From 64c5062c738e6e3a7f040611631a87b75f133dde Mon Sep 17 00:00:00 2001
From: Kai Fricke <kai@anyscale.com>
Date: Tue, 14 Jul 2020 21:56:18 +0200
Subject: [PATCH 1/3] Added Ray Tune Hyperparameter Tuning Tutorial

---
 _static/img/ray-tune.png                      | Bin 0 -> 59762 bytes
 .../hyperparameter_tuning_tutorial.py         | 454 ++++++++++++++++++
 index.rst                                     |   8 +
 requirements.txt                              |   1 +
 4 files changed, 463 insertions(+)
 create mode 100644 _static/img/ray-tune.png
 create mode 100644 beginner_source/hyperparameter_tuning_tutorial.py

diff --git a/_static/img/ray-tune.png b/_static/img/ray-tune.png
new file mode 100644
index 0000000000000000000000000000000000000000..febd6de282e1d62a08b5ebbdf9806d6ce6fe566e
GIT binary patch
literal 59762
zcmdqIg<q6g)HaL-DhMhnEhs5nGQtol(kUPvA~BMqbPS;=2#S<|v`QU7I))rTl#r6{
z8M<@m{`L*%^L*d;{sk}R=kJ`u+_Ues_qx`#uC><IUsYM=(nacv1Ox<^<YXVI5fGe1
z5fGe7K2HdKgAL~K1^*&^rXcf(;1K%fOGRoJ_|1hEvO4wz1jOXfpHl?Uv6SGKL=JLF
zQbeO?X|7)5#b4hDCLmxSkbCq%!=-<I$Ra3t%9VcWP-jk~E3Gi2wB8)HDV201AXF~k
zhM7Xh_n+iEax4${g|61zHzyRhKPhl|^}^(nVTP)vs+Z$TA4Ma|e})w5etQ(cFV!~K
z$K{kCQ=#GPxG>Nc@A^1t<V~}b8qa#{!9KwQH)iq8lEoN?1CNzbcB#kLP65LRzO1Et
zo`(Lzdqj1UoWM+rgwXR1#}~DA#A^bA1K~ai=hL48D>IHBA-V912YR$B`102OeZTOR
ze|+t2!ttT(Is9L=!T)wI^zt}4?aVk3C5ubuKUT_kqOJ(}F$?%}x3<MrLIUL}jN8p1
zb{**|jIAt=QAT(XRW+O6UCn1KS;}e<#jlizE>}|>d%x#HRiuZBeOY8`f?f;#!FF0X
z%34fspyHa8+I+`SCR3V5Xl|REp-53gNQmwt#^ZK}y3((+;B~{%m$&ke-;rzh9~>Fm
zD|0k`gR=_*UpBd<Lchrhq1~WgxsHT)fhVqA1<!k1)wT<51S&fRkEIIb=awASU@J3a
z^K$zdSJY#443^lLpiLZ`!^@UeIHXc(^l5Rxz2Y{ab;(p|m>7Iob(d+`3^%>`zA|b+
zPe>{>yil1tYyN+IGr5B<iERd*ja<t!_dP;)PcYbDau0|6?!pk3A!CkC>e%=zTY1@Q
zI}?{s-~}93=0^Z|v2Y&Mr$y-VtrLeSgR`wuK^C#Wz_H3l)wF+;1@AI9>7X@&ew}p*
z`gMgNmP3b0IKMb7ZV2-bilP@vU=0{`prtP31SXV1%PGGkxrPFK;B6&2+=$`pPfr#z
z<XNZc7IM>pwL|#ywt56UXuFw#FwS~1EASoZkY9v-^1odAs4RY_!JM60yk<jL!iwhf
z|GZy1|8eu9vOz{VC%BKk{tO!nSdO6Q4cKv_VYaz>dX!~hxfllQo91MLfxtM4$Xwlq
zc?W^6tGM5umGyW_);MLO2aqMjvu82$91FS9mW6nyEmA_M7K7e~+6eCB7m9-o|8CM?
z%nZ@BsH|iQB?oS-NIc$Kh6>mMKU^#{#N0L8EX4Gs1#FH+1a-9N4zy@urmnQ!x>~}>
z=x<x0KnKXqVQK)NY(-^3{=049tc!|vbq1!{kG7x{2ex2V@Pxx-xC{Vf^ex>n<kHE(
zHwx-zOBN<qG~`jf3vr&B-N(WGV*LW}9ZUcEAg@F;*mejqTcW<(Ny6vB*Dui9^q-tI
z+6XsKj4tUj2iqgW_&l&Pe+j3Fxq*hPeBlUveRkcFwCRIsk`6XQnDUyihb=sA@&57A
zq=Q{#6YFG}8fS%f%Vu^?Y3uhl?Dz^}jutlni^nxMF^-w6w|jbpn6CV=)$BR<zd!Sr
zW~rS^7O%lu${a)a2`hjy6@8|eTe{AyfVJ&<o{KT-KWpzo3u<OW#E=ChtBrz>H}i@M
zVwCGfg`oFk-2}d9`RT$4!wE(lZ&qu(Gb+>Cdi~lp@VR$}H1J0Ub=qXDPhwh6BhA*v
zuGGhyMy>x07K7F|&yUz}5o?xq%|?WPuiLyIfEF)?wyHoxR>0q9`)P+{if8fSu_KRu
zFQ6W%s1yNTHqcu#U-_X-==1RSeAuBcg^>pi^5W!^e<FH&=glli!VHmLzCdg0UkC3?
zj!H8pz4;vM=O?;{etdp=;k(V+iM<bjy@KuEC*t8OY8m0Mm4(gs<t`3fF?uIxCj}dB
zH!X{=fqgVf(&1R33b+kL-MZ3=%_7k2o6WiG^BJjZ!tX6B*^P|aMnu3bLixVx$t`f!
z(b3OMZJm-j5ydZF!WuK;NC@A)K8o01!1d@!vJc=>CT8qFMCO8F3-_YwzepDa@9u^U
zy9up!=jCYG0M&S3q-N8~Ik5C_I>ds3L0$GZgD`N??n=A+x9gOxO<lWAc;e{W|18H^
z=`sm>IKi#vPND-34A(m`oIWFRhZb1I$u+lpwD@lb?@CI7{1R|AfD!9;yO+sf%Bk63
zz?%v5AA>V3+&;q?6=j+7(|sT>f1nkj(t+taY|!UgpFj}VpbFw+kkkr=y4ZU30Nwbn
zM5`0PL%Z%_=?@&VuQQDfyvk$C5jw^S%gbXqdG%HB>cYBNVY@MHX}!Ys_^+~di$83k
zgW+Ynb{(wm%>jX|+gj6ptXlKNq+YUk-LLSWSh315WyF{NAPe;QfQ6gqGBxdLs@pi5
z3gt1yHo-l(-lp1!Up7Y%2!jVI()^`uMSB5i{E^wP`iwdlhy<BqR6PTy;|_ufwRNV$
zLa5Q*fJh48-d3oujeVjDEs#YCe1iq2t4A7ShAJ?ewFe0=tgFx#B*=gx*I$QRZzW_E
z6+_jqrjZZgR&QxydL?jv{sh(hXEC3Cbeakfk$w2&_Hw{BH9^ZzuW_>I#6qgmM)Jl^
zrBQlhL6L4R^(u|Vf2L)|ilZ~X9BtqV^f4e>GOwp1jLo_ht-NfZkzVBy$rH*VbS8Ar
zGsSz*8*rSqG#TMLv=mw*!Y?fa;ALl|j$E8azNK~%$z>aJXM3~=@Y%<kxcYW{*0&>M
z_FX{T2~2A+w-pwRxAdRdaw_xLTi%|4sQ$>sk$@-Li;XeiaHLJDCI@m9&K4S$inco#
z%{n0_Ct#n+H-pv+XXWI(YKy{{_X==fr)P{-G-4<3zq+MHy^!m+SezUOumqe~YtAEq
zvK4O3s1<?Cf8h!l0jR>qp0Q*i#jf=zHh22nG`mHpNG`rT-o=$IHFkRe#**!nTn&ZT
zts8RKd+0b|O8X4SF<2U<hZWL3fH)Zlaf%MKGa5>s7%F$N6QI$XJkGt_=TUdUbK7*q
z-Ea=(sWHHiNlhM}^1SF@#PDNRnlUaXYxIu=(tf6D;PR_eALYZ$tOtm0;w`y3mr^=r
zI5;7DBj18E(fN#=IOD9CV#u>ei&ICS@%?i#eSLT9juLUB<gmg_lj7ag>atY+{&e7!
z46C8?y|F^hY^M`rAA%>KJhMN!0^!!<W8Bb#c#oG-$J7lO;-H(UOFG0GElHf17U2@S
z^i&lv=mMu<FW<>az!G<qRD|q%Y{(&Ac>A6S89Vg5jhqGwP<F^t34vEU$$|o1O?48Y
zF%3=?0Gk8T6Mn5+h?O57j{vYN0MH>Hws7Pf@TNN}8h#FCKG2rCg-ZMS__~h-L_%Oy
zq?#qbxCCz~qhGBW*9rsfcEemMZ1qQmRbMav|2)+m9c4M$qQ~0)Tmx)sBHR{iin%d^
z?idS*+snxXAcdlq7t;#rev8#PqdQvjd9ucvMxFkmd20s1A#+1p%o*F?vH(LXfTvWM
zc<A&R+s0^)5wk1+(4Y%Ajo3bO&|~)zHE{Ptc$aOwyDM{smBBG$QdyRhJIzy~OLjN>
z3#5O0gu}rJdzo=fanG9~>a>ZV1M(CB=(Rz8k9S{+{XIfcCok6hRe}w_*O&pM?cL+(
z+DZfV6mD!`X>nz?F|wJOch+?xwBW8AzrrDR-T=#ZCj?YQ+8GU7G7G3TqqgGHJ``Cd
zi#GzL+=?j<6?i`|<M6=|dfzZB>vHlMFlxI~auxXADXmU~Bw#$*#7fu=RdZq!JFp39
z_btT7J`*(^pC(8>fobQ+7y>?#*X$|1cqCI*6o(sCwLe=y>8zdl&Sv<c1Ci^@v>SJp
z13%@plV^bHQ>CQOm;&zL7wF<#7bf(9GROEhl*g&(AcLw@0L7|457e;VZC)6BsQ~Y~
zN`~hO1@sd~FY3+*&Wh~_^AcBH@koB{`)&q^N=95f5-9exRv@7p$yIpFEL>qg;Jt*9
z;(|MiAqjk2qA9-dHEPhI=DGafc+7(kS}{yKeI+I&Zs^qyh<BDT7f@8x3oC9PX{iTj
z=2Rz3Nadh2-R^aZh;Y7sbf%k1!;Q=Ola|c+;7sQw1*qr&yK%<o?TfA_zAq_mZ+NQ8
z#trRZ>3vn?SAMTA;k%3RKt2KnrX&fVh3AHZA&P+dSJ#Ff!cu?@0D@`;h}x_b`Q(x=
z88W_BvamPY_g>i1T*3IrdLv+cIAEf_8S|UlTl%%qKu|SyJMrMQ7M}qMFgMB1D27%m
zX9cSPH60%0@tg$)f|NhomgZVR+t}Q81|l-Ql>0cvRe)WJ{J(dp*;D2^E2AJ>SzzT#
z11+^A2exT2Ysn0e+tf$@4}y&zzu<8X3|$<#SXK0}Q1hV5oqUV`Qwj^s_8+m1sCdgy
zJG4@v<Vfqu21EoW@(g%lSPZ;M9%5OH&29Q(6aK=`AV-jhCg&oK)aSDJdGI+DF{_!V
zaqP?yK;KUs<;ZfQ(?-7CO&S99CC(zNMAB+^wsnA++Xn&)kU)jQJ$@Zga40w&C@wdL
z?uPq`yD>E1(#CF#3n+jZtkBw8=h)I@V25s|CYK+Qwz-C0EI7>{&x3?l7R1ntSH?m_
zdjwx*@RX%Gd7!Avq4i>^&247f5G0B)hc(HuI)g3II>*BHHE?YAO{6IRU8|?vtuW>#
zJl(Yhg}QG10&DA3>~Yz^{b0E94;(e7?YYR9U$VaJG82imvEIN}q(<iDb?25}0>|Tx
zzOu!?)<3pRD=Zj6h826VGhwsu8KUJ`9*>AOY5<b?2)$cKa7qUXhMD74iUMS)=PY3h
zkoE4K@SwwC6aXklN9Q1i41)(txFLpLmLVHo9B8a8&+MQYpD>wo-Cxrb5^_J@Chqc<
zodIRhgiSW4C?3R98~#XKAAkxgi>9Wa+{UpYX6az~^|2rJU}rJ<Fkc(j0ceL!s~kGI
zeNXhk+XzCz+v2%*V;Tp<*Z-hwKO@6S8_}|MJA+P`l_C58!IB(U05OTs+wS#4*g<I_
z0_CoUhSw8kXH6uXI2*vtF}W+<%q=fUMf3&3m`R~pf|)84E4H^Otd(;DXnY3|^KB?d
z^_ReM&Nn7?kH`wJ=%?}qnAP0SUPF)aRCrH!etG|qdERreRgrTo<Bu4LrgTdVu~C3+
zaTB{^Juv)wT`R;6aCqsFz|TT}jVl1shIp+>8K#bB!GXGDb*Qp580rqR|IQjcdUG9k
zb7U<@kB68U2qV$Wb`m}ZU+NWPdhlIKvCZ4bLqJ@ghLEqH3Ox&43nc<@Jk^Ryg7RMR
zK<FIoBzMJ%R)hyO$N$fh2^5vO4V{k|$*>YJWwsoL_@W0PKqWA$i06M&M-;Wse4c<1
z`a<mLmfetL5R4x_)YR%>Dg@PC2?)o`7>VuvnR0x;n^2L%@hJs?_xF&t6xtnnbO_(;
zVc$Pm@}O2T-qlkWWTX7!txJNfTR{O@X2l+an!jHwZIvw1TC2$;;j;0}P@*C{)5hQk
z^*uDNczOy!Wd<sd$&jGd316SY<R$!q6A{#I5k35E34nN!1Tr%23NSJz`uTWBQ1ir;
zgm`4PY)KjFJrm*vdUatbp>@U{uA{fa&I_Kcas=7HF<T5<KWHsbxl<wur9}+|>||LP
zydIXI-2|(2=7#A0@597$!1I8R8uotFtWwcrM(%J*g}&#AL+v)K>gddwB2u%fz`6uI
zpee*;A!H?7z#B~@E*J(xoFrf4CMW}3C%3~nGv_aQK-8K=e%i=I-Tiy~?qc+BKn%Rs
z@=>^ZWpRf~ttt~yQ(AesUFuK4bKY~_RTS7x%PzaE+BVdviG5U@Z3Lzf4WdhihkFVT
zUf%NvsP9tc_+XgWs}XXg#0v)cQUv?DTL~jjB>)?H6PvA5KvHyfmmM7x@Az5FGP<CS
z1}rv7aWlU4Rn)D|69XLh`S%>#8(~xSi$8#vK6aFER29Vm9AH#*uj&<QuSUQ^@b}>(
zAcnU4$ENJ&fBGJClce%6@nIY&ZqYYyw$_l*QU>T?AVx9LYb94Ps_*+1H$Oj(hj^1x
z^R&^{61`MmXcWjQ<%jnd%M0UlGxzdQkrD9i{Op8=)$H%H$B#a9lsJV{UkiPvx+tKL
zE;o@{i9J!y5dQ(K8u30OPVJDk!ZQR)r>qGjpWPQfH-YlRJby9^3I`I<bwl>x<eXZW
z;H+Uo`Ep)dR1TY)FfAH68-A*iTNZS3(zc4T3kf@iSs<N*6lS#dvl=z$zv{Y2G}W<i
zJ(+tBG;E9x7fVFf+WI~Ob=%ctk}pC+y(;+17T!q;)`#K~KCEk~Rgm-IXz)zf)m5-G
zk=s4fu$Rr!T#~ST_!4Lrpa48aMz%qq=(fZT$VD;Y9)v*9#VpIP+N~;$>3m*F5yQT~
zLD2AhM};XjXj2fO2yNLm%?aD5-!NaIA%eT(21|0BeH({|^*P8q*6uG>I<qvJZzm|E
z3e<PBavfoz6kAvY5&YBo^C;J{2e6q;ri#EMw#N+GYnzx7v|XRHDp1+m^%d&d>#ev?
zF}9RS$>Ygfp`6?c<CUIsiXcXRorZx0yl*isPh2IGa~7tuUfN&c_t<>uzBIto;f&TT
zY-%mS=aRLKW8thF&p#bD%rBBc^z4d{N(OTL<<Rjasv^-GZgO6$d4W=FyJ+w=MHQjd
zG`~p!*Z^%bH?DH;U>nM9C94miZkM<5t>uXY!iowh8>tIZWjFRKko5Z495>()6^Xpy
zR?)m*&;(jzUfc+PB+9mrgjm!ng)`7Ij!Q@H^R9|im|h5!W+R-~T^nuAB`qm`wFEfW
z@5V7-Pb_PvhvT(et;!;iC}+MP3CCL*b4R@nZ(|1N)kCH^F%@3hmDT~!s}Bh5kBdxp
z`om>wAbcix>_Z?KxFqIrE49^T?iiq_Pa9#grXwYmD^8EGfv1G+o^yH~=KJBlx~H`N
z7g^ctY@ZQ{6niPIB3kijq6>6ye(Xa^n~e{^Qfn<+)^Glaz_ejV6<Rmb)Uzl&VYl8|
ze0TaG%Rt~)?O~^lJ*NZw?t9mTadVJF7bU<q@B3c>p8XbSsycdx1S?uI(F3^QX-ZV(
z=6l%W!ETB={cCRA5RmpRZg-$frnXoocg(m$u|UAgeB{$KcP);R&{ym8YqFs3+YWEV
zegYi}`T%a6z<c<1jf6xLD9Bq&h~e?jm+s3!FO=s>`?O`Lk?xKteoczI7a>C*-N@s$
zQ>&;XByWU2PDQH|+2tY06Ccmn{(ad(0F#t|r1SM(*Kx3(o7q!Ej0>O{ex_vmkG5qO
zKFF!Je{o`kUcVllAesQ)4eG0Y$eCZP&w>bGmP`J5$~@Wb7ps~^MmSU)E@h<w2Fj<h
z>5@0v04iFdFiK*pBW!yvQZp~t>=mRR@Am<1@CNS}c_$0yJgY}M=s98UXV*$cPW74v
z2P(O82L~<{aeF@CLcL)~x~TF)E>z>%>|)!I<11$ON`ut;i#k)!{Z-gRRg>(Df)NuW
zueB=eg?Q-o8XG$Fu+VAV_lH7Eej$KItn6EWXJ`ibmQYL~E#@UB8P@BFfCJsO2T;2{
zw_7X9qj~n){PndL$47r>!JsXwUOjEZ?ZL%vhN1rTglc9P>I*{D*?Ksr4|PP(v%r%8
z$GZdWOz_xu0FAX9#nAUYfnGD3CLn}V3pt`)c+^%-*dKo=zVGh?3WCUki`PKmPzG@`
ze$B|J2`Gp^PznC!EXc|`x}RK=YMxL}Eu)*8KLVM?il31oovi<w3$WHb0|jy8_C8PL
zkp`^QZ0!JFB<P*`>kET)snF`KORO1P@+mCT+v7g=+i(dl-i4JNRsFY_3!VqFATF^T
z6p#)Rp*v3AyHOx4U92qhbOtnb<+40smhR;t+kKh`ec3~yDiU_QjLmq7?Ar}}tMKjj
zm7+0`L75ZdGhD6E0Y9ZD-?H=9Ym-e+ykj%0cMW)l#cSs6wkGIZ{RY@mwT1JGUrVHO
z+rQn4DF{7WW6ef*<VOZJXTc!Rl%sNlTgEgY1(dfH#TJYE*Cm<7xZhCipW8M<x$tL(
z1SGtSql9PvvhJAnJmiNP2cB5UWj4`ae3FWt4E{4oJW|P(hkcTOH#;mk8*!7M>jP#g
zt3+OI{9gO**wg6hFUy+HX>CC1M$3YwOVgi>S2o}P>y`B@F<p~iV&FHVnkv`Q-U6K`
za5jvteSs8etU>Ytgts&ZZ|qU~OdRm;0()VS8vgOw{Y)#7?yp_x6F`1f?nOzuj+@w9
zkT#q~gI3Po3{agsm0sdU0e&1ePbHv?6h4a`Es~Ft5n0Qlcx(vI?_i9Iv6zG%RC56_
zI@4)Y!`DAj7#0I4)Kx6IyeQtD!s*SC(%O|BQmGoyu#~f$5U+W<86D`bH2@0I#z`lM
zgu}NqkYBMtqU5Py7S`WGEs=vEiXg<Z^7uAQG5h?xOHepHrT;V1$gF{l9ntHa<Lf0x
z2^3EnwOe=FJvSCi*y`06KM3jj5J)P+TUyc4N)})78Hqyjfi0Ms<j?g%CpdrRC>|9o
z87IUg+Y|_Q8Q)k3b*+ZDr{Yz-)+YzOaEjFHz<bNJeG7YDUY_j#!>j{PfE3pPr2mMa
z!-R(#27Qy=p^?j#IE@4H5ugsLh_;hexS#^+_vOJ*KwJLmG24^me;GEO6s%Bnmj!Y`
zSCY`iBqxraJ@lX(^Lb=90<3XHpm`6EAe{jEGP#fTV=B6EsC7&w4*W5wPZQA7(SsfF
z*RbwY;0Le#g3O;iiPMWQq&Rewa*(NQN)&F32t9t_o~oW5UfFtFvK-(fT)2$(TS{Wf
za!lfnq#PK!P0-m&i=}<K8_XGcpsYr+F%(G$N;SQp4^xtxie}#o=mXkFeQpbTS_&9#
zC5~;ks88jWVw+o(aWoM2iwQx{`nZ*3%e3`5(_!S)vbF4MP#i~@sUabzwfDIVo2>$<
zgg|9=#M-Hz4txq6Wf4alqkjg-u1Jlv2xkkuP!f@br-xiuGBi1&5^N?$h)Box+HF1e
zcBBo(i{d>SYT9l$^t-`uI|t8LSgD0wo-%<l7AQ+BG=Z#ol2j9^kkOeBt3!VXsCaAh
z>mZhSz_JU>pXyoZV;iTuMtp%P@O@TLV>ms~s4h|Z8DXg1Fk=rjh@8Lz1jW%{BQ?#>
zKe_lKL)SJR;yluUWZ$#qI2`pV=Q`YkT{irC7hqiLqv<y>+UAY@2O%b`LU<xCJAR(>
z(g@HKEoTNxd1uHDHzq%sMLKOqbbcv=)j8Aafu`l3AI}f4JlM~>OEGeZhExF=9e@b;
z>A5Fx9xE=~oX&QOdmlHU3FeD%kXZ`yh%i#fC5hcZFqhjL<cl(LaS;b2kTjBp^}fD^
