diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,27976 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 7.0, + "eval_steps": 500, + "global_step": 12698, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0027563395810363835, + "grad_norm": 16.792909573837832, + "learning_rate": 1.2598425196850394e-07, + "loss": 0.6602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31006354093551636, + "step": 5, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 0.005512679162072767, + "grad_norm": 16.31603541982407, + "learning_rate": 2.8346456692913386e-07, + "loss": 0.701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23665527999401093, + "step": 10, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 0.008269018743109152, + "grad_norm": 15.289678330204236, + "learning_rate": 4.409448818897638e-07, + "loss": 0.6823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3267960846424103, + "step": 15, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 0.011025358324145534, + "grad_norm": 14.279750905086196, + "learning_rate": 5.984251968503937e-07, + "loss": 0.6217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2942882776260376, + "step": 20, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 0.013781697905181918, + "grad_norm": 13.671644658668507, + "learning_rate": 7.559055118110237e-07, + "loss": 0.6323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3043963313102722, + "step": 25, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 0.016538037486218304, + "grad_norm": 11.4158146233281, + "learning_rate": 9.133858267716536e-07, + "loss": 0.6308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3370758295059204, + "step": 30, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 0.019294377067254686, + "grad_norm": 8.354178699756039, + "learning_rate": 1.0708661417322836e-06, + "loss": 0.62, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.30885541439056396, + "step": 35, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 0.022050716648291068, + "grad_norm": 6.5400220687367625, + "learning_rate": 1.2283464566929135e-06, + "loss": 0.5769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2766607999801636, + "step": 40, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 0.024807056229327454, + "grad_norm": 7.481522487728226, + "learning_rate": 1.3858267716535435e-06, + "loss": 0.5818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13531705737113953, + "step": 45, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 0.027563395810363836, + "grad_norm": 4.553513660511177, + "learning_rate": 1.5433070866141734e-06, + "loss": 0.5165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24960283935070038, + "step": 50, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 0.03031973539140022, + "grad_norm": 5.639050164188589, + "learning_rate": 1.7007874015748034e-06, + "loss": 0.4989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23326541483402252, + "step": 55, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 0.03307607497243661, + "grad_norm": 2.9165738464531294, + "learning_rate": 1.8582677165354333e-06, + "loss": 0.4615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22679060697555542, + "step": 60, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 0.03583241455347299, + "grad_norm": 1.8685520274481575, + "learning_rate": 2.015748031496063e-06, + "loss": 0.4306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21953804790973663, + "step": 65, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 0.03858875413450937, + "grad_norm": 1.4055742204527895, + "learning_rate": 2.173228346456693e-06, + "loss": 0.4163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20238222181797028, + "step": 70, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 0.041345093715545754, + "grad_norm": 1.2025265389654956, + "learning_rate": 2.330708661417323e-06, + "loss": 0.4175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21335455775260925, + "step": 75, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 0.044101433296582136, + "grad_norm": 1.0244843782448096, + "learning_rate": 2.488188976377953e-06, + "loss": 0.4118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1873829960823059, + "step": 80, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 0.046857772877618525, + "grad_norm": 0.8485834208782914, + "learning_rate": 2.645669291338583e-06, + "loss": 0.3965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18655532598495483, + "step": 85, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 0.04961411245865491, + "grad_norm": 0.9360790762670237, + "learning_rate": 2.803149606299213e-06, + "loss": 0.3936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19727565348148346, + "step": 90, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 0.05237045203969129, + "grad_norm": 0.6893561143656093, + "learning_rate": 2.9606299212598427e-06, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20135433971881866, + "step": 95, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 0.05512679162072767, + "grad_norm": 0.600154592425302, + "learning_rate": 3.118110236220473e-06, + "loss": 0.3938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20234315097332, + "step": 100, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 0.057883131201764054, + "grad_norm": 0.6189359380729005, + "learning_rate": 3.2755905511811026e-06, + "loss": 0.3518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15832287073135376, + "step": 105, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 0.06063947078280044, + "grad_norm": 0.5426616732316848, + "learning_rate": 3.433070866141733e-06, + "loss": 0.3527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15709516406059265, + "step": 110, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 0.06339581036383682, + "grad_norm": 0.6040270943129593, + "learning_rate": 3.5905511811023625e-06, + "loss": 0.3675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20795878767967224, + "step": 115, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 0.06615214994487321, + "grad_norm": 0.5531403206110393, + "learning_rate": 3.748031496062993e-06, + "loss": 0.3531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16832762956619263, + "step": 120, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 0.0689084895259096, + "grad_norm": 0.5515875158934355, + "learning_rate": 3.905511811023622e-06, + "loss": 0.346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1948687732219696, + "step": 125, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 0.07166482910694598, + "grad_norm": 0.49384266360930623, + "learning_rate": 4.062992125984252e-06, + "loss": 0.3058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14836347103118896, + "step": 130, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 0.07442116868798236, + "grad_norm": 0.5354176939789098, + "learning_rate": 4.220472440944882e-06, + "loss": 0.304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16761240363121033, + "step": 135, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 0.07717750826901874, + "grad_norm": 0.5102781217988436, + "learning_rate": 4.377952755905512e-06, + "loss": 0.2918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16363434493541718, + "step": 140, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 0.07993384785005513, + "grad_norm": 1.0549856037766225, + "learning_rate": 4.535433070866142e-06, + "loss": 0.3548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1325737088918686, + "step": 145, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 0.08269018743109151, + "grad_norm": 0.4720517094871231, + "learning_rate": 4.692913385826772e-06, + "loss": 0.3415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1472627818584442, + "step": 150, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 0.08544652701212789, + "grad_norm": 0.5575202941023607, + "learning_rate": 4.850393700787402e-06, + "loss": 0.335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17119142413139343, + "step": 155, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 0.08820286659316427, + "grad_norm": 0.5061579926847838, + "learning_rate": 5.007874015748032e-06, + "loss": 0.3413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16755619645118713, + "step": 160, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 0.09095920617420065, + "grad_norm": 0.5180748181747808, + "learning_rate": 5.1653543307086615e-06, + "loss": 0.3331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17087745666503906, + "step": 165, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 0.09371554575523705, + "grad_norm": 0.46742974593558306, + "learning_rate": 5.322834645669291e-06, + "loss": 0.3158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16173171997070312, + "step": 170, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 0.09647188533627343, + "grad_norm": 0.46061095391798, + "learning_rate": 5.480314960629922e-06, + "loss": 0.3198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16262826323509216, + "step": 175, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 0.09922822491730982, + "grad_norm": 0.4882548484316861, + "learning_rate": 5.637795275590552e-06, + "loss": 0.3112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17065249383449554, + "step": 180, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 0.1019845644983462, + "grad_norm": 0.5650759180248933, + "learning_rate": 5.795275590551181e-06, + "loss": 0.3159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16450242698192596, + "step": 185, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 0.10474090407938258, + "grad_norm": 0.5243564773229903, + "learning_rate": 5.952755905511811e-06, + "loss": 0.3083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15657715499401093, + "step": 190, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 0.10749724366041896, + "grad_norm": 0.5271600001572089, + "learning_rate": 6.110236220472442e-06, + "loss": 0.3118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1518152952194214, + "step": 195, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 0.11025358324145534, + "grad_norm": 0.5089512056531876, + "learning_rate": 6.2677165354330715e-06, + "loss": 0.3114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15263652801513672, + "step": 200, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 0.11300992282249173, + "grad_norm": 0.4694611914702037, + "learning_rate": 6.425196850393701e-06, + "loss": 0.3079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15220537781715393, + "step": 205, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 0.11576626240352811, + "grad_norm": 0.5194215910173066, + "learning_rate": 6.5826771653543306e-06, + "loss": 0.3079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17008039355278015, + "step": 210, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 0.1185226019845645, + "grad_norm": 0.5292466972694881, + "learning_rate": 6.740157480314962e-06, + "loss": 0.3, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1381671130657196, + "step": 215, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 0.12127894156560089, + "grad_norm": 0.49118404652110687, + "learning_rate": 6.897637795275591e-06, + "loss": 0.2969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1455690562725067, + "step": 220, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 0.12403528114663727, + "grad_norm": 0.5109318438443031, + "learning_rate": 7.055118110236221e-06, + "loss": 0.2994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15448932349681854, + "step": 225, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 0.12679162072767364, + "grad_norm": 0.4515385820410411, + "learning_rate": 7.21259842519685e-06, + "loss": 0.2891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1368420124053955, + "step": 230, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 0.12954796030871002, + "grad_norm": 0.4885904301434691, + "learning_rate": 7.3700787401574816e-06, + "loss": 0.296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17897871136665344, + "step": 235, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 0.13230429988974643, + "grad_norm": 0.5029750822532669, + "learning_rate": 7.527559055118111e-06, + "loss": 0.2962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15558336675167084, + "step": 240, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 0.1350606394707828, + "grad_norm": 0.8341466739747923, + "learning_rate": 7.68503937007874e-06, + "loss": 0.2743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09492023289203644, + "step": 245, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 0.1378169790518192, + "grad_norm": 0.45934107663422186, + "learning_rate": 7.84251968503937e-06, + "loss": 0.2858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14476755261421204, + "step": 250, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 0.14057331863285558, + "grad_norm": 0.5224031151309098, + "learning_rate": 8.000000000000001e-06, + "loss": 0.2854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14522704482078552, + "step": 255, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 0.14332965821389196, + "grad_norm": 0.48715390649914586, + "learning_rate": 8.157480314960631e-06, + "loss": 0.2906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1330384612083435, + "step": 260, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 0.14608599779492834, + "grad_norm": 0.5083705209441465, + "learning_rate": 8.31496062992126e-06, + "loss": 0.2847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14883795380592346, + "step": 265, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 0.14884233737596472, + "grad_norm": 0.4669624706565387, + "learning_rate": 8.47244094488189e-06, + "loss": 0.2697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11565808951854706, + "step": 270, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 0.1515986769570011, + "grad_norm": 0.5216754602866563, + "learning_rate": 8.629921259842521e-06, + "loss": 0.2851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14072370529174805, + "step": 275, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 0.1543550165380375, + "grad_norm": 2.933235131794009, + "learning_rate": 8.78740157480315e-06, + "loss": 0.342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24734140932559967, + "step": 280, + "valid_targets_mean": 1886.0, + "valid_targets_min": 1111 + }, + { + "epoch": 0.15711135611907387, + "grad_norm": 1.7250168287713792, + "learning_rate": 8.94488188976378e-06, + "loss": 0.4262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21419323980808258, + "step": 