{"item":"sixteen-symmetric","batch":"e02-residual-head","status":"complete (replica only; no contract/submission)","credit":"Replica (replica2.py), Gaussian filter, fee rule and qm tables by Taylor (gtmcknight), github.com/gtmcknight/amm-sim @4812a6382311624e1506aca131ca35c378f930d9 research/solved. COMBO/TwoComp/qt1-ref (QMRef) are IdentityMD items built on Taylor's work. Every authored file header credits Taylor.","hashes":{"amm_sim_commit":"4812a6382311624e1506aca131ca35c378f930d9 (git rev-parse HEAD verified)","artifacts_sha256_equal_url_suffix":{"qt1":"8a6e303b662e1a5ea9bb47f31acedd3988ef65b3ddb66d074517aa59ad41ca40","TC":"d3c849db2ac31e567b57e52b80b86e8c24c842e5b4125b9d75c78e3cb07115b9","COMBO":"d360bea6c8c982717b7b43a8069a87013169c89ee3e4da25622db4d880b936d4","QMS1":"17d249ab0c33fc44931d9b105cf3f9e3846a560bd518b67310c2a56defacdecc"},"extracted":{"twocomp.py (TC first python fence after ## Code)":"9d3c1833321493fb51bbde4e1b4dabc2f8978bd6d774ee2af8dc8aed13f0f0af = COMBO published","combo.py (COMBO fence after ## Code)":"c5b711a0b4fb4783e8879a484bf87a427171f4a11aa5eb5ec3513c24ead1a796 = COMBO published","qmref.py (COMBO-cited raw URL reference/qt1/qmref.py @31ff8ed1)":"1dfc9b7a3c2237ff7a93dd9f0af54b9bdab16f2ce0a7eb00784cb4432598db92 = published","blob.bin (same source)":"3ac7cb8811ecfb6a17f499aa4e9b70dd55f7fb6432b3aec3b904e11a3dece38e = published (10,167 B)","qt1 Strategy.sol fence (+trailing newline, 42,990 B)":"c68887cff61180e12ba2b1ceb333edf76a830f91f63b5d775a8806fb9ec08aca = published"}},"gates_output":{"gate_P":"PASS: unmodified python3 -I gates.py, CPython 3.12.3, all 12 values equal README to 6dp (7.8 s)","table":["10000  448.261189  443.455598  443.522608  443.663520","10001  576.779199  571.463409  571.623551  571.977293","10002  531.886497  525.168028  525.865859  526.411631"],"qmref_check":"QMRef(blob.bin) edges 10000..10002 = 443.101741 / 571.454483 / 525.908441 (as required)","combo_check":"COMBO 10000..10002 = 444.195330 / 572.172682 / 526.274760 (= COMBO artifact); COMBO-qt1 on report recomputed = +0.597391 se 0.013388, identical to COMBO artifact","head_zero_check":"Head(V=0) edges identical to COMBO on 10000..10002","invariance":"1,845 fixed callbacks (seed 10003, 1500 steps, nonzero V) replayed with random arb flag/p/s2/j: 0 fee mismatches; head(mirrored odd inputs) returns exactly swapped (dAsk,dBid)"},"frozen_spec":{"base":"COMBO (Combo class, filter+quotes unchanged); head applied after COMBO fee on every own callback; stored fees updated so filter no-arb band uses fees in force","architecture":"16 fixed tanh units; odd inputs xo=[(m-ln s)*1e4/20, (ln s2h-m)*1e4/30, ln(s/sp_prev)*1e4/20, +1 isBuy/-1 sell]; even xe=[sd*1e4/20, 2*pi-1, phase0, phase1, phase2, ln(amountY/20)/1.2, min(dt-1,12)/6 on first own trade of step else 0, bias]. h_k=tanh(Wo_k.xo+We_k.xe); dAsk=CAP*tanh(sum