{"approach": "Distillation of the current rule entry into a plain MLP, seed 3, middle-of-the-road choices (hidden 48, lr 1e-3, all opponents mixed). Model: 224-48-48-3 ReLU MLP, 13,299 parameters stored as fp16 initializers and Cast to float (ops Cast, Gemm, Relu; 27,120 bytes). Made by: (1) building the repository simulator under .cache (rustup stable; gcc-14, binutils and libssl-dev extracted from Ubuntu debs with dpkg -x because this machine had no C compiler; cargo build 37 s) plus a Rust harness over SimState with random spawns, threaded scoring, teacher recording and an evolution-strategies trainer; (2) recording 322k decisions of the current ONNX rule against chaser, dogfighter, ace, brawler, Starter, the rule itself, a noisy rule and a random flyer; (3) behaviour cloning in numpy (Adam, MSE on yaw and throttle, logistic loss on shoot); (4) two DAgger rounds, 470k more teacher labels on states the clone visits with the clone executing its own action 50 to 100 percent of the time; (5) a 10-minute evolution-strategies run (pop 24 antithetic, sigma 0.02, Adam lr 0.01, 12 matches per member, pool = rule, Starter, four built-ins, self) which was REJECTED because it raised the built-in score to 1.0 but cut the score against the rule to 0.04 to 0.085. The delivered model is the second DAgger clone (bc3). It is weaker than the current entry and must not replace it on the board.", "blocked": null, "current_vs_starter": 0.9926666666666667, "records": {"starter": {"matches": 3000, "score": 0.9511666666666667}, "builtins": {"matches": 800, "score": 0.88625}, "current": {"matches": 1000, "wins": 98, "losses": 765, "draws": 137, "score": 0.1665}, "exploiter": {"matches": 500, "score": 0.068, "how": "evolution strategies on a copy of the final model's own weights (224-48-48-3 MLP), 767 generations in 420 s, pop 24 antithetic, sigma 0.02, Adam lr 0.01, fitness = board score + 0.02 per net hit over 12 matches per member against the fixed final model; the final model then scored 0.068 against it over 500 fresh matches (33 wins, 465 losses, 2 draws)"}}, "entry": "CAgSEGltZC1zcC1kaXN0aWxsLTM60tMBChsKBFcwX2gSAlcwIgRDYXN0KgkKAnRvGAGgAQIKGwoEYjBfaBICYjAiBENhc3QqCQoCdG8YAaABAgomCgNvYnMKAlcwCgJiMBICejAiBEdlbW0qDQoGdHJhbnNCGAGgAQIKDgoCejASAmgwIgRSZWx1ChsKBFcxX2gSAlcxIgRDYXN0KgkKAnRvGAGgAQIKGwoEYjFfaBICYjEiBENhc3QqCQoCdG8YAaABAgolCgJoMAoCVzEKAmIxEgJ6MSIER2VtbSoNCgZ0cmFuc0IYAaABAgoOCgJ6MRICaDEiBFJlbHUKGwoEVzJfaBICVzIiBENhc3QqCQoCdG8YAaABAgobCgRiMl9oEgJiMiIEQ2FzdCoJCgJ0bxgBoAECCikKAmgxCgJXMgoCYjISBmFjdGlvbiIER2VtbSoNCgZ0cmFuc0IYAaABAhIDbWxwKpGoAQgwCOABEApCBFcwX2hKgKgBrUEMu3m4HbfPuIGz+yjVIy0oBz0wt0U4jbowJ6A3VbtxtFmt8THNMQ84VrCIKvs6Bzh+vbS63SkHMx8qtSxJIpmw0q05q48qdDMaLIkiuC+Joc+rO7SLr/yrNCxbIMWwCLM/LWMgFTG2J2K+Zb7Qtre4FzniMRSuDjlqs/CoNLWkOnw3e7fQKP8z8RYCM4Msx7DxrKkwDrMhqD4yMynMQxM+L7wGslMs5C3BLEOuWDD+r1+w6CgiqScwJLLmrQay0ylZp6ouBK6ANJMs9yw0iuEgg6cvLsethqvSI8c4GTFzvXw32q4bMZ8s8DLmn0av3zaqNUq19avhMLMwPDhUopStFzLfJgcujLgmLI+oOD8OLPe2QDlcqYGmtK6bpDswqyx0rZAwL6KNHBiwWi20LPerVK56MfUzgaz3L4ut3qL+m2ksYi/+MBaxYzZ/OFCsrbxjsS4rMzU5M2k12KE+L5QqUjbzJccskqSRrl867biPLGs4aSXkJWAwhqGNLps6bT0Pqz+xmKilqzuxri2OI4Kt+DC/qyksFKSFpTssNii2sI6qrShoItGiXyDBsqOoyqsJn0Uxky2srGg8cDmWp5Wz0rn+uZEvOa0err2u1asYOt20TTgIsno7ZC6Evr05vLYuoII2QbkIr1Kxdi1rvrZB4ilYvBEfiaYTM5euCaqmsCcpYK4Fr8MvADD/pTmlFaCEL12qnKxELUQuMywzMnAuJatiKV8xbTkDuXY/6rFYM1K1BbCVMBK4kCmNL88oXToMuJkuETZFKtuyw74TOeGvbba0Nag1FDLVoDKwarwbNS+o86MtsEMiTCKHFiWq4Q7iIgWsR642LD8yPS4eLNctGykiqvqwaDANqAUwqyNKpmsx2iONOo41P7kyNLawUbOwMLwwWKqRKVWqdS2vo8I/xSMbqyKpmrNwsEC9FjahpMCuWzX7MhA4ui3XLFo9ljNbMzwn2C7RIkMo2KK2LMGkFCtNMX4FsKqMLaqq0LC0KJywhzHYpBoo4a0SsDCsDDLJKO8vxDUxMxuyzzc7rmq5aDC4rDwvv6mpNv2sjy8TPPy2Y65itLmhmChWug0ulyGTraYwRrhGpIawgBOQPWUmp6xILhWqYSkFIGOrGiqhF6qq0STfK4saZqQBp+4wNDKWItQgGK4DKXitiCHgqriugK4FJKUwCCpwp3qy8yrvOtafDDHRNCwnGjOwJFE3+rU5OJSw3LG8sDI5DDjDuuM4NjdlthOqC6yOMHaljKK2rxE0QBzvLPiw7Kw4KbsqKKiiqjakwJgBLtiwMqScKkkl2S0vI1QulC7LIjsorqlzJX+toKfHnsy2hzGrMk80taqorkGfA6/dq1SvUaVZL7AuhbUrLfypziBZqUi2gzYYN5+yr6lSLIqXqqL4Ebk6FLuTsNOhRayMqGiyC6wSrGSiADJRsH0tNzCbsowt9ichp3AvGrPIMJKxLi6XL4mf3zBYqKURYakDqhCzL7X2rhi34y6FsXqyzCYWtcqZarN5MdQrpjM1Miwxlrc6ugA24hxut/i9miApMdGqVKZSOE+4uLGHMzKp2qe2pT6psSjmLbGuy7C3LIooNqnSI0yx2a6brv+snCYtsHay3qV4qd2xRywqMSYK577JLLM2U7YItWmoWSwxKTiu9SwTH5syDC/0N/4vE65TqnautTNVLGYViKtkKIoaZqyWMeEpZLJXtgAxrS3bK9Or7it1sbktOypDqBMv0DHhNNUq4q9OtDEnK7DGLGygxCounA0ctKzUrKgdWjBApt8iiq3zr0EqSTQ6qXciFrOLNoSu1xkbrGeoNLHVNEQvWS0RMdG2Fb1RuMqsqTugLVYuAKn4ITA57TeQtp4r3Soqsv8nUrHEMA4w6SnlMKAj8qwPsuIqGZ0qKDmhaClFL96yDTL1KuutTCvPqYwsy7J6PDEsHTBbttmyViyIqLUmYSgTsgUlPi0/tIa1YS9BKG6tBK2RuN87xauCL5A+QbAvLlIgpaa5tHCyK6/HmBmu0KRnrw0oHK+Vr4Yjwqg2JmGmQjCnK8SgQa4iL9Kpdy9+K64tK6vgLFYdci+KKpOuVj4CsTi3oLE0rPYgMql5MkQusLC9puotEqoMKzAkKa1JKZKluLUPLigiAzEjMbAirSnbqsuiErRYLsUeXp+LKZAQ0BlDtG0pFKr+KdCh7CILsRyrELF7MP4heCmQK7+pi61tMcuv6C2lKtQo/RYALBKvLh+SM7GwJK4VqtwpZCOcL6MsHB9eKriabivksvsl+qsgKBqt5zA/pNAytabkIaYYsqi4ITY5ircKKo4roCgiqLUr0KrqKqOqbKAQMvAs/6dGql0b87DwqwaubzAwLJgtIav+LD6tladjMBklOhaMJzYvjyGhqD+sNDIduH4tJbv7suOqwCpROd42GTc2sjmtdS38OkSxWjAhL5Y10yG7u7iuKK6wNtk9kzMArjwroJciLvgoESpeK3EvxqSonCow/yzzI84tIqzZMFwwxqwHJlopZ628MUAb0ypHKsvF/jWUsVG4u6ZysOi0FDcmLCC6ASympRI0JjnHNHg1PCdZreCqGTpbNhSuYLEpuMcxCC/MMVUqvrBFugowvjM/oYmvUSLUoRQoNSntrE6izKgALZApGTD9La6t9zGJregoPSkdIQ4vgDA3KxIlmjBfvnm47bTTMCwzpq0cr+whqas1u1esUy23LiczEzEaNVYq9h3drpU5EywJjjixrDKaMnaxujRcJ0W2WbeBOLw02SjCreQnajDioLQt4x3uo/oeiiDVsKMpqyhPrSgxRCfwKZKnYSsCMASx9zCUjk2o/jHnLPaxvigtLgK4cquuJ9koKC5grX6kCKepNTA2CTBLLUmlvKAJOW+11qV+ryqlV6w8rbu2wisMvWK7Dr2qtHcvBhSkqPaswjDjLD2oQaI9qiQrGy2TMFooMDDwquMiAq4CJN2w8i3MqIkoUq5eKiez2qcwq/azepj0O0g477h8tQi7gLPMLnWrMD8GPX83WbnstWUwwLQ6vAupbjTaMWa35LkwORIwO0CQsWO8gLg2Mo4seKxsND4uUy6pLWQZZ7KOob4unSivsxGxgShELWcj/a8Ksc6sYKfVKT6yyrCXP/0rfzV4vvws7jJ+sx+w4K2xpsqyArJXsZO8kDoiMK+0zTYuLN+1PLe1ICYy2zDDsVWrbq/iLFM9mDuOL0Ey/KPSqHopx6aNqB2k5qn3MHwjNx/FGgupK6FOpgCswK/2jqQoM6mwIfEq5y3CrlmqqDLMLzi3GrkaMNIoJLBissanabd6KHQwO7KAtYE7gS2lKme0BLCPuJ4wBqMPNn6ovq5OONoyc7AFOzu3YDesN0uqrS02pucsVyQnLoqkEC3wLbAp/S1lrEQssK5xpeWmiq5JKc4wIqSwIl6wGSKyK0GwjLGisEKyJzBLsIstiyhUNAEwDSupqniqdrU3PGotCCjqKGOgobf4NDkmejSWo/y0rqrfse6mA79hvCi0OLVinzgtbLDJof2uCSwTLOSqTS8YqHmpaiWyIriw+LAjsMCurKgMs+uuzilBLbat9yO6rPKenrhCNGEuIT7FMp6yNTasNn8vbSrerC67rbKLNOi15C+Ltb9BvZPrrmg1bKpHLQc5M7f2sTxCFDitvuK84y++II8sRawfLfOwf55drtkwP6gbLlEqU7CYplYs3a5AJQmqWqH0LLwoThz8LYOt9q6gs7A44LnsroC9a6sGMIYnJDFsMACsK7hJrwO5qzDiLs4wSS9KPs2sZKknMyewq7JotPckiSR6tGctqLAwLrMw0yGzJlcsdaIUskIyo6nCLNun16oHoeWkH6p2LWEYCCxKsGExdaqhKucslTD4qRC1M64Xs3Q8nLSHuwYkjzLjq6Ut+SR7nTS0Q5nFtZ2o3LD0JOMwVjf6MOitAzDCLQgc97CcL3ammrBbvtewDLmVqMutJp+Fr5SxJKDUI6+usyx6LE2wq6doJCqwjKjlqUotji7vLzssmaQAMCUu/y2rLz2wUrAgs4Syx62VpIIukrREOIMk8iI7sUy0A7jcqYOwH6g4NEAxTiafp0ExN6SVo8Sz7622Kv46bz+Mp1YytZ8SKRsu0S/HpBAyZKgtq9IuGg1GJACojBq+pJEo96b/KGwuB6fiKuysJ6I5rssofq61oY6yU5zKsxs5ua2Sub8ykSxgMocpCCs3vGG8Fi7rMyMy+DKStPQ4wrRntL21Cb7UuVvBgjdCvfi54zYMuDAruqwTMOAoFynUMlspa6Rfqhctb6FDqtCyHiVfJd4pFKTcpl+pdSVPMJ+c0yJFJSS4hLIUOvc17bIDL4Q3h6nUJaO1KK3UojwxZbxhu0e0+LK5OaUqILucrVgsjrQ5rHEt6zTPIlOqqr7OQSy46rGlL5aw2Kv6IlIvIJZKKTgxq6yZKkuxCjBoJsCsDLHWrxEy9q/DLh2jYKkGJvYtixQHup+yVDs7NAqwe7ngMJ8z86PQtJo0HKBvISS95rsDtEcwNTUZrsq8YTYTKwy03yUDNZM3RjOxrc0ouz6Zunexlyk1KS6w7BITImOehTKWs1EwE7IHKHOpnLBILIGnNy5cJeEiv7Gsr/KtFiw7sNUoYizCKE80zDyXsDO8CTjOs3ssDrL4MSydCSjlvtm9Gbi/M8gzw6ACvOGzeKnxsfsqAbntMfAw0DEsLRY5YLSMsPOsPiy3nTetaK+IK2ei3CymqyOuBS9xs22wSKdcMFooIS9CKTetlqghnwIk0y2XqE01YylVN449la/uppO5fTZeuKC08jJ2LUkyGrwEvYApPberNS6lDjgVJgAsvbPnpVHABzR+uOI03cGGv4i1kLGsrOOx0DF2HGOsHCmHsCutUi8cKxSwKqp6rQetVDDZKfuvmaxPsIcnc59/MZsv9qAyPbwp4jXWN7Ay5S4fuDAzvTDJsHQfeCyzqlgyfrBWL/O0UyjksIE0qrdDL6yx6q2psIM0hbIoJii9oL4msdezzjHLrskwWjI8rGMvKa20LgetSi/1LgswO40PJSYzEzDOL2Uq5S13KESuCyvwMQciNjmHLHEkjTijLHA3f57PNCo0BqrSLKmwEbBiMsYzjDA1sRcyoanfOBWwyCsNs3SxrLDbuPWwaKhCOiGtorRRJX6ypCcepOkof65zq7yloavUsRqw5RpupiMm5ioSKZYtSqtxouof9yBdrQUpFitqNKyyM7YdMFCtFCPLNVQwkrCHMYqv6yhtH1CwKTHDp+s1FSjrLc8xFjnUpuWqRa7OrB62NDAdsaEkAriauoE2mDMGsO+wGi2drJSzaKmjsCkuvi6KqiysMC1qroCriLICLCOu/q4aJxesvi7krScvDzH0NvmvuSx0sp6svjjxOMe78K+ntPIswqSjM0W+kbV2OKOuz7JvMhA6SruTrDsxua5LtM01TzANKwdBNbr2sJo58ic4r42oTK3AriQpPK3orqirKiDDqiQrXqpzKoyrhrBMqJOsWzN0LCqw1is3rPqxSsRjsqA80jiNLVoxGbLKNHqozq8hsoUKZRq9vXYm0yZiI4qykS6vNUcxuxjGKh2u+DG7LwKl86hTtAu86TI9rX8woiZjrQ2wgi2TqNKo8SnsJmAsQqT4L5etIiwTrhOprqiUrhksfyoOJpWwX6SJLf2u1bT8L7O1GayJLxggETLKLIEsWaeHq0QsZb7TKdEw/iWDrTq0JDE2LpwonKy4MTkZDzVxI9+qMb3ruY4kBLGeLDWltLMNnnovoZ7sqeOp/6tIncWtBqxdKGCvqrErrzit6qx1pSwwkqu7A5ohYiwPNPwvxy0EspMo4SQxKs+vRihKK1Anwyp9Mfi9BbPaLYgxfqBVsAUzryGApRGyAi1QLDIsNSDsJUEyBD2HMretMK4RL74qhSpRITMsLK1dL4Ko6q/wIOyfXpTSrayoaJaQF9swtioPJzouGywzKIOr6TRXJlQtMrHgrr0tETiysOQ2RCzjsg0ufLhJvJk4f68rKUCwPrZUs/GvPLTNMwM3Czj0sq0wCKkotWqyzzmssbqlISXPp2AtvihUJAorOjDrq7EqZ6EUKIchyS7XmXyp66GgMzwtxZ92MLUkTCk6nAhALTTiOcy/5LaSMCY5PCj/tPi0869rsK6unbYbPV6u8q3dKT8pWDH8MzCpCTU1rMy1oDb3MzywSUJ6uDuvu7PcIT4sNy4krtiq56FbqM6h4KDFLFKnRDAUqg8oPbG9qFAljq7Mpj6vxygJrmsoFBh9OkAsiDLbs5uyo7PqNFAmyqC3MbioSCUepfgoDj2jp2C26a4rMkk3ubAoqTY23yyxMEE05KGdpYw/+UB/qMU6pzDvIgUzTyxTrfAwe6/eHG0mmawCM1Cvfpn0pdWoLq0MqhktTSuxnX+tnKlmsNkhNKzCLG+0KLnUsFu09CxiLU2o6Z6SraolbLSlNEE+GbavMX4seDRSOGswwKcJNpEt7LRcuBUzijHSPe+3BzGXMmor4S7qLvWlvSqaK2osu7Lerv2VRy5dInSpJiwlrBQkUC6GJksl4zA0MNqpCrCYrVkgzqXQsJeoZLOaPM0tyrgQL1O5rbJFJzw0RTKoNnc4nLUwtsOswjv/s2yxMCpptXod860nwLmdvz2pvwS2ybj3pESpb6zdrfwt0izJrXGsHy67rKoxFjCgKlsmpSgXpJGx8q8vHB+sJi9TJDIxFK+3M/O4HBnFLI80NT08GWg39hz8wimxJa0sMZqzW6l3KKwpxy09sNI8drU7rDqxbzUQqTsqnTR8oOdBIjH1t/A0Dqy5gdexvrLeHSIuWLBMn5mgG6XAMoyoEy77MK4r77CVK9usQiv5qF2hmS7jsukpAqpWMjUnM6wjLwotFK3rK2elSb/OpiUgbiJzpiqrATRsLasss7ESPf2uSJYwslI0Li0rNU47fRRUQcs8lz9tvCeq1i5mrbWnlapWJ/YoVLJrI3ukGyyLr9gvFC0gMpio4BWJpxyudjDNLOYh8KPnpGOwxzaoqvcs8DEhuQMsnywBLSG54jGgJlMp6aczu3UtuCpkMWMyTT4XNsap/7G7qYiv2DR0NuCksEJFOGC44zakrbYh7Se2LwsvEia3sIOowilvk+Ctey+OFO6r9axtqEgoy5tGGwmrvinDpnktV6lwK5soDLWcMq8yQLt5uBS5rK1QppitDCmksOS3+DmjsTEyQrCsrPm9YTl7rYGw57kQtuWw9izgLJ67nz9rNda6k65nqrkoOigOMHuud7D+KM8u1abWsVcACibPqTWaFh03NNsxmi0YqSofXyhUMDykd8PHu+420712sAg0irHeNSWqezH9NHypLzUDtWE7aTXLMN22Vi/GuV403qgfJko0paeVNEKxuaolwFO6bqsxNt0smKjPLo4scilDpFkrMwrkMOqhLrEQq/utySp9Kkmt1KIGreSwjJXEp9OvL7KOMKq4Eiw5NU2q/ytYMoEUhbLGLaMx0Siyp5qu/ri2O6AwHTF8q7uutbmiJ+Ii3CUNrDQvPTIgpQye3rWcsgOnf67urkux1K6QsZsfviqnJsIjTbAfJ6EsfyZHK6opQrBNsIGvLKwRptoxtSt9poMtlS3or+IvRTaQoyytCjQwLYCugzBrtBqxgygpsV23qT0uqIaiFy73KTe5liQLpZ6uLiwXsh05QjKwq9y8XTNlM80sXrAKr1uwciipMKipIajuriqopKkrKVksxyu3rF8mXK+yLn0rviokqbAplaw8LnMkkinMnBUxUzTYsNA2kLT9Mrs1ljgcsg0siKDkMUysCzy+NFs0i7LGwCS15SyIL/W3PsBWPOW1TLWWuknBwL0cQFEzih9vKGwpoDBAsFiwGS3rKVUj0x2jqGst2CeAMUskgS5vM94MuiWgr9iyfrCSqom847JVtnO7bK+QL4e0ETMnLvsvi6xpqouwPzerKLo4UTAGLgc0IL9lLSmjLDE3LRQ2A7ycKvwvZ68Ru0k1KDENr6CufigpLEcuYzGoHeAnWSkvsakpFS1VqCsuM6SDK6+XUjCGscGvlqQzLQEnyCSALewy2qI2NzGrC7ZCqzcu47RIMSKxB6z2MTY14y/2NHcaR6/WKAq+SK1GLews9KqTMGk5KybtqAQ6DzaNsL6tEasVKzcwu60RitQuLy5iM2yvqq5GpLisKyt2owwp7i9pKI2sLK/6sPcvKi0XJTux/yGzmZI0MTSfq0uzTyZytPiwJS2+L6Enq7CwOCorRDclMb6p/K6zvsuv5CpUMQQpK7K2pSGy1SyPuQ48dzMTMwUwPSwTrpWqLCyoHjeod5/EqMssZqXrrGWb/6WRnicrtq3crIUpx7KiMm6wFauHLz8vtKfQNAAZ3iKlsxCvobs1LeG26Sxiqw2sFbrNM+QzLyEItGUrzbSQsu23/yc6v/GvHTwGL5GxXz/ss/k26DTIreGvJS2xpXwuEC99LhAsE607M5ctzzBxpHGnkSGIrxyamjS5MtKutLABpz6wBC1JuNm/fDDfsBUstrl8LSM66rH3saqwfi3usHwseTAlGBWwjjNvraG8jinVsbWmaDmHs60tW6kaoAg7mz0FuV24eylErYWfrTBvsCUnvCbUqy6u0ysxrRGueLBmLU0ohKvlL06o+it+rqexwKUSqRms7C7FOBAyFjRpqBWtsylBNVwvsKjaJVSsBxkIqCAqnzIOMCsvDa8Hu4IoGKlnJj88Ua5SsHYjDypjOK84sLfOKHWsaSUasQ8iAqxHLDSahiNpqn8hoqphrROtM6/KrCcwrS8qpoEwBqqfsLUmCCzhqn0tDzzPMM2yvDF2NvqsF63ZJpyyTyJUJR8ys7CMMswydzAIKAgpNLvhrTioODOpqKe2TbLguC6sVD2sPNSzLbDqp5sw3yfWp6GhVq9dqoI0LaN9KsEuOqFPsOOvUSsEMekytqTVMP2tzyBWLD2cP6Z+MBataTS2sho0eLfpvWc046Vzr+24Xh6Btgo5R0Biueq557jutUzATLwlIHypYC1otq25NLKtL5U2s0L0vUm4VpKCqBauFKy6KUYhLyOMqnEoq6SjMk4jQq4TpYwpTK3ermEat6oPL0WypCyGrC0fUboYsXC5R777tKEomyeSJgYvky9wsdilY6ikOr85E7ZUsywuQSIHvnM5Y6yZM3QqC6/DuTyuo6k6tOk68b1WulSwLi16MUMyoKhQMmUtRayIp3UkiCTKnBcuA7AOJhWlM6mILWMozajBJ2ki665LL8AwvC73NC41YKlpOMI3l7clobyoj7A9JG+oRj0RPOWunyxqJ44yX73fKHet+jUbqsWtQzmWrQ4kvsCOL1W6IagxMv+pNbMzpdCkcyubHfmtujG5IhwlY6rdrRMonKpCKKMoUbG2s9KqV611Kc6v8jDZKUMp86XYpDKx0Tm9LN0rdaWurC6ycCG7qRg91DwyLyKsHLS5MTy92yd0pZAydy61sUy1YKw1KRK+/KjRt6Os6yhPpM8viShsJHIw+CzuMHCwiy5grIKoLaYBsYyweavqpRSschkVMs4q/zEIJVCvczRqsOu2tbevKnq0Bzc8uqy3P7VbKOSQhK0MOkA6bzOVOP20Ja46PV0zOp9LtVM0SKfEvcQwWDREwEW9QqTAPOowuK2SIWWop6piKISo76QfscWbCCL0oNGqa541sJ+nFyitMEylMa83JnkrCrPirYU4uihjsP62qCq5t0cuHjcsNUasKySUpAwmJi/LN0E2mzBxrpyrdz4EuIEmCy4SsVI2Fz3+sAazCLsXsrc3i5O8rYIrsKGAMeUt5Cc/rNMsYiovmyqwqiButJaw762YHqalUbG0Juevg6+NL2SRNjIfJ62xvapqt/Q1Ly29q2GtCjDrpvedc6NOMAKqJDarNJcqRLE5sOw8BS+Bo6im4LR7MGc2qRadqWE1Sr8QrPW34SkBJFUsMaxUJlItci94L2wx4y55GyiqRzFNLTEpySv6KYCqTzEFHAmwiatdL8SnSKiLtnEmgrIhJCg3aSQvKJeQfbHWsQIk+yxOo+s5xKwDL+uoZizoOD00ZZN1sq8sGLNWPMmyNRq6OA+9hCwVKR4u865Os5UwjZvChS6w+ayjJJEwRzBAq5GVgijQKsukZykNpHau4i40LGmsY6h7LVAs1628M7i0bC3bslcxiSlYM52+AyoRLaKc1rJTnkA4GLEqJCcwazj9sI02lDTvMeg0ey7CNd6uaLfyNgi2nrYEJvyrMJ0ssjglAKT1HZ6tdjLpLzkcJTAuq8cmY67rpBUutylqJeoswjQsMCwvHyboQUYvxy8+LIE1+61utXStYK1GubsgMK8WpC+tFLJXr/8n4CfeKwY4BbSxpgCwerxDr0Uy4SzbKfUxsbz8NGg1LS3VqB8pVjElo1wziCt4Jt8x6604n0Skwqt8r48v1JCJKOwppyM8L4ewjycNMNetiq69vBizirFAMsEtlR0usBGuELhKsEkt0LIENlSxe67QLwaq+7CfNDAt1CUlBFI1zDI9trkz9iu2sks3MKt2qmIiMa7HMKQncq/yJrupCiMKrRqpUqKjq5Sx2axBpqklRyWUK8OfQKBtL5keRC5LqfqlhzKdrSWovCj5NWMs5zGgqxuylS/SqjsfGDZSsSqwG6purcGqGjhepfoY2anOLSU1arHOt78oKDKTuby0zjTurxSiTyWeMDWxBLHpKZktkqUcsNCpl7BZLT6sJS0ILkmtSK8gLyesm7IAraEtWpPKMfIsazAxLUApQrZosBI5VLaONxc1B6afLjw6/r2nL+ieLTrxJae0pjsGLqG4YiZZvnIw7zN+s73BxUc1r5cxbCZcKMQr1bG9KLyoUiqrMcirpKihmTGs+ytrLZwvly9KMKwz5SGOMGyU/LAILJetujYMMwetbbeVrGGuDy63PAQtiyrLMb+lmzJnOLu9ai+IpCo2ja2BtIWtXqZMuLYwwSDvQGwr4ybbOWM9RbSeMJ0vD7AtMGSmrqQ3IoQyCyUPo/epF6sNqmWrmLIYsHamMakypWusea3MJ1EioKsVpQ20xCbZLD4zjbSGrSGkdTnNLh4xEjADK9csdjg1vZawK658OZIoh7XUtVaxxrMHK2GuNzz3rbYncrldr6WyUDEBLsww+apMKi0rMrKQKNYqL6sktF0pRqm/J+krnKxmqfipEaiwLGoryKhVoB0sdqypMn4xkrGEMbex1TMlpAWzOzTIsOAyNSVOL9s3W70tK7eskCyZoxezZLcJoHy2cbEEujI3TDHeqos23D/kLs0yu5mArr6wd5ntLOANkCXFMrcyibBHqjAsty17LGewSrBwIPmsL6dRpoucHjH+Mc6pAizPLEqmPimKrca1+zI1vEMxnj6SKC0u1ywCuo01ZzaBt384ETWLpTM5Ti3hOVssNyjfNJewGKA7xEE8AaiFuDgvRiymqP8zSBhLLyGxI7Bjq5QscxgQqLAn0CxmrvUtky8tMw+eoJ0pnyMiPiVnMEdAkh0Rsqq4zjSDOmMtRDYGqFs38bTeqIK1XLPeOOinOavRtYQea7dhvH6eITU1NNa1DzCGr9AoFL5QL7MiB7GZrGqjmjC6qa+ooq5OI9Gns6x9LBSp1ShsLYew0DFDrMUonyiirkKlwS+4K4UoQy4LNjCjazGNNdiwVzQqrd0yCKzuN4W01K2QstK31DdOrnEtmpWXqase8bLdrFA3wC6rrDSyJLLiKU+8zjMILyGz/66FGQKtMiXVsNogDaqXr2ImkiqQLO0i0aoGLxkvvSv4JI+umK1HodmnSSdCrrolrK6KL7KkjrPbtQK0w7G6rxWurzcdsd8nd7MIuUs8G7VdK4gqHbJQNXUydxMFN5mmXbYtsUQxrShjuE4+FzrGq4kt7i01MrEvvikKLwOhGytYqnkeby8OkD+xoyRDMIsoHRYdrt2vGjC2LQq0NKsfLKSgG6zmqgMus6+dsh42Ujowtby1HK7UHC6ozD4ium46QKrqM+qsVcDtNAMrbbANtWKzorNAMzIYg70TQzqtVL8Gq9QtualkqhmoWbE7nsUqdbDfKKwwr6k+rUGp2i33MKacF6P3p36viqGAK3wyJLCzxPyxp6IJPPWrzKTXtsi0+y/Mp98ochraLkQ8LbiWM06yszAnLgjA+Dd0pTWxFLassuswVyTpqHO7iDiTL1Gqeyb3tKav/aRWJxKwkamGJ/WqWK9Xqa4mSq5KIrscN7CnLcmfKCyMKH0vTqdJqEAqUbl2tmOwxLPcLOs4TrKGrjgw2KCZJd4g9CxnPOG27qyCnPWqAa0cv3wvZyCxtHatGzDgsFMmaihBNuyvCSudMnEpfrQBIYcmFbISqauk2CVFq4Wo9CcbrdGwxI1nLFOlNLCSLgutLqbHptYv86Scr2mnIZK6s/cm+yvVNaUyqjBUMASw/DFRp3o17jqquYOy/KkAK/SpfL+NqMypLLWyJditzTCEKgSsxjMHO4ovdbJgrXYY1TB7LuOllw1yorEDrys7KvstNCqFqpgrPq19MHAlhittrrqwnDAkH6kplKs9Mwuso7UapwIYFDGZKfOynTH8LUqvyak9Me01LjPwtAKZty6LL5U5B7R5MMCpkTy7q3+wP6jwKPC4QTGKNbG0ta0PMJexrS53L4op+quNqbms26R+LequBzB7r84wY628LSCpzCD/Iw0fwaWGsAov8kiPPNO2dClbLo+waC4TMFOngbOtLHMqNy7+KI8wyzBDI6SrAyWVOFIypClIpwe3nzLysJQq6SWmt7A1P6rqKsOuY5TNqZiugSRyKUywvSmmrFYjfxfRJzAajymYncosFzRtKGmrma9kLJSuu64prR48erc2HDgwki9wrwewlCj+sruvLLFApKuppC6brm0MsaXApBGunTbsrpSjFDMWs8CsYDCro1yhGi3CNGg0kaVyKS6pdqLnKQoNRK9rrlewtiAZLrCtX6SyLM2rpq0eK8CvYarlK+AUcaUGKOWngq81rWOtxDEopREwZSpYrAsh8SGTNrshKiHoKIwsNq0HMIcpCa0gqgg1rq1OIwCrlKgxrtafr7d2Gli5NTtFtSGrxaWFE4QwzqfwKZYvuy8ELiywGispJQ4pEbCpIyiwZykRqcqwe67/pk2n3qsTsJSshSgCoRYxFrYVrdY9qrPKtjSt1agDMbOshjGmvDe/XDDgPHkzlLJ/OQE8FTRus9q5hLE5Gtg2Ea6JPHs4xDiiuXsutSz2sMsoRSj/JGsx8KuTHkcoqam+rRaxhBxNEy4mtzJmlwGtey3/MUOiIZkbr+O3brl2tVC+b7jyuHmizTZQFgy6PC9vK+qtlTyBJXk1yqq0L2gxSTjJrnAzUC9dq+UwVDtLt10qS7fJuJMwpjYjrIcpnK7eKdCgHa36Ka6oS6lIJoWzb6wHrKAsq619joScVa6aqcqlHi20rnWtLbC3JPaygzklPM22FLu2rHiwSqiTuGEfLSFDGt03KTZENQmzhDMWNFUw/6qsNNAuKqaVsZ+8kKhLNf40WrQ6obMqeS21Le+ucS5qorMwAi6VrRoum6D5JkgokiyqoxYx4q6EKbcmYS8jHogqI6H9r5ymv7K7sCs5Gjgdtw28IKyusKikwjIMrhMVMayXOIm3XTiXOBIvvCmNrDK16zAxqPAoBiw4vJI4/itmrWU37rYOMOijnC4sLKGxe6kYrFGukjLqqL6uU7D5rXYxeJZbqCCybZUFLwqsrC4rs7wvwqw6Lu2whjBEOgs1qLhxQaay8Srtsru+QzPUHvIgoK+oN4I5MjDaHwG23T7buBuhm6UeNNE5HLUGMcunpDhawC8uuLO8JBCwAir1K4EsPjBYLsejsK+VFKEnrq38JsOsqCgrLFOsHCxRMeatQbB9FQuvc6nENY00CLesOcgqSbwMNiozNSpisME1PKmit8U2KZVfLOMYfjGsqvM5DzXjrAIovjAaL6SfIS6sJUK33LZ+M1wxryvjqVKx8ayeMg6t3CReraOtFKloM2WhBy7epzWp07EGsVWtNDMYsYwtBDStrqKpLSjhpRuw9C9DMSi7EzMmsCkutrlbMxspuLQiMEKxU60QobSwoiY1uPUwFygstI41WDKhIVo32qFnuvK47S9bMnSu7S0sMBymLyFhtN6t9C+EKt2rGh7NrTCu/KuArKQv/aKtqbwurqpDLWWsBKG+MGQm8TNAsDqy3DSzuBewUyyLMEa5HDUqFEW0ZiwntIq2uC56LKM1qbxXrnUbLbHfMFkrwbGCOHgsNblSQKCsiKq+LCMuYSttKS0pNKbKKKMwHTSPMOUlQSlRKvYmlanApxArMK5xMImk6K2XrJCa7KwZrqilFjU/rqYwUDiNM1Y+BzHGNjAtSyh5Lam0ksHeOGq40jzirSkuTaxAr9Sy0TG8ohw0YS1yrWVAOEFoNLY2sa0TIjytvSxiMRAeGrIrsvuxHS4KrpIqqyp0sJSvOCl7IIqpCrAvKzSeEK/gLFknWqzGmdsxKD+/sEooIbB3LUKq+zCnMtiWsK4ZuDe+vzICrDo2NqDeMCc4U609sn6qY7WaN3epUS+EO/Q8uipJOpGnwysAsX2tdq5gofqt2pLqrIghgqjTJHmxFa4KLM+oiiy8qgineLEgqMMwrKvYK0Ao2y4xLdm0Q7AWt8uo9DXXqm4sN6cMHlQm67R9vL8kTCY4NWMsZTUTrMaparM7q2uzghrQLUctO7dCOnqyriScKqUlP54LokutlqJjqj8vfqj9ImosV6gkMPAsMCotJEYxCCecMDcxaKRiqbAl3q0wqL0mNbTYN9OnMDB6Is0gkqo0uO8tEqewoKm0d70oq7SuHqpmpq02Q7ZIKMCvxB+/J324ljeGryM51sCaOLk04asQoC4pByksHemlVBiPMhwtQDJnshysri1OrFSoFSisqeQlBKXcpmsvqqoarCKqmahyqMIwiLUyJKcyg7uGtgItG7JLoAAmLTNYP386cR4emI0hLShqP621uZ76LsGsayJrLK8gMqYmO7K687Rers+v3KT2KcmuwKi7rAYu1a36MakwFKf6pLQp6bOTMiaqdS12LOyrI60OqNgvT7QALdTBZhu3t9a1DKjKMRy0IrGwr/mzIKx0qPKxYT6uORufzCkdopcqQD4XsNimgyghsyKscal8q1kozjRYPdqqHyWcrGioQC27KmihBKmqNE0p2zBqpp6pOSMjHKmouiE0DSIwgyhBL0WpwqxOJQotniz5vDyyJy06rri0OzMxsfOppKPMKu+t6iZSpK09STgdsXkfHKcOrGQ+QKv8m5AnjzSQoh8dJSiSIYe8YLZJstgp+6G5q3uV4So4KNCszqwaMUmvxawgJ/+m1C5cLo8unzLzK6CvA6nAK8YbkaiNLbIvFDDsLfAm4ivirfevjCySJfwoKRqFsYujITAtPXs4/Kg1KDAoIiv3PdKuEKEVMn2tvhgZqaywrqSQPtu2ZrI1rNWjH7CRsCcuJyA2LVYqAq0/rNsrTSk7L9MujK1ioSCt1a18Lo0sujEVqsqksRYmsNguvylUMacusCSQsSW1Y7ZpMc44wy2hKTYurTlbPxoy1qs6ttcorKyINsyofa/3rFm3v7pINs2szzjuyA8xgqw6qCYqRDJDrjorqKraJKuxUrIJrE8qXCxTp+wmaaytKFotoCwsKjQqFLSmqxifYrDDspsroSbQMC20zLattpK6XK+IrNOs2KxNKCkwkT3Uq3W0Cbg5KhC5cbVLqnqxaJ9/LTC/UjL0JeAtvL/ZqeGwtItJrQcmhKbRotQv9zCqrEmsNKNILN4uSDIYqCizkyrzHeEoOTBOJxOqGDNXLMQuVrhyLtwsoLdWsWiw7rMlsISwx55oLjsqmCwaMp89OKwnt/64cSW2tFUs/a+FrQyle7FGve+oWiCQtvYxTrakqBMmCC//rIUjYrMLsPWoiiVEqgwvCjErs5CkGaQFn3Gy9qtfqZOjX6yTp10v1i54oASz3CxgsiQ1NLPxKcK0IC9MMP+r4S2LCRsrYTLiO20zrLT3q98xC7M3LrKdkKayrhe4zbzDLLglCblSOlS4QSy4NHWqZyxBLBywaa8dLCukiDAJKAAxuy8xMCIvFyp/JyCx0y+UrL2owzDZrsOst617rQiv2qH8OVm01KVDt74y2TVwPeqyl6s5tMK89jeQuka2bLkMt6m4VR3OMOQubLsrQI7B9Lf+PXeyMjl0ttU2lKtopGqwHi/MslKrwi00LKKoSyL1o38mKagzL8gvaiwIq48tdSy1njert6zZrtcjwrwquoA3rSxFsRwyT7EuM0WvgDTmqTqhkB/Fus82hq0cshKxHjRWuM40IB9OM7u0yTdRtA0x+bFptJK6CSywsPGxW54ho/auL6oRM6Is9Si1LQGsw6EcLIQsvyaGLzQdbqoVseyjQTDAMDqs766aL+U1yrSQMOCz/7DNs1mn2C+Qr3k01LD8oxim0LbWMKgqai9dMPMytLl9soGnKDM7M5s0fbzcLYKtl6XgvQ3AkLFrFWOwDi8SMIEvH6l2ksgoEyGqMhCpQhzimxinMLH+qyit+TLssDavG6WmpKkcva+vtCU0vp0Lp7+xDzGJEIOxI7QdNzkm7y/fMN+4xDEfn2QhxK5WMae4Oa+EqAs0Iy0cKn+9rq2mJv08Ib0Dtc8weyyMIG6hKSx4qVCn8jB8sHSqFCp8scQqWCo6JnCfRLT5rsAvxCsQrOoonidjLA0iWjD1pP+1gjVQow026qX6tE4s9DWXJNcg1qClI3W8uLXBt6o42awstkK+jLZALB81zbAnPYstLLSFvH83z72bwKqllbDGpOstz6oIqGwwUi3bsNOuby8qsTwrMayOJpiz3q0yrNOsLzG5JqM0MSz+prslCjaGrIa8mK1XK+q2sTQ1KlY0Sy4GnxApLLO6vX62Y6AOrain7rmXLQuv37FOM3ysDDbFIvItjbssOO2lGKsbpe+uAydcrtcxn5nWrTskYy8QMeCpT50jMCYyxiZfKo+skKzcrGeupCWBL7wuGKw6vBw0hLs0s62wpbRoKXQubq/7LI8rF6SBpbgxg7yPtVisNrDyKIa2iJ9trCatyDVWNDg8xzGzLDs5m7blIP0yoynkLVEY7qmeHaClliuvKwCXmSihLMEoZistsKOp9i6+p4evh7AaqlSeCylJqVGqx6QXMiou1C23rcyzJxySsS0wVjJRMfoZCSZQM+q81KiPL18uqjOysjE4L69BsFYhU6/vO2gwEh7eLOw6czMpMZ4rprAjqYMoJqXWrpkkPy36sAAsUq11KUEw+idCK/4yIrFWsD2sn60HKAEYbLAyNJskZaWMNMw2M6oKuHIpUbfkLIi7VbTTrVG15z2wPk685q9ov6WwGb+nur+4Bazlrb+aJryeOuqwxDwsQU874rDSoQAwTaJBrHCqZ6kujfonf6imKEQtHjHYGKAsk7JSrxIpoqvJKWsoTqlWJNQwOC3wPDuoDTSQusq4/iCztGG2+ac5NnazfC64Lys5JikXtlSr0Bvytbe9vjgctoWpX60dNZQgGLrlpVRDJrqPqFywZyxvriIl7KuuHCKmR5qGpDmrEyhxMRupSypuKhGsdaU0LCetRDGgJUI0cihRMCivhi7GsTo0cbjctYI1D7IxMUYvnyckL0Qm8jTrPGm0cbK7MDs77TWkuko71bPbq/+42Si+N/Uy5hddu+u6bbTJJ96eTrAzM8EmuSxNmOsnmCv4KdmhhS4Ep52viipBKkUylahnL1+sm6JarpYqsbBarxY35rO0tQa4c7R8KJ2tdqzkJycqBDTpqAU1ND7WMfUytrMqMSEsb7niNnIvAzJdrDU1ujgsq7oqrr/VsIefSDa0LiMxXLOoLFwnM6jXMAE0NJlbMb2rrzD3MqKsfrBwLMowVan7qNyuCqpJqGurQ6UnNOArPrkYPN+0TsVSGjslN4C7oO2wSg4OHqefKLTOo84hn55dqLBACikDg0gu56swFQsdyR+fFbkfTiq1HvOdiZsELCivayl3MQ4tvSlXKEsqU7DXJsAvTbK5pWqx7KI7JsuwICY0lwS1wCdtG5WvMR4gLL0jRaaQuHG2MJkVmaUUHCKxrNeUtxwHJnsoFac8pZ0fjimptU4qhxZILbIt048gHgqhipFPoIoSF6CuFIIrmK+5LXopVC8WrT+rtB/5LhqtSTGirBQm7CkMJ4Qp3Kzoqm0vfC33KmesRKgkq7GZnq27pm8oxiLLnE+gGqTDkLWbFa