zDLoa=$<0cK+X{_keZBK;>d+pvUR6ap7gI5UbViI0z+!F!wm<Hvi0?MMMQwx;M^S%s
z#bnG4q4rnoAvA(P3*#E3m5VO|FzY6VXjo3{AAFM~j8j9+4$jL$Q(>>EFyfM_4S&Iy
zniWp>{o6c;E@G9t2}X_1@gfNmT_2f`lCYLG2x;ag>2ifSqh}x;9l(<bJPAsqHlnF~
zOk3mc{tVD+W88o6c*m*}?0mGXaBmydI-c^jO^~LLv#F;XLZmkbsFsrVn@Mo;8wcn1
zL+=D6v9+Fmwc)}@Y~L_-(?Aw<>48K(no3i>ZUUpo3TQ7npRLRe?n9$UfDJ5kk)!H5
z&{}k#%4^Wmk?hM#p$-1hqzKL+alyzigwU$Hm&<Jm?!}b(C}fB!7-sQ!Oau<1I<))8
z{M0iS-wuKW2=oojxA3H}#9VZ7W&y`fLZmNdBARO8L|p6)qr_b{HKe6|$NX=|wQ$-_
zNgS-q{R@weu17={83VFXMrW8MNAe}qd(U<pXEQCX(lxkCz8_3JKrO!4VN@YZk3onu
z*9LRGoV)HiQ8%}Z*TU9A5KRs@_<9tkX68v}A~h|T*3**I<b)9yuQfcK0>X2Gn(hzK
zWDwzbINSo$JgBb*q%QZqDDJlVDynj&;ujD!g0o#vNhKK5rIB$R2@%gQKj2RI;0I~7
z1h%E?cXs5ETzA}(_x8Xs6PZUJ>9p3py=u^_l6;xD@epSlk21f5b9=^@9XD8CYu<V!
z@`@E5w(KmzVTi4Wq8gYSQh>@qUhj3^ObBQ!;42R%;{YxvfYi4xa&DA3ZLA84V`qFa
z5N;r`*2>HY$L1<3YmY)yzMS7a0s_WT&XIE>ZbXl1Kaj~sEpnv(>&l=@PPoy2atyxO
zou%VHvbt%g7m-U&vyC>o7m@8brGGTWHt#8e()bqD1rRs#chdqF&s8RC%sXz9{-(dp
z-Gs)EkSR_G1aU$GUSlff0s3qIA37`i36(O#2#5_eLT8`*XUm3=6e)<ck8pvcJZRV9
z;<*rM`&(kq-A};lDN{|}=yxEQl1=^(u4=6!G2+ftQ%G;V>b%7=gJhA~K7+F?3@zyk
z0j}61^O{mM6fKQd5_J@Zm-)0OfO`@}*k%|8{TVzYT@md^e=pcPKYBK&fFIhEzLW^c
zHFAJ2RO9y!;<OZhuhPyy&Eo(|=99^R(f*(k7~QON`Z8VS3I!JoRzY8IyZb259Hn}C
zut=5413#Xn@BOmrnRrhxFDo`_k@Do@8njicG#K0zSlYw3L5<skFKqjLgi>I9&T6#t
zn+D8U!00W$hN^M=Q4Jj8rmYLp%M%8+b-h?^zdZKRhNp=C<oSi$>Ol)1KP_p0RO``@
zk@j3gL(GV*e8G8=aKeCEay}1MSSG)`h37^m7Q(u}eh*j#3gTIDgm+=xT*-ngymHIm
z$8|c9h`-I2r1>fuI2(*Tw|~&nwL)amwVriMHWH^^@hsWFXXQNj?^wm&n`Ww5PogBz
z(n*KJ`%%r=?B${$zF%Lhs7#_VJplAqd+s`Fn*%wXMsM&>s}6hc*F-vN(s|hXBAMMY
z3guJSHl#(aRWQ}uB`yM;I#4D-?&w+vvr*frC<@KIfjeHs227x}Gf9KfQ!La4MBWT~
z5VD1(usVo!yq!Ttk9#sJXCs)ru-wcrgCrS}^+Okcnp8;;Kkc)sgPS%pTj-({6Q@-e
z3}Bk084}%YQiHV5qq2B++k~bc<yDZ39v%Xk9}%vr79-0Zt|qjVGR<01SfPWFgtF2P
zeSC=wMf(!Pb{q~KZNloyxCXkjAen8ec;VgByT%g^69+<>h*!Cg;Bp(mFx}^o#<IbW
z=4v!ND|hdN9!#ndVdVb*UMo_g3w?zFxxTD847XcDzCFZ)i2$Xq-Al8c0m%6-L_jg=
zBu5CQgvrZg%klNjEz_;}Vl+Ub-4@>7!s9vm^ybd=E5!moguh<M_^3Rq*P0QoKC%)o
z4ay2|YHkT&6@o~>%yCb!^UIl^97p=}6b5`HXuoT6Se%Hjn8W1=)J}*`R4iruAG8D;
zbeH3ByH$Cz5;dl%9M|_7r~sc^lm?Indm$VJh%7@RY<3z;g3FslvidB{ebWbR@YVdI
zKGOFrXM3lY@YeF2(4m&6Y%ORo;V+5QT16i0w?RDBm{7t8#$C$p<s+Gpo+$@i+Oq#x
z)>o4F)!LVMLJ!>di#^RTMgM09h9(VgfX<Zk!MIgwJ}>Tf?{hBn&8(EZ@2*+UMuA4Q
zWe-mhiO^6Oe9(te+JW~x=>t<v&L-vrcZXY`YlYjcNCY!mmxs8a76ugwXFe07bZF9}
zTNt{1CR0&)`Y1VrfLTm$PU>f0+9Y$@1PI-)V0P&7|DTiw23^}^MxJ8O7&7NW`A{F-
zIHV7L{2JP8t$ZotIin6s(hXcW(wvU1?Lycw-w!N>xqu9EW7agop1l|f&k{3iwUO!H
z{-r|O#47H@0On~L95wbB+iXa>;g>KjqBz%C6c|tR_072FgWXj*W^VMF1c^-|hUwP3
z!@B61%2LBV<%kuu%d*hR)-G{xJ`YHkoDP>y50?j%N6w}NM~j4toUrChL*Jw}7}$Ma
zrlFHFQIy1=-%V?V|JkHKx$Z>;fBP$EEt?3CxzQB56g`M3M8=awAjE=DC~lL)!tclp
zd^jlJ3fC7^-(aLd!HodS!H|6)b_QX!+l`w2x-Cd^qXYs+5117#WvQv813{$Ql6(Fa
z!%QZ#)MhD{amEvc!f2>%eXuQ2dKc&+(6^+>&xg~Ra6>U!0S&r9$I{%^oT*tFOm2?4
zH^V!iU>bsAyZk2+CDZ=;oXE1$hBPghBHMu*BHI$4yNiM}U635<qYjNiOND|q%)#gY
zLI4Jbyy;N74~8I%mmpg2xel4uFH0@pUX5(yz-?0e!pizRm@4Hl%hzcPku*uu<k*fp
zoZkSA1gLP6yuQQD<%};g5{(aRHazG;;^NE`BcjrLPkajQpOs|@hEDEby%4EyAnO86
za}p!^h&_;{#)3W%LX39zluB~SW2OuZn{5#;weatm&Z~7s9*p7gc)i2%qCw{#jeZmN
zL@r<)fv>?rZrx4yrq(KWx<+UmEHr#6942U^$f{1p%1AsLm86-GqMK$4E+<5#g@9rr
z3|xtD+kf`Fr5HpP0G;*HS_ioEW|<daUJ>n9F$)D2fus`1L0=;}>o#wo!=E6<sOH%x
zaQtFjiWP<p2zrD1T1}W0%}_bGeNb94a^3@UGeEI6@!rdcAKbKCjj7OE9p29edNK)g
zFCp)Ge+50Jp(pG>G^By+BfizP?KtHCPqmfo^~z*GlCPp3fX0enhjk}D+iMdrE+*vQ
z>Ai=b9SZJPjPqbTxPhasZE(>lVuA_giIzv`<1GMH0NPR@Eiim@mnbd?U9^cSe7wG=
zXPzpfmi=V)gmz(v2h*x=G05gMk69TC0WjyEUo%YI36UHNAOoh6FrVRarAM0zrm3)`
zeOlcmg3BRZX4gcbqhg<rF9({qu6-@pOEeOX1yxq6W+qAVb^sH&8?+aRjBi--nt7z0
zFwoG@5i{d32zBtDZh@3r(WNWrA>!^Ou38)XU3LfAyke6v#awWlH4_egRzsUosdN$n
z?IwgQwTC(rxOX`zaX8cP{ktQ+a(}fflu1BcNdOuNcz?YO7JNc2=kW{~kEA+N^)1pN
zg45QWqRWyQG`+=ls-r#xb_?)f>%9*0!-QSkte)4xnHs<iMQ9oeF7s_?dZF~hS67MD
zv!8TWM$JHT8xkUb=*<3Jni<u!g}13@>Lj69J2i(+ya7mlN76H#Bmk;bw#TwDpm(S2
z%M8Zd5CzvvMF@!X$-0-v6D=h)cRrF#g?NDX8srIiu{5Qma>j)BN$B`BSQlOlXeh;^
zIZ%!?>}B9Ea5=mmD9piV?CS*&A}Qz!Rpr)33AmS<SLk5{jqIWWEu)p<AT55s1axVT
z;nMteCTcWFgagc?ReG<MT7dotP+b}ljDbp!_L+^<Y!JqZDtGbphg<8h27^@TQ5dlP
z5hKJ~np=L0+FfL~>$sG7)J*=;d5jkeMb@rdCW=N^--Ai)sL-pWU8nIsF42Ly2h}*S
z*lLFD(S$hKQYqWU40qz$B4@X4BhRrG5<;m-a1~G3xE!JcRb`M|^^zkrWHv=mjg~oK
z4w0b`5Ij|r9^ew#IHoKm4-`>Qj^7!Z$}8`uO8?_*k`XC!IN1k{+Pt<UhC+;XIxnSZ
znj9fnRlX2}@I0TynWeO{*zfziePS`6?4qrhsmcT_yx7tTJ6HD)E`8HfDpYF#(gU<2
z)4Swbj0T&*9U&<`(W!Z|0Dv5DT?+;V(_kc&tHHsCh2oBljt)~UQJ7niOB!yeM)G@e
zUrJ&olSN94%j@C)<XzfWtBJ{&TKroH5vWFPU$~?)3k|%$K=i0=l{*qcpB2FP5ELa8
zM+=_=2vTumN>d5nBJipBW#eN|xOe9mJq6O1QIy#Tn~KQ8<q2}=df-8DlQUZ=xMc`U
zq!FNm1Dz_=VjGR!ROZ5mz@yQZw;f2F2O2V=wbp8BtWw!1oDs=}2<jY0V*mPonvni^
z6&JGCEm3k;4*!;it`&@qzDCW&Fu{>dJDw}8%HTGl0vc8ivSO5VPoN9fJHdvS0LnFi
zuwldBbYUiG`qS$9z$nw%>Lgd*wCV0W3!q}}1dM~5A$h#8=pJQTbF)IYhJXhsn!!b9
zs9u=a3}|W*6cF#0bstO!fp)zMD#XpyY1X)ohFP-EYE9u*RU76GGS*C*P4g?O>Q+HL
z0`1?;P~RiunO(;Qow{MAl3g>PE7(=2{4t?lu|l%}+^guAyTmV|rjR*#02#R(1}&*3
z2gW+#OU(Q0cJW_DLKQtZ!2~M((%pTkyVULi9x~Bg4(?Jnc)+#ie3-^pPzIaiM^<`T
z++|fkLY@;4(pAXRJ$89-<lb&Wxk`AFcu41foKD^F20CoWUan1vKRrnz5!5Z<ZUX)O
zuwn{Jj4-O^`N6P?$|Sg1)7M8w(k%R*DMqgXwE&W0@9YHnV8v#;sg+(aw4iRk8rpwO
zB@nt^pbaS!-I*Gukw@VZuGl75Iu-UnON=KBbi&N-vpbxDBmzaz;gm#01Q_(Gpkbn!
z801@w*Ln{{$A4NK3zv@JG(dxVoEd!qD(2?NKsZ(Q_hLu$<w0>fv3u|>VXYg8&Ivn^
zTLBR#fqSYw5?=Q}g0Eb4msy>)nwweoV}+FTZ%Y_%5Bm(NE!*+7;+xZ#N>;m_YNg*o
zJW_Q4<m*02w*8L)m=ykF3&py%<J1~Co;tZPt?S-}lpRQ_UWaw`Q`=gaepwyP?&t&x
zLxX1egGhBMnxS32Y2Y$}ldylW0BGqjJrD**%`q>8(lHIacTc1%z76%bVsjxm(`gyL
zho!&SN|s=B6OfTyUOv3u2za%~d1q$xJy0Bq-+S!G{}d|#;|TQCp>4|xShO&Ane+E@
zkDLmqSXx9b1v>xN*|~P{c~Fl%2oR<QaJ8oF=&H-zvNE4*k`684@;-e<PUzw4p&K`J
zcj|E8Xb{|~1{y;;TFDt=_+lZ;ATgl$JPEOp&w|&uSu#{u_5%U9`W0CuB0da$(%5&c
z3s+6k06jo~{`3iu=t_(|1!H0=z$`kYgDUF}OwP$^S9qn?9H>ZP)N{@-@@5U(9^@^8
zEFpi2C8NTSPL?p;vU0>Ul6i=DLSIBHB||OK)Ga;Ats=B?_dP5L+#~wJhX2{20GbPz
zo=k&4nCP;_?8Mop^<-=p8kODwICNmCrJ3A;M2A)!DRdflXEGNYx13tS*MGB`IJ1n=
z%UYUPy1l&{Q>ekwP+g)4UEKkLU7(`-u#1O>^Fy1F`kNk;+ndeT%S5K683<0FJ%5$t
zn{0z-4rfECbz`C+KDN-=X)1O<+sVlZRlcG9?X-+B)5p^?-(*9soew5>7fU@fP0QXB
zxLaf6{fSHM(W2gZc}w-<vC9(Edbo7=jQE|k)rE9$AYXKei(e$SB=-?UnDO>5jh5Vy
zd!6DT_2}ql9MkqDdyu;09D;43vU&8A^;Pdml9w{>t_hz4NV7~vMQ((Rxk?k-?{#cu
zzwk(t8k9;yp0?IFW9Vt}cKg?jNCXXPq=TGR{7;a@H$$T<$3G=J*S3CV=3X3LJ0mT0
zc*x54hUPN5Fmf|wPWMA#(yOV%mB69-&+N+yhXYc|@k31K%QCup$ewW<X0HBtZ$G0@
z%invbifdkf|L*u_P$AtRSdZNKi!0$QYyN<bRrpu@l^2BVa%Lq+c`2IK!f6iS52RDt
zI&(RB-M2oJa;ycKXycYk%AYJP|M_^;n9}6wu&Y~(|M6<|6^T_^wn|qYB5qpN<-?h+
zKC7b6;sq?E-VSPgG*RR3yGAS&kH{+?G&T`nn9chGJ7pOdY{J?j?Dj0~8P?q4o}cl3
z@l>O%%g37yahFW<8KkZ`<$UeCnr>jQ)MN73*WY>M1oU_8HVIz3=o(|tA$72muTHYz
zE_#@#<GDM@J{RMuwZ3!OL4*$(ub+xNpLX{&X0OD;x-m(kfKN$M<^11+%zzG(k1l7o
znp?7tA<m@fb1#wCg>Ra=a}1k@7N*o|TKd)l!>b9jsh`m~oIXYsMNQ^+zr^t4-UzOV
zd6l$Ccy3;N5WGC)B~rT>byc`ENyB07&0oXIq{B(=-jlpbx9~&mQk&k5w)gvr@9aM1
zfUhd5F66!RnVM&L&BRB;>J$ZnX3KNX(dts-oA5%?rPyr#%HAKmB<vpCA8khOvG{(q
z_~&~2i>QfiO6mR~JuezvidP>9?H3u=#Ll&-_RnyT*>dQ=S8HP-F(IX&uFD&q)en;v
zNNVh-dq@6Xmn3k%{fH&qfDq+>#h(;rc=WU*;@{ss{;@-7%rKx^UR1>>l8L^5O~TDU
z=ymS(rFUm?I2UGwcm;HRI&vONu3FSB5c9OWT(OK}n8(x%MrXbYL!^r$4Qlw6`s3aW
z2{6}6ujyYq@f#e1L0q^Bt4TBMqd@DdTi=t#FmnIx4@wiVlM-z{dG}3l++*aIFn6=i
zvd?0=i>$XM>*)HJ6(=XU1S433^K@nPyv!#QTjO|rFW>}P3y+h121}WIpI%Q#J}i!E
zzDs%I+UA21<D85A9Se33@)Q+oo|^tK+5C8WEYWo~FgP$hXsOOXYbBT|N9U5u+!|s2
zs*;o9^*iHS-`H*?bZox-yZ<==1RGff+Xwq+9kLMi7kp1)1wWDb7<5P2G8Wl{HtH0A
zY{-hbU51d?j;Q|a=4Z~qw1`<9JpcN0-!=<ttL-x#mJafBbdxM63!<f7We*aPkCF;1
zKb+(56D@8gaK5(l(c7S{+GaqN8!;k$!-P?or4C=ve}%YlA?uPy4X<(HtN^<A^q2Oj
z;8jk=nyRApy}xHkCIglesgXT{+qnK;!tdu~wx>0({0@30K$4Q^G30AK@Uw^{;GAD7
zDH%Us1j6>~uh~+SYpe!nEQ6x9;lhgzwLnX*B#oT3HR0ed|FE>J3T-;zs{mQy*k!ap
zDdCxb-CLF9iYj`g20UsJa%s=UdRfR5g#nll`I7?ZKex}9B0FKqUo^??c_p{J<#4R}
zl0C1=-Pc?6@Gm~MpNd?UwGA3c3O}!T;Rob(+ho7iU#~U<&g}dq6>j}0c<CVfFemaE
zc5<px$JDh@5`Ag6R?n}bVx^n<aqQCpi~3zaDIR7@i6`t-zpFmXfv$6?T2vtuouj-@
zNRIBsT|~kATR*%oc%QZ+T4eLbTFFF*uo7kF|8_#ixYGbosG#nZY&yL}?0NrhXM6bW
z^jO#W+81Xfv%Wc@ygeE~c>|{RHI(3}fu4HPVdCMBsYGbuGyUz~^$R%38wH{0T$c18
zxm5h!W(&lqaj!=O=5cIJ@!({5B8PsQ6|dOiq+s;BZ*Ne_xW2#Kw{&DIZ~G#4XW&oi
z*++f)ejP*}DzD%qzG)uK=X(n4r#RWo{Ng&KoRldDd;-%|@~^kOohoUY`Ke#hJv`p0
z#Mx--_RZTQhDpw?IPS#S^K8rE@GFI1#E1i(8CC>}z0Z<Ru&X48VNz0<^MCS^h_?Q)
zS8`I0?7gd=)b`=C-t#s~872eD9be-gJ1MPQC4@9x1g&@Pgj>s2t!k_=oM03%$;Y36
zN=SkCVfiIzYkb=igK-y9Qn{KLdJ1#6e4bU0<jSyXr&JftDa{7vaIkv#@EW@o*31w;
z;nuf#8cpc4&mC`N(XVx!>K*|!(aAnL-`!pX8?2wCCLzX`eOc!1WnQq*<S~Wd6vnhX
zeQ>drmg<`>88Gt7*_WFYYu=jhICF9LT<HwD$DF++8v53aRe9HTP9lDd{T#-Q=;}K?
zi)2anc<Y;kH+<k1J%W4;<jVY+^W~*t)T&vgbU?CepM9inWZ5?J$YCUh=W(oR<~hzR
z2U%H=xq2hy_>gzjt{>CV%ww=^($B$X|J*y%Ol-P2Ty*hM0s(>{ZY$i{sI%yFt6E#F
zJy-g=ch@@U0{114G0JWs6js>C3~ao~^hyukh1YM?9#?dpoMDAD2u$%B@8V0%jj9q+
zzLim7*6Um1r?J;QXKfVHa9c<Cn^AuV9Phnr@J-FB@B2B8xl@lnIZQ+k#C=|koZH)3
zzrM#{-kR0kUF~=wD}#ss<SgCZo;E5GkUM)2=rf>8W$@<ITQSvMV$BEc{mlA3KkPXH
zbI|q<)O9|J*MvXw=vcy9;LcjE2Rpae*iAkRj77BLrmp&4$V#+53Cs|~FI8HLidQAI
z-}dLu-y9?ij(OFd_%yt8Y(NUI$11024}e|O{a<GJH^I%{uVZz3<T(Z#p^EKihPaKa
zmUmX}f}4|yA17uwQmeD4{y9A<<_>_S3FYM{9T#vdw@n>w+v$Aq#xHiN|6DUgzce83
z@J?Gf-6}5p#w13+44VW3vMhh(6`S&NLCHx9-KG|p|L7a}zG{1}*Swm|KF4^MjZ_=v
zaIW|MkF=2d59Z=eDt*aN<&6k3ANc8%M4N?&Sur=qBmQBjjYLP=wqxd}LCNI3S8j~s
zqt-rQhz!omB0dtzdFKZA8@|R*pMNtrKG>uTD3pM`g<l-0s1Xbo1WHM10?0t<>$omh
zQE*(hkv|s`i^xSfM?I@HcN|L9C3{yUu*n>7RVe*bzMgt>@<VDA{G-23+Q13ylVYG1
zDf}*HN$#IL%5(k<uv5RzJl(WVDck>nh`P3V<IMG-196H*er??a)eOH!2K^u*sUm%2
z%QU8X(=J6X1Y<&dD%Q2x{(+RV=;rAX6V03D7hVun8L@<c#2M^#xo*KhzVDAh;y_83
zR2qQ_E%j5=dXpZ-kfpsX>N@4De9Vlq4LY0tXKaPG$;;&Ci9W)^nHcwXV`qErMn!SX
z`rkSRZRpErU@?O7ayR5N$GpBV;$A5>I*lMCS*}f(e#;L>MW^O#edsd9<asr#vz@cI
zj6XXp^bf=dZteFqzW;8sBbtYPZ9%#9;E{MLkLU?v^W;G!oRb!z8i;>8x$U4UWPg=c
zk!{gukDFJ_#)_<Af>X5anH^{As|pVw02MfFgwj2={4f{(Sig<MsE!4qyU!=S$sM7U
z3q8>4DO?*rQBe()fB>dQ7&9MCH$50Lr9LpXqhq<Yk>ZEe+3DIaG#OxtM;KG)IBWS!
z?(PDeH%X!AAo8*&3w|rpwHVVo>5R;Ld0-c3*Jf_bM5ISlM&<jJ_1!x$7-8q%y8uU+
zla&TuEMvS-ZS;Ca?tBK7fn=!Dl}OT82Fhoi&IoVEDq-{4#<|2dI?ua1MQ9sr>KJVL
zf8&ndQ2H|DudO@t!Vx2PndMV}kZ|j55^_<M4$KKZ=s;ojkx4m&mKssyhRUb>z&vxx
zVWO^z>a4n0ikQ1`fg-yZMQb5IDs<qqtlz&oKQBzo{O!YyqB@hQ{B!Twd-7CR34c!&
z|0UlY1AtC8`5aD<D<|vc#ew)MtKsdaLOW9FQndTyn&hfpeAzlbp3*tSaqpv?SAwM1
zuaV)J>!f1(UxBR#OT)W-Ln2Ewb*jiWO<YCVL(2_Q)}!~T1ux;_-x^b<yPh0;Ukc#!
zp_h!y3;I4gcizFN4E)Z$6`P@7-<q|sq129$!{&dcY!>KT;#Yln;m7A@-TN7ISDY;>