V_k h_k(xo,xe)/CAP); dBid=CAP*tanh(sum V_k h_k(-xo,xe)/CAP); fee=1-(1-fee)*exp(-d*1e-4), clamp [0,0.1]. Shared output weights V (16) are the only trained parameters.","input_mapping_note":"brief lists both \"belief mean relative to own spot\" and \"w\"; w=ln s-m is the negative of the first, so the w slot carries this trade's signed own-price move ln(s/sp_prev) instead. pi = arb probability of this step's first own trade read from COMBO comp (1/0 when unambiguous). \"inferred phase\" = COMBO phase 0/1/2. No rival-pin input beyond s2h (R).","hidden_weights":"random.Random(16016): Wo,We ~ N(0,1)*2/sqrt(12), generated in head.hidden_weights()","CAP_bps":8.0,"V":[0.060908662028026195,0.02997183764594874,-0.4158719998856596,-0.029048105411809605,0.19880175209919515,0.022149587220239277,0.1382260184888461,-0.3081631178168128,-0.1168933772467558,-0.15606486085914498,0.034433109869034556,-0.4475294509624875,-0.3356700322633093,0.15556911886865982,0.014435005705924434,-0.07573903818716743],"V_sha256":"b7bc57da2308dd69b7340f6e9293ce0e553ee7e08dd80142a1ed566366b4fb33","file_sha256":{"head.py":"f068b488db198f0a7b4b27a48e70942b6c18848d5cfe46c958028c104298722e","combo.py":"c5b711a0b4fb4783e8879a484bf87a427171f4a11aa5eb5ec3513c24ead1a796","twocomp.py":"9d3c1833321493fb51bbde4e1b4dabc2f8978bd6d774ee2af8dc8aed13f0f0af","qmref.py":"1dfc9b7a3c2237ff7a93dd9f0af54b9bdab16f2ce0a7eb00784cb4432598db92","blob.bin":"3ac7cb8811ecfb6a17f499aa4e9b70dd55f7fb6432b3aec3b904e11a3dece38e","ev.py":"37b643256b0249528f9a1ba1054bc0216fe59c1193f082d708d910192b039819","train.py":"19a0eec57ac404057c361c258194dc09961926f2b4f7feddb96dac23916a02a1","final.py":"6443589ccf35b3456a27b0059e3ea1d27211edec66b10ce92ae534bac0fbb514"}},"training":{"method":"sep-CMA-ES (diagonal covariance, stdlib), lambda=6, mu=3, n=16, CMA RNG seed 424242+gen, start V=0 (zero output). Common randomness: each generation all candidates on the same 256-seed block, paired against cached COMBO edges on that seed. Blocks rotate 64000..64255, 64256..64511, 64512..64767, 64768..65023.","budget":"benchmark: ~270-280 s per generation (1,536 runs, 4 cores); capped at 4 generations (one full rotation) to reserve full Q scoring. First run sigma0=1.0 aborted after gen 0 (all 6 candidates -0.59..-4.86 vs COMBO); restarted at sigma0=0.2.","curve":[{"run":"sigma0=1.0 aborted","gen":0,"block":[64000,64255],"evaluations":1536,"cand_gains_vs_COMBO":[-1.320783,-2.334409,-1.243097,-0.591969,-1.325505,-4.860502],"cand_se":[0.051642,0.06421,0.059372,0.043848,0.06525,0.0899],"best":-0.591969,"zero_gain":0.0,"sigma_after":0.92815,"secs":270.5},{"run":"sigma0=0.2 used","gen":0,"block":[64000,64255],"evaluations":1536,"cand_gains_vs_COMBO":[-0.121889,-0.1899,-3.6e-05,0.032217,-0.02342,-0.276473],"cand_se":[0.022717,0.027723,0.022532,0.021549,0.026147,0.032226],"best":0.032217,"zero_gain":0.0,"sigma_after":0.18604,"secs":269.4},{"run":"sigma0=0.2 