285, + "valid_targets_mean": 1497.6, + "valid_targets_min": 893 + }, + { + "epoch": 0.15986769570011025, + "grad_norm": 1.2936578360013578, + "learning_rate": 9.10236220472441e-06, + "loss": 0.3766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1324019581079483, + "step": 290, + "valid_targets_mean": 1356.0, + "valid_targets_min": 914 + }, + { + "epoch": 0.16262403528114663, + "grad_norm": 1.19752648970076, + "learning_rate": 9.259842519685041e-06, + "loss": 0.3588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.152134507894516, + "step": 295, + "valid_targets_mean": 1324.5, + "valid_targets_min": 680 + }, + { + "epoch": 0.16538037486218302, + "grad_norm": 1.0622790771522501, + "learning_rate": 9.41732283464567e-06, + "loss": 0.3482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16145223379135132, + "step": 300, + "valid_targets_mean": 1931.8, + "valid_targets_min": 876 + }, + { + "epoch": 0.1681367144432194, + "grad_norm": 1.1340136821153812, + "learning_rate": 9.5748031496063e-06, + "loss": 0.3342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14782202243804932, + "step": 305, + "valid_targets_mean": 1292.2, + "valid_targets_min": 734 + }, + { + "epoch": 0.17089305402425578, + "grad_norm": 1.0645038411641383, + "learning_rate": 9.73228346456693e-06, + "loss": 0.3379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13295873999595642, + "step": 310, + "valid_targets_mean": 1469.6, + "valid_targets_min": 655 + }, + { + "epoch": 0.17364939360529216, + "grad_norm": 1.068949353247098, + "learning_rate": 9.88976377952756e-06, + "loss": 0.3079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16226810216903687, + "step": 315, + "valid_targets_mean": 1548.4, + "valid_targets_min": 737 + }, + { + "epoch": 0.17640573318632854, + "grad_norm": 1.1296664318778888, + "learning_rate": 1.0047244094488188e-05, + "loss": 0.3248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13937769830226898, + "step": 320, + "valid_targets_mean": 1316.5, + "valid_targets_min": 759 + }, + { + "epoch": 0.17916207276736493, + "grad_norm": 1.012822901800621, + "learning_rate": 1.0204724409448821e-05, + "loss": 0.303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11625304073095322, + "step": 325, + "valid_targets_mean": 1249.6, + "valid_targets_min": 753 + }, + { + "epoch": 0.1819184123484013, + "grad_norm": 1.1188873356683422, + "learning_rate": 1.0362204724409451e-05, + "loss": 0.3011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19144614040851593, + "step": 330, + "valid_targets_mean": 1833.0, + "valid_targets_min": 927 + }, + { + "epoch": 0.18467475192943772, + "grad_norm": 1.0549085915259502, + "learning_rate": 1.051968503937008e-05, + "loss": 0.3108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14709064364433289, + "step": 335, + "valid_targets_mean": 1338.4, + "valid_targets_min": 679 + }, + { + "epoch": 0.1874310915104741, + "grad_norm": 1.4507432308924364, + "learning_rate": 1.067716535433071e-05, + "loss": 0.298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13247793912887573, + "step": 340, + "valid_targets_mean": 1244.6, + "valid_targets_min": 771 + }, + { + "epoch": 0.19018743109151048, + "grad_norm": 0.9220488125081328, + "learning_rate": 1.083464566929134e-05, + "loss": 0.2791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13263286650180817, + "step": 345, + "valid_targets_mean": 1732.5, + "valid_targets_min": 986 + }, + { + "epoch": 0.19294377067254687, + "grad_norm": 1.0748813023760544, + "learning_rate": 1.0992125984251969e-05, + "loss": 0.2876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12777942419052124, + "step": 350, + "valid_targets_mean": 1389.9, + "valid_targets_min": 1103 + }, + { + "epoch": 0.19570011025358325, + "grad_norm": 0.9662253513600024, + "learning_rate": 1.1149606299212599e-05, + "loss": 0.2906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1572059541940689, + "step": 355, + "valid_targets_mean": 1593.2, + "valid_targets_min": 675 + }, + { + "epoch": 0.19845644983461963, + "grad_norm": 0.9900851305975743, + "learning_rate": 1.1307086614173228e-05, + "loss": 0.2878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17566180229187012, + "step": 360, + "valid_targets_mean": 1926.6, + "valid_targets_min": 974 + }, + { + "epoch": 0.201212789415656, + "grad_norm": 1.1162299549240424, + "learning_rate": 1.1464566929133861e-05, + "loss": 0.2934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17048877477645874, + "step": 365, + "valid_targets_mean": 1953.1, + "valid_targets_min": 869 + }, + { + "epoch": 0.2039691289966924, + "grad_norm": 1.30055795445194, + "learning_rate": 1.162204724409449e-05, + "loss": 0.2906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12949338555335999, + "step": 370, + "valid_targets_mean": 1327.2, + "valid_targets_min": 1023 + }, + { + "epoch": 0.20672546857772878, + "grad_norm": 0.9828747183043868, + "learning_rate": 1.177952755905512e-05, + "loss": 0.2925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15447880327701569, + "step": 375, + "valid_targets_mean": 1717.9, + "valid_targets_min": 709 + }, + { + "epoch": 0.20948180815876516, + "grad_norm": 1.134636534370077, + "learning_rate": 1.193700787401575e-05, + "loss": 0.277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11985255777835846, + "step": 380, + "valid_targets_mean": 1139.9, + "valid_targets_min": 658 + }, + { + "epoch": 0.21223814773980154, + "grad_norm": 1.0618190848913298, + "learning_rate": 1.2094488188976379e-05, + "loss": 0.2874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.136907160282135, + "step": 385, + "valid_targets_mean": 1444.8, + "valid_targets_min": 662 + }, + { + "epoch": 0.21499448732083792, + "grad_norm": 1.0368755067730058, + "learning_rate": 1.2251968503937009e-05, + "loss": 0.2924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10148206353187561, + "step": 390, + "valid_targets_mean": 1242.2, + "valid_targets_min": 859 + }, + { + "epoch": 0.2177508269018743, + "grad_norm": 1.0592652507591802, + "learning_rate": 1.2409448818897638e-05, + "loss": 0.2648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16245663166046143, + "step": 395, + "valid_targets_mean": 1569.2, + "valid_targets_min": 950 + }, + { + "epoch": 0.2205071664829107, + "grad_norm": 1.0365178685435077, + "learning_rate": 1.2566929133858268e-05, + "loss": 0.2772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12986180186271667, + "step": 400, + "valid_targets_mean": 1539.8, + "valid_targets_min": 686 + }, + { + "epoch": 0.22326350606394707, + "grad_norm": 0.9755497564382646, + "learning_rate": 1.2724409448818899e-05, + "loss": 0.2707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15090861916542053, + "step": 405, + "valid_targets_mean": 1610.6, + "valid_targets_min": 702 + }, + { + "epoch": 0.22601984564498345, + "grad_norm": 1.086646924732427, + "learning_rate": 1.288188976377953e-05, + "loss": 0.2644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09992213547229767, + "step": 410, + "valid_targets_mean": 1190.4, + "valid_targets_min": 779 + }, + { + "epoch": 0.22877618522601983, + "grad_norm": 1.2463012565868825, + "learning_rate": 1.303937007874016e-05, + "loss": 0.2818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14419332146644592, + "step": 415, + "valid_targets_mean": 1467.1, + "valid_targets_min": 822 + }, + { + "epoch": 0.23153252480705622, + "grad_norm": 1.1003042710807185, + "learning_rate": 1.3196850393700789e-05, + "loss": 0.2593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11426033079624176, + "step": 420, + "valid_targets_mean": 1350.5, + "valid_targets_min": 868 + }, + { + "epoch": 0.2342888643880926, + "grad_norm": 1.2487754226742487, + "learning_rate": 1.3354330708661419e-05, + "loss": 0.2698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11017901450395584, + "step": 425, + "valid_targets_mean": 1283.5, + "valid_targets_min": 781 + }, + { + "epoch": 0.237045203969129, + "grad_norm": 1.0268796247770806, + "learning_rate": 1.3511811023622048e-05, + "loss": 0.2706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11194470524787903, + "step": 430, + "valid_targets_mean": 1303.2, + "valid_targets_min": 793 + }, + { + "epoch": 0.2398015435501654, + "grad_norm": 1.0229084758034368, + "learning_rate": 1.3669291338582678e-05, + "loss": 0.268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11475589871406555, + "step": 435, + "valid_targets_mean": 1309.1, + "valid_targets_min": 806 + }, + { + "epoch": 0.24255788313120177, + "grad_norm": 1.0041334736478857, + "learning_rate": 1.3826771653543307e-05, + "loss": 0.2664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11097784340381622, + "step": 440, + "valid_targets_mean": 1397.6, + "valid_targets_min": 803 + }, + { + "epoch": 0.24531422271223816, + "grad_norm": 1.1148243301496727, + "learning_rate": 1.3984251968503938e-05, + "loss": 0.2633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09967643022537231, + "step": 445, + "valid_targets_mean": 1187.8, + "valid_targets_min": 660 + }, + { + "epoch": 0.24807056229327454, + "grad_norm": 0.9610716353135176, + "learning_rate": 1.4141732283464568e-05, + "loss": 0.2454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1072668582201004, + "step": 450, + "valid_targets_mean": 1547.0, + "valid_targets_min": 853 + }, + { + "epoch": 0.2508269018743109, + "grad_norm": 0.9762983619020633, + "learning_rate": 1.4299212598425199e-05, + "loss": 0.2651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13858476281166077, + "step": 455, + "valid_targets_mean": 1704.2, + "valid_targets_min": 719 + }, + { + "epoch": 0.2535832414553473, + "grad_norm": 1.0185390363334275, + "learning_rate": 1.4456692913385829e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09609312564134598, + "step": 460, + "valid_targets_mean": 1255.0, + "valid_targets_min": 765 + }, + { + "epoch": 0.25633958103638366, + "grad_norm": 1.0638300507371012, + "learning_rate": 1.4614173228346458e-05, + "loss": 0.2612, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1289488971233368, + "step": 465, + "valid_targets_mean": 1550.8, + "valid_targets_min": 875 + }, + { + "epoch": 0.25909592061742004, + "grad_norm": 1.1193336413962336, + "learning_rate": 1.4771653543307088e-05, + "loss": 0.2589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12733636796474457, + "step": 470, + "valid_targets_mean": 1362.0, + "valid_targets_min": 739 + }, + { + "epoch": 0.2618522601984565, + "grad_norm": 1.0107071006244766, + "learning_rate": 1.4929133858267717e-05, + "loss": 0.2684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14933910965919495, + "step": 475, + "valid_targets_mean": 1666.8, + "valid_targets_min": 760 + }, + { + "epoch": 0.26460859977949286, + "grad_norm": 1.0786718616367155, + "learning_rate": 1.5086614173228347e-05, + "loss": 0.2507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11007131636142731, + "step": 480, + "valid_targets_mean": 1319.6, + "valid_targets_min": 1046 + }, + { + "epoch": 0.26736493936052924, + "grad_norm": 0.9811501311014607, + "learning_rate": 1.5244094488188978e-05, + "loss": 0.2706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14439749717712402, + "step": 485, + "valid_targets_mean": 1727.9, + "valid_targets_min": 851 + }, + { + "epoch": 0.2701212789415656, + "grad_norm": 0.9993243280560129, + "learning_rate": 1.5401574803149607e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13957354426383972, + "step": 490, + "valid_targets_mean": 1643.2, + "valid_targets_min": 907 + }, + { + "epoch": 0.272877618522602, + "grad_norm": 1.1210014568989635, + "learning_rate": 1.555905511811024e-05, + "loss": 0.2748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11695282906293869, + "step": 495, + "valid_targets_mean": 1478.5, + "valid_targets_min": 764 + }, + { + "epoch": 0.2756339581036384, + "grad_norm": 1.0576266686412625, + "learning_rate": 1.5716535433070866e-05, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11889296770095825, + "step": 500, + "valid_targets_mean": 1522.6, + "valid_targets_min": 822 + }, + { + "epoch": 0.27839029768467477, + "grad_norm": 1.0762586402411454, + "learning_rate": 1.5874015748031498e-05, + "loss": 0.253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12489502131938934, + "step": 505, + "valid_targets_mean": 1589.0, + "valid_targets_min": 982 + }, + { + "epoch": 0.28114663726571115, + "grad_norm": 1.075895253831119, + "learning_rate": 1.6031496062992126e-05, + "loss": 0.265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12909933924674988, + "step": 510, + "valid_targets_mean": 1574.8, + "valid_targets_min": 816 + }, + { + "epoch": 0.28390297684674753, + "grad_norm": 1.111508613383734, + "learning_rate": 1.6188976377952757e-05, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12755581736564636, + "step": 515, + "valid_targets_mean": 1422.0, + "valid_targets_min": 721 + }, + { + "epoch": 0.2866593164277839, + "grad_norm": 1.2832527927150066, + "learning_rate": 1.6346456692913385e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09376775473356247, + "step": 520, + "valid_targets_mean": 1150.6, + "valid_targets_min": 741 + }, + { + "epoch": 0.2894156560088203, + "grad_norm": 1.0441644189569994, + "learning_rate": 1.650393700787402e-05, + "loss": 0.2556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13824179768562317, + "step": 525, + "valid_targets_mean": 1684.2, + "valid_targets_min": 734 + }, + { + "epoch": 0.2921719955898567, + "grad_norm": 0.9670568548403016, + "learning_rate": 1.6661417322834647e-05, + "loss": 0.2588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15274253487586975, + "step": 530, + "valid_targets_mean": 1937.1, + "valid_targets_min": 744 + }, + { + "epoch": 0.29492833517089306, + "grad_norm": 0.9333423962700258, + "learning_rate": 1.6818897637795278e-05, + "loss": 0.2551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14364033937454224, + "step": 535, + "valid_targets_mean": 1842.2, + "valid_targets_min": 620 + }, + { + "epoch": 0.29768467475192945, + "grad_norm": 1.1707048315407131, + "learning_rate": 1.6976377952755906e-05, + "loss": 0.2485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14644601941108704, + "step": 540, + "valid_targets_mean": 1637.6, + "valid_targets_min": 785 + }, + { + "epoch": 0.30044101433296583, + "grad_norm": 0.9214988635000847, + "learning_rate": 1.7133858267716537e-05, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12463974952697754, + "step": 545, + "valid_targets_mean": 1969.9, + "valid_targets_min": 740 + }, + { + "epoch": 0.3031973539140022, + "grad_norm": 