zkf#gae4uVA_e6!iO&0Pn1GukujbOidf|H^<z@eK=SGF=#;iQkb%I4~oQq?iaQts?C
zZ`=8EImtVCd&Sg8hcOq)O@wt9-sy@Zjma3m7h;(HJb$BKtTr{;B@!;~SM3?;#Cm&4
zC*9?*z2+Kf?9M?WuVE1zIHw!lBp=0DPB~28vV1u2Gh=Bqk#Z2AK+8HV`hLb=RDk^4
z!OYu?^1U7YxtLRr2jg;=0&Q_POv!JM+Ya_rYPVEygJ-<wv;lMfLt-C`;0T9vkn#P_
zSKY$T4Q_?K;(4j7@)+fkQK$7Pdv7ka<kghUXF|n53YU$BTP)J6Ks3a3dGpR=ZH>H@
zStr}O2(#&ol6A@Y_@4X}B-vxQozUU@?!=E|(SYyWNytZq3Z}lg3e|{VgsyE$8pE_*
z7c%x9WLZ6C@9!<DP9)o0^$+@Z^$lNskDj2OF^C`IPdTO7ku5k4tynH;(g+#|hYP($
zmVZzC<~4BI&dkQT{;CqE9y7*R@qWm>iQ9jAy`4dC++*BcDEXTbWi4v0{xZG0+Z%HZ
zQ{xh)RC4Df1-j+^&maYAW!>%TC0W)FQ)}=&d#>`YweNtnWufQ4u?x0%SB7>;c}MxR
zKX~ES?r#(#0rK3P>(+Mhzc52)0q0^0FD2-T&G&;dfV8NM{p=pl%bGg6HYR$$um5Hz
zqDO!{JaWm_JN}eN3%R4nbM)D65~zq_?1(8e8k0VwDgSeskH7Qsx!1WGUV!00kiLoD
zWw~EQE>PJ!`ZG!iO3^kob@obx?A%3WPbp{U|JHbD!BRHoNaI%qavWC_$-eRWyq9I@
zebe$xSHU>4ARGU+B<t7HN;#tj^XFB+?3H$QX+00{x3#8Xwmdg)&ffTSVJ=2Io8HL8
zEVE`OMWy0O9s=P%<A2)2Oh)Bzq35nh4i1@$<>#$OF{<!WkHr#+{i~_G@y0m;7QtM2
zV<SS}S7uQ_NVTNhCfeA@wJ((oy`Oor-BQNkW#gI0z=MnSqfIPR++EWgaFUI(0{9iz
z)*pHQl$u>OASZXmbdaUCGlGA))-s*qAo}H3O?O5;%-?QKzvp&jzU{`&;?R?CV|ARF
z1s#9TG3rL`p16`ZE~t^9pfv|j`GO8qgBr^m$3C<Rf(Wmbx1Ifs;^tPUrW$~or_4@%
zy$xQJcz~bn25r2EXnjptBZy`z8affr!NMU74Ezs*r+ceGp{L~Dz*Ece`?rv7wsoHZ
zEN@&rKJJnjz#3vt?~h8(i574_ylUz|b>rH`{exTUTfgIsthvG<c?C+a8?K9Kw*W=}
zz3GZEd@OU{zW{uyU*0~vj@Gn5o80Vw{KKA0#Ph@0XOT{&W9k^^Rt2&k!TA2*DX(+0
zsSLK}4!8L3)u@Uj-83q(((n1EQ6P8KWmHunVZY-7L6!ls-BQY78>qogIXc9}E``<G
zjdz8tP6@X%P3gQb{x*D)z{g1eg}O7!F$}(`wK%VtUq)D?n2M@++$B<P@Pxfz<2Scl
zXpVbIeUs@cT`^OS8dj1r_(tHjH_et+B?^vM(cG4(_jf@ljKZq@J6u2L6s|p3y<LFl
zNDx49EH^K``62oAM*ECeNEzFb4hfkX&{D4RA*a|qNNJuV=rS^WEEU}fBnKK!A>Z+1
z+~Dx{f58Dy0XgIf`6z=}L8hxvrb|7$^Vy5;)!Z4~4;Bxkdih@(ya(y+d6mhRQ8mt9
zhrQ2kj6car4^`3ke1ROo7ZTQ~WZEy<1%6DpIjzav2cu^_361SBU}l9erNuB!mgE@8
zn=b1tM*nJuFU4LN-PK*#-xVd4dMqi%y!_KEK(`0YXYz_pU-9K@Ht!^_UJCTbmAt+2
zzR^ZeAZ}ji>azao<QUTlKP;($aMpXwC05L(eQMM?tJ<OcCT@^42j8z=g0LQ7ip*c8
zeczBPDnq(%{~V{eOvBt6a|%Q6ykBkWd0D`@;+IvOrm!RTeqMbZ2}#YD&**D^>j4x2
z;IkmAKcys@?}Oj;=|2p;Z)=&+{q@tvPQajKk<F;<jVv2b15d5^>??kl{O|5V@E4&S
z_7)R1-ewLve{E<(_W<e<xNBJDs7T!$T%P(51!RwhIW|8UCl6JONbAmBzRO=4d@pXE
zi|#DqvZ9_l$L6oV;T}N%`Ko*s(vVfoES|!&?lYssyYxP1pBFMJCuHNpBD3sY7FBn;
zyn2ZJHuUk;*`6opqGLr{e+g6_SmN%lFR5SOtsmz}WPR#p5rnkm-`5!zw=;m6Hgd!M
zlKZHYRxoa)yAjdfNnLS$FTPTLilS0ybSv$WeqB;!D>DiCwZFZPBnV;=_Xm9^rKXz@
z=<ZX*csRr1W$j;Usv;q~-!%%-J{c(13^&xAL2Zzee`e7)vhJ{y`6;HldRgj1mu}TI
zhohlz>5rM!y8-9k=Z4TSS)6%qlQH|?IM&=E0mE;GGF@_yL9_?>DF#&2y&650(`aGv
zw*HJA<CaqB8;`O^(bCG-HZh<+E`V2D*{kIf)Vo5gU;~Q7oPz$^A`*FKz3S}a16UIT
zenSF^JoV_we+jR1qMIF=E>MCT_gdhrrm{N+nLak3lD+}7drWe0zwEi9o8L(bFa=yK
z728z0`rPq0kae;5@*5iRpZrZi?W^Db_CwD85haV8yMKdNhFtdcGRv3c%i3=Rn{K!U
zQ@m(YJp4=uPYs|IS2bSB;`d<oGlyUD?C>1mt}JKUI;jU;VLX6UHe-!bULoCY=u+HT
z5hx8l8!N9nH@pS{tmqOb*@1NJF>U>LHHdy5fm`JFF!R$SBdP>#y1b&1u@aUu*C_`0
z-+epV_%Hs46afC+VW&YwB&aOe*|f}&Lhx&UnAB&+0%@k_id-&3YDLwNZLBUY@=vHO
zovU0>{y8ciV+E4>z%66Sj>Vs!QyyUR|N2_C2h>BGGFKlA%3IOBzMiKuclpc(m#|>m
zCCw`&lI|P>ALn#k45{}Phc0^V?&^1Eoa%DBf0>C%>_v{=HRkWOTeR?!YNRLn$KEsI
z9zMCNFX#Z^y2DOd)TJMRt5rU%Ef;mWJ~w*H)HKN-$`rTAb(5g%yDm{*x9!<V-%}v?
zmuG?k`GqyVfOhJ->M{+hs?{4%XpCh#3la44Jz*j-NiX!4@9X-vtS?wL=-G1Q7gy5J
z58Wb6eE-}%z&>Coj!^$6X%O`E$-!t)#i=9T{CwO_9wbRy)9>4wA*1CA`hyXg1M_`4
zEOJ-D`sIkU$rFU~x5|bxwbd|eO68m?G<mUb=(}@=;rFCS>@CW}P2L@?;g6QTwfGS)
zO9JU1ffA+?<>>FVw|M<yf^KSGz;rUrYm2Hs-gO9Z@!G?K^&wifVzQoGwi}%CQZgzh
zUAG1%C4exK&oghV5v2WE?LLG2cTS~-c8lEGzldVC2h9ihJ^)q1T?0s4t2y<4KiBE@
zd=z#{H}jo+q}ARiOac6@lUH+}*EUP^{05kdvkY;^1R2T&=s=vmV-frG0yW~6rTJF4
zx_3L{aNz5-1C<7rI=-1DdrG>NTp%@A>FJ*4;7sywtZ8&fUP}%qoyzWXotVCu+S&uG
zt(CS5zF!x#BKkKfLLs6#;3Hx1&AAAAMcRIf2d7<6r<{4~#{fHL&K+lXqhU~?y(&);
z<laBtja$#-zn+Rzt4<lK>!i8mdAWn*m&8kwW=*K<2t&YJr}V?6Ig+x+hf8kINz}en
zB&qomCUeXXMO+}Fy_hdBi3_%-c9F_SyQ9V~&x_YfoOKeu<SoTNE4^$*O~~kTGiWh;
zxgLA5#^)wLwzaev`5wy!pd&!$z?P^^O#3^w7-Z$@5ZAj}+`H*}LGE->CXd_kD@=DG
zPL4EAx`oV%_Qky9h?$dpR^|O;acx+LBfBEJDgUTkwdbsru@m*oN$}bAl-RoT<8!eL
zi)Rg+aX$5nMYT9wMMdpEDlE##+L&6Ey)XFw<=WZp<KPdmfx5ItFj8CMxj&*r)znd$
zjtr-eB%mMgD%+ASNueCJ6-V1>Kdvg_xA6p2J7f1)nl)up$&+#JYu;?iAiHF>+!c1R
zj5A?{ff<nU<)5DvNI%-7FP>nu=+Ze1o`4B7n@S6w%w0BjrgJ)nFRPyE8L?@1JZih>
z512mVZ0!iTgs;bnkKQx3Vf@k;@Xl@gDa_j?W6{_+?N?fUOu4IFT*p*KC-tpgua!zn
z&sa$PQ_ikX4|g9ysi%&K`BP5tSO7XA2H!o>W%D9Zq<TH$&=l~49^tRoHTEi(Rzm8C
zKp{Tc9BdAes~hu<5~RupmO*_vyrUMwt2piOY>y#st&qzHDl?T1j+?Oc`^i9pRMK=2
z8DFAKy`{~XUr0(gtej+k0R-geDo=sq<pj!(Wl2Y4M_+t>o-JYb=RLVl-pbx@eF1JM
zIOn;G0Hq~P6%9z7{sbuXwQ+0y-`?;l9y1a738WU6rkufn-DkSM<S8;T5T!t7*ElY6
zo7bj(=``Z<_t*{1(!<)^>w6IaLdKM<tn^ePjR`R+?Nd1g?O(AZ{_g+Ho@zgX)$9?W
zg%?)Gu)<`$>C~lu@TTYz|37TK1yq$=7dA?Rq$r(|N~5GwBBHcP_Xg?QbaxBV-GX#?
zvniDh3F$8B?)ulp^ZobUG43Ac49DTfUNz@4pZU!BuD8oI?tdk@6&Pm}z2q@xmwWu%
zAvLd6>h`6h1rAYVPnsD@--FVQuJ~5Uv6#~1te>mA9uGei23{YIG29+Bs_=W5rW&$j
z5<_3anSeUqA%MffGstYew~&G6c$U1w28)18`+woS`AVkX0mS(UHrGyTXwt@_MH}Wr
zBU`raKZ_PWZ@_<Ix@Dus2J^;NgsiIr!Vl2rP4+ebRPH1i-3g6yXw>UULhDWQ*t=Vi
zb*I(8oBf?RvtNvB^~Qw$8$c)ne7#ksGTABnHCE9rcBc9}h&XQywQv62<I6OOMmKR6
zMT(~;p`t=7QzwjAy%q!7MF$<}gGav(c!2Qo;pBR*Nd}?Btir6JYaCQt`0iZL>3mI4
z`(FQSvp*4!{5Q-DB~XgcJ2{6X=XG$BjA#ms>}Vd4$hlZ)SYr0gG@_4uIiBgA`oCzh
zn@FqI@iG8i$b9<Sa%Yt~&8L`SO@LYqMbPu>-C`c?>oMB+QXZ1Owq5INQ#&LQ{|j(k
zlt4De3)9Wff_=hDT7RLv_0VS){UWfH-8Vl>ZsQ>SOnKi-4N%cCpvyhQ!-N0f9V{J1
z52`1+3CE*9phzgX?LF#&wQ|t?qj^^A>ru24y?+|0$HU0M{Qo(mWjGSVx%F$+qc({5
zAMqf8NzHeUa?O~w<?t-Iv}7-W&4G8=yu=#hnXLT&gX=w-vz@2NGTlpx&aBxlf6t#=
zhuaIi{q&}q$ZFRnHC|h5Y3zF;{R4PzWgZ<yolCe_Q^N2M;SU!<Qi0xE_T<46(=2aj
z8DiqK_#R47E2-D@fL)JDVC|P{s&NpmDd9Aj+Sx4#tP0dpU;JzerT$SR@M&;N=m$Va
zd~o}VTxzimSP$4ZPA_;C4usO#|AbPvXP9mVu0c~K3tJpHCm4w_UP@BAZmFW2GgY~&
zZ#}{-lAY4T2P%OGE_f>;0)+Ra*0dKlgv5`Gt2@#PT<boal)*M^fjkEAi{HV<7fWRm
zMO3DKWhUj)p5n3CIMKgE!I>HilPvM1C?^9t+V}U-=V7LK-7Z;QEyZHrE_6<8<iW~e
zk&^B?Ed$}p9}h>ConLv#R1*yyjC`5be)M!`^DQfDHNU+>G=Sm(f%QUA-nP+pd7@3i
zm~{Cc5Q|(p3rI%2#q$;XD2en5hgaLb{ti3T_H|5uwJHP2HNjU;WX!A99y3pIOVMWh
zA>()FqCW)I3*7c{ZJUla9|4sDaPTFnCOuNG<pkhR+n#xSdf~@M2f2aHhx<=}RU&{w
z_Q>eye{>Y&zB{z0Vn-BxHzH*R<U!M<?&6al1rZ+Kkt?Rj3Am4~ve@*xiF|hFLeOz8
z?iXpXiiO@;#r2MpoqNcX&eh2vAHD>&C!`O8uUgZEx!QrLObRIg;Z)`^oIpIe-?vCA
zkhs+Z()$b`vEDr-N(YtY3bKs}ch2LNBzLb?3|pCvJ;irhmu$#If<S%N<6`sC-SG_z
z;0$1mr{?=tx;Y+No%Hb|#B%?PO6V-{lXs~(c^jHJ40rraBrn~IzUyyc&t%f!4WnY(
zbkgBisd~svzA2kd<=XPq^IVS6^U^`c5=bFFTXZ}Pc0E%*oy(tpf2QKJxsNa?#M8Hk
zd~+OYj$nWDuiSF{kaCJ~S_0`Us0yX30nJbQJ<=}5N-d+CtOjV44ip^xC+-01GwA*~
zrPM<=;8r*gkdhNa35pZ#{bU5ugtD4NTj!d#u?xnNcM_f9#*V{cd-tj(uoE%J>wkqZ
zJ04V6g*(Z+%cm1wYt9&E`Za%O9*k74IOepKZkY6_tXR(0u|`WxX3IM~$MFSn3!Hj8
zwSNDDVZ!|;Gmq~nJiY@(raYSb`Y@JZK)XPf1Z~VkqL6ZtpRvZ^+4VoW+V~8rd^-$9
zZ+0va?$Gm^6Bn5v$W4+YSD4uDL`~M5T05m(T5DJM2xBf?pvYyoo(v+#K)9!Ms=|7~
z=q94j%&8Y5kPNlB#D#B<Ie`OxHmqnpmk8VV;=@e+kEAplRDMH(Oe%k-zFsZvn1&(W
z(k7-J^4NZKMq|7s3J&)gDZL@ASBVefr@__(wH}Z1pY1lu(kkRlHg^7S3;_vxs+<hR
z{Jsnwj;Es>@x_n#@6XRiIsobYY|h`=r|(jiWs#haljHG^uvf^$;#M@15Yaz>D{Blv
z^dt<bsE?yZ{A~4}riHwSf+XiubwykA6*hA3f64%;>kMqr(uq;U3Zsv30!IPvIrE*a
zg|ERw#C+A)V8i7fXhn$wHz#kfcA7@2GNJ)3#*_~xva5Q>Hx+=J-_(_>F~%8uzNd7c
z^svyCf#@(uz(izz{RHNmpjIjer**IJ<4btYS6QT+9Nc0-D+7!g>c8}{PXqJ6ewd2$
zdR1V-0(`j85P_jYe;vx8%KA8dlsEf~Pxx>QMFB3(JiV{Cuh?#k?kxis?u!&X-P_R9
zAL@YXLF&l(XI8(VBn+%j*_+$T69bTkCsMF1ykZPBlxU;`Ifcuce<r5y_X=QRFxe14
z7S_Da2}D$}cUOJegX99~t2>SjSJEM&v)5**7pKg=Z-R6sj<!HePcO7+f9;F?8c|Z!
z)Ta;BhiDt3P|&FYLgc)gM-udU6*mBVWJ@`FdsxNc`0YQ%YjPkR-aAl3fc~(gW0ON?
zG5wo;Y@gf(87>IM-7SH`x>n34D&;|fb<#2^KypW5Evb5Wc}m97Qr@nyXKuCjITR?Q
zPyX(R%{>wNTdZm0v>oXIyBHTZAFccj+BB;6`t(12p#KLb77}3dAi6l}D#NXgI}Aaa
zlUgA{5pow{9;i0aI)qg%w9{{!@gH`Y%ijEP-kBgPrP5m58A;Ymg@Uq^U!-Mhm<1<q
z$}-2DTNG%G4NRcuMO=6VZ$uD^{EyT>fcqr_z8#V7j#1fpYBAINv+t$c#8T+G*U7-5
zQxNSzkmTA+y)5QIl1rOhiTz#(gy&yiB^6~enIpePOaJp1&~L+}B>ay?nAZ*h#$%v?
z#hY=<%j!nX3-PT>{OzBx5Z2xW$%%%|5vV2wJ)SI?+QAT!Fw+Oh-C;5Hz@cw?zxAS2
zs=5%Ms&+@$3q66LJXBJ6N%n6uu4KVxKtGZD7^q%B%FcJpKU^eH#ZU)+xkFfe+PXUu
zQ!0VR%3h0@_Xk;9sG%J+t{kj&<HRXXRZT^<&$HDSaB?pkb{oCFxGLRU197p$3~BS~
zA8ypAgCjkSqkqWr%^Gv<4v`lUSeH5$Pwo`Wc}hrf$?CMM(=eqM`8AIZiLCz^EB7p*
z1=>_Suw;`QMvcD}hMDEwvP|a;J5GFX@*lF!7=aDA3PIkF*Ewh`JJy2hj6_`MvEi-f
z^KT9i_z5|WSq5p<j^_3q0YDsyq^-UeNnal}_jM7A#p9dT{|5Ufl}BFPs>yhl>ie{y
zjg9NxQjeI-)Awyk<ASXB7K`MDyi3+cY3Mdse}cNe?R<86V-(u>J%65DIB!p7@n7ug
zngxA16*Q)zM}n+?I<}zb@-&(~7kWnNQx1K-N@{i4T2|djOL_(Z#cDOLh_i|J-R%-Z
zE{CGpJT``pszM7NyDz5aKPcKIMTJG-NMhV#7TvuPYASXX?ofa+I_a5&?{&pLsYNaV
zeRWx&U}%TogCY{d?_AS!cU3HXy%a01Jnr5uQ9C=|NXz^mR@Z}GEGY*+d*d`;e`wof
z>5~EjM>a#;o)4|(<HL9M=+!N>T3u&!L@L8Qf$Bml@ulnsgB+Tx=lS-Vp&#w>?k<PV
z5f}F+tZ>Jb*6Z)>u8SJJQaaoxnt@}G)^-umE#`+tC)wz`7+*a|-`!%~2Y_-3;=(^j
zyk^1(M<Eusy|wIz9qC#|y7l5Dn-R61{WaJDXqv(o+s{1j#%3ud6n@#uzW(1vqJ!*o
z^3k7f_6#^~<JvxXX^Sf9SnlB+KjuX*o|gsdY^ubIyR>b)AtTI(zjV*|<>`L=h@-h7
zt0(X*WB;G{ywU-P?_?B6MEt7L+m;+h_T!%wa<tg>IR540ycX};Lu<?uBHEzYQ@Pmx
zBS1d7>CH@i>xGBuX~Y=f;&tz{S?hplZa=tpEe~{F$e>3&bVCOenw^PeXgJMx$|dkz
zpsO7{NrNrd3(WI(8x){8S0xY$z6ha{7iexr*@H6s)GfCUqxI^O?_~d59lY@(O6cCZ
zJW1oUCGc#deTrZxZ5x5!A6h5IGs@8K54V$zz4pNOxR!YI>?)%qi$AP;!YYiz&BdqX
zs)wEN-8}g%k>zdCCHlR~Zd?vZz?G9Il+1Tz$Lzv*TP;3im};S~v5BSb9bwf3l)FDZ
zp`m3uxtV)4{UR@hGR?`(tL1zYrO_xDDCUV|gIv7VmJf^m+rNfLBuIA-zn7H|j%(>1
zpRtx7@P=N_)I-Wk!LK(0R-JA6#^K^CXL);PH6<lrcKhFY-&0X9DGY{H=oNhX-?v;<
zMu6`rDnm<kR=uPuy2iuc?b?pvAX+if**4=mUSGiW-hh7(*va|Q_IbRE0CfL<wXk8j
zbWdQnh79;9TV)q$tk+zlA4Hdo9EFrg(S*_T=Dn6Y%4}Y>%D%mQ=HF6#uogIv=P7dW
z`;#Ff7FZ4m`9S{BKTD{%Nd!G!`s$-TYtE|IyhV4V0ME#HR~$R`iHF5znykyDr}XeI
zh6rxweq|Gv)x7eg7MU1Kzw71~$i)H~2np$NLYfob^$yDxAIH66&?BeYf~n;&(44cB
zUDC2PBNR)M?jEA{3C7^I{f*DKQ?o*|_eu_$i39M(VkUfd=v(M^must;W-+w;_V3!A
zaJ(+E=D5B-XD4iy;dGbzXggXAR0@ao{9Cd!LrutBTJOivpN9JZv3ZsPOnP2A*R8`e
zfLPB*07)#}|96Ye47MujPI=Hw<b67gi0KL42)<6+cN`pe1ddejWa0^6Rr+T#55}J(
zf(n(U9OPdgY1b*JoL6w1(iDNV3Qa)Xs0JP4OBb^!VaxLpxo$fifV@=(%@BewJQ|iE
zyTw)1bdV#Qc&W7lD#wFDFMnj*5cd46+7<oJo(1;@IY8FU@<SMuWSHhxsl|Zhg|QXU
z9?v0iCjYM&z!UoyEH0qML2V3|)@Hn!?;6ysjvz4mNa~yGS0Ib_d7^xN<pE8TaP{nE
zNZvR=x;l#iIpST*15a7N{7-K0d32ZiZC_@qN&_fPlp$?#bS%L;YnS8#Ac?#nf*dlC
z+zQs|>8w@{(%KAy4jAAd`)5S#<4xqZa$nb+jo*K1Zj=(d`tk4s6_vg@zvI#T<`Cb#
z1J!e+D}W;%$1od8^BqfX?fH;ViSTw(D_ki6_H-C|rYGD&N#Ggl!${XXJpUGv>q&c}
zI1Nzh+CTGUpUn{+;36gMDF>x2kf_6x!9+SSnXu-opIYTQCYS%ra2F7OVWB9_ox@^E
z->o|w(@q*oLnl#2eqYXWNs=40XWJu)mb5nWvlyQ4B5Qp=$vB$$J4+!DKx@UD@tF~B
z^;B$-LAo4KOG35pbgd)ec{9nB=5CV!lP5d>59%WuCA~}S;`35Z&)DxC#%X&E*uM~3