used","gen":1,"block":[64256,64511],"evaluations":1536,"cand_gains_vs_COMBO":[-0.010964,-0.025693,-0.054154,-0.054142,-0.089199,-0.007461],"cand_se":[0.025385,0.023367,0.025649,0.025383,0.02638,0.026068],"best":-0.007461,"zero_gain":0.0,"sigma_after":0.17954,"secs":278.7},{"run":"sigma0=0.2 used","gen":2,"block":[64512,64767],"evaluations":1536,"cand_gains_vs_COMBO":[-0.14457,0.043435,0.048588,-0.081797,0.028895,-0.027344],"cand_se":[0.030736,0.022498,0.022032,0.028646,0.022174,0.020112],"best":0.048588,"zero_gain":0.0,"sigma_after":0.17062,"secs":277.8},{"run":"sigma0=0.2 used","gen":3,"block":[64768,65023],"evaluations":1536,"cand_gains_vs_COMBO":[-0.0536,-0.050719,-0.099251,-0.019056,-0.088254,-0.05584],"cand_se":[0.026826,0.024867,0.027586,0.022599,0.024494,0.02802],"best":-0.019056,"zero_gain":0.0,"sigma_after":0.16608,"secs":276.3}],"selection":"Candidates on full choose 64000..65023: zero (=COMBO, gain 0 by construction) and CMA mean after gen 3 (+0.016350 se 0.010841). Picked the CMA mean; frozen (hashes above) before any report seed ran."},"seed_coverage":{"choose":"64000..65023 complete (1024) for HEAD, COMBO, qt1; training touched only these","report":"10000..12047 complete (2048) for HEAD, COMBO, qt1, each seed run once per policy after freeze; no report seed used for training/selection","replica":"unmodified replica2.run(seed, policy), 10000 steps, default cfg, fresh policy per seed"},"results":{"choose":{"seeds":"64000..65023","HEAD-COMBO":{"n":1024,"mean_diff":0.01635,"se":0.010841,"worst5":[[64574,-1.804488],[64884,-1.493134],[64861,-1.211173],[64901,-1.143766],[64715,-1.135867]],"wins":551,"sum_d":16.741929,"sum_d2":123.400825},"HEAD-qt1":{"n":1024,"mean_diff":0.625954,"se":0.019466,"worst5":[[64101,-2.19122],[64651,-1.156817],[64859,-1.016143],[64452,-1.004846],[64761,-0.945666]],"wins":880,"sum_d":640.977087,"sum_d2":798.149327},"COMBO-qt1":{"n":1024,"mean_diff":0.609605,"se":0.019401,"worst5":[[64101,-2.776983],[64845,-1.163868],[64374,-1.023434],[64795,-0.983171],[64555,-0.946189]],"wins":875,"sum_d":624.235158,"sum_d2":774.814573},"mean_edge":{"HEAD":536.34645,"COMBO":536.3301,"qt1":535.720495}},"report":{"seeds":"10000..12047","HEAD-COMBO":{"n":2048,"mean_diff":0.017715,"se":0.007374,"worst5":[[10256,-1.323861],[11651,-1.308247],[11176,-1.266323],[11672,-1.258856],[10671,-1.222146]],"wins":1098,"sum_d":36.280655,"sum_d2":228.623094},"HEAD-qt1":{"n":2048,"mean_diff":0.615106,"se":0.013826,"worst5":[[10146,-3.440009],[11808,-2.001334],[11376,-1.65557],[10162,-1.603538],[10519,-1.495166]],"wins":1757,"sum_d":1259.737348,"sum_d2":1576.260725},"COMBO-qt1":{"n":2048,"mean_diff":0.597391,"se":0.013388,"worst5":[[10146,-3.552497],[11062,-2.334849],[11808,-1.822738],[10519,-1.782691],[11376,-1.750296]],"wins":1753,"sum_d":1223.456693,"sum_d2":1482.295158},"mean_edge":{"HEAD":533.877937,"COMBO":533.860222,"qt1":533.262831}},"spot_edges":{"10000":{"qt1":443.101741,"COMBO":444.19533,"HEAD":444.328626},"10001":{"qt1":571.454483,"COMBO":572.172682,"HEAD":572.638843},"10002":{"qt1":525.908441,"COMBO":526.27476,"HEAD":526.195351}}},"findings":["Frozen 