY>>5iO#KpR zr}N-yp+8?P2g`cq!$5z>g#^6whHX%3j%azLJy z>BVUmSQ3`YGi>^#doCQC<;&AKb^SweyFf=$*T%O`a6w#!G9VUkiwzH}!_Tq5L!pR3 zRtc;7IvChZ3lhKsbp!QhRvyOu{Yym%x;88FMZ$hI^1*ey>4i^3`*~EcJGgRRd0`t0 zSSP1Z%r1VW1)-mq8(lD^KD`iwjt-VF9y|c)i+3G2c^`O)=Jbitg0&!nujmI;&Kex% zYB~@QD}w93Og7;%`gNgXw-;@AJ#BfPKcgZg4h8 zMCwRE*wv(>U?ngO6p<7FJ?2*;c{JoRp+7Jz1XF_#c@`(plNSBO*V&?0``z&?#J9{f zDs-Z-QRgVP(4YK%EgHRygmtDw4!eBf@M$Y~M*&iD`IV>_iMg0qtEX@jSADXkN4$gJ zdZqpv9I4P;X-jkKgVY5P;W@kE>Ta3ArVPGFjWYHu+zq^~Cgs}d@y`EA!ppNrSE#>L zjus~YF3p7G&0%2fc2SmA1Z@%Z?IWS?ayByjO|X>SS9NG z1%I_!2n9SwTKCz;^c*=!gEgW_h*<{IPZ1m4xFE+b z1kT;9UJUr(5+DDdI~#5q;iwLPBTHY5?h62LbIIyk~9uix;B z&Kaoxb+G(gd_dF_RE#KVlyCf68ahL3G4YfTdeKoQ2;dmhk(b!!-QxAaD~5H7eOecr z{cdqI*Wtaf{&1I6oM zLos#j?TJ7td==VKkav^`N-9E4aD0mJY9K752SybII=xvr^)zNo7ACvb^R@hgF$NTt zvo2Mx&(cS$)5{58avc@FNwnc+`dU6%A~w5(%CKn2W)d1PVoS)QEyN0D5-24H$!;gS zEJ0P#q)|5IAwFWRLj>G?R#V;x?+L_5l4Qt3>G!vm*_ig`w?p?pI-AR_t5)gQ zb(^j=4?B^^=$}~e0EOp!4clVDn}sUMj6=T&I|nV&5qx;+S_!1&Pkt&+1!>)`YdeBh zP6r=2e1I&J&uu^K-8g!~EpUhr=8I3XT8IZQgX+GabQg@^BWJ%yPTP^U_M3i_2bVjf#mqI_YJ84^gi%4034F(d1O`bmxUsOMV%@SW=p1FpSQeW9^=>?B z?=H=?;~H{#SmE>v7|7JbD3+F3`2yDP@f*e{IM?DS1(X5@&(#bqN7}p?GDlTsJlEPU zku3VaHJD_#JiKAqI2 z6RDgIVKn67Os3yhHlwpxW_z;NG^78HaYmnU#`cG8-GXgpVB@;Ejgx7_zW4C1^5@Ey zp&*)W@RH~9Y0>Wxo!*Y&8B_=&%KbV86;5yYd@WNYWR6z|$y>hztlf6pn1pMB1n0>Z z9xZ5^yT{m^C_8jBS>T074t{j5hyHvYccK+6UTRurWSO{A4EJ_Z+GK>Y4uW>gyyZlO!7+hjAojaciOtxH3M+LzS23L-=7bv zpNJM^&~3|iz{9F&%=3~O^1|jLod;LU4i29&5E^Rr1=397W)NHvK zp1K+z&V+(;NQVE!*bfiC?04Qv=ns30SG9~1(`QMysP`nq2AX@xWDg)~Br{uH3;DcD zt!6qu_T)uT7Q%3s&0>F9fji+LrjSIQ>56REEE^03+SmLF9Aa{JMkz3V-Tc^dUT14; zvIuN0=^xbFDj2Qa(53))lfCjYOwPy0TRAmSImJ*row+aaho{YqkxYv)3>HGVb;MI2 zuK&zb3*sIs^Sz{+bf(Xiq?=6=MRZtMw>tcqF{9B&8EGCuT2Bl@&doVWAYty^6)R_O z8;>ubWm|Xh!i2HdXKD|V$#CD#rKarRKwGRhOnIemxBL8GKbkqk&U*YawNVMBU}qk^ ze{4+bk9dO3?FSAMCI>nu`6oB$?Cs2wfhVJX{EVIEN`Wn-!e80E;TmKpgx!?I{q)O! zcy5fFB6ERJK&?028b1LF(PAcAubN%)0}#6Li1yC;Mt%)P@a^i?JLL#xAs5tDes%Qg z1ZA?d+J1&WL`$8q`Q`n9GguuXj|^-l;*|uv+MbiS5q2tyjy1o44GBD3|4}2p&}j*iR-)85n%HBrZu+ zJB2!sZ_POyTZI}M9E(}Jc))pt^D?3@VHp!zKii1jmnn6u#82qgcj|3ntm7Q~T?S?A zjMPiI_mQYo^Fa*PNU!Eda~Yx@4Rt$vi@Itk3Lt>qD7D+zTkW*u|71al85ZrN$o-PF zNO(ieP|677)wE_m%AJ37-Z_jv_+bT4-a0+=sh3UHi4cA$qD&L{D&s=98RlBnjWOwF z$ZEY$Wg~>yyT!+}@Xykim)E9F1tH?yWyW4t@4m7x#nA$p{sO@_m)RQXS2maN9{Out zl=~+xf4QvhAA$pHryUXP#PfoJ8`Zaa3?Cl`pqfg!)6N&$zAVmuI>?1Fp~TW+r@HJ< zVi%_?Lcc0vpyl(N&rz%T(j$7Q0RWMLz?p>{7th_#?w%~*{a;}tFP&LG&El*R0k*PG zO;%r&o0CO@9+}&*fYU#l9Sh&iy@r(oI2sg_waS_>D=!*u(XtoK8=kX^M|^bgEr__( z%|EZfNV?Km$8TjLLI#j*!A?sDNFEDpB!R<)AX(6JxV)d(H1wsr1HAO9xz;E;xiUsU zi2^Julzdo26^c6W#r7@uH8bNdQ4P+MX2#$$WbGN<<~?@zd%qYaq18}vt^N&CXhmI4 zxvcreeDf**DQy)Bdn?Mz@YwiV=*FtTe~$5DVKJN~UMmQWUio@^`UI(qs0lbm!4kZ?o{zP3%aNoT- zF>$G$ZmNy%F#AmyEHznYK~TVLp%5L47-n_rJ&7AV`H=3q(y=lJ0AxzxC zbqoqC3bs`idE@vgoyjqj^c$cW-g6uX&|d#~m*zOqSbiDBm{H4R`x{3;UcsCpW1b53 z=c;w=u7b0L9QIBB;|LAl&c@ElY~%QhGqZ*CPwM&7U~JKfkB#>N86T;=tBt-#ZxrgNT zZ_`t$xCe{djp73rZ($7>=JNy$IjpEuC~s8Y4+k@tL}-FAP?0!U&ED?O^Xx_)>4V%G z1D%~+ADa1X(9j$KkY71z_HUP+*qD$YWH3vVQ&B1|)>AvhUH`>h%>;fPtILeN@eMe^ z)CGktm*hH+ZIsx1tK@bWrr$rtxc<12?roaT*^3O+sk7ZF7 z@jAW4X1&CD7I8Bwdb?YwkAGKEj%JKZFrZNTrD%CSU0UYai?fm|ug>tef?82_K^==4 zQ2+^E@(8befrp#n`6G>~7w+*(n&5Px>S%Rp4vzE_VJgEau6MS<0NY@Xl3~-A4u97` zMk{Vpnhh8~>`;sRlq-2Npmnd5v}D4d{7;ak<)Pp-y}_^f^2Q>ZbWrm^qCoxo%kt($ z_??#$PotmfYf#CgVbPd=(HN*~a>0S)ASrW&2IBKa^#Iex?tOLeHPZu$9eDJ2a~JKv zr=`-&iL_*Hf zCULcq{utM@3D_cbT%y;mT>F&-xL1d9r($ut?k#MKo39~cMr8-FepLPz<=I@Ho0)ER zaVfth@F{*NTbu(58IIT^-M|4#wGqZABGTYYAqYuTwS`v#4F-ZD73^#nFBisKO)|`H zBB)YQqIsym@xIdK4S2s>&k;8?advFqPcaToXO$D;d1v?R1pv^TUvQ;gF1yIY%o!32 zq2okzyaGEfAN>+o(8<%Mo6GeU)|4enysWhOoklfdJkt= zChNBpOz!f>oSm?0tFYq!mn9SlW?GT}h6*;DKYC4(#&eqM9@p36cl z;>1gl?b+x%n^>unD+IYGeWn8~{F!lGK?$_*m)B>ehj6Xuh#po?~X69Bepc84| z=tTCzktT}ts-5^Iw`^*-nWrzVcR$I6w5nICO2n^Z;Jxx)t@;`^O1GDFWp? zKu;U=#CdcS{3FVOT#VO6Z@>#qe52@bd-8V|fg>_kFS4bdw6VlOqAO_w+)|0(@@^|@ zaB{EpOfduiA*bec0>niwqB406x z-ara0jaD@@TezV*jHuy<)l6?>1hYnqzZ->e{p`wA^Tuzg)NrKQh}?S4i2O z=Lg%Bqv?4*EY91?++%R6Eq9wN%CpW5)bNdb{dm z{2z$1(MrvyfMqsqh*8@9N2MpJnV}d7_&U}}_o zm@0S8FASE3POuMYO!TR7+&=5Nz$HHx-v>qYzF^8ST3i#=N_uQs{HE2dyE=B^? zJ<>~hvvm|lPM*Qr$Et@W+T&!V0xbtZD1H2(_o+Z$pk63`G^fA=waXFvi!S{qqkI}P z%pf`9T%T&*i?|&6lFu9QK(dD-K7Rs;e;2w-2b~4E)8ZZGGESLsChcbGoN-?+ zxI6}lY$xAaW`#%@J*ye4M>nn1)B{J}k`j~nSQMI%qh?pdIZVPOkYZ4l=7B>jSOb#~ z?zxb+ll%1Y%?IaBll0Amsu(}^jkg%#mVnNd1{C!XylDxqOAeRYfB0P!`qZ2c%vvax z`s>nPI`kkR!0CfEZ{~oFiuK>g0N9TRlFz0PgJWYQ_l*VM$Cd(hISnn-8X+4m#*`tiMiyaL*mWl_|^cbim7JBX$ z_eSapYU5))*(iv#^okf#?D{&0y;)Se;aOl^LymphN%E88YtNs0yGvYrOq?;4@=WUH ze)`sA0-P7?tkJKEB!TrApUp7}{GS&L#`*!-dK#>0mE8aWpfiN4OBf()&R%%5Kss)W(Yz=JpxaSP#dOBCn5YoNmrsE9_G`3EF5I+&OYmj+V-_`S%hNz7Cu6%;US~m>}LG74frsj#`)u?brU=%Ia!mhzaeLlufbPd_R z!Wf0fK$1~&M9}n77$*2i_zS2F2NTjl1FGRrF*4^v8P~73eAU_xNoMeN9UiVYSMUX) zFqU{)-`maQ$<8+?#W3GQ)NH3-&iuN;B~imj#HaWI)GQOqlP9)Q@jVxG|C^S4Ws{L7 z%*+0R!1S?`$xOU6(^tX^x61d&xh>_85K*gh;E?PW`fvK944}4!0b+U%m^%dsiBYEl z;USD|4d;&-Wp7A3xk8?o2&{+ILf&3=#^ZA` z``j?l-T)HP9#IB!XMi=N*vKGZFEKX{s-A!y`x`Rl)lcoW5dd*rcCWpB&f6 zU;BO7v@84&_{O$j4aJ#mezx5(V`ON!AxnSC2%D}#dOU=dwfw5OXWd|$o{Q;3>a1*x z0wrNnQgesQ#k5f^FZpw;#06)hZxi%IB`@A=x+p9Fpou(%~J*|PiteOG}fFm`!@4mR;fzTkD6Y2PULQ@Tyh!0%Qw6b#8`eM9RN1X-;^Un^? zV6WPu8Js$+B2zg>CDxZB(YeX*FF$6>f;v`$VRYdAgqaX-m@6_^i_2psLZ^60|G1M7 z9_Byuq*y|GV`G1d2)aL$az8jhf{EmtkT$jza)mJlw&1 z1I-?C%WM`vw`@KcA|Mfzv>j<&9nC`0K`ciEwFm%sJhZLvz-;C|%@$xb(LCJ^U|{_hh_(0}KxLFfE8;sO8GUI_kgqW<6WB=HL|qriXv z@&EUShzQP6wDaKfbm!l{)oqiJa|?o^ja60eq%<&BT)cn<6oJF78jVabVK+y3*QTya z>Cg|dewO?aYuDV+(x5(5Xm1bc(9oYlhJQ*xBv`s2U{~hmIf@E0rpZ7F$d%S!V#Wt1 z`A0rkW4ldFb-B7TPCiC;?T2ovtdO2?D^vHgx9?bC@i|eN4gV}nToa_CDZt8vEBPp3&>@(CIP$kN0Vp4fPZh> zBySPy&b}HBxp4Pw^@50g&RonAcx*0_kx^8nQ}1;vQJfc**@ICW zVil;V>`g6?Fx{@L4+;61urrDo8Q<{I{3@{m%E^YVuh&IJbr4_WTfD2P`eki>0d!zu zxSlt5TY>y|o91BWUPyuQsrRyrF5R*eU`?a9sva*>81H}ZP7ip_-V&9qhgee!D z8AlRmch7=a(e;#sTMRi6e13F!`90v>fm~*P%oHg(u`i*A;OueL(A&(B(Y(A+wecPR z&9+ApX2)oPY&fgS$&?#=-xocru15M%Pw1%TnK|@yUymL%=oMN%qW5P^O}&0w@+|k6 z#7ArgbXj+frImleVAT-aBGv6No!T~qhU&~~erK93FG|V>!hasOcQN&|<3GE%Yh?74 zR4Xc&5nf-`km9F|W%oMog~{HUEriI7EtYHCO=itASEMe z-7iTRTwS_D{c~R@T?%2g%Iu`$?TYi$7Khg@ z^pA=6<;J_79Ey>FoaORf02^cC4W^{9m_BT7-eJN{mG1~-izxgx)nR|vL=O@u%31ELPp!IoO^dP)2R^48qk7XUpsaP{Lr!Rhu!svH1AkE-n#@inTH_ z*`kf6jU84xo0Ny}y^IvrUMu82wYMJHR?Uv-mP}n$6%(1R^x%C)TwIL6R>nop`DDsa zNYCQV4hzs<9h1w*#wk{w6W}*YSZL&Q!l^ReRmwFC}7%k9})k0P-XB-rml{ zq@bG;UCIKB&doQ7Q@C6UN~7fB>L*?rw`PYC%wsWb9@uN>F?a;wf(Tu9R#z7nsTYkD zkqmw6Ic^n!fdCP1u9DvC_T~D=8d~sBoKbVZ*J}UrJ3F}zSt@naBe8Nxkx)~!LA^PB z-+02bW9YzfHhd;BQZ}0oKnjOL-}l1k0EWV^Kj~sZyMc;ksGoAW>B}k(^h|7lV2pyX zRaZkq^{wsBlQGf2gL>ffCK?y?d0T^&A|q99 z<)3;rwLU1_4&09~f#k#eY`<8)tK{ePCM5nC9X(dTq=G2xd^3$m(hTB2W424 zg>f;$IyWx;uacM~{ydQE=R5oT8}$$OoplBu3BDG;R@b?Fq`Er@-Xp$IicLYbN4fNwX{39%0$O!Rqn|SefX6egy-Ch1oGNldt+N#*wRQ0h+$QNce;b zVfHcI-S*zGierIE5@SI=L`neAJHXa*^)}* zwLWfX8wX%fcMs05J~`4r8LWJGKZnFS0sJsE&Jt`)Q=|YbTfxp@F4eI$7o^xvo=^nP zY{UP`_0uN`TEQyCMGBJt+}Hos8;#V9k_bDf2;%4aA~EH5N-XS}SOKv>^Qcl4mD>lKI^d$PvzD9pqgFPKJMEize1e3V_;=lBTp?-xJcu=AhJoNMaE&GE; zc>=-G%H~u?PR9PB9PyU&=+xg?@MmqmXVE)K)9ZU z)!cur_L@{^yR&u@S!`!u!2W5!+26)b-&U_JHDxK^?SF8up}HdwC{f>dNeX*Bk-Z*w z{$-cBbP5>NQ!+CHeJ|UV9CMN0UG)$wtpv~mNpq&)eC$8uj9*F~UP|S0ykU3jXb_gn zw)6H*T|TXyai$;-1r3fo&1@&I9)9YrH;OP>L#OC@`$n2Lqp|Ly79%74@KR20eZBM7 z7V7wllxt4Nuz6B*wJgr)8s~VzykyMV+y|UH<`5HE1+q^{?73%F#%1 z@Ym|kQ6<^T)K5od8~G#U%c+bbO$!=0fjVxdQgeHm^hzp<3T{m49fv&p>&{klb0!Ld z%LTa_S^zhuI<|N?KcUpv($i?A?to+gbhBgK_m!4S_(s5EC^t6?{7W8$K_#6L*5@UD z=9^Tl?+7NjZeL@7pl{}Emmsis>-R`ln7Z1GN;exu+y(L?rswCEBL^oITG$^cMb2a< zO}yq;kYf7}h3|9r)2Al&Kj*8IdtYSH^%Ks)%0Idu^K$aykouvGS%**~ot%Zdz|M&d=y#1fA|L0?OP1p4Q1@I@`Bgh!k{!t_TudeX_ zBK!ZN5&VxTpy4=XB!gMcKP&ooD&D8TL?%d<-CQ9tkbk0c^xD>uaH|LXb83m|?TVEI8z-wG>#r|cTKQ_rpO%$p z_pRWD7m10jJQXQvthKQA4EFu>&t9YgHGd#hR2(>qem%rd(a`FIHz+Eo7 zI~M;?S*_KW5xh)eqn$;L_Z1eg<0c4nkdm~%9L!mD)zi8|cjffYX;s1d4Ivg5z7dPvWy-2GHm4 zQKQ4-qyHV&gvsX*S;k~hE>28uFJVeFJ^=vg8yYScqi3vZeNQ(M1AMPec8*2y{f3(s zW`@hW-)NANIE|hus}x*Qo3V{ZUXkWvve)6+n+X4Zs zs_LKqip-+kNf+ezS5TD8`QehvlMG1=GqJRklk5+>gy@(EWGdyg+*UtPq4}Foo2)Z@t^7Z^Ms2Hu%WGspkMu8w6(R>)z{pcFVhnf6ZLd;;}r5bzHaaBt&i$C z0(s9)PTmfd4w_Z#o_gNS&+EsmUGA1Z@2=bKY;0{Km$qHSyu7?Fd8(_cg;Jv!@o!Fg zdU`G|Ar~Buw2m>8larEi@~>Zu6prtF;neUkSV+-M)7S2^wLa?j)zFIZk-cUzIAM_g z=jlGQbDeffM-`Wr-i-U*VJc^jOWlr3-L19V4YpyiUZy{OOnHZb-V>=vhu+^Z-^X%N zI;n(puObY+NSB?rm!-jHOZWRryMbh~M4iaH7O#SqUZDqX@6n_8`WhN|5Eu3Fw%Zl* z#>U2Mt@8HD4i*1`<Y+J!|yDd`raTe@56?v9Og2-2y5lynJ5 zcXy+NbSX$9VW687Pzg~$;v1;Xd!BQyFMfXh;F87KYtJ?3m?Q5o@0J!R)uB^N=cNve z6w?H{^nD+9y}&l6H)?SnzdL=Do1XmdN-8&#| zuVezZSy+M`WUKYu$1gVy&r@rxjRziUd*~1e>Y%?8jlXrebbdq?-Ul%F@B1sSl00vX z1K{Gw_~FkNz6RUhw|-jpm6x5g8&km+K!Vw;L&h%cY8F5Y@)dCp{%N!Vq@hJaQLaG>-=3; z;uQ8Lv@kLv(RbHtnYvDTzkk~G0S1R7$@9}CZEbB0uFK|qQqo7)Fov*qqWq5e4tM$D z<6Qefto^S*bo7X{#WSI_DE5s<;)MpB-LLH6Fb#g%-zCn;i5t;<7XtC)!+Oh!z4-GT z+=Qm)=ER?%UG0AS3OxDUrLU();Dd6v4H6E-2n{k;Of3J8o268N7xGs+{bhS=i;j24 zAG-@{CZ;F;8m3}q)!sM zyP>GRA5P^zT^$buk$5tQ9f^v^CC0b1Mg8 zB%RP$Vq#)YP>|%=ZakbMFL6y+RJ48UMn+S7EJRHw^$a!6d>+Dsc5g29;eq4g2F{I8 z2}46e3EQ)fYq;MQ8pA6k4jY$9Nl4Ukgj~OWGaYVkZ$DqnJvYorT{o+9r0D>T4sqty zM-{BO=6#~io%cmArN!M8$9zJUo{;hcyUmx!Z;#9uZxx_fxcu7C(~E@)PL|E4 za+)FsCFeM)Z@moXxsZdm4(>PhpJYkmdVH!uz00kO&>Ixr@(@8$pEz5rto;}hb26QK zhKeEh>izQWrL$q$+OD&%^J^G}p0YWyadGhfeMybH>V1XI)1Ss0->-1`pvSsg3OqZ1 zCyM)xwW8(vIR%zDUP>rI#Z<8(Cu_vx$G-NQH&JAXG?RZ?XjD+s%Ch9oX3b0eD z=8*?8nmOPw!wzObQYAHkvie#e4KCv7g~CMNHL(XceifSrL2G0b2-SD{;J6L2mb zMSkPPmJ$7mIxRmAebN#MyNJ(DUgtev6eA9u+k1#8i_wl$sgmci`-oyap!||E6px@b zst-7?<*}*GhN6f=mHW+damjYz#O0w}4o9VqP&}mWdh-d{Dljwyj~t0fsoK}PJn!I; z8~hGLgoKvA5HDPUV#i1y!%4ykKDwxwSXhJv1Zr7Au3s1V=lTp!e1C2me(-~1hHE)$ zR~z!-wBg89_@SBaW14zB3{Q%Aj{r^~Ci2CATcY6Wq!ccG{Jn{0_u_AVez1o4hzFcN zX8AYJdx7#URnL`hMarnCj4m|V)|#|F&CJ|7N+;07@ZI@zS4~Y#OH0ed#6$)ESMl9H1%Ms0mdvMRJ=-L&iw{+5Xhhlb&ZDf*Dd_Hvgzaz`d6m<4Gm7xo{KJ*3H zJb;hBK`kW#Fy=@x%S0f$Gv$;MsN-g5k42?tV`kQf)|ILvzuTB|Uyi6WdMcW<2EXsi z-s*^tYNz(v>S{KdRIN*`&iht(CJqjqkp1@oOei=4k-{=x^~#4|z1p3QSy^8niHU#@ z%3ZaMq*@PneVD z4nQt1D>tD!T}3!JR2a4MJeKWMt{Pt_}p&Ajh;L7e5YYf($Q@WOTZ#e$0_!3wM^(>*!t!$i}&SP1^#5;Rt0RaJUe*XodmLB<4N4cq~DJ2KF z8X(DV`P>UK_1tk`Z-leUq|ruIPfxE^fVBA?oc^tx?oUvQd(sX7n^0l#9}#bNJ@wy%>}$ohE3-_C@UuPhAP z{pb$5aypLrP{?QJ9xZ8m*yXO>!5eTI`NLffkx9aDu9tp2aUD+QwLLMD=fT6LHqfom z`LfW+bUZ-)dHnfDzpeG$KmiOgH)kNEh0S_8x&xGHFb%#e@A84E1yuPjBm5Fv(`h)> z9_H$y(b}g^_k!O>+0bgta+tO~+Iz)h{qzQ1mdLZg7?ZAm592?s;ofots$`%9kb^TR z-Pcs;t1kEzM7`A%6meB6r(J?Jq_~kU9jvOG-~CpO4g_l&do^cY(X#9u3r(~)or2YGXzY7jLJ{gbFNUmyeUiy}` z47{0r<^esH7${QGS6Z+=IBeboeztu77T77KJq~xw*N4rMkOM+uqWB`_9--+xnZ#bprXgfdjYdO3O`G*Eh~GT*AWXRB0wn zdzsFi!21&Y(jX}@@h-Q^^6!JKA>4LPJpn2$H^iXrHsMJ$&i9 zJc!J|+U=#D0ir7)2OT5uBZlNr6>whOY;Oaqg!gKjvX0hq1>ff*xU=~0lbA=iDhVF| z1;XyvSFut#68?*xUO&Ppgn=BDz*0^|CT$b-s;W1T7C-)e!>ZH$MQ){HlOj-9;885y zN-Xu6WE8>{qKEBZ-0D`$cDVs58&1&a52R+Obvbk2ndvq6fp<+d+WdxnH_+g{3qxB% z4n!0L9tHLUqm2`D3cvv}wa^%P{yUev^4a_M@1sdMC0L9oin&HFzjjo6eIE1ZwcJBL z4prX7KzJy8G3>A?bCpJ3qpXn>op2*iy#BDORt#H^2fd>Tetv#17FgA-#j)~iCn@Xn z1DFKU7cTi*ZMRw&BqTom=sh{wC&w+BAN82ttMQv$lj|VU{}Jcz?BT)pHil(Trqk92;zjtXH$K71v%37qDjXiPNcfpWn`sag(r zwm)Zkf6NyF=DW2!+gA7ViMhXcxOVH)3n%oEGX-#rO>hD?Qv7!-j@@1vOX-LWd$~oz z`n)Yy)9o_34q*Li0p)WXW@C%@R7gzheYvZGr2CC9VVUrWpP^U=>DQVq2; zgj{Kb@u*3e5R{$w=QxhT9yrZb5Tn=tQFg+Ot<5T!HQ3_VvK4oRxUrb8Smw-!P7Vl_ z#(_6u{az%T!FGkY!3pTk?4_in$au%>do_wWDHT3o7?)wOj&rfG?Xnh=b6bw8r&#%E zxlw7}YFm2w(0fw>Yx+qOms-_xu>pa*A>Q+Z z!kC2gpOk_RDB0iV-G#1VLx;H<{~sHgjigJDzZKG@`Fpt9*)iX&U}_jK>DaKVCCyN7 zLVPXNR@MIr%!1F!t$es@>j}tEJ4jEoOboFvKO3UIH^XgXA}Gn#Q_U;j^6Ac5e~V@3 zFiO1LZ~{%)C4T-}`0RZ`e5zK#6xgInJ^e$3g;|ICj9Ph3pys@;-b5#4!T9kgf|R(Z ziR?2l3HcIp;tV~Z<~s~c#$@5EET^=n>Hmz#U^7{frg;<0${UPo6Sq7w$vDPl5t=ZQ z!SyB!r!chJoVT2_nU0E)G|1rInWDWMy331Pi6h`Yzf+z2ka(5_5pyT_e$!QqyjuJ5 zzhq2Py@Ndh$S@~cEY`VV!`p?XhHJt3tOz;hf9Z7p2NMpsKpY{&?ASWf)&1E=%=E|K1Bhm1}ZXyus58 zlLKTW4rv-sk_^45TM9ah@;hzBG_g8FHQao z^!#5YvP>v$A>G*${uv}Cr}kQ#4{Bw`Gx(!PZi)GD#W8DVKFyS1kZB;CxD%>O_*sLz zQ93E(C%w4%aH#b#cj~0}SaAD&66p;?t3X9Wg$xh*Wz_gN_La zN@2O%gF_66tl%%wC|QCd^{)L9_$!XD3!|OS}z;OGfF=E@)&) zraa@1fp5xwjDnvXM z5pSV=pY}GPBKdAqOiU97*-qMoW})1IVw6MgJB&!OVLoSUcKw=v(0)O!T(XzEr6O7vV(ss1zJ zTHmCn#GEn0IW(P3TNzjnGN*U<*%5UvpQY9c4ZG=G!E-=2j^W>W_nh;(l%kx79c+8@ zt2-vRtMtz>4!tOFiW~wYsO6X}NU}(Wx$vFE!U6>KNwSb(zf~K3K32-<<>%+?b=cv} zc?x0<2H9x95ZX=kUk_6~)P*@_dxxssA42aHMJlD@vTpxlRMw!8-nxHeVe#1jM_`ar;z^`(&a+(up!d?D!%){S#_V7pQZS$uC0N)Gm+2GMw9fKz`}hH z{3`*wiN}$5u+o@gYF5-sGTnE03GH>C)Dv^ol5z%|9odQu+fiPRzWSNSkhC|76mi|@ z+kxWsCwPR{hwN7x!O6ULO0Qyl!rLbFQ@Fb~E5xoBn#B}NRoj6=e zznw+v1B3WCk)%|FL8Hxe{)T*#uQA1|xBYX91JBRM$;nYnnvf=EOWK>XEG!tg{o~3u z%skCrZPJu!x~M(q?JC?byCf4hIT^vV!L(@xmv;A#vmjFF8^evc@pakgUC~OD!zE-` z30W(I7^T`B3QtgeP}*!4ZwWfi-y0ZdeQ=L4%&?<@&-o)IX#`>Z-K^otF57|o5C^Y( zYAd{?4f_6913~RnS|XF|FTEpsxeh;hI=Ps<1{?dZR(0!#KQ=xKAHKOBCD+kmM052L zYY5@+cXs0zYO9|(*Gqlyhdz_~zTAl|%BZaI7po&(l{J(_*d&r{5p@&$QnH`tn)0(* z8eCdFKaA)ALfy5+PR}j9tEH|_6WIeSNwS0qQ&oZNPOQK3!pAa>DXTq4T0g+*$2DoP zOXS@v2$iv_`4Ymyp}8U}0o#=hF+4W*ZN9#+`YfqHc^rme^*h8C!$)_hli%+dB>5c3z!&{qSIydu{!1N zC8oO0wt0M+D$}HsYiHZrdn9+IPeF0?r3eEt=RHn1>%mNBA!-ge5KSIh6Rm`e%fkYP zc-iy^at=(#B(EKr8D+}dNjX$S2j81lwxHyHgSJK20H)*CHaO&+%RnH^o?Y1HZHmrl+^>UrnSDrIk_}IX$_x3stF0MDIj}*5>=~SQeX73mRgy8fW6BFKh3HBU6J}e9$lA%ae z64#l%p*dS(a?ZrkM{V^^CoSr)^vRDA5`#F%yQGvKeW*k&d$VJ$*=<$neh(L;&^)D` zPP|YUO;C{Av(Peu$M+Fq8f_O|ZKgnRZLH5{w0t|D-sJs-W7sDOks}a#rM;*vLRw!& zDM{IX_An#(PG1!1*%eg>s@)bGb8^JKpScYxT^(@a%@t5B2ceup_?A{v)Z^-rE$;dI zr_&9eV?54?sHlQz*K<4WfQQ7KYt)PpK5zM_d*0s&*=aTLaEw^LvR&{4X|2K#Clh(= z@te_6E2CDR4jgJ8oB~R3pI1*Au#NUzQ9X*lq}a@|(a;3dd|zy4ObW>z{j$_sZWIr6YjmtwMa!0gpyCsK)^Yzk(FIo<&#fm zz*W(`-CbAs#d0XlxlBpm5*z;z~7GY2bO=JmdCXb@NezzupychWn@s- z{2{hL6>8%8dM78V40~W}APxTs>h<{oSuDS(eA(GFeDJs!i})+wG+{_E*_Snt*YLS5 z(VpOQc2HQj5KQ!r@u>l&d<^E3p%`-BQ4z5#PwUN(BbJ;Nn|hoV+Pyg~Bl}+@uE+&3 zsI1Z;`1ZoENH_c?o$YOgxos{#%x07$<3&7wV9MJyx~~VR))qkpI_V!0?J5~{bzC-0 z+V~T34?eV+ePp;yMTm0Q`w{PtP}N9r&&9i+9Lv$elI)T6FHgT9L0|mrhn{7 zUqxkpxmxw$i`i|ACzpQ>N}NZ(05dZ$`q|!PsxjRG4{h=D_Ca4j2NJhbLsDkEO%tWEg8Q@yOaZoO9iH?wZtNT)j5!U~2OE(=53q-LvpSm6TT`za!aMzqMS8 zac~`1D=98+9hLa1a^HEO0gZoM(6b18EJN#_4JBcZ3ZKK>bkigpwjsG%We!R;v|mJb ziIvU~EU)DL2%nwFlL8m`5~Sby}R(2%6=tDJ%UdrxvDDXfP#C9?*a6&`#`UM9*y zPZ~CG`(}NoEk-z78K-8HH`*mu{0X!6)vwg=*E*VV+?x`vEC;nsUmj&qQ;L0klI{35 z=5TBw17FVhUg|45!d`>Rw5+QOoI1}om(>y;Q;of&=Dyjv`gBi2K)>8hv2A`fXOKrm zPEN~u{g6*DB6b`2BJt?7m)`%#eHuV8pvD)+V-sF4|2^$ikv3V z_el_%5vQdp=~KnZsc)-|e{D>L=&{GEvvs!N2$-|z0U!uczy5-Tkp4_1@Wmj7EZ6Ii zk#_JY0s=bU_CSsq2|0QOG+W3oCe^D)Mn^wHS)%N5Jvi0Pi;0P$0Y;KvN#qwdm1T`) z3f4Xg$;-=|n3y2vg##WOR=~nT>i63+8Zh-DF4%VWTW$Visz)IE2xYi4V0nHc3o-YD z?fLo|Z_-rZtYRQCl!&Ro=iDc1Oz?qxcd3uC+2YTl4enSlgy7)Uo~Ru|$`vZxM=!*+ zkh*%wXAh)lLVXzeu|_jrY%`q$Z>cX(A%Y(ZxwKCC_*-jeXbTGq(Io7f5G4Y)vWVyA z58%#R)L+e0$_+eg-I~TIn}zp-l*Dn-#%uFkr?!{`s}tVYXs^-XPV7 zfu0_Zpv7^a!L;2A@|ko&9|n8_4?l;CdVi}jZEu0BDkx)?Z^hylIeH{NN$o&1seZZE zKm?@LmW`VoAjihsTMyRDX}+%KJ=5qj!KFtR6s%iwwN%^|s9%80)jqdD-Ws<3>Zm*T z#6{ekYsX)oFF_W-Niv)iAAk4fJ7@1NQ`Sd&tLWM;E?k(h7Ho*}@^WA~EzKHUg&dM` zlb!j;6sMnOkepomec5gUvcZs&)3INANDV~-9oAi*+S%ZXRCRwppZ(MdxCDF{8cAS~ zRdXdieC!SKQBCJrefjI-yL!v;H$eHyOL_2$DK{;g&~`lM^sTY;f-%qN@6Ua`y`v9n zCaBpEQ11iT&lL5pd9&FcNvJ@apNr3la|G!}^zv;{ILNpGFb2|Ob}f+WkYqLm3la4& zGpk?uD|M*PX+KuVU8UgYGVM|vEh#%oGl4KL{$sz_-=xk!+IrbfhftBW_(%+;x8Hf+ zG=)i)GaVZP!?4~WOx3=Bt0&|d7&5sT6HXmSsTUL1g43%>|2rG&N?^lel?p9_J}xd! z&)KiwK0+#DPPZdCIJmd3Z-|83eu_S{?}%TkzY_Sa$D|WVA%m2mUN5@v+?h(muG3r< zPw-y%W4&r4@*v|EE;O#`{+2PswI?Shz!FWM+xp-mbzCL*zWB9}WGhh3s_zW(Uh+?M zPgmEcM)W`NV^&E_#VgQ?Om$zF*vGlkAAM)EHH@rOE|wza*Hj@t`fEsQzR8`dHLKMs zehp-sZtm`pPqUUP4~RJR^?=m(IkMS)nvu|eK|K1XDF~mF_U_A6gNsKtM5b;%xxR$- zTz#Dax$9oOxdwCaELbhOrowQ3eqYKRO$9UU1GezBb_W!h!m-Jvtk#VLOkl!H1!7`j zS6v7vd}Wcv!^%)G@~iRb`P36~|7{gHDXHyqD;pb-_|&Sbs1}uWuhN*pqkX zLNN#evwtZL~G_qMC|B0fNbwKY@DYtbJ8dEirsxncX* z!|g(^J8x}nFEn1zt4^1^`SE2MSXtu@8*ReiUwC-txJ8EXDg%OmlSXFM9QOK^tJp39 zWFH1#Hv{vv%J^3qaYvxE7NMDfXNtq7tIZw!*knHqH0CBZ;wgovYD`5l&w$$-(y15h z&xTN3#8)gXF82H~HQ>7?x+miDsar4#9Yn3oqx|~n)1)NHvqQ_dU!7t7Fmq8?W&r|l za2n+!Ns}R+|ELQujnLZ4YU94r881+7UzF}5^szoTG5^z_QptjCYoJG^U=5|v^Dibp zRAc3TF8rLI-GavdJo3Lr@L%6?C^zy19{gtrGQNe@pilokw4h`1zp9b{ zp7iJInVp*);r)=!@f~W}V#0w?W^rE^IcGQQIE3uk_M4HBkvDJH=PqBw0!&Op?xJW0 z?OslI?*<*~Tof`}$@uIv;9ut3>g;ooCg;E}Qw%=?gdGhus*R40j)I68VOgDwpC2pi zfQV-BJ3M;Z)m173b%5jSPoKceELVrJXksc>Xd#e(FO0i^lR6Yn)%a`pb05&NKZ2C$ zy|Bobm@3`IwbS3fqLuIX=vF)Eo~fR%WrpKYiFjuNz2q(|^W%&0&?9n?vyJlc$e#$1 zJ}8!~{embS>R&O){=vZuh|+C#DZpXuZfpSd0?A+K!K~@v-~d~%mxzig@L*FnRX}*4 za;Y@;t^fCz{ttg0lW_jvJKD3h;8t@|asyKVqOZSu=z4bIi!(kt%Gb_J(0or>PVR{< zxq^B!spqF|FjwE+w-(aShvISOJmt7?Oj7Gys_wi{5O#EWQP*;QyA2y6Th!Cm%RcDt8Lwn!`mKXM#nc?kF9#J&C3>Drt4o1$;?rLPqw5NH_z zqT3&RH}~U9Vg9i7^KT#8Ok3S#3l-35SOZpOtMAWUzBZ=Y zS#LQ+OS1N>!*}0dwo3R@<9GlRNmf>@;FRKzot~AtBhQkURFzmP7N&~rH^e%kIuy& zdV#=1=(M*usUEkwugz3x7FvemG}?}Xss9zVI${hZ0FmFhJ39d2foitv?-nPZs`fz0 zx2Gh9{Pf(OckO#}9i16Xu0z7-+}}d(Iz3Hj*(YQeJwGN^<9UP`lu*9ucHm`zW?3-7 z;VmjIrXeqjKT0wM6c*^BHFf$B=?MQQ(RN6Si~0RvW;i-)|1LJL1u{hAxfOSde-n(9 z4d>)^O}41;TNVU=g}&KKzh-ufW7%NfST_B+I(aMXT zeo9JCzR;q216eC(dO{!-5yg%WZ}Sd3x$RY}+;RA|ja~)`X?Z<%VTpkP!P}?=ypEDcQc)A<~JoKMGtL@UBg0uGZBlu19xS!o`O?8=d(U?6l z4+wt}>f$vebnZ$^QPh$9fDlj)O1pco48|RsGp-dO#dUQ%;NLX{9+$&5zfQqdXQ89D z8ypjZZ9x^)WIKL~4UsAEpi1+lNU6)skEPzLYt>D9fEK$-g&AhGJa^{ zLV-R*@;iTf-xG3`fsrvo60ouqVPU6GMp_#`{rxjvHMz7d&0w(7NBq&Di`wJ|(rq4O zSZB-)QGwgHW3c!4_HIcAJc8Yf`>xsjL!GA9u1Mh7BiNLZR9w_h>Jxh+r#TcyZuuRU zFwl1`&_4;+BE%n(Jog6q1h7on(*F7zte<$bGcz+Ksu^`JA3lGO#H9Yy*?3an3oXs9HD-)Zhs8zYgUe z*vn-(szM^T;mn8u$ z>b?#@k6I6J!l)_6`uh5KD#^E>;0d~~2Cz(r{V{uRg00U-$HxN=bpiWPxo@q9OC>>% z93C0L5|Z8w!?}P$5!B8}cVuOWcw+8vEGl#({dRe@f>(qI#W;egz|Y#=b1=oo+l$`0 z`Nox_iVx+UpPz{2c6qidAsScdy+U4T0(URIIE(rF(vI{fv$n~{FKqg~+&R=X3$wB& zSJWGr#u1?f3TuMrOYFTQANg>MJ94tEtQ1-X__IrW5^=-Rp8eWPctO#VGyJSB5B=eb z1&h~t=?wIU0uz?~^$xKCtV}NuF&ys)V0Y-{c;9AT`U;h5VPWC*v20P7PnY`cL7oso z9sY;x;Gs(bGR!cF{$)JcotS-a81ZW%dfN?Yynz8}Q>2L;x>}A{WD|ZyI{$cD8kuD} z%N2RhG9eZ+^75Gqv9XlOPaM~;dH!IfT1ihxxRXOPUInQXcR98@I$4Y9CoJ!v;61zp z;3h_}+XU;#BJg2K$$xDx^x+m#LnKgI=D1P@sxOs7D}aaI7*l4jN;a3cFFiG9gKB6YnV zi1{D$eQ+}I!2SYmeQfpMMRDZ-0yhwzp-_s+o;6!&YejREcTHg5wzaZaNnz9Pc!m-)dIdp9RdZ=EF7$Gjk$8x5~89dC5A7c#NixyeJiSn9h4S_=8GjT4+~d@|C* zm`^^mDkv12AO~{!BDjB_sQX2|@nVx(QGb<4oJ7sR5tp>+)(~b?$T`4Lg)eXHis}9W zU7M9h-rEsHg)Z-tUD8*f*UN1`<#!b<`^4i`xA{`-oxuAqpKm?U!}ykN(=*`3sU{ymYTyQZsrc&qS~~eSAwrG2ZZ01$-=&ER zQ3;K3nTT2WwSZffgv#xpR=O7zUIB1?-tX9t3q8PiAUDlN0u7w3wdRu zN4)qY(T_E~Ze-EhMnH}C$h&FLo5Ld&r#Gm+tF0#j7*PQG9R5QnfnB zJ~xFZcc9xU(IF_+9!qB~{p#L2U<9{G!EPYhh%muTME^m{yj(|CP$tGL%{${fb^#a{ z5;B0*F5hqdtQC7r9Jj_=I<>7Hi(M$-M94-!R5VkY)cc`adUR^@KB$G_gtZ54Q_-o0 zCmslyMo(OWj!~xJqv>%$332_7QCSfxby}Y&(rX2KO(`}>Fq@=6|V_wLyTZq{3eXcu@kq@RBKntf?3$Q?#YL$ zQp4SXQP56U1|663X`}@*!WW0hKd{95q)Sb4dK5%@3_qLJR9k01K8q$z$9tU^h(P2h z_WcBZyK*rHc~G=rl)uXnv*}80J7mhOIwU93`-8}shnLrB`Ox5i7@|zXgB3L9%OpOa z_|o2k6E!`#HYm2ri#{*gt+z<#VP zrO(2gdp#U8Nn|UKEqXiVj`)$hCu;UylXe%W?nZ5_7DP!$D3b@ZKeZJiPPOnC8jcyH zOhI})o0cYFPow6N3IuhhTGqJvUR&mL6vjV_NcH85QKRP$CsPlpNEfr`c(=!xUXH^* zn@mc;sd<-{QzduKn>k-*?TN!}TQwjaCAo3G!}Vd`)wK2Kh0Mgn&utFU8hBJx@dLc+ z2JS|b=m}c|X&+P5A^|3)&_&d(91uyu}TD5;1ZMvyBfIgyiyK9+js`)60CG9!kkK^2ci z-^OFv@ASTLNaS7FaRCJGS0<`0Dx%-F=!UFhy5L=5{Q_9rSAe@W{&3+3MMlvf)T21a zic3pNi;8>@Tz)5q+&!rh-`W2Bk?|s>uuEVqo4$0Buz(q!q~-jez<6QE^%XjxCS;A^ zPJ~|zB9fDd_Aw;O6~l-*_w%78t;`8N8U=C2LkkOwbjcQ$oapnp3t)1DKF1%x@*nT! zZ@S`tqhg#SWX{~g5N3x9^#D`+I}|LSf2Z%_J{Q`rv; zIOS}`w0>aa?%`%_?sV}#XA1{Rer{?m>WlyI@^Ntq{m)gNe^*sWx zAHEse9@|!1*c)wG$o^oPG^!>gg^MSWb}64wN}0N8LKFjo;SRGj0d>M98hSiG&isXE z49d?|7#G}9GU)|hJ!Ln3#_l~tbL5X~Y;8@9v2?IH>;Ru~YPPlXQSGDCkM85bzeYGO zQ0*W||Ns0LPE}P^75qDd*3vsX-?g;6jyEfv|BmP4N_qL;k!2|Ty~dyNXS5pHe?OF8 za)G}O#dBc&KmNd<^6wL6{~qw4PyD+NKJn+B^Gme;J?eiR@ZVnf$kO zU9vycvZX>GUby)5duh>>$~AqSjK!EH<*m1`hY8`X867F~)|AwbO7gB2j*catMK`+& zzmX_l$oSQ_3k`i#GPAG#&uAZ5mr5yxrDEQ!E-z9s!XuHw!uH4ZrPu1PDD6aj_)1~c8-R=$;;{4n=@WaBw($>6hZ*64-#`kIZw?38mfL` zASAq4;l;wd;|~kr**3Pe(d|b&eW1MCKOhVZF_>f;u9=7vMHSq{x#np20lS}HT>Qfq z4?P{7!$%T_6*Gy-W(=Vk+8`VW;VK%gMwK-+9IUK76tPOh1i=|;Y3RiLKR=c85;eVe zF>l2K{Z;u-A%@edY%VH7MX(?17w=Y9V~bJ$6&=*pyeQ3Vn3M<4MHv`mYv<(US;5{kE{EzCNmQ^%kA=BjaY zTpTQoUr{o4EIWh+k&$=zp%w5Zjy$xw!NPcY@>|V1V_)hL_`4Wlm*k22G-&MWv}5J2 zXjj+P2J&5*nwUt)&5ftFYhKLCqU`Tq-c`7ww^04}jfO*9btWTW5>vCYtF_9JNF)UC zWaPsYK5jv;tQy{Zeem44%9z2kQfZ`;NAK^Kyf?l}9EZ8g3XL}L@pE>3JN0>fF*E8| zQZ#zP6I%xm0&}D)Z|beu>Hd||Z%KtwV|TK$_xAT8<7a7MQBhe*o_>eA7=#KoFgyF3 z&n^)|DAalu1FRYU=Iz&8O;@ep``~NzX~KTb$jBf|*Q2#*T%e_;#l9+yKpCGQ6x(0r+|YNzFog+%&O|G#qFDz|k!)*gYiK~?=jXqE9awSx z4K1>;x8a@%=qQQY_(Xfr=H_NlS*$6q9R)-waky)h(kg_;r~mW#EAqspJg_R0N~)?( zj*eDV4@gnD=;)-&bs)whB_fKD*NY6(T}`SWWB#{8zm>t4kwi1_@PI3o6r%ZkP4aRt z`mDNw!tnSwDcdC-UEN4UU5o1fd2c6=8EC=;m%G{MtBcSl&6v>-k1st-PL`3C9T*&h zwk42R`~SM#d*;cLC*V(|DCz3x^w+&-d;dUCiUL~28W;MDSko#Vod*7AN~?t&oSd9M za3O`+6lLD}JUKNLbm2>cbF+~v26f4Kqx!$?NacC))c$^%Feyzft=_u#Oxm$AF*6R$ zT)OpVk}4_w6@LYx30Zb|vs0PIt%3X3cl3%hFLuKPtHTmZOsKT>1b)8Ma(spTBR9gI6og{D7fVB9=N3{Pdr@$a{t56-@<}6H|#j+ zF`)zpo=bWHw7;dMuDi8%AUJWNxO8!}F7ze_O?+NSl>J5j86?L%Jw0PysiS%DK^K`k zN{JB^qhLUqnlvQZs~)U>hP{U_!n)1O#3Y}u;c`@-o3*#+Mu{HMlaQTlYG4pSM+uSpAHE;>!A9BN-zQJHmseJX9!|&4KWkAR!BA23Mdw0&f%f0q5Mm`#sGKb+ zD!LOy_KcbaIL_Ym%NzLkiQp-uryFW$P?MBe{F@*9xj5D~t0YTj4hYgm{IM5%?7a z(!``UO{D)Wb(6L{+C&;y&SEW$Vsn*6P8ZD-C2Kwk*>G0*vOkt8!K=C)S+DK)QUV~- zwd6@&u?vpRM0AQ-&bKxGZj`S$^?DWWzUna(fqeYxf4JoT%XEJ7>o)^aQ-5g7{kfqj zSGWtHBHTRc-0@$9G*T-mN(_9UqjjtDjbIN2G81?0|4}ydhly=kygWR#bad~0hPpaC z&!990#oA0j;Xs2ClS*3MqoX}2*+#VLgZy7qG|8}OxgW8BkIBAJQVNF;h#j?z*$Z8} zDoqB7(S2kzr^Msnr%#_E0*+SSN{6O_SC*}2%aabi=(bn;&~gvMCDFf{UqRGOhugQJ zrA2^|@v74FvzM%FY!@kcDE}-E#aog-f9KS=AaVL*dOhdiH$c4XPwT8u5)d>%`z896 z_}mO6_yJBJrj>!Yf1;_1CPeGdylBTqTjcBO%fP@Oo^%s zEv@h0O+xq|%!1R@6%VdW!nh#H%W_aVU}Bxh$GwsT^E^#?5 zgw`xQ$}HfInjkpIh?-V#aA37+U~h+9Ix9WB%*ICb&h!jqgC$F4^TCWVK@Awt4v>lP zSniZnt%i2Gni`y{)(;P0E6Lnlcyp(F1p}NZ*)x*pVox?{*B!yG*z0d1=IXqgHZs?S z#H?*>hTHh@yKtplMD^>hi1LEP!muz*{Ym zDvM?1w3jDzR$>m7*|h8bBX*>!0~n|jB|5B709$Mr!Q$+|yLIcMABChU^vN6p%%w>E z!-HSYF$|O43Vk0?-nlg;3Q(NfaniDKJV02@L+#-{A4L}}wJERd*!H{U-!0?C1zx1X zVksYW{am~cHC-si^EpHZ1N4}Txs5<1`@v`$cfbw6t%*2)6x%hbodXb_@yvCGg5D2p zo)4jPO_Tn7YG!fKh>zU!4%&oDhe3x=y^ARZdQq9M5*63IZh%ZpPfaCs%{rt^XEyXN5Kvm~`Nuk83ub^#n?_Ykz?D=VWGNAgka*V;AZWmi~trj$8p$bZ~;9rgOo{m*Sr z%VDLZyJY)rP*7}p{f3E#ydzfmvyx0`LPYVRASOnms;RA&XJpf>A`^UIa)Y0Ew!1UI zT<3}}l9`T5A`#+G@c5gsREgKABC_-pB;kbfvZ%+CNHUvYs zSNQq)_?iH@7qCpL!B7O9siM6&L9hiXioRXuBX>&WBUXKgjAVk7N2H_yr8yy|Doepf zo-!1~mH7ssBOTxWdNGTMB~s}Hg@vGd78{~pSy8~UVt*auo83C-U5$h8lFFN5C2wbD zs?54>wWOz|&9^BX*|SpKrl(&C!IeN)X|TDpU6|mL z!PpyDu%t3rv`g}yfu~FwuSc?cOQ*M||x`Gg#LS!T&9r-E6rr0wthv*qyMJ(jjwS z4vzKs4(f+)L0g?5G)JRSNkCKfR{V{fX`ScqVxP0?IYYJ>a2A~@B{*bU-OhAT0jnwc z^r|RH{CBTOsis^l32yoB1AQw{n7#V3?m=|P z+u7MSY21YaH;^E4E1O$e<{B1$Gl05$S5s~`37Ag(8WVE65Buxm&YCyxK6>;BFrF*Y z?)AT+)(6whZoip(_JgTww}y5cYPuK;N%uFIav$vh&gCjelhv!IXjmk9)Ej&qr25wa zFn)f0$7Ii1@C%@5*Kh_6S{x1E@e`-LS%YH$U?)HlgZn&s>*OK>6Poa84qWl8LI!}9 zdFLJK05j+Ty>|ikgJ_>an!jGcfsOYR4rG`rl*a|jxx@`lrb@+L<`*go3QVVVz<*Zf z+*@^Py^2F=U~J6h6!i4^O%a`Z4FzM08#e%-^RpL=GhzVs^4)y^a-I*6nkl@DF2ml8 zsZh!|p3zpnTer%lk^dm%>h<8W$medVVYW{~L%X9^{qq4Wvhkt)xq%_p_4lAn0b8{i z{W$+?S!+@GJZVikSiaqp1?LQ=Jb>9<)QVg9nqRyS;^84MS5#60_eiaDdh3AJ0vUMi zB9j3c0riFj3oBW)R=MMHw+_bk0f@fz6tqwr0%o*37>7zi4D|AFP8!ZqR0;AC8Vx0= z+}9~=pMxe^05eHyh+&RpB6t9<#D{ioToXavJs|pGimEDl2?g*c*yIfs1DJqLVf-W6w4~P(d}o@!kT}%?CO-f zy!>xrb7|>5c`AQCi+*(aG#+a>OF+fn{}iyWD~-Ri)z*!jU-Ea5^A*+AO?>sD5^~A= z>KCci-&l69BBHVBb_!X}>)C(rTxVZlkf|f4-pI~Y@)5IMue(}r%=vUiA3VV>vq0w{ zX(|5CmQK0Mp>=r@L07OtOO)&<<^?KlK83N!v3cHkAdy>pfL)o{x*W*;;t}K`D0BYC zBa6``QablI8IvMW>~FS~w8%q1PK1e8Bj3NHlR7f+pYL}OshH3KHv+7fv$=}VHO&kO zf4|?=fp_ns@bPnSaBy>Lit(pZmj3mA^JZrt(bdqv(`ZhDgI;o(AxgUp(9z{%UayDOD#2 zhu{(jiU7dM%@}_f;-W>Hm88?7;lCj#fldbNtt2BeGYG?8P|VUv1q4V?@`GPWLxb}A z|6=X0qpI${FJM#!Bm_}PBo$N;1Vj#kq@aLw9J)oNrKD2{QIJv*IUrriT&Ao?X~8bb1pW9$R=L&$@|%uo5L0m6RUdAmIl#l znu^qly-1gnA0wGql~34{bg~*%U(rxIYiInPu@}J&(#|t&2c~)J<+%*7z3OVIt*?)2 zG1iUw`00~4^6~b+s+UX%6|i;|LD_Le!-ar`5@ed>USvCO^AahSV7ZVqyE8Vk8KP!a{4NQlM(lFKH#)1rHC{iV$hU1fh#0_y|ax${6R^ZU# z7={nM(yj^^i?Xt<$qvP(OW>o@-fDN%IZ-O)lQmW&G3s}z1js_b4u{IZ3dq{4WB{K* zRGf}jgUEE(dia)p4Z6T;h=>QM*>x|C&wRO1wzpV{f1d6_f&)@Zadi3PY5@z37c__lkRYY5~87cvw@ zc5q#w(r|%1Z<#wt(ns8%gw}Dame6?v@`kv?Z4#bAa2l~+{n^~?V`vTP3J#8pZ{NrS z{Owk#f;oZ$p;D&<=>!IF(4gt7CP`vbZL(pcY&Kv&0qQ;9*RK=0hz6&_hQm95qGZjjoS3kx9=;LdoulLNN&y^9W+3J{H6kum}Ja zB$WddUpX%c%@AVJ7AFRZl<D<8pymfe-QFGhC`j!rQrHQ zsD#~*4z`vwl@1WR&{Md69S@H2LkYNZu~1f33L47oEj1*Uei;D|@keY$3mZkR$Up}{ z`MR$p@vQkQsQL|G+CA_ovg*&1*1t?J`$18}>^Ajs|I<5=jC0r!_JxQZYIbyi9Cd4^ z6aV~7gk*ZR$Vx5w*tLBSdSqzkoC&jSdL%3BGg9FEslsOT%sYh59*}z{u8Iqf{&aE~ zv`$9fe(_&5bXPv;&6}d}9JD9~4IVrNUDsLh`+uX6EGz5j1*Wq=zW2<#uI*_K*O?CR zjCw0N0r^g?BgG}n=f00<{03y+gdPoL<@>#4r67_plXy<< z!CU7aLGcm6WiY%cmIVKcft&l!Cd8iVO^G*) z;VIRRl+}jUPw=t@aU>w;B-`H0ToDPiR%VN==9xSNGyN{ zJ;t%04QzUd&MIf7Cm~w`Czmz&GyoQZ3|H^l)3@$^YW_9t`4;eWpn(QDF-J&YTyC&F zgDSj(f(~KeqeDwe8x;}Z zWU_PR1EiwWao{ux08Zo6e!vPqI12#LHA>u@-UU^(E9g2SWwa_7cVWrs$%m4r!;nJ) z{_#c*wgzB9jtl*GD?T`)kL2-8AK8>cZT2c7mEjZ6iB(u=8|m-DWJ59viusCa-G=al|r^b%9q#yRX~*A2PSu!@I{ zL+gf~6vDWvDexTuaS{AQ^g0*9!D*bc_kZqgEqMVA6BNThm{-jF2G}H5`8t^#F{l5? zQu$WQ2zeazP;!_zXZwhgR7J)`1O!|G3lr7;06L~>hPpYZfxwn$`wL}QhCwCgAQYpx zQ#YUIXlFnOof|CX!Mvi|x>zVg= zHj03Y%#v|A4Y*An5a-onB>XO1^;Z{9>Y0Pw&~2-pULVtKUR?Hnr66kXr=Bs?F$7iQ z+eWY!zUyArU=T|tC<(VW~w)CfxBGxGVGw>tj(9KbV) zf!zguJu>Ot&`gC4It@;c(@}PkVM;gLz$Igb!MS3T+!Hg3XUJ}Y4`grvOJcn z@JB~Soj$1v3TQ;h-N>_t(}^t%a!nN7%)YG?u$f}8jcqC_s!!qJVn}7L|3yeRjnJi< z+qiR!lusmD}T;a&mH@q}#NC#>T?J;(wPkB@2nExr78>*Y)KY*@>$)#<_2YRgbYzI3<9%LpKCL zc_apwXwv`PB8g}~lfmN_tU4tVbAsk>ggLs_0v-kqy3ov>+c-{3yC>~NnUXUzg8}ox zwo|>EbZX5~0na)^PES9?IfeNkn21IbYA_Wstmq6>e*30!3nlZvcj)Lq1Min6NYa)6 zrpxxrz*+kes3~(y%^|ZLNkGrbp2RA#WVF$@%KC9w^o}QUjXqZa0fvoU^%0LA?vp~=^VcA7V$B{p68;z_lRP&1=agXRD>(boPO#5WAyA}vFIU+izMKnRDJ z{A_P?0hBb&zkXGW@bX2<0ULse=}Zq$#iaq%g|=70&mH8+7X8ed4~Vmz9mW%|v8MCn&O_NDh`5F7I&$6zr4*H@0?bK)#TT24;e%ehD3aM*V z(TiQ?8v&H;Hvl#VH7Uf^04d^YEG&VzHdCPUk8O2=U;r%diq$4|I1K^8!B)Zp(b3UP z!!B4iKF9>>&*k<-fByW2uI(qd{lZN*T4=BaNrv;E@26*=?e52?fI^r7DIZ?mwpcff zvg~~9SR$OmO;EAI?ctmQ1_@?Aw7|I=t~_Pd0DbTagw~6d)gmzCeYahNVnC+(_6DFI zb|CSDemEpV5+vZRk_`no)@JMJ)4`sXr9)N(=1+M5B?LOkn{4;PkJP8DD+x|&C@5%* zm0im9w?8*Nbv-o+kr+74Rc(gvvs0kJEnlua2X3eWA62du*-a8Xw4G=OyHGqoZ>c#5 zg5>x&RDqM~E>sj0_8_1E!~l?VNaGsZ!4IoBf{Rhaoxh*paumOz3R1nPId zpnYC}yP4Rv`_GifeX3rTh@F3&k+4Ve*uSazr-iw>&6&!q(^ApNJIFC^v|}kfFFG@G z>ifUCEt)`{GB*vn5OLgFZUTz1f`HdRKF$7I>}_ZbU~}`pTsLY414*oD30y+byVrz} ztD2%$qT#*(@p4evD#Wb$F7yXL>(>C0A*Q`6Hylso;Hw|N1o0Wx{{{GB?c8&L?VZ;B zi9;tTXn7cLx~9NkaGwM;aKgd`jBVFJ(hS@ zIy{2lTXN8uY{X9q_^YV=KO{sEcttq_3xJP%a>-&vtpxES2To-RgT=YJ$gm$v(XSPI zGIV;3VwAYjtPNv8=0b&HTQ36P4+=if(ZGshr=@)Huo6Wa%S zwEnVMbRydLxfMi9wxmvCt3rPWzNsZ?%Q<^6pDazMaPZ|Rj#yuZ1a zZSw5d#&r9GJHT=^fxY}Y>tUG(m^kXRtNK)(3L&A`gY5Ber`78dA|IH-TYCkb?BGTE zK&hs>8d`zw?qz+JKd8DQ&+Yw(5vpD)>5XU zNJZi6W^64oSqJ0H9x!Pw08gl;{eB#JDk0W3U%+8ueG$*UFv9qx)8o;gs zku4sk9kSCUO{DT)xRLSilFUq^h_bPcrOSB$mO)u9Jpc`gT*Im}xK)Z6_^LwzDL`=` z0UX$>aE~&GDaLv1tznz0uIs~x4}Jap1C!GL>HVtC9(?CKcf>zXrOFyJwvW`&nOkuN z_6lGdO9r;!cIC0Vz7mzoNKbzNFzZN8int(1?_4S=(}9J8PeL=1iUtVtB?2nO4#jY| zsLqly%xkby8$+?Mk8r39pj>S$U+<6wJZSzs%-q2qH!(`2|N$r z*IaVi|L@*3Y*|6<_7p=nnOr8KsI^eUN6{|Z9)QYj-O3eTn&-&Q^uGl`+$*lvpt6sc#0Sf~G zA7t9Enyap{Kwy7djq5xE!4)iNxyZhMJ;Q>iDzf73Jc^nU`3cS;Hc`~$6ap#pI!Ffb?TUe9;*cb%ti zejdD!k4&ijPPcr~94`^>TPnc>_0`kqy^sk)-Xx<&Nqcn|$|KYBY*%95%&zSuy>K(6 zXCMxpot`$;wg0Y4a@ObbI2VIRo#5f|eVkNoim@MIJdadWO;SOhVafjBh&!k$bcL6R zNj$4(yaQ)3h`IURzn}Fm!?=I(3aNi`n;JGHb>yZjR10)!@k)RA&Og)E*0!?B)aoZa4>r{5*{4zuYeG zT&;%}cNiuwkHQtmWVPh7Ze?T8yGmeJ6LD#pu@^VE=kl-HCRKis9%IB54$C(C2>I%f zTYN$>aobzhCYN;S^={VLx_0%cja@Wq(F!OZefRj zs{1O%NyDfoh0Tj8T(ch`BBEa*$4WSLgiV0V^97ig62dc8|E0NRA>SV-g`>;#twHWB z9&f`>_fBpZh&^m6o3e>2lX_VAFj`3Bget)W?Nl>ga}|~M`dJa8@805GE&BH6?ORA^ z;GDmKX7{`d<(!hO>c38VC2E#WJRdR{xV?CJ!>b65K@0~JA;?$r)G?p`a4pG9E}XQ` zJBH=Et(z+QET?oahG#X%0bg(gTa_COf_}|_QDa8ZRI#jD$W!y2kd!Rfb6#VbC%tz7 zz~s48g(V8YA#rx*&%AL=FG_Hpr{Lt~ulmPXd5U85Vhyh!f3*h9GwA4G6td}W0>w;_ z1gyOSl}62(paSCKgJB!g5!6L(f3#2b%f!_+Mq7QT0n_1kZr%h7b030HJRfo);B=*{ zvH||}fVm~Qsj<^aky3V zvj8)JQ0uFZDhmWP;H~ha+Spb}oS?`JHRhPrVpVuGqu-T{PZQ8*YrKM>9rBC^Y8aZJ zvrVW^pMakit&rL?M;WMOEBueql#K#LiyDAg&|B8rZZOZk!!ODKG@D-Z=OM14#d=5E z(pa9UdIX7d(GOpMc1u z<>&zNkM?sTP+$_D&DSQOH*#kX;wLJS!#;B$*|JhBt0lIx8gG^Aa?Uu4{Wm;j)s5OX3&p_wwzJ^*Z)ds` z$*ZlS19Xn`v*5|}n&+G6Ykz30HK7&0eZidiW)}ijih%lMsMr7a0n|Vf6O(}j$DGfi zv4Q{SQM~fuT-#h{(bd7gX8|IxDTefiG91E(ESS%CDe38hntcHi0Hcui@5y$I4*x?G z(;_kMp_rOcIOWg^PN}^xWc^j$;wg`fYOStZly?zap0^?Am6?{s#XP(h@lch4qzaS z)!rW>dBfpNC?+9PSMFJy2!%8JE^_bQ^ON&=p;0xN_h@czu8vHNGvZtdDkg^W_u5ws z#wc!)xWNQMbeC$f%FfW-rQlh|(z;dTD*E8x(u-^Jspa;4wNCl^_1&uyy$e4;r5Pm2 z{9DQo9dHln89ZCKPeFgu$qfB1ul5ev($v&3V-L0OjD1I3KSL{I-@v*^L+8B~wHz@; zKvzKP7FFrcum8T#;!o8&{%(+w&(F`p8UcN`u3GAy+T49X9v+@v0lM>1EuX@|9^Z+iAo6QB z1uBto5$*}EcDVyo`~;YycG6@%YT^^Z-37OZ=E)UjAgE0L@1>ezl_%GBnwpxT6-U42Tx2QoP|lq%lL`?IG&z+j3%y?H~B z+oy{}u1z~@08tQDLp=V1VmO?%I=B_->FG(_0GXlW+pRO~f@&}aUb7~b@!y_tV@iMZ zKrLP$wk6Hk%d(1!KoM>;&(G%lOiLy-cp9Ca-rU%DaRF$!zFz{u(h|tkeohx^XMlAC zq|s0+u2!X~h_Y2`{CgVY$ndkL-o#^!PQBDTecwJEf)H(xBoO%?*H9(7H#{Y75lb|j zf}6}7(=!L~P+e~FpLC;ZnCXXLxbi~a3q;;ALL}Y0f13;Ewwf8dQo4+Nx=f-P#4AV( zuG#^1!NQUXqnk!eMe+tAbAe{?U7xe!N~041(hN@eCkKY5EjzZqh3pER)hg}5KGVPT zikF!