zHLLsh3S$}K+rtHS7b=qZaMgM2;dI|1#{s2_xex2-CuY5x8piTaKahbJITG%9N<J<2
z>~e<Fmo)xkXa}V8Im!0fvCcjH>)P==1-iagC&c&91^2$z9ig96g}avq>1DcadFdts
zAO0j&{N&jKPl$(mBB*151m@unOiz)&I3y<LTdx>FYa3|vIB(B=dOq`%k%Y?aETlQH
zJClm_-#Jtdk^^Jf`XP+6^-%~jWIo_h66BB`u~Q4OZC%80uQC_d!|_A}WedT(Wi^k(
zVfzwIGx=h4FvRw1TGg#CVD`kqLT&S)KjQKCJ*rDy-lz_H83Z5-a^naY-{*UWP!SD~
zsFp<2o3u_hU`x*jR;T;6ApMvOMJS5GwR#*WqpIm~hQ;gO;`r+YP7gyevg_sfsO@)(
ziIL6;iLir63eb{zDRtWKxz%vlk9R+FNf0b{WBeW#iSAecs8b9Fw0PgA2G@%G9OdgR
zA!w|jDa+Vk$Df4V+Mi!dl`Fw{arPuyydgPjX^lxW9cnA}8oSQxfy#tL*8TF_0C0(w
zSGjKZmvX_!-<^L2%%`HT6I#gL-TT2YCorbwOrrCUvO@*c3;Zg>?o&idT!PbV1sJj?
z(ryrZ5q)`o=8D}_HBD@fcX+hjZDT=OoMw~y`lURm3=5shLS=6Gd%3%u8NYYDbdV*U
z<k4Em??iK6>v{s3arp|VHQ^cZ@jbVZ-~dQ8%kt$(k?1lut1upUslHoJd@=G&nt!<`
zx<9f`Ji%qK#M&r1;(>CnTrs}YVIN~Dh1TkiM4SQ5JZp;#$!7dV0pN&6lV7&l!#fEC
zaxu44*lnyJ{_ANJOGCGqt4oc9uiXvC5bv$b?#2iOEOG2Jbvw?48x9c+2WalQ`d50*
zTnNaXm~<LrxDKS)vd}LQyw%hA9uYY$z$iYXPOtP-qp7SV6+ti0J%I0b%rcY}Rb4rL
z<{mtL>kZo3fx`raQ;~jl$+S9}H~r_DS<)iJaBCpu0S&t>BHjSO)?X!-I&|&%Ee>`J
z8G&v9mQr`3$Ud`Lx~bIT%Wd@M;Lw^J%wypSc&wT10Hmb6Vy^AXet$a9fNbM<Z{Mpk
zET2|k*_p~CukU?0wzUqF+CAHDY1|qc?Mg^U=iUglUISlkq4%+RYzhSgCBub#c?|d6
zt1%6a7ecMY*CZ2!N&-O=DqogY&Rjf8*|?o=&f#gtD#&U<uQgHJ*p(}}<3rZoR{LSh
zARz&1x7+Atx3v$yNU5C8gylDEH&`b5?sglJQ0}#+?!GGzu%VALiBQ19<a@(dLxS-K
zrs$7TL`ly1zq&oZyc!XPzz2+<nmC5uYMP<+anz`(nm(_cSPES>yIyl-0f1`aNGp3c
zx0x*3%K0R{5)`La^giGmtt=vfI>G-)L`vhup2fYhS>3O5`a}qZ1la)fhfm04xw?yx
zhGhi4j({0Ck_%ewTT&-IkY0_`4xcQ6G=5}m-t>REDpCA2XkS~6FJ-m9e8ZD^t90qQ
zRZHAzEfavQ1@EA+{ea8G8wJu=JvVL~Z~KQSaBAAc>+zwqRmJ5g&EFd283Zo$?Dx9c
zEFYc8ilRU#^+vJum1{dhBnZ?jX0bgv`gxrWRIr+yKc|8wD-v+wl|{1}3XoV8M@9e5
zJ<$md{%t+3cvc!0)EkJFk_kIJiKu^FDc;?oU8bQm;>pNre35`2-Z<i|rldmWuAL2L
z*Dt9155eOgnyJ!-)W>Lld7%Q-hfnyu$|F;#IN?|kc?A-U<G3@bg_>jcyj$FMGW|=v
z7Svr%>}#g4*9-)wKC#j)Km<YYf`|5qfVOIvu0m+Zq3_+xPK{hXaL*-Z?eojP^=vHW
zi_SuCds!n1nKT6Y$-2nPHkB66juqA0wH6k5$<3XH34RDRgY(~6_G8m*j}U6-+k3(@
zZgGF~o0oR?WsB|&t$(}ITL@(VEt1aEC+}W=2fel)4-W?Fndh%IC^%YReV}CJ!u8`!
zFA4><wboYp#jx)zrgKOJVmQ#d&WF=w4?*#RiFm%5qhAK5ftumFaL-<Lserw?T2d_h
zX+S9y@vnwBh*tX|8d3#c>He)8x?^>E3<5PM+flck(f#4`P4<Lnn;ij>0N~;;&;bv6
zW6R2Br0~2#xDnmZ9hq2rzsWnm<6gRpWla7-`xO<63iA}E*L}WZd3I9^Y)sNRA^;Z3
zRd!4{;e^_5dMY9}ee>}G<Ez_qp}u}n1hK&NpddF@u*Q@$#P|OXN*!M?j}s*Uc2)-W
zV!hAJz75d=fA%O*_+G{t8D_RBfFcqQOB@DctSFDxTJ~59G2g(KxFXgvue`t_#BxeA
z<JHfnb07d(!K?+ALn>Oy0bP3!!`S7YVDJYWi?tcEVf()4YxAWb1E1C_Y;v)S1(GN5
zC{UaI1kJlU3!9HX3ey7}fAH477c0-XJ}m$eA!w^Fq1AftMzx2x>cDEM@*bp$t6se<
zy2emfd7-`YdTM%0HV3^9#~p3vQyC^)4%!|1mE(D|u1*J<-kIS-#b`aV$~u{5qFDuO
zBlo7orEYg3a0jX6xc)eH?xt~T)8i?8gl3TSAJ@?Pl;QSQmR<W+O|?M~m%e&R4Tv4+
zf)x1g9e*9&<`E%&eaO^Kki+_~Ht2J}+klsX$4XSyj^M>Qc%R?4TeQPd`J0Yemt{;6
zv=v(-5GvuQONWk5LV0TGfb`)*lPgJzNv4So+!r>Q&-y{YU;jyo!I$BI8Rv;IHcm6~
zWZI}Q*~AmNYC>|06a$(?5l6;6LV{+HVbrtN^`OzOC}*TqG{I`^O)T}bd74UCNm<!w
z&SWq~yJ}$TY_UKZVu{&{7YTYUw1XgP0wQCY`)7C|-~}&0C+NNOi&6&13p+kk7%jpl
ztZM_+JD#HJhe^Z@-v`GjeQwuvR-`{plnO(DxAy&Kpr!v;?|_#<ShiBR5peA=X0#fw
z6471`>gzwl@p}61Nbve&A_y@{t#iVcQhRdO1$e4(V}U1B_&`R@WV9n1W}lBu&zgn3
zCH3?dh?tL*yTH9wQ{T~vhx-Gb1YC!Qxk3R*9sI`=V(S@gv1Ex<rbah27`^U1Pj|U_
zORk!(ZlME3E7*?K@he(U29wH7zce;~gw6`04wFl0h~RwuhA}nzIv9y9m=QV!ll4Yo
zx@2Yxu9A%;s{MTmiyrl$$hH2gp}iy?#nwG#FiF7lWjZ3oec-g|d=T3x`mJQlTPhEc
zR<9H;u9d#@mA;oE+O}GSC-TpjC~%p`%3~x{H}{&YL>3m?{62pwbV&1P1y2>K#$UCf
zz+U_BNu6dY^3(J;J;|Pr*?w7k<pIk`*jN`hZpngmW?MzkSR$0YV>O5?^yGS}5g5k4
zS=?;<%I34A(ZH&`3vKSCo#g5Z!q&pu!)M(UZ=@LLrmQvMsoZs<EnDiPn`b{=B7(=l
zb=>#19-SlEH&y88W^cDiiOiqT8%e6~@hB^6Zq_N0U3+H1o@V_;0QW={JwNeXpx@}L
zZ!%nyN)l8V+IEw?!hMdho7AGIe2EC%3^8-d5)(^))L<f*ALwwl#npJpJ<&3Qm*DUa
zzN$hDH*Xgh)a5L0Kne?*@CWzeG2-8K@~Czurg|;1^URGca`KoWKT=0r1XbH4^tRNO
z$>a5jnTG4qNeMk$v3!GO^KDf`T)aEfC9{ROZ)ljvsz)dBVa5VMX}bvx3cc%e_|kDd
z&EU7%Tef<zdUFi@aO`_nyOhO!$7OO$n9oUbv1)(Y+Ad%9VH5U&>|BYdx;N$NF{&5O
z;E~(Q0h^GFFZpgeM=}Gy8>#!Y{8e)7_Wx$C5yzpoUBEDWN`K;3^+_AOr!DILo_e|M
zj@;n-zVy=NGud{ynf|kk@<m+xF(rh~%(zDn=G4H(dTo`pNG)JhEeb?M$5v2Fa^119
z!B{4b)X&6bHpS*;2NmDUdvq_c<s+17^Wga;CvupA<;_islzw2eO-f|h=$jO5_9{J*
ztM0R^D^Yygk^G2^Ov2cSGsB|RVPkO?dCJ<g7Oc9o)?MgRKs$KUd0}ch85||?0uckB
z(dD_^9fav9<7Lb4-6oy9`P)rT`Ke=ZrgEiVyOBX+1SHzz%GI%cav@vZQpj%l$h7Oa
z-KC7&V+H5bWTt1ywk!NqK6N=9^)G0wxIe4E!Ny}eo*cyfE1ga~vcs245Tahag=-<j
zo$`T!uF4PWcx&uVAbjw}H3mKwHrUhDI}!QQ*9eyvQkde|KZs05N+oX~oHuTCOyFJv
zjBoxwSL5C$rWEHNiKiyB8(U@+UHC`gQmE;Z%!E2I#gRT7AAdYDVHL=G(#az{lX11T
zF)r44W#1b|fXE3(D=~*ew-VQ(@NvWrNFRyB8$zNmbCHJzRbq!63%ONs+yiAqAX5Se
zuXT9NkhFVVscWix$O;H(jfQ)?%D<bmm5W}->{+FZwFsA5*gB|S<)SFbWhwZb;dkk}
zOhzj6H%-9xBvH$~!P!>+cbCnAfZ&8GzKkpUcB#=*wg+zB()(H?gM6NEl5Xy@{Q0sE
zt2T}r=kVByar2Le^7mALd&(kE$R*U^A$Z}eU-un0s`k_rhCj7&x$CGprAirJsrMO>
zBewIX$&>jLxdkpz?Nl*WQ<lire=A_QHT)gVeAt4|eyn}54Uvn!ga!fBufPufn)RgL
zM^$cRQF=2UJv`Gp+^!KjG_Jjl*B3>rC4VTFf2WeQi=KUs0etm!+B!?J#_hTxqtVvT
zZ}H}8N4fWMtQ#{jpVo~+j*(%X1P)#hwgC_b#kJ1^OP9T|wrH%D#+#cirL46tWU0PZ
zPuu?bY^>f*_))nrKCrHbqweBJ5S{GzFoVaJhNn_o98!sR^sb^&jRVvP4F1vbHjO4b
zLsx1{!pv&VUBJ`ProrMWKZg&^v~5KeY)4SxQ*gx_Fmr&hPM>0Um7aw)XMODQtN$I8
zk1_Y+PyVdHRn_vT<;lqp?g{Ggu84e$T`Bfj>r7g3?bValSN<rlRs*Y}D2<4?x#*m`
z(wLVz0S(2jaRfWD4SUgOuxHs0lD~zSX1WI{5Qdb&kVlo`vAw<P8|Y2H94;CBI~Y-@
z=3<71a{SOZzSF8^egJ0F+Aw)Ww~xvmoy<qGx1?>d7AU_a*4>Y~;H@T~^?q}X&EhiZ
z!q`3`4&1q-$#{ua>Whr1RY&V|Wudhx^-i`oXb~4|pRtp|XXnzkD0xY=X7EgSQTL4?
z5@Z3h&N*LD&*fqOM!y;+q5YgB&|^_<WKMU#SAPDxJU@e)H84+v64geJR)uw}SA1Hd
zeQuH)KP!L_3;|!{ejck*zS46hj^XC(UmUUYTpHjiuuS7`9680yl0r_<7Kgp<k(^uA
zj*~59-7mp90CC9ufloYwnZ22BHvxcWa;uc@^zN`%l)ovOmWioafz0XMrt`<(Q)%V)
z7(_&H4P?@^D60V_8#`M|fT2y;+>MSlN7IXJ-}l5kEg$8}|8_eF7I;HBS@hM#{5oiU
z>gj^Ru{#T+nyoGT0H(7kC(iRKW@y1{<1q!mgX!U?;HVZZpX{TSoEwkv7E@y8W?w)g
zHiYCizOKX~?e{&5%tVe&gME<fPR+?JCFF?*bHFjMt+BaSMYNVy(|}uO{MEYMw&FXZ
zT&!2NHQy5JGlLmElyvTHCr!9V5QM64ZUmb;DZCMy{e*M;({5Up3CsmD4c(Hj&~p}{
zC9761@M&x`25;L$uo}BE05+qvZ$8q(cNNKf<x_PI(;t+AJ@!H;=8bKhJ(F^8p|zi-
z`x%f2F2T}t{Fzx;G0XpCciT?XXpp$7Oy7)MYk2x<?fBN)xql$uV46y))y#e`vqbjC
z&}&UFgHRmDzZtX04R^a9p*VbKt?okQ0nCdMKcWoEsevI_JqN>&*@Q42_islVYRAN#
zN)GrS{dO)>-<Up%;NJIdkeVkEaYSgvQDASxSsxrEQex*oZ?D{_5Eg9uc81qy_L;eK
z@*--WpMXaIB454|?6bgA(?~PINc-pk6WMdZcE_3F>??;8lPxECfE>Oz)}h#my!cx7
z9ZMus;9{R4!@J;CVpUYjIg}Rv$HRSCh6x(Za#7@7Mqi}<PG2QdZH#9POv{=DN)W@R
zkrdNEc^G#DNlj-n-Sc*CZidR`KD9|oW7z;ZGE!OPr+U;#^Fr6Ef-5t_yADLY&7ChM
zy~eq_jg249w*M-5LYGA{`NAD|Q*s@Fb4^%XeFPi8e9#k>8drgyJ((^c?b?6w#1Dok
zxfm1E(CY_lPLFMuWSU{}=l4QFD6*JLVe3?zTTM-4;jIcvkd8k&;v3xZJWUzj_>%JC
zrG~+D^zu3O7Y+2H2!aqbvcI8bcKjKNMv6Dh&;P&BH6y@3N&2av(wb3}_17(UodO0f
z0IuqD*a~6xu$sw1{P4Az{BF<+tJEQq&D+?BSC+A8E%YZsy3C%%e);F92s)`rkZ0XB
z@s3}T1AnBZGnu-S<N-J_sKBM!cj51fBz^hti)!$@drBoZmG8Q@J+{N=nBT}!u&el%
z&X<ymBB-_$GfX3N8p9tEq19!Qm=@r6<er43DDwyDNilhs@R&)C$s<J46~n}yxiF2a
z+ov*MDCO?p#{f8yzN^|Pv9h&@J&ms9v4`aW@a$)c0WLZGJDPOo?pg7d{HBVR?`xKO
ztpx!y@g46FjmVN<z}VdgX2P$Hk$T77#FUZD_q5ueNI@$<&!zD@89=_j9|Hq;d2nN-
z?ceE-8MI@^0FGiN?#F<jeDs%w)?Xm&JR#rhM5w!8Vlrw)zxsAsz$ph}eel@iIXe^U
zNuB)XkSsYS0-jZF0C9R>-qvNTn>ofmxUo|yrDsZZjOpdLqvMOVw_90>g5yZ1oB_9x
z%ILAV<cO`ffvqA+sW^;&`XE-j3j8yz-H9prJ1@IL|Lv^SCwuyKfF$-KPV#(ENmIML
z<xJ#NdZyT~55cQFyg<1UvXDd}UKgbjHB%Jhp^$W&5hK9tjEeew_bI&^rTpj0(W3~H
zy&>m9f(@PwJ`nnJkeP^P1JYV^XzRY&j)?d7#mDr&1Psz_PxLVchhdy6iYt}w{;N51
zQ<7)n)0AB4&|tIiQ>vZ0$y_$Q)L`-WUN6^%?`#=Xj>LOWz>f=?(%&p#X^P4H@@ool
z%}%72Ohjfx!Ik+-&NtTV{vko(b0xt>WD$voD?)3)s&h`Wn9}5Fj5aGrI*02kcgEZ^
zysVSOMO3(3>E;(zu7@OPJj>TST`3Qw`s&M89nh|oAM`HpmEGO!&6K-JX+Bly2nD7_
zT#VEsOt1mH6X-MmP@e;)W0dvNF{}PA5|BwtId9SxqF||mPx*V+`FBhs^W(u%k?bC)
z2nbb+&ajI5Uvsxc18Y27kl~g3rnnbjh9eA=f(?0icpSzqHF<7rv0J~N@<*WSn_~N4
z{r`IMNIn6$IhA;}1rF!W=ZlY<QzMAx;)qTq>X+Lv-3=1G_^K7Ajv4aLD&_eNn7zbg
zs?ZP?<a<_s3EypqYZx?IT4^ezF97@^;;I2$2jFRMjnJEgC)8CYFA1o&!wYMl769+1
z^L;Su8UDaOIk4x%PJArZ6w_ZJluna`8y(^G>oZD_GHkV0GL$}x9gvjXlLWD3w@jbI
z^ajtSP&oOm%-`fD5Y^Q0qgt}*{KF#}M1(prfaIS|oYOPRT>*OWYOZfzWD4`A{Fy^G
zNlkukL2J0)n<O2t$Xc(621QVokowWFDJ32g6gW1C|LKE@qDH4cwLz5vvb!~~OtQZr
zCRR@}BHR3;QbVsy;1Cb%sEPr3(+&Jsi030`G8JFZu{F_UL!u8m>7VAl_J{rU$TzFq
z*hp@W4#r7LgB_osf;Zl(_e|V_yuCxix5{-{LMy<`{&s&o^^E!RLs5G&g|732l&TgV
z56QrAJKznCS{&fmn;T)K3J}fifgqPgy(vJF;AnrQ<<7>&AfFl;_bnj~+)Z7SG428g
z$E!p0?%IB!h9w(DXzZ$HH;a9|CFqc+XRx2sEpT(Cih9JCXh9QV3IVJmdW{!<vgy0#
z9Jsk_(Ymh7JCjUoQDQ>g=)Ed@*^c7M+1Y;dKR;L9AtC6DK{^5BC2#GP&=ysn;^>Fb
z8=RsSHv7KG@r0VIZo&q!?W8TEGm~YqIo!eRZ*_};gV20TVO#cjE8%!Y`|Z)FQ!1IQ
zm8WVTh;nnO3<I81-$XZ_{eq;IeM~;Hg)c|7PhRCk0Z4&?8{|;|Ga>^<XGy>ZRTm@>
zj$89Iyu6GXNWxrb73_E``*-?Chv9gZM*J$v>Zyn)0>U5f&%c!>wW(etcTXn0f(+Gf
zWE!qXGZScm@!+wWlD{JF__JDe1<PPGMdx=Y=hPmpX}t?SK@@Fd#H+{nCnvvRHoxYv
z5dI!|{pM_Dvaqq{#8|(q+=ZX4TA6?#Hwr%hyla<t&5X^P!wT8ipj#~_*eL6GYfUjF
zqA3b#Y>8&d?DEpm<sBf5YLq!h2{3SrQ&FiF@nHP26VCL&ATMrsuRbN^%la%wk%aMb
z1yV4vs)TajoTE!o#T7f<NA&Y&5B#j1(MwISgTLdC7{WKDL{Tx7QWxj6v)Wc8e0Of{
zI~<(;;Cu^t$BKjdgB)ha{h0|L>UKsd`W!)Lw#5)6UwjdJ$-0&UJ~1K&Nb^4(We~Or
z2~xzqNreRdAmZvP-m{hO><6#Tn<7T=dd)<nMJU~CQP5M9>z{4YVGpwG%l-Y#IT6lM
z5MF@K4u=@@XG{02YQ1F7;8eEY81?J;B>g)m+loI+f<G(ltty}#sg#k@Xg{O4d1roG
z!L&XC7AQq;tc8Z4id^amVn5E>RTPAmjS6g6_F0y{W7ai(x3;|1<)u2~aVVa2eq**b
zheW<B=^7eBD;i)QHYuwe5t<p+-%>_}3DTsKZ`&p*ZhWmNPqoG5=|3}(Jy7!bEd9kV
z8^8wTm$FCJ(jdB7&ZUeBIi7B=^-E?)g<e&;3qAk2S@`vP+@%ZI{3;Qc(GPOuHHjaY
zJtPQLKLMOnw^z!t1P+rvKF|5jV|8$Cr`b+~85wG&e|z_E*5)y_fkcm7+xA&X_oTWx
z*$c;0w@=g*a%OwpRe+4ef*-0@kunEXmyMeV2hjIo0wHHK5~Hw^uE=<_Dw6~-lU~(a
zH1wrd8b{p|hfeMx;t?IsH{AojOcz0(B~BOGW~EYK{VmlX(c{+%MmVAlG7K009LX%m
zC^^HzQdz2nssCD{hcbF51?j}_*T=WPc9N-%5?bkz4SVsgETd>=<)6twfghLAVugej
z(<X`OPM@A^=*wM+s1LP$#{%L3lZq9Ff4PT4vH{Yv#E)aPnm0L9{8W_$;r@hQ{NPac
zFS~vh_^8_B7co}2LU&$6QbA;NYH@Q&qeAgsd&o~?4MYU9+Y?{FAbnH5ZY(xdL6UC1
zz<CQ6rd&4E;A|{7d2?<wmGO9|;e9Za*Mf@v@hu3Knlyt))uW1pGTEHG+ifi}Kdt#G
zq72y3yUG)?o%)XX!rrkm2++Y(8NlNKvVqhlUms55Kysn3F8`J`BhKQBlPcwxKUAFu
zs(St}Xt>I*UB(0ns#OmaYshP3drCl%v8eZ<E47YZ8_tf~d#=T&e=^B2|05Iq=4CpV
z@AsJ|AZE?rGc~_gcT-_b=Ob+OepI4(HOGD;%zPsF>}Nn9FIod4U>6JWvF-9#ADx(%
zjXk_YZC|tL<areJTA3}{XmwE(`?W*{ZG@%(Ld6|1w6F6f^-I}jFx!EkOIJM*GI+Qj
zLIr&=G&VO(&5eFX8sj8>x45LMC9gB`S^w4f$c${}DZXIo(;Xy506eukiq$)77A4y`
z(`$#tKUMmh*BO{*k6k_sq!GoDK|W}XkAKdYu@NDPH)k`jSUn5_jwwGmCjaX<P}x9k
z1laGEp`EPr|9Sx~Z8SgB?9{?`&)00b-PpCxtGn)w?^fgK{|<Hn*CShPj8ci)RvY<4
z12$pG><y$eONa8EpUmCd#JBw89Z&hqDF%R6m<dE6xm`dr!}aIjHr@sLz)0oVf*{0-
zJD%Na$y_^n{jf)$GPWn}5H<I(UYZ}fXyEtT<%<Y{f~<ixTZKuAnm3qAQ=i|`(bzk@
zjRA@QQ8lygh0IS|S}J8$(?o@PWLJDK{D{Cct#lyFYUB*J+OIF;V_x>14gGKf8^H@2