1uDU4vSCQwrxGjHySVEQGjCrzNJzMVtSy3rbaNlhexGKKV3hxxpFMYkJCOK80WEauLsLWsSKq3LCeohS0VLp6tkapFqxatvqgkLzSozTA0L4Ex7LDxIX8aJrMglo0tYCWVl7s05SSFHw8cAg0vGtGwMIVwLkymuK9XqAOedp+CHxi8MiK4hgUxJqBZlfAaXZyFijKigCQhnhoN6KePKZqsrqiYr5apQaqALMouozIxLcYsGCW8r9gssC1YrYOz/jBBLKuqbrBeqbcxDJa+HzGhNKE4NI+99KhGO8+yrLSLMEAr7hW8rk25YadQrx4sbSkHuQsyBqz7se2z0TVfriAofq70wUDBWbPtOC8tFbBGqWUs46EDMGgvQyk0L2MwJy+kJ120PygzLBYwSKgAIkOh+qxyL/IrGK+Crsy8j7XVLsKzYaVyrd8y7rAANK6kyzQsqzMxOKNOMGwvUrCvsZ4wfLf2sAStca7uq3Is0SfmLGckLj9JvU4y/LFnKe6viKzGqDitaCd9LvamuSoTsSQwMBymrj2yIyA4Kp2hODAesa4tKagbLaud5bG5qNcwwa6wOCQwii17MBCtiDD1LQAwM6UQMeC0yrPSLq6ytDJToqi2Ai8NroiuXSlhMRuuXzDuJuk82zjZNJ0sYa5xqvsxKTBJMEWl2KiYL1myAq6dJ50yl6worXMsxq0jIoYvKKGUL6kvoSMBr/IopyWYM+Cvm7bBszg1erLcL7uwrzD+kC8ohCfKuZK2OjH8JlutG7C7ty8zialSsf2vxS73JuEsiCWaO4s90DG4LCkm4zEsqHoi1qjEKboh3asxo36qBKcJrowuDqpgHJ2tm6Zxrvwqb7LAKWcxmy/iLtgYpCxmp7uuRrOENkG3iLURHPO64igyLimYTr2OswOw0jHuLW+yWTVIN60lrrQxN5c4AT4VMDKsfEMmPFe4Fz3SJeWjVaqQMMwwViy7qGsvH6RPKjMrXSjSrWcgezEJLbmsZKX2IYatrClXoTkq/S8JtHc3FLmbPf4qEDNRNAk5tjE5uOgli6siLa+ma7qstkS1k6WFLwo5wyIRJM+ulLadrBy4PTGarjy3aq3jMFC2ri+aKWkrj5rDqZSpMSHTGAmqqaw4ITysZarUs6QtiKkBHKqt2KxdLMQs1qZtqpGu7jR9t2YvQa4iL6KxMTQTMKayc7wdKWKkvy1eOAC7hqqjrhCzy6xKNCUy8KUtHPiiPzGYnl4toawAxJsvNzXrshIqRCy9s/WpzjKgrPusMim4rUCpAK/tLwYktqjkqqgeVqAZIpwuQjHVJa2gEyzQLqQ3Si1+MVqpXDMduFE0GxkVFlm2Uy7npV0xJjXGutKxeLKzMc4wVDQTsGkk+a30MKWrC5SjtbowXj+2PN+6nDSYneIcUyz0rBgsNzE9IwCqhauvIL8ih68KsVcrbCnCKx0mDTCOM6izjy3IqLWs0CulNKsoQi5fuAgzhjPuujiw0Tj3ORky4BwyNoA5d7qeM1y0Vp2lJUo6qaUIsC6wVi4hQI8yRreNOnpCgbX1sjFAjzDvMBIwJaRKGRGxIyGKMbqucDDrpeOb7KF6rpAnUqlrqtOsZ7HAMNawcq9rMe8wMTATNHC3gz6IMt23tJ+oNguyfy9frJIsayj9Oau8kCHImu00SCYDMTI3kirVuAiZ/LbnsdWfeyhFOp83hTVFNp2q6600rU+wBbBtKUAvEobHouSs0CSordatsiplMAsxYjNhMQkwFKYXqhsiOimPpeI3dazyKwY00rREp4wvNDNNqxoqpS5iqs8t8zrcvJC2S67pMtGuti58rS2Yg7bDL5K1QrcoMOYrRbqhNaS0/CocrCIpm55oJwqy4as2qiqnoTDPp8IsB6shoBSwc6rWsdsnDSlLq9ox7SYIpJyq06gwsVEtpDanMqSvULFdKkctpLVkMLsyZyCLNck7wL7ituewOTTAMcAvU7hwLhW4tSwCM4Kqpi11qH6xCjxIrsQxtS8aMAUsLKq2MYAp9KyNLZWtDyrdKCwxxKMMrRqsDakusNKvgSr+LIupzxyeqDmvOa5SrgcthzKfoce8/DcYKmO4PLxDpVmpM6+5N9A6Yqh3sWqsGDJmuR41V6mqrVKuwLgKNMc2p6XwyAU9ZzRlL10t5a/1L4SrFiwWshStXy5yKM+lDy1+pAKr965LoUAvwSJPJGIwUy/nrwcoHK6UrjU6za8BtL85XRl/KUGxdbiWnKG09SiRpewwgTjwn/utsKg2L1al37W/N8skVaieMjMtL7WUM70joLpDOKoy4qwLrzQqGq8+HNSsXKXMIFqcBqZ3LF2xFaiAKFUihSTvKgGnrSsZmWgqJq5QG1kqmaq1rpwwuatmszcvaDd0LBy3XjCEJCYqBJ+DNLA1Za5aqusptK/WrgozVRg+CuKwxa9DKEEqqKA1qCs7vDLoNtEUrisnsRUxgCghMFmuA6jDHRCvCyJ8ISEubCPmL3Cx6qYwqRKwOLB1sOAknJfcrBomBS9hLREvyq4prIA0GyByNZo1HLUPKlMrZjAINzA1RrZvs+Wx+aDBNtK3yioptBMvR6nFNnM2y631LM4vQzToKFqmzCfLrCkfxSQ0Lbktby2lrYUwZitxLjmsE60wqfAZuZ10MQwscik+qbCl2iW1HZOkDqH9NY6kdScBufMx36vEMGyumSzZokqicLljPwYpDjZ0vAqyb7zztMk5yLH8IOWsXji0Lxkv/be2NcawTzOdqgatl7DWNOsshDBupbKt6qpOKHwnIpogLZWsD6RSIFWw3ytAGvurFamPsQWyFSxvs1woDjTnNdS21zShqTm4fzAKLwKveSVEMwA5szdRsdgwPTiEHuS6tScFL9OhEjbas74thayLriS5RT/itNs0m6KOJ4KqvbDgJfEuG6agGA4rYiUjLQcwfjLyqecpt67GLqAowKUcLKssqK/YL0GuOTZ4OZeyjLJ5shI03DAHK22z0y56NPQumi1wtVS4TbK6sDA1JackuccrHjQ5spw16p00OhGwfa7gIcI9ebojsJEpCTLIIVyo2q4/MHcQGCFap8qx6S2EKmyxSrBdoK+ycC24rFasubAoKVqrHy99qtQqbzSUqAS1RLamLSmzqDILqva1dTBWreUyFLY4uTqy9S+nMts0GrfmNOQop7A1KTYpXjlbNCCoOTTzPn20YDFrLa6vdy8yrZMhDalJsh6jEaRpMfYnpi0VLBus86Pknn0pl6/GLmkp069lMDYfHaeqsEMuGjQRNl6yxTFdONawIC59uZSgVqD6NdC93TlxtCuinCpMMhxAj7UgoQqwQqo1JMEw5zVWnfSkwTwSMiov0Cn8K2YobqgIMpytfzEuLYatljAUL/uvCC0jIt8swafbm2Mo+S7JnrQvcaw1LIgoWEC2L+A0Vq0KM7cqgjX6s26i2ygDLMUl3rAHvm846i74KgCuUKnePvmpAyX1pCusyqrYrhe0cp5fto46jbK7MzOumawBLDuwqa18qY2xhLRpsFQlf6sOs9KsOKWNrZkpYaQtKuWo+jNNMMAkgqWqK2k6OiqELrqqJrMnMEQopCTHrBUrS6xspWqmAbxeOH0sF6asKjEvRD8FsHUgLxx4l/ObaSNLMKAZRDYCtKQvFKf/q4AsxSw6JOirOCwvMWQv7K1jr5oUOzAsLo4ndrM1MeQwEiz3LmypsSyFJOqTLCz3sBMqoKp8LOyts6z3KeOmYaUZMSgoRCE4LZy5yDOtrlerYh7uslA+pyP5GJKw46i4JjmoELPPn3K89rhQtLCvoKcesPUg4a/zsfMlaa6GKH2xMqwbJcArRC45JOSwoy2OLGAyMSLsKzgw7qvoKG0rdCvKpnOw0ivNr4g8uC8atfkydzziLJspgK1yPbc5N7KqLn0wG6jqPcSsxS3+pXe0sjDzscqsji6OvcQ0ZjEvugapNpt2saEoSJq7q+CmiKMflwgxiKV2KxgsprC5mgCyxKhIo/wrKa32L4UqICnULDTEpLdusUu2E5oXt3un3DMuoxczpg2XIZCscjiWMLgqQ7NPsN2lWD3Ttwqr4yZXMoaokLNaqTop0bunPK8uQ7QtKeKsoS4yLzQqDCsbK4+n+hsNoGyqZSfMMaYsrCtysMeeIqZvr3wrY7BppUMxqiSOOWkpV7NvsIorgLipKDk0QK5mL8MoH65+JfYzaTSAJBwpdCg9K288NCg3rd0saa8spTO4WbK0Jmm1w7N/NaSwgSxNsLAuAjHfpN6tlayxsIsv+CndqeusETC1nJ4xDStpLka0qKzIrE8u2B/YpAotETVLLG6xrTUfMJKsFq2wsMKysDY8LEwpo7I3Ovo1PzhVoE6o/StzPMEuI5zyrBWt9bVwLEywDigEMFO2NDixKyYtOzASsMAw/iafpOAkTzBUqAchRCxOMpuqySzvJ/6wDjGJrEStj7CjrTgwFi9lLBU3ISyDMqEw0jEtPIm1VTncLwsxxK/6nVIw0DV/uR64qz0vNbCyKjq6vLC02adGNAEqEbMmrxAvXKyGvSY4SrYnLLowkijKrbusJatOMbqrkyoFqUKxMRl+rgcjJ67grQinEC0Io0ecSy5np4GcJSpfr4wwhrwHNBqxFDFHOTk14rLIr6yyoKkFrl80HD11sRS08LB5MNk8FrU+sYAr6K1mKQyzLCbrJAc8LjyQv3yvuqvxn/uwYilnp70sfCiqMm4q4aY0LTKvRalpJjopbTDqMTorrSGIrVoqbS6vJN2gSa7RMRA4jromrSS2jy/msDUuVCRIsXAnuLMlL847JC7XLoiyFrJDPLo0LykjNiyh+jLMtGOf+6afPxS4wK8eNEsfd5otLLoz7an9L0cySyzxL9stTSJaL8SwtC3eq6CwwC6WL8ioBRxMrTwm3q19q4YwcLKPNEMxkyIGMNetKzE1rFiwCy/xJHCxKTBYPQqsqKYEr4gzhD2Mt5cnpjYnLjeqqTSoLfggRz4guFer7C5ErTMgS6wpKF2r2C47LxQnj7GirUsqBDFDLGQy8KgCLuct+SkorDCsBy8JqIOs3C98J/isJzkWsOanZ7hiMl4svTC5NCSwAZyrsM40arSDtQmxIi4Trp2+LjMGL/8xArbdNXcwwzlAMOhC/7XAOCO5grGMHSqqd62/LCuqjDG/qggnhKaALfWsrKzqrSgnoKyvrvUtSybiIjGpxzAnIlak/cCjsGQxc7kttfqwzDTKsgyraSwXs4UZdLGLt3yBA7fOqmK0kzBwvaQ4hyhUNlIq97qLuHG1FSlnue25rbXsOl8kTjIYMK8p4CN5KRYxK6i6KSupFjBiL94oGqqYKYmobKsir/wwFy5BrietWSxPrOY6YiI7MsQ0cjDXoA4uNjO2LVU0G6wIrXiwmLnFNvusbLLKKfIwA719qdSwLDimp/S2pziFudSgdrqHvbO8vzczsews+iHqMDsfFjAeMKkoJjApLu+xEi2tqhOuS653Hzsk5DCmpfopq68sLwKouLD/NU6foDE3tLUd1iuLpE2jWrOrMYK0Ry4ytKO5KDhmMJav7rJosLe8PDNxKGQ1+yvYMtY30bsNKva/TMEYt9coCC7lKdcmpaCnJSedSKoTK5Aw1KwmrqAk6yxNsHYrZ6x7rrKsPCgHoWUxk6XHMYGqhzb6rQU2Ma+oMwG/0TFgMxwygzo2NF8q8qwbuPOoqLLhL9iuNDUquQO7b7NHsfsxyjfiMqi3vZ/ntz/BTrhJrZ6tjq2NJm0rWyOcI20wQClIklMnuao3GeQYTyibrTwz+TOWKpqydK4+J7ep/iwAM1YlijMvs7a4YTBgMuE1qrOsLdA5t6ajrNslBbqaOqwz6DUeriawq6gmuBWn5q1DMMUtuLYcu46YL7G5Qf4z3y+3MbcsAKs7Ihws8ihfJ7qk8i/AozmtHCtcKocpSakkq0+wqylJKews6jFiM2yuNqHbqdEu/LKtuHmjwixXmj+rXLJapZurvh57LGMwtjs/p9ozmLPVsFA4FbR8KkGs6yubrZW117pqLBMbU7j0uoCwmS2nrw6sYymbrlMoe6pOrRCu8C1PLeWtF7BzMOsuiaz1qyorfiK9qoMyTijALnerhTF1Mbk0RbEYswWwdi9KsmquFjwhrn6ZTyl1OO05JzXPqjGxuazCPrwzPqiBLgWpXjCzu5y8nS55PUY7QbretESiEaTPrKsvvzLKFe8r1jBAMaixVydrqwImP6ZeMAKvCS0bqa6pdTDGsWelRCunJVmrmyoPNYW4TLF3OeqvsbMxMew25KmBo8goGTMAN1W5bjK7MCo2QrdwwQc60jJNvGIxvjcGpqOscDeeuHYbgi/YrwKrkCrJqkQxby1pr5srBq3wMkYno6snML+qGrA+rACphCtdJuCr4Q5VsSEk5yozP/m7uLGstBulRzWQLkezgK4BM/OvoSoALLez6DPztpYoDq79L5i8mrXCJRQfJLeCq+SyNaiRIv23izhwrdoWv6QeqD6v0q8PJ2EjliiEK2OnKDFVnV4lBKweI9wlgjKurFojSpx7Jmud7CmzKXQqxzTLtos0wix4tDmn3q3HJ9qwzzAPsaAaJq/atMQ2XbGZsIciQK9XvJ04+KiXKn+66yNQMSqtz4wctxa0mp9/pI+mRizgsHWjua1HKmSqeCnWLbcwTislqLUrSihvqBAwiDGLocCfMC/fL64v6RsLMCswY7nGMok0BrWeKUIuyat3LeMsMK9uqfarJ7N2OEExQhu3LdC05bugOHYfYDULKyiwcbYwLosr9bVbr7ksaqaLpnoikK4RMZWsnK5jFlWmVBJXspSqjDDMq2OqiqoHr30z0iQYqYkuNzBzLX4voywSsJ+n4bIrMe6xY5rPtpg1NTjdPMWxGqyqpWy1rbsbuMIwNC98shC5wrYPsXcso7veQvJDD7eTuxg/EEIpuHq8pKJJpMupg50GKJGRU6PtpiAqda6Bn1IrKTL8G2stm69CrRIv1iwAK8WjhStqrxAujbqqvKS4EjixsfqxLTBdKjey3za7KAws5ariKh67OLhJsemrLzMruWEw5LLQnwAo37E1OVo1jqoTNQg9jjVRrfYhhqaDK54rzJVLpiim16ITJKwe7iG2MSCnwCjvLjIwWTKspjgoNCmYKRIqBa4eqwSsiyoHLPw21rAtK8szlLUYscs28q4bpRcrbDYJuIe0SbQRtDQyn7ZmNAmQMCjBOpS3zzo/MAewOqwELLe8Vq10s9ooaq+krT+uci02JEGpqizwn4MtQC9BKaWoQbDbj+ykCy5nn6ktiSnQLfQoLaw8te451KyluHMpN7U9qLKfMragOMgqHBw7KjU0lrhtLOqvIqidMKS1xTR8IO8zDTCAvNO0C7FVLD2w1rujuSYvfi2kGQGsBDBpsrIpm6AHFw6dxydpJp8iia+7NJ6wRCaVlx0w9a4TLsAsw6EKM0QqMjSvryMsIbqcK827DCysLEK3C7X8se0nNbWStOQy2zUZrIOwkTLHx+KyB6iCLhQ80bRqoW8yqS+tQDA1fjgpMSKuXqBmpZWs9KqPsWMvqyldoHavOa5tLi0sYhh6Lbcxj4x+rfIxd6wFrMkrxK+qr7IxdTz1sqUhLTHMNResFyI2MoMsCC24pDAW8jEkKUApPKrkKrk0ycYcLJuoAi4Pnssi9LfkqgwkLTF5wGE0iTd2rJQxDq5dL4mhQSo2JxUyICvHrMMw/CsvoE0tvZg+tfmtPKvysBqttCo9quiwvi6vJUiV4Cr6sq0mRDl+LGqtfjHNsSWuHqUPqOAy+qg5qIkkOiTzs7XG5zLpn8IrLaQCMKicGzCJoZw+jb3UM/Gw3igAIx8pnrPbMSMyLi3rLOaqbKlysGEl6CXhsPWpYqtmn+soprAIJ0Ct6ayxsKSniyqdmUQw162QMV04WaupLEovLbAhLP0jBqarNGcrWqBnLiUrtbULxuozwaaCMVcmbDCaragt1aPDNsw1IrIPJ4gkZbB0KkgwxpKxpZwzhC6ILVUqUDB1qjqtJS7cKSepaB18NBkuPaVireghsyIgsCcukqgKLmUyIDVXOeG0AjP/L0K3bDXMKiSrtbUxJ/6rIzLntH4yGkAuN8Y0XyQ2O4Ow7zWvswAj1L1owIO37D6gMfupr6QHpFqwVybVMYSp7inmJ28oki7hLRYowaCSJ7YswCW/riIvtK6nMgosKZVVJFg8rqoFN7g1f7wcMWQvGTBwpMMob6+Mtsa2RzatsHCykCG9svg9R7s5rGcw5bduMPO3BLTZKBTAprKewIEn+a4XMBEt5CjXriWsgy9kIaWY0ilUMOGtYS3tLiCyZayBJXktXK09L8Un6itmHDixzK3BuQQtNLXdLke6DS7aq+GouLgVLgMwarRlLWcwsqmAsIcdrbL8OkizVy01LI40oi1BL+MwTixlv79CTri0sP8uqTBCrGStH60eKI4rbSgaqOKfi6xZrmwwFxiGnemwuqr8MRymd6ghL0csDCy9MNApbjQlMQSuazA3LM0wb69Br1i04izqq1S0PazrtcGs6y2bLTYj8DiQo3yl7S+ELTAk9iRSrX+s0buxPqe3piigrfwicDJCr6au/zDwLCscoq4TsNEvirAAsfwwlywctKmwMLJDKaQph6m3MGQusCDfIgomfLDKsNEr17NmrKostCg6umY0+iMwKnM0jTCKqtYwqy0lMUo5PCxIJw0qmrbKq9YwtjK9lsi/zUCANNa6YDAfMH0vyrLkopMpYSA3MjSwGiZKLPwnNiQOKWswc69uptSraCFwKnqoAyvurSisvMjLsLOk2jcVMiazvq6OqrUqXqgirA0qiK5xNkAr764lsJoyDq6sNbkrPC0nrB43aywXKM2tAiOVF6M1GKpHpUAqOCBtKlExl63ML06s8aSro+IpSrDFqgGmw7ATMMOttaWlsc0x0aX8LUkriaNAqH286zaDqtesGirQIfowNSsDG2ilTbFOqqarmjQ+J7iwdJbOlzSuTifurl4hV7QLtmul57LGnk8kRDBQIEUxr6W9MVQrzjAdKFuib7EeIaaxOjCFI6+pf6zHK8EqiqQOqgswzazXsoEwWq6jrRuxS608tGq1ja1+MwIfAzeLqDksz7B1Mis0xqA5MhG1FLGatRio0LArnTcxdTAKnj+00SlFrvszHbbzol67xbXvs9gu8ixVp8GlQairpsektSi+rH4vESMEIRErL6REqz8vHym6Ji4xSbIBKjYgkSv2HtCzuDXZpTmxC7iTKUy7X7RAOuqyhrPBMhQlDC7NNwK9RDJMM2W3arEMujE5P7L+rV4yh6L5N8oyR7HewoDEqDGZvAErM7Biqj0soSVxrPoaOq6lrvIrvaVPLjAro6VzLGYeCTEILYYvpzBdrU6w4DAorCBC0in7rEG4YakDMyK05iXkMTAxNK/bqWU0TDoEtss1GyzapkKlG7nkuOetoC6FMyMs27UKJfspML3Muxc2eKFGKyAxwR3MMA0w3iwtrFGk8zIttKyxzZzmrw2wY6lDoBMqFrH6qmosCLOJJAarJDAatOYW5CzJrT0tkzeTsJ0dtii2peW04qlOLS80zrQCNtYpNiAALry5dCrrpU4w9y5UM0WzmiRyJySzHbzOOfCrQy5KpcEq5K30sEox5Kx7KtGvv63WrKmsKay8MKyoE7E1rscvlC4FsWgqdSB9JoolrKepqsouIzESJwwwdiqds8WdPKg4sm8qSBw2Mau2XjThFH+srSyQuDaqB65aKOsv4rOEOf8qQKpir+8r2jhbMdCo4iWcsDkrhrIXraopBrM4KeIkWinNKgewxqxuLpAwxLFRrKisbyC6qcQvAiiHFFaslKpnoOs2aCgjtFojmbZysMwlG6sXG00wcbhCqZ60NSM7qyc0xcNxOOAwcq52qTIzFDBpNDqonkXnQUI2jjRSsEosKidQLSI0PbCSqCqsIqwSLm2uJ6qvrQyhf7EXsTarsK2QltWvILPTLxYnpC5dLgmxVy+tNZc05TGfsJ82ujBmMmMyYKCQNLc7RrjBsCgvATA9onbCELCdL9CtRSVfKp02ubE7q2o60Luit0KsRLBHKuOuLqz9JM6twyjrKNQsXjENoOMwd67jJguufax9KBmriCyRrDysMSy+KcevwyaTqOegvSc6MgStSTN3pR6otzMjplEg/TG9Mbi1LKOVKCiuuKmjwkioNyuHFd0osKwct2mwA6sMNSHCLLn0LImZHqRHJogmfC7lsVcsNSSyLvCtua3WIWwvhipALhOqZKnELGQxb6KpqMyokTAGI9+owaR6MVeo5zGGrRWtSifALQCu8K37ipSw4zSbrhCwXrCGMUm0OcNNsgGfubTqLYqwWiv3sYArJLJ8v3+69Crpph6suykJqRgseSnrqiOr9C3ZrZOyJCtDrWKjU6h8snGqVCXYIxumUy8iKN0qOizDqwytxqrXIHYwKmwIMBAKQgRiMF9oSmDruJe5D8IewjG5ULqluOK1RjhtuN82M7vjMDi1kShOOc8qmiFCMqI1HjAWuEe8mDSzt3e81bPDu+O2UjGYO309F7PttBQvbjyhvbS5yb73NCA0jr/auxk4TLrdtTM5/TwqjyQIMAgwEApCBFcxX2hKgCSuLt0stMCfxN+eHrP+tDAxDjMKrYAs/rE8tF0q7LJgKe4kSDVLqSA1OzPIrFewVTc2KoGmJjTFpHg1mTVlqjivdDF9sfstAznSrDwzTrfduJwzpreRr0Kq67TGM0MlxyE5Nxy5NkDZPI0wL7omuoCwRyrHtcuv+LWoNMG0CDftuUo1yTQorlk0MDJ0tC2y5DLKOcUutDA5tiovqawyvHUwLLPUtyY207hXNEAzvbiltfejxS7YMRi4cjjBMC+M1LjzNKg1Sj4pQEE09DR+tDk2SLbvI/mwlalTt3Y1sTnLtQypPSejLRg2AbZkMr4jnbKNqhKmSLbrsP82XDXAp+mxFzKuLSwwjq2rqaKyIjU7Is6ydzgZOtu0IDa8sdMvZ6dRtIwsfLmDtjcqBCn2uPwyn61Br1YrmSuSKqAv27BrJDEkByyvJZswL7GHrDC0WbOIKr6gHau7MHIgdSwMxnM1HTCCpfEpqS6KrJmqOqkcLD0rWSxSqV4xoK+TKAcw8inbJB6vSsRjvBCyiLRCsFWz6zUkJhs1XbUBsK+4XbBYNQI0vTBdNP0lr7Q/t2Y5ADMsIxswRjLxtqC03bYvMKwpGDALLoG27yjWMuS4cjSptL01Bq4HtgGwNbcgvPc3czTyNIg5eTdkPlO7DDQiNYq5HzjvvY42UTblv+2owagJtxye/a9Kqhc5WsB5szA7VjNNMVUuQy2BvEi6xrhRtJ23zBhPIuw2kzImuWu9sDlKNKUykDxTtAK1hK8wuWs6k7kos3IyIcCuvlguAbI2uyYuG6u2pzcupTStMBQ3srctMP20D5/ZJa8zF7IFq3a0MrJ8K+qsiiRHsc8yWSwNvd014zHDrS806SzDsHaxqC0KMQM2izDYsJhAkbh2G14qxT7NN4K0WjKHs72sXLImKzk8IqY/MjEteyRYNBezpjAMtmyqdKjysc6l5zj5KIa4qTj3NOIzIrJ1NPAx0Kcrs0s6PLmHuByrC7mPNnYu6blTM+uwAi/zt8MoESl2LNOrJykAP0k0c8ZowfWymrxwrIe31TBstd62OTZDNvuyQ7h/Oqg0J63vLASw2rYquwY+jDbVN6C+Zjgau4a8aDNNuU20mTfbtGUz9q5TxEo8RyDDuxgolMFqvD26YDgWwB828LDDOyCylUBAPdgzDS2SsCUvj7TtJyi4QjFXL3E46TRqJ9MzQKmisbExFTZZqEq1tKSPrE0pJ6WUNTi0jLRlsXU46bUvlZQ0C7XXMq8voLQLMJGo8jefsNm85jaBMIu0e7TFv925m7wZrhUvPrAztJa6/K1Wue88QbwUtfK1DjSqNzw1QDm9MrKt0r8MumDAhyhzM2DEEbhpOacuDTcRs9gvFbNArVa01y9FxKW7B7CFOFGnXzrnN8Wzda+2O+o3KLHAwRg7jcG3whY5mzfEu2gvlLchNO45ALOQOEEslTZfsli127JGtG6zybAauXPB47CjNJa0M7Q1t6I4fi7nLy81FTMmOTc5m7SgMpGxKMFMtAo1Wjo+ux0wLjhDsEerArniufCsEz6Bv/Uz47dqK6C3PbBBNAo2xifBM9Iuaa2FthotazgdM601HDNQt8C5arswtkC2lDO9K6g2m7koIVe3VDQjOaq/+7VmtyKjvjrhtQYy+DS/rtIwyzRLqrqtMrf4r+Gy98TYwbC6M7pPtRE6xq1juOA0GbPqsa21CLVSOsQ4p6FGOl00jr7ntdG08raAtjDJYSzZuLqq67DEt741TzYTuPo5JipfwmCzsL+ZsQUnxrOEvdmzT6CavRY2KLUbPQ22rbJAOHayu5yxtmg4bbVQOZcnbzj8Ico5YrULOkOtHK6Ssvk2nrrxvFg8TrL3qsIw0LJjv/Yw4sCgpkEkCbaysKe2xKZSQrk02jpTNZ+yTLn1LLM6zzexszq0TqKKr5EwVL5NvaYrtrFhuIYoDKQdImgsCDNnLls15bVTLFy0v6bDIewwZa5rplmxDauKJLCqdigpsjgysCpxv3Mxni6OrbQyLCcnrr6vfyckMHM07C7orpU/b7dZoMkIqj3oty6wBj7MviY3ii+iuGE36bF4MDK3n7Vvn5A5LL3Fu225QLg2tdAsxTeRNj9AS7optHYuA7muJrg3Y7kbPIK0DbSluvi2PrjbM0q86zvjNkqy16UNPUM8HDutuJ0y9TaYN245b7OCOKq6fKJ4NRewbjnVt426GCf8m660kjEgu9A3Vi4Bt8AzL7klveKyiDR3KSG+ZDRtNGWz5T1zvb0mPS4AsFux27WUQaiv8Cz0rFcyKT4HNha4ciBxvYo2PrLGtnctSTzLQDExlzCjMSiz2S5/teCwuDAiKn6xnDMgteGr4SzmqjQxuLgAMmK8tyiHrFykWbWpNByvmTsRtn2y67H6rb85ErVfrs2xYzCQtECyZavbJG0plzTsKYWxJ7hxtpM4ubqBP1cywzMLtlOsXLEjNnwxHTHQKzkzKrqDvH6wEzhwtx+vrLZ+uui+MbE8M+Q4b7aoH5KpTLhFMs8x2zkqttK1WLp5uDeyAzYitfg1BTmltZSQw7PoLLssIzUVttM7OrzexXc3HrI0NtA4uK8VvEMzkTJctDe4srSst941fDihtM2tY7AnM+85hbQoKh45+62Htk63HzW+tLQ4jTdQrj4vDziXMQw4GL89t0K54aFkMLi9CSE3MVazfbDbs4s01DthPCa4frmXueEwGpTXNHK4yTbBuN2p7bRlMO4vG6/iMF0sdzNGv1owkzaRNVNB0zVtt4s59bzQumOtxrTNrGI5ibHeREo3wLVVtqg1IzqDv/KyaSwuu20oMzhkvP03KLMQvSqyVjWCrTCyBKQPtAq1BbMktFqtPaT1q+s4Gy1ZM045l7a3tV680i9WqJojoytVt1AuHDSYG5wx6Sf1Lus6PDF0M3G1GjR6uNsecrEgs4w/mqnzubmstjSYJumyCsdevo6t1LXANEMs5jZOKUYvNy8dNXi30rcuofwvGbd4M7+pJjU9vIMuTThHtCifwDVCuOCz/Bx/tWY1lrXEsMgisCsxscww27RmtcMv7LuIueOwCLdWrUm2Y66MuUwyirxivTMlXzH/sb8wAyrdJwyvWbMLJFI0R7NhtnSy2TAKN+StxLTesyq22C1uudYoUjMgsoA4zTXrsJ01Ax+GNVI2DDGOOb8tFDpnruMwnC1EMnA/2rXHLPwzbTtNOWQ6ib8uOg4sS60LuGQ5IzBdL2w4KLGbLPuyNTM1NgU0ujO2sYkw4zDMNtY6aTXcNhsu3jAVrS80DbekNF0xcDmlrhMyG7BPr4i2B7cjsYY1qbIGuBu697Fusl45ZzRbvKOxIbmXueGz+bgjOc2zKzaTuEY6hjs5teq1grqDuTw4jbYULzQ78rxGwao5kKScNve8GTU1Omw5xriQtiI0n7hIsHe5RC7ZvEA0Crr2sfMxYbekObO8tTlNrt+r5TBPOh+w4jx0tRs0YrpZOIwqJ7TMOMa1XDi1NsQ6mDStNdiyF7SQuLY66i0+uJo0y7iNOcU4B68jL3ouJa9KuP82UjmhNE+0L7WiQJqkwC6guCc1ODohuHs8vTfUvI8v1TDJN6krtL1+PGAiEjZMNrSwxzDUtOax7TRWLHYr6TRkuqgsNzO2sAI4+bArN9IxC6dqql8d+CwosiO4WLrItn23ICreLj6uMKzsscGlFLqrN/e2QzhoOfeybDKTMzq1TbPaMQW4FLj7P4MxGqFmMss0rzGNMiarY6p9LX64gSyhuOg2hSi+N8y1PLCQOViwxSlVuEI4TzNRsKsu4LdeLeoxv6+uMAo40rh7OZ2p7LE5tC2nL7AjIBCwSjDkKBKt7DFBuyOxpjlXwuCz0bUfs5owNy/AHyu2XbAysc8kWLEFNV41oSfMLQ01ujG4vI6wBDb9KEkyozXCLgUxj7kbtBElG7NtJDMyvTl7LnwsWrUTsbQ0bLF8tpIrXC/wLDKss7dmNsMy5rWvO/cz+bb7KAy0GjjhrherWDEpt14nx7UHLJU3pbd4HWIwKLB9NBg52zvsLT00JDoVtmAbgrY/rZ2uNzejt/o3F7CPpaavtLk+LmQyRDiCuhG+9K+rMBSw9bL9uO+43LwhvZioqrjFqkMyATICKtM0hKiMOS6u4bJTuA42BSchNFE0eTB8vL42UDO1tqqc3DpsuBu4tbZJuDc5j7kKNsW0GzA+K1kx4zo5sjIsFLnft6sk36/frr0wfbMBt78q+zsCQJEzHzP0NNy4ICUbrKQs4DKvLuU2cjZ8vHWy8jDvM+Cd9Ks7NL+7h7jgvO63BLcMPK+rED2Iupa767AJOTcugbZGsWkx6zTxILS2o7XitSO3wzIHNNCw+rcmO0C4fMUxwNe0HTu4OkG2Zp/xseCrwjjUtUq4prsJufop7TIerz+12rxFtYK7lzUoMFu4tqyjtgUypTYYu8o2r7X/NYyqVDSIOlg3Jr1RNmawUKtHLE6sbbJPvQSuFbXktAOwkTu2tFcqIzlrtnK5bi8Gs8SyKSqNr9I2FqPyNy0oXCL1MXAs3rMcs/Iz4rVatdYuMDPnIOaUM670vNE5ubNUONY5njNlN9W1IC2MNPA2nzK6N7c4ojeENRIwWTbnuMAxID5gPk8sMS4lKHQokpEiKIo3gyvzpKI11q3Vuc66wyxLt4y4RqhHNForf7YILbIy4LKhLj8rOTfsNCkymTVZMlG0Fyv1KTKzqDqJNE6r0rbFNps1yjW+soE3RzJ0tECvrcRTPEm4Z7B5uJewWjUQNni0czj3rACz/rZOvAixpzYAI+o4NrFZwAg5XxlfrK3EjLKdtVY0GKY2NhY1YjHnvnE3CbP+xKSrqbRGsbyrrsEvOH7Ao7J6tms1BTQJuyYvBT5hPaCx8zeQLyutArvytzYmQrDmL6QwjDEfszy11ziptiGo9Dg4uRU2/LTmLMmwcLm6uHI0O7dFtAoueyuct0K8CbXPNbY3xzuuN0W0ZC8NObu7JLCYsO/CkbSjvHqymDu3O82pPzBFswklC6PSs/2yrqmerC6xlzU7uGm3SDLcMD2s9rK8uve//KgtuxC5iasjqr018TiSMey3N7AYHDys6rLAtP0zcDs9uP2ySjnBLMnCijmKNCYtEMCTu8KwYLEgMKW09jWjq8A1R7S7OAg0+a9QNTi8cTjOua6uVbT2LHM0QyTwtEC3MSibo2+x2zX+ufmvuryVtU2x5i22N/S0IrlOvasxk7yCuPWtnDgRu8E1WzeOsaS1+jkbrjg9ajVoPE2zEDhgrwQ4y6+TM8KxYTNNOMc1z7EsvGE1cCiRqSM20i8AtVK5Fx+utn2h87LXqpk37rj8uFE4kzSxqIYt9rf3Myo6rbrJtnUw6jD7r9AsLDGrIsazZLg4urqztbHkwJCqOjRLMy635TK6KTY0iLN0tqsx6LQmNx45vrZ3LrU2brkltQG4kDfvNdmmpjLWuW0r+DXsuz414rGLMeo2cjYDKsgsVbc6stoxsLWOrv4mILSCu78ygTVvN+Y4E8CCOBA0xzYbsos1mjHHrw80gLTvufMqkTQosFiy0zOksTs2Bbl4thEnbTl/NJAv66QdLSu4ZrBaq3y4N6eGsJI5s7UMKl6x6q6gOaGxVzripRi7uLhdMdA3gzQFPuk3Kb15OjvA/bePwAi4DbMcvCelXz6TuxawyLqts1e3b8GktbMxoK1cLrE8hbidMr85da2FrCU1+zB0O2W1U7vcNvY1Ma0St0G4XzgtOHc3OjAXNqC1BDztt6oxiy0ztNa1YzCrvcI4SDs7uAUx8TW4sAc1MjwYrjy4nLfQMI04tDRDN5y197h1pHU5Djk+NHYwmrB2uVA5NLo8OpQytzXFupQ1XDROO+Wk8bkEuKY1IbiXvLa94TvsOAqqornmvDK5Wb3IQpM4/C3RIekx0zUtOe03IbI9LFC0w7uurdM1gTW0r1Kw27Uxscws9CkbMmY5cqNgNDKhjrVAOu827jZluRc1OroULyk3arfJslK7RCs/NbO+arnvL80zObaLOA2tvTxUOCg2/jkhNBW4cLszNOuvyjFquJcxz7MLvPy4rik7vfY5nrSBMCez5rymOK26IL1NOck6fTlzs1si+KzxOLe9+rA3MIw2qzinNFe8+Lw5NNk3qjtZNgIkSTgqbAgwEApCBGIxX2hKYOU17yTosvs6RzYuqGM8CzVzN5qwLjaLMf+8/zjKLVs9SbaVuj29qzXDNV24lDcVPG40JDK4Oci00TCFtq44QjVQNdS1aTgJtsUvdjXWuEUr2jarthQ4zC49OU00YqXotCqvAggDCDAQCkIEVzJfaEqgAqo2eLBvMxCoTTaLK8EiDTKeIqO1qKA3LI04HCtPpWswK7V2psu5m694qM0ihzZquN23yrT9LKSmmLJmM/K1nTE3NBs2GjEBMawzM6yUN685sbBdMJa0kTLdrD+poag2rYOefyH8mWG3nKLWkJk8eSR8n1yKlZQPmpYnCBSgFni9ehxGGJweAKHoHemcNp9KoqIoEKTEnjCbViPzn4UhgCSlkl4gup4OqgIlJiBVINgpepp4n+Ml/yIDGfIcGJ6vhNQuUrnXujUcdjDtwlszHSyYQJ4xlcAmQA07zkD/wKIv9sNov35BIT/8vnbB/DoiOg+59Dd9P+i+2zH/uV40tbxBLavB6D1ouCw48EBnwIyynLjIvGo4zzkvQf++Zr97wSoSCAMQCkIEYjJfaEoGHCh+O6k6WhsKA29icxIUChIIARIOCgcSBWJhdGNoCgMI4AFiHQoGYWN0aW9uEhMKEQgBEg0KBxIFYmF0Y2gKAggDQgQKABAR", "entry_sha256": "a0c17abbc37505af815824a58443eb96e80709829c533af37c1a5350918853f3", "entry_bytes": 27120, "ops": ["Cast", "Gemm", "Relu"], "tried": ["Harness check: current rule vs Starter 0.9926666666666667 over 3,000 random-spawn matches (2,978 kills, 22 kill losses); its seat measured 0.9933. Harness runs about 57 matches/s with two ORT models on 6 cores, several hundred/s with the native MLP vs built-ins.", "bc1 plain behaviour cloning on 322k teacher decisions, 150 s numpy Adam, hidden 48: val shoot accuracy 0.985, yaw MAE 0.176 (the rule's yaw is nearly bang-bang, hard to regress). Matches: 0.9367 vs Starter (300), 0.085 vs current (300), 0.805 vs built-ins (400). Compounding error: a 1.5 percent shoot mismatch loses most fights against the teacher.", "es1 evolution strategies from bc2 for 600 s against pool rule+Starter+4 built-ins+self: built-ins 0.995 to 1.0 (200) but vs current 0.04 to 0.085 (200) and vs Starter 0.951 to 0.959 (200). Rejected: the pool reward is dominated by easy opponents and the 13k-parameter net drifted away from the rule's shooting discipline.", "bc2 one DAgger round (238k labels, beta 0.5 and 0.0): 0.9733 vs Starter (300), 0.2083 vs current (300), 0.85 vs built-ins (400).", "bc3 second DAgger round (233k more labels, beta 0.3 and 0.0), 75 s retrain from bc2: 0.98 vs Starter (300 tuning seeds); FINAL on fresh seeds: 0.9511666666666667 vs Starter (3,000: 2,853 kills, 146 kill losses, 1 draw), 0.88625 vs built-ins (800: chaser 0.925, dogfighter 0.855, ace 0.79, brawler 0.975), 0.1665 vs current (1,000: 98 wins, 765 losses, 137 draws), 0.068 vs its 7-minute ES exploiter (500). Best of this seat but below the current entry on every record."], "next": "Do not upload this model. The 224-48-48-3 clone cannot reproduce the rule's shoot timing well enough (98.5 percent agreement is not enough when one wrong shot opens the rear-armour window), and ES on the full 13k weights for 10 minutes optimises the wrong thing. For the next seat: (1) clone with a richer-but-cheaper input, e.g. Slice the current frame only (56 inputs) so a 56-128-128-3 net fits the byte budget at fp16 and gives the shoot head far more capacity; weight the shoot loss 10x and treat yaw as a classification over {-1, 0, +1} since the rule is bang-bang; (2) run DAgger for 5+ rounds with the student acting (beta 0) against the rule itself most of the time, since that is where the clone loses; (3) only then fine-tune by ES with the rule as at least half the pool and a fitness that counts the score against the rule separately, keeping checkpoints judged on 1,000 fresh matches vs the rule; (4) measure the exploiter with the same ES recipe, which found a 0.93-scoring counter in 7 minutes, so the delivered model must also be hardened against it by adding the exploiter to the DAgger pool. The build recipe on a machine without gcc: apt-get download gcc-14 cpp-14 gcc-14-x86-64-linux-gnu cpp-14-x86-64-linux-gnu binutils binutils-x86-64-linux-gnu libbinutils libctf0 libctf-nobfd0 libsframe1 libgprofng0 libgcc-14-dev libisl23 libmpc3 libmpfr6 libgmp10 libjansson4 libzstd1 libssl-dev, dpkg-deb -x into a sysroot, a cc wrapper calling gcc-14 with -B to the sysroot's gcc dir and bin, GCC_EXEC_PREFIX and LD_LIBRARY_PATH set to the sysroot, -I sysroot/usr/include/x86_64-linux-gnu for opensslconf.h; copy the repo's Cargo.lock into the harness crate so ort stays at rc.11."}