% zG5T2{At8z~GC;z%u`|1dDI*V0S2%ge)$elHkW6{%VWBy^T>bb>=+}Y_rOy(Ia>(R z#=n`~VcHF>Vb}>~#HXG@o=v1BFDD0%-oUGU`0)PbpY-g$IN|1aiER1pRKzJteT=E3 zkPe%GfHr3u@VMHRAdYL6wEi7`^e=vSOQ|PIoT{Zh)nE?cIaCd8!kZHbpKwMi!8rGF z1yS)qBV$RxZX!)~C){Hh>gTv9d2!)%S7+u3-?;I^@HWguYQ&?Ie;F6BkuuS_^WQnd z_4yjWeTv_syQfDMM}6`e+mEl+E52W05>@qXEuBl5UsE+%&K{14M&j!W3fy;2J7S;^ zq{(&;8i7K2>`;X4+O=!2o!@Y$0U1e!6r-7EK)|-NdR%B;AhSf#@EdkRd&8=*X7P{n zZQ<=0YQj^s-k^1{wWb#*u=YJILux5r;A9~El!>AQkUbI098KWs0|fHz@-%Do<;on1 z-kV=3n}{Or(gfZ3KoyQVQnVE6SzeGjpXllazt}qk$)`8wl3jdx` ze*~;%`m#W5D}Fusm-Gt_Wo1GQCXS8;(*}N2;_4tl0%BNljGSeaMI8qL=^ypR+O{*c zFBo-dpL>j^Evu!O0q1$T->(hAy|g0{nrbK&n?F%iQv1k={fkalL)Eyeh#c|jKL}_H zSrt+fgIL2lw=d2w^Z~1po!y;}l)-Gj@n7Z>1yK;Z^?P1@yRPIbnKQfOhcRSBYUbrK zD&aJ7nHBW1%7o9)_$qPv(MABepjLou{#z%ocOT_J3R3dSE|ZO*e9WX`}r7lu>D$^5dl5iz)$f zz7b41tlR;iT@(exCXNDM5|&hZ(VDnw_OU8T*!v%nq=~CMYQ*bbl0Aulz&w9`ujSV-$=7$I zNMv+@=G;XQ0}Y+gpZZxEy=eOv2B>PqFVb?fT#3A`ay7Z3cGQM|i?Te$jyHU9Ue+VkGf~kj6&o_4q*}FU z#>8gHe8{-=jI#V|wmQKz1lbtdYkVlirmtTk-<`EgnHKyLD{*HzEd4U9gz}@C-@FN# zb)KH*k`Nm!sGxRYlYh55CzyPEtsZAVDPP}bHI-`OWKhazb-%Oe2{0~e`FM0;$%%lT zQj$$T)&g}=FGVE_&A)OsSTGrdy^gWz#xu25wc>9?7U6;EHX89+WcjBbF-i-ty;Hri zr6Mi=e;-(UQBL=|Q}}v5$D`6mk+?E9%RfRDk_+Pj86c1V+g{qbnV)dHsAa`|;*iEePN6V(uoLDPhc#Fsx_rgX~bh2c7ZR>9knixNgdk=)b zzl`UP>-FN1`GL}{ne+6+u4^rf<+A#^x-j#Nq6k}A zmOHKo*^PH!fjIIjMmVsxBN%!pfRAJ6NP*9>{`Qbo1)%%TGJE#o7jJiW!wvw2m$-Dn zMf_8-x1hk!dQE(6|5(lPHn~x&^v!Obp*y&!%h=*nEj}p9*%$S!o=Gj6^O^;Suw(p z!VX#0Q#P`gtsFxTwHZ=*n z*hL_)#)*(rK+C3ZcDv5e9WtEd{sl+&>Ka8lQO}t%iKrH;qI2fn3kJFs?yI$gwd(Ki zIoUbtMn5p#au4*i9M?%wWeZag$)qNy_eO7ka9 zdAUc^BUy}&UbBrXVc&FmGNGzXdbJUMZ<~P4&%MNc@ttCpL8+r6f-(jt@xc>`_Ew<- z#HO(Qbm)E({sDua@w+lgF2ygiAB%;BDixQ8Y56Q!?fBd7UyS7gJlipDesvN{JXEQ+ zRu-Ok=^mPbtPfUdqrxM;FC!H~=>QY(HSd)Bf$Md`sq9}UL=DR84 zJOknpg4~T{dc=S;_DruC1PUemB(Wy&Gy?;v=U|P>@7_UC@I)$vi6jq89>S8I-jCu- zfJq*HJ)Ve=$)=_4^vc?P=+)cil`6kt?yc!&>8sof-9fS_+Fi#z#=qhw;it2MzX&?+ zo63Ct&dyayymRv?z6~@&6WIq`USS3IPlej;e0v5xdLIpt@S2w*=*rS7 z4_FT+f5asu&%frjR9^VBk*v2N?iJ2skEO~j2FEfM6zctZ@04-@j}rCtmd3_cGA+^E zI4Jzc2%ynJmBKCFCMgT5Dqw0*RB42=a zRmUwB%{fdsyYlX##~zR;L53wVT$4KxOK4AXwD&lBETBFk)Bmi zV!foV%X>>dkKl5(cJk?N)=Bs!g60Y74gag=+~vHsAbk>O38 z_OZz=9i1Hq`+9TEi4sM$A+Wqs@9ESrs@f|7?COD)GbrC$aGv~N!S-|<2=Q=(PykJZ zwx?|4cc-E#heMj@i;*8G45E(2M$EBqfK-F$6n=KyYUmpt8|B4|SBLcLca$zN43GDj z#Cuhvd{e3u;33TMag&U`7udS;H&HT}4lF_=IvV`5DgnQ;cves`ts1@A#O~p`_4;0- z>CM9X@)8fOgg&~lq~JRtIuwtitHJ3E3Drv2`+QqVOj{S={!p*UN}Yb}0z?(1ye=X- zG;kCAeSII*h*y{~INbIc7Rnj()kUsQafeaOix0tc%z&Vx=vRcY%uE*~b(tDNRKaYj zf)t?t51cOdwX`mOrF>l6twwx&Y3!J}Y;p5E}D3>+P z(+Y&g-{v`kvsT;B&1RwHw13D9kcMTY&8WBY9`Hw)lcYqB2CbbMT%el)N{O4I`^yhR zvm|Ng#44>-eZ^nZrYgixYz9AgT?i1kr~la=v0eshFnWW;k&O7OW(D2R^rT+c)V;(+&-d?Nmzp4zq%^`3E z67TAIjfdZ#iEijK&c&OunV)VqiScyZ&VGE-mOrmM`KJKk&^*=RiGQv|svWxR7p;3T zwS1m(a9+=WX;-51b@{Cj0w&FX&+)b&IFx1}Mvm@HY5 zKKuDvu=U-)dxPbA{HxHXD~cqS@EXk9zQ1}^{|un;;SiBAmzT#BR)U)jrKR?Z*nwL| z({f(iix)S?Qu(0dvU#}$KAOS2Y8mQIv%R*!q+(6DaZ38{VsQ$P173i&HPBmC(5f6a zKHq*^-S#(js~IbZaLY}y93Dt(T=|t?O zlOy!KsjUaIM?EcO`C8bYK|^eu%ow`Q*qi>K#6jixQbtOq%N$2>V3QVATa+s_98HHDKTrw49fle_b`nVjRK&lg6k*DR8S$nUYR8iqswZEfRYQIvdbxV@AIuV-1AhtKG z#3rs8DM2GF@&_q941Am5-6Lx_0LB9=7bmoK3Apw@N;dSiGfP!a9(Je@^n!FurzZv^ zN_PmqcD%0(RMurzE|HZdxkYjLS2TB`NeF|>NQSYEHSdPa+V|I;sS0@dVnU)%8-&|c zRSiMn(=ZEsiv1@npoNpap5&pdsmbr&o6P`%3AZh)qEE4?YCc1C13sKDZY2-=fm$NF z*xuHb4O~6&q?BZ(0Qncc*)8onf{K+xx248Ol4O{=cmRoLVLDprCc@;f%tf^S389+L z`S?RO(cMRfxSag6)5?k2!hq3Zeu5{e(&ggagq@)hM-Qhx#}!9dPgqtWYYWe!hevP# zz?k^vuOLwH`Fen82xyI~LBMl(AgdF(AbZ%oy$jDy#0_0iVltO_;;iKn@Jp($YOxc&Eh)*~4i?NVbB zp~qmvy2@EG(5p#c+AZG4&X1MIlgwsfEpb37?C5_UC`*wMAdwct+lOdHn z{=PcY629c0N`|CwbiyOQn9wuz^CErt^R6z_@(NiP_q@AV4$+l51lmGWcbso0m4tJS zm3DQ(;SA6+d>kJ?_6Ab>TDlP;vtpL$(H!eXj~;!L#EGnwm6won^jbcxa-NW|*l`0> znuL4f=Gm`f%R}t)5;xr(mn5w`_SWxt&TA?whYeYYLa|kmChWGpT#8^mS-#h)U+1~J zcFyTu)Mpd;wv5vkC2aBgDC1!GoebLDtQwcV=TTdRg$BWAgVYgVr`*j$_;vt$qErEI zaFi0DVW9q0r9Av)U2q>h6mEwjXsmuotx^9ACIm+YYycg6WSVlLD2Q`@BmDsuU%3w z?H-(>TuuIJXb;m{)Ap=pKh#)BT^M%x`{g(OwLhJ?CcnT2mMZxQc@71~gu&s{-upB( zH0rkuMC@t_L^r%S#~NgYvY|xtUp%|X#AKDtPeJrxu`7p}s4nncuE}^?!-x2L*@D|w zpYEyiifM%&DW0cW4!oMl+qMt%9uLN&C_Yj;{!@HZ7k%Oq4pSAsR9EnS&0LB}cAKA8 zeo75vjWyanL#36WP3b&&rukrh`SFhO559-G58rs(6u}?}vP-nhIp~x1ChcmNv9Fe9 zQe%7L3n0BJZf4ZPigVm)AeyCiYH?7_L@ts2N=VQddIX{t(`T6iP2R(C_SD)SI3BW! zpuO^oqgMrF8`F`H*n75?O6e%Mbdd-5Q;}9JnieUflaP`EZtf8AG0&OJhlhCpH*p2h z?Xy0o$7ntilRVT^W$OltKge4H0|F+7ij}Hh%;qX%RQ8R2yn&yMjc=7$!La#N!5jZi z1;GOgk{SCbT`jG*=CVfb_?cTWUNc2_Xqal6vVJWBeT}S=fkDWPrzX$v7|yxyJii;o zWoTflyC1XnGU&WNGaDPxMOnE=z*@Q(!WOE&dX(bQ;^oeFrZmnrjI5VF;3dANgKbhM z&uS&ytD2hB=bd!Mr2h|5`5)AB0-Rv2Ua)^VrZ}EpF#rD%6>LlG1Zp|KYL1O%;CD~Z zqGJr^{{h7Lf8{Mh1#5cRK9Aqz6VU5FQz}q-C9@$?Ut3#U$8BvcYw1d1)9aL4qmjN_ zTbaJqg$~8(yuS+*EOm@lMWk;MHE;B?&7!P4TR#8uEYd#Bg!*%gcuvGDyZ;uMMf`kJSG8 z@5PcT(9#;F5xBfee2h9Y6}xOLdTfoe4(S!O6FF(OmPsC6e`e#doQ>^3^<(m*Yp~=;Sh+K~DAcKM!opBAm zkf=Zni?4*726J2MD_Rl{I??am%eT9vyk}zoCLtCwdoxHmsh4upV{4v6slGceV(>1p zVToByzbe9DsBF?~VX7yBBbj!iSZgx>`Pcg%oh{^3V;d#9frY{TJ|~V`Os1y0v#+}y z?-&ZrW_bR3_B32|WR=0=Td8@_Zz`nfOyQ8a@}*CI?|o6WyE;Ixr>fk(?6Z=l7G$u% zdrw4~bIcdx_NHc3DZTq0&I!pPwFrZmq;kE>+^nO)<$+S3B1ClKn%D+vb$jk~i+p&Z z=&k3O#VF}21U%L!;YXIijk0zc9@lk*zax((UG=Q&q#Xt$ zkh~4Z?gNWP?gsM7lX2lz7hwEj<<~==!n&NfeeO{9>~*;#1`ThQ0jF+ZI`rZRM#w})i?8*>_8tK7$A2xHn(kG75uN4yUIdTkyA19ftLXyjlBg1s!O1vCIl-y8vF)pIRs z1gNSbYtevrv>feb9hEf88MmM|G`u!6wr+kdiu)d3XK=J(aI{%i=}Ag5fZqP5wGhm0J=v*Z09R!unpA@?b(@KrCz?UEUFPTk#yyMYh7h z-vGqCN~t*59N;G(wmi!Pm=TP2SzU`g+7I#Ce*%q-<@%9HJ5cVT3ko)4{<{an*zGOM1ylMjF+R{DMq0peySG;FMkjj4Xrwu}d&^dcVX)_f^jv zZ={^OsO}atuC#AanO(wq7)hXjqz{*Y#KVU`8rhqChYO7|$g7K8H_WxQX~-UMCCl9h z%+A9@`2FPHzu+MPEpOA@?X>v#D$Bc1o&YCDNO9t2$*4Q$h-u*fa1)_G0W+;owK>}~ zG`fAp9w?&^>)uHwB0TSRbit~?B(ua6w@KAeJi)^I(aHR`qx%y$fH1Uqt=p%eIM3=fJfED;0O$c}O(Gv)Pwoz72l zAI_dXtJ2owfgWwYe)Br$3~rtOz1-bnkzg5rw^h%a4N8N(NiMH@Pa7_bT0$DU0F#6C zhe?01!Fa)EFqKH08JM@b6CV`C<*`!eEw7`k-Yve$ofeSxuuG@hP7e{;oi;d}G&n4P zv0i_@^kMV?2#7;-g<8R5!W}>=ul;dGuf2g0ru6gl0w8yMGX`$~lb=VPj8`5ERKipo zJpnhE2=a~sYZCCs5H8|k2ewXyOQ@wUm4S0Pnfy^lV>;9q4W1yRQ z>U|^N@8>+L5$AfeCVGUU70wp2ZhFJq8#nD9nU_c4--xPK;A;A2foC8#fIYh67hZuZ zM59+Eeo^ImJEnZ{6vb!jR|W-`=ia<(kbF!HO~%4GSsj0|f&}y_?d_4HgOM@gU64Vq z2(Yuu#+<6Y2QKymNuk`2dO!@H3ItHSa#zY>UJxQ%YQfk@{+8+#PJ;^KKH7VB*ARdq z*@0a8Xod0UG`(N0v*al^;Crd;mM(3R+k5_60Hot8cE*Sc)w2bbr{xfPlt zd&kAW9c=WtJpi5=UEtkv1AIsNVJ-z$gg0_&UiduhIrct0p5XMqY@zxI5|%F*%eJK5n*=d~$$ zXZp?6&MwjYQBsvcj;Ge3R1rPeX^r=KZvwOoRbpf6G_-m;r4iv8g`Z4h(a_>|kw#E3 z1BiyF?lM}9XU1%f=R}$DG9|&a!2tCzME(99$P%v%gX(k7sd8@!M4-t!YrO~XJQ|R* zmEjhcKFdRYuWf&rsMUyn!1R)-LCGm^4n&S&huralz`?&-2;PZsR2HmP zo;(MYX8K2>gP*nz5J%RriT$gQCjRkCVMhnf3OW))10{@?qg=b@-*>ppR=Be_Q9aul zCYK)qu|ccnCeKyT=k2Q0#?Vj}s)CXoxa-_e3j>G4z)G#Uo6M>77G$NO$3*N{jwe!; z37{>zYd6twUw#PCmXZCI5g?Q|OYL#}OBeU#DFcIKeWWP#4IU-hu*oAfM*SHT#}$Z8f2uT!s|*33c)wJTcJ}+336{^*;Dc!^Evet zv>-$51M7@&2{$)EZAI{B%(($w6#jj>;oi2kxVA1*k?lXle`BI@ZbgTHmiY~_?H|%- zabj`k#l=O~DA>H%&f!m`bnRqC;H;9FR`bh}$*n-av!js*PmXpfyY3)TjKv*5ZaKZa zm8qbh&#&EX6{*B4lsO}_YL1}H+<=i`BEWe!;&VOFvcXW30qCG(qfj#RpQ?D-bsyT?UNh86Fq0JSjx&dnhzixrv60a-= zjh2o_Lwe&ji%#t_oz-(1!uw!dLb08?i=_N_C#9L#%o{|?w~StYhViD+RwdKW7~w{I z74aW#w{88odr-%*x+KIgN7Ik$j=P{bkTmP_?77F2Bdv@y>v`Iqy#CMqoGmyp@<^5& zH{3lutV^zy_G>~4Lbt!q81z5&Y==gg(w zNPJfSAw)7x$iGd-**wuja;lgkgM}R<9w2iiO-iPje2&&tQp!X}{ZUolE@xEj?CPSd zNKx$PXgTXypRd}rYR?~G2ya>48xh4cSb0eYx=9Z4k$*r*2}Y#=flFkw_sQ0q2s<_j>HT3h57l1uOY7HIqN?9Gd6bSaz!wzzrEcSRLd|eYn*89W7dW1 zCP2wg^D8VYbWdD&|KKW*2LcYU{j+D!5*63|fHS|6s*402j3->e%tAW>-^QN$UILKu zfm7MeRMct~+w|gMK|uk?`%~bu11xM`<0V&Lf3GR4-H`*ZV+ySL?E1LE{B?p2cKC`( zgA|>g@mbyLX%kuJA|bsT`&me*=lNxW@ZVid$t~x(To>|tEK|98c%nqT4sl3@PRWOh zzQNV5cy6Eq)wwpuW7Rd+C`6W)YXzAD zZ1J;C9qPAh!(Bvvk_hYyUxqgKU#xoE(D2HUF>9k)ykg;yBW=>$R^`zF>6*=kSaq1vItCS;ARLTki#iXqjG1!V|UWL8@XzkzReR7>S@16IJg zeB+b*d8+Rt*YagSulG%6EvTp$es2ea;iAf4VCMWD#0EA_&X@(JG~m?L*5ZbHKOft& z_FSXwO)NTl4@T8oGH{Q76p{(Oz|wUV%+lY#TM$o%Me49sMlAXw)Ae4+LBOhlOY}Kh z!$d*1{P6XxEiFHKhXe-tG$y{y-jWjC4a>=~m6Gz`WiNs(8%ErC6kL9C+%2%<(A__u zICA8md6(AMq;zqxcELD>57{J5RDh8&!3T>Jg|H@K4Q;7^-V zGcbTfZM3b6n*8_MX^}<+`56@NLP8$ML;YqfOIj!;B=i8vvu8YHs~K%wF)=Z5aaS&x z{M9==Y$LS*<|ffSTPpIArsN zhw5k{MtXR< zIL+9p#6#$3fwSfoBqYQFf`VadPzZyReX_NBkbq`a(kD}5fiB-W0ou_6}M?dXtE^@H3HMO=DW@kI^ zhIr|gn^Rs}17Iwik`l~;Z1mn_&IUzz!=vG<+$Rg5rKD6d(9i%INgA@%>a|wpSjy&z zK*ad?IFzE018!%GvXu2S)K2(ihOJ;?M!gGO&3`(K{(U$WT=nCyJ&HN)DEP4vD51%L z$-?NlJ(?g9C?q5VtP)^t^4kMN2qn|Nk&XtpxkXyO!U-LJjL?KlA>2KoF=Mtl%*<>R-Lvd&3dfW}00IXf>~@Lz6&u0gycWbqOhWvl{%B22{x2=9vXYHb z+q=hQ>tlf(Et#vQ;x8uCT4gxt9hwgSr<&mCMl%c8Vv|_7>n+Na4(o)%z+vPN= z4<9_hDL`TgQ^1;vPhlILqMDlGD&ID8Pz!f=6L&8g8}72`QJr}GfI>Fj_~}!jqrn&d zdwC-8r#;&LEBb(62NhLUS^)#h;Bc2NqO?_1R$+^0YUU)EnXkBPn?IAN69v;N|6^_z z{85ZYsFaSr{R80n2zZ695jRD*78FF31260HXXZxU9FC5T5vJzI5BuL?Z-Y;KYiwm2 zx~eMoK=z*J{x%X)1>WwS&Q9%|GxvTT?Z2SFcY4Nm@%Ta5rGp2sktjzr&mC^=i?pz`)-6dWQ;7d|Ks+5Id&KDebo5({=Ippe1Z zL%@G16xh0f(j6LHU|u`W-%layaPke|a!bt3C1*mx0O?EHiq|V^9PNnq@gb^8pnyO* z9DB@Ng94`gzL<`-HVB0P1M6=6$n?zG+PaYfG*p%XCwmrNC^3;4SW)@;YBf5F&|i;7 z1O5X?%u@KjpD=`JyFi=H$yo^q227ugl{GCR;~9>=6y!3QdCVtEv7-Qstqiug;17JH zpy6t1U|;|c`7MpL2SDWptfioFd2Ow^p`nOS)>BVPDG-$6JLRQ4=fGaF@?gQb=X2xo zvW>O%63O#dpb!!i6eK7A025iZ*4LLk9-h<@f=VdtB-+f(bmSmV?cv5~@gsR#xmCV2 z1auNn)Kt7bL&BKD<#>&$wgw)WeFC;rD0W0$I^C_vQ^fzj+oQJ~U_x^dcGTljku$Ho%4(2*{4}*>D^RcU)~W6?;Xl1D zU8?DGj+Qc=(NNF`StPtFc9H2_b0#}`-Ma@9TGA-CfcvOEh0DEy2tRZ7w`*PRg055} zTad#K&8sD~2&|!NxEkc@>RP!!$#ss1XxobE7Ot3*;B53-q& zqr)wjA;CWC42LBayxK9_SrZ!ib zxgP9X=mo2{$;CxZOQb2=1S27q`?S*x+k;Z;>TTGW*yQghat!8BZ*FXaE1EKQ0ozK> z^ATLu0Ht_tm5vC*wrLgK8pNojgH?F)A2>N9uo|F?!9oPpfcqVY7Uzpc$?0YzlseM>C4*6ypMq-5pcMi$lw(#&%+H}+th@rat5 z`n$!YOA*+W>DY)G+5Z9N&&|zq&}T1!^b&L>tjX4p!~-b1@yCaBc4h|9DV)1l1LM(9 zEWD*c!%8(c51HP@YgDhmx(SKRW9@uqgjXAwK`GGR-#^CC4~VDj8-M@$R@kZl?5?or z(mbxou_c(N>3=PbEtC6bet-Y|Cs=UtWlOa}*dgSmC2PL~{piGzw)0r%h9$-sIFv?! zOF;rpD*j$w^*%eyHYS|OX@5>sG&$rn67({8AJVa~`@7^+0l1=vDKQGVzuDl=eD*NY*ZSt<39&BM2sW42|dwW0o_}~P; zx0ogMT*jer-ClmY(pbt*gaJ-c)+0M=Mn=+eym#+Xd5Ke8Kd$&LV&l}LqN-R3@CwwH zw!eJ-Y-4Zl>jEn$Lwm6s_SQp=I5Yui|6qIj)y>y_eqvER-rm>#y2G4w_&7;AiniOZ zP2y0g3SZtMDBES8d-q;=4Ei5c7zkQYLc^7Pakm^eab|a@y0Lz+%7QMgyfB+ud_H@OGH1{Ko34_xrcbh<4R&hIN&gQ{G#(-5!P_A?*% zAkj=SXB&HX9t>M()Q`fX7+^?xoh9Lg0kBX`=_SYl0}pQf_07$M4RCo#;SYHi9PA0} z0I}XnRyI%^2_&tffssSUf=(F9D;qYd=(+k~q~X4Ak*1BEor=LFT%Siu)09J3@1L~I zUPC2Q{ROcftg42JiptH+yg!)#)P5+2yv@S-p!SHhFf3kwUT!f=gBc)_R&F73SO zioHd0K0|e*8H1ry^a48FDNZQ%uE*#}`~ZJUY5{ zfEmgPmmMvKhTwy#iBRU9tfo4&uQmhN%5LVV<(! z)`~DSHI-qBx59|GF3eQQ*MKe=+3#GKB)a&`{=+>qe^>yRdcZ%LkNN4y-GpxO_V zY|kR5#>XKm>Ca(jW9y!nIQN$Di=dEDvM#&5>IaZF9R-1@_o49^kvR{pAyFFHz|U=8 zFhYl*&9llmy9CEJw@b4!LZ3$!53Nu-oB8x9{%dDfSD3Tbnf0Z-Hz=ZsNlB*^kz>Xn-g|a_bZCe)qL2qS)Gk;=4I^GrE+Qz1={|r~3!%G; zr^5(l*^d<{u?M%eOPgGfX6Ms~z=}a)KPm~-ss^>yV_0x6*eKp@Il=%;l#D2z&tmNB5MWD<4E3Wi?8c3x*`Plr8ld%MkNsz?2cXH^6j(-GH&ww@KvQY5|$#W?YoEYB33Z^A`n!E6W} z9ZC|9?hw4yPR?K1)`jf&xCxM#u?ebaJo+q$CPjSnJ5=#m09(7gyZiKv?9YQ)Tuwtz73^d_^iQYR(F|^n}JY4^+ z^-tmQAR7V@V7Y2VxI2Y%{$dvX!dL!aQ~o3n@YA1|`VX$<&jbG#N(EBhDk-S#fg=ln zRD7lN?w{u&vXL1?ztNohEX?$`ScXLV2l||KHKNs@$+OKz3hP0`1RaM<`a=@X6wRZK*8*9iZ+~wta`u04)ySbzNZctmhR`&I| zA+y$-qF5iUE^{~fR~M;qLvo<9mo@O3Il+1dly_Z zrXcZs^XBgBcnsz&KHi!(9^A+Ur~pC3{tdCaOVQST_ZSx!*NT>k)#TFo=(NY#?PEiT zU(X%aUO`HWev-VpYJUH|W_*0kg$sS>r&6sQoSmh74>uq{=vuGF?<`Q_tGyW*KrVj* z^_H{y`-HRI!(_J1MC9&nZcoFBwk}Z9HQn3PwebNG3dl(m{SKVryEHK~C4FyWRC5@b z^@LFS&*&NDR@RFi@`zs{2q8j<3W>x<$38NP!iAW;?dqEED)BVxW z^~%Z}QhqrDgCtTOSoeKcFDW7}K}tg*%xQaD0jFO++?P@8-A9nne9o*3kS`Ys)c zH$Yhovi)T7g6VL4{6qz6g5QU;bvF%FRD`BtV`FIso_;?$**)A39)hi(M>Jn__)oDcqCu2y0)5kZ+9YtNapP#_$df zP3R(vE8ay&I!}v;i47ZRdm$eeHwTpQNN;Vg?ZjnigPavLpj^KRG!&3kwV0OqI5VZ}h0E!9OtCaUu55 z`k1voJ!m?^Z9P9T zxtNvpUOiR*`#R&w?VL=pH=H`lDh`3?aTEk*Ap>9kj~r8kCcpu% z{YaLeaig0J%C;ks=946!m_coOW&PCT%S%_JOxyn~G&sGmsdWrZO^`-YYSY_4N&jLKGP}d8@=VSkbwa7KZs~?d76J zxZK3Vf-Kw(O_Ea8X0>Nbs=kd)jG?Q_s4}8)3H5UVKpY?`Kfi7kodvU}sd57_@8WDw zh&4AiJA-rPFEB74Wnq9!xFF-=? z-GNQ5Km6}n@*|X(SK4Pw`v~H(^pw%X70yuQzf4on zqh}fO)6=MWYo$B7$%_-LCeT)|u148qOW8Z$z8Eym_vQkLM}tZ!`$hIcAD{Q+>}Y71 zRx2Yz&)|sP4J?z%21RgGNg*hHmCusJQGb;;L!-`6A0F*4MaKbvR6_CkC#`?K6|wI7 z4q$cQF#TW%P7HE$AGVDwoJnx#|RhA{3@eg0-?sdvs; z7t!{xmE$r;nG?|SH%%^HY<3wCrP1kDf`Iw!xlAXIt(EM<`r^xH&PA*x0u#(6EmL8p zS`&HqZs4GX29)fI;k4Wf;|1-3KDzQDW}z63FpQa|D9!w4vm;XO{x6rakABc^Y`;*F zmcCxW!>gbmuNECsOukso#ho7$9W`3*w{V1$?X2Z{8Y)ev79U2-pWX}dsjr9C)8_Zd@r_Hn7FZhi33Q|O9GTO(crlwzU zNOGyUTnk)R)wW`#_O!CHYOk!EK6oF0d>l_kwn{)BP4HdmT-c{&@*PFuyy@An(xQ}l zkV+#xJSIzO?^!syc2$jHb;`vimV;(3)izAUzl9H=yYlVLjl|Cj8F|mf?;`h^I zlQ2>SUm5@$>;8Eup$M3n-2mx17t3?MmnJ5mB?YFOZ#AKT{D?y6HMqk_09m>JlcmRR z<_SGLv1`}TPc{EM_LcHn?7u(2|9rAPGYrf}`R7CZo2LCAp0NM?+J6wX|E{(_zotI| zX*1{#L9(T;PHYNTQ~08j?qMikUHO|JToeHtjxMjXU@5G4Xq- zu2FB@ek?1{SJTmPp*L!Q<`RgQB$12I1OrPw>L+IQ01Lp&00~tLt9!K8o_d=U3$zJ|!oc zALo}uM-wGk;66=E#A2pt-IMPiuC7gWC&DZrZRE#im?QheBDGfn5$jWm>;4xl$Vo^b za?5`KdY2uN-C*6o+&u1yDtYJ5)>g>rtCueU zEfRx*0kE<$_-sP!c4%nGT3+-%(M>7+yV&v3Ce+l_m?UqZrA4w1E6_~im>s*piiGte zWOa%mS>I8A5)DxwK4&hMy=TanI|!Q6x0zfsdeM?5(j6MepAB1Q`JI?aaL)QkbE~vXDqxuJ^*rRcuBZ9#LQx>UjDkLs$xG1asuLIt+ z1PvJ=T=K0KO-=|Hh(cAH z#{AqIM*J8Su;>wd3LG4qy;tDN2cLVp6b(bL)oJbCkl)brIiT0nuaZYPJ3FtuAo>jp z0)*8Rr;2v|Zd+|+r}@B-n4>~g9|8T%tC!m7n3-w5K7qIc} zD}j>n>=}fTrRqeetWq{&NBHQNy-~c7Y>3nKrO? zHsDzT)lJg(k7b^z;Pmr)dV5owf^>F~I}&g{k5;3%U|p(|vOefXcXr+o7Dl>6Xce$J zo~Io?+X)Vj3GK6-52)2sf`fyh7)Y`Q?L26W6u9PBB;J}ji?Nu81AEJCTCX_-(yoCS z)^@qP6DS6w-PPL9C(^VrbT3C5{PU@P`TcWz{<|jq_wo6kR@ne~!ff;Z(6RV`Idy;2 zAV4Zf9b6+A8EBF^-9M)@M!Q>9;Q(#-h=Dj+?@6IqWFd@n{BwUZnD}O#QHAlbW z5#tPMZBa@2w*YJa4T{6x6tulf*ROf5;?7@UV_Ug*`=Zu4%$m#|(@EtwZFunee(&75 z9DJ4ROmW6{lL-?+0DBOnd6Jrvo|%C{6`6g_Ld3?kwzgL9&O#vS>5+HL&?BwSLw!Iw za#i)3Wo1W1M7Vo*zPqDkKa`=I{%|QHg9xyV(+jg*dc6xooj{!$9fP9HRLq@>kZ|Mk z=Z}s+@-0OZc|3djX>+sQ6+&lPhK)p}sxTXs#DW@9SCw`6$E>cN5%`N<8oy}zDWn$E zlrCSnEGjOp9sy(KA_$R}u9R!eu9ik!^(c6)1MP-`trt3=9~V1Q8S2hzdekvC_Ll6y z3oV#xMjFesUS_W@uSBS63=R+P=AQi2(MRe8dU|?V%bzLZhz274Ds`o}fIuSHXk^)x~nlt_O@|mCd7`0Py6|%F4Wc4L`KCwb!TE=pyD5GD3!`k zum7RwoTu#YYA}|hIlf@`+d0?PJH13N6D*LV7aZY`VEyi^$aew;i^{)}s%FDhM<)*~ z;M|6sU)N!)uaD4#E^6!`y1Tjzt*WhWOb-*y)YOt+4Fn`OW=Vi0gf4czS-`LZsW%c- zKS7lK!?`1nX7jAAb?ADVbK$!$h+Y^tI@%VM6kb;Pp`H;P96at^R&K9nAfl%faIFlf zt(|AWJoX{74Kk#$;cxlP{sd?;n(w8roV_XTbkQMqRB+7_a}$pFd3p75ugUcG$}n7b z^{snloDcUgIK#5gMtpt3{$i`F%tu)nbf#8EUWMa-W|kUdAUi&W7`fq=!vSYWMp<4* z$6Hw~cjbtrbDmpIKvHe>W#h!c51^Q?ifU>=D`$^hg@8ca%L@oT4pG|nKo1ump8?=X)#foz22yzyN>@K^ ze%LB&4?Id%iZQ4(i!xtmbj_?BS|L|OSC=$Sk5ceMe+?!pNhRLTUd@SJlJFnenu^4| z5!lj_NVKdjN@KF!;dwDck>jML1-9t+_SSkvAm{b#A|ljNkABXYn6^qv5x?2o*bopE zlzSvW(OJpE!|JHpAi@86??65u^az#1!yQ<@J+vW|02k<)`TEh4IX;$nmK$!*+_KxC z>+xd@Z?8S^^v996_dEP=aT>5G)?fSj`1VkY@)TH0FO9P)=-6@ zrC{s29S4U>0+Wu*jnwq0B&;2|1xcxEH!tE+y4~@6ZS_K+X?7kTU*PGRt7MPD@3-Ft z31`NwZ8m{OqS|~vv6kkezDwHqXa_zs;|&RvZFQa1c_GNnUi1ClNQn*1qC(?g3 zIiHO!8MR8)^?HFZ=^cS>c6%Lb+65C?J%a2IHe8)lDyvV%rY1di1?2}d5Red#y(=qZ zw_29bDfrp*dZ_oZZD2n$rQ2K^eer^(A3>Gv`tAEKPL2q*NX+$~+Xc2TzLO3aUobH> z&&sC6sY=#%W{M#*ceb&yc8Sf8#|LZ)T_UOarM)jlK6Q;M4{o_U3JbfXx5lCl=q>rm z7Qx(DIC8%mICCA|d0z6-_OTp^RcQ208Y2P{L|c4pWY^6HB?(%dI+V#b$4t`FcNSV{ zFR*_Q=nW8%<|Vx%lg71ZlOE)jWY#7w8~OKcMn$Qz`mS)+HPmrmyLxj9znI>!l&3Xb{dbeJ^IWFx} zwP6>xxdfey&Ol#vQR-U+WUMYWF|R0bXwHyxE091E z%t@JUz+D;g9_7VARMIFzz*bw&zDXPGsIDEcg^{w8^hJoefv2MBj^W40==FKp*|1!G(dJlJ1#ts~8 zc0`?vB8KhS#A)i0gG>dx<5|KY;{i^5Ci}+WA5n;ccyM+QSm+zm275Y`%_I@)o2G}l z-gTcR>mL@OBe7|t4nz`1z}h4==pC7%)`n0H<)2V;))I^co;~<|l!4GUSEKGp=WX>t z3{G{0=^}OSJbPZu@{RZ9+T@$ZzjF|xf;#>cX=yt-9ijWrGE6Ki+-x1uqXJpkU+Fq1! zOOaqTEg8rsj67ONYMM3r)Su;;dWOmEwr#OwFA)OKG0AR~*&->~j8spC?gnGEdAWzb z;{_=}HaP_ZUIaPw!T#~n47blY(%+IyZjS4DNOIU!Cg}v4%)cw?=)#3)Uzz)#W)F(h7f`#(9Tlk6=vU*5W-`53Zn-%Hg5BKIPLr8Q?Y)4O zXg5+)GRbQaky5616O*9-7OjMz7vGAyp(gvoeg9BTk3c$Pr(kzfFIM`9oZLw$lYth( zx1EX8R#)+D5?=RPZt;!J+40lbj<=>(J1Io;Rf-cvIsDa6b{EeY8h#lea+gI1!7b26 zPK65Ry+$Esl#tMq({6t~wKzy}#rgYFh$%|M=ew>|^76jFF`F#Ft2&B5b=j1cm#0N! zdh*+>kvl*C)QAEcg~h@=!)wBtCimN&Q%5HuRpBO%p*PI*^YTn`=DNRbD|c2~MIk6i ziCKq#Q`5xTK~WL+$qQMnk>!!<`zLl-wcOy(^y%QsQDtt>(LB_ts$OP9h@p0GwlhYL za;n@uOa{eLT;c4G6}KqZ70?<2UmFy@T+dy8jUZo=>c@6%G#Sx11XUr%K@*crIJph& zRQBUlTp8YqH35$7q9pr_o*Aobr~U#x?bV&<_kM@SBa^Na1-+Vn$f*8&`3?*-4M9Vd zOlJ~L|8a}|hbyO*71W;#pRBJlBcG?j*E8WZ=)$L&>Sp!!^E#K?r+^&mGk40}8&^b9 zDrk7u?dtixlUHE4??CYK`r5)dd_MLFl$*DYJ}<2T-fs;7H?|hK1!1wxxrvff-`#ck zU`$?|F{*9p-sYCQy5gswS19bfRX5%x`)TKs-e;?qXUq4;X(&OL+M@&cdvv0&YpA0HPN55MdJ@fXSr$!PWe^J$cy!P>OZ_aU8W747R4|4v^uJYHyDtA*msD)Dxu%N1IS&5lQpeS-;O{ znVUe$oR*&@LMRPXh>B$GPdLSXPgG}7q{_PSz{2_aov9z#+S!$v;_5zQ9&%+*o>1qb z{S#j7+ig2zi_m<&dX#-f2MIpE?>|Fw98G?rvt2i-XUW2HQQ-xkPGc6-;M9ES^V$Vk zI4apT&Hf(??8lDiuGe#o{8+mY&MttdQzz#h@~tXQ~df!s-tY7k=me@K$J&JzDvl3mUOl*lp9UmVWnHrA_n=aS*m6eajHJ~5`#-uPv;rtVpCjzg<10?+@-$@h~o<6E-lFujsq zt!PFQ0SEEf`42YsBeTo^%3;^@eywb1c=Q2=9hXN#OI1sw*Y5`UbGk`g!IXmf3a+la zxcDXa!zAU^7#SHJ1+4zWD}2oF5!g}d30SM-GRG>k1o>)Fo6o@c?>}k!qhn&WbAPpc zdzL}8Y2tnxX=YYo_mtcq^jz^h=~pn>n++>Mmxk=m*I&`fO62zn{5dOl6Ipv1Cak)= zxwOm-4Dx#ewm!K!NMT{2chLt42T52K*}_474zjWj&BRPHGIh3V$8qrq*HeNuAPDyz_08jcJy%l?8`1%%t$#dnE;Av6$i7ETrXo+&I$G-{a!J zy97&FYT(Mw*pJAPjw(%kWGb(5NIA~;+lsReHV#Gtr+a{Z0<40z8#p`sBpqLz6j_)j z#Nyjof5-Yh0jr>}V7kIY|Dq`$3HFOa01Mc)dSS&!Q*w6FGe+d;=ZENWz>s#^$$6&# z>7AAel!2jsdWC63$77*xVv>hX)@=kGL?6tCrWJ_JCHUYK;&@fUu`DPJV#A-zGd#kMHxP_26}3WhiQI& zcF5l5g~?euu^2i?9#2u{WqL3bfFG32t7plG`iF;Q?jG~>g9{}8yjnYIH8xgEWON1q z0$rP}HF;JKJZhH0Nqg$8@Hl(My0L5JY`M9)+S=Pjdk)O^d_{VC&82)*2co%TW13r< zI4>N)HH_XqBau>{GSyupP;}kvLu^SzL}7c-c5wo8enyQbUF1oo2&+j9E#H%6-n>!0 zHf5Sb(|ayFuM`4|(`Sx%G8dP68*gxm(4ank{ut3-Yd@Sf1h~%8UmFfr?VC~2Q5%0p z_dSKtM8v(?kCcaxrLCLTN2kW*gv=ULU!2VjBNkK)=}vgM+~C6K`)BmD3zBk-z>u6T)JrL zP|^pfK-i@?$FcOLB!Wb&$2RzxhT7VDjj&@(BI9gE~Qr*;?|9RFPRG4hxNYw^;U}{7TY=j{NG9h&1t*X10;!Rcpfh>j-^)nol~G zhBKpGT|fhM3Vw|Ksh+KkhB)ZsJhO$U6iX16QF|10^SO;SlpVA5II3D&mhR5nPtg%) z`|3Zg4rP^SDO{^DytmdYvL9n!!{bQnA~)oBm<3G3)z|exQVh)W`}@ak?cRb9yjiO7 z)~POc7@n>&97sw!usoXa+Wg9iJu4dcdgistsiD--(Q%I!wE}fLAsN|luEn);4=_UM zv173tew%lN`8A|pr;51dZ^YN&5K!=%7%@asT>C!SLld?dx|Xr!L~Cy^Ldz8rLu>Y< zlx2r*%cJa~N_?OG`D482RF%QyDGkOSbl%gvv?XrYTxB~sk!ouL7!$RlRSF%JLjdLP zy-|wr+ErI@gk|>JJY@q41~fHj%L}U$ZogOGUgM1_=R$OB(iAh1k*jRyOA-nle=-gr zh`MHziQavvzGkgPNuw+}Xb>y=g!A{WCLlRQrsqh%q&d%k-OQ%tEy71n9g1t1k;Jfk zRsoX*ORnLX14F`PH^M7@m7BI2uk>nvOTNBBX(+NZiasS|v@BOZF%+Pi=1Qd!MV-}J zafa{KW$tehDIJ*q&ah3r{y<`i@47YbV& z&i8VAPQtPJb~uJqddU;bZ^v zX!Dycd>FXtb*g=6EklDp1AvO{|LL&@L2;_z8JhFY+x-W+%(moD7yjSP{(nzR|DT%u z|M^AweDZ;{R9T4#Vra|08yGli0N1b-y&Ql5$1h;uk>+q*zYa}0kY`W6_)#+C7;sY{ zE8R4X1K8`rf&v(e>iPM2dtc(E5&3zs%dP#4Wt}8TJCCvYFLr!0ac!fJAw-f>cLDK;WSg;Nn`Fnjocj0e~)e zIn{3{$?FmtALMZ$DQ}UQX9Y&Q@%4xOFQf9PZAR!zNmEu z8oVlWU&8p#o%Q|v*7epcY4~+R%D_EC+CGATw+VIA>r-fL|sEqXB2;ma8i% zc89>xVq9pb0-$8izn;2?)&BBY?*@J_Fcu<|578x8I9=@#YW9ihf6>SX5 zYXZSj273DZn?PL&wHO{Aek&UGv*!vcD;D-e(lh5BN5}3Ulw>gSi=Kpxjuc`UnpJ4% zb!UpZhhyF5WasuSpELluRIp4N?~;Pd2R}+rQqmb{Uy-EEY{T!~Sp|FsWs3rB-0J^O0P|K~fQIj}&z`x?2Hq4FM4};y zk&fb=1e3%`Rv4k1{A}U-Sfndw$;nk7?cMrDQ#AAT?S&qkk1vDe*I@`nA^d0pHUb`- zz%9?nb0MTxRMuf~1xH4tieOBn#=GtaTP}jk`6v|bYFCcU)lj26xXuSj1E2>o$ItP0*9^g|GiJyBON#|nC&qIAgboOj; z8{73FYUBK)L>iOX;bF2F+6@_1eiMb3^z+v$%v;Z&hHTS5i7ZHB{OR#ZFm_i~xS-Ip z7yDXlWTdR^SWhX3P|{l&Iz81;wnyFJa$)8UA}&cHG>)Tgr80r%L?uC|aono(-uL&Z z$tZTJzDo>{R|Q4Cqr7H9x5KK;ld%gbqHyX0D)u$ulNPM+Rw4x~f`s`^rSaZFVXN=H zK!xPboej$Fu9sh3!-NS$`}d4!29gRzNH{MgIL$=~$#C5p+vWnyB=3rOyDEI@l~!g^K9l~8z^_y+dd zOP8X1RCG90_?L(DEYnQFZkka7z5zy}Jud+yRMZapoPt-60ulUfj2hG_9;b40TvO_j zM^e_@o7cXt^w5vO#auvZvf6Jh zpD{rf72%3fNr&xfs;VY3fZ_OO!e19L(He1D?Z8Ft+(DA{b;WSCg4Bv1A{h}|ck^9P zvn4|C^6@<_31()q=M27g5%(-3M7m3erM7_Q*S%(`sZ`HdW;k#iYJlF$ucX~Dv=n*% zG7(}4R210LfNpsPzqlzCt0CMsF=qifFSm}OIYRC#g$yfRNFbmW@hX`uEXBGgm47%SuQ^1n9)g4UKl=)Z^Nk2GdM`nY%XTEAJyDst5-Fo zt#T7NM(Loh^!x?Q*bH<_)xegui*0D_VwJizJ*Fn2v~ zbK%mpLL`1v!WJAKPgddbzPRhf7%c?oWK{Hwa7YE@7Ax+cPugYGrKE^NPPlMm$vT*u znNe$9SSjmv1L*#9DcfMj6pjrg+ct!MoX-~zXh}PnwqvZ!!AJl!1EpI*} zabev|A1OGvo&i_0OoF0Nz6uTC+cG}{P7$;@eJ&*__I^ShnpPL|rCrs?zS zikqdq7~!!v312ud`tIS@ z<_#lLoSQj1Z%0En$eMEtE$`jCCt!66flyVIjR^yt`^spEBz$vzX?Y(ouP#bTc$A;d zCb4HLO5=`QM<=QCOO_>G$t#NT&Zk*HTvkzu=yThlx7;-~he4)&owwP2S>}5g;h%HE zXV+m101}Oj;!HR)#z-X2ZJ^FP!k1MNbD7OkMna=I|B9~sbt}XM9+R`HD-6V)_JcGS zoVYXqW?ocQ=G4%NJC^KmmI*Ha_9}RKf^((mGn=Tb5IrIU&9Swe9sZPDX(Yj$mnV9F zFiSpU_m^P9MYpvvt@?pUFR-XdP+V_FL}+N_80v6hY0GeSx^wR|K7ra>`4{lN+#DQa zUB0=xxydYq=+UAsG{feaXGz7UP*;^XAx=>;#?TiCvA(~z`A`u}C@@s{1Fe5AW zkO(AbPsr)*dZ-R+e$-P>_*KKYqZ*X_`v@F8B0;Nw!4mu_=?!tQD8p6@xPkU5xaYwZ zdfuqsDJ9JsLW*!mF#g(5$QG+TB#^bNH24>sh;Au6ThVUCWT^RF#T0swWTVj%QkdLR zYhq!6IOd^1WHA;XZl67S#wwqvkn|43N(2(=KjMH*7Z7(}&X@FcFD)^%Um>X}EK~p} z(8`@i7WvDapDCX<*$pxH$1j@GbiF)-gR^qYgHF-G#s>HK!QP&1l=V$cLNOAU`v>xb z6BP95Xw$yDDxW;U)PsK47>WHM)CME{7tZE?6#wy_Zk|gG_!B{TkH>!CS_LDc%rG$fsU-@FMbd{BX+>AFKNsw*TAfWrE4i4S1 zfl}5oEK&uAb=EoNd7=n!Kfm&3y{Nc2r5$Y~6LD`?U3q{hj03s~C}SN23BQ0g9&7p? z#@Bc~AS^XXSTK;6G^&&)hMz($b|5#a`tY7r4eaebWz~xe+Lr*E`&vJCEA`^L(=fh? znHd#Jt94eY+3Gr0EJbb0{4ksUAk=yyY zVt_m*YMiV}<3NYIrmSat3puKD*z}DI@hx_XG||cF3WtmRj{@%eZFa9-UrtgH-8SVP zT~3oW?EJO)`Erk+oY{YuvWoVdI^U(=<+ALE_fz4qjxin8*k z?G^#e0eC#MN7EFHS~q2M)jnTJjk&Wwa_iM);!!ClvP-c40HCXe{3ztXre~%rQ|Ag) zst`Ql^~TA!66^ni?Nb9R!VpkPq3XuQ#s*0{b1=_=D!apeM&;eWu_i1zdh zqGga`VP<5`=KXXeCI;!bSIOlcCnwhzmq66T%_Itt9M7LUt5P?Byatfi_oV(@g^WOB z2%%Zb%wX#le2@>(-9{fETyFju=2y(Qw~G5V&PP4x2ng;F%Y6eGwkkPwmlba!;}HFyff zl;Np3Fb$9fNs*EW0+YL6tIR!I{)Lx6)pRNom~sb|A=3<^smY%WQ?PQN6#xVEJtrgX zWopcI1;$S+s-eCV|Lheif6v$9-V z9nX>w!Wuy`C8sA*-aW4HkF$YxC-#qH=+}Qb_^PF;27s(lZD{mhtR46?5Qg%<^~e?X zlI;rffC&oFC+Wo{zmsoSGC?hkjfUXo*gw#Jyt5R8`|JMwNE?5Dqc~Ins^qqH(QO?x zK;9?ulKfrfsTmleXJ1xaWjeAlO3SUTe*XAe1Hq@Ifw_Oc-PJnC(9jUzwcN@C_#3D| zRf`HaIXEPalaP_k8=1b)0S{3)r>(4kh6kjwaiAaW?|T}(m%IpbTbqD}#G4m!a9|o` zXleLW+uvVF2~T9!e`mhZv|-S$Ec!0;>+T#*d7FS}rIVdCfDlVn-QDe>ht6X$Gc%Q* zk(MVJhKC10oEmE{;padaxGH!?z@pJ@WOpf`xf#4-Gg-?!I-Er6D=G$o?9&z=nZdw# z@d6BKzVUj~J}axg*PLkMZ3zuacdZdSa`LwD+2z4rqCYFR<8DTaL-(1q1>l+&^8zrm zsC; zl@%3F?(&bnR#b5=MSr7gjc300%yrwK`Ir8aRQH;CrEveHr72rCrwp)P3Hf@sC!4hz zdbjCo##H&r8(lI9ij%mO!7_AlF`uU9>Jfkwb>(aJ<<+plUOnC4T%g)xw|$Djw0<$G zDhb7B(L^zS)vpuAelW;4T2@-oP&fU{g8j)!hX#y6GRn8Ek#S>Nomh~Ko5lTeB(mgX zZZI}B!HXX1_jYzRM#d%xdBpp2+KyiPYrd{F4n~3Onp|rPh>lXBt3sd>przR%> z3${2vFL_N6yBJ0&?W-0O$H%@yqCUk>4M3hN1Y*C#&kdgKAWYu{y2sTfNZ6S70K$>) zD>*;k1H^J^qI#sX2Mnf)&|M=lU*C0yt{%Z=1QXZ8^p2$bl|a)4-kh8R7L~g{Qn(FB zfUcN?Dv$uO$&<@%brlsZbIF%M$199c>Y5~&XzA)u+^?IdfX)d$GrgyuS7>7SQwK-4 zs(ciCP1V)Asfjk9ib9k~PC_UYMoMm_~rRX=~O;I!3^LojIACGt>< zj7j46(adx&{``b=t6zE58$B#6S=6t-bSF1UOBuh2p8*l=yQg0v^WHfJUG_$|w=Oq$ z1EkrlC&Ed=8sp?3hM&4rCn*VLe8E!?I&ABV_ z7o?qjyW-_Uokc9RIX-4EB|i@O25r8TnUBNo-+vgEN_+bB>BvY{ZL-B(T8{~w+>w&Q zYz`zg7FPNRGx`vU$$KN6w{`st4ZBXCOa=p8RNeXoXP0h8fyQXYqCQHcrMudkFT{YzG1<_N+yoVdowhJ#C0~Ys=yI%K$rT%$$4LtbU z|N2=Cbq4bZZMfipK0V9_d#G_4a}8xt$(Ya0ihRJDIv{%OTJx;MC?IbPJvVaf0m5Bf z?QUlm-l>Lp{;8*sK8}NBA#pZb*+$ZS% zwSZh`wM{?&$%_M*Z%chp?{cmXw3ClNW}?efQU{LC*vz%60&#(@Xbx+UECN1s2WLVvDmAAaTy;ojPC8A|=N3Ku% zPL)fRb@M+y{SxSE@`Zdb;O*|L{pe3(`>n&nx*!6dyt+t(a!)6}I%G!AvK;2SbVO@F zcM1%8r`PK0M5GCu)SiQ!1km9@kPPZ?VLbx_u8}&E&gE+RBR#l9vZQ57q9sHVKP5y& zil$J5V}}s=@!~aPQP7d3rlHyF-rK8r>Iwsjv`&ED3mf5pR-#OuGk7W|-Ra25B<1 zA;|pI;kR2DyyVM5LrVFfSD(hY)ui(D5!j{cb931a`A#j&TZSFJV?xX%7|=N5<2Cqd zgOI`;y4%4|q+#&d$bcRV+NUdJnZXo{63CJoip|Tbt7otPtoGd*_D0`iD%?!%^0PE| z03hll_C>Ixm3{9d6Z+a+{CQ?BwOfc ze=%LR(zxDbXmN6~=#vO?GM%CMLnB#hL-8#z20uDHOhfrdFhy$!-%+S78~ml2UiLNf z+G7{7eEPbCpQ5;!Wi>FQr~fplzpLw^?x;v+pl)1=1c{LvcOB-%U+-oNEk4kn14?0K zTn0D6$&VlP4X~`wW&$-1!Nbj+`L$;ZPe}jHojX~YCEuBrKYl@9gKa&O0OdMLb!UNr ziD|MyL#bR(BmTfUPKWlyZM!S+77%Hvyy=MHKdEck}~~ zHS~#4MMY^ps<{mGlbzHb$3sVuOTkO0Jo~(%>+s!s=ZE!Zs;5zA_pbZNg*23ZA40-# z*hIj}l%bO&W54vuk;FIaP{{iKycs<*+L^tR8TMZy`i2P#0HEbK5#ixyo(i?w+m;ax zt#|+vtsG0ghDd39;PVF(dq2Og-iaT5mU_~iIXhA7&2CkCaZr8C&R zHM(tn!<=()7|`364ZPR2$tc&@fl^{8{ltIw01IyiG8OBo`!vB+Gm>K`tG6v!k@}%! z6BtHR&2$AxLym6P@ZG48-W^|O%6h-9@(RJn|9ZKtwPt6z6p!O-?R2kX>~yZcvwBh3 z_LisjJDY_cN}qwgD5zNOEI$g|*QhtqSs2gQDN)u>0AeRYjr7GMDu5fM)jM{X=y!cq z615y){~I7~+m(H{Mc=rDr16(*fb|JS7MEnD8aEFoC*iQ(m+Zhe-gm27m2zL4K?LH+ zL;*C5edV}88tB`7%r~M1@n0vUE@1vz&hN4csA< z&>YdLnwSGt=tyN?5>7pkGko{CVS(Jd>p4Ti`P$U9`HGsFQORI57xL59-)G!1tCd6q zq#4_2q>E?OGahGUee#xsMl;U&#&4(he@GWnKjxZx7yvq+qD2oi#q=^ctcTF%yo`IM z1tOuoql0;pqMeM}u%o|J4Xrm$y^xe(Fzs^h$w*79aib#s8{G9gyDeo-s$Ng5?YBkW zrqUp=TnMx{oB8l#YSy0*D%uBGz`r6w`k(OdPek||9{wwC{VRs0U;+y(5|cn+7L_6> zk@`-5{`#c_4r(4dwQjIX4G9UkoQ#pDfy z{@&;%6PXGqp>j)VwMqzMLZAB2J7@k0n(!xC%q4yJCzk!^hyA~P{`-_a;qFfi{`Uuj z*Z9w$|A~m0H)8#-&_^qrJ7TD7HfRb5lhsYzvgOa1vT@_HWnF}ujg1Xdvcw~69unS> z{+SRl+r-khjsnS9t{Wxh{`MNg-0zBt3eY6$+ny0MHBb(Jy|yxhkA<}-X;b$4%a*ncLy%)SJLf>Cc>ANuhXzt4-`r{K0`?XZLK_wU$Luzak3IpIp?kzxK!8~)mw+z8GZfrgJ>S0)qC8A zIjLxdFDgHdFJI;W68eMv?;k<^Sp2s2{KCA+nrLTXmPEC|o^r(>h2IO$`%*;pr*Q006rTp$6?7yZR&dn7^W@cs_2kpuA z6{{bgd+z7946@!IT;Yb)C<@h=mpe~I(_}Y{oUjZDrk?;&argH|VN}%FZR6sKmcxye z7av~3T%38KVY73Zq--MmEJLGUh7M>820+vPp;CB6M5D`F7d}B!V$!Mp|3lqd$5pj$ z4Z|n`3JRhE(kLR*vDtJfNcX0@yE`|S2#BbNut7jTY9rF!AR^t}BHi8WZ*KJ5%el{U z?)Un8|M|Y-Q5I{hG1pvkjWOov0pc-10PANF8TnR2;n*s_#lA`cVRj9|0YA4DUB?u2 zGjsc~d{Cr>#qbWmRZVL&CmNZVho$xeN)mux71D_4=v+>>{b2`yIapsh7KL2b4As;k zsHpe$rgr;3@6Xi{V4nD9TZA=FbMhTblNb!JsX zfCq5)DoGRy`FheaT8{@(uZ-rU=H+!a^?=G10SF6y}GLEKo%D0SZ?6I9TAbP!{e~BxR@8Wz3*!PvhhSvS|$Bupd^!SU>&^D)!G^f z%EEpiRI9NcEd}yLueDyQmab~P@nG((Uw2hv`O)$2SaW0JWVb)yFmy5L;2wcOhp=*Z z1&CTOGz5J|e|!h1;H|oufFc2vXD7!hoIF-k)zrbafx~B7z?#@7*jxd$2r~N0+_$ZP zx&i_!H=SJ+sYKvHw`&h3DnSMO)s<1;u9!A619|YBJoZaNB`~_9a=Bh7P!eJ|-{^^U zRsK@m0w@rT>Pl`{lrTsp&DPofa~z(s`~;NdD! z@5GN5RlsLK=^sFlm(eGg!IznLWpzZg_cOzQEfe%a8nNg3rRn4WeGE7=0Ph3a36QM+ zrt5$rcz>kb{B8fj__f=;PS27|0Y+}`iZKn2myu}|6&0tv^tlHodVh$NN&(so*vNsA zx&CMa_?v-TEvKlssQ}2_bm3ZK5BRd6ECHx4@Rv4qx-D{U2bwOlw8(*i)nLjX7|G3T z8$_EfEq(TCnszQ3wQmL0x<93Y>T{raMFFhF1|%DM@W%p`4#@XZVGY_(pmjhdHSf5~ zRo6oylwHcj3)?ORU2!~h!1AIWtU7HBiX4NYPgFNZNGwcEfpbJ>`yR3zC?j-fw;@0U zi2=#x*~Ght$J*);ORa)0b(0-4s(I%iQCQF?CmE#gJ_iK_0WAxnw%9+z72_v%I&fE` zYA8DcLprE%bi^wlFx=T`WVY}Ghu}u|UC1-I7(@D+oLUGnbJkIzT*j z1FSzL0$m;pL;^|JU_Z14`G8|%6KH);URt`brR9?{*fbHPrVuw!0QJh?&=A=gM@2+B z(8d5)-w6UbG}u9bv)OxTb026%s>mHu5C=nnABb%QSM^`WIW~uR>#D1BJESmr^j3ph zgkZI?-c#JmMOfqOftnkA>Ef5$dM1*PxoXmk%-f#6zPbhJDl9pjFtC$=jTv5LAX$Af zS(%)f$wC7wXahMTkZY{c05Jh#ENt$Br1$((-o zjg&1wm*k-k*Xg+oOzc1mp7MESBbu_%04&&`zD*`5OA=hfb&55fDZ+7O`mUgEBM58A z$x+VH*3+{yL)J}Rj2eV<_o}8vyWD+&(5lZMEiElQuigb#<3l=H@&uOfXnh&GXCBbg zaconN3PewL@LeJdpbC=|@vhLfQ{6r+ zcQ|NoX|dFb$ky);I5-sw4PxX#og7e-JuNFMepd@vY2?%-7=_&~1Mg;UzJ1Yfld4@t zVq&c>)cPbQKAzff83Y7oAiG9N837ykfLE~UX=0Le*BU!i+@`}nqSf2buRvzg=H|}u zp-rsz*AYd4r9i_y1~v$^er&-RuYm+Ki(3m8GLV3p(IzwX9gsM?kMPsM_bspol6?Q- z75$**>8{JcikYC`sK_33THi;_-pfGOCVR<^RMVnf1ci0(X)X6tSq=aNXM`VcFTfXU zQckYkg?V`Zy!~x50W1T)`26u|%&B!}JGHe3kPsHDHF$Gig$n;iv&#R@c>ekJg~ix( zP?iYD%K*`S*>(m2VuD&IT3UmE^t1XwA=meB-@Zlj&wBa6nRUbd%l+MfEOip_7 zy!-G@Raf=vA`q#8t#J>r5EFM|IbI27Kau*1LYRVxbf1b!!k z9R6-QOM(dRxsI^#bRhi!p>N^}kSRn{$f&|BHot{}6z|3UFU!mG`{|s(H{=TBxpJe^ z$t8iZy~j|Sa($4|=<`KVyd@?bARqX!MaCvjF+ICwz?Z9Sqk#d4i-AMABqsW8;#TjC z!#p6XoC^V(SF1U`vY%f?U-?pIQT_&~3$*OE#tTIOTNBmXV9K{|wU5t7Hk%b1v)%*& zQTDdBJ+?dY7XT~>I9{*>%F55GqpJ!pWQfqC*y(T!kJDQP}#c7Zn2L|_>lSFTPP$RUh1|vv$f@F z^Anhs4uy}{4ZDFv2%K&`Pe5@PHer-*4E(T_M?}PBX0u(`#@-s2${JMU!137uWidco zi3c$q@mCi6ZCrp*zF=`>bt%$-O2FH7@jZ)mCu(J)<>g~{P^fUWEiSYie%#%i3T;Dc ztzBSuqk)kqrXbgMRtBfGy<3%fJpt{XLmadmTWu>?%GzJj}62PvWC&CN~c z`#yw+=K&AX=Jc^cYfl54qaBjdZ5ycu6#tzdvgM)O@%Xs7!ob}?YF~CHwrJ4fwThXn)`L1B0&c|bQ z!O7_UqqlCBmiG*&g9kzv`}=$QD!KL>xd-hO}$pYMw!#z_MtJpOb{rY zmN+6peg{AHH_!R6(BE16wWhbx^4ab_0uoQbdv%`~I`&>Z<}*QSMEY#i3ICq$!XR7a zi!v9^L2wUm|1O&}RFC2VKeCmDe>%F-z-%*+hky#=^Z zI~@%8tKYq?(5H8r;uaBf$h9M32y5r;M8dFVHcrt;*DK+4{Pd ziw12JY0I+$GXM1vqRbQoW)lS(MUDwM)t;3&)jbd1*zk%6wteu%oZ~AZpeI1GY$QIY z8vVXCW@=?5sM5)#uGG!D4_*cjWoPC>b295r-9U7{IMaKUGScfNRXkR*E_rw1`3?r) z1AAM8VieR%qXHnKqP3&Hzi8)pSCK-*-u?&K=fV_oNu?BPK|BlOT+U!VQj;aJQKAcag?$F*(|>6k zM>2Q82d!_PcLl~@)^xe)hd4kVKLVKvzfE9gcF4c`Q%o0ZKBPnP@~*IY+}9}DPA#wU zxAi`l-B$6g+Lx%?X>Yk80_BA2%{o_fWriszlDO?^;$#=67Ww!d(T(jl3-OW8xPGYu ztv`zQ_MU&EU(EqUic^b@C^z@H?~Y_Gtoqj!xV;RW*Oae%ygTr^zf6ds#K!jn3J3uJ z|0WU0kJ@gKouXR_eevUg&yNkhZiA|!4>vf|cLm3HbZ)M2bD!K=a#mJ!Sc54O5&2&( z6Wm~|JKV&1Sd~X9Ds-|Ku|}W*lQig(5R>^(#p^#APSFtb`0=|y0;ri^?Qy8bs{Ed> z_sCCoR8>~H1s6}%M2)V?;f@YZqJU(B{(+MvHP!xfy{y(UrATer15|nR4-Z^S@KO8u zVbrf)9?||dh~6vXSzKt1&)x`j^Y{VxVz`r3;Kord*J{7JTFc6Fg@BD+QPnI$g{2~s zWiP!W(O9T=b-lcl5FUltj?C!)jCoT_&xIGV$a5>9!Npm>=|E#VXxLMvQ+WExyB-3S zZox|W?Xf*QPPdlX!`WUN@~IyUv)6XbEEcjkp_`J8J>*cR!}!Z-)K?&QEC3S~Js$4z zbsze>CqzzuE=QvbRVaFY!~zZuYPPn%yzNWtTwKGBHvWGWeRH4aKu_WF0Mhn1MMLPWIGS%p??W+4JzmyS`dGqo(tue zZ%rc~Rm_5@W{ufdYP#%>RhihYNVngcpC?`O44a)U3I(|~WokOJOY@Y7xL04kXt~kb zOE&onZvezm-P!ezic9DldKYy=!$3q3YG(H^EXl?inmV)3%XApelUNW0dYU-QeBq=- zL|1BSdC@>%q&p`SPuW6(am#$*hAFT1eLX-m{V6O>obiQeX+g%)b_2dCFJZG-`t&g4#}A*ocV}8v(-&4z+sUtsQ|`KK z+4uD*93IQF_SaA}0MiVeP({jvMMVd@(Ws~`e;#A?hI)g{G)J!rlwYs6H@MmIh|Eq- zUjBht-N1-e$IG3aj$CJ#caI@zSUy=f%90zQQ3J@?0Fezlq#V;=VJYXlt!Kk!xDfl({mQR8r#bbwroapQpS$v3;u1-l0lM$D+OG zW#P{ATao&$Joeuzuai)#=~4QwizPL?G4rf059si>Mdu;?t50kfze~BwxBkpSi8tBX zGyjs0gH@NV49C9pW^PCrIPn{phVAdGE3PQYANyH?Vqe$Xmg;}Jg?a7Yotka!*%%w6 zC$V-HJ$}_ z&{N`%>!!co+NmV*@6Rvl%ko3XOU)&GUebX=6?cwRzSq2Lnbjjug0io*zgW~Bj_cuS zJt}&!*qmsRe^Mo`TV$~?IU-n*q&`+@GR9a{ej{;eKCdE~GBy_F!}1_@|GxIn$xFFR z$t)>flW=HVDnT%pSaV*a*AWw^^Ak#d$91Cg<2@zerk3zV`Dqj zjX{mE3HOZ0TuwXlj}gnu+G{P{Z@%{QGjX4|Z)+dUr6Ibw`H@s(p1X0o^ZhCMKLSs# zuEmUl<;R77$t+VFXBY55uDzhuxRUZBH%`bv_p_f|6q9JdLFLaIl!1Ef5*m&v z))u|-xa?~2d}+==Nlp+5$Uc_#l)_tZx8x`|CZ~!0uzy+2fZD?Jg6TpPZIbt_@P(psauU?ovcS3V zZq(xWV0HQ?%;p#+5IuKoN)NFI5~OIKgtghiBZ|O<=A%oz-2+1WJI%7UtnNeAIz}m9 zr^kOv5z2km(ed-swYki+v?PR9LvGx@%Uz!S3ys}!Zm4Rx3N2U$Z@$5}<-L~5mgau; z&tXvwzLhpa5jrb$Y14p$jy~0OLU-;|$dtw{wUrj{1tcseiffhZ#yH5OoCx3H+WWTG z)Wc(9uZ`nq>&-t=p4`~8l4D`7rajP`9Au}M<({3ET)-ATGvgN~*GuTR;g%=hao&B> zQ~kY+@ZRvTcf|@6>i_!f2H;**_?Jcz?{rUWWp&0X{4CsTZ5HaImmiVLvQe~r zR}xm++CJvKv!ojYF-$u!d9TL{ds6jaI}?FS>yDzevBorKjgQ$6?%oHVHz`vuOe{9J zeK^ERUdWs4=XOd#{z7)ubsw9t7>{K+>E6SyBLg+|y?{Cq2lILT^v%h+^aedmi%UH^ z#&){pCvH~-qx>Zr$|+sCbzgsm9zl<^pDa_9&76#9^cS}JJqrd_D`0p-a!hX8mHMn^ z_1|9Vk9{St=ui@w?U>Ru_Ohi)zam8{Pf^uMGoG@_FvVKz6&9pd2)aGHmeHJK(qoJA zyJc+4U|<9##b$>kw!2ezo4#AWtLMb)krRkQp0l7U^uI>suk9j_ODOwOet=Oam2;WjCZ)#~+Q2jez`yE@t}LAlrZos$4-zdL<7U)e)nH?XeZfvRvl!b?9DhAid zG@rn?wO#FENoT+Ej5CK-6!0);+4RAa^Yhbk3+d=I@GnPa(a;s;SNhQ&Wc+xm0l1&B zic4O~s+>&UquAKPOk9hQV1FNmpQ`XjeijCTFuJmeDV_u)-~iUZ7aqhr?yT{LEJ(<8 z&?QY*RPeVo#Ontn@G1TzDR-NaXwMse0`Y}F0wH;DYWIC_`PR9ut?*v%qN9Bo&o!Ro zZS_v7-ZI^xnwiPTFiuXWvX6D!N8$ZeLY_nvA>o^jo+U)tyW4>OkSp`_eDG$9-cHLw z-Gpyf40~wLq{tl{OXj-3olYJ+==Nt@DAv;1%?OFydT>Z36g`boJnR@e9z*M5Q?*!~JYu$PN))QCOJqkJv3~*J{JKo}r<<3(w3#JR53-8+z=Y%1pH)tpfjpA8oD1i8`uw4A#ndB?na#{fzlL%;IMhQY&y0iPTwV8} z461ZLo}TGacg?=Z!1S?}){ZRJkc31F)3&*5OU7P(7TOjj;+1khCB!(tLbV2OUR_td zrAlyl1esp4#4K-Ry+A=8vvKG$JS;gQtg9|`!_BQgSSRNF%6ibtzCIf>GeWcJSA{#3 zK`89Tr|qd&mlP9{Z=h zO}XLjHky$^z~W>rITU}49o}ak^0U;yz<+SQTt`ErtJ{QHC%jqhU~89W&@D8ymt~+` zI}>{<-eE0IQ3iewRCSsP`hsL!SlJQwL9wQj-kkH=VcJd^v zY3`Qy5-M+f`br8Ku#AjO`LqehbT=OyxKccjk#Ut{Jj5XuhNP>s+WL7W=c|gb%)>Xx z7)Wfeu`^eC^T?{AAxR~4x_N5q5j$Jh@&0{>EE(IA1-Tn*UAg%)GZULdLktCe+ORzR zrmqLjH!3RwBK0A2Q9yau*x40{p7v!#%e+@k!*Kp`2mC$U=hW8m-y((n|8l#ZR#Q_` zeqNSwD=_$q5_vMK0ZFMVIo^>PAx?dpuBN%GVso^ypbb{Y>oi&XTUrT z{MG^jhiY64#%<_0C}y+YMj&wfe;tVg9{aC=TyXrCrv@DTcDMYF**y)6`h5(~0A3U@ z>-qHbYnJ2-N}RcTlYZ=zP<5mw*AS2l&3l4MEt=Y~_*jq!;k=98K(71k8?>zM4#Ug2Ur5 z0?P9@kzd2Wm;jNyz_aSHI>y4pqG#v?H00ZfAQ|OAIpg49q(m&<*LkOz4bs&Oe99nN zsydc0MBlNw! zdCC^Z+SOC>%Y!}S%F5_GIm30}&jxWW>b}v`YM*TSKL>>d%`dOOl{7uq!WC~Mf`BDu z4&gX~MlU*Mxd_jtxz4>}-iKChYP~^I4I+Gw)mxvUxtW-3TFX{f#*5e>v`NKj-)A?m zY;!=bY$nL&GSB#Byi1*epsr3Ip|JDb8pxgj3Bp{8V3C2Hy+`n_WC2xH|M1utaO=+j zH=06!0r>vFJHUdjtE`c!+1Wd2MWC@blK#`w!eJ=8=IbZ(G2nSgUtY=>UlnoN`1DrA z$cl5@S#uz_Xp9lW4=b4CU%|Id2c0f#(qF%Ry?T&PmAly!blt%BWFLfzK$VqCK~NGf zarg15nPuPIum?Vuy%Gx~zsvA@7OV91nTEKP6<}*;B5Vz7ssgpnzK)e!!OLCOjX}6v zHK^jcIPSkH