z``((IE&q;Up+xt2X#SxlQ?F@dj%=lp+dOn~tpTQ4dqCQEq)}U6!ZB_(`=`&^%~=*-
z$(E;3zd)0GFIn*!uVp3WhwO@6^8rDX>9Aaz$><%Y*d8@-o7b*c{x(x`trDvX4o$)G
z<)NVGOTm-)j<JBf7OG`5r1=FA4uG0w7itsF+e)KjgVUu#_7bG#uOEz@(tKp2qVIlz
z2>l!~dX_MH_L9juT3`Hu2cYvwDwdcuf**Q#{o50ZiYrG@fna9JuBFVrXABa?4X&5&
zf{`Qb@@oE>g>G%&0G0E%6I;DkQ4xxiM_(_HJAu=2S%~e=7i2TQlf;5bB>F}PDdeBx
zB$W~DMm*AxO%4R&ZdkJHg4xfF9dIw+pFK;h<XTb=0}Zi7nG);i`VAjp!HREZA~IXo
z=AY`DzJpO_-bhN59w|fAr4SI5eOmd>0nphWechHwW1Bo09u_vy4Uf6$@@w@rJ9v^f
z#P^AGvqoXlU32sHE5QzUNBtOxSUSiimQ0^&gxFYlsuQ~!UkPXp^=V2TVHvDqh7l3T
zN`Lt<uu`R(Pr@EA1Pqigm>qcoL5J=a77EN8yp_DD(8yriuTfi#IVEnmRO+ZyGMmd#
zW5HouEGLt1ii2S<2B<U*yKzg&*%Add&N1`w3;&9XTqbIS`uq4`VyqA=JIkEB#QG=@
zcUG~&;wox#28d5KvK|B<p{f{6V?Sh$!w5#y(R%;C0ehP-Djs;urc|9vo18K*x$=K=
zcV`%x=A-!;m_0fxKX?NKDFX7y_Eh<<s5kf?@GD<S^v5e?Kc@5x;2?ivCm^5}O8gBI
zGfBPVF1$c}OtgMzX@L;6m6#z>v)o-H<V@j?)Qa<df!qy#U~BsHM!)@<v7l1YuGxnd
zUNue;S!D#0Ng0+}XHv6x65Rq>e{-Zl(0jj@gSTC@@~E&jFvpRF;MkzP*H$PHu!l`p
z34<jr*(p=I#r(<t(EqLBbGK5)U*=?6#b*sRGrztiz<wZ<xd6E*93a^E;BZO)X|+-I
zjJ3&z5Rh>duN7@<Fb%7r+5KXGbj`7y2(sGG(TKZdTLmwG``mRx&4>x&d(wIwmTM+2
zDtb<B63Z=IlVsz$1t7zk4O!*`pGp>t4NJ1PjIfeYIhY`JEBr7_ag__7?|^{V_tS7>
zJV#K>L_wNw1&Y|88?I_SAIQDBdhvaD$yNUJ8|mM*HNrkEjb4-%kad{Vg0w6DH1s&}
z7KUvPQb~*yoBHPj>~T#2A627loZr{G>Fq6n3WUSRNzS^|<==@+Tq>>6V@*jgqWq)+
z``$B%mA*hu&xF1E%aR9Sh3^L`-Xg1xKS%P0E`V4lx;SY#KGK~;W0{DTo$U^1{e$R@
z{bF@uQs%KPA}l9DGz^w6I%bOa+u$cVrLM}#w<~w$A^kC>H(B{{=rn~va!I^4E1=j6
z?sjgH>xRKSYuoBhDZLzlooVa6*0O^umvq*+J&2kgP+)Dqf>A<McnMohn;y6kpC(YX
ztR_nR-9kI|9!GBpmvRr6rsPX*<s;y^SD&qL^lfV>au0@HMOTs!#P_DfRfLLKgQU+A
zGwmprmk>+ZAe-1CwujL!O*)<%Wbi=S276tz4AMSDW>*^u2+YknZRM($y3p7N@f(#l
zbRR6)+wIgrwr(pn8;Ey{I9#K}`(IXw2>uR@2TOy5;>aO?-6Q`>m`PY}ANY=!ecO=?
zs#WvnZZpb=cqsU{`z0F$%C2R~4;Ao`?$UhL05l8}-R?hdifI9xzkZ{(ZWJuVk<{ig
zwz&e%yvE%bW@0Pag{=uWd2F`XIqnBfIxa{sC6%9L8FOc;j>_{>$a8<rucXWTUV8f~
zgyeW09|`iL%Qa+GYTr!y=ZCCgGx<xLeEv`Qrzc34tsMB*JyGosJVd~GAQTJgv&RC4
zSu7Nt%x16?m)5-@-ni=0uUoZsp{`k>0bBq4OPSG3-6W>BWV=TP2ArXOd(4HEIJgA3
z%+Jh8Wt@yuDS=RRyfrC_*_42XVdkv;dN47X$1&`l?EW1zKootG@sWWOC!Y@g6Eo>2
z?nhAu{si~rCN?OLFhqTgpmKDYQURniLR!V*?{At9e$h=><*m9LBxJu8oGD|&$19}!
zUB{+=krq1rCb*NVbR6*O_QecY3Ow}Nv*cFd2@F@Ov4z?v2Dnm3EkJhWv!HSlXl>-x
zeZ2k7MaDI7bk?v=W8Ff&uJVwT#Ns+0{#tdfYTIWcj9*U$yR9{lH|38$wqzE{*B`j}
zXXYzfHWg1F3LX0YaTH&}3SF#OFJvG%<0vbxC!4l*i?yg-wXT;PKXM<rt54gyWP5L=
z%QAW>q}*OQQ#0F8y5gcd!rN=wteq!sFxAv=`U*GXt$K@xZ3I(314?z#AB!DJ@=*2Q
z&M6D5+L@Az`&B9oO~rogsbF{a)GO!o?5f)ZQ6cn4zqH&tAVY=hB`M=wL>p@8ObhyD
zeSC|VoAo1M)HyJwNkM8(m)V03uuN8eR$C$XaRPh$3N}RIS}Jc#KYN%IAX5HuKULgz
zy;O4G6A=bQW{>a5GO<w8+|hd3%{s|oT*WCN;|sd`-)(&*wt!{+R$KB=s9TNOcP8!>
zR1UJ*nYy(xoRZNcD`fJ>H)0=YWUcTjTueJ_WN-2*IR|9T0#E@TRS102J!cFp=9gDf
zw(_?zF+#?2{v+h{=QxQDxQlX3ywB`+BN}#j5jl6L{u}Qt7|#x-SK!`UVRBSfr=sqn
z(vu$1GscUzi>4Y>rvEehJYv*38HkW5EtGtTL=VCLp&{G7AJMBiq?OofaWW@!oJBfY
za%4^Sz&+o}W*N0*5f{`pK$&6oOggnzB2t(sMbq-^Nr(d=P~fA}4+MdsJmGV;lsKX}
zpq0d6rb}6BlhYl?%6ub_%xsqLWRoBCl^?fcs_8~)StL+u`G5UhnhLK{a~d0=NaV0<
zmVa&5(hoE&4;FWiVlXD7{mzRvDERHF!Q(reT$yypmwosvJP;6OfWnHvKENU^r))DU
zzgt&zoECHOky%z!`}Z-z9A`V@P^g_WrABs%tL!;BbzJ+bEF1OPb9m?l!B7CabbxsM
zDvDWbcBW{BRWiFWr*m&uTj6F)pME~J>QkB=vr<b&*~ZSab>*mSOH9MK;>|nxoK3zN
zHOI&8ApHOzs~rGbs)?riTds`f_D4}dEC#rV1@klb-D!MnWlf>S-0BZJ_2kVzZZ^pR
zvJo@GF7Cq8#u58(>G!n)tBY%?J8gKVk-Ugz-0!F474qQI-)CV9Z26?Hkz6OaS^q(5
zR$=2z`hOE=0~1#ns8ljL(;1ZX{{Zmv@%37m<mv_v??Un7M)jlHcX1u_yH|ZfwG!zB
z8&O3AzxKz24+KD92A>j4Zw{}*b%9=`7P&vzcD<CKIGflrKdjb0YwYepo@T0q<<%iM
za+$GA1VEQb*wh7~v-5!`s3?HAt^|T(SwmCB#ox7To9fw)PmiyO2F({7@J@<)u{_1%
z?nN{37wb^K9LEEUqz~h#g$P)7sDNg7p(pSB%dZ_)*+yI-dp>O)eIY2!%upls+q)rn
zK0;so;s#_-;A4S*UgBf-if!DUhm-T$$7er$5Ki3QAF_-!$tm|xMiI?NR=m#|&H^W<
zZH-*(nrD<~s?7rSRVEwSseJTY*i<+?RAMo#!RX-GeY8SA$fMcULLC&jYl0o6?mZ(4
zF>!O#)YH==qXu=huAq1C-<##`?phLIktZkfQ8!mSaOI+wO7xXqRLoxD)5ynmP-8hF
z&?3Di;vsB}v{cw3r;d!8&MI*QiZMaK)*1OOcZT(jw4d6$`|0yOOnOt2X;WVmE^@0%
z?Rjn8;V^$ji0L0A!)WHcKQWO2A1`Ac3>?)>9}=X4<(L$nVVrH>#x~qKk~0W3Oy4-s
zIDc!iCd<D)IW^JjaU<nlMdys#ofeBpEnSf{nV!JOH_X7mE1mC}H7YlP*)El^qx`hl
z>;8l$bodWja0@J`Z0b|OC<Su#I>V})wV}|`U!9$1V{K#MbaZrNR8-1>R7TJ7v6U6&
zt8u&;fPX#ZGjq9Shdo{WAd$^9a+vXFYg}(<McPId3mYO||0LE(w%YEQ`J1_k9KvHH
zqW|_4@kit>Sw&gdH-CRr%2#%IJF8-1V$}}#C&jey-o4}I=g*Vyqkj4)!z%N@ZBR`9
z>?TTx!vIez5jkhG)=kcaDDT>$m&-lM{r#a-RHZSqw^_5-`>Y>CkI&y-WlfajV}{`-
zm~Jt%{2{^MdIEp73P9mNKP$?_w6s6R$5p3$^Yb=p{1hchs4COqqqSdkzES+~H|<5A
z=NT)wN$1NoUELeLlD}2VK8)W#YiC9*+mf&xC>kH0>oJIJv#OA(Bj4b4_j;@$Lqf5^
ztJL<TSl&&w;uTp-RjlzgKCARn{k_9@@%Xm#J<%K=TUp6seEM`+TbuHk3XUJ?vp^m;
z0&QH+LKmtRPuu4U6(9IAbh9O&u*xQ$VEfC4ptI2UE6DS+t4T5m)7#H5+RvH3(Pgnz
zfA1Ha^WMrm)@r&y=Z)^Y$&?Ty&EqUdq4g-Sz`vI~Z2#<49`!>N?k&|=9LY=P>~-jb
zUt2jM6R+Zf0EO5;_#LsY>pJ{e7jcg&vG@YpU&Qk~c?N&a7lIDY!*fe3?3j)Q+w%Qd
zeg;B0>Fi4R#jJ}aLC{>i)`QY2Z00n%iT-nv@nH>M*T2_{VphYeO&5(<-DHC8665fH
z!+N~x{T>8=N6t~_Pf^7UF2+@moEZ|%>{0Ko)JT@K&^~;la#{VRcv>v-bUI0F>tj*B
zSa6NU;97*Be)Ywr#V)$&OZZgGjZt8#k+HZj)6K56xBGZ^5vUy}pQvvGXypr3q;_Ib
ziWHNIP+%iDhqdSV5ByS5xhDD?#m^h3u2;ihDh~p&dWomGR1T_DSX!}_c8!efci-o+
zuavYD5e-}xoXFyeZ<SD&_|26|O_;F-E-dP2tJu`-Iz;y5vrFKKvxuFt+RjNO9A)r|
zIzO8sd!o_;?Kq(=uC>@*)@-yGY2ZfR91Gt6MP9P)l2NAiJ<h_k*`l|G(=!-8bzW{H
zNKx?S>v~Q0_X{E99R-C+U$e5h96+1SVNaBh$^*Bj*~2gM_sd^G2PG{thDYKy=d1f6
ztEJ13=|;jJNcIqFjZzJt->m7Bwr8`ILY~<e>Ueq2X_w93u2ptJBT)>W|5Ej|I0FkU
zlX(fXTtZpn%de+5e=NpIr>`f9axBL7#><LjpJ|f^Of<30tko1M9%wat;;iSN!%zA3
z0@jnDO|j$%2DR_<Zi<@@3`Y6#H&^-j{A{U(g|}Mf1-A?czuv7u+%&#*?7Cz~Ybrw~
zDI2;R%q}Vdf}xO{rAGC7O0H=$!uix#=B*(cE2;uTr@&3-pJLA+dyV5ks+M|V7jfDx
z3R<U8N6UE;?ccJ+E-GIy4R%y_Qd(A8@UV1zFBaio?f{Teh6o}cM$Je4i}U%j>^ugL
zk&SBQT#olA-nLkZ+Psjzt&6#M&72ao?QBaqpI?ZCR#tw_(qi7>w_uTO&QWpmCc|mA
zlcfL~FFB&=4;q7dedriLutqQ2$2XA;TR)izSIc>YVAH5Q%2)d{!Bv)K<2m{aV@nxj
z%A?dXrYZ-sC8k2d4WYLSU}L;N8l$s};oqkL+rkU$NR2}4cjkloEM(j%y-bg_d7qWt
zEWh|TUy!a|I_G_vO4cJ11XJY%hsVJX{VS=0pB%rboBnJGG&D75s5cE$H>vYjh$~^`
zm1e2CWp^-7D|z#wkB)T?H}SG9o=R!t6_u$sSvbj6^8=DLkKu-Bu9*=*h9)N7ppJ*+
zO8q=nonzrazMXQ#h`e)+jmNMey6c!nt6wXEmtlOQ19U>ity(=#vp~J+1;$7)#~xNy
zF;OYrU@?|rhzhzDBl}0T5?-Y-i>ZR5I(5NeSgPdoDNB3<HTZ#M9;Tak2ak#>o7^*8
zkZOZPBM?8@)x4c4P(*%OlW=s>u^?q6dlh&nqb0Jy(9My|tn)?(=ptT3%q*4drUwBJ
z&*jIK78^FVKVA_rk1ZxCR}%(j$0~)qkJHE-MYGh_-HM7cwJvlwo_*nWWQt~4Qf71>
ztX^#FKekl$5fMR?5(IbE-{neK)-A&Nl7E5-sGqMlSG*ikME{W*8*D`H+Ex{m7tp?{
zHX<@nD+|?irhN$PHr^qYO_AMRjniR<-Fk9WF+{qTO>wiZvXF58_bRAbc?3o?l9b}^
zVEZTgSew=N%mgQHsh+SYPGR?(3tYdLfe}0P)|~MARcfKWv`|7=052awY9*w6|CY>A
zHQ6!39TI#j1u%TD!!S9y6#GlBgypA0Dfx>gQUaCpABqpgOQN~r#{+)iqXC**ggv&T
zQXcIZxifBsVI+;}TxXAAR9)&=T<TOke{PDs*wlJb73KBH#(PNJqJ{^3Z-;qu`j`tk
zy(~Dqwgu5GBrNFhZUuzb%dDc5!jvn5QFG(5W-(@stUvT_odumJ4OG`bASItr<3ud0
zqzB~gv}k2TrVyi4layjzF`(1pFCF~VaYdK`R@%RC?+i58O*M-1`gH~hO{#|l6S!t_
zE*rG2VjT!qSd5K^ON^WvzcpFxHw~Y%SOf$Nd!Uk&cyl0V*1EUnqy|{2%4b@lPsX=D
zU=n7v_xxpE7|?z)qnZ7P->kq(XxhBe2y|$z=*yC9ugQ*;RjNJ+wCJ{Fwx9=s5{RiR
z>yTS}Uc18BDbe2sn$YYWm7>*N^{LX)YQ+aXnfZ!qAd-ojR%0g|;0cp_v^nLsm_N6e
z2Mc#pmwLs7$aib9YhwRc1nSr155vIh*pT_gTtqfmIs`1&KJYI9ztHL5zVhoE_>ryM
zM>kvS8mO)COcTd(%Nh0YqekbW`aSl*_H5gns0kJoOO1#Wne1Xey@GJ}Mw@z;5ySRE
zb10e2dSh(3_6QYg+Y7K`W#tzt4nhSz$#_&MPLrqwz0}U}ke^<Fn0nctP0g+Ch%c_?
z9)<A=h*@sc1c=%aPp#D|llDS!V%&32Twr=~{nJykH9I#JbG}E)WtGddmW3y8U_V@n
zoB~HIj)W*4M>P6N5mq-gF;<_Fk9p0MC_)9+Je=hiA_kS(ucS~Ns@oe~?%W|!8W&_9
zvq|T%M;^;JCpS5rulL~`rN}d0jeUToH2~y%2;}0^-|F-XrptWXOL!40_KFu=Zc%*>
z9T4r4E!gHe%x<W(>rZO5>-*1CyG0P1PVb%>J7dExTr4FIL2c@-VB>J$ls4{3b7uZC
z$0J5=6$aL`9($!#53L~W@vtn9vJ42d(^c_VuEB?R1kfRe@OmHk1VR}ZJAY`sTCylU
zrcT%Hm1%L;X>vi>7{P*aZM{X!q$!T|&Y!)l>il=m{1<1k7_(N5yl3CFoV1IFEH9{b
zv-gtslrAla<5`-_lon2Q6qwl*O=Fj0qvQDqS5AtR3J@Fix=~h#$CSLKcf_oGtiiZC
z&7fcjIt4fg<hQ<IRGaV-&GMU76eun1)Kov{8Oy<JBE#h~x1Az*9Y3uY@_*WU?{KW&
zKYaK$5*d*~sA$-$B1&eZjEu}|l`^uDL^e%H3l$<IE2AQ#jHnc!L{?c*A!H;Y`#G;$
zeSgp2&%e*FqoYII_jO<6HO|-hI<L2U<|ncFR;kO{9DgemZ;Nz38`U))Dw3h}x>!B=
zz>9*}Tj^bey%H)zPF{Yy26#%d))k^17Spoy8~s;&mYE`K3T4dF^fGCe<ftymP{GEM
zWo^#+E=wdeTU>rO3gtd)UHVzvDer(&v75_GME-L(*W_cTJ*JfZ=x$<{mSbpPZTuP~
zQ(`VT)EafQJ#A)3lt4gZ#%re6WLR#*Hk=}>covJB@MDjrzDpUeTl2IFRX>cZG7eXC
zV$H+~&YW$U>B&oDacee79uzRhe}CAmqv@gfiye&h;?n0^OH8L;SaLU=5n}g#y5Ikm
zc*S<-#=ZcnoDQQk0q>lCG3V(E@<5+t3Z|1RNG~hh+%MF@RI(&V-%_04wl6ye?k>~f
zY;&(9$;l&Szx3GXlJe7e0g=&_-4c^r*~TIH8C-77>fKryuLpElCu5y(2qw%32ykax
zM-22;KhEjbLspk(R(tYJw^ofXht);!iDl~F8M3GzZ#nx_d`-b;p~T6!s$s+WW@n!A
zz9!Z#3Hc=*y6uTIT1$K~znr#OGPUWvd(+ZxmM(gVB|mj+%JJReSzLQV<&;Pe_L-uk
znM)3t<dY4n_x#Vr^%Hwd?UJmelmYyrGE9U*5r55IqiwNumjY7fj5teXt8wacr!QOm
zHbwo+P`LG!*EjpNiMUH3)9EaVs{nP|;oZJG<v!g_2Qs3f{YtD<UGh6jOZRI>M_+As
zQ+f2%#eP=noR{zZOcVY_^Ow0S1Gxj!(PyeBvw0twF4%QN=`ct4tMEI8Di0R5PI*n%
zxfMkQ)IY0U{YlgIN(OOA@+XI@)~UP7(j#M4nv2{t#<R}Z^1_Oz^m?>}OZjQHC4w_r
zbmKkxyuliO-w-rDkZzut%c1+^p;@y;G<6XC6B0~4KJs7MkpKJT|J{fGpBTaVa`Fq|
z@u$t~$~TdO58|=!#bT;O@i+E;Qw{gBC7+*3R(gC_xMJDt^2ed|)Zp)aO-M~H;BQi0
zYI&FYpx?$71`7PHpVmfbM1d7aF%mPL`Skg7+L0}y)k&w=%?79u9-m-x+oknRLW1dz
z_@@8<yWG>$BP%bT^7%8AG&N#3-ihFY@$Zdysrn-OuKRs46<KNiG*UcDEA9Dni+4v3
zSXo*1_x06vb|y~G%;Y)N3$e1WIM~{%#%^;tpr^N-X0tmTNtYA~jd3PdhRG#4(<_ap
zSG#{tnRnW@p31!R-|rXEm!9-bO*=iA?$7D3Jk@xlvHH;CTSY}hyy=zc3G${<#^TlM
zIoNC{kJfm4dfthPYXC{`@Zm!zwIR&+=&@sR^77t$Y`i7U7hiWWt`z5BJ1?i8V5UeW
z;_j{w4yxsa&r?t{QCi7=>}gF_pV7f|%61rvMainDWRsPC_|Po&CcU*^{9P7zvsZUf
zmR9n|^nmtht^AUb5_1Rr%=H;2QN|hXrl+S@hQI!^i6%#k?30{a^44QD4u=k@-?(vO
zdaPbgZu91QmQt0pnVorEZH1mYyocJBWN!t*CTa1^VW26KkYPRVWAl7|AB~8J07y2y
zu$Y|=Yb9T=iC0W{_f9i@yXV7Gcfh`2anaGy+x|0eFfB^1cf{W{il=@i<)<-JfM3dE
zw>y4Z=i9e$AewpxUsKE_BqjU){Q9=TXY>K#q;3f7t?0KFZVc2P`chij{UkV;ru}67
z{-U{wR<Bl<e8<*jE=AqZ7Aqqw4=Qx7r`kdgU!qm2B9?IRBGdl;``=CaGj(P~{?h_*
z^)0iK;JKo4pEc@8*vqmvo)bS0#w&QJ3+u+!K0aJ|^5>_=Ex(7%-BNPuj4C>d&rCKw
zI~bj8;Z|(in(x-U#=GsF;!we^{0`p?-!o?p$|Np2a>1MQsKaE4U*9UdzQ5&&-{Eyh
zf07}`#Ue#F{$rY9in)~b#uY6k{?DJeeaUKdeV%Dzps{InPQd&3@B7Mv>2{nRNhHg?
zzc<o)#A-!Y@jz&=PhMF`2dKYOKR?}bX?^C{oMqKg=$Qg0kog?z#WL<P<L&;_xiHtp
z_V)LOus+A5M>Q%V*JT?=tWokcGBh-l_-}{41lRb>rR#o#E>})<c`Biu-*zJWnM><K
z4A`jNSyH0R#bOo;s>-XgjK+WD<<gdXw*=*>@0?urp{Wzg=3h$(7@V0e`CX%!cV^PH
z-4nl|d}ipm&(NE59HXM*1y28n(j4~dLj~X1{1;}v924Aj^$UI%c3oe63OrU$Svl+R
z<IS)W$CD?G?%%)9yLxrq+qaAh2bWOj52P7=+`G>I!Q9N`0NI5T-!FXZlVyHN!@|l+
z)|)5LHPPy3s=$BLE@<HGq(8Pbooi&M{kds<eZ2&W)Q=_$w?_Qe`9pg}MTNvN+KbC<
zC>iTxsvE^_xAYATHvai@T5@I9^3e0=&--L-q}W=CtX#Ik?@!LeU~3A_w!iabDB_XC
ze+;rpN|~nCK6|oXj9%I(dRwRe?wO&w;>nJTGsDHE-W@)|GxB9}GKeTP^P@4o-y-D4