1.0512521983711394, + "learning_rate": 1.7291338582677165e-05, + "loss": 0.248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1134440153837204, + "step": 550, + "valid_targets_mean": 1432.1, + "valid_targets_min": 711 + }, + { + "epoch": 0.3059536934950386, + "grad_norm": 0.9969359194700903, + "learning_rate": 1.7448818897637796e-05, + "loss": 0.2504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13817359507083893, + "step": 555, + "valid_targets_mean": 1695.6, + "valid_targets_min": 1122 + }, + { + "epoch": 0.308710033076075, + "grad_norm": 1.1018226718170476, + "learning_rate": 1.7606299212598424e-05, + "loss": 0.2475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13357987999916077, + "step": 560, + "valid_targets_mean": 1524.5, + "valid_targets_min": 722 + }, + { + "epoch": 0.31146637265711136, + "grad_norm": 1.0215280564169613, + "learning_rate": 1.776377952755906e-05, + "loss": 0.2336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10690191388130188, + "step": 565, + "valid_targets_mean": 1289.2, + "valid_targets_min": 661 + }, + { + "epoch": 0.31422271223814774, + "grad_norm": 1.009788651717605, + "learning_rate": 1.7921259842519687e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11379726231098175, + "step": 570, + "valid_targets_mean": 1519.5, + "valid_targets_min": 1011 + }, + { + "epoch": 0.3169790518191841, + "grad_norm": 0.9758846485639607, + "learning_rate": 1.8078740157480318e-05, + "loss": 0.2512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13338914513587952, + "step": 575, + "valid_targets_mean": 1731.4, + "valid_targets_min": 1277 + }, + { + "epoch": 0.3197353914002205, + "grad_norm": 1.0971214159067513, + "learning_rate": 1.8236220472440946e-05, + "loss": 0.2372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12739843130111694, + "step": 580, + "valid_targets_mean": 1486.1, + "valid_targets_min": 807 + }, + { + "epoch": 0.3224917309812569, + "grad_norm": 1.0340468485614716, + "learning_rate": 1.8393700787401577e-05, + "loss": 0.2508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1405021846294403, + "step": 585, + "valid_targets_mean": 1549.9, + "valid_targets_min": 679 + }, + { + "epoch": 0.32524807056229327, + "grad_norm": 1.0766613252507802, + "learning_rate": 1.8551181102362205e-05, + "loss": 0.2456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.117359459400177, + "step": 590, + "valid_targets_mean": 1656.0, + "valid_targets_min": 1294 + }, + { + "epoch": 0.32800441014332965, + "grad_norm": 1.018931445094327, + "learning_rate": 1.8708661417322836e-05, + "loss": 0.2471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11087405681610107, + "step": 595, + "valid_targets_mean": 1368.8, + "valid_targets_min": 790 + }, + { + "epoch": 0.33076074972436603, + "grad_norm": 1.0768523205951555, + "learning_rate": 1.8866141732283464e-05, + "loss": 0.2421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13643892109394073, + "step": 600, + "valid_targets_mean": 1765.8, + "valid_targets_min": 592 + }, + { + "epoch": 0.3335170893054024, + "grad_norm": 1.0355116282229733, + "learning_rate": 1.9023622047244098e-05, + "loss": 0.2481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13067127764225006, + "step": 605, + "valid_targets_mean": 1586.8, + "valid_targets_min": 662 + }, + { + "epoch": 0.3362734288864388, + "grad_norm": 1.1275466559561504, + "learning_rate": 1.9181102362204726e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1199248805642128, + "step": 610, + "valid_targets_mean": 1336.9, + "valid_targets_min": 622 + }, + { + "epoch": 0.3390297684674752, + "grad_norm": 1.040973194282113, + "learning_rate": 1.9338582677165357e-05, + "loss": 0.2487, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07834835350513458, + "step": 615, + "valid_targets_mean": 1097.1, + "valid_targets_min": 625 + }, + { + "epoch": 0.34178610804851156, + "grad_norm": 1.2659090059798843, + "learning_rate": 1.9496062992125985e-05, + "loss": 0.262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09738148748874664, + "step": 620, + "valid_targets_mean": 1484.9, + "valid_targets_min": 888 + }, + { + "epoch": 0.34454244762954794, + "grad_norm": 0.978952834643859, + "learning_rate": 1.9653543307086616e-05, + "loss": 0.2528, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13259857892990112, + "step": 625, + "valid_targets_mean": 1757.0, + "valid_targets_min": 635 + }, + { + "epoch": 0.3472987872105843, + "grad_norm": 1.4490528611981357, + "learning_rate": 1.9811023622047244e-05, + "loss": 0.2552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1431387960910797, + "step": 630, + "valid_targets_mean": 1476.1, + "valid_targets_min": 522 + }, + { + "epoch": 0.3500551267916207, + "grad_norm": 1.080515327902184, + "learning_rate": 1.9968503937007875e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10920877754688263, + "step": 635, + "valid_targets_mean": 1230.9, + "valid_targets_min": 697 + }, + { + "epoch": 0.3528114663726571, + "grad_norm": 0.9732638133829004, + "learning_rate": 2.0125984251968503e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1304725706577301, + "step": 640, + "valid_targets_mean": 1671.0, + "valid_targets_min": 1010 + }, + { + "epoch": 0.35556780595369347, + "grad_norm": 0.9999569199714109, + "learning_rate": 2.0283464566929134e-05, + "loss": 0.2447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12042930722236633, + "step": 645, + "valid_targets_mean": 1411.0, + "valid_targets_min": 769 + }, + { + "epoch": 0.35832414553472985, + "grad_norm": 1.2654246183762357, + "learning_rate": 2.0440944881889762e-05, + "loss": 0.2467, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10285435616970062, + "step": 650, + "valid_targets_mean": 1298.6, + "valid_targets_min": 745 + }, + { + "epoch": 0.36108048511576624, + "grad_norm": 1.1068178358048435, + "learning_rate": 2.0598425196850393e-05, + "loss": 0.2424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10256120562553406, + "step": 655, + "valid_targets_mean": 1118.9, + "valid_targets_min": 680 + }, + { + "epoch": 0.3638368246968026, + "grad_norm": 1.05681811596592, + "learning_rate": 2.0755905511811028e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14364585280418396, + "step": 660, + "valid_targets_mean": 1905.9, + "valid_targets_min": 1347 + }, + { + "epoch": 0.36659316427783906, + "grad_norm": 1.053963973101003, + "learning_rate": 2.0913385826771656e-05, + "loss": 0.2515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1761883944272995, + "step": 665, + "valid_targets_mean": 1636.0, + "valid_targets_min": 851 + }, + { + "epoch": 0.36934950385887544, + "grad_norm": 0.9790558512605747, + "learning_rate": 2.1070866141732287e-05, + "loss": 0.2361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10620364546775818, + "step": 670, + "valid_targets_mean": 1457.2, + "valid_targets_min": 903 + }, + { + "epoch": 0.3721058434399118, + "grad_norm": 1.0052522325253757, + "learning_rate": 2.1228346456692915e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13085365295410156, + "step": 675, + "valid_targets_mean": 1674.4, + "valid_targets_min": 1240 + }, + { + "epoch": 0.3748621830209482, + "grad_norm": 0.8070922606987252, + "learning_rate": 2.1385826771653546e-05, + "loss": 0.2318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09653821587562561, + "step": 680, + "valid_targets_mean": 1697.1, + "valid_targets_min": 917 + }, + { + "epoch": 0.3776185226019846, + "grad_norm": 1.0565815558781528, + "learning_rate": 2.1543307086614177e-05, + "loss": 0.2178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09989050030708313, + "step": 685, + "valid_targets_mean": 1580.9, + "valid_targets_min": 930 + }, + { + "epoch": 0.38037486218302097, + "grad_norm": 0.9300928739483568, + "learning_rate": 2.1700787401574805e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1240275576710701, + "step": 690, + "valid_targets_mean": 1655.1, + "valid_targets_min": 1227 + }, + { + "epoch": 0.38313120176405735, + "grad_norm": 0.8995472482177691, + "learning_rate": 2.1858267716535436e-05, + "loss": 0.2362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11895745247602463, + "step": 695, + "valid_targets_mean": 1565.8, + "valid_targets_min": 870 + }, + { + "epoch": 0.38588754134509373, + "grad_norm": 1.0126634115584547, + "learning_rate": 2.2015748031496064e-05, + "loss": 0.2403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11521008610725403, + "step": 700, + "valid_targets_mean": 1564.6, + "valid_targets_min": 822 + }, + { + "epoch": 0.3886438809261301, + "grad_norm": 0.9365070086520886, + "learning_rate": 2.2173228346456695e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10756641626358032, + "step": 705, + "valid_targets_mean": 1540.5, + "valid_targets_min": 1214 + }, + { + "epoch": 0.3914002205071665, + "grad_norm": 1.070324168474326, + "learning_rate": 2.2330708661417323e-05, + "loss": 0.2333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1100616455078125, + "step": 710, + "valid_targets_mean": 1268.6, + "valid_targets_min": 591 + }, + { + "epoch": 0.3941565600882029, + "grad_norm": 1.017877238521452, + "learning_rate": 2.2488188976377954e-05, + "loss": 0.2477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09052532911300659, + "step": 715, + "valid_targets_mean": 1232.5, + "valid_targets_min": 646 + }, + { + "epoch": 0.39691289966923926, + "grad_norm": 0.9001108413914594, + "learning_rate": 2.2645669291338582e-05, + "loss": 0.3429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2032996267080307, + "step": 720, + "valid_targets_mean": 5097.2, + "valid_targets_min": 1971 + }, + { + "epoch": 0.39966923925027564, + "grad_norm": 0.7192288790584641, + "learning_rate": 2.2803149606299214e-05, + "loss": 0.4105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22589395940303802, + "step": 725, + "valid_targets_mean": 5217.0, + "valid_targets_min": 1631 + }, + { + "epoch": 0.402425578831312, + "grad_norm": 0.6531507311496912, + "learning_rate": 2.296062992125984e-05, + "loss": 0.4034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2564042806625366, + "step": 730, + "valid_targets_mean": 5966.0, + "valid_targets_min": 3294 + }, + { + "epoch": 0.4051819184123484, + "grad_norm": 1.2676364981675718, + "learning_rate": 2.3118110236220473e-05, + "loss": 0.3827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16092795133590698, + "step": 735, + "valid_targets_mean": 4246.1, + "valid_targets_min": 2736 + }, + { + "epoch": 0.4079382579933848, + "grad_norm": 0.6872336127979666, + "learning_rate": 2.3275590551181107e-05, + "loss": 0.3791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15573278069496155, + "step": 740, + "valid_targets_mean": 3896.2, + "valid_targets_min": 1439 + }, + { + "epoch": 0.41069459757442117, + "grad_norm": 0.586434755181198, + "learning_rate": 2.3433070866141735e-05, + "loss": 0.3957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15307369828224182, + "step": 745, + "valid_targets_mean": 4481.8, + "valid_targets_min": 1512 + }, + { + "epoch": 0.41345093715545755, + "grad_norm": 0.5828009991520775, + "learning_rate": 2.3590551181102366e-05, + "loss": 0.3638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20178546011447906, + "step": 750, + "valid_targets_mean": 5127.9, + "valid_targets_min": 2405 + }, + { + "epoch": 0.41620727673649394, + "grad_norm": 0.5539669808893363, + "learning_rate": 2.3748031496062994e-05, + "loss": 0.3744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15149244666099548, + "step": 755, + "valid_targets_mean": 3253.8, + "valid_targets_min": 1654 + }, + { + "epoch": 0.4189636163175303, + "grad_norm": 0.5495100016493063, + "learning_rate": 2.3905511811023625e-05, + "loss": 0.3696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1736399084329605, + "step": 760, + "valid_targets_mean": 5750.2, + "valid_targets_min": 4644 + }, + { + "epoch": 0.4217199558985667, + "grad_norm": 0.6277772887593315, + "learning_rate": 2.4062992125984253e-05, + "loss": 0.3878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1839992105960846, + "step": 765, + "valid_targets_mean": 4141.6, + "valid_targets_min": 1544 + }, + { + "epoch": 0.4244762954796031, + "grad_norm": 0.572629060566468, + "learning_rate": 2.4220472440944884e-05, + "loss": 0.3683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17725741863250732, + "step": 770, + "valid_targets_mean": 5869.8, + "valid_targets_min": 953 + }, + { + "epoch": 0.42723263506063947, + "grad_norm": 0.5629248777629676, + "learning_rate": 2.4377952755905516e-05, + "loss": 0.3693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19953767955303192, + "step": 775, + "valid_targets_mean": 4973.6, + "valid_targets_min": 1265 + }, + { + "epoch": 0.42998897464167585, + "grad_norm": 0.5473291917635997, + "learning_rate": 2.4535433070866143e-05, + "loss": 0.3652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1739855259656906, + "step": 780, + "valid_targets_mean": 5359.9, + "valid_targets_min": 1982 + }, + { + "epoch": 0.43274531422271223, + "grad_norm": 0.5865344626852554, + "learning_rate": 2.4692913385826775e-05, + "loss": 0.3709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16023752093315125, + "step": 785, + "valid_targets_mean": 3892.5, + "valid_targets_min": 1759 + }, + { + "epoch": 0.4355016538037486, + "grad_norm": 0.7603763970297652, + "learning_rate": 2.4850393700787402e-05, + "loss": 0.3784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1775597184896469, + "step": 790, + "valid_targets_mean": 3185.8, + "valid_targets_min": 1451 + }, + { + "epoch": 0.438257993384785, + "grad_norm": 0.6405024107052186, + "learning_rate": 2.5007874015748034e-05, + "loss": 0.3738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18925821781158447, + "step": 795, + "valid_targets_mean": 4283.2, + "valid_targets_min": 1861 + }, + { + "epoch": 0.4410143329658214, + "grad_norm": 0.6390168764891554, + "learning_rate": 2.516535433070866e-05, + "loss": 0.3695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1907697319984436, + "step": 800, + "valid_targets_mean": 4340.8, + "valid_targets_min": 2345 + }, + { + "epoch": 0.44377067254685776, + "grad_norm": 0.6433952814149125, + "learning_rate": 2.5322834645669293e-05, + "loss": 0.3807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1938508301973343, + "step": 805, + "valid_targets_mean": 3923.9, + "valid_targets_min": 2767 + }, + { + "epoch": 0.44652701212789414, + "grad_norm": 0.7334218348595662, + "learning_rate": 2.548031496062992e-05, + "loss": 0.3869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2073996216058731, + "step": 810, + "valid_targets_mean": 