8STS75=6y6j-wo<-gD6bU*<@U+#M#i0#;_2HL{^k}gyHD%0kjK< zcT150jx=DUFL1k#pbw)SFY)uSk!_wTQS91X8iE3k%*>JDs($M7r!;hT3X?;3pOoji zr8n5y`wje(MZn;|(dDr^L?jqfSZs8B0CJhf9)Xdww9x%EmPadmoSnO%T={m*@4-ix zX$8)g)|M7vo(9Qf-%EW}6EtCZl-b3^s!f#(a`MFnprFGle+J% zO{QD;uY&TK-;hL1m9UiO&!2uceuV(7H%i=Fqszg!Ufn&c%hKg@+xWS)#r11sHW+nb zVP&54bQ`4r2s>+i2VBC_&F?G9ybF$>D}xA=kw;Brpy5pWw=3lNOCC)<#R47*@q+1k z`i4@`B7A&0puLlu8+cLIeOa!#2g3u?1F{!eB|l={@v18$tH+>(vEje2VHwshe}*fm3{`kI|wR@ zy*-mnHMFniXo*v8o|l8eXnS_k^X}dA!mYcWf;GEi{#_L?7>K_H!9oXF*6NGYjrUxtO1jjl}au>jTs+=O-ZMdR*WYS#-4L82R?8!*vEp(F66;gx@j zEdp96pf&K%Y0T(vj=T_tK*zX(ZYbygI?b;jZUwkM-z)syK;B12_JffO2|+5moUH8P z^0MA-1c-nIEqRxHxSge z-VYoiKsJM?Fw#b(Gd6np$h9Mtw}F$V4gj{ayfj|*x+0hbA|~xRS9ekQtE6rP3)-}u ztZO_VFaQ8br|?s}0z?S1h!4>@n%at>rVgXwCBh9R@? z&)y$fL9{H$XxGHKpV5m1iC#p3AVy~~pPG6)ad>$Jl($%3?zKH<0(x%q2&D@xEG=p7PwZ${gL7mr$8nwf^=X<08r9iJ8J_aE!%b5@JCyFW`4|!y2 zK^A!}v#8r+YYN|Yy#om^4KA~<4`&9#o^#t_r<iaG+I&Yc1$(77+IgK50fbmB?|i z!J|~t&vI&-ppWt}o0?h`=xztvL9`y`RH~;%fCxPVH3-0ga$_O;5L@r`DK|e`D;3zb zjo)q0X+3K@ySrOp#O>k6VI->!YHx)6%49ediI; zZ{K!-B0H@g<9ZGXux-{SY+!0Z!CZ$bLuqeSTCG-YeSF=0J6PyY#rrcmk4mps4jnQkwap+1F-F8pXOka|&21@;`|bijIhO!|zi zHAuH9wE4o|M@7k#skKNDH8wa%`Z}FVv04yBU1)O~4)AYyE6W+4t_p~Ck9d(-3K`t# zf~X-Z{F0hFZcOLwdn-GpEL;WIX*M!4SM7RqWvqp_G|mla40m-Ie__6&DRUia$RyA@ z0fK9wg&@87Fui|5u2Mevr4t>2}E++0)xQ@2G|8^G2@sD2pEk4O536Guse`r=PPRD}W)(faDdB>?>?g ztpp^&{C@nvx|8HgEPJWsRKRXOC|H2n5fM$BRGN!rCd8Db0di1{x*@ApvKH5?HvWJC zBT6qo)UKGLs=*9EBb>~zp(~5c8(H^Y$)LH_{o-n&Ea1n2bo%lwK8zYKN>+i^`Mfzy zQIM0fTw7?F=yyWx;~hC70bML(uyHqO^w-Dk_?SJXTfe(^%Rzpv%}l zIyN&ys;50%dNbo*rWU4<3Tw2m)1XX4a>=&Fq$iT-J?X!>WboG7&JN1c(`DUCnYcs_ z%0fe7`BET^M3lWKKi|?|i3IyyLC98EeFxUcChNB@r^d^`08IvMCo?l!eFqyx$hN>_&Z}jn-BY`)6KifijvRbv z4YJm7r8cpWpFw&Tw~24uuxtbMze3fSddTwmT#VJr-U~@~7o33aNo1m|{HCCoxYJrK zM*Y1F71pO$*>MZ2s~Yr+4YQ;4GRsIp>jCX22kH$nG3tH5@MNt+7RI2qoTX3cBEouo z;m8deZPv#{AdOYOYk9;Fq*$s-YH67*$i}db&vh#I=E?{pPi*49shb2ffuJJDxyL%P zbmmn|@2H}hWtNw%wN+!fB_dwonmkJ^pjv*0>*(p}IY!B~+t~jYX`q&PESR}dKk~L` zjRh33n;TMlraca}r*8u~26;-lcc}-QQq?%MA^Zp{wxJ>tL5Q@rAGsL#2k28d~ zd{W*F(hj3mUj^-G1g@x+qr++urM!@BA4t`7x9?sYXbI@&aiRv3z#bP|&|#2Nd)_WE zq#a|uBTzxCo#1|i6}TR2j(;<1N5f<)4}KaTx2*#Ldk_29S}70rlQ+SDrVENu=~Q9# z5J$eZu1?*RnM`j*vQba5OmZGBX+9h@~5JUMh{PD`+!MA(F$mQ1#iWk8a+1RB!cty~km~GC|pwM(-kbjjZ95 z<2N^?FA;cRNHw(;ZW826o%h5T>u>k@8aCBD?zcM=-#;!4_ua40YmYyi45wF&7a@0v zd-KA@!14aI=zXrBi+CmNwXa`D?e&kpLT!CV8672U#zeB|KRb!>|H<;8_O=M~i8|^I z>1Nfwu)knr&-TuKyrnxrn^foMroPVw!rv+_v}g-IvKvC3DuWlAnIlx-Twu zkAzP7DJPk;@<}+`#qF%|m+|`3Pk*wAp%%7_=cX3err6lr;lw}L^p=*en|&{_x?}gc zIap{Od(huwOIk?n7cAD%wc4!J5D7SR)7W*t+w1>P)?qedaiIdG`5>4UZ2Ar)gdV5IANV5;)*_(yHFXk17i(jrbZB<#7ub|i2N)so z7ltKkW=8j0?M9BhX`+eyO^&x-M~rKFh@w=hgy{pQK8<@+&s>f%qKI5q6**$Cgg)TS z@~)^HeYwNBv?dzo=C!++F&-vT#u!m8D?P$@d!>X?mbOo8WHw-L(EVmb#oDg<$?lyL zlvnSdmfw^V@pGYPA<-up>ll7w`}4PG60Ir%WQ%HM0-YlXkZ)=N;C^+;DflhCl|x>g zOeopir!Z9(j1BV{c-YTQLehss30I#cNm$-lJ*R{xrqR9jNV(&7#m9%2Xu0GMhI*y* zTDCv6Vd9KJu$Y{X8uBjLh>ZvHq$aKIp&hmM=60V4v}_-V&lXGf~+UfGy+2>v$7BE+g^>wz7BF~jhapX=ZKe#aK2Q5;M{IO}tYlvac z$z+i#UoR2M&1&9!Q9SwZLay^@O~QNIubJ#NFv~A(Ew%*&>MA>1aICo)TD%S(ZqUyd zOR_4Hj4gZe<)y;Q2fQm6hGvvtT1`pZ%04Eut)lY>CfsiE)ivR+Xo)8qAgB2-5PR^HeG%T)Hb6?x{RqvA`&4`ypq5=nw} zatYms&G7O4@vJl4EqxxU;zn`WmNS`=OQuo(c=`2SS%Bvhy`jCr7gi)br#PjO#M8M~ z4kes5x>=WTstM|5N-g86PuiJnXY4Mth+P;janR)>TZnv>`r#9T$@o4LzIyE=I}>Mw z*JpilwaxU*R`Zw-mzq6Rpdm0e1G*U+Qx1j5+ufm%+oe+wYHTg4o5(p@_8Z=m<#;8R z4J`t-+Y9jYXARS?s=3~JTzj)OhtW7|bi%#ZT>0fy+j$)Uh<>(GKHtpZ%N9+l zS~elO?MH&&%c6=n>wc;dCI;zPT62$d=ycxh@++$~ZCj=2zAT$h> zBy64ZSC`q)H;C%g9?#J+&|)qn^#yEe!{)XmX({iNQE`2PJhsW!3O zT`PpFm-6%5uHqf&8_F2R)0}H6afh}#ocn0>s!@T5OxMR_@k)+2tU5OM{3H8GEv23t zHDeJohZmou$J~87L>V2m$(`##yVfQy;P9EZR<2q!O#7DbH_SVtf_d{7DpL8mtY7SC zZOSr~*4C3ey2clJDWpvDlMziT?l|uYqPTO)JM3?=@9%DmDZG0cQd?oJJQPUWB7bLS zbEVux_MW-S+a$k82fe#xk$egIox?W{$Vn_NILzFlBMmF{>3)=ZV(EHLkRMlDL7Q0E z%bF}}QYMXYNzhAjRh_xxHDx2WS3oE+xw8#vnO#8*-?gMT4E?9kJ(u5A>kZEodZNZT zp9d5(%S%z*^^>0oRn?cf+HPx51!3+ury$kAQUBIYR{I9s9TI_Z7|ye-GE=K4Vml}w`z0PuLKog?UPYu|Kf!!4s`l2f$bYZ2UMqHOw=QXcb%HE4^d|LS!XIP%F$ZO{D zJ%--FFv{-q*rZyEie1i>(3>^9?*xr*Bt2aJmoljV`9})>wT3}JyNQx z9t`BBT)y+jawVzb>1=JA^t93z52zZFt-k_oYLYif9EzwPMLbOMaf@PNOt}=7YOrh;X66^@9a- zyo)X*eb^#}wv%JeOW%x2U$fjk`a)XjW!pI4lD#ywe9^f3C|p!>d*t26@gITSvpR#r z(G=zCsXpF1iu2cH?T){zD6hT0*icU{-Ks+N=5qtc1b)-ok9fQAdDq#U_VV*~l18&m z_B&1^)Q6Q&u>=<6cJ|Qut+mH($=5|=zLy?UaS>st+v&KV#4Vt@A=m7mrBET%Jq{cZ zV%HS3qeJCfc1+*8WMPurThvzq1eyq%82u_uqMj%r~p z`*q3e+F1;{&OGw!~VO4@IQPW0dE7yDmud%F08f7LmZbCK?aZq?ve` z9Ocs-&&3GMd5-Dx{C(+KI;E)|kx-Pf;49Hh)u(qBzjO%q(yG{N4H3VXsr5d1vEe(0 z8`Y(u9(rSE!L9oBu-8@f)E3yITjWAbIhTcmR2r^_-En_v;YJkrLivq(IlrXswtqMz z_}-;%DPdLSt;oj_uU*Jerz?I~xV^h6DDkqd;Y9_r*)3|CSA1sMmKoH9w9I%V@oR~i z@|3S~KP`ReV`(P;{=(KpJTj?~*O^=Ewq{6PW18aiYi7BoUtmfx2Kb{^w;%TwG91XN zOI)9a(JA5!;MaJ_hiJRLUyzh(ZM?iaYBHafYU1hodZa#R%lWuZpk~I5=J@=$;%LO` zm#_&fYw7n5Wd@7qtqN01FSR5aGz^6m603MzSCcbMc~Jblh4^;kV4#zMIBC#R)G;p) zp;4P^<@a7c#@i@|dZsrY_d`eUq^;mnSyTOCt_|yInB>?;Qz1uIw~zcu*fxKd#eKSS zB}A-9_(GRix!+-#`)3aL%A;AmonbC4^U9J+N)z~#7)zz|3ygtEe>JjaB!Ypm*${>o?d5@ z=A?n1o&$G2^mLVkxdPD|+$#xVr$IkK1srkEaQqU#C1IR29BA1beHKCkJv|f#|NS*7 zZVqq}EzRrD@Stz|{V&I_UpP(&q~SQ7BMs;21ZX(XASyJRr<0)JM5_bfH?ChlP9ICd zdHNh0&R?T&{u+hz*C>!*qdNjRN^K3gj0!CZYHh|ZO!c1Kmk=N zdmA<-n$yAmy;i3~{3DDK0Dzv(gH6eT=GTEN4TSRyJo@B6!o$EP|Ld7-vNXTu`)}aW z*Z4| zGw|q-_owi@XW;+q%KyXjdCtJ2G2u_2&wU0SjjewQ4?P3_7nT06*!uU6r$h69OdEP< zo`5E+|2jj+DLI|}!kk?HhCijwKZS>!f&Z`QiT*M8FD#g%6|%fhFng0`=7#} zA(zuF;!okvkjv>#^QZ7G2nhu~+B?^HbbL+~@WcPixkA@~{8I~A4w5c~}0oeErk z2!00fPQ|uA1V4j!r&#fi;Qz@)|34%np5fj9RZ@|#vT$=Y`mj1T1IDQdtl^Zc6wS_;l8$<0@HnGDhsloB8p+Lhf5qCQiLipWI=Us%kJZI7d|knV1B<9cyt)O8#`Ke`_UGA+`6ElK<0v< z(3Z;q4d|SP`)tVzO2aSm-3~RvC_VU{pQ-qy+!5+W4eB&i{HZi(*;CY zj*mR|^(e#J;-><<{e4Hw{WAkMi$_zcj92XN!|3k_riwh66%)~oGydk{U9Q-jTJ`OA zV=1#1P21Xfl&`UJ$J~<1tOKn3pyM)0)zFtIHG6_}Mh*Df3Eu20tyGu8)Jr$*n)X~% zE(~Bf%U-R1G(mjN`N4B3iSH!YHV+;dk$GWVG{WrY5(t13g-J8C_B7!8G~B<`L@=(ZgQT>$ypiL-sD@2kgN6snj*H8tp~IoW z4c-U}EC#24*@2}39CPyU{_A%T*7dJrP)Z6I1b}7Y*ZsJE!JWMyeT@Fg&A~y#`x}-A z4FkqTpFceZj={KK4Nrfbp8pFD{gZZUQ(D~W!qgjnBm;o=Fn31ED znU#el_>F`6HxqKDVd3OpKm9VNy9^6A;883PAh&U%Ul6<{x8CVh1tU8%w*O#u(ym6f zRwnoDEo{wb*n!=`%ubz#mzPcR_fw!8e5as*WtnlJ-vL-g{>^1ifB%h!=cdV@0CvA5#}&i+^r3VA**Wq5@+6ub9qGms|K>JUad8J~=3fx3%_*V(Eo#;YON zErI7XA`)+Z-F>$CV_#PNi?T(ad)>b2Q}!y)`43uBK2495YP5MHBU`ldWfKqqm>tiDl=QUC|>4UCrsABhb=U}2v7rXzaD-g$m# zR+_lO>x1nhYQtcxe34M%Pv^eme!Tdx@$rRfX1os=c`P&-F3f8fx+ag#vG<=pH=&JJ zjbBn8DG}g3=X-IZ^MdxJ zeHKT3$vAOxjPVP_)K$Uu=LRm%)@cnob6uNoPBXCWB_${(pTIy7E-(obJt5);TAG5qF;VqL3igt}b5x-TNna-z6`>CP|w+IRd1OuV|m& zcfA*LuAe$(w`N~8fCHmI=>#tgf-^LPn|*VIT4M3sQOm+`h0?6PDY#ZmxfeXb$b~5}P5c^H10#^RF0;g{M8L&k zmP)Kj{IFzvwlv`RgX%n&j)W17T0lIQd_0(=PpCv2)<&CLXcE3$$jpsDVyL6`V2J}1APB4VcaF%F%RKIZNo9`A4f573C@*UQe z@1c#9=YvMbi&KckI-l>PYjsG;~?Q_c7pmnLcUui4q z(b*|c_0YfJmhPj5$@EVoS~Bn&7MwP=K%c`$`Pvv$ej<3!zL3Co3ENQIoc;QE zY5fDH`JHD}=lS=p6pAUDcw(J!z;XP=C0-+b-rBge3pSo19>cBu83#96!Ld%<7g`I|;lLcaReS?ywj&AQ2`uUQiggxt6~_3gRcxml*D zcgG%IIEx77F}yViGOKWo&wCeAQ%bDXI9>L8fn{5V&uh&4Ds_Uo#tBl1De~B3lHN(6 zeEQeE+I{RC^7t2JSAdz*D{2?poPeE9G!FGwpupD)lns;sH%aZcN^;`Ec%X zN!1fmg7fkZMPdC97AZ@XV2vDmW|P4I+r-2XQ&J#D@A2%?X{& zVb)p*yUgkjSckblP&W7Ot>Z>VPW)FHjk;YA3d75!1-t@%8>BykcJr}md<*>%U0gCE zvx2>%&stYeQ_=4CJ$AY>khUSi_Z{007|83WZ~v*WUNs>yEka~|F$<461X(OTa?kw* zis{Ph=Q2~tmsG;@<%j4i=y)*QDqqHlweQ||xyM@+XEPBe=JA$MSn5RTg#b-lXK zTcYWvBx7>+wacG0HEHp0iLGh6Ml056R>`=(&NIEpgB|uQyyHO^8;Yy5PUIn!7sbin zSas3wkXFzuG;K1bkwIC^?xr&0sekkKboa>O+&z!WvPX4uS>?gUdw0AZlM|?}*}(4& zXgjS~%xUQ2RbCR3Bom#F4R^rdrwWNbU&2nR@x!5*;>UN*kBN^)BQ4)MO zhDqrarO}Bq0_ixY`>8Q<^CX|0upI}#_>!GCOUd>ewhT2@ov121mh8!m8{z3MP!!Dv&?7nIoo59%j}Jy`*-?;F>`L`&U<0_Ul{Gr+0a} ztXFACDp{N;ipQdb1(9$l81VbN9=FSB=nlJPQx#bf1)d`ql{X|tHpeol<_w=oTzYA5 zdY+gC1+#h-G-qN^H1gs)nOHNc7Q48n2{o_%K}|DX(>z}j8?jnc+q*p0kcedt)hA*d zc>08CfZtJ{XpCh}GJFifw=>)hcsvd_#2Nzg>th+*9BasoTs?nFLvuYmZjVhJkr**o z?Ol%cs5U@tfMxuNWqYuHC}s%pHw@lGkZgxWWkQMihD@D^(30kbv5b-!o|wO7VKYok zXiEl7ydmBe3Yq(J$5E&uwq^3PL_AX$PBhM{kL>ZFEfa6p8#H6Vi?8%9-(xuQ1G~qe zTZm@wiOfBJjbW@h#YRs&KeCv%C~+AOL?$*T1}&MEfIbbP$=pF`%bY<_4StOn?#f&P zGqNC~jA-6s45sDu{)~_`60t4+LO=nDlV4t3GRrCvlg7UgMJb(}CpEaDYS7YJMZ7Se02v&B$gR5$YA);co^GRje{#-_~UUp!CN~_G#gGbtrMo@&tqtI zxDA;V>BbCcrdK+E3MsE&fngg07`CWCZjhar_TTBhR@wFe{cr-{y#3y&0ilq zDKUD&^l7n%EzQ>0j-GtcS^oN<^Vcja;~p`M3lj??VUAq|qMUKQGCi55W%wlgiHxYP z>5ycgN>+knBQr+xrTMf`jmQ5VZ|caMrgyRI!fjSFGlSC?pAWg{{G#S*TR3PisLIik zr*GM!UUUseE&syHtPCg~O`aByjmV%Wz?n(->&SHuqR<%5m@s4`Xa%6Ze35l}QN^%@ zH^PrnXYcZnAn3Mi8JUQU+|s;dR!6RNb|PjZw(KV_k(ahCYiRB!oQ~Xqt>Mhb+Z$o5 z=HtP@6a)?5lfauN>@o4?NzE2WzdgMi}7dj3=xL$4Pg6Mv(gezzp5;Z%x#9o)TQXkfvivj&qurvngQh?n62LX6A zlPiRB1JZQ>tpK|L4gwqj0ACbRdNl^H6krFyae9R-;393Ys4=XB3&OoYFcx*rk4^#P z05}wdv;trpz)XN`06PFgR*#k}1y~1g5a2YsVsieryQ`q>{o zF$ObYC+9^O0K$n@iOIS47N;tA9OQU}Il^Hu`=hx|4!7BzmBUn$Bd5UqkCLy*$@~>^ za<9W#xntPncv0DA|u0&EA^36KE*DudJjh>_O--T-(9 z^#w2i#skRYNE^JSgB)&44UZ1y5`6WX^`sgGMBN_*qkAqA&2+Qvf*t5o#Af!?&?ihqMBq4qzt0I)GCE zB5Tem0D_iUP29~IR$(=I96cj1!$BS3IEkADM#M-PmvY-8STDrm3Uh^IAo9_0vCAnt zJFtDV@bqFTM{I@<4UC2BmNk@<|gggy-Wm{w_+@`ms%A*IcgWk0-`fcPy^z%pu0ra!T3(w(i(TS<)(M#cZ^^w_8a|JweU$icAb+k5Lt3SPGU$ieYA)QCdps~Fo?4^l9_P{9v zJMet7T)J19CXEAcxl$^Z;!;#9kiwEjcFBh9kTqGAWm%MYnaD`?bmWekX&PPhh%CrR zLwy+W3^xc(Nj@A&uuQmN`2q(ScR47KFrboNU?ZNoYMkvjY%^et2eFx6bq77kcm+wayJ2%`Brg`axI~p5N zz_+?CQ0E%ztQlGVUy^1kO<(v0F2V~k_l}-4E%Q`CW2TZKxq`;gnV(R@eLwyhpKhq% zkN=lujnnpXL-E%Q6KOd&w7#)%bO)Zw>Y*6^U#JHZ^S`W~42qf7LovB9U+=+uy(HWZ z^@?c*^%MmqS-qs9uzEbE_4ceNZm8c=T+HhDV`v4dv%(+iS?9GR)JY~;9bYSYjn#S0 z*GlVThO#P=2vjMIuqrr&BCJXThgg-V7phcPRc*v4qvDf$=9(T@2+D%4t3K1 zufMs&(`meSNaLIt)ahwXG|UBP&TL&hKagpi9gFRm(`b1^spi>p=F@!E+)QI)ZhdA> zqCU1~$c+DTc?P{aBvHQy&1jfBZO;sIZhhMjb4WvCR(<2%QRAxzUi7Zp&cCaA{C|1a zc>1tv`mRv}|I6ip^ztbBu7UJj1L?a)nWNad5To(&)Aq<{cq7=@EZ<9PD$r|lINmti zXDl1ah<8XlaC7(oSS!1b4O|Wl+*dmQz0`YH?_u;3@D#n|pq{ODDRA?Uc=!O`WnD6$ z*qIoP(koZ3ScL)&3+nTKE8qu3D_7A0&!_1X|F<7pYXB#;e#J_JMrX<WW zptPBGDKogc%w}r<+d5y^2MPz%LXJCMk(SlcGDWc}{^$I!vhop>3tGtwd$C!FSE3b- zT&8gJWCEgKvgJY?09%TxhZT*`!4)`-S9CwaYIZt}^5=;5xf@olvM_5bR$6)fPPk=7 z_mG_b(ZNceAF-9`^!cDs8A!3Wz{YMvNPzK??nX!%p9l3TgdF!Gr2h;dXEW%RzaZp> zx}M)7n1#J*n3Wn7c@Q$%pBQ(Amp^1=Bg6B`ZAE7BD0lq+JYF~t|oQ%-4EJBUf0epkd zj6QVT7$i6g)C+X&O@bacmyMSyQvgc;ywxE3a%4@_E0$pet>_Pldow?Yz#((t+G* zgmT{uf?p$E-qMt(-2|?}2`6#v6ZgsZp?kgM`5N6Q~-A5Ls-!CVxhm#AP9BQOq4{Ua9?D2WLb23R2=NC z_0@*1@LdsVlAG+)^d{fcp@s5d`+WU6-*utR=zH2x|Iy%w?l1jc20tn|9?eCAF`+`Q z@bnexbW^xOA1_=h94+`4|DB;3UI#A{6pn}zR=p92El_;Kh7Fr(Yqqu8_^4^OlPNY$ z*C7osyd!tEoz5H}r`v5dks@c>b)6x{+bQYL60?GmmTe_=Rw16vK2Hn48%h$=iSc&4 z6KC*g%tvt@9)~&Hk;}FlK$d`fZx%QlfNAuF*r0b~m(4~?=nF9=T*DCaReki;I6&Wv zJ@gegSU756AU)PNb3vdqwJfh4V{}2u38U*m;Yn0?vJU>7H8AP@`l2Qb6;Vq(;jD5F zsH!X^UIQhHN;r?-S5-Nne|0KRB;wvrv@hQ?dv}YO{rcB0E+W-a?ppoKFITO8MmW&* zuWjSDy|yBID*MjQ@x2G9Y(4b)kyj7F?AF8VmheN79R=~Ceck|Vrn@8e8Dl>k5OW2^ zg6x7zQiIwk@ltuJ*eK5v7szt8G1xWO*FVr;jCPIoH3Vh|Gn9!&ldH)$F|b5fqFiGv zaV_y(6Sy9G6;ZHX#Z49_t5<1@xw*nzb+M-UBfR8{*leESFq_Y?Ne5Hv$bDft>C8%O zW+lTa%}*;sXnPn!2bm#soB`=H-Nni3zQEs-A;lz4>US)R!?gUe!bEjnz#YX%bHKxq zi#{A_j0Pg~705xKj~t91G&+&UM<0t!cw`jS!GQG(Rl^h-XPeH?;7Ow;-E_9;LP45M z-bvugmL}8!+RIcX36qrB!fb`dO^r+wxs8D^jmYZ}#Y7QudoIjk{bRTN_OICYqkn8Y zmOZ(@ZPUiKy&E>Q5jQTmb9MIbU5EbhV_b;sZ@m7-pMLxLYoPN&AZ|b7Q%J>M>>)JB zWcmdesj`Sg6$y$gU?LRLu@oxO?;kSWKjf^cLUna@l&r$f_ZKiKa@MFFxerXcvqtf` zBGocYK=gjG7jkUnP~$JAQW&pBrI4tUGL)iZHS$3MXGhJOOZ!x#7$myZ3zaGb_*$|+UZAW&*W>HS4e}c0dUX@tL^g7_Nt@*@%FogL%3bO+=n?hT z=y_?6`Umt|^(cB<{StkoogNoK#eg zqNw0Mk|a1X7(+u;q$;vZ2o@z(;Sd%oK++e%~ znNkjj`mUy~rcmJIi6$!1;Sbf-8gSIrI%}xV-*gj``H(dN>w^*8($v!iK~0TuT;=xp z2MlyqVf?G?;{Q33j0VzQ?$2Jwr@C&Pw{-GqviaOmD!oSoe-{FOdqH|nGW}5$@q%h= z5}K4Lwz=FQVW~3LCVM+_Cm2IH0Wl|1tROB$zMpGwuwP`DYfNZZWP)o( zaAIVZYe{HUWR1ASdzPFH7|4fpyWc{>Ebz&^Aa{6ow zJs(4vxC_SpMBPaskEYtLmRg*jE@12?L}vuIEXRJvz)k}JH#YWkJ zhN^tYfZwbdP#rSqi2F-f>Y|#<-;w(g*ae2f&vc)kvV@Hej^9J)#82%X8yz6OMkOnp zpEd1t*m8c{>Bs)DAv*Hj>H*c6(CH9a7TOuggmNK1q$L$Lv+&x}3PwsyIieg_cqOkA z$rOs1PTB^Ay;)J|TNV0zrYIG{>^wr>t&9zh8c%uUygmde3(480mL7ai+jW9z$=aG4 z%-D6r4Q7Zp9CqDKL`fE9LFPn**TTpyJHuc)<8)csdQhcsZ@fQDTMA~WKPXgCqIA{< zP@Jn<`}Wn3jx%iSHs^H{Cfqrs{b%i$E*aN<1-ZLx?;ZU{O_;Q8GpPaQ!l({rXAkgJ zU(Rp-h%fRK4OT8!>Wilq%`IB1+^O7J{J8t+@|QTf;tvJkVP7aOR2HfWZ40%9_V!eJiC&T_E-CIcpm@64XuBqLP3g+S%Hr1IpQu06 z?k>Hr{GPs#sZVN;mONIvH}%_;uasIjbkvHvkwiCA+>Nk0bbO1tkwiCA+>I1a9pEag znJ$+kHI)y=QeNKHry$fpo-&Jq<&>qO!MfnM;LPCe;NhUC2cyBI!DB%_8r&8n!C!;= z^n%D?oNjt(bpxB282AV#2pgDC&fn|tRp%vu!&!}SpBV*<3rIo4EAg}$^d>+)VKx*+ zOgAMBKGMe)4dGC6&~yi?E9vbORHFq0`IM4#&_{_X7^8OvWAv_|L3^W3(FZ?ssO4zEgM57wSwC=|#s7u0!KMXb1aQ zqN%B6j3u5<0vM1jAItcF7PWwkx?o72Tn@YddELnt5Le8uq!uO1fH~j~c*LJ)60@Xt zVM6efr<{h%a2uRhWRHcBQYwY9&>NCM51hswi7+Zk*fqJA8pfq1iYli0Fp3%lVGKHn z>Z^Q0*(qI?UcY`l>S5L5Ch$Al1ATeJrX*F;hxD%=FmM1^JwBgbN>TCX1;&E|j9F#| zrRv)B+kUimP5cAiT+>? zsc>V1#|gfgAHh%JujNwo#7*M5$7`UB+@z%#Gco$O(%7ym+JG z${0F=j)Q00E#A<1@n$&{A~Pt#4`395C`B+iheT#k%8Yczx<CNl?B5K zmc_QmE;x;f3-dZGo}GzbM=pd(-oN;Qi2>g(o~G3XjKfF<*iQ z+VuA42OA^#%Z=&A$F_eIWDT3s0fqn@Odq&!$Pp|)qGCfe)n>I-0jPd|HnV#sqO9U+(TRE{bHB@e&zBt4O_pT%m3r>;d5R-#Wejb zAVvlX2Ve$1+9wR4T@$D-8#t(%<<(USnD6p5ZOIsw$rulFT4hVjWZ8%Vl4IJ8 zN+45Z*jWA|qbjc>K;FRE)=`_5p3xu37UwPPG!{hV#%AbyE$h344w+Xgto!4X?aEFi zqjW0Al+%iYl&G={{GJ_F(QzfGs8O&2vBVQj5xJW&5(SZ0MJXvDeh0sk&+wi6abE1? zPxA!vG5!dg@qBkrdD>Gy?J3WC%CnyGtfxHn3@ig7-&y{AANjHJ#eJk4%Q^@olaHSP z(9SMzNxLm#*bKAZW4fWw!?P2(FF;lnqB6YHY_{1zS=y4GD{KuOQ7H%(l-p9C@2E+&Ssw*F98yRotH90=H*8*uy zd`_?C^49XBC7;E=N}MWj`hB9egY0Q9jkqODO)Vpa`l4pE47H+8n12a%kekd(ArjHm zhN6h3`n*-iDm57h9PwkrZ~B}4t$w~77@15dXTsF)V&*F@N#DNuhel86-A2=>WB4XdQRt~dYwK_=XB;(=^d!9O68RSTFYCoc$0S^_M5;w;R{$rIGaIzjVVMuDiQ#ZPSlkdp%klz`X- zqrpXaxIb9w%tq)LD!F#It#ZW5n>Gg=cy;Elr?30d9WUPa_}stl{GTr#c>JcdyPmyq z&8}&o3CYT9rVq?)#kKF>kMZsIw|>9yo5O3K=F0xmdGL*wUwxTc2%A8zeggCD!#A03 zffL>Lh(Pt5U9~$jM{$7p zTLGroY|QJjdFbHVsMqCWK}tr>wjx&JyzgaWz89Q-oAt2A?knyP6x=cz|KG>&>7$Z9I5j9m^?Y#?sVp*#Rak9{|~XjGqFLhIhkF;{4XmO zN?+mnp{X<0Z(QVn0b_x{{tvedY*9;IQ=Nz{8SN9DiYlv;3NBg7DLTSfS6$SS1oN?M zJ)L@>3ue_Y1JUPoCY;rIE77^>kyYHrWWn(ID>RLfIUbke} z-EVZgNV#?-_+!UFQaM34PB9--32#r@tL^o6p}(hpeqnMbY><|=bM&5_RN zTf*D!_k$n1KlYsR|0DP@i%9#T(R7H0u}6n!Ok3(hitT-TgGqn;XwqOG>A5U&r8?C< z&;GIasqZ^{#$jMD=dc+%3P)^`6RBSCzynnnC7pWGFpfB};WVAi&Q>QMWpt4bTss+) zI?pqylQF52SwBvyEg4-n9h5Gd-AwAFHs){^s9EVMJ}4cQj!8L*?~ZyD_K12EvZx2M zb0o%}65~%A_2{X7DRcgSGl5lWnO{wnZ0&{kGmT|hUUQx&eSVOgxCc9OgXX@n?(J0z z-@2vw-ip0lv1eAT{^hP8t$BFkFK$2g=nl+nnJ|nvz8gtgZ@l)KSC78&3Z1jdbD!}M z;Ji{akn}do6}wUvw1>)i*~`jm>;t?5!-LB%D{Hbhl`XU{C~NM!#lErEgT9ACPujhu z)L*0hE1`BwkYbMqpDNuKe4+G};Nj9gd*3gW>wUP84KeMz%XMK1sXyJ8Foj}Ke>9LT zFRQNMYsxRR(Z7o=Bfo3uY@-`c-TI|o)fFmF^8SNkjDp1{mrOM8)CkqSqh zW1C}#Bj*ryICeWuIXDMfi*V$%6GP6JUOImcHUSQia+$*s;rt-x_XY0pL?RM%3)sw{ zh7z?BNT=55R%H z3+6wPD!66J$fq}qT|d_2u!oA1>UF&@X>19!Y#nXRy1dVt)8{r^G6=t48Zk=8R9xD8 z)woNp2XR@P`krb15N*SqBIwFY@l>?L-q`Wi4kgwT9O8FI23t1Etg15pK$Ublo%H9$ z05!3l1D({c_E1Nk7@;d_XD=?l_13-5J?BoB7CyYg7&`Y6GUs+IEzaI?d)H6Kl!vp&C zf_+S``k3xjnATO8)>Y1HT`S@kqV-tBkx5sDX)A~;P z$THtfU&fd7@jl}DzefZ9FDh65fy(vuP`M-@3h@5Wa-|BEvdfvCx|N1P5tGdzn$K%j zhv<-!4p9qZyR2JUbvR*j+U$tob9bnvdeSLS+V?XPf zNgm$xqe*w()b#@NX*_7tlOUNxc)GQ&RP9)&W8H>LG#<=T4)HF~c!3CS!w#<`Gcwfj zYKIYEe$kHc@}ak0WlLhlD@~OY=*r3{MZ?j^2=@s8B=;nLv%A^<5P67u(Egb5SV)t> zt6fMIa0`W1+A@2q{c-I%WuN+-runpu+DC+Q6wTC^>g&KLWj6B-eGzRGwVJo1oiG!p zk%DyHhAy-*0!=S=u&9fpC`|d(mX3lei|JxLjRu%36mTgE0EAco;IfFf_^^bjdL`BZ zg336GDH5q)xVk%b1EMNFaZSsuE#CWS6m(GI$>nGAkvX=A;H;@InohtU6=f}8^E7rZ zA6N>tJ1jNTXc7D-So{*J$CEDJMjkp^M`Q(R-eIS{qBdu zix$89<~uL_nDUhyRBEg6Ch}vuS?E!)9;^uV4VuAa!H2Y;*`Kt_A$zGk6YLD~K}x=* zp=fo1Z09sRqGB&ed)z!HB6Wucdvb1*_a}M8k-Kr;MeS#~sOdilM9*gq7=#GYc!lF5p7oKNK@Ms|TB0hlP=qy$dxDRK0(mag? z&@jPCde*F(ii|R0Hc17;S?1~ltD8YiYttqqRyfKW$V}DG}tZ6tzf{w0Y3-_ zwC&j84&Ac)iW%WSl@sd^ALbssy=77L$SYkxS4TF_zWw`aVN4Iez-~f^fZmg4fYgH1 zUON*lMeERR#P5XbJNbw2XA8njbj7G&-zxAb4;(slh|Wh}Aix7`m&1N=FR>E3=8nPFopvoY)fEjSV9SeV+zM53L~PII)&g9B|Gy{m$?Z<;-S$(d;UvuZWnn2NGiRLnh+y6Zw$>_f zlc@!g+rv1_9v43UxG;MhoPT9H>Epso?uHrjLTNTH?!#<>E8IPjVfs8DB2|ebK7ufn zBZMybkAs<~_vGU)Mn3M(CWF~L$Y)V@k2tS~k#;`llHW@ZEKYXdHGAWudV0p)0mZHp zJwuEqdu|Tt8rv|p{*xAvAVBgkfn&ftKk4+dh;X;)*F5f&M{|a;%kJ$K{ni3yca+%Y zr%6wjs)s6gxbpFZtM7~6{Ms*`+MAd$blFear(JW!`oVnap0P7$Pdl)CUsnnF`Qn*_ z?|H22KGL>k&G-lJ>Uxjb?w4j4aL0jO3>3iw%sXu)P09j8$Y^qdCf0fDf}?}m3wIU@ z)$ZzWU17buJ`93BJjXpJ++5gN_?Gy#>l5*F?TdiXixg>TZw=|MT}DP~)5!wzp7z(k zN50R4pM<|BI_B-35J+){=m9y79DYX?qMntG4c*k6^;VrPWZ{oO=2_{?v(nFdRyy;n zbf2~IlFbVw-A89j?_Tp^)$3RsW2N&yJ*#4-%2?z>Vv!GtZ3~e?`IY2?!i!@b|HZ57 zqPElz!^EfsJ1zM?zzT9)_*F-uZniCUV<(;a*Ax##`w_bS;B!sbEHYeo*MDQ5ei14=&?-spN9!Rn9w|oHNFDY|!mmR$%XV_r`?N zrV@=;8#jp}|2|VIOn__dwE(5taj}(!^wgN{nN=jvN+EI=A zOM~Shj(TpCXi8(`(Y6u#DCgy_tMrMkMbb6$Jl754jnYbazj#33=lYj;PARoHOHrx4 z#8Il3xGFq@&_LJq@<#c7?mq1ayo>Czfk}e)i3c3N=id?EQ$FKA(?4~c5x-L+Hj%bi zW2qtLT|Y)yEOl8MRl=%6=UvDtOL9`ulMcF!;gC2x){=Jc%FKaOtk_8smoe+ajy-Ns zwK-F2+Bun@sLpUMcCK}9ajH(0=MbjTk)N0gL*2yUofYXb@JG*$6ZFV?RPYxz!Jrn1 zC`p2%sxokeYB-(12&4B3$QA3zU1naZ>W>EK3sIEaTF3hXbjevxXfv+o@^C**owPrWzZk z;-%_3m8czLim8lq;-$`YPU56zHbcP8EN;#TQ29B0)_wL`7R3mTIn&e>=xS<#Kf0}^ zDe%epy)@QhL9Ht@!BQNHT zgKc*lIdVtZQC~gg0vg4GtY72kO!Xw-I5~G@kJJ~lg7~D-nJQ)y%DLlvq*%VxWo_8p zPam{TkI`p?u07Hw^`%d2LxaeH{F}}{{QP}>cAqnMe6JegV~Ad81ka4^lXm3Z+UKf4 z#e9@WoGMONINknF z;jG9h8WqK2euDwYui-IAtQ+B|=wXZKQ@HC@;u2%-cr}c|>AkKzm$yUi5?Dbz&-S-M ziOBN;FAh{j@j_DUtxi*~=T@mlxsOEYaSHpx~T{}izgMy^ei4@UV@QV3)}73Zc8HI2(TlpMSZGz`u&Fukq^bXm<{3Tu`nFe z?0=7sy0)ja0BVG(|FK5P*>n?K`#*V}*6V7)D4<)%c)DL$09tkcA{lb6T+6Z45ug(q-u&gB zaK1X-DcNLUEzFC>RTw9*<2Zt z5i}rOu^ME?cmcuu60w%&YO%=+~Dm4gu^3Q=V%) zbm;phz@*3Mez1{sx*M9D=+>?>jVI8_lpS^_+r0i&J4FO)f0j~AbJ5F!u5k*&L8!_OYb_1LZfs^K$~B903??N!Z+6@Am~1$<*hDV7qj7E@i- zI!go%l#j!PWMGhICIrhIRD ze76xc*CC?I9uk)M)!Ihw_b`;&W!hyr*Nacu%N^6WtN7LSHI7Yo*+vAp#y-F?j*RB& zB~u<_AMQ}^Cl7GyY@PVGzQW^pf+&J0$sKFJ^z4*&YzrULILXC8a<8>_`DZy3I!YL!iN7}BQ19h zXz(3Ga__MH{-ZM*Tq|Wd!k+tPk3*%UjA8Ad9O5;O@^}r{7;o(xSmUT1$k69{Luv1B zK3(3z_`IpHvC0|8be}1XgU0~O=e!>mq_{-PDO5EA3lt&_vP{m(Lz>mDozO68ivc zfW5!t0jI6hRqDRf*XU|=H+mPi7PuF9ZxC18Z*bn|xzW49zQuXF>vs2Tp8M5Zwik^T zoCiE#sGoZNW$!Y+_2eRj-G-XaZHt6?yXM&MO@J1 zb|+n`2TpVi%%G%A_1J8xn=U}uMEU>}F(RZQaxg+79i;9#9mc}+bdbrWt@ zy!3v$D4PDLJ-c{Vao@F5tF!Z-G)jxZi}V7%wCjOY>({O(i_ZOi_wdF^)W2Dqoj{s} zH^J+?#8gXw*$kH?8%77N+KYBLfH|v7r?kUyHR22|#&O&;&Y$1T;xk=mPa0=g?3ArF z;}mgLL({6DMUv<>Fg|wAAID68as3S?mn6VF$xe6?e}f%g9ql@Iq;bo=FZ?xyprYC@b_@!CU1e>~W>9^Yw`D_mFdjZ&j+n&Uovzk0vzdD5Z%-u9Z~ z4dW>Hw(=+YUyYAdS2v~FG~{yX0XsOhA~mcW6w^gS>_}CK$a+qR6*Ov|X0H`RPLdS` ziy~Mv9B0#Y!(q2$UAG%Hj1*$Gahjouz-p@TDtc8RMiMC=gcOe0Uj-W{sc{}n<5WfA zI3j`~(=?=xb79wI_M5dLRi7m)H=Ak)4nJ>-_Uz0_sHM+h@X0~}gg-l|!TWb+%yCYCgW2573E2zyP;I#%$9?G&@L+CgnxYqc=1>mB*M_zd~d)M4VL=#Hp^IG%A?bW9ui16p<`kdyAQOqEp8g)f72 ztp^2KbVtCEc=5TfbK92wX;{JQLO#kz_uA^UNzs}0g%B|T2B*3DRBh56=r73^movRx9{CUdA20G0JpO@)tP=?NK2x@ku<_CQ4lcs70)H}H)U~U|@XIH%U9=;>H;2yDrz2m9u2ly6r2&ri7b1ImAPm3LRyqUsWE%==(rh!`y zPLMcA8axi=L%2JUfOgs}8;~t~(U;MbE`y)U?<%JOU5_Z03>YwQ-=XnWR@MyQ4jpRQ zni>rJyfGSuFTtYTfYr?z{ z%&WE3tLJ%6RcnE=gKbhRxaYy=VUSxhvY-{+TCg2GC_K&m(!QT- zxBt$51f3}OXMxkpU?>p>0#Z)6vKjO=p0 z?QkOx3KPf!+rAx9e ze|$3g?nArrh?o9`%ZD7Sdg&)mel%mrCmSFAkdS_-&iw{o_verCls(5^@4fTxN3y5x zdLjGy7Pj|g68BFs9Y)SZ*PYE7JA%7|RXbK;X^>n?FskLaZ3L5Gpyi4DX7>DE&aM+cfVFuH z)inWEgWr{>kG%f{f5`X)U28_mTUuIhi}eGQ?M=`{oPmCD*yBBL?D-$UAsxWOF2h4Y`X@G>t31EByBZK|f)RZjH!JU+LN6T_T8j$IB&0&)cG zb=4V1EWla@zx8ylc)XHss&jcg#6c3tI5o;>jJYJ0i2nkA`}Fji8&|Fvd*fY)He~nU zn!A44uVKu6i^o2jeM30lEx2NK_V6oDWV27ss(f}pzlP5r|K#6gg)okzxX;O0;SY@B z{%Ve86n9!a?IBpk9&%hd?j}d1BW}_ub-GDL%DBlR(j#tiw{*9g{8;+2n=F%-xyf94 zu7^yLCs`DyYc>vfo_13j)9gTc4j?)#KP}OMz8DCRpjwO_daVX@S7P@M)igULJNqg^ zK<)tPl_1Kb3mMe`JJ!<8+O>@2P8bYl0-lQZ?)+k+J5O&dEkJ*uTdDY$^76l`2Y%&m zqUo#32lVIu(v9%n0HF<;&})=$=A;W)cT9dH(42-B?vs_l96tmI>C?R~KMi($YB~Vn zgOG4Ep$lz$11t{q7jp<8d?_TZ6yHlOQ!5<&d^}&AZ~H*xsmUhFk^`abz1IgqA26Tg$v`{{*9S%_8#;Eixy8eRju7@#bCK>$h|TlI|yC*(dphTjLD# z@-g@tebUSU5!eUr%oU-ZwHuI(&iKLqh|mw(Wlhoe1GzJZ%bj7N(8m1g3Hz>A0h?L| z;AxyGcp=tFe{mWxV$N$7c9J1ouYZZ-;~R!wjUyj+Jx`W!W3nUHu3fPm@BV&o*H7py z9RopnPUr*+Vz(Is7t>k2h+hW-@PN!e!(#=3(&H2XYXrZhvZZh}o~Z7NtQFhi-7y>M z(`6K4gW|Bp=yd*jAjBes)`vqOO)%|xUhzBVvjiLyOo0eN+W}mQHy|oEPf(Al=a+D` zW7vjf8rAOp6fvG~f*9;iDU_URANJJ#nr zzJk&h--K?$Yx$LOi*32K%6_ANE82o@=QqmhZMSM0?RWU!aK7qx7179jBo?AsELK6Y z-mw(j#Z=fUrlGuU zTN4J$mVTmw5=96~dhV{~JiZ?PTVUa`#h)DP{9@6PO?PC^f^~oPuGt$G&EIg_wevO) zzHIxX^}C+^@wz9taIgCp?mT*I=e76sDt~43i#f0^JGcD?Po974Ei>nAy7l|qnC;^p zZ~gI8yC}ac0c&qR*eprxHbY@g*h`vA@YS*#ySU4ulCj}O{}$z3;}9!lz-!Cl9f zuh@^T{>wXfMf<#tpY~nWI%dMH<2LVjCHqb5tby1G8htuKyaw}8h>FmyX2tEHtzoh@ zv^Gp;hvtUKB5js~OxGqm$N)#ZgM@>!#3Q4`=|uKk9$X00f47;47sYF%YP7bfC{`Pf z7ow{RuT!t~FDy2$jybV&VdBc^j9{trdWIT(pn><|9?^>;Cve#^80A<}UOFDP~`95Z-G z`j@i@FW9mD0pFpcUp(IY$jY&oH!sfK$NY^MAQk@sJ?=}qX2~3G4!?q1$@9sQ{#;FD z1b3NqML|QfzIbHGB(71KQE+AHZEi<`5}36mH;*K{kyJNQ(v2jTEts#IN0Qx0sv9Yx z`gA14O6{p)Qp}Yk2k6y_`eZ}J^w`wIl;mRDLi-}ewVt_w8*De)Z`5xxRu!*EZsfMu zZnJOE?=UtL-;%uBey@J7x6ld%^^T`p;Z#UT^};EHdWBqkWxo`f3sY?GeM9)RFbOAp z_TGghNt_gX0_BhVf7dDOtrQmeI7aqq8r#eheUlaD^n<+2A7Qh1ve;p_330HT!?Gmu z91(G{xCqKbp)lM#WK!1K7Q&&EKGd7BFk1sLa14*f&3GB!jz#b!GN!vXeIW~cY{3ROa%aLPrUTzbjXZDqU8{}UhAWU7dR{VbMl8gqjEa}Mg1 zrPCMOG6Ota`eAH-Q5@)vZ7>Fu*%duLWWY@=?tz7beHcz{<>C@1G=KO3EAOZBlLj@5 zQ!}2o&;0#OOP`uFe#VgO;t31p-Td`WAN_Wta6o_d$;`twgYbLPT5sHV?&rVD{__F+ zu5sNRR}NoM-!L!XpOqf?=-j2hxn{u|>m6I~Tz}QLs;Wh$L!Mi`>hOw{pHu(tVXz>d z1q(8OitzVl++}lM*MP|M=(X~aC^)BVjfSO?VSV`B$)umXd#P3HMr_>(jNFHNU7>0h zU0y9Jsdm!yf|6>(%5^JuQG2u?mA?-44J)VD&C38uIxde~9-CyF5m^#huB>s~pl?t& z>-X88)I0Rg9G~ijL(^hT-Q#rXPF+)6VGo-SaL|ty;wPp0j-wVpFVmS#fJIti5$yoxAccjJr#5SU_=i zcP*|(io?R)-6>9ScXuo9ZpDkc7pJ%sd7=C4efHV+{O*0<&wCfFd6Ic1lSx*R$sd_- z)U+_}UYx#_%ZBD+YTU`6(J~pSO}u>nE^0+D%?BY5L0{hLJgamq*r`xnqW25oku441 zYZ`2!_P%18@5L$@(}78KamK>Yu=+b&az$8!I{YnBa*kdkdbtLcZ*Rm5wiux=rEepT z=m#~U1&)I3hVCKPMDB!b&N%sd_k8rxx@`_l;y--Pq9VY=Qsa4&1b9EJT61tUX zFlZp(nR^8yefpw8HH0}rX;*C9?kjZFL_C4m%DslO3PYJYTf2CM0-AmjrN}UfIIw5h z1dfFIL)=)KJ1W)7ew~U~IL4Z))b7rWjL`7ZleWJmp%_C&t~g)ev2s~~CJ4?O+IZ9rq4lhR6rr+yrAdoC|HFYr{fc4omMB_Raso-Al>}@^8 z$YPFC*-~3WJt;k**iJfT=`@_uFQaN$pHpXMb5w_uR5e(Wn=)E7N_C@ADzKb3WGl$q z{pcZH4n@W0rxbDC8DzMkV-4P>1enr@Ey=JgsGE%DGzAwJF(-bA>?glN>W7hY)mz$f zsQM~+vSeXEP;6lG2sjkQ#D}&d7L*vzC)eT02j^=i4JH(7ADuKT0Vg^!#IF9B*%Z$O z0_q(YC&Jc1uC6Uk{=Yn`A+dFN=K!;;zL0Qb|%sTdXpE_B2l`>Db zvTiH?bWuNd&}(y>BVl+%wA}HSgP4Q$GeEkKLAR#IblK%>{Gnx!=)mUn)zUt7 z%ctNHAbeAqV!h!qsGBKBsR*wW(wr=2iFt^%z?L9i{s{M0SY0|^?tx-t#%H6%NM?@r zhnz&vNCcD5$Nue_3MB!=Rc0Z1c<)&}MINCHaaXl)@7zl=M-k@mkfZNQ67s2>lJUbJ z-ji$K?NHCcrD){Dv!~;$4!R&R!$q=sqWFH4Q(>|)K_e&!AB;(zJ8DQd4Z$>ywxBVe zNtR{nGjqhubD8Q@nO<5VX2w>Qz+tiXs-x$oXknG&066={k+p<4Z)3v6g> zZ*IL+KbcF7!pdB3h;$Yt6U#uWFO9ls=1SimK>0RtyO5>^=z=qmWuRnN&A?4Q2;9VS zB;cP__I?N*$LDJYcS+%=mBc>i+msWP8DaradLxdfBJUqNe27P>YCn8juVlW>RJ3yNJ%6>Z+WR=9gpjXAp~Or;jM@V~~myJX6>?-_0pw{TE2KkfC262h6~C@`nW!1wCULd(4g1cU9;?b}J9CNLQZV9o80ANdH6dwWP<<~0%g@yFh+LU&AO-cHYi&y$}M z#F@m`sms;5Z&{DM?lmWTtS59wct;{CpJOW|HB2%B%YQu1?-$W>UsQA1GiiuaODTL5 z&XUrVx|1Cmz2`Y#gO!gn{*WvQ2hGc()FdTJy_a3EPbEnqtS_scJ8yvDK*op+Gw3zJ z(1x*`tqwJ`tLf{FD21f6+8w6n9K2QT{R0eubGi>MKi z(tq-Y-Vk$yMjppqRgl4?Vjm`7WU6vosDbZcVxi)TRQc$_ExkYB5_TU(C7 zu{aMCk6|A+(O8a0R2nS4)~-q43e+8zlN@h8T;L>+BoWnJf@j$=75E{!{=l$|9;)8>bQ&n1u##m;N@T8ErGjJi@ueW;)&GGA2;ElCk!T%6`VVIh0z zX(umxDSwOYJuxsji!%IpHb4>cz@@v|>L9_DTe=K zi7N&19hE93`aLS`Iqx|fwg(x>PpJM~Yu(W14m2w_?&;AI&T>B(uHI2d$ zxO3baREdO-&cLJmp8;#}ye)+<+z06vye1Mbwa(^P&LbRgZZdjO9C)y8-RE9|y0?%c zA)DYe=2T8oe)Ql@&`g6&yJo+d5SW6O`!{hy-Q`dQtiF8TvTcPqdsmxVf?@pLCxriE z^M}L9eVw`<7Q87FLcfj6PCDI;yuTzt6>11>H40?v?Rk@fZWWocj;E-)yMB80@fLGp zPsiChr-&L(W-Ec_^TBwub{Xmjuk@|y^X(%$;Uzq$vD)Y}G`&x1Z_&Y0ZIB#vj*}un zh+xa*(g+f{$FN(hesrg`hJ|co_)Bw__yEh5O=Tge$c}Lue;}j|7FQ2a!lv3NFYKin zekQL^$}n$I5Jd`kFbU`<+-M|$?~wv%JPe+z292U^i18ak%ue2VZl7Hx5M!$@BjRmh zi+;*JqUvY@h<3k>@P>5DzLO%LtwjA8{wat>G#=71!&bdZC%JV^uO)(GBlJeQi~&8+ zMd(>JU5yiCTC+>%wgHOt=X23sGH1otVpD|kfo(&<2|ABZ+6#ZZ1!`*s;?EVwi9oqN zu%1|9PLfM?A^ekSMdLL_)OJd1tlO|bF`>4gE{_pLSe$8Qv(`Rof~MgsssYR!QzTl? z-J4=$${?*Z$grY4FVGf4TUQ#IlS}O@AyDw-)w9Ek*NRXpVpVd_mkk@gtY;dv$mN`E zhvB-!!g%pjIC=}B^hHejL)#yrj#VYN<-E5j{kWIjQqr7X4oOb(dKb0G77i#OcpMHY zxlG_Y(gqb-SmU8C1#w0Zoc+B963tOrWM`rA@c@l)qCO2WJdyj27}?-EhVOy%+q+&` zH=bV>Qs}xRYOAjKgXJ@7AeJK24(_yJY4C0XH4agogfnwNM7LbK2W*vCuI(`+(8D&K ze7aMeaMa@@*4=OldarOP>=h8@q(AoDeZGwOxiCs}#24CVLJ^ihH&>i^%=ZF5C|XbR zzaroM1`vJ26lG@mJNE4xXy_Z4g^7)e8Sc-ESAO4e^FKo3{sZP&L|9x}Q{@ksXF6dU zOQZiI)-%T+pxgfo)-%WNyIlST<@#IjA3&&oT#+Gd!z#@A2N{k9R7Llye2CMaqg=m&~)Lwt=+E(uHR*ora(v#2}3@>~o<590xmsps_xY$hHW-VE;DOgyBZEY_nVp%5gZ=+IUh(S{8FVZhuYk~h;1x45{td7AzxDcm z11)A^Xa65ii#dLSyvll6ZBX8#Md_|>{UppXATEiUkYbyZQV{mJ#i+RLN8 z-R46JPlhAKL7*jpFd!CBk_5%CIu(j^42eE^sFX!m6a#^Jqa(Wy5hXq6JE)BMK*%<8 zIO#U~VT9o`Dzx1Y1B*OrjAJKvBM*#?=h?5dm-RyHjDka2r$d!lE{-?k;ymX*z}aKk zj|P&wN(?C-VVHGRpKGZz_B*;Wnq3iTKc;k5IO(wLHI9gx|ovB!5w8 zct@ONG2GJ(&(G{7fJ3|m@_AD_yQ@yl3GL!F#8Jcka-?41SHgqGP*vcgEC*ed0M@1& z1;`oFP~#e}i&dFzx;EnVQ}k`aT5a`p7J5fDv5_l0@ph}iXMIZZF3r0sON@d{1jpTN zS#W~8YF2fSfkGj$Vrg)AMkYiRTv@a1r%_pOXhsG^XWx4>MsQk5@QlkIR&0=WMM4nj zu~1lD)2uKmP-c_hE9p=)UA3&mAc;kO<4tUca=xFYvq9BkpwPQUSbu_)P3u9Z$3aC? zw6GyUWV0Xw%7|>?qUB;BBpO=e<%!q_6`m?aN|Yde9sih3!RQ=l;J4lE=@)`6DHo&x z(+dJw>)G#n&-zgqi(acqR`}q8h=>DO>@nPw#{+&kj(UF&@d_n>Fb*pQ`U1H&OJwzj zDK`f?->6ZxY;-1EbZIN(EZCFQ&^q)1nlDbB){x-Q9rb%*u7WJ2YRC=H0vC=4tWxn$ zLL?9$JFv$ifDB{ZEil$T~fq7kzh*_UXn zA%^4+qt1J(4gLhZYeOAl-aV^hR;YV$q2aA)$mHHJcZL_cJ48Y*SCif#Y`&srAQi5l z?fz>4a{}U7xt)qnem4@E8h7sdK*+TSqMpvKs%?R-D!oL>dtjVmK9ak(oHZzy6#zLl zIQ!OR!^PiIn4jzgu+EGhH-=!clx&iW{k=C`?|EFx(wAT$u{Lf|!xHUdVMJ?c+Q7-? zjoo1LS5LN1Uz^=d40aei)lMQ>>_-bQvc&fWq1Kbyn5NTKfY73;ui3ZIdWhceXiwk~ zO+;ScWg9*G_MOCS{mbw>XUtqT7-#R9X|%H>8}~eG3hdc zSDQPthimt*4e*bj8Y5Tx>{M=Ae_sE(LZd}=D{)HV5<}4zKCuygkM{~)v<{(zP0tE@ z3v&kll4IVF(*z=Z+P>n5N^#sxTbrc5aP126x`l*q@aBQ*=9UT2J|`;v{GqF5-;6I} zOe&rB8Z{(o5&PQX4yLrLD({+Czl*u%i8AU%_+fE=NOd0^zhBOo<7>n(y7L?_y~u`` zyd4P{Z?eJ_uC;#w#oyn-p;m)>W2Bx&(&-odH1Mo`ad)&rhG zdUe7pmd%6xvgA9KpiqxayK|WYMIe+Y-jVp)nDcet!!eZy0hbMm_MoIs>Bj8l(g;`! z+6tWoED~Ny-E2-NXd*h?Sa#Q;X=&9Pu zvL}_K2dO(6>w``9n0a!y=as}2<^xiho>h7=N|1gVsowF}J~^6xR8GwTq|O6h@&_{9 zxXfWVQy{pc4fOF3!)lIk|BNQ2CpfP_*oNMSJ&_k`(<0OzUg%<%Zn(A3)*K^H^%Jlw zh#c;G1 z1eYGsE=Y-!*y)j3&>fV=3Vb7?sw+Q|vwQmTcVKGcxuE#TVRth&+E@0eSD^A3yLh2u zFy+NDLrWMx-!k?N{A?YeB&2|kOG{=8O`ANM46i0Vuf7Ki%&oi1ls>7`>MstkxoYq<;*rtLcbPnW2(s<7ST4;wYeoY0vmQK!9KXD3a2o}f8|m}D`EwVpEuF--kJDI$!L`0O zl&aQt7vavC)y`CREs^^`wZl5gc^zxSL=fY{ODQ_avkp&ccGH_KJ~`imT5R|X0|=n- z_oHb5`DkP&gO7I-Ek12OLbPp9*4eoI2?BS6h|>W$yy6I(MdKB28hbcQrW){&<4B7U zhKZqoi8|1e+Jh}Yb03!5Gs|bx)Tw#_26mo99(}ShG-QC%IU#}ND*Yv5Pl$Yuzkn;@ zHrMTxGd5f#20L^*wIc`j2to5ZzluXB-yyUte>TMUTr#!oL6%T$%2858g#Pl2LOY}C zI_KCrC-CFf#tzcdM62y98vBw5)TG1dl4LhamTmF_3_BJnI?7$Aq!6=04Ln1KO`Kue?jk;h`13$+NjW zsou4*9g%75F0Rv%P~cJOAN@{bb2}7$;eqs#zw6oF@?5A-_hf zY;<2C=1Zl4(npNSw#pf|74*75scs+7h!G6cYq*LodwI5d1LE1 z9N3s0S^d&IC8CWxlonuAOQeQO~Xv-ycly8wzrRR4#b=wa)uh zP(v1gj&H$-ngdXZals{qS{pS~i=|EcvlFGRIzl+3uC&_NSnIWgNIe`oRtR z0J}gbk#Ad*3KcOW!FjESht@{Sn_WNjag%8w&x|wu@*;!XN_nC;4|sZbi*5M?