16-unit symmetric residual head on COMBO: report 10000..12047 HEAD-COMBO +0.017715 se 0.007374 (n 2048, 1098 wins); HEAD-qt1 +0.615106 se 0.013826. Choose: +0.016350 se 0.010841 vs COMBO.","The gain over COMBO is small (~2.4 se on report) and far below the +0.5-over-qt1 increment needed beyond COMBO; it does not close the gap to #1. HEAD-qt1 >= +0.5 is almost entirely COMBO's.","Most random directions of size ~1-4 bps around COMBO lose (gen0 at sigma0=1.0: -0.59..-4.86), suggesting COMBO+qm tables sit near a local optimum for smooth fee corrections; with only 4 generations (24 candidates) CMA barely moved (sigma 0.2 -> 0.166).","Choose-window gain is in-sample for the selection of the CMA mean (same seeds trained on); the report figure is the clean out-of-sample measure."],"runtime":{"machine":"4 cores, Linux x86_64, CPython 3.12.3 stdlib only","per_run_s":{"qt1":0.23,"COMBO":0.38,"HEAD":0.7},"training_s":1372.6999999999998,"choose_scoring_s":247.7,"report_scoring_s":691.5},"limitations":["Replica numbers only; not board/engine scores; no Solidity, no gas measured.","Training compute small: 4 generations x 6 candidates, sep-CMA (diagonal) not full CMA.","Only V trained; hidden layer random-fixed; inputs exclude per-run sigma/order-rate estimates.","True arb flag not used anywhere (not even as label)."],"commands":["git clone https://github.com/gtmcknight/amm-sim test/scratch/amm-sim && git -C test/scratch/amm-sim checkout 4812a6382311624e1506aca131ca35c378f930d9","curl each artifact to test/scratch/art/<sha>; sha256sum -c","cd test/scratch && python3 -I ext.py   # writes w/twocomp.py, w/combo.py","curl -LsSf -o w/blob.bin https://raw.githubusercontent.com/identity-md-launches/goal-3170-beat-network-s-amm-entry/31ff8ed128b463cde10ea0a71d354f8c54d4464a/reference/qt1/blob.bin (and qmref.py)","(cd amm-sim/research/solved && python3 -I gates.py)","cd w && python3 -I bench.py","python3 -I train.py 1  (x4, after editing SIG0 1.0->0.2 and moving the aborted state aside)","python3 -I final.py choose_m4 head 64000 65023 \"$(cat state/Vmean.json)\"; python3 -I final.py choose_qt1 qt1 64000 65023","python3 -I final.py report_head head 10000 12047 \"$(cat state/Vmean.json)\"; python3 -I final.py report_combo combo 10000 12047; python3 -I final.py report_qt1 qt1 10000 12047","python3 -I agg.py"],"versions":{"python":"3.12.3","deps":"none (stdlib)"},"data_regeneration":"All per-seed edges are deterministic: rerun final.py with the commands above; state jsonl sha256 (row order is imap_unordered order, so hashes depend on order; compare sorted values)","data_sha256":{"combo":"bec15fea3e62a28aaa5b132e87529a01fd0afaae11db6dc90bf908b94c38a70e","choose_m4":"1660a80b1c96bda724598dc49cf25cde357445debdef926d1ff6beb6cae777b0","choose_qt1":"1f182c14852c170b90cac1d14ea0cea16a3629a3449a803955070c289ca285de","report_head":"33d1b78878512c322f8b030f0e78e9d82ee86046910e86e70dfeb9402e1a1fe2","report_combo":"79f5d8d9248461101315b8803226e5ed95f593d0202bda6cc9187cdcda3408d9","report_qt1":"a841ca4904211b3c32a6d1ec651e31814734171c7f8adb22fefaeaf18badac5f"},"source":{"encoding":"zlib level 9 then base64; paths relative to test/scratch/w (ext.py at test/scratch)","files":{"head.py":"eNqVVm1v40QQ/u5fMSofWLe2E6e9cGrJibtSKBLXgwOloKqK1vGmMbG9Pu+6Sfj1PLPeJG0PEESKsrM7O/PMMy+bo6MjU2ysUnVstlWlbFvMz+nyw/t3H+iEdK2oVabIO1nSQilaKpmTSMdkZb2kri6siQjnKm7VolRzW+ia9obCJLhsVV7Yc/pVbkvdkniw1XxVFw9LG1Icw3hTFnM5iuh72RlTyJoWRWlVGzl3bVcqknVOnyp4zEplAvFQ2GWXJXNdDQ7GBrKqYlNU9M3Z63Qkx6evIwauZDtfDowuH1UeJj6sAf261pe6arD6ZFNGTuLn9x/VIiTZquCHXNW2sNv331JhVWUo6wAJXPggvjS01u0qoV9AQi7bnMoia2W7hUq5vaBaqdwQ8GU6abYR2bWG0GBNGXOpYDYJgh+LR6zqprPGXSSh1zXNZVlmcr5i8iEuC2M1LD8WskcP5z1BZKy0KrygBgkAf38wdioeag3Gk+C6yBEGlXIL1UWxUUhbC7i6ArHSdiCHFi2kjzffz365uvoWlhwKu1SE/JolzOWkOwuAtFZMsqGpc1KC1pqdEOk8P9ijjaYJ3ZGo4hJVEA5GwwzoBAujZVyFg1O3UdbCDEwza1r1uFM6SalYUGHedVtSpVEUp3QPB+oRUew9bJRzYHJ/a3TcFHEaUbOcTIb9j5dGzo2sdFfb36EdDtJkxKVaIrmDcUS9+eVsJTYhwSoXtLjVs1WCKE7oVvFK4Wj3EbecrnnZ5QDCLGWF5DSXXVWHMJW/NSvWm9Dl25+OnTnTVbMVTfF1fnQEgwOcstX8XZH/s278XPkuawzDRU982ftIY5HGEMNjtWlEnB+nKj4LI5qXsmqQOKvpDowMk/SeseckUQ9um6F/6rTl7C4WRtkw+HjoXZF126/jN0aVZUTr+E28Ds+pKtqWy6ovV3pA5Roya9mwJ5EVeUTSrEJSGzm35Tah6WSIXPp+2+0GR0dHQVE1urVUSbtEh7qCDFwdun4hf3zJQhDsapPjHQ/TcXBzHdHNB3yv3FZEZxG9DkASxNfJMHj/9rfvsETUQRDk6Oula4OZr19h0JqTfcWfB8xliws9kOSj+3FaaAczZyfJEGOC4SbmU2vFDY/Fm6vQXb119X7XJg88vAT4TkM65ovHNKIF5t0MlLH1B4Wb4f1ne9fhfW9J/R9LV/9iqVVolRrQIhgFC7eg65bpekFFuGMIE11MI/QuvspzYiQumMxROXQ77G71zF2vyZ8103D74W51f4FJwesrrPfHvLNWd8N7xLJxvycsp15OvTzy8sjLp14+9fKZl8+8/MrLr7w89vLYy18dEDC+tfYItEegPQLtEWiPQHsE2iPg370lEHMyoSnCw5GrCte8isc1l0z2D8cxjp+mh0v2qQbsDngTHfzZUeaPcPsLdDLGTOSmB2cQ7W4MXXMOXc/4rHBiZ7MCz/NshnIuFxFlpc5mjWu76eQGD/uTBJquUa0Ik/2VvXJ40IGVZAoiSzxJYhryuJ5yk9faEtvrx/YdCoaDv7l+fvNP1XIWWFnWW9FbQwQ7tLqe2VbmyqMtzAyDCGNlgy+/oSjOiHijzSbfSXjCmJ+kw2HCr59f/DEZPgkqtzzVwbzzZUGURYKaIVc275hmr7rI8HeDS95xmPw3LAdqwMQ+xvNdhnubBxJgfYtEbi6o5LXLb6kfhEHZzHeY+H/C/kpTYHuOauWChQ+8XGIe0mSCieC4FjyccOB0ULQ4GfYnSMIBXrPkPt6jnNf2iaZIXx6k/cHoYMA966LPWQVCS8Oz6ZXjXBwCcaSOluGO8oq1UnWGqE9Z9aCJHSTiYMTH0RO9e/6fhuDefXGYws7+48HACCvQFRNe/0WppRUcNIIMn8vpC3kURgE9++xBVnIjONt4V8chAI+GCf+4fxGiKmqR295fOuL9ceI6Aptv9hQ+y0Iu0bU8Ud3A7et/P3UPlSj9mxP32Y2xE+7GQf/SS8cqHvsL1Nhn2tkL7Wyn/XmxcxD8YHJmNgJgI74Ohv7uQIYvRgGb6Rds60W1P++BvwD9SIsT","ev.py":"