3651.8, + "valid_targets_min": 1205 + }, + { + "epoch": 0.4492833517089305, + "grad_norm": 0.5948927293790542, + "learning_rate": 2.563779527559055e-05, + "loss": 0.3728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1528070718050003, + "step": 815, + "valid_targets_mean": 3835.4, + "valid_targets_min": 2004 + }, + { + "epoch": 0.4520396912899669, + "grad_norm": 0.6476987998976456, + "learning_rate": 2.5795275590551186e-05, + "loss": 0.377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18259508907794952, + "step": 820, + "valid_targets_mean": 3778.0, + "valid_targets_min": 1020 + }, + { + "epoch": 0.4547960308710033, + "grad_norm": 0.6373375236939519, + "learning_rate": 2.5952755905511814e-05, + "loss": 0.3562, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20902374386787415, + "step": 825, + "valid_targets_mean": 4307.8, + "valid_targets_min": 2308 + }, + { + "epoch": 0.45755237045203967, + "grad_norm": 0.663173001184542, + "learning_rate": 2.6110236220472445e-05, + "loss": 0.3516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19900460541248322, + "step": 830, + "valid_targets_mean": 4553.9, + "valid_targets_min": 2046 + }, + { + "epoch": 0.46030871003307605, + "grad_norm": 0.6911890514646282, + "learning_rate": 2.6267716535433073e-05, + "loss": 0.3552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24456271529197693, + "step": 835, + "valid_targets_mean": 4660.2, + "valid_targets_min": 3373 + }, + { + "epoch": 0.46306504961411243, + "grad_norm": 0.6872620279082434, + "learning_rate": 2.6425196850393704e-05, + "loss": 0.3677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17736120522022247, + "step": 840, + "valid_targets_mean": 3727.4, + "valid_targets_min": 1747 + }, + { + "epoch": 0.4658213891951488, + "grad_norm": 0.6519247337851097, + "learning_rate": 2.6582677165354332e-05, + "loss": 0.3682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16886401176452637, + "step": 845, + "valid_targets_mean": 3326.1, + "valid_targets_min": 1570 + }, + { + "epoch": 0.4685777287761852, + "grad_norm": 0.595128132824858, + "learning_rate": 2.6740157480314963e-05, + "loss": 0.3498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1713506281375885, + "step": 850, + "valid_targets_mean": 4600.8, + "valid_targets_min": 1537 + }, + { + "epoch": 0.47133406835722164, + "grad_norm": 0.5399559924976055, + "learning_rate": 2.689763779527559e-05, + "loss": 0.3509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14568282663822174, + "step": 855, + "valid_targets_mean": 4328.9, + "valid_targets_min": 1934 + }, + { + "epoch": 0.474090407938258, + "grad_norm": 0.6369406881664786, + "learning_rate": 2.7055118110236222e-05, + "loss": 0.3602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.153453066945076, + "step": 860, + "valid_targets_mean": 3494.0, + "valid_targets_min": 2562 + }, + { + "epoch": 0.4768467475192944, + "grad_norm": 0.6238940918728475, + "learning_rate": 2.7212598425196854e-05, + "loss": 0.3582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16792476177215576, + "step": 865, + "valid_targets_mean": 4076.8, + "valid_targets_min": 2659 + }, + { + "epoch": 0.4796030871003308, + "grad_norm": 0.6299130709272105, + "learning_rate": 2.737007874015748e-05, + "loss": 0.3477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1690496802330017, + "step": 870, + "valid_targets_mean": 4040.6, + "valid_targets_min": 1349 + }, + { + "epoch": 0.48235942668136716, + "grad_norm": 0.6960127350254371, + "learning_rate": 2.7527559055118113e-05, + "loss": 0.3526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13297244906425476, + "step": 875, + "valid_targets_mean": 3000.6, + "valid_targets_min": 1567 + }, + { + "epoch": 0.48511576626240355, + "grad_norm": 0.7663162226415476, + "learning_rate": 2.768503937007874e-05, + "loss": 0.3629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14324843883514404, + "step": 880, + "valid_targets_mean": 2628.4, + "valid_targets_min": 1500 + }, + { + "epoch": 0.48787210584343993, + "grad_norm": 0.6427398749729412, + "learning_rate": 2.7842519685039372e-05, + "loss": 0.3422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15800535678863525, + "step": 885, + "valid_targets_mean": 3413.6, + "valid_targets_min": 2033 + }, + { + "epoch": 0.4906284454244763, + "grad_norm": 0.7846331928093135, + "learning_rate": 2.8e-05, + "loss": 0.3679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.206691175699234, + "step": 890, + "valid_targets_mean": 2852.0, + "valid_targets_min": 1346 + }, + { + "epoch": 0.4933847850055127, + "grad_norm": 0.865182537406447, + "learning_rate": 2.815748031496063e-05, + "loss": 0.3587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19867077469825745, + "step": 895, + "valid_targets_mean": 2981.6, + "valid_targets_min": 1821 + }, + { + "epoch": 0.4961411245865491, + "grad_norm": 0.6765655442401763, + "learning_rate": 2.8314960629921265e-05, + "loss": 0.3529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1874970644712448, + "step": 900, + "valid_targets_mean": 3951.2, + "valid_targets_min": 2535 + }, + { + "epoch": 0.49889746416758546, + "grad_norm": 0.7708927026384667, + "learning_rate": 2.8472440944881893e-05, + "loss": 0.3472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16506461799144745, + "step": 905, + "valid_targets_mean": 3048.8, + "valid_targets_min": 2080 + }, + { + "epoch": 0.5016538037486218, + "grad_norm": 0.7214683086697998, + "learning_rate": 2.8629921259842524e-05, + "loss": 0.3478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15880101919174194, + "step": 910, + "valid_targets_mean": 2975.6, + "valid_targets_min": 1153 + }, + { + "epoch": 0.5044101433296582, + "grad_norm": 0.6618775371754808, + "learning_rate": 2.8787401574803152e-05, + "loss": 0.3446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15499462187290192, + "step": 915, + "valid_targets_mean": 3207.6, + "valid_targets_min": 1858 + }, + { + "epoch": 0.5071664829106945, + "grad_norm": 0.6426410857061894, + "learning_rate": 2.8944881889763783e-05, + "loss": 0.3556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1667099893093109, + "step": 920, + "valid_targets_mean": 3643.6, + "valid_targets_min": 2197 + }, + { + "epoch": 0.5099228224917309, + "grad_norm": 0.6231613963817665, + "learning_rate": 2.910236220472441e-05, + "loss": 0.3532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14109091460704803, + "step": 925, + "valid_targets_mean": 3769.4, + "valid_targets_min": 1394 + }, + { + "epoch": 0.5126791620727673, + "grad_norm": 0.6106380999313492, + "learning_rate": 2.9259842519685043e-05, + "loss": 0.3485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16766898334026337, + "step": 930, + "valid_targets_mean": 4348.6, + "valid_targets_min": 1844 + }, + { + "epoch": 0.5154355016538037, + "grad_norm": 0.9835472171473024, + "learning_rate": 2.941732283464567e-05, + "loss": 0.3461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19661033153533936, + "step": 935, + "valid_targets_mean": 4341.1, + "valid_targets_min": 1734 + }, + { + "epoch": 0.5181918412348401, + "grad_norm": 0.6115524881867599, + "learning_rate": 2.95748031496063e-05, + "loss": 0.3418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1919013112783432, + "step": 940, + "valid_targets_mean": 4252.6, + "valid_targets_min": 2077 + }, + { + "epoch": 0.5209481808158766, + "grad_norm": 0.6430914580283145, + "learning_rate": 2.9732283464566933e-05, + "loss": 0.3466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16576319932937622, + "step": 945, + "valid_targets_mean": 4177.9, + "valid_targets_min": 2349 + }, + { + "epoch": 0.523704520396913, + "grad_norm": 0.6041369377800954, + "learning_rate": 2.988976377952756e-05, + "loss": 0.3395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18523333966732025, + "step": 950, + "valid_targets_mean": 4160.2, + "valid_targets_min": 1607 + }, + { + "epoch": 0.5264608599779493, + "grad_norm": 0.72287805683052, + "learning_rate": 3.0047244094488192e-05, + "loss": 0.3576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18509402871131897, + "step": 955, + "valid_targets_mean": 3273.0, + "valid_targets_min": 1377 + }, + { + "epoch": 0.5292171995589857, + "grad_norm": 0.6204788679508508, + "learning_rate": 3.020472440944882e-05, + "loss": 0.3571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15396207571029663, + "step": 960, + "valid_targets_mean": 3304.5, + "valid_targets_min": 1902 + }, + { + "epoch": 0.5319735391400221, + "grad_norm": 0.6086738253334372, + "learning_rate": 3.036220472440945e-05, + "loss": 0.3445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15169401466846466, + "step": 965, + "valid_targets_mean": 3324.1, + "valid_targets_min": 654 + }, + { + "epoch": 0.5347298787210585, + "grad_norm": 0.6189638123557771, + "learning_rate": 3.051968503937008e-05, + "loss": 0.3457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16640031337738037, + "step": 970, + "valid_targets_mean": 3907.8, + "valid_targets_min": 2607 + }, + { + "epoch": 0.5374862183020949, + "grad_norm": 0.5947590629426671, + "learning_rate": 3.067716535433071e-05, + "loss": 0.3488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18467682600021362, + "step": 975, + "valid_targets_mean": 4467.2, + "valid_targets_min": 1739 + }, + { + "epoch": 0.5402425578831312, + "grad_norm": 0.6634199576526302, + "learning_rate": 3.0834645669291345e-05, + "loss": 0.3451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15196575224399567, + "step": 980, + "valid_targets_mean": 3224.1, + "valid_targets_min": 1140 + }, + { + "epoch": 0.5429988974641676, + "grad_norm": 0.6422159271595398, + "learning_rate": 3.099212598425197e-05, + "loss": 0.3242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1691552996635437, + "step": 985, + "valid_targets_mean": 3543.4, + "valid_targets_min": 2501 + }, + { + "epoch": 0.545755237045204, + "grad_norm": 0.6245847496071618, + "learning_rate": 3.11496062992126e-05, + "loss": 0.3383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17896845936775208, + "step": 990, + "valid_targets_mean": 4364.1, + "valid_targets_min": 1298 + }, + { + "epoch": 0.5485115766262404, + "grad_norm": 0.5977488876845884, + "learning_rate": 3.1307086614173235e-05, + "loss": 0.3191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17076364159584045, + "step": 995, + "valid_targets_mean": 4377.9, + "valid_targets_min": 1101 + }, + { + "epoch": 0.5512679162072768, + "grad_norm": 0.6592354485368468, + "learning_rate": 3.146456692913386e-05, + "loss": 0.361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19205403327941895, + "step": 1000, + "valid_targets_mean": 3894.2, + "valid_targets_min": 2079 + }, + { + "epoch": 0.5540242557883132, + "grad_norm": 0.6015288095411865, + "learning_rate": 3.162204724409449e-05, + "loss": 0.3341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13418832421302795, + "step": 1005, + "valid_targets_mean": 3054.4, + "valid_targets_min": 1743 + }, + { + "epoch": 0.5567805953693495, + "grad_norm": 0.642527116303665, + "learning_rate": 3.177952755905512e-05, + "loss": 0.3443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18287159502506256, + "step": 1010, + "valid_targets_mean": 3688.0, + "valid_targets_min": 1234 + }, + { + "epoch": 0.5595369349503859, + "grad_norm": 0.6755807279045383, + "learning_rate": 3.193700787401575e-05, + "loss": 0.3474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18014730513095856, + "step": 1015, + "valid_targets_mean": 3790.0, + "valid_targets_min": 1822 + }, + { + "epoch": 0.5622932745314223, + "grad_norm": 0.5936624451702214, + "learning_rate": 3.209448818897638e-05, + "loss": 0.3507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17635107040405273, + "step": 1020, + "valid_targets_mean": 4106.9, + "valid_targets_min": 1651 + }, + { + "epoch": 0.5650496141124587, + "grad_norm": 0.6604897255140217, + "learning_rate": 3.225196850393701e-05, + "loss": 0.3426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1873820424079895, + "step": 1025, + "valid_targets_mean": 3287.2, + "valid_targets_min": 1374 + }, + { + "epoch": 0.5678059536934951, + "grad_norm": 0.6223617007124282, + "learning_rate": 3.240944881889764e-05, + "loss": 0.3268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17855298519134521, + "step": 1030, + "valid_targets_mean": 4399.2, + "valid_targets_min": 2725 + }, + { + "epoch": 0.5705622932745315, + "grad_norm": 0.5942152415830797, + "learning_rate": 3.256692913385827e-05, + "loss": 0.3319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17899192869663239, + "step": 1035, + "valid_targets_mean": 4269.0, + "valid_targets_min": 1966 + }, + { + "epoch": 0.5733186328555678, + "grad_norm": 0.6516891639299425, + "learning_rate": 3.27244094488189e-05, + "loss": 0.3366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18318672478199005, + "step": 1040, + "valid_targets_mean": 3716.2, + "valid_targets_min": 1561 + }, + { + "epoch": 0.5760749724366042, + "grad_norm": 0.6281568277018019, + "learning_rate": 3.2881889763779527e-05, + "loss": 0.3292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16948923468589783, + "step": 1045, + "valid_targets_mean": 3750.1, + "valid_targets_min": 1444 + }, + { + "epoch": 0.5788313120176406, + "grad_norm": 0.6244269828491872, + "learning_rate": 3.303937007874016e-05, + "loss": 0.3165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1449054330587387, + "step": 1050, + "valid_targets_mean": 3042.1, + "valid_targets_min": 1504 + }, + { + "epoch": 0.581587651598677, + "grad_norm": 0.669093972403944, + "learning_rate": 3.319685039370079e-05, + "loss": 0.3248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1615549921989441, + "step": 1055, + "valid_targets_mean": 3153.0, + "valid_targets_min": 858 + }, + { + "epoch": 0.5843439911797134, + "grad_norm": 0.7155883655076967, + "learning_rate": 3.3354330708661424e-05, + "loss": 0.3371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18785884976387024, + "step": 1060, + "valid_targets_mean": 4063.5, + "valid_targets_min": 1288 + }, + { + "epoch": 0.5871003307607497, + "grad_norm": 0.6350403359252221, + "learning_rate": 3.351181102362205e-05, + "loss": 0.3592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16096770763397217, + "step": 1065, + "valid_targets_mean": 3914.9, + "valid_targets_min": 1537 + }, + { + "epoch": 0.5898566703417861, + "grad_norm": 0.7924388201918854, + "learning_rate": 3.366929133858268e-05, + "loss": 0.331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18221747875213623, + "step": 1070, + "valid_targets_mean": 4236.5, + "valid_targets_min": 2972 + }, + { + "epoch": 0.5926130099228225, + "grad_norm": 0.6326468377027029, + "learning_rate": 3.3826771653543314e-05, + "loss": 0.346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18635350465774536, + "step": 1075, + "valid_targets_mean": 4411.0, + "valid_targets_min": 2013 + }, + { + "epoch": 0.5953693495038589, + "grad_norm": 0.6635522188313777, + "learning_rate": 3.398425196850394e-05, + "loss": 0.3243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17104382812976837, + "step": 1080, + "valid_targets_mean": 3548.4, + "valid_targets_min": 1744 + }, + { + "epoch": 0.5981256890848953, + "grad_norm": 0.58747304322224, + "learning_rate": 3.414173228346457e-05, + "loss": 0.3364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14577405154705048, + "step": 1085, + "valid_targets_mean": 4129.6, + "valid_targets_min": 1914 + }, + { + "epoch": 0.6008820286659317, + "grad_norm": 0.6233242217619444, + "learning_rate": 3.42992125984252e-05, + "loss": 0.3287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14406809210777283, + "step": 1090, + "valid_targets_mean": 3826.4, + "valid_targets_min": 2414 + }, + { + "epoch": 0.603638368246968, + "grad_norm": 0.696604652466041, + "learning_rate": 3.445669291338583e-05, + "loss": 0.3147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18018591403961182, + "step": 1095, + "valid_targets_mean": 4195.0, + "valid_targets_min": 1937 + }, + { + "epoch": 0.6063947078280044, + "grad_norm": 0.6755391622885568, + "learning_rate": 3.461417322834646e-05, + "loss": 0.3448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19086028635501862, + "step": 1100, + "valid_targets_mean": 3154.9, + "valid_targets_min": 1649 + }, + { + "epoch": 0.6091510474090408, + "grad_norm": 0.6939728375429693, + "learning_rate": 3.477165354330709e-05, + "loss": 0.3266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1444491147994995, + "step": 1105, + "valid_targets_mean": 3088.5, + "valid_targets_min": 1463 + }, + { + "epoch": 0.6119073869900772, + "grad_norm": 0.662206877577478, + "learning_rate": 3.4929133858267715e-05, + "loss": 0.3428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16403332352638245, + "step": 1110, + "valid_targets_mean": 3241.4, + "valid_targets_min": 1426 + }, + { + "epoch": 0.6146637265711136, + "grad_norm": 0.6905564284152957, + "learning_rate": 3.508661417322835e-05, + "loss": 0.3425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.144118532538414, + "step": 1115, + "valid_targets_mean": 2768.9, + "valid_targets_min": 538 + }, + { + "epoch": 0.61742006615215, + "grad_norm": 0.5990270812749439, + "learning_rate": 3.524409448818898e-05, + "loss": 0.3358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17306233942508698, + "step": 1120, + "valid_targets_mean": 4483.5, + "valid_targets_min": 1646 + }, + { + "epoch": 0.6201764057331863, + "grad_norm": 0.6797719345603883, + "learning_rate": 3.5401574803149606e-05, + "loss": 0.3677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1944427490234375, + "step": 1125, + "valid_targets_mean": 3761.9, + "valid_targets_min": 2169 + }, + { + "epoch": 0.6229327453142227, + "grad_norm": 0.843170489479794, + "learning_rate": 3.555905511811024e-05, + "loss": 0.334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1481870710849762, + "step": 1130, + "valid_targets_mean": 3168.4, + "valid_targets_min": 1625 + }, + { + "epoch": 0.6256890848952591, + "grad_norm": 0.6053795002191212, + "learning_rate": 3.571653543307087e-05, + "loss": 0.3429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13735730946063995, + "step": 1135, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1997 + }, + { + "epoch": 0.6284454244762955, + "grad_norm": 0.6528165349849502, + "learning_rate": 3.58740157480315e-05, + "loss": 0.3466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20795869827270508, + "step": 1140, + "valid_targets_mean": 4406.4, + "valid_targets_min": 2949 + }, + { + "epoch": 0.6312017640573319, + "grad_norm": 0.5446809452387671, + "learning_rate": 3.603149606299213e-05, + "loss": 0.3428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18867036700248718, + "step": 1145, + "valid_targets_mean": 5358.5, + "valid_targets_min": 2652 + }, + { + "epoch": 0.6339581036383682, + "grad_norm": 0.6195595425028796, + "learning_rate": 3.618897637795276e-05, + "loss": 0.3434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19956643879413605, + "step": 1150, + "valid_targets_mean": 4314.5, + "valid_targets_min": 1381 + }, + { + "epoch": 0.6367144432194046, + "grad_norm": 0.6214315392118165, + "learning_rate": 3.634645669291339e-05, + "loss": 0.3412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13539068400859833, + "step": 1155, + "valid_targets_mean": 3395.9, + "valid_targets_min": 637 + }, + { + "epoch": 0.639470782800441, + "grad_norm": 0.6380895127441484, + "learning_rate": 3.650393700787402e-05, + "loss": 0.3418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14906813204288483, + "step": 1160, + "valid_targets_mean": 3256.4, + "valid_targets_min": 719 + }, + { + "epoch": 0.6422271223814774, + "grad_norm": 0.5293262106509432, + "learning_rate": 3.666141732283465e-05, + "loss": 0.3385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14705684781074524, + "step": 1165, + "valid_targets_mean": 4725.1, + "valid_targets_min": 3210 + }, + { + "epoch": 0.6449834619625138, + "grad_norm": 0.7065284110432264, + "learning_rate": 3.6818897637795276e-05, + "loss": 0.3307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17961552739143372, + "step": 1170, + "valid_targets_mean": 3775.4, + "valid_targets_min": 1372 + }, + { + "epoch": 0.6477398015435502, + "grad_norm": 0.6267224028848232, + "learning_rate": 3.697637795275591e-05, + "loss": 0.3168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1890057921409607, + "step": 1175, + "valid_targets_mean": 4266.1, + "valid_targets_min": 2768 + }, + { + "epoch": 0.6504961411245865, + "grad_norm": 0.6023798014472563, + "learning_rate": 3.713385826771654e-05, + "loss": 0.332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12609617412090302, + "step": 1180, + "valid_targets_mean": 2540.5, + "valid_targets_min": 530 + }, + { + "epoch": 0.6532524807056229, + "grad_norm": 0.6028739147617584, + "learning_rate": 3.729133858267717e-05, + "loss": 0.3251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16367727518081665, + "step": 1185, + "valid_targets_mean": 3903.2, + "valid_targets_min": 1841 + }, + { + "epoch": 0.6560088202866593, + "grad_norm": 0.6055504683204002, + "learning_rate": 3.7448818897637795e-05, + "loss": 0.3417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1499730944633484, + "step": 1190, + "valid_targets_mean": 3476.8, + "valid_targets_min": 1728 + }, + { + "epoch": 0.6587651598676957, + "grad_norm": 0.7012826477164547, + "learning_rate": 3.760629921259843e-05, + "loss": 0.3356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19046294689178467, + "step": 1195, + "valid_targets_mean": 3712.2, + "valid_targets_min": 1362 + }, + { + "epoch": 0.6615214994487321, + "grad_norm": 0.5764995810896218, + "learning_rate": 3.776377952755906e-05, + "loss": 0.3302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17078228294849396, + "step": 1200, + "valid_targets_mean": 3693.8, + "valid_targets_min": 852 + }, + { + "epoch": 0.6642778390297684, + "grad_norm": 0.7044889851436874, + "learning_rate": 3.7921259842519685e-05, + "loss": 0.3364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16281527280807495, + "step": 1205, + "valid_targets_mean": 2822.6, + "valid_targets_min": 1195 + }, + { + "epoch": 0.6670341786108048, + "grad_norm": 0.7571872846291527, + "learning_rate": 3.807874015748032e-05, + "loss": 0.3301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1942022740840912, + "step": 1210, + "valid_targets_mean": 4747.8, + "valid_targets_min": 1397 + }, + { + "epoch": 0.6697905181918412, + "grad_norm": 0.5786887713889898, + "learning_rate": 3.823622047244095e-05, + "loss": 0.32, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12428903579711914, + "step": 1215, + "valid_targets_mean": 3487.6, + "valid_targets_min": 1975 + }, + { + "epoch": 0.6725468577728776, + "grad_norm": 0.6561078693898419, + "learning_rate": 3.839370078740158e-05, + "loss": 0.3129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17524409294128418, + "step": 1220, + "valid_targets_mean": 4387.6, + "valid_targets_min": 2564 + }, + { + "epoch": 0.675303197353914, + "grad_norm": 0.49783479932338043, + "learning_rate": 3.855118110236221e-05, + "loss": 0.3183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1519559919834137, + "step": 1225, + "valid_targets_mean": 6404.1, + "valid_targets_min": 2595 + }, + { + "epoch": 0.6780595369349504, + "grad_norm": 0.5724182004106165, + "learning_rate": 3.870866141732284e-05, + "loss": 0.3273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17201152443885803, + "step": 1230, + "valid_targets_mean": 4013.2, + "valid_targets_min": 2868 + }, + { + "epoch": 0.6808158765159867, + "grad_norm": 0.6666547370621647, + "learning_rate": 3.886614173228347e-05, + "loss": 0.3149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1399753987789154, + "step": 1235, + "valid_targets_mean": 2886.0, + "valid_targets_min": 993 + }, + { + "epoch": 0.6835722160970231, + "grad_norm": 0.6441855795833656, + "learning_rate": 3.90236220472441e-05, + "loss": 0.3181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15389788150787354, + "step": 1240, + "valid_targets_mean": 3805.1, + "valid_targets_min": 1575 + }, + { + "epoch": 0.6863285556780595, + "grad_norm": 0.6931968048487174, + "learning_rate": 3.918110236220473e-05, + "loss": 0.3231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1897139996290207, + "step": 1245, + "valid_targets_mean": 3722.0, + "valid_targets_min": 1360 + }, + { + "epoch": 0.6890848952590959, + "grad_norm": 0.5919301600890774, + "learning_rate": 3.9338582677165356e-05, + "loss": 0.3166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15014256536960602, + "step": 1250, + "valid_targets_mean": 3669.4, + "valid_targets_min": 2414 + }, + { + "epoch": 0.6918412348401323, + "grad_norm": 0.6029508714742269, + "learning_rate": 3.949606299212599e-05, + "loss": 0.3435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18831568956375122, + "step": 1255, + "valid_targets_mean": 4327.4, + "valid_targets_min": 1454 + }, + { + "epoch": 0.6945975744211687, + "grad_norm": 0.6476480355297924, + "learning_rate": 3.965354330708662e-05, + "loss": 0.3326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17568446695804596, + "step": 1260, + "valid_targets_mean": 3619.9, + "valid_targets_min": 1720 + }, + { + "epoch": 0.697353914002205, + "grad_norm": 0.7136853184681312, + "learning_rate": 3.9811023622047246e-05, + "loss": 0.3317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17106367647647858, + "step": 1265, + "valid_targets_mean": 3105.5, + "valid_targets_min": 1278 + }, + { + "epoch": 0.7001102535832414, + "grad_norm": 0.704188566824314, + "learning_rate": 3.9968503937007874e-05, + "loss": 0.314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17615364491939545, + "step": 1270, + "valid_targets_mean": 3147.8, + "valid_targets_min": 1858 + }, + { + "epoch": 0.7028665931642778, + "grad_norm": 0.6333636556365725, + "learning_rate": 3.9999987908526714e-05, + "loss": 0.326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12883499264717102, + "step": 1275, + "valid_targets_mean": 2984.5, + "valid_targets_min": 628 + }, + { + "epoch": 0.7056229327453142, + "grad_norm": 0.6794824245772948, + "learning_rate": 3.999993878694153e-05, + "loss": 0.323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19361761212348938, + "step": 1280, + "valid_targets_mean": 2602.6, + "valid_targets_min": 1039 + }, + { + "epoch": 0.7083792723263506, + "grad_norm": 0.5834104462006615, + "learning_rate": 3.99998518796201e-05, + "loss": 0.3322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15409792959690094, + "step": 1285, + "valid_targets_mean": 4317.8, + "valid_targets_min": 2530 + }, + { + "epoch": 0.7111356119073869, + "grad_norm": 0.6117204104155071, + "learning_rate": 3.9999727186726634e-05, + "loss": 0.3269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18922315537929535, + "step": 1290, + "valid_targets_mean": 3711.0, + "valid_targets_min": 809 + }, + { + "epoch": 0.7138919514884233, + "grad_norm": 0.6521034177228493, + "learning_rate": 3.9999564708496694e-05, + "loss": 0.3241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14246782660484314, + "step": 1295, + "valid_targets_mean": 3450.5, + "valid_targets_min": 1780 + }, + { + "epoch": 0.7166482910694597, + "grad_norm": 0.6552026801728348, + "learning_rate": 3.9999364445237255e-05, + "loss": 0.3323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18096542358398438, + "step": 1300, + "valid_targets_mean": 4356.6, + "valid_targets_min": 2185 + }, + { + "epoch": 0.7194046306504961, + "grad_norm": 0.8182069341774965, + "learning_rate": 3.999912639732668e-05, + "loss": 0.2916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08992710709571838, + "step": 1305, + "valid_targets_mean": 2424.2, + "valid_targets_min": 876 + }, + { + "epoch": 0.7221609702315325, + "grad_norm": 0.6677152822496627, + "learning_rate": 3.99988505652147e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07074375450611115, + "step": 1310, + "valid_targets_mean": 1550.6, + "valid_targets_min": 669 + }, + { + "epoch": 0.7249173098125689, + "grad_norm": 0.5704167467627762, + "learning_rate": 3.999853694942246e-05, + "loss": 0.1538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0822879821062088, + "step": 1315, + "valid_targets_mean": 3182.0, + "valid_targets_min": 2264 + }, + { + "epoch": 0.7276736493936052, + "grad_norm": 0.5183536320369988, + "learning_rate": 3.999818555054245e-05, + "loss": 0.1479, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06996433436870575, + "step": 1320, + "valid_targets_mean": 3035.2, + "valid_targets_min": 710 + }, + { + "epoch": 0.7304299889746417, + "grad_norm": 0.4843667416262553, + "learning_rate": 3.999779636923858e-05, + "loss": 0.1623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047072216868400574, + "step": 1325, + "valid_targets_mean": 3213.8, + "valid_targets_min": 1223 + }, + { + "epoch": 0.7331863285556781, + "grad_norm": 0.5052461811148854, + "learning_rate": 3.9997369406246115e-05, + "loss": 0.1467, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08826921880245209, + "step": 1330, + "valid_targets_mean": 3550.9, + "valid_targets_min": 2558 + }, + { + "epoch": 