@q$hz zic~Bd8|8H8)M%LVa@Yq}kl}4Rr|%n1yI^3D!DR@kK>>j6>uAj?N-yYqZoOGU!HuR|T13XB{aAcM zWM|{bjAvmFbmgOS@!M!1?!EK59%56aT zC8;MVyP&7<=65Jn?Jk(PAf~s*dv2Ej9jBa4jH6)rDsWvU%(Er7quiXf5k(`Hm zSF2fatv`Mz5hLN)psjh0#HPR{sqrJpZ2;)v2fdeAx}TMzI$%rS0B}pv+rGsS^%KBo z=y-Uxv#Ser31!=-1dR6ec>38)RVB6^dR81?LA)c7Iq*906;P8>s~vXVEYL^%4d((Q zvMy?0Ha%HQzdTINSE9OU3K{ir|Ku2LMEi+AA!%&>e&BE#1-<6p9(fR)Qoe5lBMNPU67-EI z{>edNuyyt8g~!(kdLtV@!GhH`UkMgx8(c61KANp$7ctcz576LzTgj1!a-gF zH&SJL1uh}*Mml@`;K+=bI(75t(=WRIclR~aXwO~9t z)Jb;>h~Qy(;+3lW#KllxtSj0`7|QFhlP}1bgCf^s?!x!o1>t#H=D+}yB`zTR;F7U$ zTxo1DaX}e9-kX?y%M4fWeO>Z#MS?0GpSlxem^C?)eq zJW)Pc6Ky=%O?)VR7K3ZYxWxypqd0{HHxGpl22CYE?l>u#-NYH{Rhm06<3-$)J)dbZ zA&-^~^|$~H+Uf==p!bck&tWIx$cyuJ_^!oOlKE&iK`FRsCy!b1F$bXSI+{lUbj1hh zm*GIzKhjwlHPG0``N_M_1%CVj((Ui=d3>S|+8(zYU@~GpF_l~7*EEM-rK;@y6It5~ zKkqKvMD3I>lOl&M-DBvo$t3y5b1oq!#ya$Nhr|KN=_1IFDzDnb@26Df<+R)f8k;E_ zDMu-XX2U@jO+UCN6uD(Ey2p&xiRIB*ISGwp@NTs-mh%v5x*f_SZXozn*YZ?4Ohrpq zM{@I2z=~;SJ&1=I0gb_jcM&6mmhtd`Uwhc~8yLm0$?k$IiKT}yXsP3*lnzui6Eq8y z&Z$3&sYn73KHPjGemLpxp~=#7P3+JhqeFclFZsd-Q3jLhYhYX$!zzatfM@7+skqH+ zbG*XpgZ#1g{PX3PZYE_?^K^c4c9NL|()UFJ)FP2gIcCldSOd2b*{tI1w|cDY&CnEP zQ`*+D!MN*@ATe5eHH#)|lBb=#b}FilF5EFwwF3E-O!-H|lBRqI zKy0+t0ND!n?M*34o5r25r#CyMqX)s?ixCDvh-ekD?sQ9;)NB!bu>bxZGf9#scf2q+i$<^ zUYatqm9b%@vzoqPT=&NbwbuznscznE!E+1!mFCmcUgi8wkkh$RbKROiBmK0t@~W1H znC`9CVOjG71#VekB-{56b90~Emc)%pxJhb6l z&?9jjIS+5d1{U(?rc`n~B<}H#R)RGGZvnEDXx|?O`y4^)N{btc6F+wI=c&58^CVT; zm*ed^5Dj=ttD<4xJsD0$o1&7LN2tn}Mci-L%Bzx*Pdpv4eG(!nSn@ken(G!As!{>* zPEyXr6u2*cEHRlaRi)XtP&I`Vj;!%Fi>%pH)_d;{qAhkrqQGuh9HAdyORrdd_Tfed z2m|G(HhpSJ9Vh}2(m*Vt_jA0+0LzgGN?5HNt=%2$3aEnib->p^XZRR3VI=F~OwZR@ zeoS8}sn1!dgzs`ck{x_1rB}eo5d<_uec!NaO|$fIF=bg}Mq^myAV*tae^E@^tj!fG z;N$0y2(VzWG^7Aa>~|+;DX!v$({>cHfxp_vhfgKkT=S4Nsk0Uvd9}=o0 zIlyRQ4!vekmP=XVl$xdVrPyLewB4WX1W%_2L({q>Z;AV&mM+Mmwus6ulh%O6AD8ONkeX2`HioHMHk=@|N0KY#V<5;AUJgG5y<=vWd!$gGzC0y zRrvE0)E(HaJOSrJBM2C-gYO0e!&NlyVgx69H=w?YoYB72m0l=0kJDEazZYo-=EILt z=rW2A>nig2<>GA7iS#{Xb$upRI1ZbTnb6u-%wAYPq$l4tOCl$}Wz?;V75#iNB#k_M zuu|X8`Z%Lk@0Wp0hDPcq1YFd#{-j58hWGao`U-FiIw~asSQEt>A63f6-UlRzep!-- z)2LD}*O5uAhVRFlu^7_4akHy>$aYotl55LuI5BDr<(ah!=~R3P_3ZPe^e*QY>oDp> z=~OzJ3$d7WiS=0I>Gzmj4w)?1khIpr3YVTK8?GooGBH-aD=eK>WEo-B4-_cCNY7K@aYOl5xrW zeh;+<5rV}&e3eun%BkjGBok6`FKm{v&yV4}|4Mh3g-Y;!r~ZWZ7EDMyF(fS?qkH z39z-*Izg&~bG_8~-(A@>r$xkv$S2`&U}6wNi%0YNHJ2hP9WiMiIjq2QqE()J3e1y~ zaV_yI%bdCLyA@#I^5d)+UAq{)u}{dY)lI*jbgr?E3Mo=5X@b>!cEfe0RITll-X>ro z@&6QGbdosp+0udvdbV37v>F+&ahp^<64H~EDf*7}+-<9BWiuRU?;AYg@n@nrMUaeX zio$B;MG)rgpP!0$LcU|#1}{fpTVlPMlrq`mkPU%4YQgS&=CzZ$D^ zL2XH`;U`qj1l)cjj}%fXpAgZp2xT(|j%(H=XyzXaCL1iq$6z)rozORg#Sn}$Cg>m| zjgTnXL!hU!Y*N2BK|uCUgbmb@`&Pw6K)k?2mlGNw=h`bUxRL)5E(bn zi%^GAv=c0=IzVm2s+v*?Al@fi(4@4=)WkJJTiA=yOWQ_SBaq7!<8*5XkNFAXTFi44 zl<=%HcND>9JbWCOG^_|7?%jaLj4?=@Wlm#olQPCnyN~1J%xFr(fTvS3nA7KWEWg6S z?s8mzlvXHXA61g<(jJG1a5&saAG?<5rUt}&$^C?y<0E~AqW-11I*%+wfBxI9n^xw7 z`7dGToJ5*xwZ*|jk_IyN$`SsR_TeYa+mO$f*bc-!O9DXYdT z8gyXtKHdu17quf(`6T!-5a7?jl6&8#c-(Wv62i&XxX$joZ*A9#1gh{kidr5D&GL}# zgIT}^(TkpKe~ypeM?w*k|3^gPWfn26Z2H3_&}>LU5E)wSveosQh%-0Z$;)3Ro7;ih9!e=headZJN0-C}6f z2+JqhCt$t}X&( zc^PZ=2-c+|>KEst#4djMo{ve_4b^jOToJvB3%a*SdJ*3)y&R73Sa>omvkY~uO^(Me z3hGP9KxYX+;wG6tm~&xC?aC=7fryi+C~Jd%_A`I@psZ>q{fQDtwcVf6M8_OGa+RBI zwEpdM^ygJNBv(Eeh8ZZ5f+v(?2X5lK#`?)wL?j{|9pkSWIN@Wb0lNqnZ2~+BsD%7& z_O4{*)dFdQyyY#kU{VNTy)-|2jP_SuqHb83PgnZXa6jk`@U34w)}+Yu|9oJR{f1Yx zeH84IpuDK%BJ0(hPwazt(j{tK7Iz8y#TP=aiYErH8_ zHqST_lG@=N$;17IbZv$Q9M{ycJ%2IvP1z+P={5QlYQRPv3AEu?lGb6%ZfEowJ!n!m zbD^XCZgqBmM-iW~Nv`jMU@Tv9oY_wrm@Q)(3#+5L>X$|{}LB1rBHf7 z;E{Y7a6{uVhr-B0%%>2V8?XoG1suJzyvCzrrxxZ(V1Q}q*{`Ck8e)rGeqim6*nk*p z%hGlDlpx-=jdW}P{uFwECya9{4>wXYlf)mfe&BO<57F-AgR=4@t%qf*Qd5W(n%&PP z?i5MO9-6Nm5$d%Cdu|bP8KP*{z5ok}-wTgo&j=L=bOgHwH-U$s;2nf7PlW*;oz4I} z*!Pqxt14HFvcx0t@7q)tRHSYS0`u#pSJ0T18Nhx5oXO?%DLi1lhK38s=z$A(#)gS4 zz#N<4Em846d|l<0t{3g^!J#jQlXfcQ?llYDJ}@`Jg=~+SH)q8zfkj$~J|2TTgN}ob zY{GX`X*f>*$u3b8P{%#aJv5JWtY87A`LpfbE<;Y53ZZMdq~Ao}h3n*CDqsTq6FXpv zF^s#*LCTMrxZiIML7^3j<@`QEuUoL&jb?PfBq%aktlaXw9kq$n>|FwJGJfN3dm;t7 zhvb|zyacON1O>09-k@KNH2ZUeX63;KabDOi^+4Q?n*}om5H0TGyrz3&WqUm2ESYy9 zL{gl+_<-TCx^V-sLFmr-XckTw#=0q|4X+J*WIzsB*&np=8h&1Z&Dy;iA zMmP5ry9{BJ^3|}W*e2B>gX0*6<{<>PICbA;$R$}=P|*jCO@T{%xxmjF)|gI5qgBuL z`=GSk9PcQv1c~FjzQNI$zg0rq8W7)2l`$x0g`K`Rj8drh?z&YK10rF|Wm(f!dU?TE zqj9Esrk2&k?m2gI+Mt8XU-LM9jNizHTCcauIYz-PrM<_bZa;J#9%@7ear$R`Pz&aK zDf6Wo492UAJ}u_`FYFEM*^$Y_QHE=sSEaV}o4SAtcnldNwG%T+4Sj z+8#lx2WWS?-fhiFAg?Q|dUR&4+;1sQ@IDFuP_Ax34tSRp>}4!yT_--i+pHs970{g8 z8{Uk}Wx%|5c;&v6Xixl}36OBI&RaParxA-YW;td#vBOLjfK}xGdP4IxxxIU{>~vEl`sW-%Gg`1M)H!_iYkBzZu3d2(Lq7F0^%okyLjRYA zs)j>VNJuxB$7w3%Wt5XG5!Q;2V?iurA#BdNNyhh~)uLr*c(v{|y@);Ewv zgPRnj)15<^j^gQV;ErT#=L9cc-vKV4YlS#B`6yOt61?I(VWszW(n&%kS;knbPzUpNn4Xzt^OgwD!7ofs8MdI8lvk*p zr8DY2UGrRcJhRQ>X<8>75_;!_F-@Q3&Mh1`*Hrr#NZE?5R#u5L$ z#8(*o7Hu3ZT_9E{)=+@-PPzWGI<=)g4sjXe{Z|LD=A?t4Gf~|HFMXSRHsid}ca`8C zL>djpo54wR9)U*d;rCH}X}T{g=Su@SnMJq92P-uu$b4$Qivqg!&x165fl6=PS?SZg5_v2nGBwt%ia_I$toY*2k;3EI1YWx?=rR z>|_*cL@`?RHGvwVx#p}D$dKxf830>9m z)7DUB9&C#x6UfxlHBP!8^GdVv1M&XahD}YyQ!Sh}y~4vWpTgfhQ*YXs;4XF!WuHYvjoVspOp@t`vsNBPwiTir9k8~caWmCF7?OpsV>7jEc zSM-h1NF5k*Q_h^M6L5`r#Pd;{OB1Ko9HYDuocy+1_)u2%p*VA&fc#JqiHfj4zbQh- zX2i+pN0DtYWe@=lm70O?WK*1ulQiQo7dbj&rF!JBPy#@L6hK7Z@Ic5eA1U75$K$V)-W(8f#{qqcTPMq%JBy8gs># z`ELL+z7rMs*4G5>E*Le~9w!Fa%SS{LLi3U(1!GA5hPvBHDjP49k4_$zg1#d4Z1=DD5ysfeYcl#COnAgGfSCKm8k8Te1LF;v9q;<#T zhZ;jQh4!>G-hORlkEI~~c#2iJHN)n?26VoGPO6THwyOI?REK_Et4#X!SesA%6=-LB zD*hVP%VW+S{8W1OT*O~I(mhlj5ieUlmvzT=qsS+qoJyFKoPnm`5R^cw9)WCY)UGP_ zmFl$=R(IG%fy%S%1z0)-BJy#{kF#Tz;PlQvr0iSdv+hz4cj3(G75O6iOHiU>A0D)< z9!;N?pM*IPJVGJSQHWrC7Y#>qsIRN@{i)uGg+nHqk(Qwh#+WB6HZ-RJr`fAf?JAo zelaW@r2PU#M2`X~9T1cpweNmy^U(1#Yd+)K-lY?ZBfJx`1tHtX_~|TMy*F18eKn_t zkY1cw-0tZyk3-*xAm4g+N_Lv1>CedFceM`weo|IJRV1xFx-q0b&BrMp@LRIkd3cr? zc8tCS(bQScHEOE;sG?r5Tvh)Wc)jTlz6kx;;@mdtUA2Agj|FcK)V698=k;7IJN`3} zkNpuxvO#OtVq`T{+8!uj%sH=Gtd;Bsw4&}K3u9J%G}`HlrfIy&6oV-D-GFB(z@I%@cjY7-6NBE7gAvro<1pPf&b zSSSku{bei*IHt~;bAAfF z7THao3ffu>W#!N_p10IX#J%^N;y`53?78l+?!NAt z*M?A^P(MR2RnNj!v`)IdM^d?^J@4t~FVC+p z)D4Bv<=Cw8XWT(IAmYFHr&Fb|d;MB7u{BHO!)S*xD+5uMgyd|xA=W%k+$R7Pi@c@SPD#Yt4$01eAZ4*@wZw@2 zF^E9lhoD4lmgn$MlR-IMKsI8p@yFSyPv$z^@igHMvbo_s#l^47YsyZ}PFQN4;TPJh ziK0|(@O$TiSai?wjgdO4sa8Ro(;eE6(GdgLJfx&3Dk5#>(O~3xyk#Wjecy<+j)@U# zyLXX{j0G1j5bz;wpn=f#;C_HY7bFnV(PW?&vsunA4|BrvujWgxruOkIsY84w>gxkbzvtxIYtvClSaR~x zn3ycsHU{q36CkA|9Koe&H@64SfiZukV|7eA->x6v9<@C9bS3$r_v?IxJ39l59W@Il za3)qiR9SZ7gh4Kg$?zfl=#UUEBAQDcxs0%JHwJVeX#HHfdRAHHrTZ__UYHxg&D4z(o2sP(` zwuYaF)wpKriH2g4sk1FH;q`X!rb-&j>%HR#p_)mUv z6196PUiba;rW`qA$K=*un>5-#{9<}Ye)9b#_^AGR29D+m(<7FL%DaHS0?2kGzE%^&R4aYACUqM|%`G)1=(^L2AM4UqQT!G+I#A^?n+X@B&?@Rx z%7q;gIzw85E325~teL7q=PH!m1%-Po#9Pcqkgvo$;>2@W8IIzo1>QNBIW zIhM+((H;*CpdZ}G{>HVh(BW05^`Q5@z-PoaeuD5SGGc_1^y+f3;Or~P`e&5-U1-+Y zN;qRr{>n0c*eqy~%9#4P9a?OA@*u!G(fp*=OSFdCBN*SrFUF^5ApyTTY$ypdCQHvt zpymDpK?C^>+4J=UC_5u()Rn^BCVH)nEfhp?S{J!M{z|h^c9H3*j}HB;ifT)hGx~QK z??U%Nt3uM)^NcRF>O9B1QOr~K2E}>PN*96my!55@?l6im;ioQg+4CK>8hK2B`qd)) zdgD!6Dk3=(DBL46I~wyM-6&=Y^I&>PD5$a*e#bz0Z3{$gg>EXHuOYhmI!2#%)Kj~s z4BPCz10qepXy7ubOxZpi3lQ#b=>Pb9cP?sB?c$td3HPP5es?@F9BXZifROi>!j{U` zufD}Tim#+RlHKrepIlCPu11wdv`I$2tldwScSw{2zZ8=e&O{Ie9Asei&z0ZMGRMMN zNNDzs!^&hh``3@(H&JTFQ(0(@AU29m;_3C4p4_`@`)blH>Iqg>I6m`p`fIlgORsKE zX%wBhLDR2>QQz7O zN?qA%b-MRzuH_OFkAHye_>|!f5z&@~mfKVy1MkZP3y*%sCp3#Ve=1|-QnFLNsg^*gRPW-z32m@^ewXDhhp9L)Hn-@hA8db#+lBZG$Jk=7_Dm)_Ix z3u|oI_V%$qZLuq=5Hhrr918sh8WG1RHt*P#^O;Af$L&`heLKQzj-Bc`c8nwusN-(c zt5|j1fI;k+u+G8UuEi$&X-h4G9Y%8F@F^o>S_qn=BpGp(S@UpJD-p>&q+-$C1(L1& zkp7p`1%F!5wCm}b{;&gy`IqMf`-4hhJAOU-;}Htd+356$5{nC(wg}X-Wlzh~sNoIW zHswll^^cjHkcZFx<8~EkjY!rM1+6BkGpi1_8#u{VK46nW0QdhA`15uK|Ls^l6EhigS-O3Sl`kC$Y5jpuY|mU zEP(0X$N!n$A}szM%JA>9!mau`@9PIJlVq7O3>5`|quJtJmSRE~mG8 z8KljO96kYl2QK}s49B0bSFgs(0If|NO#!d*BEM~T>v~zBqrQ>8qdtSOlY!$KLtR$p zwg3JVC&4S=YO8N(0dxcy08PxS`AE)MJ4pa$Mtme{Y_g29wjw}NGf8)Qpt8H1ilMuO zA(s(}06!ey^~3AR?P_If^{O4wKp>6RuUEe zm*8uTkHpl`(UzNm!NtXe-i3wU#@>X1iHnPifsvVknVIfYg3iIs+EL$?&f0!5-`WuPL6qOg$;^mbh>cNPlu3-0j!TS@g-%SEnUjuDgiVZ&RYaUcn3eC*V!xj z`sHV4WM-jbVy0u{Qek4@X5-*yVWVT@ zV>4?rM>l|ilYym~gDKD+zzkrfXQBVQvVWEMr?x)__Wv8B{zLXJMgKp$(e`iS_&x!@umZb+Who zy(kzNG5{@sZ*#}t)ikDmbg7Xcx8d8sw6Xv91o%De-lm5#(C$CE|C$_@X8a=3fBf{X z$-j*MUGDe1`bR-thW}m-?A4l~cFJ9N-KTLa_u79rJ{BZyF13>H7 z#ozvIm48kB%S*hb-n`OpxAK-?Vq<6da|iGigZp=v^@qn(u?GT`Y-}9=20QNt^ zb^maNZyY(GtD`i~?ai|ReuwP-+plxJ`NzMj6?t`GmNq7@`d&TazkKk2R`t4ayg5Bt zeS3=!*2Xrk-ta96Gy=T3U?4yLUyktA<-Q62yRO$r%RlP+#PC5xL|EV9wG7i+{Yu(@ zHqzA|Xbi^);ADb(4T1dE2f)t8#=-_L2K?qRGcj_py>5}-J^<@KdCcr@p~rvcF>$bP zalSV8@4VNtoGkw+%f!LR#rnFp_`4hvJ15iI_UJ!&9RE+@R;=1U3`6&PMV8KyWZ9OW z6p|)^PK9>toFULDbn4glsZhvC2fgjYvm`%RmNdATp(VJNLEq4XBN9WqvB1So)+a1B zlHFL_^4h5+ASu4T{bFy!(0bg<;LD`y-|Q=ECHV|}Jr2#iKFHJ&z$|OYXW&}AB2Ks# zflV^e!hPe+DfFT7Eo&+I%B4PWFFuRo4QSJ#$JXg}-66;-ALlv*e1k<{(Y49JEpPi%G1+Op{+TKB*7){qw<}{zt_4h*{_M^;_8Bj}vbm9!{_4%@5O}hv)zR diff --git a/src/test/regression/regression_test003/eval_questions.txt b/src/test/regression/regression_test003/eval_questions.txt deleted file mode 100644 index e69de29..0000000 diff --git a/src/test/regression/regression_test003/generated_data/eval_questions.txt b/src/test/regression/regression_test003/generated_data/eval_questions.txt new file mode 100644 index 0000000..68dd52f --- /dev/null +++ b/src/test/regression/regression_test003/generated_data/eval_questions.txt @@ -0,0 +1,11 @@ +What is the correlation between the distance in the latent space and the MPJPE in the AI Golf swing analysis tool? +What is the title of the paper mentioned in the context? +What is the title of the paper mentioned in the context? +Question: What is the title of the paper referenced in citation [28]? +Question: Who received the B.S. degree from the School of Computer Science and Engineering, Korea University of Technology and Education, Cheonan, South Korea, in 2017? +Question: What method was used to remove the background pixels and isolate the human body in the video dataset? +What evaluation metric was used to assess the alignment precision of the network in the golf swing analysis tool? +What is the main challenge addressed in the proposed golf swing analysis tool using neural networks? +What recent advancements in technology have allowed for the development of systems to help users understand the movements of professionals in sports? +What are the main contributions of the study on the golf swing analysis tool for self-training? +What is one of the main contributions of the proposed golf swing analysis method discussed in the study? diff --git a/src/test/regression/regression_test003/generated_data/finetuning_events.jsonl b/src/test/regression/regression_test003/generated_data/finetuning_events.jsonl new file mode 100644 index 0000000..a8405d3 --- /dev/null +++ b/src/test/regression/regression_test003/generated_data/finetuning_events.jsonl @@ -0,0 +1,12 @@ +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n\npage_label: 106286\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nReceived 20 August 2022, accepted 18 September 2022, date of publication 6 October 2022, date of current version 11 October 2022.\nDigital Object Identifier 10.1 109/ACCESS.2022.3210261\nAI Golf: Golf Swing Analysis Tool for Self-Training\nCHEN-CHIEH LIAO\n1, DONG-HYUN HWANG\n2, AND HIDEKI KOIKE\n1\n1Department of Computer Science, Tokyo Institute of Technology, Tokyo 152-8550, Japan\n2NAVER CLOVA Voice&Avatar, Seongnam 13561, South Korea\nCorresponding author: Chen-Chieh Liao (liao.c.aa@m.titech.ac.jp)\nThis work was supported by the Japan Science and Technology Agency (JST) Core Research for Evolutional Science and Technology\n(CREST), Japan, under Grant JPMJCR17A3.\nABSTRACT In the \ufb01eld of the acquisition of sports skills, a common way to improve sports skills, such as\ngolf swings, is to imitate professional players\u2019 motions. However, it is dif\ufb01cult for beginners to specify\nthe keyframes on which they should focus and which part of the body they should correct because of\ninconsistent timing and lack of knowledge. In this study, a golf swing analysis tool using neural networks\nis proposed to address this gap. The proposed system compares two motion sequences and speci\ufb01es\nkeyframes in which signi\ufb01cant differences can be observed between the two motions. In addition, the system\nhelps users intuitively understand the differences between themselves and professional players by using\ninterpretable clues. The main challenge of this study is to target the \ufb01ne-grained differences between users\nand professionals that can be used for self-training. Moreover, the signi\ufb01cance of the proposed approach is\nthe use of an unsupervised learning method without prior knowledge and labeled data, which will bene\ufb01t\nfuture applications and research in other sports and skill training processes. In our approach, neural networks\nare \ufb01rst used to create a motion synchronizer to align motions with different phases and timing. Next,\na motion discrepancy detector is implemented to \ufb01nd \ufb01ne-grained differences between motions in latent\nspaces that are learned by the networks. Furthermore, we consider that learning intermediate motions may\nbe feasible for beginners because, in this way, they can gradually change their pose to match the ideal form.\nTherefore, based on the synchronization and discrepancy detection results, we utilize a decoder to restore\nthe intermediate human poses between two motions from the latent space. Finally, we suggest possible\napplications for analyzing and visualizing the discrepancy between the two input motions and interacting\nwith the users. With the proposed application, users can easily understand the differences between their\nmotions and those of various experts during self-training and learn how to improve their motions.\nINDEX TERMS Computer vision, machine learning, motor skill training, golf.\nI. INTRODUCTION\nIn sports, it is dif\ufb01cult for beginners to improve their skills\nwithout prior knowledge or assistance from coaches. As a\nconventional method, people go to lessons to meet experts\nand learn how to play in the proper form. However, in most\nsports, players spend considerable time training alone to\nachieve outstanding results in the \ufb01eld and retain exceptional\nbody\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What was the motion manipulation technique used in the AI Golf swing analysis tool?\nAnswer: "}, {"role": "assistant", "content": "The motion manipulation technique used in the AI Golf swing analysis tool involves a decoder structure called a motion manipulator. This structure is used to restore motion from the latent space. The restoration results suggest that the network can retrieve intermediate human poses between two motions that do not exist in the original dataset. Instead of teaching beginners directly about professional forms, the motion manipulator can provide them with an intermediate pose that is more acceptable for beginners to start with."}]} +{"messages": [{"role": "user", "content": "You are an expert Q&A system that strictly operates in two modes when refining existing answers:\n1. **Rewrite** an original answer using the new context.\n2. **Repeat** the original answer if the new context isn't useful.\nNever reference the original answer or context directly in your answer.\nWhen in doubt, just repeat the original answer.New Context: to restore\nthe intermediate human poses between two motions from the latent space. Finally, we suggest possible\napplications for analyzing and visualizing the discrepancy between the two input motions and interacting\nwith the users. With the proposed application, users can easily understand the differences between their\nmotions and those of various experts during self-training and learn how to improve their motions.\nINDEX TERMS Computer vision, machine learning, motor skill training, golf.\nI. INTRODUCTION\nIn sports, it is dif\ufb01cult for beginners to improve their skills\nwithout prior knowledge or assistance from coaches. As a\nconventional method, people go to lessons to meet experts\nand learn how to play in the proper form. However, in most\nsports, players spend considerable time training alone to\nachieve outstanding results in the \ufb01eld and retain exceptional\nbody conditions. Therefore, it is important to design and\nimplement an effective and accurate self-training process for\nsuch situations.\nIn the \ufb01eld of the acquisition of sports skills, one way\nto improve sports skills is to replicate professional players\u2019\nmotions. People watch the movements of professional ath-\nThe associate editor coordinating the review of this manuscript and\napproving it for publication was Jinjia Zhou\n .letes on television or the Internet and try to make their bodies\nmove similarly to professionals. To accelerate this process,\nmany systems have recently been developed to help users\nunderstand the movements of professionals [1], [2], [3], [4],\n[5]. However, in these previous works and systems, users may\nstruggle to re\ufb01ne their movement with no idea which timing\nof the whole motion, which parts of the body they should\nfocus on, or how they can change their body movements to\nget their form closer to that of professionals.\nWith signi\ufb01cant advances in machine learning technolo-\ngies, many systems have been built to recognize differ-\nent objects, make predictions for decisions, or even predict\nthe future [6].\nQuery: What was the motion manipulation technique used in the AI Golf swing analysis tool?\nOriginal Answer: The motion manipulation technique used in the AI Golf swing analysis tool involves a decoder structure called a motion manipulator. This structure is used to restore motion from the latent space. The restoration results suggest that the network can retrieve intermediate human poses between two motions that do not exist in the original dataset. Instead of teaching beginners directly about professional forms, the motion manipulator can provide them with an intermediate pose that is more acceptable for beginners to start with.\nNew Answer: "}, {"role": "assistant", "content": "The AI Golf swing analysis tool utilizes a motion manipulation technique that incorporates a decoder structure known as a motion manipulator. This mechanism is employed to recover motion from a latent space. The restoration outcomes indicate that the system can extract intermediate human poses between two movements that are not present in the original dataset. Rather than instructing novices directly about professional forms, the motion manipulator can offer them an intermediate pose that is easier for them to begin with."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106290\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nFIGURE 4. Three types of datasets.\nC. MOTION MANIPULATOR: DISCOVERING\nINTERMEDIATE MOTION BETWEEN HUMAN POSES\nDuring self-training, it is not always simple for beginners\nto imitate an ideal motion form, which is very different\nfrom their current form. In this study, we propose a motion\ndecoder to generate an intermediate motion. As mentioned\npreviously, the TCC network is trained to learn a latent space\nthat shows motion similarity. Therefore, we understand that\na high-dimensional data point in latent space can be rep-\nresentative of a human pose. To retrieve the intermediate\nmotion between two points in the latent space, we train a\ndecoder using latent vectors to predict human poses that\nare the same as the inputs. In particular, we \ufb01rst input the\ntraining set data into the trained TCC network to obtain the\noutputs of the latent vectors. Next, using the latent vectors\nas inputs, we trained a simple decoder consisting of a single\nfully-connected layer to produce the outputs of the human\nposes. For the loss function LMSE, we take the mean square\nerror (MSE) between the output and input poses:\nLMSE=\u2211\ni||Yi\u2212Xi||2(4)\nwhere Xiis the ithjoint of the input human pose XandYiis\ntheithjoint of the output human pose Y.\nAfter training the motion decoder, we retrieve a new latent\nvector Vinterbetween the two timing-matched latent vectors\nusing linear interpolation:\nVinter=(1\u2212\u03b1)\u00d7V1+\u03b1\u00d7V2 (5)\nwhere V1andV2are the two latent vectors synchronized by\nthe motion synchronizer, and \u03b1\u22080.0,1.0 is the magnitude\nparameter. In the above formulation, by increasing the value\nof\u03b1, we can obtain a human pose whose latent vector is closer\ntoV2, and the restored human pose should ideally be more\nsimilar to the human poses generated from V2.\nIV. EXPERIMENTAL SETUP\nTo evaluate the accuracy and effectiveness of the three mod-\nules introduced in the previous section, we collected golf\nswing data via the Internet and created a pseudo database\nconsisting of raw video data, without-background-video data,\nand 3D pose data (Figure 4). Next, we implemented three\nmodels of the network (V-TCC, S-TCC, and SA-TCC) uti-\nlizing TCC loss and conducted statistical analysis under four\ndifferent conditions:\n\u2022V-TCC using video inputs with backgrounds\nFIGURE 5. Key event and phase. The impact moment and the top moment\nare labeled as key events. Frames between them are labeled as swinging\ndown phases.\n\u2022V-TCC using video inputs without backgrounds\n\u2022S-TCC using 3D human pose inputs\n\u2022SA-TCC using 3D human pose inputs\nThe following sections introduce the data collection pro-\ncess and the evaluation metrics used in this study.\nA. VIDEO DATASET\nAs the training dataset, we used GolfDB [33], a video dataset\ncollection for all types of golf iron swings and driver swings,\nconsisting of 1400 high-quality golf swing videos of male and\nfemale professional golfers. Although the GolfDB provided\npreprocessed video clips for a frame size of 160 \u00d7160,\nwe rebuilt our pseudo dataset with high-resolution videos.\nIn addition to clean videos, we created another video dataset\nwithout background information. This was because, in our\nhypothesis, the background information, for example, the\nhuman shadow, might in\ufb02uence the alignment of the network.\nMoreover, the network might also learn the motion of props,\nsuch as golf clubs. We used Mask R-CNN [34], a generic\nobject detection and segmentation network, to detect the\nhuman body in a single image frame. We then removed the\nbackground pixels and left only the human body (Figure 4).\nB. 3D POSE DATASET\nTo conduct a more precise analysis of only human poses,\nwe created a new pseudo dataset consisting of 3D point\ndata of human body poses. In this dataset, we \ufb01rst used\nHRNet [35] to retrieve the time series of 2D human poses\nfrom golf-swing videos.\n\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool?\nAnswer: "}, {"role": "assistant", "content": "The TCC algorithm in the AI Golf swing analysis tool is used to learn a latent space that shows motion similarity. This high-dimensional data point in latent space can represent a human pose. The algorithm is trained to generate an intermediate motion between two points in the latent space, which can be particularly useful for beginners who find it challenging to imitate an ideal motion form. The TCC network is also used in the motion synchronizer to align motions with different phases and timings."}]} +{"messages": [{"role": "user", "content": "You are an expert Q&A system that strictly operates in two modes when refining existing answers:\n1. **Rewrite** an original answer using the new context.\n2. **Repeat** the original answer if the new context isn't useful.\nNever reference the original answer or context directly in your answer.\nWhen in doubt, just repeat the original answer.New Context: we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\nQuery: Question: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool?\nOriginal Answer: The TCC algorithm in the AI Golf swing analysis tool is used to learn a latent space that shows motion similarity. This high-dimensional data point in latent space can represent a human pose. The algorithm is trained to generate an intermediate motion between two points in the latent space, which can be particularly useful for beginners who find it challenging to imitate an ideal motion form. The TCC network is also used in the motion synchronizer to align motions with different phases and timings.\nNew Answer: "}, {"role": "assistant", "content": "The TCC algorithm in the AI Golf swing analysis tool serves to identify and illustrate motion similarity in a latent space, which can effectively represent a human pose. It is designed to generate an intermediate motion between two points within this latent space, offering a beneficial tool for beginners who may struggle to replicate an ideal golf swing. Additionally, the TCC network is utilized in the motion synchronizer to align golf swings that may differ in phase and timing. The improved SA-TCC network further enhances phase classification accuracy and demonstrates a strong correlation between the distance in the latent space and the MPJPE."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n\npage_label: 106287\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nCompared to existing methods,\nthe main contributions of this study can be summarized as\nfollows:\n\u2022A golf swing analysis method and its applications are\nintroduced.\n\u2022The proposed method distinguishes when the difference\nbetween two motions is large and small.\n\u2022The proposed method helps users understand the differ-\nence between themselves and professional players.\n\u2022The proposed method provides intermediate poses that\nare acceptable during the early learning phase of sports.\n\u2022Crucial factors that can in\ufb02uence the accuracy of sports\nanalysis are discussed.\nII. RELATED WORK\nA. SPORTS TRAINING FOR SKILL ACQUISITION\nMany recent studies have focused on developing sports\ntraining systems to help beginners improve their skills. Stud-\nies [13], [14], [15] have proposed multi-modal sports training\nsystems based on sports theories. In their system, users\nreceived visual, haptic, and audio feedback when they did not\nideally move their bodies or instruments. However, the ideal\nmovement could differ from sport to sport, meaning that it\nmay be dif\ufb01cult to generalize these methodologies to other\nsports.\nOn the other hand, another way to learn sports skills is\nimitating professional players\u2019 motions [1], [3], [4], [5].\nIkeda et al. [16] proposed a golf swing training system that\nuses the motions of professional golfers. In their system,\na user\u2019s motion was synchronized with a selected ideal\nprofessional\u2019s motion, and the two motions were overlaid and\nprojected onto the ground during training. Sasaki et al. [17]\nalso reported the importance of beginners copying expert\nmotions and proposed a climbing training system using pose\nprediction. Their system predicted and visualized the pose of\nexperts based on the user\u2019s current hand and foot positions.\nWhile these recent studies have shown the effectiveness of\nusing multi-modal feedback and the potential of applying\nneural networks to create AI teachers for sports training, it is\ndif\ufb01cult for users to change their motion forms immediately\nto match ideal forms. Thus, building a system that can teach\nusers step-by-step to improve their sports skills remains\nchallenging.\nB. VIDEO AND MOTION ALIGNMENT\nAn ef\ufb01cient way to evaluate whether a person is performing\na motion correctly is to compare their motion with oth-\ners whose motion is considered correct. However, owing\nto the various timings and speeds of motion of different\nVOLUME 10, 2022 106287\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool?\nAnswer: "}, {"role": "assistant", "content": "The motion discrepancy detector in the AI Golf Swing Analysis Tool is used to find fine-grained differences between golf swings in the latent space. It can distinguish whether the difference between two motions is small or large. This helps users understand the difference between their own golf swing and that of professional players."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106287\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nCompared to existing methods,\nthe main contributions of this study can be summarized as\nfollows:\n\u2022A golf swing analysis method and its applications are\nintroduced.\n\u2022The proposed method distinguishes when the difference\nbetween two motions is large and small.