zI?fEcqzXnC_%=o>FO1tPY|`ZA<Etn8ymzh7?#VBKr3cbhhKH#F?Ydh@NJu!?*{Pjp
zTjO+Qs<!s(Hn$GT43?~H_O<-{MjOlPSK%%|b?@&TCx2WeQ}dsh=)=@b@Si`c7H%n(
z0&~-LD;|&DxRrk>>-kW+@wLI0JZ;1*iIwiR{k}BAaL2F}jvwwSaqqlK$$Hh$a5Laf
z#k&6UDzo<xOV^(6q*0pwao=~WZZ8g`hSsj#yk2rmz`|r;pq0eewzfEUZKd0d2WAov
zVcwx<3xe98dk;j;|A?d=zm+bzCJR!wN7t@hJJ;Xo#Itk9j&%4Tg9Ih!od-W`?dI9k
zg7GGc#=1I=PnDCxfuq8#lHPCLypi3qrCfG^SpjC+arzr;@%;1vF?7pjdjt1IiRzCy
zxpDl`>b`4x*Vakqx0NSRxb;O=lAUc-@3?(?HP@2nc6>hxzdKbYQh7S{)L=`pgKSlj
zZM#{0@bHGno#lFr!J~yA{B;B8ngiFQR-RO%{i1#G^&!^5_33Kv<F8b=CpLckc>S(V
z&uXruuL%oN33=GsGUXEHLBmZqG+jDgOob?YlCNs5Q~FSUZ`b*;>4EHl@n)OSu59N|
zJXpclpKp-^b$izlBXe8`VwJw}?nk4kO`~+cFXm!+LT>fTJ0BYxlgW`M%(V5A*Z+=-
zkc*t$M`t!!)%`?y{?B3m*<Y2ct^OqE6m_Z=*3}uXE?>S?pN-FC+>M^kR4S&v?*3kB
zvq3AD{CdkZmj2#n?GFx2JdV-Y1Oq(N^7L4wb>N&ABE>J}-~lHmCoST2cLV2rHmzV%
z8mryq`}1+-*ht0lt-BfdoK)sxN@tU;1E%bYCt94tUYI@Ia{Ds%zq6y&T0%Fb8*S!{
z{eOkqG#pd@ecP*8C9D&Idv>!%f$-oH6nFcFDJgRB1vtku%cj|(OTT{oI_GWn==D9}
zDL?Bo6F=)RvP;tCzxWxxK6Gu}mZq>nk5iki&+N}Gp18Bov9~7Px99e;u?9muM_b#V
zCF-60J5TGY%=PKvm%B?l+t~##d0BM5y4%S=5FR=H-_{R5`ce{_T=LuBe6@WKPs!=p
ztj|V~`f|3z%Aj%fL+-S<8A5?ov-0VsZ2Yz1(j}=0^J6Jh-6v*8a(et>&BSBoC)^fH
zmaEL>kaJf0^@{U_IMYDOt5;zG)Ab4c$uZU+U%jGd4pCJIoPSZWFgH9f^JSrg1VL&4
zK5gIrv=nq7-hS;ow?3H?q@8L~kY}i(Kja-a={;UXw>-y|k{G_Q_5Blxh|8CAbHh`g
zUH&mJ@VtLm1;HT3Cg7mT{I5u99xXcOgCZ-h56}3{4Cap`Xo;_l9Qd7pa|)PtR2nIx
zV_9Ba)TwZVoTSpi?C+M+Kov|Wqa<T&cz$S@9N&O^;BWcyw;LjBM``5g31hPaZLRIw
z(~wP-roP`m`!a|2Zm%N@0I}UgdickWf*&%4SFT)PW?&}K!P@WF9^zvstYThk*xrQs
zjD6Jh?MsenXl_n{#fJM%_oa9>TX{cLF!O{}%+HiASS<Wnx8U)z*H<95f8k)heKnWT
z%+IIOc{Ve09m&B1xDx}A;jDE}pPJcb|NMFuvz(2ssb!B*XlH$j-oR^an<IaCw}qYD
zVa507nP;yqdAslEJK<H~ug&<|^b`C>FV6o~SuiBq)OUmJZSIz-*wQIJ|H<cH8r^2!
zxcv~fnYx5F*FhdjPpl9nuIbUEM`gEet!NFUxgAg%IBpu4jA-NEEpAf>)3#?Q?kXDf
zw;3*7D4CyJnEwL*Sk6qFd+q(v@z8a%stYiNhaUpxfBQ~+KIZj7X);Hq;OHVow2hCA
zjjdOitx)-qFyEIj2Bb5PA25)g>pydfi=p}Q@XYk*4*?GYr(y!@E^-Q+vD=;vlqU(n
z+NXO}yH9WYbiZk7y0l*x^VrnubdX4xZBP($y!4Ba5bM)jO9nfNElTHFOEZP<o>src
zcs{^P$*pwmS;<WMaQ~0G`w|T9XW7;%KPvqbURql^KTz6V*n790gs~*s#*y^p7>+`d
z8@LcK_@&gw;@R=Kb+a$m{YXo>n{<iAv>W+xddRyp1D^SWr%Y`7sh_vz$MhDACfmIe
zkgT@1`%kt@EX(4}#=QpIK(yuL<eaW9oz<S}7@oUJ%#MpbX7KyF8`C}VrD@rLa|2<;
zWpYovhpqkR25g%1OU5Nz9(1M~bWsamzA^xr@!r7sUXt&}+dMz?mCOtsHgGq9=}mV`
zmZWgo_!*Q8KNWC}8=hHqR^8NpO`BPE$#lB4|G4qoboN3~$wXeZ<MHE{_q??KyH%{;
z$r|3sm)twQ)_0%H!gMp$7)2cPrDCXO_hqJI^1A{X^9j1M>VgC@oIY(fH#gTmFyOr1
z>(nVR-Uu<Sa8dl_jcBTwDUoh{=A44&>4`##)@M;4dZTnx@(K!4r>1-saVGGvv#+l?
zK<((xNIchN6i*KPA*?cW@vi^GLH|K#8^dGAgu_EtElzJ+;_01P07T<sW2yQ1`4f*b
zGPK*<+b_L7bZPf`{LLFloHq(Y4>5Y6D3ARfuNWLGs@h&sN~$R&tgx`KrLwSadz98D
zu^QK_01SJFU;Msu*Kbr)X|m0;sPAE{8RSX#c1=&@m%g{C%6{6>VcqUgxu)fbefTri
z_Vi!BJdlkJ0v}zf{AN0yx8d4_9XobN&76AXGVt^BQ&KM6-g-=S5!J@;^f|C~>(<%Y
zEV@dMF=F*93sWD)Gu`~Pmn;7f2oH&xYkx`)D^PWRb!c-$WTf-|Fm}_(NDknCLwkGX
z-MeeOyuEF_f^y7J1vsgdrz8ZljMP1I2}o>yNlkFS-tbMgG?zTg4j9kCK1ftC7$H4J
zF8qm{o(x>@ZjrY<jM52cz83Hy?{FpC;HSrj_m{Y+@J2YHOlT?ca;~<=L<z1VhF@`h
z`M~{1<&Pgfu25-nK7v3NI9nO`NIGz;nrarLvA%?R50y%Jv=tQ(**94LNP;{J0|OU|
z(xcb#@YI%-1q~rJo*TUK?Z>xBm8`Q&09GI2vW!Vt9IemBF}X>qE4u~!j%~?v<_>=n
z`~s<zY<?!9Dc3jNeL^>i7UpNh5qF&@huVv#dSdMlHWdZvgocLl^7A)v>g^PKH^!js
z*>hQWc4SvE5HN}COhMVf>*jv!r6Ry%ffL6_s5{7R$xfxQax}b7o^(J-GQH6He6m1f
zV0P|Xvkd=kw#~GH0(Ut~cXn=OF{nN%ng8tLqn!^v_q)zNX*w=xVbMHhaj?+dJ}bHD
zq07=dm;9e)EiXUHw(mT=R(~`7?%fwc85w`iR9sWrwPbbY-P_H-e|6pqN%m~duQ-<F
z^CdxX#}0GAK|_3Q<1-hFshOGQ^USOC^!DuCD~=x##J{nxA<b}yduPz~8#hSR=F(mi
zC2G4=TT`>{f#wa_6W{mkk5e!R2nY!KT0Z*wcb<Ef`dqKdf*{M%Er7xeX*aS8dw9qD
z>h}VGUyq52saEMpO%>U*XOH8j9J}<?)Y^{XGkM4z#DrlE`aV9}h0oJ{|Nh<K&E?fa
zNJ(U$#T688>5o)<t>hp9DnfewdUaHSMl%zGjagQTdCtwcU0q$eDoc9y?T_7NobTG+
zkZB<w!7Hcr`t|FU$A`}YGe3&plNXbdi@M5R=9JeD6R0op8W=;$6cH2**0-!hQXhN8
zsV5>Vd|rnC!QRNVygN_lFtG3lgunjUbnx`))2W6jx@0|XBKT6u%hfY&O06|DHK*ri
zh6`86(fR#)%}pF~ZNMKp)-^kBs=BX5Jwc%)a^$+Snke~>>7yoT%yTw9d-g1cmgLm*
zw3xE8H4u_7DyOL1x6@epwuy*}YCb<TxIS^m4n*j+KHoW5S8v%JIQwl~dr8@eP2#3m
z4IxXIZxx&j(v9EY`1YzGuY$Y4hi@^R!^7D?piT%9Rqx+7l68Gqm^Zb^cij{xeq_sS
zBYpkE&rgquZ{L0pKAu>ucl~~k2qCwFe0#dO?xA=TlakVHFY+>sk+Kr`(64s-bb-cY
zo_ib{G(?n?tZe*#oxOhj`pJ<_weqyk2M;#7d?~TgQyRZ{=<s2tR~Msh-uyw5(H%Kw
z#`|o3etsthGV`7OjC5VPbZPxn^CxB!MF99-_@<b+xHdJZk7oypDk{dhXVs<~XWYDg
zJ*m*M_sMIzjRsVjWPH7tm=+xai~ev&abudHaQ-Ln=6&loY#<r=2dS-s>6VU7^(G8d
zELSlyFyQF24E|ObIe>#`MnXJLn6f2K+r%V#Ei^HMpP!zW$<nvAwe618dg0j1V`pbq
z1aFq}9dpRE$k)Y|eqR<LPO|v^TQb4&y#g1@32I5*n$ZZ(=Yf<}QEK4Ayw<*b`e1Vy
z-w4#S&$_e~c$gRa=AqP3>hA95(2<Oais}OvH92wO2|50$KY!RS-T3s(C3b$cS4CW2
zK1L=>)xjZksAR^q-S7AN7e1zXFN%xhPLJ5KuH9)yfbr6Geu+Tlmd9%11mnyVCVSct
z>(cEi&T(AeyH5oR{Ufw_dI`!}Iyy$`>a;Niyq2n}L2pp8D-ORfT<1S_ts`K@WsOQe
zoUKzIBD!98as^z1{*e36<}=+%NlCIt|KpjPZZ7R%y~1#E=%^b>{iLuW1;!KeOnSpz
z_V)I76uq-Z4Jm5)Fef`Zc8g!vfW$)Tej(0dH#pZw*TDJQlgZ2-Y7?sldGBm<(Gt;p
z)7)&jVegd&x4;<-|M{t2^T&tIB1$tp<eK>T`92e;kV8$}_Ws5U?qHq<=j`H#@*Sr`
z1)S0#7~<V}^fCh*zsS?iIsy?JckZ-=lBqx=a&6`&?Pv!?rLY6c#N=OwE|qiZ*eO)s
ze)XMG=Q$}Al`o>|v(2UR$)y0~ar>_$N~f6G_(usyl^@?${o%uh<hAwf-d{2s8jRSo
z9)L<1c-HE=fl=Y??5rOgDop<V@eaS=`(M9$Rq*rMKx@AA($6%94jnpKEt&5y()n^3
zBclL2_W^6`Ea}qO1JyCo&9FDd2iv4Y_6sS^j#Th&K16TF^WXJN+PNqD4uqthU903v
zHA}f>%UwJSxo%&SB@aH?J<yyjeA)a$czFHq4=TeatmviQ*K(Ew2M6nNBxPqyAG;ob
zGNzti9$r|@#K>5S1TS~|lT1v2VOjg__eY6k$hWzNHf7sncUQ-F)+h4fnC_?qmQ+<&
zd-cY9uehKlVe#zplYFj_&`>$Aej`?1*$Xl(5<)^6@RK+|F2(l=DmJTEuMRsvNBF0)
zzWVXm-yiC23OOdICM8w-+}n*9WdD~^m6>ft95DD-uXb+|5^}14xSxE~m9<J*GIAqd
z!=<YbGr=CLzpOFQ6YJKGI#>7tFLokKX#bs8QGB}JTb#3vEiL1srL59{V9sZ5a<)Yp
zr;-LZJS`jfE;WKG%03tS0Nv<jn&)Xr7mr<MSiV;0^T3<0-@YM6xH*DB8y)GiH{35_
zmZOT-v8l->5dtHeTUzJ>?}aN@ng+5<vmEOm8cgiC_i%r#Q`JpT^Jm9bn|3zTMoU`A
z9)81uaKwI&&$gQouKD(P=G_AW8Q?hj!LHP<TllkgkA}tymW}c4MJ}V?-}(N2FTaY9
zPg_84<fFE9-(db1p$i&0hh8s}G|#;t;AC4K#$;Y}G7T)I4v%gYY}WkAk&E*q=T$~}
zELr^)<|aw~EaKlzS>VOTbw<<D(aAY}-01(O>-?RaK2LBmOiLOsVcK6`hm$k%Idg{d
zt^JG9D)E~otUsTAP_8dO%gpq3iv(9VmxNg&U<gh(T-7oxB_)OA8A2<)eEIVDFZb1m
zKOYkoW{)A*aGYbN28AFCCe%}O%XcO=W|%r-u5IpIa)-9X${{I}3f`$MX-Tl<8}aNH
zN#K+c1b9`QoYFzQ?q!eR;O0&tiKV|jWw~8JDkzEpY|eS#@%}wLF^}yk7|8qSyUM9)
zqI27uD{C^M#7&*Xe$<UlP71R(nCs|V#)dzHYpi%<|KiWra3D5uwTAm&zaAoG^bY@N
zSv$*~Lfn+HF>!DZu`4L(?6fV$>v&ZHcG_9KxwJ}r3{J`s7M?_mF{t$iX5yWq5x<0<
zp42TLZ;Bdz+B&-8kqMrw6jS=+)kV%Q13gkHB|mt;{%X&y_ebB~Rk8rb8=aCO99+Pp
zr>8eJKUvyrsHgP0X-&-LV>SGUXdGe8hrEa>Q1XxnWt|#CBrOZBx~NGB3JdRjdi4EA
zAY;agmxQ(?lmNX4hqAJ=?@(c{?2)(Z<?e>4c;kCMBxJ+P+28E>-re1Y@<)7Tsz`vE
z-6;j?9v&X%xegpLmu$+LGA%xfe|*WdX2*kL3I5RnQtJf-REd>OPYi1FxMq0|clb_q
z$M_r`T0yAXW}Cox0ZW#%(qwuKYQSK54t48YdrmQtpp*Y4K%1RS-_r6CrkMWVL6}TB
z2S5KK1Uh4@=chKY?7I%DB>#<)?~(e=U+nQsBg9>BB`Jd``g?(TF{^7y_wO@(Ti^5F
z?fYY6Cqq*2;|%*px+>@mB5)e8@qL4>`NAx5DHzc*!eh0b^3(&}yE`@6r9!A_kisvp
zapP`M#*$QjpLfZUC4>iHN@5!w9nJNdIuVk3E;2F_CY1tASFW0eEKOnzf^eja)adLe
zd>SX5NDViXrA#a=4WNPd_MPNVnH>q;cKnknyQ#y}Xt$Vz#KsHEw{(Qw-`;v&-opJL
zRGziy9&~r9tucv*af@%*5Gr6vIR*_Bg`#lv^`W}iso$$suH4PjXoh9NX_%6tw`|>-
zccNP>B=rqoKoTNZh-$>E%1Xk)ZMxtTx7|~tI$ClABe&{eoyx|wYE?4M$)uyBBg`W>
zbm{W`8+yuu7tD#rS5;Lpex*J9=5ifkf-ZY%l!a^I9-s{23r>#?KGhHv6{YX`PIV4S
z;@XpaANRnP)?aXX0lW$<g;=Epd`uA0S6Udzbj7-PRrZ4QH-z>d_)_A}l*CH{8olt|
z#hHwM2Y1*!95AUEQlNV3)H8xCLQ(~*Vr)jo#~c5<z14nOEEApSt0P++fr*!~uxz}{
z{0Q(<(R*kkqp-%`K^%Zf+H}%+#Z4a7zPllesPeJCJ}+R-H>98p5PsF_)tc<7H)?D3
z_a`XF3#31Wt*N^o`fD(eJC&5i0e|0CR-Tth7ZDTt2uPyJ9`g%~B}e%Z;NZkqx02fe
zo$R3utRIo9n7-1Iq=MwkRIa&*+f<dD8`Gm6>fBam=iaBKwLYx+&v4*^Hls29k|lk5
zfzzC7W%nNH#;1b?ydxkrf*isba)avqlpqFoR7m2=%6A3Q_i1a_qb6Yby8a6P_9W2z
zoV!<X3hNrVyUSnz5@8WYu_mwxYK*#<77dj*=-rWz$Z!B;dw7f=Sv`-24G;3yaIapS
z_GPvQoS5%s+cL_8MQK@A#re-oFaOP%j%oRF(wb3YH=dm=T`=CWXNj6o=$2VOoYnRG
ze0jA*YuM#n*YZGt3(Ps0p`QOHOaRr?)D%b9d(a)IK%_7wfbd(7GS;JVLh7c&TgUqw
zwVN$HS=EeQ^6AwqVoM4yvENjK`;1A;mYuaR&v#uTXtdem>u#UZ`gV!)`|rqJWKZ9>
ze}DITVc}O#HxY}M&AGN=ub7C)1%cxwf4(wD#l&Dw$m!F^{`ld5guQ~EcQ?94vhJO#
zA*mtqiE+ypql&CdU|?Xx>aFJlj-T-KjJk72i+!6(hVi?1@96C+B3f!|T}$Uqsiz)7
zAr;nn<gYz*w4qD3CH6hN-P2<q)~e@B0o$U4P-VaV-_fmSCX-%{|NLzD9uGdbAok%y
zn`M2WJ{7jBZpbl{uzYB}<ZO03kla4TZ5J<IbZUCES}jq{=k#d@lza4chDJsYa&u)v
z+#h=0B9Hl^Bzllf3$uNS9+7&^z1e5a!5lEr#GPTt0gzI6U%zOewfg`OR`K)euwOrL
z;DC{t**^9aGm|Z_EKy=KwIYZ-jes*DIb9JVz3PiD5%H?JT6@=dwoQz&V7`j{88&U&
zr1SCf#*G`_wzkF!)H<9Fcy@opOX}}OT*xY&_f^ladtyw=UBr~_t2cTLG)1u2rut6&
z)MVU-T{3iZ6jf7TS`7XUyAh|Z0?eshVA^t#@0oF9vNpH6>iuQR%!YU(s@gs=Ny$c-
z8$9IS7LAUL9l2h26lAw+0!{MdK{Yisr^fVEYKe6)m2mdspIqPg{Cn;F5hZbHX+1|r
zN7wk5$k5w4Hlmz;cVr6(J&T9bSBWgtn184Kw&>KL7UQ-dOOfwAJ%lfeqyM^}-Cc^8
zW9E_K^1rh^Qaat3{k^YGll{c7B<sXbdo>eb`gSqOgk8A62#g&kuo1~MwV>dZKpf?-
z0d29dvGMv*8!Yg>!-ChnFxM)aOOlc1TwRI(8TG42^2|x|FT|a8aUQl5mzG>X`uO-0
zlA@v>)s;2##m7($iEP{$CSc3crdlhtY13Kue`E7dI3te1mYo;o=Uk^gwU2w)q}o?U
zx8<;&Qxx`Sg2^+!WLn&eAe!BK_Hcx}N4RMp+RV+xC5AFR^h4<q9{nRnHVS^41s@yM
zNxx{oWfYOZpMb9Yry{o_Z_v3_<SJf1$7J&y91>>;=ih8Zm-zjOwzjtC^UAopcWZ)s
zztFOXA=#nZO)n^lLKO@Jy`B5Nb%9oBL@3<Po;2d+y|P4x_r|+d5v}R&l^HT|tczR4
zvm1byN$ps5*riL$@_qkXS#k=`3=N}H*jX%9ocJZDym$;UjLXhGMS8gz71eDLd7HLd
z!p+T1sNHU#&Yw}gZk@kN5D@VjL_TZp|5fyD8&{(v0;`S?0;qez<A?3ko$37FD_8z(
zbBufo;AMa8m}l)KLNL&>h;G?(z$7CIC;IT=ijbF!5jsds|4&O{{VQh;4Thy{Lb9?q
z*=_fsB-Tl2<==kd)!@WI?`?lq%#Lr97$7)#XC!$fkC}-{ke=r{8aF^&i-rwQ%~)Rc
zZx+ZOqA!lOrH$;@)9Vdtb)T<D-MMpT!wVlzxa1ob1{Ir&#N8>EL(2yqB?L^rbCFv8
z5aY733DNU#ozY2H;Aj1|7+WcH_ePX5a&sG8T#xkV&aCpBueiOr%Q@!uZB0gozuAuA
z<D-LH)pI;z-kC15Q55pCI|3vmC-^B&!6Ss7<?>=uU<oU4u359}%k}Fkgu0JoI!7Lx
zG18uU9hSZ3Z1&3py?--e+N<b&%TEcM%Oex<rLCuiV&5MP39>tT?%axyi08)(zdc%9
zWy0&TasRXcWvaJE@;uMKYmSTKXUk=csSfpQfamh^{WS(gL_uU>tNG3I07U!_d=G`~
zPQoOTlamq11sFsB&sy`UtE)YyD))6cSI!s96a@uQP(y?X94}CRS@3szRbdVY;25oF
zSDKZfsg>6%!Ju9TVf8L&%fAOE0Lumcwt+ETa+kQ6nC0~5EnUvs?Ch$HGA|Mn1+Z7=
zUA*H~{hc1mncH_!97QE041bKgR^BWnBXgDA_Own?D|Y(dkTRBAj`#MTO;tQ@VPL59
zr+hW62YLr=@ZYS<a!zxv;xe0>GeN%_ZhiXrzv<e3s%WlhWVUf_Ia(Pe9Y@1ia!Th9
zFZbWizKn<|_(`T~Kv~4Eiv`<QT~i}O|8GM=9b0lm1wR#x(*K0^!N}aaO6}RbeW6=z
z%bTw-+59t+(0kbZ?(*iwlo;m<luYkVDJ&7H8wBOV_%iYDrrBcCEDKi>>s_{NJw5dW
zIABvzo|XT0;cWIA&XS?v>hCWL`{P9XBK~G*mUA#1+_Kc@s@}f5x1wo7uehkFAU)mR
zGYh>p`^2-G$;LF`c-(JyV{2>cF#RQ*C2w{wrX7Y|Q}*7Y-!)_WkK@3R&js<>C{NT=
zL&}?P{hM6Dkxy9#1)uDtS4&Ar^-x;>087xug#dC08oi8W``=&{VZ&1h7nY+Qkj0z}
zb*_lzcjr3#^x)dH4UWAqIL>3_J}iZYrMr2`GH;~&fP><?b?elF>TRo<#}g9b*Zf;6
zXGyPTqo&!V6*WEz!c?dEKg)eymhizi+yLKw*z(dd7a&JY=x?}cVV0XXzjL|8jsgTQ
z1<q#zT|WPf3^1-IA4>W5Yiw*IJ)e-Dl@0I`BaLn-!Yq1woFO!SBMuF?4A@?ylAqPM