0.7359426681367145, + "grad_norm": 0.9265320705693709, + "learning_rate": 3.9996904662371736e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09092816710472107, + "step": 1335, + "valid_targets_mean": 966.1, + "valid_targets_min": 559 + }, + { + "epoch": 0.7386990077177509, + "grad_norm": 0.4519939987985544, + "learning_rate": 3.999640213849345e-05, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06382960826158524, + "step": 1340, + "valid_targets_mean": 3475.8, + "valid_targets_min": 1013 + }, + { + "epoch": 0.7414553472987873, + "grad_norm": 0.52868953122772, + "learning_rate": 3.99958618355607e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08017554134130478, + "step": 1345, + "valid_targets_mean": 3533.5, + "valid_targets_min": 3049 + }, + { + "epoch": 0.7442116868798236, + "grad_norm": 0.5909736612715547, + "learning_rate": 3.9995283754594256e-05, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1172465905547142, + "step": 1350, + "valid_targets_mean": 3382.9, + "valid_targets_min": 1949 + }, + { + "epoch": 0.74696802646086, + "grad_norm": 0.4072460274622218, + "learning_rate": 3.999466789668629e-05, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0683635026216507, + "step": 1355, + "valid_targets_mean": 3985.5, + "valid_targets_min": 929 + }, + { + "epoch": 0.7497243660418964, + "grad_norm": 0.5889068096734931, + "learning_rate": 3.999401426300034e-05, + "loss": 0.1416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06982085853815079, + "step": 1360, + "valid_targets_mean": 2270.4, + "valid_targets_min": 447 + }, + { + "epoch": 0.7524807056229328, + "grad_norm": 0.438553402495516, + "learning_rate": 3.99933228547713e-05, + "loss": 0.1379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06671202182769775, + "step": 1365, + "valid_targets_mean": 3307.2, + "valid_targets_min": 754 + }, + { + "epoch": 0.7552370452039692, + "grad_norm": 0.6290688318288891, + "learning_rate": 3.999259367330546e-05, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08816584944725037, + "step": 1370, + "valid_targets_mean": 1430.6, + "valid_targets_min": 604 + }, + { + "epoch": 0.7579933847850056, + "grad_norm": 0.5239820857737189, + "learning_rate": 3.999182671998044e-05, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07441706955432892, + "step": 1375, + "valid_targets_mean": 3397.1, + "valid_targets_min": 1862 + }, + { + "epoch": 0.7607497243660419, + "grad_norm": 0.5009124919331982, + "learning_rate": 3.9991021996245256e-05, + "loss": 0.1425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07202465832233429, + "step": 1380, + "valid_targets_mean": 2834.5, + "valid_targets_min": 574 + }, + { + "epoch": 0.7635060639470783, + "grad_norm": 0.552659237491683, + "learning_rate": 3.9990179503620255e-05, + "loss": 0.1362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06499902904033661, + "step": 1385, + "valid_targets_mean": 2592.2, + "valid_targets_min": 523 + }, + { + "epoch": 0.7662624035281147, + "grad_norm": 0.5776388290188639, + "learning_rate": 3.9989299243697156e-05, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09200268983840942, + "step": 1390, + "valid_targets_mean": 3489.2, + "valid_targets_min": 2189 + }, + { + "epoch": 0.7690187431091511, + "grad_norm": 0.5144762222574663, + "learning_rate": 3.9988381218139024e-05, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059957727789878845, + "step": 1395, + "valid_targets_mean": 3051.6, + "valid_targets_min": 812 + }, + { + "epoch": 0.7717750826901875, + "grad_norm": 0.4876371871087572, + "learning_rate": 3.998742542868028e-05, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062310487031936646, + "step": 1400, + "valid_targets_mean": 1836.4, + "valid_targets_min": 597 + }, + { + "epoch": 0.7745314222712238, + "grad_norm": 0.565494120341878, + "learning_rate": 3.998643187712669e-05, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07612445950508118, + "step": 1405, + "valid_targets_mean": 2217.6, + "valid_targets_min": 718 + }, + { + "epoch": 0.7772877618522602, + "grad_norm": 0.5894448172627752, + "learning_rate": 3.9985400565355365e-05, + "loss": 0.1761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1525067687034607, + "step": 1410, + "valid_targets_mean": 3171.4, + "valid_targets_min": 1297 + }, + { + "epoch": 0.7800441014332966, + "grad_norm": 0.5103366929265124, + "learning_rate": 3.998433149531475e-05, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06291086971759796, + "step": 1415, + "valid_targets_mean": 2718.0, + "valid_targets_min": 786 + }, + { + "epoch": 0.782800441014333, + "grad_norm": 0.781863093774017, + "learning_rate": 3.998322466902464e-05, + "loss": 0.1377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10495224595069885, + "step": 1420, + "valid_targets_mean": 1962.9, + "valid_targets_min": 954 + }, + { + "epoch": 0.7855567805953694, + "grad_norm": 0.5213711487325935, + "learning_rate": 3.998208008857614e-05, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07504158467054367, + "step": 1425, + "valid_targets_mean": 3074.2, + "valid_targets_min": 1304 + }, + { + "epoch": 0.7883131201764058, + "grad_norm": 0.4654518308823379, + "learning_rate": 3.9980897756131716e-05, + "loss": 0.165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053179800510406494, + "step": 1430, + "valid_targets_mean": 3154.2, + "valid_targets_min": 794 + }, + { + "epoch": 0.7910694597574421, + "grad_norm": 0.6993416597477006, + "learning_rate": 3.9979677673925125e-05, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08588965237140656, + "step": 1435, + "valid_targets_mean": 2348.6, + "valid_targets_min": 717 + }, + { + "epoch": 0.7938257993384785, + "grad_norm": 0.6345487150731937, + "learning_rate": 3.997841984426147e-05, + "loss": 0.3635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1532469242811203, + "step": 1440, + "valid_targets_mean": 2661.4, + "valid_targets_min": 987 + }, + { + "epoch": 0.7965821389195149, + "grad_norm": 0.7621866664835062, + "learning_rate": 3.997712426951715e-05, + "loss": 0.1664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12519598007202148, + "step": 1445, + "valid_targets_mean": 1800.9, + "valid_targets_min": 757 + }, + { + "epoch": 0.7993384785005513, + "grad_norm": 0.6208006354383474, + "learning_rate": 3.997579095213988e-05, + "loss": 0.1452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05883906036615372, + "step": 1450, + "valid_targets_mean": 1464.6, + "valid_targets_min": 673 + }, + { + "epoch": 0.8020948180815877, + "grad_norm": 0.5471218480132252, + "learning_rate": 3.997441989464871e-05, + "loss": 0.2536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08253096044063568, + "step": 1455, + "valid_targets_mean": 2554.9, + "valid_targets_min": 666 + }, + { + "epoch": 0.804851157662624, + "grad_norm": 0.3919104785303273, + "learning_rate": 3.997301109963393e-05, + "loss": 0.1245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039561767131090164, + "step": 1460, + "valid_targets_mean": 5195.2, + "valid_targets_min": 3529 + }, + { + "epoch": 0.8076074972436604, + "grad_norm": 0.4893023564038904, + "learning_rate": 3.997156456975721e-05, + "loss": 0.1416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08282340317964554, + "step": 1465, + "valid_targets_mean": 4099.0, + "valid_targets_min": 2336 + }, + { + "epoch": 0.8103638368246968, + "grad_norm": 0.4665120623278485, + "learning_rate": 3.997008030775144e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07420054823160172, + "step": 1470, + "valid_targets_mean": 3430.0, + "valid_targets_min": 1226 + }, + { + "epoch": 0.8131201764057332, + "grad_norm": 0.5207551516497763, + "learning_rate": 3.9968558316420834e-05, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07871238887310028, + "step": 1475, + "valid_targets_mean": 4274.5, + "valid_targets_min": 3275 + }, + { + "epoch": 0.8158765159867696, + "grad_norm": 0.4557712439833038, + "learning_rate": 3.996699859864088e-05, + "loss": 0.1302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062801793217659, + "step": 1480, + "valid_targets_mean": 2150.2, + "valid_targets_min": 723 + }, + { + "epoch": 0.818632855567806, + "grad_norm": 0.45470663322194593, + "learning_rate": 3.996540115735833e-05, + "loss": 0.1286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07046838104724884, + "step": 1485, + "valid_targets_mean": 3443.2, + "valid_targets_min": 2234 + }, + { + "epoch": 0.8213891951488423, + "grad_norm": 0.3678775320150355, + "learning_rate": 3.996376599559124e-05, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044194869697093964, + "step": 1490, + "valid_targets_mean": 3956.0, + "valid_targets_min": 3226 + }, + { + "epoch": 0.8241455347298787, + "grad_norm": 0.49516447784551965, + "learning_rate": 3.996209311642889e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06538870930671692, + "step": 1495, + "valid_targets_mean": 2450.4, + "valid_targets_min": 739 + }, + { + "epoch": 0.8269018743109151, + "grad_norm": 0.5542176136427859, + "learning_rate": 3.996038252303184e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06214669346809387, + "step": 1500, + "valid_targets_mean": 1556.5, + "valid_targets_min": 606 + }, + { + "epoch": 0.8296582138919515, + "grad_norm": 0.4388587913593567, + "learning_rate": 3.995863421863191e-05, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04456140846014023, + "step": 1505, + "valid_targets_mean": 1918.8, + "valid_targets_min": 585 + }, + { + "epoch": 0.8324145534729879, + "grad_norm": 0.5065222899814672, + "learning_rate": 3.995684820653216e-05, + "loss": 0.1321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07136686891317368, + "step": 1510, + "valid_targets_mean": 2679.6, + "valid_targets_min": 582 + }, + { + "epoch": 0.8351708930540243, + "grad_norm": 0.4479748762945977, + "learning_rate": 3.9955024490106875e-05, + "loss": 0.1171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07549671083688736, + "step": 1515, + "valid_targets_mean": 3722.2, + "valid_targets_min": 3260 + }, + { + "epoch": 0.8379272326350606, + "grad_norm": 0.4494772871941269, + "learning_rate": 3.9953163072801595e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08701890707015991, + "step": 1520, + "valid_targets_mean": 3233.8, + "valid_targets_min": 999 + }, + { + "epoch": 0.840683572216097, + "grad_norm": 0.8146319136515292, + "learning_rate": 3.995126395813309e-05, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10312506556510925, + "step": 1525, + "valid_targets_mean": 2054.0, + "valid_targets_min": 696 + }, + { + "epoch": 0.8434399117971334, + "grad_norm": 0.47196486034201035, + "learning_rate": 3.994932714968934e-05, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07035676389932632, + "step": 1530, + "valid_targets_mean": 2716.5, + "valid_targets_min": 635 + }, + { + "epoch": 0.8461962513781698, + "grad_norm": 0.481777881411327, + "learning_rate": 3.994735265112954e-05, + "loss": 0.1415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05764827877283096, + "step": 1535, + "valid_targets_mean": 1854.2, + "valid_targets_min": 816 + }, + { + "epoch": 0.8489525909592062, + "grad_norm": 0.5415841793910088, + "learning_rate": 3.9945340466184086e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20636659860610962, + "step": 1540, + "valid_targets_mean": 3260.4, + "valid_targets_min": 1183 + }, + { + "epoch": 0.8517089305402425, + "grad_norm": 0.4781043653653037, + "learning_rate": 3.994329059865461e-05, + "loss": 0.1407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06641293317079544, + "step": 1545, + "valid_targets_mean": 3414.1, + "valid_targets_min": 2342 + }, + { + "epoch": 0.8544652701212789, + "grad_norm": 0.7312764329443081, + "learning_rate": 3.994120305241389e-05, + "loss": 0.1704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08996740728616714, + "step": 1550, + "valid_targets_mean": 1638.2, + "valid_targets_min": 687 + }, + { + "epoch": 0.8572216097023153, + "grad_norm": 0.5593807273017817, + "learning_rate": 3.993907783140591e-05, + "loss": 0.1341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04166106507182121, + "step": 1555, + "valid_targets_mean": 1557.5, + "valid_targets_min": 449 + }, + { + "epoch": 0.8599779492833517, + "grad_norm": 0.45493133241465117, + "learning_rate": 3.993691493964585e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0439620316028595, + "step": 1560, + "valid_targets_mean": 2998.2, + "valid_targets_min": 658 + }, + { + "epoch": 0.8627342888643881, + "grad_norm": 0.7481745005677496, + "learning_rate": 3.993471438122003e-05, + "loss": 0.1393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06205380707979202, + "step": 1565, + "valid_targets_mean": 1299.1, + "valid_targets_min": 577 + }, + { + "epoch": 0.8654906284454245, + "grad_norm": 0.43953952047774836, + "learning_rate": 3.993247616028596e-05, + "loss": 0.1179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0671330913901329, + "step": 1570, + "valid_targets_mean": 4539.9, + "valid_targets_min": 2883 + }, + { + "epoch": 0.8682469680264608, + "grad_norm": 0.4480983074437537, + "learning_rate": 3.993020028107229e-05, + "loss": 0.1141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06733305752277374, + "step": 1575, + "valid_targets_mean": 3626.6, + "valid_targets_min": 2252 + }, + { + "epoch": 0.8710033076074972, + "grad_norm": 0.5049401846610271, + "learning_rate": 3.9927886747878826e-05, + "loss": 0.1267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056880634278059006, + "step": 1580, + "valid_targets_mean": 2025.5, + "valid_targets_min": 654 + }, + { + "epoch": 0.8737596471885336, + "grad_norm": 0.4236882525710864, + "learning_rate": 3.99255355650765e-05, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07104510068893433, + "step": 1585, + "valid_targets_mean": 3139.9, + "valid_targets_min": 744 + }, + { + "epoch": 0.87651598676957, + "grad_norm": 0.4624213069104655, + "learning_rate": 3.992314673710739e-05, + "loss": 0.1414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08039423078298569, + "step": 1590, + "valid_targets_mean": 3662.5, + "valid_targets_min": 3014 + }, + { + "epoch": 0.8792723263506064, + "grad_norm": 0.3550421590757934, + "learning_rate": 3.992072026848469e-05, + "loss": 0.1296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04613850265741348, + "step": 1595, + "valid_targets_mean": 3219.4, + "valid_targets_min": 2366 + }, + { + "epoch": 0.8820286659316428, + "grad_norm": 0.38855674926497336, + "learning_rate": 3.991825616379271e-05, + "loss": 0.1209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055348217487335205, + "step": 1600, + "valid_targets_mean": 