\n\u2022The proposed method helps users understand the differ-\nence between themselves and professional players.\n\u2022The proposed method provides intermediate poses that\nare acceptable during the early learning phase of sports.\n\u2022Crucial factors that can in\ufb02uence the accuracy of sports\nanalysis are discussed.\nII. RELATED WORK\nA. SPORTS TRAINING FOR SKILL ACQUISITION\nMany recent studies have focused on developing sports\ntraining systems to help beginners improve their skills. Stud-\nies [13], [14], [15] have proposed multi-modal sports training\nsystems based on sports theories. In their system, users\nreceived visual, haptic, and audio feedback when they did not\nideally move their bodies or instruments. However, the ideal\nmovement could differ from sport to sport, meaning that it\nmay be dif\ufb01cult to generalize these methodologies to other\nsports.\nOn the other hand, another way to learn sports skills is\nimitating professional players\u2019 motions [1], [3], [4], [5].\nIkeda et al. [16] proposed a golf swing training system that\nuses the motions of professional golfers. In their system,\na user\u2019s motion was synchronized with a selected ideal\nprofessional\u2019s motion, and the two motions were overlaid and\nprojected onto the ground during training. Sasaki et al. [17]\nalso reported the importance of beginners copying expert\nmotions and proposed a climbing training system using pose\nprediction. Their system predicted and visualized the pose of\nexperts based on the user\u2019s current hand and foot positions.\nWhile these recent studies have shown the effectiveness of\nusing multi-modal feedback and the potential of applying\nneural networks to create AI teachers for sports training, it is\ndif\ufb01cult for users to change their motion forms immediately\nto match ideal forms. Thus, building a system that can teach\nusers step-by-step to improve their sports skills remains\nchallenging.\nB. VIDEO AND MOTION ALIGNMENT\nAn ef\ufb01cient way to evaluate whether a person is performing\na motion correctly is to compare their motion with oth-\ners whose motion is considered correct. However, owing\nto the various timings and speeds of motion of different\nVOLUME 10, 2022 106287\n\npage_label: 106286\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nPeople watch the movements of professional ath-\nThe associate editor coordinating the review of this manuscript and\napproving it for publication was Jinjia Zhou\n .letes on television or the Internet and try to make their bodies\nmove similarly to professionals. To accelerate this process,\nmany systems have recently been developed to help users\nunderstand the movements of professionals [1], [2], [3], [4],\n[5]. However, in these previous works and systems, users may\nstruggle to re\ufb01ne their movement with no idea which timing\nof the whole motion, which parts of the body they should\nfocus on, or how they can change their body movements to\nget their form closer to that of professionals.\nWith signi\ufb01cant advances in machine learning technolo-\ngies, many systems have been built to recognize differ-\nent objects, make predictions for decisions, or even predict\nthe future [6]. Researchers have focused on producing\nself-training systems with neural networks [7]. A recent\nstudy [8] introduced a climbing training system in which\nusers can receive recommended poses and movements to\n106286This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/VOLUME 10, 2022\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the method proposed in the study to improve users' forms by comparing their motions with those of professionals?\nAnswer: "}, {"role": "assistant", "content": "The study proposes a golf swing analysis method that helps users understand the difference between their own motions and those of professional players. This method also provides intermediate poses that are acceptable during the early learning phase of sports. It distinguishes when the difference between two motions is large and small, aiding users in refining their movements."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106288\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nXis the input motion sequence and Yis the\noutput human poses restored from the latent space.\nabnormal images. However, although the network could eas-\nily detect surprise samples, \ufb01ne-grained differences among\nstandard samples were not discussed. In addition to image-\nbased methods, recent studies have focused on systems that\nuse 3D human pose information [29], [30]. In this study,\nwe apply discrepancy detection to both videos and 3D\nhuman poses and discuss the ability of the system to detect\n\ufb01ne-grained differences between two input motions.\nIII. METHODS\nThis study aims to create a system that captures user motions\nand provides \ufb01ne-grained feedback to improve users\u2019 forms\nby comparing their motions with those of professionals.\nTo achieve the goal of building such an application, the\nmethod proposed in this study is to \ufb01rst train a neural network\nwith professionals\u2019 motion data. After training the network,\nthe system compresses the user motions through the network\ninto a latent space and compares their motions with those of\nprofessionals in the latent space. Figure 1 shows an overview\nof the proposed system. The work\ufb02ow of the approach is\ndivided into three parts: motion synchronization, motion dis-\ncrepancy detection, and motion manipulation. The system\n\ufb01rst receives two motion inputs X1andX2and uses encoder\nEto embed the input motions into the latent space, where the\ntwo motions are represented as L1andL2. The encoder is\ntrained to learn a latent space in which similar motions appear\nto be close.\nNext, using the learned latent space, the motion synchro-\nnizer MSmatches the timing of the two motions in the\nlatent space by measuring the Euclidean distance between\nL1andL2. The motion discrepancy detector then captures\nthe two synchronized latent vectors V1andV2, measures the\ndifference between them, and passes a distance vector D12to\nthe motion manipulator MM.\nFinally, the MM integrates V1,V2, and D12to specify\nthe key frames where large differences occur and create an\nintermediate latent vector Vinter. The system uses decoder D\nto restore the intermediate human poses Yfrom Vinterfor\nusers to gradually improve their motion forms.\nA. MOTION SYNCHRONIZER: ALIGNING MOTION\nSEQUENCES WITH DIFFERENT TIMING\nFor motion synchronization, we aim to design a network\nthat learns a latent space that shows motion similarity.\nA common way to achieve this is by constructing an\nautoencoder and decoder, whose input and output are the\nsame motions. On the other hand, previous studies have\n106288 VOLUME 10, 2022\n\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What is the purpose of the proposed system described in the context?\nAnswer: "}, {"role": "assistant", "content": "The proposed system is designed to analyze golf swings using neural networks. It aims to help users understand the differences between their own golf swings and those of professional players. The system works in three parts: synchronization, discrepancy detection, and manipulation. It aligns motions with different phases and timings, finds fine-grained differences between golf swings in a latent space, and restores motion from the latent space. The system provides users with an intermediate pose that is more acceptable for beginners to start with, rather than directly teaching them about professional forms. The ultimate goal is to help users efficiently learn an ideal form in a gradual manner without the need for a coach during self-training."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106287\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nCompared to existing methods,\nthe main contributions of this study can be summarized as\nfollows:\n\u2022A golf swing analysis method and its applications are\nintroduced.\n\u2022The proposed method distinguishes when the difference\nbetween two motions is large and small.\n\u2022The proposed method helps users understand the differ-\nence between themselves and professional players.\n\u2022The proposed method provides intermediate poses that\nare acceptable during the early learning phase of sports.\n\u2022Crucial factors that can in\ufb02uence the accuracy of sports\nanalysis are discussed.\nII. RELATED WORK\nA. SPORTS TRAINING FOR SKILL ACQUISITION\nMany recent studies have focused on developing sports\ntraining systems to help beginners improve their skills. Stud-\nies [13], [14], [15] have proposed multi-modal sports training\nsystems based on sports theories. In their system, users\nreceived visual, haptic, and audio feedback when they did not\nideally move their bodies or instruments. However, the ideal\nmovement could differ from sport to sport, meaning that it\nmay be dif\ufb01cult to generalize these methodologies to other\nsports.\nOn the other hand, another way to learn sports skills is\nimitating professional players\u2019 motions [1], [3], [4], [5].\nIkeda et al. [16] proposed a golf swing training system that\nuses the motions of professional golfers. In their system,\na user\u2019s motion was synchronized with a selected ideal\nprofessional\u2019s motion, and the two motions were overlaid and\nprojected onto the ground during training. Sasaki et al. [17]\nalso reported the importance of beginners copying expert\nmotions and proposed a climbing training system using pose\nprediction. Their system predicted and visualized the pose of\nexperts based on the user\u2019s current hand and foot positions.\nWhile these recent studies have shown the effectiveness of\nusing multi-modal feedback and the potential of applying\nneural networks to create AI teachers for sports training, it is\ndif\ufb01cult for users to change their motion forms immediately\nto match ideal forms. Thus, building a system that can teach\nusers step-by-step to improve their sports skills remains\nchallenging.\nB. VIDEO AND MOTION ALIGNMENT\nAn ef\ufb01cient way to evaluate whether a person is performing\na motion correctly is to compare their motion with oth-\ners whose motion is considered correct. However, owing\nto the various timings and speeds of motion of different\nVOLUME 10, 2022 106287\n\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What are the three main contributions of the proposed golf swing analysis tool described in the context?\nAnswer: "}, {"role": "assistant", "content": "The three main contributions of the proposed golf swing analysis tool are as follows: \n\n1. The development of a motion synchronizer that aligns motions with different phases and timings, which has shown better performance than state-of-the-art video implementations.\n2. The creation of a motion discrepancy detector that can find fine-grained differences between golf swings in the latent space. This detector can distinguish whether the difference between two motions is small or large.\n3. The introduction of a decoder structure called a motion manipulator that can restore motion from the latent space. This can provide beginners with an intermediate pose that is more acceptable for them to start with, rather than directly teaching them professional forms."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106287\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nCompared to existing methods,\nthe main contributions of this study can be summarized as\nfollows:\n\u2022A golf swing analysis method and its applications are\nintroduced.\n\u2022The proposed method distinguishes when the difference\nbetween two motions is large and small.\n\u2022The proposed method helps users understand the differ-\nence between themselves and professional players.\n\u2022The proposed method provides intermediate poses that\nare acceptable during the early learning phase of sports.\n\u2022Crucial factors that can in\ufb02uence the accuracy of sports\nanalysis are discussed.\nII. RELATED WORK\nA. SPORTS TRAINING FOR SKILL ACQUISITION\nMany recent studies have focused on developing sports\ntraining systems to help beginners improve their skills. Stud-\nies [13], [14], [15] have proposed multi-modal sports training\nsystems based on sports theories. In their system, users\nreceived visual, haptic, and audio feedback when they did not\nideally move their bodies or instruments. However, the ideal\nmovement could differ from sport to sport, meaning that it\nmay be dif\ufb01cult to generalize these methodologies to other\nsports.\nOn the other hand, another way to learn sports skills is\nimitating professional players\u2019 motions [1], [3], [4], [5].\nIkeda et al. [16] proposed a golf swing training system that\nuses the motions of professional golfers. In their system,\na user\u2019s motion was synchronized with a selected ideal\nprofessional\u2019s motion, and the two motions were overlaid and\nprojected onto the ground during training. Sasaki et al. [17]\nalso reported the importance of beginners copying expert\nmotions and proposed a climbing training system using pose\nprediction. Their system predicted and visualized the pose of\nexperts based on the user\u2019s current hand and foot positions.\nWhile these recent studies have shown the effectiveness of\nusing multi-modal feedback and the potential of applying\nneural networks to create AI teachers for sports training, it is\ndif\ufb01cult for users to change their motion forms immediately\nto match ideal forms. Thus, building a system that can teach\nusers step-by-step to improve their sports skills remains\nchallenging.\nB. VIDEO AND MOTION ALIGNMENT\nAn ef\ufb01cient way to evaluate whether a person is performing\na motion correctly is to compare their motion with oth-\ners whose motion is considered correct. However, owing\nto the various timings and speeds of motion of different\nVOLUME 10, 2022 106287\n\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the main purpose of the proposed application for analyzing and visualizing golf swing motions?\nAnswer: "}, {"role": "assistant", "content": "The main purpose of the proposed application for analyzing and visualizing golf swing motions is to help users intuitively understand the difference between their own golf swings and those of professional players. It provides an interface where users can select a professional's form from a database and compare it with their own. The application also offers a motion manipulator that allows users to gradually imitate intermediate human poses, rather than directly imitating the selected motion. This helps users learn an ideal form in a gradual manner, rather than trying to directly imitate the ideal motion. The ultimate goal is to enable users to learn to play sports without the help of coaches during self-training."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106295\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIEEE/CVF Conf.\nComput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 5686\u20135696, doi:\n10.1109/CVPR.2019.00584.\n[36] J. Martinez, R. Hossain, J. Romero, and J. J. Little, \u2018\u2018A simple yet effective\nbaseline for 3D human pose estimation,\u2019\u2019 in Proc. IEEE Int. Conf. Comput.\nVis. (ICCV), Oct. 2017, pp. 2659\u20132668, doi: 10.1109/ICCV.2017.288.\nCHEN-CHIEH LIAO received the B.E. and M.S.\ndegrees from the Tokyo Institute of Technol-\nogy, Japan, in 2020 and 2022, respectively,\nwhere he is currently pursuing the Ph.D. degree\nwith the School of Computing. His research\ninterests include human\u2013computer interaction,\nmixed/augmented reality, and machine vision.\nDONG-HYUN HWANG received the B.S. degree\nfrom the School of Computer Science and Engi-\nneering, Korea University of Technology and\nEducation, Cheonan, South Korea, in 2017, and\nthe M.S. and Ph.D. degrees from the School\nof Computing, Tokyo Institute of Technology,\nTokyo, Japan, in 2019 and 2022, respectively.\nHe was a Young Research Fellow at the Japan\nSociety for the Promotion of Science (JSPS),\nfrom 2020 to 2022. Currently, he is a machine\nlearning researcher at CLOVA Voice&Avatar, NAVER Corporation. His\nresearch interests include mixed/augmented reality, computer vision, and\nmachine learning.\nHIDEKI KOIKE received the B.E., M.E., and\nDr.Eng. degrees from the University of Tokyo,\nin 1986, 1988, and 1991, respectively. He was\nat the University of Electro-Communications,\nTokyo. In 2014, he joined Tokyo Institute of\nTechnology, Japan, where he is a Professor\nwith the School of Computing. His research\ninterests include vision-based human\u2013computer\ninteraction, human augmentation, information\nvisualization, and usable security.\nVOLUME 10, 2022 106295\n\npage_label: 106295\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nInt.\nConf. Artif. Reality Telexistence Eurographics Symp. Virtual Environ.,\n2018, pp. 171\u2013174, doi: 10.2312/egve.20181330.\n[20] T. Halperin, A. Ephrat, and S. Peleg, \u2018\u2018Dynamic temporal\nalignment of speech to lips,\u2019\u2019 in Proc. IEEE Int. Conf. Acoust.,\nSpeech Signal Process. (ICASSP), May 2019, pp. 3980\u20133984, doi:\n10.1109/ICASSP.2019.8682863.\n[21] I. Misra, C. L. Zitnick, and M. Hebert, \u2018\u2018Shuf\ufb02e and learn: Unsupervised\nlearning using temporal order veri\ufb01cation,\u2019\u2019 in Computer Vision ECCV\n2016 (Lecture Notes in Computer Science), vol. 9905, B. Leibe, J. Matas,\nN. Sebe, and M. Welling, Eds. Cham, Switzerland: Springer, 2016, doi:\n10.1007/978-3-319-46448-0_32.\n[22] P. Sermanet, C. Lynch, Y. Chebotar, J. Hsu, E. Jang, S. Schaal, S. Levine,\nand G. Brain, \u2018\u2018Time-contrastive networks: Self-supervised learning from\nvideo,\u2019\u2019 in Proc. IEEE Int. Conf. Robot. Autom. (ICRA), May 2018,\npp. 1134\u20131141, doi: 10.1109/ICRA.2018.8462891.\n[23] D. Dwibedi, Y. Aytar, J. Tompson, P. Sermanet, and A. Zisserman,\n\u2018\u2018Temporal cycle-consistency learning,\u2019\u2019 in Proc. IEEE/CVF Conf. Com-\nput. Vis. Pattern Recognit. (CVPR), Jun. 2019, pp. 1801\u20131810, doi:\n10.1109/CVPR.2019.00190.\n[24] S. Haresh, S. Kumar, H. Coskun, S. N. Syed, A. Konin, M. Z. Zia, and\nQ.-H. Tran, \u2018\u2018Learning by aligning videos in time,\u2019\u2019 in Proc. IEEE/CVF\nConf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2021, pp. 5544\u20135554,\ndoi:10.1109/CVPR46437.2021.00550.\n[25] H. Seki and Y. Hori, \u2018\u2018Detection of abnormal action using image sequence\nfor monitoring system of aged people,\u2019\u2019 IEEJ Trans. Ind. Appl., vol. 122,\nno. 2, pp. 182\u2013188, 2002, doi: 10.1541/ieejias.122.182.\n[26] G. S. Parra-Dominguez, B. Taati, and A. Mihailidis, \u2018\u20183D human motion\nanalysis to detect abnormal events on stairs,\u2019\u2019 in Proc. 2nd Int. Conf. 3D\nImag., Modeling, Process., Vis. Transmiss., Oct. 2012, pp. 97\u2013103, doi:\n10.1109/3DIMPVT.2012.34.\n[27] F. Nater, H. Grabner, and L. Van Gool, \u2018\u2018Exploiting simple hierarchies\nfor unsupervised human behavior analysis,\u2019\u2019 in Proc. IEEE Comput. Soc.\nConf. Comput. Vis. Pattern Recognit. , Jun. 2010, pp. 2014\u20132021, doi:\n10.1109/CVPR.2010.5539877.\n[28] D. Abati, A. Porrello, S. Calderara, and R. Cucchiara, \u2018\u2018Latent space\nautoregression for novelty detection,\u2019\u2019 in Proc. IEEE/CVF Conf. Com-\nput. Vis. Pattern Recognit. (CVPR), Jun. 2019, pp. 481\u2013490, doi:\n10.1109/CVPR.2019.00057.\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the title of the paper mentioned in reference [7] and what is its subject matter?\nAnswer: "}, {"role": "assistant", "content": "The context does not provide information on the title or the subject matter of the paper mentioned in reference [7]."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106291\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nframe showing a particular moment, and the phase is a time\nseries between two key events. For example, as shown in\nFigure 5, a key event in golf may be the moment when the golf\nclub hits the ball (impact), and the motion before the golf club\nhits the ball can be considered as the phase of the golf club\napproaching the ball (swinging down). Note that all the\nframes in the period between two key events have the same\nphase label. Following the key event annotation of GolfDB,\nwe labeled our pseudo dataset with eight key events: address,\ntoe-up, mid-backswing, top, mid-downswing, impact, mid-\nfollow-through, and \ufb01nish. Phases were labeled between\nevery two key events, for a total of seven phases.\nThe phase classi\ufb01cation accuracy was per frame phase\nclassi\ufb01cation. To calculate the accuracy, we \ufb01rst used the\nencoders of our TCC networks to extract latent vectors.\nWe then trained a simple classi\ufb01er on the latent vectors to\npredict the labeled phases. The classi\ufb01er was trained under\nseveral conditions by changing the percentage of the given\nlabeled data. After the classi\ufb01er was trained, we used all\nlabeled data to calculate the phase classi\ufb01cation accuracy.\nIn general, the larger the size of the given labeled data, the\nhigher the accuracy of the classi\ufb01er.\nTo explore more speci\ufb01cally the effectiveness of the net-\nwork in detecting discrepant motion, we investigated whether\nthe distance in the latent space could represent the discrep-\nancy between two input sequences by computing Pearson\u2019s\ncorrelation coef\ufb01cient. Because 3D poses varied for differ-\nent camera views, we could not directly compare the two\nhuman poses using the results from the 3d pose estimator.\nThus, we applied Procrustes analysis to align the two human\nposes to ensure that they were seen from the same camera\ndirection. In our experiment, we compared the distance in the\nlatent space with the mean per joint point error (MPJPE) and\nmeasured the Pearson\u2019s correlation coef\ufb01cient \u03c1:\n\u03c1D,E=cov(D, E)\n\u03c3D\u03c3E(6)\nwhere covis the covariance; Dis the distance in the latent\nspace; Eis the MPJPE; and \u03c3Dand\u03c3Eare the standard\ndeviations of DandE.\nV. RESULTS\nThis chapter presents the results of an early qualitative\nanalysis investigating the potential of the proposed method\nto detect discrepant motion differences, followed by more\ndetailed results comparing different modules. Finally, the\nintermediate human poses generated by the motion manip-\nulator are visualized for qualitative studies.\nA. CASE STUDY\nIn our early study, we conducted a qualitative analysis by\nexploring the latent space to investigate whether the network\ncould trace \ufb01ne-grained differences. We \ufb01rst used the V-TCC\nto synchronize the swing motions of professionals and begin-\nners. We then computed the distances between the aligned\nvideos in the latent space and visualized the overlaid 3D\nhuman poses for qualitative comparison (Figure 6).TABLE 1. Phase classification accuracy. This is the accuracy metric\nshowing the ability of the network to classify any given motion frame to\nits corresponding phase.\nB. PHASE CLASSIFICATION ACCURACY\nWe trained the V-TCC using the GolfDB video dataset with\nand without background subtraction. In contrast, S-TCC and\nSA-TCC were trained using the pseudo skeleton dataset\nwith unit vector normalization. After training the four mod-\nels, we computed the phase classi\ufb01cation accuracy for each\ntrained model by assigning 1%, 5%, 10%, 30%, and 80% of\nlabeled data.\nThe results are presented in Table 1. As we trained the\nnetwork properly, the phase classi\ufb01cation accuracy was low\nwhen the given labeled data were insuf\ufb01cient and rose with\nan increase in the number of labeled data.\nC. CORRELATION\nWe computed Pearson\u2019s correlation coef\ufb01cient using the\nfour models. For video inputs, a 0.69 Pearson\u2019s correla-\ntion coef\ufb01cient was measured for regular video, and a\n0.72 Pearson\u2019s correlation coef\ufb01cient was obtained when the\nbackground was removed (Figure 7).\n\npage_label: 106291\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIn contrast, S-TCC and\nSA-TCC were trained using the pseudo skeleton dataset\nwith unit vector normalization. After training the four mod-\nels, we computed the phase classi\ufb01cation accuracy for each\ntrained model by assigning 1%, 5%, 10%, 30%, and 80% of\nlabeled data.\nThe results are presented in Table 1. As we trained the\nnetwork properly, the phase classi\ufb01cation accuracy was low\nwhen the given labeled data were insuf\ufb01cient and rose with\nan increase in the number of labeled data.\nC. CORRELATION\nWe computed Pearson\u2019s correlation coef\ufb01cient using the\nfour models. For video inputs, a 0.69 Pearson\u2019s correla-\ntion coef\ufb01cient was measured for regular video, and a\n0.72 Pearson\u2019s correlation coef\ufb01cient was obtained when the\nbackground was removed (Figure 7). For skeleton inputs,\nan over 0.76 Pearson\u2019s correlation coef\ufb01cient was found\nwhen using the SA-TCC; however, the lowest Pearson\u2019s cor-\nrelation coef\ufb01cient with 0.51 was obtained from the S-TCC\n(Figure 8).\nD. MOTION INTERPOLATION\nFor the qualitative results, we computed and visualized the\nintermediate human pose between a pair of human poses\nconsidering the following three circumstances:\n\u2022The two poses were from a single person. The two poses\nwere in different phases (Figure 9(a)).\n\u2022The two poses were from different individuals. The two\nposes were in the same phase (Figure 9(b)).\n\u2022The two poses were from different individuals. The two\nposes were in different phases (Figure 9(c)).\nVI. DISCUSSION\nA. CASE STUDY\nIn our early case study, we observed that in most cases, when\nthe distance in the latent space was small, the difference\nbetween the two 3D poses was small(Figure 6c). On the\nother hand, when the distance was considerable, the 3D poses\nshowed more differences(Figure 6b, d). This suggested that\nthe network could distinguish between beginners and profes-\nsionals. Thus, we can use this feature to retrieve motion fea-\ntures that help users target the key moves they have to correct.\nHowever, as depicted in Figure 6(a), the distance in the latent\nspace remained large even in certain circumstances where the\nVOLUME 10, 2022 106291\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What is the phase classification accuracy metric used to evaluate the network's ability to classify motion frames to their corresponding phases?\nAnswer: "}, {"role": "assistant", "content": "The phase classification accuracy metric is used to evaluate the network's ability to classify any given motion frame to its corresponding phase. The accuracy is calculated by first using the encoders of the TCC networks to extract latent vectors. Then, a simple classifier is trained on the latent vectors to predict the labeled phases. The classifier is trained under several conditions by changing the percentage of the given labeled data. After the classifier is trained, all labeled data is used to calculate the phase classification accuracy. Generally, the larger the size of the given labeled data, the higher the accuracy of the classifier."}]} +{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106291\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIn contrast, S-TCC and\nSA-TCC were trained using the pseudo skeleton dataset\nwith unit vector normalization. After training the four mod-\nels, we computed the phase classi\ufb01cation accuracy for each\ntrained model by assigning 1%, 5%, 10%, 30%, and 80% of\nlabeled data.\nThe results are presented in Table 1. As we trained the\nnetwork properly, the phase classi\ufb01cation accuracy was low\nwhen the given labeled data were insuf\ufb01cient and rose with\nan increase in the number of labeled data.\nC. CORRELATION\nWe computed Pearson\u2019s correlation coef\ufb01cient using the\nfour models. For video inputs, a 0.69 Pearson\u2019s correla-\ntion coef\ufb01cient was measured for regular video, and a\n0.72 Pearson\u2019s correlation coef\ufb01cient was obtained when the\nbackground was removed (Figure 7). For skeleton inputs,\nan over 0.76 Pearson\u2019s correlation coef\ufb01cient was found\nwhen using the SA-TCC; however, the lowest Pearson\u2019s cor-\nrelation coef\ufb01cient with 0.51 was obtained from the S-TCC\n(Figure 8).