zhZU2MUzZMy11EX-tK?xagv{4)IX%4zja)nJrOo+h1a$ZIa?;1CgV+=NG~!>C==|@q
z>~JWU<g=4GXTX0j(op!rB-fB3%7Q*^_*i3{uS`OF!_vPK$GFR_v*IhlP+#Sl?P3St
ziEUaATg=+gRlT-^5;Xm{643B_{3gt8lWBaIJpzZm^#4Bb*Vy;9t-S?Ko2%LfX(;2t
zf5Xb#<J&kxSUFK@HSqu4UG}rt@flqAf5HM%m%}4Z3w{;I^4$u?tZYZd<3gt{3j>3f
zFPBgh@P?orT|;%JbR009^Cf2A9=b%qDxDoE6sMwd0<;vbu5aIVF>XuIOAsQANy9B-
z5)=ym6?}TQ(b9XQm{N6-*-5kbN&N_NSE&1-&v*Ck-7s);3fKWJga%VfOC18TY$aJK
z4n~+&e^_<x&rf*D5_aU}{#MH3k}xRh8A0lY@mTDjQ2Ob}B(dH9#y#E#N|Y?i?7<F3
z@(SF)mZc;<#u=w_O6_{ONKs-5s)zBVzVPurTWV<&)M86EbRBFyXP*~4^<WwpuyNF_
zPt=8j7C7=@I5TZJ3N>_crLmppZxB_-_f#W%e0(@)DP>)YZ@-GPx@y%b4l+;b>ydZK
zEGPYt@vv+>!Q&$(NTY)77nvqIIUY*tFXj`jEheHeGWuwxEmx!H3Q0*_VZZ!m?kppv
z;kB`;)gDG;8RE7<cL=dT&nGP{9kEvFjsT^M1{<XsZzjP-{FacykCL6U6BC2jDP?My
zQG+H*KE`{?78hk?Ybzl5>1K3vdT#Db@@DEQR8=Y7+*u*-s9U?T^?v*?3NcKT4ThIk
zL)f90=cq%=ky@6~*q!bO#hrzmo(7UZG|~B|iVFFCAe(tbjl4%)U7ay_j~IU7bWzFg
z19?p+4>{S(&zW0^=Mb*L=~Zj1#r*s{Fn;5SzduLSdNj<o0%ngfS<eaoyahX|gv00`
zLy*D<<lFH7XM5?W-!5C<7`$jHkCL-laOOdW0mV|Ef;Ujlc}D^^-qG-hx`W$@eJsu|
zBqW6K<-<phE{l@pM*|Nv!2X+}lHVZc9NjntHS(P5R1VDIQw2ALKeU`%mGC}q5y%BL
z-qsO9lk&O%UZTHYac<Nl@~<a&<v2DLKQM%e{crL7AuenFnzbA?xl2<N7~q)aFQ5ZK
zABzmVXis54Mj-^PE&eodB=7Lw-Cwa{J#cQw)iH3j*g^$@e$P96ezHmM6BU3aZsWPD
zkdc~tfiIsXhz{R!XIy6U{O8Z(WMGPOpLp4X14SRekiws|C!7l#O`;CW^l3Ch&>NH^
zWrBYG{`j?WYV6l9g@+e*k==8;OigK&RHRS}TYb8*xO&cZpV5tQc1ohgVwC*%L$pJ&
z?Qh=-P`91YwCXc_PqiKYxbwJfvx|w2)*zmqdhO~}cJ!83gcZ<ip(xOjWpo{oeGYBl
z3St>%+RZEQy#|sL&cnfn(6?PrhRli1N(Q}wwKs5(pE0qr0voJTyFlHby!xa)%<8qD
z&)BWx=H{fQG|-UOe0}3Gan?ywlSz#_VK_6-T}n_90#v6y&aJPnFAP5j;<a0D^Bf%l
zc8pEl$CW-FhHb<eAL2>=j4y@Ilw+qPvf*GAj516z@f}IY2x1}n8w3RfL4j;yB#X5o
zizPZPoZ+H#N1*sszk2e|h@CnmXa{V$Q{TYA01YR4LDy~gfS{JXw!*D~pPK4`y69;r
zrdUl-1+}Dm^<Id4fFhcb>*+NSRxBl#QoP`nWnH&)6vROeqAicNVERBBil;{e1Sw^!
zkpDZ~t%^<x(#JyMw&-i5adGe7J$gZ+V8VhIKLicbYGjG92WV4b;pDJ0x@#cfRQ%sK
zz-#@V9&b>~(I)|2(%{eCD0ITmqX@xWDH0&8-M>A18Zi5dbWSe`Q1|2U<HunF@0{{Z
z<niMHU@)8U5I)^FV*`VGN@ph6$w4|jK*0+RJ-O)JkwfcjKv4Q|)96bVtbFeZ_G|G#
zevOYCqwCG6{t?U_6AdK`<{sqB!Nv73Cr1WOfO<du#f!Mbk0fLxz2?m1aP$1!3=<wV
zO)wI3qi=M4{QH~4$ek|x`wbET6^CGs9nyCzI@zbrC}D)%)7K_7C*#YWdOa0ms->t(
zk}x8#BlO9CZoIkb!v|qvx0WU-D_0^0vR@}XWI|d6S!_oG{pf$U?tM_6(LpJrs_#q!
z-F{Tz-GhVs*{^G9Yv-LBG!Mc1%j^*~Tth5h6UvRo)gGMUG&nW6GZV*2cbYU+S2ipo
ztNb=Nn2B1NGt8xKe(3pdIzkm5qDyf<h!(8?eJq;9{ZN|F)8yxX`#>M(!Dutsk~^}F
z*N?uxll<R*I<)E$WCf;RZ<S&aKRrlJ?!)3XQCHD~w=&(k^g5iIlh)GO!{aewUV%w%
z{7?^TdbfMU5HSro&K!45vKaq+P>yS|Z_|#JF!1o$tcIC%UN5hxxILPAsNx)J%QF`~
zwX$SF)TX7SanJcWI;IiU{rON)_?*87bqHfy+eE@Ivv2$S`SWyV=yFD4ZZV|02x6x$
zX-^m>X8iN91jp{GD7gTuck?9$D`7YiMtq;3Oim67p`~u%CRQF9)uT^u3mhj(N%#l{
zy=4tPYYcjbz`Xh4PhYsXcflw-HvMH!j^m)l-|)FntxoFQp~vodZA-W$&G8I+cTr+4
zmsf8+iS>)m2j-MYO92}vc7QKSLpzn;;5E9I_typd2^PRy-8FZeV-XA!BXl3NkJq^Z
zr7uylrQ%P`g7n1X9%Jkl?<-dMHdRLMTZ-ON7I>&Pwra(S6%7wGIYUwpHKrR+<2)H@
zbtUEHP0_zH1d+OHq4ei4GYnRhPRa=Ux;DEpzv3w!Xs&3fTb+a!W_qhzX}n`G^Um=G
zLny~Ls+EbNZ{UdL5X8Rj#dE8F4i9u7Hpwyq!nX%5%+Y%U>#58cp)0;*aS+m|DJui<
zwjN;v#wwv!*2ll&<UYdoLUI{8>w_Wgk|Ayd%uT03T7V%4@x1p6zaX(2BXQspP9YsI
zjE~TiUO?BFiB|V81Or58;sxa#!<^^mFU@l>JO_QWJcPoSzxVbIbe&&kqN~ehhanQp
z^V?6wl6d-+7H!KXO<7jj>@qJPDsVuH3~uI@TS_-GQ$)BRR<Ui{q7_+9A)(&|tlqnV
zATXn9Fuz?F!%st$GLFVNJ3R~HjMM8&tL`ZK7ino}vHbEr6ll{7D3bURHH{{)CYFJ)
z<Iwcz;116o7BY=aZo;8?4Yl1CkkU6Xc}Uc^z?NY&c-;H@#$3nxT*yO=jf{APhAZbw
zvchpPq2ZS=H&6-KMzhj@;(6$79;c9-9|14HcVqHsRzPk60w%(74-Xnqin0N-r_qG3
zs}20%IWrS5-e}4%%Z(hHXwc^U#UH&((5>Gf5sM*xiP6y$<ru8Y2Mk#{Nc^(`=^nFx
zJoew+c@QoY7OY{2;7MAH*jCq(*r!%Z8I9@aYB_<f<lnhCEJ`!Sp|(P&j}MsajJ`oy
zk$<!%&heJgT?}l?Pt>E3lB%*dDu6*4DqRTJBrMFyz-x(S8^~~i2Uee8KHm@YLFju7
zO8lq}vXE#ui5NnIJGbOIk`@Rap-nLw{q?I3J*?!QAWC_vhPCxwBESZ%c2TBQ0LPGf
z<lTyhHHsz>cX-_?%?Im^m6|SLU^u~Ug;r9Kh8!AK5j>ki1f;&6Ul&k|Hrh$lqGS(L
zli$AGH`gs4D0m^9=&Fgp1dJ#LLx1DK{G<vQRXaw?tBB7%a-E7b5?rqUwapbV6ufk;
zKImn}uj|1YW<b@aM<Ob9y8X@B2@q6ShZJ1ER=>X^r%lWcp_z%EE$YUN`%rd}CM~|<
zSyUPbla{k9r)PUWx_-4kAQq8DpH6B+VVLKcYeW=HqzEB99-C)EOT!HJ>ZFBJD_eQK
zm!tP!B91W*+2(@b(W63ki8afXEz3Rq>kw9}jt8JTHdH<b9`k*?qA6^XU!QvqvboF`
z_oE-~zL$^F7Lk^2hI)pmtTTY(kxf8^e+ih$6w<ksh!$HFc-3VTif+yEZO*W0r;Zof
z?JUcYD&$<-6#UUA4ht4PfNmU`@y-ypkj}B!^q6CcrJjWZSMf+W^D<V}?!+P)IM_Lk
zb?dTI?)qN(IJu={e%d(#+xgskScs)zUt^N@^jKfZb8qQ;Se!cgzC?MuEUsEaM5LBz
zK9(?ZrJ~uVs=on8W1i;}MearvPxoi&D*1iu?M>mfKCOnUG~dgLNR@Z)`mJ00`Yh<s
z<tHhMI8Rzm&Q(=?$xGI_IJFmWig+BM1kt6%{u0f_HkX#&JO!zU{BuwlH^h5)q)q+q
z1?H(tTNerWB2k1w9O3TXPGE3_rjAYnGTi>YGpn&<M3I5s#Te-eG1e%`i}>P~3#b5g
znHRX<Bk||k7ErJ0pl<pv3g6W~b!uC1;sJB>J7k%2m|kPGbm^l<k0eq8b=Iz3O9UrH
z-orICr#KPdS<-Au1BeJ5eO9!guf@k_R;~-Q41^TBfa7eeqQ}>;Xla`)(4Pm!Xepf!
zNw={|JX>VTmg_Q3wor@^!D*gTqZnzBT;OZPl^KE`@Oq;6GDJSs-YEr<15R6P^X6-z
zKQ(v~=xoiHn3$-_b6h7#0~}D5Iy^T$PA*KjJ|Emhw7bieFBe`ZuMb5Cl*86UY*RAT
z!-p`&4uf%(0sv&@mTG{fq5jhlh%rEq0~<=@Ne!Ulp_UOKegDgR3b>erz?>oH+~fex
zTWa#pUk9kNmPdXptIaB&FsF*3A<D7pS<^HP`YRoBB|V%YX}Zawf1QW8VQZ;>^youa
zDJ*7sAMWi411`JY1$skwTc2pW4j39TJz!v4yE9cd!S4XNjZHW-S?Ehh%j$ACVsahi
z6Jzg$6C8juq!d4|L2udTcaJDiL+H>sBr6_kKff2z5(ksW5s*xEm6x~>09NkN_q+rx
zpr|l`V2tt2JK$$%GWJiyw+i1OqZK?tn~@kTRpO}Fpbj|}Y28a&6&Y>NxLk`CbR+oN
z5RXH!D3$p?_mvlBJz?*n($XK2*STS}vVrq{&rW$9hE>F=lg>E89-6~@zxe%T4Q?oh
z#4XpgeaHBZx?Ske8epbi$6qR3L`8<oV1QReK=SwB1VDbG37m}Fc!1aFkOuHo^S8L{
z_QbwA7dnr2-+@ft?I$)axpQMpDY>GCWwaD{+&^*RMB7)7%Ga-{!Y+M%cHv+m#Ojot
zotvZHJs@fcE@^8WfxKL=0TYNKjqB>_cok1Qf=XtGkMF2U*#)xu;{(kD02sz6PHYKA
zF|lKF?E5<k9!Y4Y6KqNfb}dO*P5<2Fm*%P{ae~^2Y;p7E&Gd(5usH-s;h>N?g&>gK
zWs=dDxuu}%9G|Rbj}AI3?&<~Gt^4TQ#e{BXqI@&0ys`$yCBc<p+DY`%)i(1<2#G}A
zT{KiQuoC?<nFxfJ-H1j_$dxI;d7?^6@%orrh1HqCJfck{L(kzP=?`dQ0+2W&>m0xi
z6GfnPv9Cdv+<Nc$uf<oP<s|71B0+aKVItjgKS1P+WCX}?VqU~L?Rxw6Z7mx0ayU${
z!4_lm1OF-MQtpEB5P={OUCajL^jx(6qN1Z$Ag82yqisUO3nXL^B@O+Fvo-s#ZFBs%
z8@EL=r^g3Ua5d|o1Xp?5UY%d_2Ox}*#a^P~@(f8N)oqIPqxrevz-B14#%dIY3+Cjf
zi$3A@+o_?pw6o0IVo(>Nt89AcP!bdr5>TJ1dfkkVx4Qic(l^r1sxigy^l*S{j{UeF
znU0j`3t1b)O*$f`KhR>H#&wi&$Xbvv(vXFtLRUh}v|eE8qx*&j4xGgM*C@EF!el*L
zMLYW2PPMw+*<ybQ)jtzmFDZ>1@THu6)k-4p0*;ieFJZeH9sLk;4+|pr2Z7R%ZKDFa
z(L>MD1m?)}V+$ygdr-5X+qjPy1f;4ruU%6^-BRZ6TGS^*`mupCtzQZ@KcO?rG$Ha#
z(wird)+Ezhbn|8d1hm6fziz?OVQdd?-du@3p97Q#<SI$`&>ngQhW^KuYwMw@-^-KA
zg*}YHg~q)^#x`6$NxcCe73qET#fzOpE=b%Q6<DUS|Fj6kMosF}UX*7cZO+Wh?1f@{
zjk2E^5mlobPPPLrd5^DvA1S23M2K;O(A2l<EH@fpdqmhy;Kt*c?I(6awV<!Bf0Y3G
zd^a1f{sv9R7hPi^GrkD<x8?S{=j3ujYrY<M(Vhos!B7Yc;Ht*c^73*<I%=x7Kkeh=
z_}{I~yfU^yB!#(resWfyKQ%Yk2vWRen9U|l;8p|Rm<F`2sv=RV6g2Fog_7kY03k4c
z9~9_d+SB89+B$*nE4WTW$B=58ZB6tb9sZM6&l1wldv*T}fSLy|Texu}_6f-{71}TW
z0I!RcPUxz{#H9TE`SZ8(lP6Eeos2cO$b=9Xw=xeN_=sF!fY?Nc#ge+woKT@4>HmQC
zq)uD9_yGt2^|Yl}WfSB?d2rCG52S3u7fzp_B*QQ6Vd6-62at0i1(E&lOEH#+43D-_
zEi|M^u8*oBQK*7>&;c&^c?^$$XuFh<kg(QoM2$#5o1i<(Abf$A(!1*F`g5#&xlo!`
zeOR(|DH=!p+|GT#iTKRbUx%%b^8k$d$*o#s-zQI>iU;0}iZb3VVPInNM(q^)^5x4(
zhP)$V7lO#hL@!6Uu1zS$QF-Vf98mRgKp7)aC+MImIRJ?)bmZbH$MVeWx7=RI&YlRu
z3cie}r33J2h)qSf&MN8`Ll0~>IxC|S6A*Z%Ky?Z`+5~p?V5SKqa7ffT_y!ck^~mO=
z&aC=iI_G0&cMoOhhnYW+2a7XzBqgn9<ARs$dG+emkGy8<dZM#Mb^i2}lfd@v+pFHa
ztAo!6{*nuSodja@phPr&IYwh%Ze*lh-hcz@J@QG|U8E*4QPH$nAJpHv3CjK*z0`k)
zl_m`!tp=Jw=k<3{It>nKKPf#=o|LsqG(i$A$i=Ya_9PXVKrTw~wicaY{_9<8YQey-
zRPu@jT@Zz*pC6Ii!LW=bQXCV(o{|Zlfvam48h=dX`-zAKIJ|bheEatU7Cra6PNSpY
zUHimdo9N)k9j@ugVLbtY^I6Gji5wcTTcUI(=wq3_9uA0z6N^z(@@_lMV=tY>15eLi
zv1Uys+_De*V~9(jOoT;B4F=mm{WuIkZrT{k(z1Prl9F6$%`nYw0-nl1g`%z&PsmVh
zaAI{tW=PO0Wb*X17^6XT-@XfLeSZj2KM+#Uol2iYR}>t8z>efelWglu%^O0i`!$-~
zU3YG!b9eTi9y5UCdmOS}D&WCQ!CfL4sLY;}PN<Q1<(zI3M<hx;qO#&iWFTtfL$4S~
z&V){qp7uOR(i1vDH+(0DG7%%hB_;Q-^BW0Pp8m06NP|=?mKspE6B!c3mi<J5i4_uI
zjEt%r5z?aOZ$D8jxw1Ib6xc(5kCD+$vIa;MF}->{A}|A)$U#POX9CAgG_dEEuge0N
zp++_alLq6ntUS0iJR+j;?X?YZ(Bw>%YJ`HG3<{Sp8wQ+bML^?|cWpqzA4lTnTpSE3
zyx4dAK(=);B4ctOu0v1b=0j9$thR!?I2wM$G6--g^gKqZM!Kaz+9M+)`s(T-2%5x>
zU|W4;L8LJcpQjXr0T7Z>{o9)e>SYJ0u`I}?0G@(bpM}^jvl^vygKn#Md9?%_R+7c;
zLYY~I7o8!T84|P6fdkhGUPYxqe(xbgs=TstFSweztAe|bF*9*z3o<uNEnQt9{sMXx
z6&0(%xp7`Rpz-V1uiv}6)I_LFCw*-IeURRUkOz~k#E0Nbn$x)GXl!Dlq3WK_ZG(JF
zE_4aQ<?N3hB?6YJw{P`*eTyJr+nz#03EG8bywmr$*MQX?;Cth^J4htr5b996%F>|F
zdJUl;*eIjYfH~NywulUAqomTI6-uBgxlI6vBgL3M?)HIKi$%|k=B4|K7cXvJy_%>N
z%Fss?%y3@m=`-lZdwqSyWPjv{QFK|aRu_P-n6z}HDjrI|im(;9mUX}K+6`|!jXOhg
zx0=zJp~Af+;o(YW8m>pU-eK7le(_>mgnUOT1{Q_-Mb&m6Hn>xOhK7b*7xF^!h<gF1
zUji43x4E{Zf^~c{NXHw|=oE!m2R1JQ?<2(!dQ#bSfzI1vS1ppWUc()8+&0n@5=OY_
zXfNoOwNhDrlclwF4LKgUkGR5v^1!4Nmj;Op9k9<EZgYb$day}EWp(xYGO_Rb>+YXl
z!O0ngg##U`2X_&AT?y(fa(3s~)}%jqvW477BsPldaPmj3Agy}w?_a-&dp4E^29hm>
zlAUc~D`Ltj#E&08e%Sa8LaA6)Q=^>CqHM~TlAJ86qLL7bXp+8)lk@(qTRiOu9$=!H
z!7ApGi$nnGpe*Ube0x#XhK11+%mcC9{^Cc;!PBIdKK>u{Ti_u~amz~1XY@#NN{TbV
z=Q(zv0e`xd61=#AKM8zyQO_Och7`SZUJzm;yhP-ejf@$=&B3HLLc?a`SDcvG$!+hh
zp>YmZ6K>+R4Yq>~u?06n5*b)|A7Hz|c8P7;l>B729ruNBWntGo_X`lUjt{n)0zDbx
zPSJO;&LZySILO*El%6D15Ko3C3m(u!KrFe&)>dc;M+^(&eQYLiDKgR&>7pUSR63<)
zmxO>o;tAhgbwG_Bps>16aiB#4J5Oy>EXc{x2gR+eh+=xH(2caw{v_P%&cJs8B<qo*
z@^C$Nu>E-qEsq?2Mf$zFN@vo+BldR`KPLcai4<CB6y9su+1XHVCj+HxQrU3=nJ^!M
zv4I&2#@0k3xe_9b+-m`{figWBw-Bn8bdgvgj2GLo<w;96jw8XFv(VFiq|nB*VmHu3
zXlZFFF_>GoL^P2IA^w`lb#Bh88kd#7{dzmPABdqb(8#3~6cmJK$<dgS#yZ1SU_@N*
z%kb>2A(0#4lSC<aZN>>&1qibe+k!qHH-V6x2$?^9GJE~z%@dJ4h=kEhO~ehTKjGOp
zVx;(tmcCBtmr3(i08yw4Q3zcLV?rh&yJ>y|jDGO_`w@(+4PNSimew-0LK>v>Teogi
zT^xbs<sJWg-L`WZ5^`A?Q6>GoGe(}o@!)BF1gc@gF><Ylcnv|>fEM&UuduZ%`S>gj
zj^K`Qb#rr!t<yU+R)M=f>8DQ&phHBW6g;te_Xv)Q&#M%Q2K_)>NFXeIKf=+?jd0w4
zRPN-iIqE^P!iPQk?Rs%V%Ne3NHI$4_UrLI}1r`*{A;A{NUKC+Y9-dU(DM`fvM<qab
zVgM4kJ7nCF<G=F8(>OppXCoWOnU3dh5a0~~>HGLVKAAwQ7`q6juZ$HJ`Al_b5)K24
z-oAbCAZ>zp2rl1#7OPT4StAzS?%C}M!vY~|SWY8l`(*Uz5{eWYfZ#kWJ-ucu20L#u
zbr{4Aq4Lt!BK(rN33sNlBu=&FI^G99;6Q&mh*Lq{$ip=})w6XEd78rvLj!3Mlf|Ki
zr7jM$1!5(d2{wH?5(UXfNFQ_ITVMsdz+UwBaLHMQY3$xDvUBH)(Mr&^;t~=%P?#PI
zM_8k9o3#7W)%b(Czsx6Xb99YA7+-N&StGn4d)5meLw(mSNJAwcbq%5I4}vF*FvA5Q
z;H*cEzX=i&9bh1|0o-7Q;}wu<3PeGS__^ZZ;#Fw<Qqf;$RrNwfKv-Cqh=j!DIWGiM
z_s`=eb#<>m5Pxv$T5POE;Tz@rho>gQhtb3a!uanfDL%xi+dNh7il4v}i3z2_upx+n
z7=OVbBJ={_1tP^|y<{-humoIpOdb8*8z3Pff==Z%Bt<WRTq6;0Q8Q#bdi3+fhxi#E
zWAK;I-bJbS`uM!7sMrM<O#Ykj8+|*(zrbGx;}~S9w^tZ5hqen?-9sM}u1!qOx~&(#
z!?-Y785NXIX;#r2d|}jmXR6J-=*5d!?o#E+Gnm!J_+rsq1PgG6>({MIsESN4otT(V
z9{z0{Xw^Q<v}{>A{~ch9&LCRqVq|jhjGwm^6``*;;5gjRVg%f~6GXBur?e4{aPdz8
z<D2BaUzY6wEL{9L=%4%_|K)~;;qNQ{-7vs^E$RPXLm+D}{&DAktt|fOTo2D({PO<}
g{Qq$)UM(zr+8w&FQR4tgTM9*UkM8a)b=#2t2d}g9z5oCK