1961.0, + "valid_targets_min": 669 + }, + { + "epoch": 0.8847850055126791, + "grad_norm": 0.46622979861671004, + "learning_rate": 3.9915754427686867e-05, + "loss": 0.1487, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0801142007112503, + "step": 1605, + "valid_targets_mean": 3525.8, + "valid_targets_min": 1785 + }, + { + "epoch": 0.8875413450937155, + "grad_norm": 0.29601007767650345, + "learning_rate": 3.9913215064893664e-05, + "loss": 0.1105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05134030058979988, + "step": 1610, + "valid_targets_mean": 4065.1, + "valid_targets_min": 3684 + }, + { + "epoch": 0.8902976846747519, + "grad_norm": 0.41874359198648947, + "learning_rate": 3.9910638080210714e-05, + "loss": 0.1368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05827372521162033, + "step": 1615, + "valid_targets_mean": 3246.6, + "valid_targets_min": 595 + }, + { + "epoch": 0.8930540242557883, + "grad_norm": 0.45707836129215, + "learning_rate": 3.990802347850668e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05286899581551552, + "step": 1620, + "valid_targets_mean": 2653.8, + "valid_targets_min": 1016 + }, + { + "epoch": 0.8958103638368247, + "grad_norm": 0.6749770194382505, + "learning_rate": 3.990537126472132e-05, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12368709594011307, + "step": 1625, + "valid_targets_mean": 1386.5, + "valid_targets_min": 908 + }, + { + "epoch": 0.898566703417861, + "grad_norm": 0.6618388520214199, + "learning_rate": 3.990268144386543e-05, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07151079922914505, + "step": 1630, + "valid_targets_mean": 2152.5, + "valid_targets_min": 697 + }, + { + "epoch": 0.9013230429988974, + "grad_norm": 0.731926250098089, + "learning_rate": 3.989995402102089e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0897190049290657, + "step": 1635, + "valid_targets_mean": 1274.8, + "valid_targets_min": 736 + }, + { + "epoch": 0.9040793825799338, + "grad_norm": 0.42029321747267034, + "learning_rate": 3.989718900134058e-05, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04420304298400879, + "step": 1640, + "valid_targets_mean": 2531.1, + "valid_targets_min": 804 + }, + { + "epoch": 0.9068357221609702, + "grad_norm": 0.4398143478665166, + "learning_rate": 3.9894386390048444e-05, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04994278773665428, + "step": 1645, + "valid_targets_mean": 2974.6, + "valid_targets_min": 1018 + }, + { + "epoch": 0.9095920617420066, + "grad_norm": 0.40817499207910296, + "learning_rate": 3.989154619243943e-05, + "loss": 0.1298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06163801997900009, + "step": 1650, + "valid_targets_mean": 3626.9, + "valid_targets_min": 1632 + }, + { + "epoch": 0.912348401323043, + "grad_norm": 0.44289623553294494, + "learning_rate": 3.988866841387949e-05, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06624019891023636, + "step": 1655, + "valid_targets_mean": 3119.6, + "valid_targets_min": 1006 + }, + { + "epoch": 0.9151047409040793, + "grad_norm": 0.4563353530517001, + "learning_rate": 3.988575305980561e-05, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06047752872109413, + "step": 1660, + "valid_targets_mean": 2694.8, + "valid_targets_min": 903 + }, + { + "epoch": 0.9178610804851157, + "grad_norm": 0.34947644564310054, + "learning_rate": 3.988280013572573e-05, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055833179503679276, + "step": 1665, + "valid_targets_mean": 3786.0, + "valid_targets_min": 3052 + }, + { + "epoch": 0.9206174200661521, + "grad_norm": 0.5915521656103409, + "learning_rate": 3.9879809647218805e-05, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08503246307373047, + "step": 1670, + "valid_targets_mean": 2972.0, + "valid_targets_min": 564 + }, + { + "epoch": 0.9233737596471885, + "grad_norm": 0.3848570304298655, + "learning_rate": 3.9876781599934726e-05, + "loss": 0.1262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07070305198431015, + "step": 1675, + "valid_targets_mean": 3655.0, + "valid_targets_min": 2555 + }, + { + "epoch": 0.9261300992282249, + "grad_norm": 0.33521190634348325, + "learning_rate": 3.9873715999594374e-05, + "loss": 0.1132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053336288779973984, + "step": 1680, + "valid_targets_mean": 4186.8, + "valid_targets_min": 2991 + }, + { + "epoch": 0.9288864388092613, + "grad_norm": 0.41825036110056846, + "learning_rate": 3.987061285198955e-05, + "loss": 0.105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048326361924409866, + "step": 1685, + "valid_targets_mean": 2357.9, + "valid_targets_min": 618 + }, + { + "epoch": 0.9316427783902976, + "grad_norm": 0.34988882813443484, + "learning_rate": 3.9867472162983024e-05, + "loss": 0.1342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04564378410577774, + "step": 1690, + "valid_targets_mean": 3658.1, + "valid_targets_min": 2944 + }, + { + "epoch": 0.934399117971334, + "grad_norm": 0.387640604931235, + "learning_rate": 3.9864293938508465e-05, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09076780825853348, + "step": 1695, + "valid_targets_mean": 3401.8, + "valid_targets_min": 1539 + }, + { + "epoch": 0.9371554575523704, + "grad_norm": 1.5884413174614103, + "learning_rate": 3.986107818457048e-05, + "loss": 0.1462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03863140195608139, + "step": 1700, + "valid_targets_mean": 2312.4, + "valid_targets_min": 509 + }, + { + "epoch": 0.9399117971334069, + "grad_norm": 0.3529589532594106, + "learning_rate": 3.985782490724455e-05, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041794754564762115, + "step": 1705, + "valid_targets_mean": 2753.0, + "valid_targets_min": 637 + }, + { + "epoch": 0.9426681367144433, + "grad_norm": 0.43792365193451477, + "learning_rate": 3.985453411267709e-05, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06898006051778793, + "step": 1710, + "valid_targets_mean": 2755.8, + "valid_targets_min": 737 + }, + { + "epoch": 0.9454244762954797, + "grad_norm": 0.3520292102149443, + "learning_rate": 3.985120580708536e-05, + "loss": 0.1139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05722911283373833, + "step": 1715, + "valid_targets_mean": 3081.4, + "valid_targets_min": 492 + }, + { + "epoch": 0.948180815876516, + "grad_norm": 0.6242959575952906, + "learning_rate": 3.98478399967575e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062430839985609055, + "step": 1720, + "valid_targets_mean": 1433.2, + "valid_targets_min": 697 + }, + { + "epoch": 0.9509371554575524, + "grad_norm": 0.3818993417683654, + "learning_rate": 3.9844436688052526e-05, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0461457297205925, + "step": 1725, + "valid_targets_mean": 3534.8, + "valid_targets_min": 923 + }, + { + "epoch": 0.9536934950385888, + "grad_norm": 0.6598181909329016, + "learning_rate": 3.9840995887400274e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07108645141124725, + "step": 1730, + "valid_targets_mean": 1486.1, + "valid_targets_min": 531 + }, + { + "epoch": 0.9564498346196252, + "grad_norm": 0.5062636880753861, + "learning_rate": 3.9837517601301424e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09940898418426514, + "step": 1735, + "valid_targets_mean": 2096.5, + "valid_targets_min": 575 + }, + { + "epoch": 0.9592061742006616, + "grad_norm": 0.46377247746614, + "learning_rate": 3.983400183632747e-05, + "loss": 0.1267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06860677897930145, + "step": 1740, + "valid_targets_mean": 3725.2, + "valid_targets_min": 2784 + }, + { + "epoch": 0.961962513781698, + "grad_norm": 0.3895112966455787, + "learning_rate": 3.983044859912073e-05, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05860050767660141, + "step": 1745, + "valid_targets_mean": 3427.1, + "valid_targets_min": 2525 + }, + { + "epoch": 0.9647188533627343, + "grad_norm": 0.3817682726461495, + "learning_rate": 3.9826857896394306e-05, + "loss": 0.1235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04753850772976875, + "step": 1750, + "valid_targets_mean": 2502.5, + "valid_targets_min": 970 + }, + { + "epoch": 0.9674751929437707, + "grad_norm": 0.48145123565962944, + "learning_rate": 3.982322973493209e-05, + "loss": 0.1179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05467890948057175, + "step": 1755, + "valid_targets_mean": 1527.6, + "valid_targets_min": 729 + }, + { + "epoch": 0.9702315325248071, + "grad_norm": 0.792707226192526, + "learning_rate": 3.9819564121588736e-05, + "loss": 0.1535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1010986790060997, + "step": 1760, + "valid_targets_mean": 1267.9, + "valid_targets_min": 637 + }, + { + "epoch": 0.9729878721058435, + "grad_norm": 0.3726797892626685, + "learning_rate": 3.981586106328966e-05, + "loss": 0.1125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049895208328962326, + "step": 1765, + "valid_targets_mean": 3062.4, + "valid_targets_min": 893 + }, + { + "epoch": 0.9757442116868799, + "grad_norm": 0.34321892530236126, + "learning_rate": 3.981212056703102e-05, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058021459728479385, + "step": 1770, + "valid_targets_mean": 3553.1, + "valid_targets_min": 2404 + }, + { + "epoch": 0.9785005512679162, + "grad_norm": 0.4499270239487714, + "learning_rate": 3.980834263987971e-05, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06501922011375427, + "step": 1775, + "valid_targets_mean": 3326.0, + "valid_targets_min": 986 + }, + { + "epoch": 0.9812568908489526, + "grad_norm": 0.346264297175817, + "learning_rate": 3.980452728897334e-05, + "loss": 0.1317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058167703449726105, + "step": 1780, + "valid_targets_mean": 3475.9, + "valid_targets_min": 1078 + }, + { + "epoch": 0.984013230429989, + "grad_norm": 0.3001994407113807, + "learning_rate": 3.9800674521520234e-05, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06304462999105453, + "step": 1785, + "valid_targets_mean": 4152.8, + "valid_targets_min": 761 + }, + { + "epoch": 0.9867695700110254, + "grad_norm": 0.3930576010919473, + "learning_rate": 3.9796784344799385e-05, + "loss": 0.1198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04717961698770523, + "step": 1790, + "valid_targets_mean": 1554.6, + "valid_targets_min": 507 + }, + { + "epoch": 0.9895259095920618, + "grad_norm": 0.5743464283192015, + "learning_rate": 3.979285676616048e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11262590438127518, + "step": 1795, + "valid_targets_mean": 2711.8, + "valid_targets_min": 780 + }, + { + "epoch": 0.9922822491730982, + "grad_norm": 0.3976393030027254, + "learning_rate": 3.978889179302385e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05783668905496597, + "step": 1800, + "valid_targets_mean": 3892.5, + "valid_targets_min": 2347 + }, + { + "epoch": 0.9950385887541345, + "grad_norm": 0.45515755096165894, + "learning_rate": 3.9784889432880513e-05, + "loss": 0.12, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07580891251564026, + "step": 1805, + "valid_targets_mean": 3665.6, + "valid_targets_min": 2473 + }, + { + "epoch": 0.9977949283351709, + "grad_norm": 0.4441646018021414, + "learning_rate": 3.9780849693292094e-05, + "loss": 0.119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05661609768867493, + "step": 1810, + "valid_targets_mean": 3330.6, + "valid_targets_min": 2245 + }, + { + "epoch": 1.0005512679162072, + "grad_norm": 0.5561326888583642, + "learning_rate": 3.977677258189082e-05, + "loss": 0.1939, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13960739970207214, + "step": 1815, + "valid_targets_mean": 8215.2, + "valid_targets_min": 6163 + }, + { + "epoch": 1.0033076074972436, + "grad_norm": 0.47556528540006193, + "learning_rate": 3.977265810637957e-05, + "loss": 0.2814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14123216271400452, + "step": 1820, + "valid_targets_mean": 8594.4, + "valid_targets_min": 6006 + }, + { + "epoch": 1.00606394707828, + "grad_norm": 0.5126205487434939, + "learning_rate": 3.9768506274531795e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12284629046916962, + "step": 1825, + "valid_targets_mean": 3578.4, + "valid_targets_min": 289 + }, + { + "epoch": 1.0088202866593163, + "grad_norm": 0.40807821166362745, + "learning_rate": 3.97643170941915e-05, + "loss": 0.2633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.129932701587677, + "step": 1830, + "valid_targets_mean": 7019.9, + "valid_targets_min": 4985 + }, + { + "epoch": 1.0115766262403527, + "grad_norm": 0.4042372834174987, + "learning_rate": 3.9760090573273286e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12571382522583008, + "step": 1835, + "valid_targets_mean": 6956.6, + "valid_targets_min": 1457 + }, + { + "epoch": 1.014332965821389, + "grad_norm": 0.4242609290754954, + "learning_rate": 3.975582671976229e-05, + "loss": 0.2485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11796011030673981, + "step": 1840, + "valid_targets_mean": 7361.6, + "valid_targets_min": 4961 + }, + { + "epoch": 1.0170893054024255, + "grad_norm": 0.4676152502834955, + "learning_rate": 3.9751525541714164e-05, + "loss": 0.264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15607969462871552, + "step": 1845, + "valid_targets_mean": 7058.5, + "valid_targets_min": 5985 + }, + { + "epoch": 1.0198456449834619, + "grad_norm": 0.3858442142205225, + "learning_rate": 3.974718704725511e-05, + "loss": 0.2608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14063405990600586, + "step": 1850, + "valid_targets_mean": 7664.2, + "valid_targets_min": 5169 + }, + { + "epoch": 1.0226019845644982, + "grad_norm": 0.4009522731456042, + "learning_rate": 3.9742811244581806e-05, + "loss": 0.2488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12624630331993103, + "step": 1855, + "valid_targets_mean": 7111.8, + "valid_targets_min": 5348 + }, + { + "epoch": 1.0253583241455346, + "grad_norm": 0.4603926255309547, + "learning_rate": 3.973839814196142e-05, + "loss": 0.246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11568456888198853, + "step": 1860, + "valid_targets_mean": 4238.9, + "valid_targets_min": 152 + }, + { + "epoch": 1.028114663726571, + "grad_norm": 0.4049015048357173, + "learning_rate": 3.973394774773159e-05, + "loss": 0.2403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12659597396850586, + "step": 1865, + "valid_targets_mean": 7199.4, + "valid_targets_min": 4258 + }, + { + "epoch": 1.0308710033076074, + "grad_norm": 0.39502425723954765, + "learning_rate": 3.972946007030043e-05, + "loss": 