\nD. MOTION INTERPOLATION\nFor the qualitative results, we computed and visualized the\nintermediate human pose between a pair of human poses\nconsidering the following three circumstances:\n\u2022The two poses were from a single person. The two poses\nwere in different phases (Figure 9(a)).\n\u2022The two poses were from different individuals. The two\nposes were in the same phase (Figure 9(b)).\n\u2022The two poses were from different individuals. The two\nposes were in different phases (Figure 9(c)).\nVI. DISCUSSION\nA. CASE STUDY\nIn our early case study, we observed that in most cases, when\nthe distance in the latent space was small, the difference\nbetween the two 3D poses was small(Figure 6c). On the\nother hand, when the distance was considerable, the 3D poses\nshowed more differences(Figure 6b, d). This suggested that\nthe network could distinguish between beginners and profes-\nsionals. Thus, we can use this feature to retrieve motion fea-\ntures that help users target the key moves they have to correct.\nHowever, as depicted in Figure 6(a), the distance in the latent\nspace remained large even in certain circumstances where the\nVOLUME 10, 2022 106291\n\npage_label: 106292\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nFIGURE 6. Case study with the V-TCC. The line graph shows the distance between two synchronized motions in the latent space. The red line in the\ngraph indicates the threshold for discrepancy detection. The colored skeleton and black skeleton indicate the user\u2019s pose and expert\u2019s pose,\nrespectively. The density and radius of red spheres indicate the degree of joint position difference between the two skeletons.\nFIGURE 7. Pearson\u2019s correlation test for V-TCC. Left: normal videos. Right:\nvideos without background.\nFIGURE 8. Pearson\u2019s correlation test with skeleton input. Left: S-TCC.\nRight: SA-TCC.\njoint difference was negligible. This might be explained by\nthe fact that the network focused not only on human motion\nbut also on other motion features. In this case, the golf club\u2019s\nmovement, which was also critical during the swing, might\nbe the main factor causing the enormous distance in the latent\nspace. This led us to the following quantitative studies, where\nwe conducted statistical tests to examine the accuracy of the\nsynchronization and the correlation between the latent space\nand joint difference under various conditions.\nB. PHASE CLASSIFICATION ACCURACY\nIn this quantitative study, we discovered that the previous\nTCC implementation had limited precision in terms of syn-\nchronization. As our hypothesis suggests, synchronization\nquality might be in\ufb02uenced by various background informa-\ntion and other movements, such as the golf club\u2019s and human\nshadow\u2019s motions. When the background was removed, a sig-\nni\ufb01cant increase in the accuracy of the V-TCC model was\nobserved. This result supported our hypothesis that infor-\nmation outside the human region would affect alignment.Various types of information from the background, the\nmovement of human shadows, and the motion of properties\ncontrolled by a human might be learned as features by the net-\nwork, thus affecting the accuracy of the analysis. Because of\nthe signi\ufb01cant superiority of the without-background version,\nwe considered it necessary to apply background subtraction\nbefore synchronization when building applications for real\nusage.\nNext, considering the skeleton version, both S-TCC and\nSA-TCC outperformed V-TCC under all conditions. Given\nonly 10% of the labeled data, both S-TCC and SA-TCC\nachieved a phase classi\ufb01cation accuracy of over 90%. This\nmight be explained by the fact that the skeleton version\nhad more compressed and precise information than the\nvideo version, where the color information might be noisy.\nAdditionally, 3D poses represented high-level human fea-\ntures. Therefore, the features outside the human region were\nremoved when retrieving 3D human poses, resulting in more\nprecise accuracy for further golf swing analysis. Note that the\nskeleton data were the estimated results from the video data,\nand the estimation could not be perfect for every human pose.\nDespite imperfect skeleton inputs, the S-TCC and SA-TCC\nnetworks outperformed V-TCC. This result suggested that we\ncould use the skeleton version for more precise implementa-\ntion in actual usage.\nFinally, comparing the two skeleton version mod-\nels, S-TCC performed better than SA-TCC under most\nconditions. However, SA-TCC remained competitive in some\ncircumstances (better than S-TCC with 10% labeled data).\nThis led to the following correlation test, in which we dis-\ncussed the ability of the networks to detect discrepancies\nbetween two motions.\nC. CORRELATION\nDiscussing the correlation test, as shown in Figure 7,\nwe observed a correlation between the distance in the latent\nspace and the MPJPE. While the distribution of the data pairs\nseemed to be scattered, a Pearson\u2019s correlation coef\ufb01cient of\nover 0.69 showed the potential of the network to distinguish\nwhether the difference between the two motions was small\nor large. However, as discussed in the case study section,\nin some circumstances, the network failed to detect discrepant\n106292 VOLUME 10, 2022\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What was the lowest Pearson's correlation coefficient obtained from the S-TCC model when using skeleton inputs?\nAnswer: "}, {"role": "assistant", "content": "The lowest Pearson's correlation coefficient obtained from the S-TCC model when using skeleton inputs was 0.51."}]} \ No newline at end of file diff --git a/src/test/regression/regression_test003/generated_data/model.txt b/src/test/regression/regression_test003/generated_data/model.txt new file mode 100644 index 0000000..9eefdd2 --- /dev/null +++ b/src/test/regression/regression_test003/generated_data/model.txt @@ -0,0 +1 @@ +ft:gpt-3.5-turbo-0613:personal::8XsV12eK diff --git a/src/test/regression/regression_test003/generated_data/train_questions.txt b/src/test/regression/regression_test003/generated_data/train_questions.txt new file mode 100644 index 0000000..2a560b7 --- /dev/null +++ b/src/test/regression/regression_test003/generated_data/train_questions.txt @@ -0,0 +1,10 @@ +What was the motion manipulation technique used in the AI Golf swing analysis tool? +Question: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool? +Question: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool? +Question: What is the method proposed in the study to improve users' forms by comparing their motions with those of professionals? +What is the purpose of the proposed system described in the context? +Question: What are the three main contributions of the proposed golf swing analysis tool described in the context? +Question: What is the main purpose of the proposed application for analyzing and visualizing golf swing motions? +Question: What is the title of the paper mentioned in reference [7] and what is its subject matter? +What is the phase classification accuracy metric used to evaluate the network's ability to classify motion frames to their corresponding phases? +What was the lowest Pearson's correlation coefficient obtained from the S-TCC model when using skeleton inputs? diff --git a/src/test/regression/regression_test003/train_questions.txt b/src/test/regression/regression_test003/train_questions.txt deleted file mode 100644 index aeffd2b..0000000 --- a/src/test/regression/regression_test003/train_questions.txt +++ /dev/null @@ -1,3 +0,0 @@ -What additional question is being proposed to be added during the creation of a user profile? -What is the mission of CulturesConnect.me? -What is the purpose of adding additional questions during the creation of a user profile in the dating module? diff --git a/src/test/regression/regression_test004/eval_questions.txt b/src/test/regression/regression_test004/eval_questions.txt deleted file mode 100644 index e42fd84..0000000 --- a/src/test/regression/regression_test004/eval_questions.txt +++ /dev/null @@ -1,40 +0,0 @@ -What is the proposed approach in the golf swing analysis tool mentioned in the context? -What is the main challenge addressed by the proposed golf swing analysis tool using neural networks? -Question: What is the main challenge of the study mentioned in the context? -What is the purpose of the motion manipulation technique used in the AI Golf swing analysis tool? -What is the purpose of the Procrustes analysis mentioned in the experiment? -Question: What is the phase classification accuracy metric used to evaluate the network's ability to classify motion frames to their corresponding phases? -What was the phase classification accuracy of the previous TCC implementation? -Question: What is the main factor that may cause the enormous distance in the latent space in the golf swing analysis? -What is the method proposed by the research group of Nater et al. for detecting abnormal events? -What is the purpose of the system proposed in the study? -Question: What is the title of the paper published in the IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR) in June 2019? -Question: What is the purpose of the AI Golf Swing Analysis Tool mentioned in the context? -Question: What are the three main contributions of the proposed system in this study? -What is the purpose of the proposed golf swing analysis tool that uses neural networks? -What are the main contributions of the proposed golf swing analysis method? -Question: What is the main challenge addressed by the proposed golf swing analysis tool using neural networks? -What is the proposed system in the study for golf swing analysis? -Question: What are the three parts into which the proposed golf swing analysis tool is divided? -Question: What is the purpose of the proposed AI Golf Swing Analysis Tool for Self-Training? -What is the primary goal of the AI Golf Swing Analysis Tool for Self-Training? -Question: What are the main contributions of the study mentioned in the context that applies pose estimation to predict amateur golf swing performance? -Question: Which paper applies pose estimation to predict amateur golf swing performance? -Question: Who authored the paper titled "A simple yet effective baseline for 3D human pose estimation"? -Question: What are the research interests of HIDEKI KOIKE mentioned in the context information? -Question: Who received the B.E. and M.S. degrees from the Tokyo Institute of Technology in 2020 and 2022, respectively, and is currently pursuing the Ph.D. degree with the School of Computing? -Question: Who received the B.S. degree from the School of Computer Science and Engineering, Korea University of Technology and Education, Cheonan, South Korea, in 2017? -What is the reason for the skeleton version performing better than the video version in terms of accuracy for golf swing analysis? -Question: What is the Pearson's correlation coefficient for skeleton inputs using the S-TCC model? -Question: What are the three main contributions of the proposed golf swing analysis tool? -Question: What is the main purpose of the proposed application for golf swing analysis and visualization? -Question: According to the provided context, what is the title of the research paper published in volume 10 of the year 2022? -Question: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool? -Question: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool? -Question: What method was used to remove the background pixels and isolate the human body in the video dataset? -What evaluation metric was used to assess the alignment precision of the network in the study? -Question: What is the main challenge addressed in the proposed golf swing analysis tool using neural networks? -What is the purpose of the AI Golf Swing Analysis Tool mentioned in the context? -What are the main contributions of the study on the golf swing analysis tool for self-training? -What are the main contributions of the study on golf swing analysis? -What is the file size of the file "eval_questions.txt"? diff --git a/src/test/regression/regression_test004/finetuning_events.jsonl b/src/test/regression/regression_test004/finetuning_events.jsonl deleted file mode 100644 index f85b86d..0000000 --- a/src/test/regression/regression_test004/finetuning_events.jsonl +++ /dev/null @@ -1,16 +0,0 @@ -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n\npage_label: 106286\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nReceived 20 August 2022, accepted 18 September 2022, date of publication 6 October 2022, date of current version 11 October 2022.\nDigital Object Identifier 10.1 109/ACCESS.2022.3210261\nAI Golf: Golf Swing Analysis Tool for Self-Training\nCHEN-CHIEH LIAO\n1, DONG-HYUN HWANG\n2, AND HIDEKI KOIKE\n1\n1Department of Computer Science, Tokyo Institute of Technology, Tokyo 152-8550, Japan\n2NAVER CLOVA Voice&Avatar, Seongnam 13561, South Korea\nCorresponding author: Chen-Chieh Liao (liao.c.aa@m.titech.ac.jp)\nThis work was supported by the Japan Science and Technology Agency (JST) Core Research for Evolutional Science and Technology\n(CREST), Japan, under Grant JPMJCR17A3.\nABSTRACT In the \ufb01eld of the acquisition of sports skills, a common way to improve sports skills, such as\ngolf swings, is to imitate professional players\u2019 motions. However, it is dif\ufb01cult for beginners to specify\nthe keyframes on which they should focus and which part of the body they should correct because of\ninconsistent timing and lack of knowledge. In this study, a golf swing analysis tool using neural networks\nis proposed to address this gap. The proposed system compares two motion sequences and speci\ufb01es\nkeyframes in which signi\ufb01cant differences can be observed between the two motions. In addition, the system\nhelps users intuitively understand the differences between themselves and professional players by using\ninterpretable clues. The main challenge of this study is to target the \ufb01ne-grained differences between users\nand professionals that can be used for self-training. Moreover, the signi\ufb01cance of the proposed approach is\nthe use of an unsupervised learning method without prior knowledge and labeled data, which will bene\ufb01t\nfuture applications and research in other sports and skill training processes. In our approach, neural networks\nare \ufb01rst used to create a motion synchronizer to align motions with different phases and timing. Next,\na motion discrepancy detector is implemented to \ufb01nd \ufb01ne-grained differences between motions in latent\nspaces that are learned by the networks. Furthermore, we consider that learning intermediate motions may\nbe feasible for beginners because, in this way, they can gradually change their pose to match the ideal form.\nTherefore, based on the synchronization and discrepancy detection results, we utilize a decoder to restore\nthe intermediate human poses between two motions from the latent space. Finally, we suggest possible\napplications for analyzing and visualizing the discrepancy between the two input motions and interacting\nwith the users. With the proposed application, users can easily understand the differences between their\nmotions and those of various experts during self-training and learn how to improve their motions.\nINDEX TERMS Computer vision, machine learning, motor skill training, golf.\nI. INTRODUCTION\nIn sports, it is dif\ufb01cult for beginners to improve their skills\nwithout prior knowledge or assistance from coaches. As a\nconventional method, people go to lessons to meet experts\nand learn how to play in the proper form. However, in most\nsports, players spend considerable time training alone to\nachieve outstanding results in the \ufb01eld and retain exceptional\nbody\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What was the motion manipulation technique used in the AI Golf swing analysis tool?\nAnswer: "}, {"role": "assistant", "content": "The motion manipulation technique used in the AI Golf swing analysis tool involves a decoder structure called a motion manipulator. This manipulator is used to restore motion from the latent space. The restoration results suggest that the network can retrieve intermediate human poses between two motions that do not exist in the original dataset. This allows beginners to learn from an intermediate pose that is more acceptable for them to start with, rather than directly imitating professional forms."}]} -{"messages": [{"role": "user", "content": "You are an expert Q&A system that strictly operates in two modes when refining existing answers:\n1. **Rewrite** an original answer using the new context.\n2. **Repeat** the original answer if the new context isn't useful.\nNever reference the original answer or context directly in your answer.\nWhen in doubt, just repeat the original answer.New Context: to restore\nthe intermediate human poses between two motions from the latent space. Finally, we suggest possible\napplications for analyzing and visualizing the discrepancy between the two input motions and interacting\nwith the users. With the proposed application, users can easily understand the differences between their\nmotions and those of various experts during self-training and learn how to improve their motions.\nINDEX TERMS Computer vision, machine learning, motor skill training, golf.\nI. INTRODUCTION\nIn sports, it is dif\ufb01cult for beginners to improve their skills\nwithout prior knowledge or assistance from coaches. As a\nconventional method, people go to lessons to meet experts\nand learn how to play in the proper form. However, in most\nsports, players spend considerable time training alone to\nachieve outstanding results in the \ufb01eld and retain exceptional\nbody conditions. Therefore, it is important to design and\nimplement an effective and accurate self-training process for\nsuch situations.\nIn the \ufb01eld of the acquisition of sports skills, one way\nto improve sports skills is to replicate professional players\u2019\nmotions. People watch the movements of professional ath-\nThe associate editor coordinating the review of this manuscript and\napproving it for publication was Jinjia Zhou\n .letes on television or the Internet and try to make their bodies\nmove similarly to professionals. To accelerate this process,\nmany systems have recently been developed to help users\nunderstand the movements of professionals [1], [2], [3], [4],\n[5]. However, in these previous works and systems, users may\nstruggle to re\ufb01ne their movement with no idea which timing\nof the whole motion, which parts of the body they should\nfocus on, or how they can change their body movements to\nget their form closer to that of professionals.\nWith signi\ufb01cant advances in machine learning technolo-\ngies, many systems have been built to recognize differ-\nent objects, make predictions for decisions, or even predict\nthe future [6].\nQuery: What was the motion manipulation technique used in the AI Golf swing analysis tool?\nOriginal Answer: The motion manipulation technique used in the AI Golf swing analysis tool involves a decoder structure called a motion manipulator. This manipulator is used to restore motion from the latent space. The restoration results suggest that the network can retrieve intermediate human poses between two motions that do not exist in the original dataset. This allows beginners to learn from an intermediate pose that is more acceptable for them to start with, rather than directly imitating professional forms.\nNew Answer: "}, {"role": "assistant", "content": "The AI Golf swing analysis tool employs a motion manipulation technique that utilizes a decoder structure known as a motion manipulator. This manipulator is designed to recover motion from a latent space. The restoration process indicates that the network is capable of retrieving intermediate human poses between two distinct motions that are not present in the original dataset. This feature enables beginners to commence their learning process from an intermediate pose that is more suitable and manageable for them, instead of attempting to mimic professional forms directly."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106291\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nframe showing a particular moment, and the phase is a time\nseries between two key events. For example, as shown in\nFigure 5, a key event in golf may be the moment when the golf\nclub hits the ball (impact), and the motion before the golf club\nhits the ball can be considered as the phase of the golf club\napproaching the ball (swinging down). Note that all the\nframes in the period between two key events have the same\nphase label. Following the key event annotation of GolfDB,\nwe labeled our pseudo dataset with eight key events: address,\ntoe-up, mid-backswing, top, mid-downswing, impact, mid-\nfollow-through, and \ufb01nish. Phases were labeled between\nevery two key events, for a total of seven phases.\nThe phase classi\ufb01cation accuracy was per frame phase\nclassi\ufb01cation. To calculate the accuracy, we \ufb01rst used the\nencoders of our TCC networks to extract latent vectors.\nWe then trained a simple classi\ufb01er on the latent vectors to\npredict the labeled phases. The classi\ufb01er was trained under\nseveral conditions by changing the percentage of the given\nlabeled data. After the classi\ufb01er was trained, we used all\nlabeled data to calculate the phase classi\ufb01cation accuracy.\nIn general, the larger the size of the given labeled data, the\nhigher the accuracy of the classi\ufb01er.\nTo explore more speci\ufb01cally the effectiveness of the net-\nwork in detecting discrepant motion, we investigated whether\nthe distance in the latent space could represent the discrep-\nancy between two input sequences by computing Pearson\u2019s\ncorrelation coef\ufb01cient. Because 3D poses varied for differ-\nent camera views, we could not directly compare the two\nhuman poses using the results from the 3d pose estimator.\nThus, we applied Procrustes analysis to align the two human\nposes to ensure that they were seen from the same camera\ndirection. In our experiment, we compared the distance in the\nlatent space with the mean per joint point error (MPJPE) and\nmeasured the Pearson\u2019s correlation coef\ufb01cient \u03c1:\n\u03c1D,E=cov(D, E)\n\u03c3D\u03c3E(6)\nwhere covis the covariance; Dis the distance in the latent\nspace; Eis the MPJPE; and \u03c3Dand\u03c3Eare the standard\ndeviations of DandE.\nV. RESULTS\nThis chapter presents the results of an early qualitative\nanalysis investigating the potential of the proposed method\nto detect discrepant motion differences, followed by more\ndetailed results comparing different modules. Finally, the\nintermediate human poses generated by the motion manip-\nulator are visualized for qualitative studies.\nA. CASE STUDY\nIn our early study, we conducted a qualitative analysis by\nexploring the latent space to investigate whether the network\ncould trace \ufb01ne-grained differences. We \ufb01rst used the V-TCC\nto synchronize the swing motions of professionals and begin-\nners. We then computed the distances between the aligned\nvideos in the latent space and visualized the overlaid 3D\nhuman poses for qualitative comparison (Figure 6).TABLE 1. Phase classification accuracy. This is the accuracy metric\nshowing the ability of the network to classify any given motion frame to\nits corresponding phase.\nB. PHASE CLASSIFICATION ACCURACY\nWe trained the V-TCC using the GolfDB video dataset with\nand without background subtraction. In contrast, S-TCC and\nSA-TCC were trained using the pseudo skeleton dataset\nwith unit vector normalization. After training the four mod-\nels, we computed the phase classi\ufb01cation accuracy for each\ntrained model by assigning 1%, 5%, 10%, 30%, and 80% of\nlabeled data.\nThe results are presented in Table 1. As we trained the\nnetwork properly, the phase classi\ufb01cation accuracy was low\nwhen the given labeled data were insuf\ufb01cient and rose with\nan increase in the number of labeled data.\nC. CORRELATION\nWe computed Pearson\u2019s correlation coef\ufb01cient using the\nfour models. For video inputs, a 0.69 Pearson\u2019s correla-\ntion coef\ufb01cient was measured for regular video, and a\n0.72 Pearson\u2019s correlation coef\ufb01cient was obtained when the\nbackground was removed (Figure 7).\n\nfile_path: src\\test\\regression\\regression_test004\\train_questions.txt\n\nWhat was the motion manipulation technique used in the AI Golf swing analysis tool?\nWhat is the phase classification accuracy metric used to evaluate the network's ability to classify motion frames to their corresponding phases?\nWhat is the Pearson's correlation coefficient for skeleton inputs using the S-TCC model?\nQuestion: What method is used in this study to align two temporal sequences?\nWhat is the purpose of the proposed system in this study?\nWhat are the three main contributions of the proposed golf swing analysis tool?\nQuestion: What is the main purpose of the proposed AI Golf Swing Analysis Tool for Self-Training?\nQuestion: What is the title of the paper mentioned in the context that applies pose estimation to predict amateur golf swing performance?\nWhat is the title of the paper mentioned in the context?\nQuestion: What is the title of the paper mentioned in the context information?\nQuestion: What is the title of the paper referenced in citation [28]?\nQuestion: Who received the B.S. degree from the School of Computer Science and Engineering, Korea University of Technology and Education, Cheonan, South Korea, in 2017?\nWhat is the correlation between the distance in the latent space and the MPJPE in the golf swing analysis tool?\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What is the phase classification accuracy metric used to evaluate the network's ability to classify motion frames to their corresponding phases?\nAnswer: "}, {"role": "assistant", "content": "The phase classification accuracy metric is used to evaluate the network's ability to correctly identify the phase of a given motion frame. This is done by first using the encoders of the Temporal Cycle-Consistency (TCC) networks to extract latent vectors. A classifier is then trained on these vectors to predict the labeled phases. The accuracy of this classifier in predicting the phases is then measured using all labeled data. The accuracy tends to increase with the size of the given labeled data."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106291\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIn contrast, S-TCC and\nSA-TCC were trained using the pseudo skeleton dataset\nwith unit vector normalization. After training the four mod-\nels, we computed the phase classi\ufb01cation accuracy for each\ntrained model by assigning 1%, 5%, 10%, 30%, and 80% of\nlabeled data.\nThe results are presented in Table 1. As we trained the\nnetwork properly, the phase classi\ufb01cation accuracy was low\nwhen the given labeled data were insuf\ufb01cient and rose with\nan increase in the number of labeled data.\nC. CORRELATION\nWe computed Pearson\u2019s correlation coef\ufb01cient using the\nfour models. For video inputs, a 0.69 Pearson\u2019s correla-\ntion coef\ufb01cient was measured for regular video, and a\n0.72 Pearson\u2019s correlation coef\ufb01cient was obtained when the\nbackground was removed (Figure 7). For skeleton inputs,\nan over 0.76 Pearson\u2019s correlation coef\ufb01cient was found\nwhen using the SA-TCC; however, the lowest Pearson\u2019s cor-\nrelation coef\ufb01cient with 0.51 was obtained from the S-TCC\n(Figure 8).\nD. MOTION INTERPOLATION\nFor the qualitative results, we computed and visualized the\nintermediate human pose between a pair of human poses\nconsidering the following three circumstances:\n\u2022The two poses were from a single person. The two poses\nwere in different phases (Figure 9(a)).\n\u2022The two poses were from different individuals. The two\nposes were in the same phase (Figure 9(b)).\n\u2022The two poses were from different individuals. The two\nposes were in different phases (Figure 9(c)).