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","train.py":"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","final.py":"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","bench.py":"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","../ext.py":"eNptUctu2zAQvOsrCORAMlZoiXqQSiD00E/o1UDNx8oiGpECRcP235dJkyAJyht3ZnZnZ92yhphQhHK7beWstvnZ6cLChCbwBn6rKUEkq0pziRYV/0CkjwXKL41hBf8PycxgnT+N+JymB4kpi6AsoU/IjYk5b+FK3sSv2mWMwExYVvcMJOLj8bje0hz8wRN2/4MefK7gMnN+UbaBimYmqXS5XYR0jh4t7BTDeSU13eGDx0Uy42e3WMW0t42R7WA1V6apoeuF7gR0XMtKyx6k4W3GOXS6rXmnBys6IyQ0RleirnMBl/juDv0MFjAtTPjfgL7SoHojzSC5qIUWum2UrPpBSVHVTd0PRg4ADbRW8a7n3OrWymxgaHrbfhnwmiW+7NMlvOTC1luGLznIS3QJSDJ5+XdOJujwnWECLaYQkUfOI4K/9PkQ0Ee0RucT8SV6OzXb5hdv5L053vkSR/1xQspmuFp3gi3lT/EXXVqtLA==","agg.py":"eNqFVcFu4zYQvesrdFmQTBlZdh23VcACaVpgD1vsIj4KhkCJtKXaomRSTtwN8u+dIWWvvbttfTDJN+SbN8PhqGn7zg7xX64zvJVDzT/vmpKX0unFnNfS1bjsd3JYd7aNPoidbEslY5O92jxdZTafrmIwxTZuTNzKniJTsuukcrzrtaHEDXLQk3cuQcuOvDOMvUVKr+NPVPKSL1kWxfBTIqcyd6vbEv64Y57VIetydW/EDqgUu3dKuENLj9565AXaPTwL+M3XllY4NTHeg9MCI0zc3g4Ud9PjbctubmbXRybU3E7ZxDB/yOrhYAFuqoEaATnS0hSqWa+F7Q5G0ZYvGAfmsHIaly+ddcOdyHPHA3wEdDV68SE5yLlWEFCe3a1W/KUxLuj/Nb0WwwEt1IldeWeIzM7QDDAWPQorzUbTxTxNU764S2dzdv80gtMUweksnf/MorriVcX3lfhASVV3ndNFOyeM47Jry26cBst+mBJ2X1teWb63eMRqrJei1lKFnSNwcXZE/NkIo1F8idFQCr4fGadVGPZhAPYnBMOwx4FlsXROQ12eEyXErnEDXbIIb+BUhsCcjYmA7CmonIuyYRMsmiWDBEXdYRD+DkNg4pU4rZUjGfEZSxJM2Y+Ek/d/PPx++/jxz98+kuwTDckClQHHkAIatBO/8QRXZ9gXiVYbTTLvFA8LBMcM+GMBCPTAEJaegL3xKA5ZvBDqbzFJ8Bp/+laov6Knr4XirXn0WugZ/g+hfsOlUA98EervCYW6vgOZLhCgPVzI3uI7Xpw4AlidQO8oYPWIXbz4U8XC39T/z9gbi5bi3FvoZWupWul7C2Hs/uHf9xSu2UwLWfqKOh2IqoN91qdycgMf5CbLfSj2YASs+EYbUecEBrLi5a6rtrj0EwD0s9wd5NB08IIBxqUDuJJGFRsJ77p4dkVIAZIgcjJDGQLkNAKldoOnhRGWn7Xt/HGRJikH4SBfrgdt/QlcwianKxcYKqAI6at9dxlyUjdItMLCz4kPkqyEH+kDDxSpmCZpPCaEsB+CdXm2psksPjg0Rf1WvK6z8FVIysVcm6pTmuK3IoGX31vtXMj3mhNbEpZYaBCU8V8YS5T2m4PAtb9fgv0j6f+GOtbPYRwshBum68bIXZiW4KkO0ySZ6OMQ5nKzwQl8SHx8zlYFailAGoTZb7/ArpazuwWAoH/8nCUB+55eltT6qJoN3MKl3n47OlJykBeU5ruU337x/seHCTkJLZRfdGV+2Yf5Ve/l142XX3Xdt8i/AnVoewq6rz7EmDpf/Jy84APobWMGet7v6Os2w1i3oZVuR21ex9kNTPDZgyfeGKXNIFJ4fHHgwq57wQdkDPz8A6aTsLU="},"sha256":{"head.py":"f068b488db198f0a7b4b27a48e70942b6c18848d5cfe46c958028c104298722e","ev.py":"37b643256b0249528f9a1ba1054bc0216fe59c1193f082d708d910192b039819","train.py":"19a0eec57ac404057c361c258194dc09961926f2b4f7feddb96dac23916a02a1","final.py":"6443589ccf35b3456a27b0059e3ea1d27211edec66b10ce92ae534bac0fbb514","bench.py":"429d500a825f3d373e63ad3845bd71c3fbf8937ea21f7147a2568dd17c931f4d","../ext.py":"94e18e0b168d481636a852dd476c5687f9bc96d3ece227359fc8be2979b8102e","agg.py":"5c8d5133de1146ca3840195874fa25e257a0b6df517af50e24bce8bdaf28a1c7"}},"compact_inference":"head.py (in source.files) is the stdlib inference: Head(blob_path, V) subclasses Combo; V and hidden seed above."}