literal 0
HcmV?d00001

diff --git a/beginner_source/hyperparameter_tuning_tutorial.py b/beginner_source/hyperparameter_tuning_tutorial.py
new file mode 100644
index 00000000000..7f1dda1287b
--- /dev/null
+++ b/beginner_source/hyperparameter_tuning_tutorial.py
@@ -0,0 +1,454 @@
+# -*- coding: utf-8 -*-
+"""
+Hyperparameter tuning with Ray Tune
+===================================
+
+Hyperparameter tuning can make the difference between an average model and a highly
+accurate one. Often simple things like choosing a different learning rate or changing
+a network layer size can have a dramatic impact on your model performance.
+
+Fortunately, there are tools that help with finding the best combination of parameters.
+`Ray Tune <https://docs.ray.io/en/latest/tune.html>`_ is an industry standard tool for
+distributed hyperparameter tuning. Ray Tune includes the latest hyperparameter search
+algorithms, integrates with TensorBoard and other analysis libraries, and natively
+supports distributed training through `Ray's distributed machine learning engine
+<https://ray.io/>`_.
+
+In this tutorial, we will show you how to integrate Tune into your PyTorch
+training workflow. We will extend `this tutorial from the PyTorch documentation
+<https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html>`_ for training
+a CIFAR10 image classifier.
+
+As you will see, we only need to add some slight modifications. In particular, we
+need to
+
+1. wrap data loading and training in functions,
+2. make some network parameters configurable,
+3. add checkpointing (optional),
+4. and define the search space for the model tuning
+
+|
+
+To run this tutorial, please make sure the following packages are
+installed:
+
+-  ``ray[tune]``: Distributed hyperparameter tuning library
+-  ``torchvision``: For the data transformers
+
+Setup / Imports
+---------------
+Let's start with the imports:
+"""
+from functools import partial
+import numpy as np
+import os
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.optim as optim
+from torch.utils.data import random_split
+import torchvision
+import torchvision.transforms as transforms
+from ray import tune
+from ray.tune import CLIReporter
+from ray.tune.schedulers import ASHAScheduler
+
+######################################################################
+# Most of the imports are needed for building the PyTorch model. Only the last three
+# imports are for Ray Tune.
+#
+# Data loaders
+# ------------
+# We wrap the data loaders in their own function and pass a global data directory.
+# This way we can share a data directory between different trials.
+
+def load_data(data_dir="./data"):
+    transform = transforms.Compose([
+        transforms.ToTensor(),
+        transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
+    ])
+
+    trainset = torchvision.datasets.CIFAR10(
+        root=data_dir, train=True, download=True, transform=transform)
+
+    testset = torchvision.datasets.CIFAR10(
+        root=data_dir, train=False, download=True, transform=transform)
+
+    return trainset, testset
+
+######################################################################
+# Configurable neural network
+# ---------------------------
+# We can only tune those parameters that are configurable. In this example, we can specify
+# the layer sizes of the fully connected layers:
+
+class Net(nn.Module):
+    def __init__(self, l1=120, l2=84):
+        super(Net, self).__init__()
+        self.conv1 = nn.Conv2d(3, 6, 5)
+        self.pool = nn.MaxPool2d(2, 2)
+        self.conv2 = nn.Conv2d(6, 16, 5)
+        self.fc1 = nn.Linear(16 * 5 * 5, l1)
+        self.fc2 = nn.Linear(l1, l2)
+        self.fc3 = nn.Linear(l2, 10)
+
+    def forward(self, x):
+        x = self.pool(F.relu(self.conv1(x)))
+        x = self.pool(F.relu(self.conv2(x)))
+        x = x.view(-1, 16 * 5 * 5)
+        x = F.relu(self.fc1(x))
+        x = F.relu(self.fc2(x))
+        x = self.fc3(x)
+        return x
+
+######################################################################
+# The train function
+# ------------------
+# Now it gets interesting, because we introduce some changes to the example `from the PyTorch
+# documentation <https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html>`_.
+#
+# We wrap the training script in a function ``train_cifar(config, checkpoint=None, data_dir=None)``.
+# As you can guess, the ``config`` parameter will receive the hyperparameters we would like to
+# train with. The ``checkpoint`` parameter is used to restore checkpoints. The ``data_dir`` specifies
+# the directory where we load and store the data, so multiple runs can share the same data source.
+#
+# .. code-block:: python
+#
+#     net = Net(config["l1"], config["l2"])
+#
+#     if checkpoint:
+#         net.load_state_dict(torch.load(checkpoint))
+#
+# The learning rate of the optimizer is made configurable, too:
+#
+# .. code-block:: python
+#
+#     optimizer = optim.SGD(net.parameters(), lr=config["lr"], momentum=0.9)
+#
+# We also split the training data into a training and validation subset. We thus train on
+# 80% of the data and calculate the validation loss on the remaining 20%. The batch sizes
+# with which we iterate through the training and test sets are configurable as well.
+#
+# Adding (multi) GPU support with DataParallel
+# ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+# Image classification benefits largely from GPUs. Luckily, we can continue to use
+# PyTorch's abstractions in Ray Tune. Thus, we can wrap our model in ``nn.DataParallel``
+# to support data parallel training on multiple GPUs:
+#
+# .. code-block:: python
+#
+#     device = "cpu"
+#     if torch.cuda.is_available():
+#         device = "cuda:0"
+#         if torch.cuda.device_count() > 1:
+#             net = nn.DataParallel(net)
+#     net.to(device)
+#
+# By using a ``device`` variable we make sure that training also works when we have
+# no GPUs available. PyTorch requires us to send our data to the GPU memory explicitly,
+# like this:
+#
+# .. code-block:: python
+#
+#     for i, data in enumerate(trainloader, 0):
+#         inputs, labels = data
+#         inputs, labels = inputs.to(device), labels.to(device)
+#
+# The code now supports training on CPUs, on a single GPU, and on multiple GPUs. Notably, Ray
+# also supports `fractional GPUs <https://docs.ray.io/en/master/using-ray-with-gpus.html#fractional-gpus>`_
+# so we can share GPUs among trials, as long as the model still fits on the GPU memory. We'll come back
+# to that later.
+#
+# Communicating with Ray Tune
+# ~~~~~~~~~~~~~~~~~~~~~~~~~~~
+#
+# The most interesting part is the communication with Tune:
+#
+# .. code-block:: python
+#
+#     checkpoint_dir = tune.make_checkpoint_dir(epoch)
+#     path = os.path.join(checkpoint_dir, "checkpoint")
+#     torch.save((net.state_dict(), optimizer.state_dict()), path)
+#     tune.save_checkpoint(path)
+#
+#     tune.report(loss=(val_loss / val_steps), accuracy=correct / total)
+#
+# Here we first save a checkpoint and then report some metrics back to Tune. Specifically,
+# we send the validation loss and accuracy back to Tune. Tune can then use these metrics
+# to decide which hyperparameter configuration lead to the best results. These metrics
+# can also be used to stop bad performing trials early in order to avoid wasting
+# resources on those trials.
+#
+# The checkpoint saving is optional, however, it is necessary if we wanted to use advanced
+# schedulers like `Population Based Training <https://docs.ray.io/en/master/tune/tutorials/tune-advanced-tutorial.html>`_.
+# Also, by saving the checkpoint we can later load the trained models and validate them
+# on a test set.
+#
+# Full training function
+# ~~~~~~~~~~~~~~~~~~~~~~
+#
+# The full code example looks like this:
+
+def train_cifar(config, checkpoint=None, data_dir=None):
+    net = Net(config["l1"], config["l2"])
+
+    device = "cpu"
+    if torch.cuda.is_available():
+        device = "cuda:0"
+        if torch.cuda.device_count() > 1:
+            net = nn.DataParallel(net)
+    net.to(device)
+
+    criterion = nn.CrossEntropyLoss()
+    optimizer = optim.SGD(net.parameters(), lr=config["lr"], momentum=0.9)
+
+    if checkpoint:
+        print("loading checkpoint {}".format(checkpoint))
+        model_state, optimizer_state = torch.load(checkpoint)
+        net.load_state_dict(model_state)
+        optimizer.load_state_dict(optimizer_state)
+
+    trainset, testset = load_data(data_dir)
+
+    test_abs = int(len(trainset) * 0.8)
+    train_subset, val_subset = random_split(
+        trainset, [test_abs, len(trainset) - test_abs])
+
+    trainloader = torch.utils.data.DataLoader(
+        train_subset,
+        batch_size=int(config["batch_size"]),
+        shuffle=True,
+        num_workers=8)
+    valloader = torch.utils.data.DataLoader(
+        val_subset,
+        batch_size=int(config["batch_size"]),
+        shuffle=True,
+        num_workers=8)
+
+    for epoch in range(10):  # loop over the dataset multiple times
+        running_loss = 0.0
+        epoch_steps = 0
+        for i, data in enumerate(trainloader, 0):
+            # get the inputs; data is a list of [inputs, labels]
+            inputs, labels = data
+            inputs, labels = inputs.to(device), labels.to(device)
+
+            # zero the parameter gradients
+            optimizer.zero_grad()
+
+            # forward + backward + optimize
+            outputs = net(inputs)
+            loss = criterion(outputs, labels)
+            loss.backward()
+            optimizer.step()
+
+            # print statistics
+            running_loss += loss.item()
+            epoch_steps += 1
+            if i % 2000 == 1999:  # print every 2000 mini-batches
+                print("[%d, %5d] loss: %.3f" % (epoch + 1, i + 1,
+                                                running_loss / epoch_steps))
+                running_loss = 0.0
+
+        # Validation loss
+        val_loss = 0.0
+        val_steps = 0
+        total = 0
+        correct = 0
+        for i, data in enumerate(valloader, 0):
+            with torch.no_grad():
+                inputs, labels = data
+                inputs, labels = inputs.to(device), labels.to(device)
+
+                outputs = net(inputs)
+                _, predicted = torch.max(outputs.data, 1)
+                total += labels.size(0)
+                correct += (predicted == labels).sum().item()
+
+                loss = criterion(outputs, labels)
+                val_loss += loss.cpu().numpy()
+                val_steps += 1
+
+        checkpoint_dir = tune.make_checkpoint_dir(epoch)
+        path = os.path.join(checkpoint_dir, "checkpoint")
+        torch.save((net.state_dict(), optimizer.state_dict()), path)
+        tune.save_checkpoint(path)
+
+        tune.report(loss=(val_loss / val_steps), accuracy=correct / total)
+    print("Finished Training")
+
+######################################################################
+# As you can see, most of the code is adapted directly from the original example.
+#
+# Test set accuracy
+# -----------------
+# Commonly the performance of a machine learning model is tested on a hold-out test
+# set with data that has not been used for training the model. We also wrap this in a
+# function:
+
+def test_accuracy(net, device="cpu"):
+    trainset, testset = load_data()
+
+    testloader = torch.utils.data.DataLoader(
+        testset, batch_size=4, shuffle=False, num_workers=2)
+
+    correct = 0
+    total = 0
+    with torch.no_grad():
+        for data in testloader:
+            images, labels = data
+            images, labels = images.to(device), labels.to(device)
+            outputs = net(images)
+            _, predicted = torch.max(outputs.data, 1)
+            total += labels.size(0)
+            correct += (predicted == labels).sum().item()
+
+    return correct / total
+
+######################################################################
+# The function also expects a ``device`` parameter, so we can do the
+# test set validation on a GPU.
+#
+# Configuring the search space
+# ----------------------------
+# Lastly, we need to define Tune's search space. Here is an example:
+#
+# .. code-block:: python
+#
+#     config = {
+#         "l1": tune.sample_from(lambda _: 2**np.random.randint(2, 9)),
+#         "l2": tune.sample_from(lambda _: 2**np.random.randint(2, 9)),
+#         "lr": tune.loguniform(1e-4, 1e-1),
+#         "batch_size": tune.choice([2, 4, 8, 16])
+#     }
+#
+# The ``tune.sample_from()`` function makes it possible to define your own sample
+# methods to obtain hyperparameters. In this example, the ``l1`` and ``l2`` parameters
+# should be powers of 2 between 4 and 256, so either 4, 8, 16, 32, 64, 128, or 256.
+# The ``lr`` (learning rate) should be uniformly sampled between 0.0001 and 0.1. Lastly,
+# the batch size is a choice between 2, 4, 8, and 16.
+#
+# At each trial, Tune will now randomly sample a combination of parameters from these
+# search spaces. It will then train a number of models in parallel and find the best
+# performing one among these. We also use the ``ASHAScheduler`` which will terminate bad
+# performing trials early.
+#
+# We wrap the ``train_cifar`` function with ``functools.partial`` to set the constant
+# ``data_dir`` parameter. We can also tell Ray Tune what resources should be
+# available for each trial:
+#
+# .. code-block:: python
+#
+#     gpus_per_trial = 2
+#     # ...
+#     result = tune.run(
+#         partial(train_cifar, data_dir=data_dir),
+#         resources_per_trial={"cpu": 8, "gpu": gpus_per_trial},
+#         config=config,
+#         num_samples=num_samples,
+#         scheduler=scheduler,
+#         progress_reporter=reporter,
+#         checkpoint_at_end=True)
+#
+# You can specify the number of CPUs, which are then available e.g.
+# to increase the ``num_workers`` of the PyTorch ``DataLoader`` instances. The selected
+# number of GPUs are made visible to PyTorch in each trial. Trials do not have access to
+# GPUs that haven't been requested for them - so you don't have to care about two trials
+# using the same set of resources.
+#
+# Here we can also specify fractional GPUs, so something like ``gpus_per_trial=0.5`` is
+# completely valid. The trials will then share GPUs among each other.
+# You just have to make sure that the models still fit in the GPU memory.
+#
+# After training the models, we will find the best performing one and load the trained
+# network from the checkpoint file. We then obtain the test set accuracy and report
+# everything by printing.
+#
+# The full main function looks like this:
+
+def main(num_samples=10, max_num_epochs=10, gpus_per_trial=2):
+    data_dir = os.path.abspath("./data")
+    load_data(data_dir)
+    config = {
+        "l1": tune.sample_from(lambda _: 2 ** np.random.randint(2, 9)),
+        "l2": tune.sample_from(lambda _: 2 ** np.random.randint(2, 9)),
+        "lr": tune.loguniform(1e-4, 1e-1),
+        "batch_size": tune.choice([2, 4, 8, 16])
+    }
+    scheduler = ASHAScheduler(
+        metric="loss",
+        mode="min",
+        max_t=max_num_epochs,
+        grace_period=1,
+        reduction_factor=2)
+    reporter = CLIReporter(
+        # parameter_columns=["l1", "l2", "lr", "batch_size"],
+        metric_columns=["loss", "accuracy", "training_iteration"])
+    result = tune.run(
+        partial(train_cifar, data_dir=data_dir),
+        resources_per_trial={"cpu": 2, "gpu": gpus_per_trial},
+        config=config,
+        num_samples=num_samples,
+        scheduler=scheduler,
+        progress_reporter=reporter,
+        checkpoint_at_end=True)
+
+    best_trial = result.get_best_trial("loss", "min", "last")
+    print("Best trial config: {}".format(best_trial.config))
+    print("Best trial final validation loss: {}".format(
+        best_trial.last_result["loss"]))
+    print("Best trial final validation accuracy: {}".format(
+        best_trial.last_result["accuracy"]))
+
+    best_trained_model = Net(best_trial.config["l1"], best_trial.config["l2"])
+    device = "cpu"
+    if torch.cuda.is_available():
+        device = "cuda:0"
+        if gpus_per_trial > 1:
+            best_trained_model = nn.DataParallel(best_trained_model)
+    best_trained_model.to(device)
+
+    model_state, optimizer_state = torch.load(best_trial.checkpoint.value)
+    best_trained_model.load_state_dict(model_state)
+
+    test_acc = test_accuracy(best_trained_model, device)
+    print("Best trial test set accuracy: {}".format(test_acc))
+
+
+if __name__ == "__main__":
+    # You can change the number of GPUs per trial here:
+    main(num_samples=10, max_num_epochs=10, gpus_per_trial=0)
+
+
+######################################################################
+# If you run the code, an example output could look like this:
+#
+# .. code-block::
+#
+#     Number of trials: 10 (10 TERMINATED)
+#     +-----+------+------+-------------+--------------+---------+------------+--------------------+
+#     | ... |   l1 |   l2 |          lr |   batch_size |    loss |   accuracy | training_iteration |
+#     |-----+------+------+-------------+--------------+---------+------------+--------------------|
+#     | ... |   64 |    4 | 0.00011629  |            2 | 1.87273 |     0.244  |                  2 |
+#     | ... |   32 |   64 | 0.000339763 |            8 | 1.23603 |     0.567  |                  8 |
+#     | ... |    8 |   16 | 0.00276249  |           16 | 1.1815  |     0.5836 |                 10 |
+#     | ... |    4 |   64 | 0.000648721 |            4 | 1.31131 |     0.5224 |                  8 |
+#     | ... |   32 |   16 | 0.000340753 |            8 | 1.26454 |     0.5444 |                  8 |
+#     | ... |    8 |    4 | 0.000699775 |            8 | 1.99594 |     0.1983 |                  2 |
+#     | ... |  256 |    8 | 0.0839654   |           16 | 2.3119  |     0.0993 |                  1 |
+#     | ... |   16 |  128 | 0.0758154   |           16 | 2.33575 |     0.1327 |                  1 |
+#     | ... |   16 |    8 | 0.0763312   |           16 | 2.31129 |     0.1042 |                  4 |
+#     | ... |  128 |   16 | 0.000124903 |            4 | 2.26917 |     0.1945 |                  1 |
+#     +-----+------+------+-------------+--------------+---------+------------+--------------------+
+#
+#
+#     Best trial config: {'l1': 8, 'l2': 16, 'lr': 0.00276249, 'batch_size': 16, 'data_dir': '...'}
+#     Best trial final validation loss: 1.181501
+#     Best trial final validation accuracy: 0.5836
+#     Best trial test set accuracy: 0.5806
+#
+# Most trials have been stopped early in order to avoid wasting resources.
+# The best performing trial achieved a validation accuracy of about 58%, which could
+# be confirmed on the test set.
+#
+# So that's it! You can now tune the parameters of your PyTorch models.
diff --git a/index.rst b/index.rst
index 9a8762e8026..8274715b367 100644
--- a/index.rst
+++ b/index.rst
@@ -253,6 +253,13 @@ Welcome to PyTorch Tutorials
 
 .. Model Optimization
 
+.. customcarditem::
+   :header: Hyperparameter Tuning Tutorial
+   :card_description: Learn how to use Ray Tune to find the best performing set of hyperparameters for your model.
+   :image: _static/img/ray-tune.png
+   :link: beginner/hyperparameter_tuning_tutorial.html
+   :tags: Model-Optimization,Best-Practice
+
 .. customcarditem::
    :header: Pruning Tutorial
    :card_description: Learn how to use torch.nn.utils.prune to sparsify your neural networks, and how to extend it to implement your own custom pruning technique.
@@ -479,6 +486,7 @@ Additional Resources
    :hidden:
    :caption: Model Optimization
 
+   beginner/hyperparameter_tuning_tutorial
    intermediate/pruning_tutorial
    advanced/dynamic_quantization_tutorial
    intermediate/dynamic_quantization_bert_tutorial
diff --git a/requirements.txt b/requirements.txt
index af5eec5392b..87f0fcfed1b 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -14,6 +14,7 @@ bs4
 awscli==1.16.35
 flask
 spacy
+ray[tune]
 
 # PyTorch Theme
 -e git+git://github.com/pytorch/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme

From f075938e0a09e28a10f0a19b757a21c589695a30 Mon Sep 17 00:00:00 2001
From: Kai Fricke <kai@anyscale.com>
Date: Wed, 15 Jul 2020 12:46:16 +0200
Subject: [PATCH 2/3] Use nightly ray release

---
 requirements.txt | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/requirements.txt b/requirements.txt
index 87f0fcfed1b..a211f877eac 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -14,7 +14,12 @@ bs4
 awscli==1.16.35
 flask
 spacy
-ray[tune]
+
+# Replace this block with ray[tune] after next PyPI release
+https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-0.9.0.dev0-cp36-cp36m-manylinux1_x86_64.whl
+tabulate
+tensorboardX
+pandas
 
 # PyTorch Theme
 -e git+git://github.com/pytorch/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme

From faccec1cb9cc593deec7c3378fd89e0760064948 Mon Sep 17 00:00:00 2001
From: Kai Fricke <kai@anyscale.com>
Date: Wed, 19 Aug 2020 10:58:20 +0200
Subject: [PATCH 3/3] Fix checkpoint API

---
 .../hyperparameter_tuning_tutorial.py         | 60 +++++++++++--------
 requirements.txt                              |  7 +--
 2 files changed, 35 insertions(+), 32 deletions(-)

diff --git a/beginner_source/hyperparameter_tuning_tutorial.py b/beginner_source/hyperparameter_tuning_tutorial.py
index 7f1dda1287b..11524618cba 100644
--- a/beginner_source/hyperparameter_tuning_tutorial.py
+++ b/beginner_source/hyperparameter_tuning_tutorial.py
@@ -14,7 +14,7 @@
 supports distributed training through `Ray's distributed machine learning engine
 <https://ray.io/>`_.
 
-In this tutorial, we will show you how to integrate Tune into your PyTorch
+In this tutorial, we will show you how to integrate Ray Tune into your PyTorch
 training workflow. We will extend `this tutorial from the PyTorch documentation
 <https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html>`_ for training
 a CIFAR10 image classifier.
@@ -62,6 +62,7 @@
 # We wrap the data loaders in their own function and pass a global data directory.
 # This way we can share a data directory between different trials.
 
+
 def load_data(data_dir="./data"):
     transform = transforms.Compose([
         transforms.ToTensor(),
@@ -82,6 +83,7 @@ def load_data(data_dir="./data"):
 # We can only tune those parameters that are configurable. In this example, we can specify
 # the layer sizes of the fully connected layers:
 
+
 class Net(nn.Module):
     def __init__(self, l1=120, l2=84):
         super(Net, self).__init__()
@@ -107,17 +109,20 @@ def forward(self, x):
 # Now it gets interesting, because we introduce some changes to the example `from the PyTorch
 # documentation <https://pytorch.org/tutorials/beginner/blitz/cifar10_tutorial.html>`_.
 #
-# We wrap the training script in a function ``train_cifar(config, checkpoint=None, data_dir=None)``.
+# We wrap the training script in a function ``train_cifar(config, checkpoint_dir=None, data_dir=None)``.
 # As you can guess, the ``config`` parameter will receive the hyperparameters we would like to
-# train with. The ``checkpoint`` parameter is used to restore checkpoints. The ``data_dir`` specifies
+# train with. The ``checkpoint_dir`` parameter is used to restore checkpoints. The ``data_dir`` specifies
 # the directory where we load and store the data, so multiple runs can share the same data source.
 #
 # .. code-block:: python
 #
 #     net = Net(config["l1"], config["l2"])
 #
-#     if checkpoint:
-#         net.load_state_dict(torch.load(checkpoint))
+#     if checkpoint_dir:
+#         model_state, optimizer_state = torch.load(
+#             os.path.join(checkpoint_dir, "checkpoint"))
+#         net.load_state_dict(model_state)
+#         optimizer.load_state_dict(optimizer_state)
 #
 # The learning rate of the optimizer is made configurable, too:
 #
@@ -162,25 +167,25 @@ def forward(self, x):
 # Communicating with Ray Tune
 # ~~~~~~~~~~~~~~~~~~~~~~~~~~~
 #
-# The most interesting part is the communication with Tune:
+# The most interesting part is the communication with Ray Tune:
 #
 # .. code-block:: python
 #
-#     checkpoint_dir = tune.make_checkpoint_dir(epoch)
-#     path = os.path.join(checkpoint_dir, "checkpoint")
-#     torch.save((net.state_dict(), optimizer.state_dict()), path)
-#     tune.save_checkpoint(path)
+#     with tune.checkpoint_dir(epoch) as checkpoint_dir:
+#         path = os.path.join(checkpoint_dir, "checkpoint")
+#         torch.save((net.state_dict(), optimizer.state_dict()), path)
 #
 #     tune.report(loss=(val_loss / val_steps), accuracy=correct / total)
 #
-# Here we first save a checkpoint and then report some metrics back to Tune. Specifically,
-# we send the validation loss and accuracy back to Tune. Tune can then use these metrics
+# Here we first save a checkpoint and then report some metrics back to Ray Tune. Specifically,
+# we send the validation loss and accuracy back to Ray Tune. Ray Tune can then use these metrics
 # to decide which hyperparameter configuration lead to the best results. These metrics
 # can also be used to stop bad performing trials early in order to avoid wasting
 # resources on those trials.
 #
 # The checkpoint saving is optional, however, it is necessary if we wanted to use advanced
-# schedulers like `Population Based Training <https://docs.ray.io/en/master/tune/tutorials/tune-advanced-tutorial.html>`_.
+# schedulers like
+# `Population Based Training <https://docs.ray.io/en/master/tune/tutorials/tune-advanced-tutorial.html>`_.
 # Also, by saving the checkpoint we can later load the trained models and validate them
 # on a test set.
 #
@@ -189,7 +194,8 @@ def forward(self, x):
 #
 # The full code example looks like this:
 
-def train_cifar(config, checkpoint=None, data_dir=None):
+
+def train_cifar(config, checkpoint_dir=None, data_dir=None):
     net = Net(config["l1"], config["l2"])
 
     device = "cpu"
@@ -202,9 +208,9 @@ def train_cifar(config, checkpoint=None, data_dir=None):
     criterion = nn.CrossEntropyLoss()
     optimizer = optim.SGD(net.parameters(), lr=config["lr"], momentum=0.9)
 
-    if checkpoint:
-        print("loading checkpoint {}".format(checkpoint))
-        model_state, optimizer_state = torch.load(checkpoint)
+    if checkpoint_dir:
+        model_state, optimizer_state = torch.load(
+            os.path.join(checkpoint_dir, "checkpoint"))
         net.load_state_dict(model_state)
         optimizer.load_state_dict(optimizer_state)
 
@@ -269,10 +275,9 @@ def train_cifar(config, checkpoint=None, data_dir=None):
                 val_loss += loss.cpu().numpy()
                 val_steps += 1
 
-        checkpoint_dir = tune.make_checkpoint_dir(epoch)
-        path = os.path.join(checkpoint_dir, "checkpoint")
-        torch.save((net.state_dict(), optimizer.state_dict()), path)
-        tune.save_checkpoint(path)
+        with tune.checkpoint_dir(epoch) as checkpoint_dir:
+            path = os.path.join(checkpoint_dir, "checkpoint")
+            torch.save((net.state_dict(), optimizer.state_dict()), path)
 
         tune.report(loss=(val_loss / val_steps), accuracy=correct / total)
     print("Finished Training")
@@ -286,6 +291,7 @@ def train_cifar(config, checkpoint=None, data_dir=None):
 # set with data that has not been used for training the model. We also wrap this in a
 # function:
 
+
 def test_accuracy(net, device="cpu"):
     trainset, testset = load_data()
 
@@ -311,7 +317,7 @@ def test_accuracy(net, device="cpu"):
 #
 # Configuring the search space
 # ----------------------------
-# Lastly, we need to define Tune's search space. Here is an example:
+# Lastly, we need to define Ray Tune's search space. Here is an example:
 #
 # .. code-block:: python
 #
@@ -328,7 +334,7 @@ def test_accuracy(net, device="cpu"):
 # The ``lr`` (learning rate) should be uniformly sampled between 0.0001 and 0.1. Lastly,
 # the batch size is a choice between 2, 4, 8, and 16.
 #
-# At each trial, Tune will now randomly sample a combination of parameters from these
+# At each trial, Ray Tune will now randomly sample a combination of parameters from these
 # search spaces. It will then train a number of models in parallel and find the best
 # performing one among these. We also use the ``ASHAScheduler`` which will terminate bad
 # performing trials early.
@@ -366,6 +372,7 @@ def test_accuracy(net, device="cpu"):
 #
 # The full main function looks like this:
 
+
 def main(num_samples=10, max_num_epochs=10, gpus_per_trial=2):
     data_dir = os.path.abspath("./data")
     load_data(data_dir)
@@ -390,8 +397,7 @@ def main(num_samples=10, max_num_epochs=10, gpus_per_trial=2):
         config=config,
         num_samples=num_samples,
         scheduler=scheduler,
-        progress_reporter=reporter,
-        checkpoint_at_end=True)
+        progress_reporter=reporter)
 
     best_trial = result.get_best_trial("loss", "min", "last")
     print("Best trial config: {}".format(best_trial.config))
@@ -408,7 +414,9 @@ def main(num_samples=10, max_num_epochs=10, gpus_per_trial=2):
             best_trained_model = nn.DataParallel(best_trained_model)
     best_trained_model.to(device)
 
-    model_state, optimizer_state = torch.load(best_trial.checkpoint.value)
+    best_checkpoint_dir = best_trial.checkpoint.value
+    model_state, optimizer_state = torch.load(os.path.join(
+        best_checkpoint_dir, "checkpoint"))
     best_trained_model.load_state_dict(model_state)
 
     test_acc = test_accuracy(best_trained_model, device)
diff --git a/requirements.txt b/requirements.txt
index a211f877eac..87f0fcfed1b 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -14,12 +14,7 @@ bs4
 awscli==1.16.35
 flask
 spacy
-
-# Replace this block with ray[tune] after next PyPI release
-https://s3-us-west-2.amazonaws.com/ray-wheels/latest/ray-0.9.0.dev0-cp36-cp36m-manylinux1_x86_64.whl
-tabulate
-tensorboardX
-pandas
+ray[tune]
 
 # PyTorch Theme
 -e git+git://github.com/pytorch/pytorch_sphinx_theme.git#egg=pytorch_sphinx_theme