\nVI. DISCUSSION\nA. CASE STUDY\nIn our early case study, we observed that in most cases, when\nthe distance in the latent space was small, the difference\nbetween the two 3D poses was small(Figure 6c). On the\nother hand, when the distance was considerable, the 3D poses\nshowed more differences(Figure 6b, d). This suggested that\nthe network could distinguish between beginners and profes-\nsionals. Thus, we can use this feature to retrieve motion fea-\ntures that help users target the key moves they have to correct.\nHowever, as depicted in Figure 6(a), the distance in the latent\nspace remained large even in certain circumstances where the\nVOLUME 10, 2022 106291\n\npage_label: 106292\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nFIGURE 6. Case study with the V-TCC. The line graph shows the distance between two synchronized motions in the latent space. The red line in the\ngraph indicates the threshold for discrepancy detection. The colored skeleton and black skeleton indicate the user\u2019s pose and expert\u2019s pose,\nrespectively. The density and radius of red spheres indicate the degree of joint position difference between the two skeletons.\nFIGURE 7. Pearson\u2019s correlation test for V-TCC. Left: normal videos. Right:\nvideos without background.\nFIGURE 8. Pearson\u2019s correlation test with skeleton input. Left: S-TCC.\nRight: SA-TCC.\njoint difference was negligible. This might be explained by\nthe fact that the network focused not only on human motion\nbut also on other motion features. In this case, the golf club\u2019s\nmovement, which was also critical during the swing, might\nbe the main factor causing the enormous distance in the latent\nspace. This led us to the following quantitative studies, where\nwe conducted statistical tests to examine the accuracy of the\nsynchronization and the correlation between the latent space\nand joint difference under various conditions.\nB. PHASE CLASSIFICATION ACCURACY\nIn this quantitative study, we discovered that the previous\nTCC implementation had limited precision in terms of syn-\nchronization. As our hypothesis suggests, synchronization\nquality might be in\ufb02uenced by various background informa-\ntion and other movements, such as the golf club\u2019s and human\nshadow\u2019s motions. When the background was removed, a sig-\nni\ufb01cant increase in the accuracy of the V-TCC model was\nobserved. This result supported our hypothesis that infor-\nmation outside the human region would affect alignment.Various types of information from the background, the\nmovement of human shadows, and the motion of properties\ncontrolled by a human might be learned as features by the net-\nwork, thus affecting the accuracy of the analysis. Because of\nthe signi\ufb01cant superiority of the without-background version,\nwe considered it necessary to apply background subtraction\nbefore synchronization when building applications for real\nusage.\nNext, considering the skeleton version, both S-TCC and\nSA-TCC outperformed V-TCC under all conditions. Given\nonly 10% of the labeled data, both S-TCC and SA-TCC\nachieved a phase classi\ufb01cation accuracy of over 90%. This\nmight be explained by the fact that the skeleton version\nhad more compressed and precise information than the\nvideo version, where the color information might be noisy.\nAdditionally, 3D poses represented high-level human fea-\ntures. Therefore, the features outside the human region were\nremoved when retrieving 3D human poses, resulting in more\nprecise accuracy for further golf swing analysis. Note that the\nskeleton data were the estimated results from the video data,\nand the estimation could not be perfect for every human pose.\nDespite imperfect skeleton inputs, the S-TCC and SA-TCC\nnetworks outperformed V-TCC. This result suggested that we\ncould use the skeleton version for more precise implementa-\ntion in actual usage.\nFinally, comparing the two skeleton version mod-\nels, S-TCC performed better than SA-TCC under most\nconditions. However, SA-TCC remained competitive in some\ncircumstances (better than S-TCC with 10% labeled data).\nThis led to the following correlation test, in which we dis-\ncussed the ability of the networks to detect discrepancies\nbetween two motions.\nC. CORRELATION\nDiscussing the correlation test, as shown in Figure 7,\nwe observed a correlation between the distance in the latent\nspace and the MPJPE. While the distribution of the data pairs\nseemed to be scattered, a Pearson\u2019s correlation coef\ufb01cient of\nover 0.69 showed the potential of the network to distinguish\nwhether the difference between the two motions was small\nor large. However, as discussed in the case study section,\nin some circumstances, the network failed to detect discrepant\n106292 VOLUME 10, 2022\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What is the Pearson's correlation coefficient for skeleton inputs using the S-TCC model?\nAnswer: "}, {"role": "assistant", "content": "The Pearson's correlation coefficient for skeleton inputs using the S-TCC model is 0.51."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106288\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nindividuals, we must align the motions to make them com-\nparable. A conventional method for aligning two temporal\nsequences is dynamic time warping (DTW), which was intro-\nduced by Berndt and Clifford [18]. In this method, every\nindex in a sequence is matched with one or more indices from\nanother sequence, and the mapping of the indices from the\n\ufb01rst sequence to the other sequences must be monotonically\nincreasing. The DTW concept has been introduced in several\ndomains. For example, Ikeda et al. [19] proposed a real-time\ngolf swing projecting system that simultaneously visualized\nprofessional and user forms with matched timing. To align\nthe two motions in real-time, they measured the DTW only\nover a short period and penalized the previous cost value at\na later time. Halperin et al. [20] utilized the concept of DTW\nin speech and presented an audio-to-video alignment method\nfor matching speech-to-lip movements.\nOn the other hand, self-supervised neural networks have\nrecently been developed to tackle video alignment tasks\nusing the latent space representation [12], [21], [22]. In this\napproach, an embedder is used to compress input videos\ninto a latent space. After the embedding process, a loss is\ndesigned to \ufb01nd correspondences through time in the latent\nspace, thus encouraging the network to learn a latent space\nwhere similar motions should appear to be close. This method\nhelps synchronize high complexity temporal sequences, such\nas videos. In this study, we base the loss of our network on the\ntemporal cycle-consistency loss used by Dwibedi et al. [12],\nbut apply DTW along with the loss for smooth temporal\nmatching.\nC. DISCREPANCY DETECTION\nBy comparing the two synchronized motions, we can deter-\nmine the difference between them in terms of human pos-\ntures. In early studies, abnormal detection referred mainly\nto \ufb01nding patterns in data that did not match the expected\nbehavior [25]. Recently, two methods have been proposed for\ndetecting abnormalities In the \ufb01rst approach, abnormal infor-\nmation is referred to as prior knowledge. For example, Parra-\nDominguez et al. [26] trained a binary classi\ufb01er on annotated\ndata to determine whether abnormal events occurred during\na stair descent. In the second approach to abnormal detec-\ntion, abnormal information is not provided in advance. The\nresearch group of Nater et al. [27] proposed an unsupervised\nlearning method for learning normal human behavior. They\nused a hierarchical representation of the appearance and\naction level of regular movements to detect abnormal events.\nWhile a network may be trained to detect abnormal events,\nsuch as falling to the ground, we focus on whether a neural\nnetwork can be trained to automatically detect \ufb01ne-grained\ndifferences between two regular motions. We call this detec-\ntion of the \ufb01ne-grained difference discrepancy detection. The\nmost relevant of these studies is that of Abati et al. [28]. They\ndesigned a deep autoencoder with a parametric estimator\nthat learned a probability distribution from the latent space\nto detect discrepancies. The encoder effectively remembered\nstandard samples and could distinguish between normal and\nFIGURE 1. System overview. Xis the input motion sequence and Yis the\noutput human poses restored from the latent space.\nabnormal images. However, although the network could eas-\nily detect surprise samples, \ufb01ne-grained differences among\nstandard samples were not discussed. In addition to image-\nbased methods, recent studies have focused on systems that\nuse 3D human pose information [29], [30]. In this study,\nwe apply discrepancy detection to both videos and 3D\nhuman poses and discuss the ability of the system to detect\n\ufb01ne-grained differences between two input motions.\nIII. METHODS\nThis study aims to create a system that captures user motions\nand provides \ufb01ne-grained feedback to improve users\u2019 forms\nby comparing their motions with those of professionals.\nTo achieve the goal of building such an application, the\nmethod proposed in this study is to \ufb01rst train a neural network\nwith professionals\u2019 motion data.\n\npage_label: 106295\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nInt.\nConf. Artif. Reality Telexistence Eurographics Symp. Virtual Environ.,\n2018, pp. 171\u2013174, doi: 10.2312/egve.20181330.\n[20] T. Halperin, A. Ephrat, and S. Peleg, \u2018\u2018Dynamic temporal\nalignment of speech to lips,\u2019\u2019 in Proc. IEEE Int. Conf. Acoust.,\nSpeech Signal Process. (ICASSP), May 2019, pp. 3980\u20133984, doi:\n10.1109/ICASSP.2019.8682863.\n[21] I. Misra, C. L. Zitnick, and M. Hebert, \u2018\u2018Shuf\ufb02e and learn: Unsupervised\nlearning using temporal order veri\ufb01cation,\u2019\u2019 in Computer Vision ECCV\n2016 (Lecture Notes in Computer Science), vol. 9905, B. Leibe, J. Matas,\nN. Sebe, and M. Welling, Eds. Cham, Switzerland: Springer, 2016, doi:\n10.1007/978-3-319-46448-0_32.\n[22] P. Sermanet, C. Lynch, Y. Chebotar, J. Hsu, E. Jang, S. Schaal, S. Levine,\nand G. Brain, \u2018\u2018Time-contrastive networks: Self-supervised learning from\nvideo,\u2019\u2019 in Proc. IEEE Int. Conf. Robot. Autom. (ICRA), May 2018,\npp. 1134\u20131141, doi: 10.1109/ICRA.2018.8462891.\n[23] D. Dwibedi, Y. Aytar, J. Tompson, P. Sermanet, and A. Zisserman,\n\u2018\u2018Temporal cycle-consistency learning,\u2019\u2019 in Proc. IEEE/CVF Conf. Com-\nput. Vis. Pattern Recognit. (CVPR), Jun. 2019, pp. 1801\u20131810, doi:\n10.1109/CVPR.2019.00190.\n[24] S. Haresh, S. Kumar, H. Coskun, S. N. Syed, A. Konin, M. Z. Zia, and\nQ.-H. Tran, \u2018\u2018Learning by aligning videos in time,\u2019\u2019 in Proc. IEEE/CVF\nConf. Comput. Vis. Pattern Recognit. (CVPR), Jun. 2021, pp. 5544\u20135554,\ndoi:10.1109/CVPR46437.2021.00550.\n[25] H. Seki and Y. Hori, \u2018\u2018Detection of abnormal action using image sequence\nfor monitoring system of aged people,\u2019\u2019 IEEJ Trans. Ind. Appl., vol. 122,\nno. 2, pp. 182\u2013188, 2002, doi: 10.1541/ieejias.122.182.\n[26] G. S. Parra-Dominguez, B. Taati, and A. Mihailidis, \u2018\u20183D human motion\nanalysis to detect abnormal events on stairs,\u2019\u2019 in Proc. 2nd Int. Conf. 3D\nImag., Modeling, Process., Vis. Transmiss., Oct.\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What method is used in this study to align two temporal sequences?\nAnswer: "}, {"role": "assistant", "content": "The method used in this study to align two temporal sequences is dynamic time warping (DTW)."}]} -{"messages": [{"role": "user", "content": "You are an expert Q&A system that strictly operates in two modes when refining existing answers:\n1. **Rewrite** an original answer using the new context.\n2. **Repeat** the original answer if the new context isn't useful.\nNever reference the original answer or context directly in your answer.\nWhen in doubt, just repeat the original answer.New Context: 5544\u20135554,\ndoi:10.1109/CVPR46437.2021.00550.\n[25] H. Seki and Y. Hori, \u2018\u2018Detection of abnormal action using image sequence\nfor monitoring system of aged people,\u2019\u2019 IEEJ Trans. Ind. Appl., vol. 122,\nno. 2, pp. 182\u2013188, 2002, doi: 10.1541/ieejias.122.182.\n[26] G. S. Parra-Dominguez, B. Taati, and A. Mihailidis, \u2018\u20183D human motion\nanalysis to detect abnormal events on stairs,\u2019\u2019 in Proc. 2nd Int. Conf. 3D\nImag., Modeling, Process., Vis. Transmiss., Oct. 2012, pp. 97\u2013103, doi:\n10.1109/3DIMPVT.2012.34.\n[27] F. Nater, H. Grabner, and L. Van Gool, \u2018\u2018Exploiting simple hierarchies\nfor unsupervised human behavior analysis,\u2019\u2019 in Proc. IEEE Comput. Soc.\nConf. Comput. Vis. Pattern Recognit. , Jun. 2010, pp. 2014\u20132021, doi:\n10.1109/CVPR.2010.5539877.\n[28] D. Abati, A. Porrello, S. Calderara, and R. Cucchiara, \u2018\u2018Latent space\nautoregression for novelty detection,\u2019\u2019 in Proc. IEEE/CVF Conf. Com-\nput. Vis. Pattern Recognit. (CVPR), Jun. 2019, pp. 481\u2013490, doi:\n10.1109/CVPR.2019.00057.\nQuery: Question: What method is used in this study to align two temporal sequences?\nOriginal Answer: The method used in this study to align two temporal sequences is dynamic time warping (DTW).\nNew Answer: "}, {"role": "assistant", "content": "The method used in this study to align two temporal sequences is dynamic time warping (DTW)."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\nfile_path: src\\test\\regression\\regression_test004\\eval_questions.txt\n\nWhat is the publication date of the article mentioned in reference [9]?\nQuestion: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool?\nQuestion: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool?\nWhat is the purpose of the motion decoder in the AI Golf swing analysis tool?\nWhat method was used to detect the human body in the video dataset?\nWhat is the main challenge addressed in the proposed golf swing analysis tool?\nWhat is the purpose of the AI Golf Swing Analysis Tool mentioned in the context?\nWhat are the main contributions of the proposed golf swing analysis method?\nWhat is one of the main contributions of the study on golf swing analysis?\nWhat is the file type of the file \"eval_questions.txt\"?\n\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What is the purpose of the proposed system in this study?\nAnswer: "}, {"role": "assistant", "content": "The purpose of the proposed system in this study is to provide a golf swing analysis tool that uses neural networks to help users intuitively understand the difference between their own golf swings and those of professional players. The system aims to help users efficiently learn an ideal form in a gradual manner instead of directly imitating ideal motion. It allows users to select an ideal form to imitate and learn to play sports without the help of coaches during self-training. The system also plans to analyze and visualize the discrepancy between two input golf swing motions, helping users quickly grasp the difference between their swings and those of various experts."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\n\npage_label: 106287\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nCompared to existing methods,\nthe main contributions of this study can be summarized as\nfollows:\n\u2022A golf swing analysis method and its applications are\nintroduced.\n\u2022The proposed method distinguishes when the difference\nbetween two motions is large and small.\n\u2022The proposed method helps users understand the differ-\nence between themselves and professional players.\n\u2022The proposed method provides intermediate poses that\nare acceptable during the early learning phase of sports.\n\u2022Crucial factors that can in\ufb02uence the accuracy of sports\nanalysis are discussed.\nII. RELATED WORK\nA. SPORTS TRAINING FOR SKILL ACQUISITION\nMany recent studies have focused on developing sports\ntraining systems to help beginners improve their skills. Stud-\nies [13], [14], [15] have proposed multi-modal sports training\nsystems based on sports theories. In their system, users\nreceived visual, haptic, and audio feedback when they did not\nideally move their bodies or instruments. However, the ideal\nmovement could differ from sport to sport, meaning that it\nmay be dif\ufb01cult to generalize these methodologies to other\nsports.\nOn the other hand, another way to learn sports skills is\nimitating professional players\u2019 motions [1], [3], [4], [5].\nIkeda et al. [16] proposed a golf swing training system that\nuses the motions of professional golfers. In their system,\na user\u2019s motion was synchronized with a selected ideal\nprofessional\u2019s motion, and the two motions were overlaid and\nprojected onto the ground during training. Sasaki et al. [17]\nalso reported the importance of beginners copying expert\nmotions and proposed a climbing training system using pose\nprediction. Their system predicted and visualized the pose of\nexperts based on the user\u2019s current hand and foot positions.\nWhile these recent studies have shown the effectiveness of\nusing multi-modal feedback and the potential of applying\nneural networks to create AI teachers for sports training, it is\ndif\ufb01cult for users to change their motion forms immediately\nto match ideal forms. Thus, building a system that can teach\nusers step-by-step to improve their sports skills remains\nchallenging.\nB. VIDEO AND MOTION ALIGNMENT\nAn ef\ufb01cient way to evaluate whether a person is performing\na motion correctly is to compare their motion with oth-\ners whose motion is considered correct. However, owing\nto the various timings and speeds of motion of different\nVOLUME 10, 2022 106287\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What are the three main contributions of the proposed golf swing analysis tool?\nAnswer: "}, {"role": "assistant", "content": "The three main contributions of the proposed golf swing analysis tool are: \n\n1. The development of a motion synchronizer that aligns motions with different phases and timings, which has shown better performance than existing video implementations. \n\n2. The creation of a motion discrepancy detector that identifies fine-grained differences between golf swings in the latent space, using the proposed networks. This detector can distinguish whether the difference between two motions is small or large. \n\n3. The introduction of a decoder structure called a motion manipulator that can restore motion from the latent space. This allows for the retrieval of intermediate human poses between two motions that do not exist in the original dataset, providing a more suitable starting point for beginners."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106286\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nReceived 20 August 2022, accepted 18 September 2022, date of publication 6 October 2022, date of current version 11 October 2022.\nDigital Object Identifier 10.1 109/ACCESS.2022.3210261\nAI Golf: Golf Swing Analysis Tool for Self-Training\nCHEN-CHIEH LIAO\n1, DONG-HYUN HWANG\n2, AND HIDEKI KOIKE\n1\n1Department of Computer Science, Tokyo Institute of Technology, Tokyo 152-8550, Japan\n2NAVER CLOVA Voice&Avatar, Seongnam 13561, South Korea\nCorresponding author: Chen-Chieh Liao (liao.c.aa@m.titech.ac.jp)\nThis work was supported by the Japan Science and Technology Agency (JST) Core Research for Evolutional Science and Technology\n(CREST), Japan, under Grant JPMJCR17A3.\nABSTRACT In the \ufb01eld of the acquisition of sports skills, a common way to improve sports skills, such as\ngolf swings, is to imitate professional players\u2019 motions. However, it is dif\ufb01cult for beginners to specify\nthe keyframes on which they should focus and which part of the body they should correct because of\ninconsistent timing and lack of knowledge. In this study, a golf swing analysis tool using neural networks\nis proposed to address this gap. The proposed system compares two motion sequences and speci\ufb01es\nkeyframes in which signi\ufb01cant differences can be observed between the two motions. In addition, the system\nhelps users intuitively understand the differences between themselves and professional players by using\ninterpretable clues. The main challenge of this study is to target the \ufb01ne-grained differences between users\nand professionals that can be used for self-training. Moreover, the signi\ufb01cance of the proposed approach is\nthe use of an unsupervised learning method without prior knowledge and labeled data, which will bene\ufb01t\nfuture applications and research in other sports and skill training processes. In our approach, neural networks\nare \ufb01rst used to create a motion synchronizer to align motions with different phases and timing. Next,\na motion discrepancy detector is implemented to \ufb01nd \ufb01ne-grained differences between motions in latent\nspaces that are learned by the networks. Furthermore, we consider that learning intermediate motions may\nbe feasible for beginners because, in this way, they can gradually change their pose to match the ideal form.\nTherefore, based on the synchronization and discrepancy detection results, we utilize a decoder to restore\nthe intermediate human poses between two motions from the latent space. Finally, we suggest possible\napplications for analyzing and visualizing the discrepancy between the two input motions and interacting\nwith the users. With the proposed application, users can easily understand the differences between their\nmotions and those of various experts during self-training and learn how to improve their motions.\nINDEX TERMS Computer vision, machine learning, motor skill training, golf.\nI. INTRODUCTION\nIn sports, it is dif\ufb01cult for beginners to improve their skills\nwithout prior knowledge or assistance from coaches. As a\nconventional method, people go to lessons to meet experts\nand learn how to play in the proper form. However, in most\nsports, players spend considerable time training alone to\nachieve outstanding results in the \ufb01eld and retain exceptional\nbody conditions. Therefore, it is important to design and\nimplement an effective and accurate self-training process for\nsuch situations.\nIn the \ufb01eld of the acquisition of sports skills, one way\nto improve sports skills is to replicate professional players\u2019\nmotions. People watch the movements of professional ath-\nThe associate editor coordinating the review of this manuscript and\napproving it for publication was Jinjia Zhou\n .letes on television or the Internet and try to make their bodies\nmove similarly to professionals. To accelerate this process,\nmany systems have recently been developed to help users\nunderstand the movements of professionals [1], [2], [3], [4],\n[5]. However, in these previous works and systems, users may\nstruggle to re\ufb01ne their movement with no idea which timing\nof the whole motion, which parts of the body they should\nfocus on, or how they can change their body movements to\nget their form closer to that of professionals.\nWith signi\ufb01cant advances in machine learning technolo-\ngies, many systems have been built to recognize differ-\nent objects, make predictions for decisions, or even predict\nthe future [6].\n\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nuser interface where users can select any professional\u2019s form\nfrom the database and compare the difference between their\nforms and the professional\u2019s. In addition, instead of directly\nimitating the selected motion, users can gradually imitate\nintermediate human poses using a motion manipulator. Fur-\nthermore, we visualized the attention maps along with the\nself-attention module to demonstrate the importance of each\nbody part (Figure 10). This may be a clue for discovering\nbody parts that should be focused on and what can be ignored.\nTo go a step further, this might indicate the importance of the\nbody parts that trainees should follow to revise their poses\noptimally. In the future, we plan to conduct user studies to\nevaluate the training effectiveness of the proposed system\nand determine whether the crucial points shown by the atten-\ntion maps can be used in real training scenarios. Moreover,\nwe consider that using a high-quality motion capture sys-\ntem can enhance the solidity of the proposed system with\nhigh-precision human poses despite the high speed of motion,\nsuch as golf swings.\nAs this work focuses on proposing a novel \ufb02ow for con-\nstructing a sports analysis tool, and because user studies may\nvary from domain to domain and are \ufb02exible for the system\ndesigner, further user studies evaluating the proposed sys-\ntem\u2019s ef\ufb01ciency have not been addressed in practice. As pre-\nviously mentioned, there are many ways to design a proper\nway to provide feedback to users for training. Related works\nthat use multi-modal feedback to alert users when performing\nthe wrong way compared to professionals have shown their\nsigni\ufb01cance during training. In the future, we plan to combine\nour approach with other feedback systems and evaluate the\neffectiveness of the system. Furthermore, we plan to apply the\nproposed approach to other sports and skill training processes\nto explore the generality of the proposed method.\nVIII. CONCLUSION\nWe propose a golf swing analysis tool that uses neural net-\nworks to help users intuitively understand the difference\nbetween themselves and professional players. We divide our\nwork into three parts: synchronization, discrepancy detec-\ntion, and manipulation. First, the motion synchronizer aligns\nmotions with different phases and timings. The experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the main purpose of the proposed AI Golf Swing Analysis Tool for Self-Training?\nAnswer: "}, {"role": "assistant", "content": "The main purpose of the proposed AI Golf Swing Analysis Tool for Self-Training is to help users intuitively understand the differences between their golf swing and that of professional players. The system uses neural networks to analyze golf swings, identify keyframes where significant differences can be observed, and provide interpretable clues to users. It aims to target the fine-grained differences between users and professionals that can be used for self-training. The system also includes a user interface where users can select any professional's form from a database, compare it with their own, and gradually imitate intermediate human poses. The tool is designed to assist users in improving their golf swings during self-training."}]} -{"messages": [{"role": "user", "content": "You are an expert Q&A system that strictly operates in two modes when refining existing answers:\n1. **Rewrite** an original answer using the new context.\n2. **Repeat** the original answer if the new context isn't useful.\nNever reference the original answer or context directly in your answer.\nWhen in doubt, just repeat the original answer.New Context: experi-\nment shows that our implementation using skeleton inputs\ncan achieve a better performance than state-of-the-art video\nimplementations.\nSecond, using the proposed networks, we use a motion\ndiscrepancy detector to \ufb01nd \ufb01ne-grained differences between\ngolf swings in the latent space. By applying comparative anal-\nysis, such as comparing 3D human poses in those detected\nframes, we conclude that the motion discrepancy detector can\ndistinguish whether the difference between the two motions is\nsmall or large. The proposed SA-TCC network outperforms\nthe previous TCC network in terms of phase classi\ufb01cation\naccuracy and has the best ability to show the correlation\nbetween the distance in the latent space and the MPJPE.\nThird, based on the synchronization and discrepancy detec-\ntion results, we introduce a decoder structure called a motionmanipulator to restore motion from the latent space. The\nrestoration results suggest that the network can retrieve inter-\nmediate human poses between two motions that do not\nexist in the original dataset. Furthermore, instead of teach-\ning beginners directly about professional forms, the motion\nmanipulator can provide them with an intermediate pose that\nis more acceptable for beginners to start with.\nFinally, with the above three main contributions of this\nwork, we create an application for analyzing and visualiz-\ning the discrepancy between two input golf swing motions.\nFor user interaction, users can quickly grasp the difference\nbetween their swings and those of various experts during\nself-training. In addition, by understanding the continuous\nchanges step-by-step between the two selected human poses,\nwe aim to help users ef\ufb01ciently learn an ideal form in a\ngradual manner instead of directly imitating ideal motion.\nUsing the proposed system, users can choose an ideal form\nto imitate and learn to play sports without the help of coaches\nduring self-training. We intend to re\ufb01ne the proposed pro-\ntotype application and conduct user studies to investigate its\neffectiveness.\nQuery: Question: What is the main purpose of the proposed AI Golf Swing Analysis Tool for Self-Training?\nOriginal Answer: The main purpose of the proposed AI Golf Swing Analysis Tool for Self-Training is to help users intuitively understand the differences between their golf swing and that of professional players. The system uses neural networks to analyze golf swings, identify keyframes where significant differences can be observed, and provide interpretable clues to users. It aims to target the fine-grained differences between users and professionals that can be used for self-training. The system also includes a user interface where users can select any professional's form from a database, compare it with their own, and gradually imitate intermediate human poses. The tool is designed to assist users in improving their golf swings during self-training.\nNew Answer: "}, {"role": "assistant", "content": "The primary goal of the proposed AI Golf Swing Analysis Tool for Self-Training is to enable users to effectively comprehend the discrepancies between their golf swings and those of expert players. The tool utilizes advanced neural networks to scrutinize golf swings, pinpoint keyframes with notable differences, and offer comprehensible guidance to users. It is designed to highlight the subtle variations between amateur and professional swings, which can be beneficial for self-training. The system also incorporates a user-friendly interface where individuals can choose any professional's form from a repository, compare it with their own, and progressively mimic intermediate human poses. The tool is intended to aid users in enhancing their golf swings during self-training sessions."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106294\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\n37, no. 6, pp. 1389\u20131401,\nJun. 2019, doi: 10.1109/JSAC.2019.2904363.\n[7] E. Wu and H. Koike, \u2018\u2018FuturePose\u2014Mixed reality martial arts training\nusing real-time 3D human pose forecasting with a RGB camera,\u2019\u2019 in Proc.\nIEEE Winter Conf. Appl. Comput. Vis. (WACV), Jan. 2019, pp. 1384\u20131392,\ndoi:10.1109/WACV.2019.00152.\n[8] K. Shiro, K. Egawa, T. Miyaki, and J. Rekimoto, \u2018\u2018InterPoser: Visual-\nizing interpolated movements for bouldering training,\u2019\u2019 in Proc. IEEE\nConf. Virtual Reality 3D User Inter. (VR), Mar. 2019, pp. 1563\u20131565, doi:\n10.1109/VR.2019.8798366.\n[9] S. Qiu, H. Zhao, N. Jiang, Z. Wang, L. Liu, Y. An, H. Zhao, X. Miao,\nR. Liu, and G. Fortino, \u2018\u2018Multi-sensor information fusion based on machine\nlearning for real applications in human activity recognition: State-of-the-\nart and research challenges,\u2019\u2019 Inf. Fusion, vol. 80, pp. 241\u2013265, Apr. 2022,\ndoi:10.1016/j.inffus.2021.11.006.\n[10] T. T. Kim, M. A. Zohdy, and M. P. Barker, \u2018\u2018Applying pose\nestimation to predict amateur golf swing performance using edge\nprocessing,\u2019\u2019 IEEE Access, vol. 8, pp. 143769\u2013143776, 2020, doi:\n10.1109/ACCESS.2020.3014186.\n[11] K.-R. Ko and S. B. Pan, \u2018\u2018CNN and bi-LSTM based 3D golf swing analysis\nby frontal swing sequence images,\u2019\u2019 Multimedia Tools Appl., vol. 80, no. 6,\npp. 8957\u20138972, Mar. 2021, doi: 10.1007/s11042-020-10096-0.\n106294 VOLUME 10, 2022\n\npage_label: 106287\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nCompared to existing methods,\nthe main contributions of this study can be summarized as\nfollows:\n\u2022A golf swing analysis method and its applications are\nintroduced.\n\u2022The proposed method distinguishes when the difference\nbetween two motions is large and small.\n\u2022The proposed method helps users understand the differ-\nence between themselves and professional players.\n\u2022The proposed method provides intermediate poses that\nare acceptable during the early learning phase of sports.\n\u2022Crucial factors that can in\ufb02uence the accuracy of sports\nanalysis are discussed.\nII. RELATED WORK\nA. SPORTS TRAINING FOR SKILL ACQUISITION\nMany recent studies have focused on developing sports\ntraining systems to help beginners improve their skills. Stud-\nies [13], [14], [15] have proposed multi-modal sports training\nsystems based on sports theories. In their system, users\nreceived visual, haptic, and audio feedback when they did not\nideally move their bodies or instruments. However, the ideal\nmovement could differ from sport to sport, meaning that it\nmay be dif\ufb01cult to generalize these methodologies to other\nsports.\nOn the other hand, another way to learn sports skills is\nimitating professional players\u2019 motions [1], [3], [4], [5].\nIkeda et al. [16] proposed a golf swing training system that\nuses the motions of professional golfers. In their system,\na user\u2019s motion was synchronized with a selected ideal\nprofessional\u2019s motion, and the two motions were overlaid and\nprojected onto the ground during training. Sasaki et al. [17]\nalso reported the importance of beginners copying expert\nmotions and proposed a climbing training system using pose\nprediction. Their system predicted and visualized the pose of\nexperts based on the user\u2019s current hand and foot positions.\nWhile these recent studies have shown the effectiveness of\nusing multi-modal feedback and the potential of applying\nneural networks to create AI teachers for sports training, it is\ndif\ufb01cult for users to change their motion forms immediately\nto match ideal forms. Thus, building a system that can teach\nusers step-by-step to improve their sports skills remains\nchallenging.\nB. VIDEO AND MOTION ALIGNMENT\nAn ef\ufb01cient way to evaluate whether a person is performing\na motion correctly is to compare their motion with oth-\ners whose motion is considered correct. However, owing\nto the various timings and speeds of motion of different\nVOLUME 10, 2022 106287\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the title of the paper mentioned in the context that applies pose estimation to predict amateur golf swing performance?\nAnswer: "}, {"role": "assistant", "content": "The title of the paper that applies pose estimation to predict amateur golf swing performance is \"Applying pose estimation to predict amateur golf swing performance using edge processing\"."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106295\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIEEE/CVF Conf.\nComput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 5686\u20135696, doi:\n10.1109/CVPR.2019.00584.\n[36] J. Martinez, R. Hossain, J. Romero, and J. J. Little, \u2018\u2018A simple yet effective\nbaseline for 3D human pose estimation,\u2019\u2019 in Proc. IEEE Int. Conf. Comput.\nVis. (ICCV), Oct. 2017, pp. 2659\u20132668, doi: 10.1109/ICCV.2017.288.\nCHEN-CHIEH LIAO received the B.E. and M.S.\ndegrees from the Tokyo Institute of Technol-\nogy, Japan, in 2020 and 2022, respectively,\nwhere he is currently pursuing the Ph.D. degree\nwith the School of Computing. His research\ninterests include human\u2013computer interaction,\nmixed/augmented reality, and machine vision.\nDONG-HYUN HWANG received the B.S. degree\nfrom the School of Computer Science and Engi-\nneering, Korea University of Technology and\nEducation, Cheonan, South Korea, in 2017, and\nthe M.S. and Ph.D. degrees from the School\nof Computing, Tokyo Institute of Technology,\nTokyo, Japan, in 2019 and 2022, respectively.\nHe was a Young Research Fellow at the Japan\nSociety for the Promotion of Science (JSPS),\nfrom 2020 to 2022. Currently, he is a machine\nlearning researcher at CLOVA Voice&Avatar, NAVER Corporation. His\nresearch interests include mixed/augmented reality, computer vision, and\nmachine learning.\nHIDEKI KOIKE received the B.E., M.E., and\nDr.Eng. degrees from the University of Tokyo,\nin 1986, 1988, and 1991, respectively. He was\nat the University of Electro-Communications,\nTokyo. In 2014, he joined Tokyo Institute of\nTechnology, Japan, where he is a Professor\nwith the School of Computing. His research\ninterests include vision-based human\u2013computer\ninteraction, human augmentation, information\nvisualization, and usable security.\nVOLUME 10, 2022 106295\n\nfile_path: src\\test\\regression\\regression_test004\\eval_questions.txt\n\nWhat is the publication date of the article mentioned in reference [9]?\nQuestion: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool?\nQuestion: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool?\nWhat is the purpose of the motion decoder in the AI Golf swing analysis tool?\nWhat method was used to detect the human body in the video dataset?\nWhat is the main challenge addressed in the proposed golf swing analysis tool?\nWhat is the purpose of the AI Golf Swing Analysis Tool mentioned in the context?\nWhat are the main contributions of the proposed golf swing analysis method?\nWhat is one of the main contributions of the study on golf swing analysis?\nWhat is the file type of the file \"eval_questions.txt\"?\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What is the title of the paper mentioned in the context?\nAnswer: "}, {"role": "assistant", "content": "The context does not provide the title of a specific paper."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\nfile_path: src\\test\\regression\\regression_test004\\eval_questions.txt\n\nWhat is the publication date of the article mentioned in reference [9]?\nQuestion: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool?\nQuestion: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool?\nWhat is the purpose of the motion decoder in the AI Golf swing analysis tool?\nWhat method was used to detect the human body in the video dataset?\nWhat is the main challenge addressed in the proposed golf swing analysis tool?\nWhat is the purpose of the AI Golf Swing Analysis Tool mentioned in the context?\nWhat are the main contributions of the proposed golf swing analysis method?\nWhat is one of the main contributions of the study on golf swing analysis?\nWhat is the file type of the file \"eval_questions.txt\"?\n\npage_label: 106295\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIEEE/CVF Conf.\nComput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 5686\u20135696, doi:\n10.1109/CVPR.2019.00584.\n[36] J. Martinez, R. Hossain, J. Romero, and J. J. Little, \u2018\u2018A simple yet effective\nbaseline for 3D human pose estimation,\u2019\u2019 in Proc. IEEE Int. Conf. Comput.\nVis. (ICCV), Oct. 2017, pp. 2659\u20132668, doi: 10.1109/ICCV.2017.288.\nCHEN-CHIEH LIAO received the B.E. and M.S.\ndegrees from the Tokyo Institute of Technol-\nogy, Japan, in 2020 and 2022, respectively,\nwhere he is currently pursuing the Ph.D. degree\nwith the School of Computing. His research\ninterests include human\u2013computer interaction,\nmixed/augmented reality, and machine vision.\nDONG-HYUN HWANG received the B.S. degree\nfrom the School of Computer Science and Engi-\nneering, Korea University of Technology and\nEducation, Cheonan, South Korea, in 2017, and\nthe M.S. and Ph.D. degrees from the School\nof Computing, Tokyo Institute of Technology,\nTokyo, Japan, in 2019 and 2022, respectively.\nHe was a Young Research Fellow at the Japan\nSociety for the Promotion of Science (JSPS),\nfrom 2020 to 2022. Currently, he is a machine\nlearning researcher at CLOVA Voice&Avatar, NAVER Corporation. His\nresearch interests include mixed/augmented reality, computer vision, and\nmachine learning.\nHIDEKI KOIKE received the B.E., M.E., and\nDr.Eng. degrees from the University of Tokyo,\nin 1986, 1988, and 1991, respectively. He was\nat the University of Electro-Communications,\nTokyo. In 2014, he joined Tokyo Institute of\nTechnology, Japan, where he is a Professor\nwith the School of Computing. His research\ninterests include vision-based human\u2013computer\ninteraction, human augmentation, information\nvisualization, and usable security.\nVOLUME 10, 2022 106295\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the title of the paper mentioned in the context information?\nAnswer: "}, {"role": "assistant", "content": "The context does not provide the title of any paper."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\nfile_path: src\\test\\regression\\regression_test004\\eval_questions.txt\n\nWhat is the publication date of the article mentioned in reference [9]?\nQuestion: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool?\nQuestion: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool?\nWhat is the purpose of the motion decoder in the AI Golf swing analysis tool?\nWhat method was used to detect the human body in the video dataset?\nWhat is the main challenge addressed in the proposed golf swing analysis tool?\nWhat is the purpose of the AI Golf Swing Analysis Tool mentioned in the context?\nWhat are the main contributions of the proposed golf swing analysis method?\nWhat is one of the main contributions of the study on golf swing analysis?\nWhat is the file type of the file \"eval_questions.txt\"?\n\npage_label: 106295\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIEEE/CVF Conf.\nComput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 5686\u20135696, doi:\n10.1109/CVPR.2019.00584.\n[36] J. Martinez, R. Hossain, J. Romero, and J. J. Little, \u2018\u2018A simple yet effective\nbaseline for 3D human pose estimation,\u2019\u2019 in Proc. IEEE Int. Conf. Comput.\nVis. (ICCV), Oct. 2017, pp. 2659\u20132668, doi: 10.1109/ICCV.2017.288.\nCHEN-CHIEH LIAO received the B.E. and M.S.\ndegrees from the Tokyo Institute of Technol-\nogy, Japan, in 2020 and 2022, respectively,\nwhere he is currently pursuing the Ph.D. degree\nwith the School of Computing. His research\ninterests include human\u2013computer interaction,\nmixed/augmented reality, and machine vision.\nDONG-HYUN HWANG received the B.S. degree\nfrom the School of Computer Science and Engi-\nneering, Korea University of Technology and\nEducation, Cheonan, South Korea, in 2017, and\nthe M.S. and Ph.D. degrees from the School\nof Computing, Tokyo Institute of Technology,\nTokyo, Japan, in 2019 and 2022, respectively.\nHe was a Young Research Fellow at the Japan\nSociety for the Promotion of Science (JSPS),\nfrom 2020 to 2022. Currently, he is a machine\nlearning researcher at CLOVA Voice&Avatar, NAVER Corporation. His\nresearch interests include mixed/augmented reality, computer vision, and\nmachine learning.\nHIDEKI KOIKE received the B.E., M.E., and\nDr.Eng. degrees from the University of Tokyo,\nin 1986, 1988, and 1991, respectively. He was\nat the University of Electro-Communications,\nTokyo. In 2014, he joined Tokyo Institute of\nTechnology, Japan, where he is a Professor\nwith the School of Computing. His research\ninterests include vision-based human\u2013computer\ninteraction, human augmentation, information\nvisualization, and usable security.\nVOLUME 10, 2022 106295\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: What is the title of the paper referenced in citation [28]?\nAnswer: "}, {"role": "assistant", "content": "The context does not provide information on the title of the paper referenced in citation [28]."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106295\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nIEEE/CVF Conf.\nComput. Vis. Pattern Recognit. (CVPR) , Jun. 2019, pp. 5686\u20135696, doi:\n10.1109/CVPR.2019.00584.\n[36] J. Martinez, R. Hossain, J. Romero, and J. J. Little, \u2018\u2018A simple yet effective\nbaseline for 3D human pose estimation,\u2019\u2019 in Proc. IEEE Int. Conf. Comput.\nVis. (ICCV), Oct. 2017, pp. 2659\u20132668, doi: 10.1109/ICCV.2017.288.\nCHEN-CHIEH LIAO received the B.E. and M.S.\ndegrees from the Tokyo Institute of Technol-\nogy, Japan, in 2020 and 2022, respectively,\nwhere he is currently pursuing the Ph.D. degree\nwith the School of Computing. His research\ninterests include human\u2013computer interaction,\nmixed/augmented reality, and machine vision.\nDONG-HYUN HWANG received the B.S. degree\nfrom the School of Computer Science and Engi-\nneering, Korea University of Technology and\nEducation, Cheonan, South Korea, in 2017, and\nthe M.S. and Ph.D. degrees from the School\nof Computing, Tokyo Institute of Technology,\nTokyo, Japan, in 2019 and 2022, respectively.\nHe was a Young Research Fellow at the Japan\nSociety for the Promotion of Science (JSPS),\nfrom 2020 to 2022. Currently, he is a machine\nlearning researcher at CLOVA Voice&Avatar, NAVER Corporation. His\nresearch interests include mixed/augmented reality, computer vision, and\nmachine learning.\nHIDEKI KOIKE received the B.E., M.E., and\nDr.Eng. degrees from the University of Tokyo,\nin 1986, 1988, and 1991, respectively. He was\nat the University of Electro-Communications,\nTokyo. In 2014, he joined Tokyo Institute of\nTechnology, Japan, where he is a Professor\nwith the School of Computing. His research\ninterests include vision-based human\u2013computer\ninteraction, human augmentation, information\nvisualization, and usable security.\nVOLUME 10, 2022 106295\n\nfile_path: src\\test\\regression\\regression_test004\\eval_questions.txt\n\nWhat is the publication date of the article mentioned in reference [9]?\nQuestion: What is the purpose of the TCC algorithm in the AI Golf swing analysis tool?\nQuestion: What is the purpose of the motion discrepancy detector in the AI Golf Swing Analysis Tool?\nWhat is the purpose of the motion decoder in the AI Golf swing analysis tool?\nWhat method was used to detect the human body in the video dataset?\nWhat is the main challenge addressed in the proposed golf swing analysis tool?\nWhat is the purpose of the AI Golf Swing Analysis Tool mentioned in the context?\nWhat are the main contributions of the proposed golf swing analysis method?\nWhat is one of the main contributions of the study on golf swing analysis?\nWhat is the file type of the file \"eval_questions.txt\"?\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: Question: Who received the B.S. degree from the School of Computer Science and Engineering, Korea University of Technology and Education, Cheonan, South Korea, in 2017?\nAnswer: "}, {"role": "assistant", "content": "Dong-Hyun Hwang received the B.S. degree from the School of Computer Science and Engineering, Korea University of Technology and Education, Cheonan, South Korea, in 2017."}]} -{"messages": [{"role": "system", "content": "You are an expert Q&A system that is trusted around the world.\nAlways answer the query using the provided context information, and not prior knowledge.\nSome rules to follow:\n1. Never directly reference the given context in your answer.\n2. Avoid statements like 'Based on the context, ...' or 'The context information ...' or anything along those lines."}, {"role": "user", "content": "Context information is below.\n---------------------\npage_label: 106292\nfile_path: src\\test\\regression\\regression_test004\\AI_Golf_Golf_Swing_Analysis_Tool_for_Self-Training.pdf\n\nC.-C. Liao et al.: AI Golf: Golf Swing Analysis Tool for Self-Training\nFIGURE 6. Case study with the V-TCC. The line graph shows the distance between two synchronized motions in the latent space. The red line in the\ngraph indicates the threshold for discrepancy detection. The colored skeleton and black skeleton indicate the user\u2019s pose and expert\u2019s pose,\nrespectively. The density and radius of red spheres indicate the degree of joint position difference between the two skeletons.\nFIGURE 7. Pearson\u2019s correlation test for V-TCC. Left: normal videos. Right:\nvideos without background.\nFIGURE 8. Pearson\u2019s correlation test with skeleton input. Left: S-TCC.\nRight: SA-TCC.\njoint difference was negligible. This might be explained by\nthe fact that the network focused not only on human motion\nbut also on other motion features. In this case, the golf club\u2019s\nmovement, which was also critical during the swing, might\nbe the main factor causing the enormous distance in the latent\nspace. This led us to the following quantitative studies, where\nwe conducted statistical tests to examine the accuracy of the\nsynchronization and the correlation between the latent space\nand joint difference under various conditions.\nB. PHASE CLASSIFICATION ACCURACY\nIn this quantitative study, we discovered that the previous\nTCC implementation had limited precision in terms of syn-\nchronization. As our hypothesis suggests, synchronization\nquality might be in\ufb02uenced by various background informa-\ntion and other movements, such as the golf club\u2019s and human\nshadow\u2019s motions. When the background was removed, a sig-\nni\ufb01cant increase in the accuracy of the V-TCC model was\nobserved. This result supported our hypothesis that infor-\nmation outside the human region would affect alignment.Various types of information from the background, the\nmovement of human shadows, and the motion of properties\ncontrolled by a human might be learned as features by the net-\nwork, thus affecting the accuracy of the analysis. Because of\nthe signi\ufb01cant superiority of the without-background version,\nwe considered it necessary to apply background subtraction\nbefore synchronization when building applications for real\nusage.\nNext, considering the skeleton version, both S-TCC and\nSA-TCC outperformed V-TCC under all conditions. Given\nonly 10% of the labeled data, both S-TCC and SA-TCC\nachieved a phase classi\ufb01cation accuracy of over 90%. This\nmight be explained by the fact that the skeleton version\nhad more compressed and precise information than the\nvideo version, where the color information might be noisy.\nAdditionally, 3D poses represented high-level human fea-\ntures. Therefore, the features outside the human region were\nremoved when retrieving 3D human poses, resulting in more\nprecise accuracy for further golf swing analysis. Note that the\nskeleton data were the estimated results from the video data,\nand the estimation could not be perfect for every human pose.\nDespite imperfect skeleton inputs, the S-TCC and SA-TCC\nnetworks outperformed V-TCC. This result suggested that we\ncould use the skeleton version for more precise implementa-\ntion in actual usage.\nFinally, comparing the two skeleton version mod-\nels, S-TCC performed better than SA-TCC under most\nconditions. However, SA-TCC remained competitive in some\ncircumstances (better than S-TCC with 10% labeled data).\nThis led to the following correlation test, in which we dis-\ncussed the ability of the networks to detect discrepancies\nbetween two motions.\nC. CORRELATION\nDiscussing the correlation test, as shown in Figure 7,\nwe observed a correlation between the distance in the latent\nspace and the MPJPE. While the distribution of the data pairs\nseemed to be scattered, a Pearson\u2019s correlation coef\ufb01cient of\nover 0.69 showed the potential of the network to distinguish\nwhether the difference between the two motions was small\nor large. However, as discussed in the case study section,\nin some circumstances, the network failed to detect discrepant\n106292 VOLUME 10, 2022\n\nfile_path: src\\test\\regression\\regression_test004\\train_questions.txt\n\nWhat was the motion manipulation technique used in the AI Golf swing analysis tool?\nWhat is the phase classification accuracy metric used to evaluate the network's ability to classify motion frames to their corresponding phases?\nWhat is the Pearson's correlation coefficient for skeleton inputs using the S-TCC model?\nQuestion: What method is used in this study to align two temporal sequences?\nWhat is the purpose of the proposed system in this study?\nWhat are the three main contributions of the proposed golf swing analysis tool?\nQuestion: What is the main purpose of the proposed AI Golf Swing Analysis Tool for Self-Training?\nQuestion: What is the title of the paper mentioned in the context that applies pose estimation to predict amateur golf swing performance?\nWhat is the title of the paper mentioned in the context?\nQuestion: What is the title of the paper mentioned in the context information?\nQuestion: What is the title of the paper referenced in citation [28]?\nQuestion: Who received the B.S. degree from the School of Computer Science and Engineering, Korea University of Technology and Education, Cheonan, South Korea, in 2017?\nWhat is the correlation between the distance in the latent space and the MPJPE in the golf swing analysis tool?\n---------------------\nGiven the context information and not prior knowledge, answer the query.\nQuery: What is the correlation between the distance in the latent space and the MPJPE in the golf swing analysis tool?\nAnswer: "}, {"role": "assistant", "content": "The correlation between the distance in the latent space and the MPJPE in the golf swing analysis tool is over 0.69, as indicated by the Pearson's correlation coefficient."}]} \ No newline at end of file diff --git a/src/test/regression/regression_test004/model.txt b/src/test/regression/regression_test004/model.txt deleted file mode 100644 index b28fbba..0000000 --- a/src/test/regression/regression_test004/model.txt +++ /dev/null @@ -1,2 +0,0 @@ -ft:gpt-3.5-turbo-0613:personal::8XZ7s9jV -ft:gpt-3.5-turbo-0613:personal::8XaasBXv \ No newline at end of file diff --git a/src/test/regression/regression_test004/train_questions.txt b/src/test/regression/regression_test004/train_questions.txt deleted file mode 100644 index 389a462..0000000 --- a/src/test/regression/regression_test004/train_questions.txt +++ /dev/null @@ -1,11 +0,0 @@ -What is the motion manipulation technique used in the AI Golf swing analysis tool? -What were the two skeleton version models compared in terms of their performance? -Question: What method is used in this study to align two temporal sequences? -What is the purpose of the proposed system in this study? -What is the title of the paper mentioned in the context? -What is the title of the paper referenced in [24]? -Question: According to the context information, what is the title of the paper mentioned in reference [28]? -Question: Who received the B.E., M.E., and Dr.Eng. degrees from the University of Tokyo in 1986, 1988, and 1991, respectively? -What is the motion manipulation technique used in the AI Golf swing analysis tool? -What is the phase classification accuracy metric used to evaluate the network's ability to classify motion frames to their corresponding phases? -What is the Pearson's correlation coefficient for skeleton inputs when using the SA-TCC model? diff --git a/src/utils/finetune_engine.py b/src/utils/finetune_engine.py deleted file mode 100644 index 81998dd..0000000 --- a/src/utils/finetune_engine.py +++ /dev/null @@ -1,34 +0,0 @@ -from llama_index.finetuning import OpenAIFinetuneEngine -import os -from dotenv import load_dotenv -import openai - -# Set your API key - -load_dotenv() -openai_api_key = os.getenv("OPENAI_API_KEY") - -if openai_api_key is not None: - os.environ["OPENAI_API_KEY"] = openai_api_key - openai.api_key = openai_api_key -else: - # Handle the absence of the environment variable - # You might want to log an error, raise an exception, or provide a default value - # For example, setting a default value - os.environ["OPENAI_API_KEY"] = "your_default_api_key" - openai.api_key = "openai_api_key" - -data_path = "./src/test/regression/regression_test004" - - -finetune_engine = OpenAIFinetuneEngine( - "gpt-3.5-turbo", - f'{data_path}/finetuning_events.jsonl', - # start_job_id="" # if you have an existing job, can specify id here -) - -finetune_engine.finetune() - -finetune_engine.get_current_job() - -ft_llm = finetune_engine.get_finetuned_model(temperature=0.3) \ No newline at end of file