0.2427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11860829591751099, + "step": 1870, + "valid_targets_mean": 7243.4, + "valid_targets_min": 5831 + }, + { + "epoch": 1.0336273428886438, + "grad_norm": 0.44584847587473353, + "learning_rate": 3.972493511814646e-05, + "loss": 0.2389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11814169585704803, + "step": 1875, + "valid_targets_mean": 6950.6, + "valid_targets_min": 3862 + }, + { + "epoch": 1.0363836824696802, + "grad_norm": 0.47691507666080035, + "learning_rate": 3.972037289981865e-05, + "loss": 0.2365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12782923877239227, + "step": 1880, + "valid_targets_mean": 6991.9, + "valid_targets_min": 5696 + }, + { + "epoch": 1.0391400220507165, + "grad_norm": 0.5203186097662312, + "learning_rate": 3.971577342393635e-05, + "loss": 0.2291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12695907056331635, + "step": 1885, + "valid_targets_mean": 6963.5, + "valid_targets_min": 4947 + }, + { + "epoch": 1.041896361631753, + "grad_norm": 0.39158298342343284, + "learning_rate": 3.971113669918933e-05, + "loss": 0.2336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11154258251190186, + "step": 1890, + "valid_targets_mean": 6436.9, + "valid_targets_min": 4988 + }, + { + "epoch": 1.0446527012127893, + "grad_norm": 0.3824826827911866, + "learning_rate": 3.9706462734337714e-05, + "loss": 0.2366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11754725873470306, + "step": 1895, + "valid_targets_mean": 7518.0, + "valid_targets_min": 5476 + }, + { + "epoch": 1.0474090407938257, + "grad_norm": 0.42015992127558255, + "learning_rate": 3.970175153821199e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10808613896369934, + "step": 1900, + "valid_targets_mean": 7422.6, + "valid_targets_min": 5226 + }, + { + "epoch": 1.0501653803748623, + "grad_norm": 0.4374534599632615, + "learning_rate": 3.9697003119712994e-05, + "loss": 0.2483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15219682455062866, + "step": 1905, + "valid_targets_mean": 7080.0, + "valid_targets_min": 3753 + }, + { + "epoch": 1.0529217199558987, + "grad_norm": 0.46899331415248224, + "learning_rate": 3.969221748781186e-05, + "loss": 0.2594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13338930904865265, + "step": 1910, + "valid_targets_mean": 6064.6, + "valid_targets_min": 5161 + }, + { + "epoch": 1.055678059536935, + "grad_norm": 0.39431566438760596, + "learning_rate": 3.9687394651550055e-05, + "loss": 0.2183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11277997493743896, + "step": 1915, + "valid_targets_mean": 7388.1, + "valid_targets_min": 5795 + }, + { + "epoch": 1.0584343991179714, + "grad_norm": 0.38335122037673797, + "learning_rate": 3.968253462003934e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10592363774776459, + "step": 1920, + "valid_targets_mean": 7696.0, + "valid_targets_min": 4658 + }, + { + "epoch": 1.0611907386990078, + "grad_norm": 0.3781366967372623, + "learning_rate": 3.967763740246172e-05, + "loss": 0.2288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12528692185878754, + "step": 1925, + "valid_targets_mean": 7799.1, + "valid_targets_min": 5508 + }, + { + "epoch": 1.0639470782800442, + "grad_norm": 0.4177115130047357, + "learning_rate": 3.967270300806948e-05, + "loss": 0.2402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1167149543762207, + "step": 1930, + "valid_targets_mean": 7341.6, + "valid_targets_min": 6454 + }, + { + "epoch": 1.0667034178610806, + "grad_norm": 0.41685566982815453, + "learning_rate": 3.9667731446185126e-05, + "loss": 0.2312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10362047702074051, + "step": 1935, + "valid_targets_mean": 6818.0, + "valid_targets_min": 4392 + }, + { + "epoch": 1.069459757442117, + "grad_norm": 0.3624464234368589, + "learning_rate": 3.966272272620141e-05, + "loss": 0.2303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10205714404582977, + "step": 1940, + "valid_targets_mean": 7857.6, + "valid_targets_min": 5394 + }, + { + "epoch": 1.0722160970231533, + "grad_norm": 0.34056845913561445, + "learning_rate": 3.965767685758126e-05, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0930442214012146, + "step": 1945, + "valid_targets_mean": 8314.1, + "valid_targets_min": 4837 + }, + { + "epoch": 1.0749724366041897, + "grad_norm": 0.331551506511239, + "learning_rate": 3.965259384985779e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08719047904014587, + "step": 1950, + "valid_targets_mean": 7368.2, + "valid_targets_min": 6027 + }, + { + "epoch": 1.0777287761852261, + "grad_norm": 0.4096735937731108, + "learning_rate": 3.96474737126343e-05, + "loss": 0.2131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11517889052629471, + "step": 1955, + "valid_targets_mean": 6435.0, + "valid_targets_min": 4238 + }, + { + "epoch": 1.0804851157662625, + "grad_norm": 1.722020733576018, + "learning_rate": 3.964231645558422e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0998082309961319, + "step": 1960, + "valid_targets_mean": 247.8, + "valid_targets_min": 137 + }, + { + "epoch": 1.0832414553472989, + "grad_norm": 0.40551946704882497, + "learning_rate": 3.9637122088451125e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1104302704334259, + "step": 1965, + "valid_targets_mean": 6807.0, + "valid_targets_min": 5165 + }, + { + "epoch": 1.0859977949283353, + "grad_norm": 0.45227138949164003, + "learning_rate": 3.963189062104869e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1279592216014862, + "step": 1970, + "valid_targets_mean": 6989.9, + "valid_targets_min": 5104 + }, + { + "epoch": 1.0887541345093716, + "grad_norm": 0.4751453307642135, + "learning_rate": 3.962662206326069e-05, + "loss": 0.2429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13115835189819336, + "step": 1975, + "valid_targets_mean": 6692.2, + "valid_targets_min": 5068 + }, + { + "epoch": 1.091510474090408, + "grad_norm": 0.4219389616204448, + "learning_rate": 3.962131642504097e-05, + "loss": 0.2361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.141310453414917, + "step": 1980, + "valid_targets_mean": 7419.6, + "valid_targets_min": 5578 + }, + { + "epoch": 1.0942668136714444, + "grad_norm": 0.3747910228846253, + "learning_rate": 3.9615973716413434e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11464381963014603, + "step": 1985, + "valid_targets_mean": 7536.5, + "valid_targets_min": 5075 + }, + { + "epoch": 1.0970231532524808, + "grad_norm": 0.3603053622709781, + "learning_rate": 3.961059394747202e-05, + "loss": 0.234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10643325746059418, + "step": 1990, + "valid_targets_mean": 7221.5, + "valid_targets_min": 5841 + }, + { + "epoch": 1.0997794928335172, + "grad_norm": 0.3834811644818215, + "learning_rate": 3.9605177128380686e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11278769373893738, + "step": 1995, + "valid_targets_mean": 7287.5, + "valid_targets_min": 5151 + }, + { + "epoch": 1.1025358324145536, + "grad_norm": 0.41623725005716433, + "learning_rate": 3.9599723269373393e-05, + "loss": 0.2386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11052314192056656, + "step": 2000, + "valid_targets_mean": 6738.4, + "valid_targets_min": 4042 + }, + { + "epoch": 1.10529217199559, + "grad_norm": 0.41645894355680674, + "learning_rate": 3.959423238075407e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12841008603572845, + "step": 2005, + "valid_targets_mean": 6331.0, + "valid_targets_min": 5496 + }, + { + "epoch": 1.1080485115766263, + "grad_norm": 0.43116215968063776, + "learning_rate": 3.958870447289662e-05, + "loss": 0.2335, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11427759379148483, + "step": 2010, + "valid_targets_mean": 6271.1, + "valid_targets_min": 4695 + }, + { + "epoch": 1.1108048511576627, + "grad_norm": 0.43664146922103775, + "learning_rate": 3.958313955624487e-05, + "loss": 0.2341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12099075317382812, + "step": 2015, + "valid_targets_mean": 5790.8, + "valid_targets_min": 5021 + }, + { + "epoch": 1.113561190738699, + "grad_norm": 0.4805925511065944, + "learning_rate": 3.957753764131258e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09358817338943481, + "step": 2020, + "valid_targets_mean": 3386.8, + "valid_targets_min": 2279 + }, + { + "epoch": 1.1163175303197355, + "grad_norm": 0.42893593219683757, + "learning_rate": 3.9571898738683394e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11254025995731354, + "step": 2025, + "valid_targets_mean": 6240.5, + "valid_targets_min": 5093 + }, + { + "epoch": 1.1190738699007718, + "grad_norm": 0.4120783365955582, + "learning_rate": 3.9566222859010875e-05, + "loss": 0.2268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11772503703832626, + "step": 2030, + "valid_targets_mean": 6329.5, + "valid_targets_min": 4589 + }, + { + "epoch": 1.1218302094818082, + "grad_norm": 0.4009504223862474, + "learning_rate": 3.95605100130184e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12054328620433807, + "step": 2035, + "valid_targets_mean": 7589.2, + "valid_targets_min": 4963 + }, + { + "epoch": 1.1245865490628446, + "grad_norm": 0.4040257052842429, + "learning_rate": 3.9554760211499214e-05, + "loss": 0.2262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09745586663484573, + "step": 2040, + "valid_targets_mean": 5577.9, + "valid_targets_min": 4458 + }, + { + "epoch": 1.127342888643881, + "grad_norm": 0.4596076700394669, + "learning_rate": 3.9548973465316374e-05, + "loss": 0.2248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10829180479049683, + "step": 2045, + "valid_targets_mean": 5432.5, + "valid_targets_min": 4433 + }, + { + "epoch": 1.1300992282249174, + "grad_norm": 0.41997434069124856, + "learning_rate": 3.9543149785402733e-05, + "loss": 0.2288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11732716858386993, + "step": 2050, + "valid_targets_mean": 6361.8, + "valid_targets_min": 5201 + }, + { + "epoch": 1.1328555678059538, + "grad_norm": 0.4267034286072882, + "learning_rate": 3.9537289182760926e-05, + "loss": 0.23, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10569071769714355, + "step": 2055, + "valid_targets_mean": 5824.4, + "valid_targets_min": 5129 + }, + { + "epoch": 1.1356119073869901, + "grad_norm": 0.8951258064175898, + "learning_rate": 3.953139166846335e-05, + "loss": 0.204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16045713424682617, + "step": 2060, + "valid_targets_mean": 1637.6, + "valid_targets_min": 137 + }, + { + "epoch": 1.1383682469680265, + "grad_norm": 0.39308444356299144, + "learning_rate": 3.952545725365215e-05, + "loss": 0.2248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11317199468612671, + "step": 2065, + "valid_targets_mean": 6474.6, + "valid_targets_min": 4751 + }, + { + "epoch": 1.141124586549063, + "grad_norm": 0.4408228394426399, + "learning_rate": 3.951948594953914e-05, + "loss": 0.2213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12203557789325714, + "step": 2070, + "valid_targets_mean": 7186.2, + "valid_targets_min": 5358 + }, + { + "epoch": 1.1438809261300993, + "grad_norm": 0.428680305064189, + "learning_rate": 3.951347776740588e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11417984962463379, + "step": 2075, + "valid_targets_mean": 7126.8, + "valid_targets_min": 4078 + }, + { + "epoch": 1.1466372657111357, + "grad_norm": 0.405969792601804, + "learning_rate": 3.9507432718603584e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.110316202044487, + "step": 2080, + "valid_targets_mean": 7022.1, + "valid_targets_min": 4940 + }, + { + "epoch": 1.149393605292172, + "grad_norm": 0.475076319316369, + "learning_rate": 3.9501350814553114e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11389956623315811, + "step": 2085, + "valid_targets_mean": 5900.6, + "valid_targets_min": 4779 + }, + { + "epoch": 1.1521499448732084, + "grad_norm": 0.40297235857706454, + "learning_rate": 3.9495232066744965e-05, + "loss": 0.2243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11459016054868698, + "step": 2090, + "valid_targets_mean": 6290.1, + "valid_targets_min": 4620 + }, + { + "epoch": 1.1549062844542448, + "grad_norm": 0.9532130059765137, + "learning_rate": 3.9489076486739245e-05, + "loss": 0.2573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10592678934335709, + "step": 2095, + "valid_targets_mean": 1162.2, + "valid_targets_min": 788 + }, + { + "epoch": 1.1576626240352812, + "grad_norm": 0.9893030829997986, + "learning_rate": 3.948288408616565e-05, + "loss": 0.256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12086014449596405, + "step": 2100, + "valid_targets_mean": 1500.6, + "valid_targets_min": 957 + }, + { + "epoch": 1.1604189636163176, + "grad_norm": 0.8847880117862259, + "learning_rate": 3.947665487672342e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11283891648054123, + "step": 2105, + "valid_targets_mean": 1371.2, + "valid_targets_min": 796 + }, + { + "epoch": 1.163175303197354, + "grad_norm": 0.9160818632603779, + "learning_rate": 3.9470388870181376e-05, + "loss": 0.2402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13674435019493103, + "step": 2110, + "valid_targets_mean": 1433.6, + "valid_targets_min": 697 + }, + { + "epoch": 1.1659316427783903, + "grad_norm": 0.7340092114115278, + "learning_rate": 3.9464086078377824e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11220844089984894, + "step": 2115, + "valid_targets_mean": 1578.9, + "valid_targets_min": 638 + }, + { + "epoch": 1.1686879823594267, + "grad_norm": 0.7539561646720673, + "learning_rate": 3.945774651322058e-05, + "loss": 0.2345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11869017034769058, + "step": 2120, + "valid_targets_mean": 1746.4, + "valid_targets_min": 727 + }, + { + "epoch": 1.171444321940463, + "grad_norm": 0.8344953626642414, + "learning_rate": 3.945137018668695e-05, + "loss": 0.2419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10075467079877853, + "step": 2125, + "valid_targets_mean": 1309.9, + "valid_targets_min": 761 + }, + { + "epoch": 1.1742006615214995, + "grad_norm": 0.9081490968433219, + "learning_rate": 3.9444957110823665e-05, + "loss": 0.2244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13452501595020294, + "step": 2130, + "valid_targets_mean": 1368.6, + "valid_targets_min": 714 + }, + { + "epoch": 1.1769570011025359, + "grad_norm": 0.8179807829143761, + "learning_rate": 3.943850729774692e-05, + "loss": 0.2424, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11519020795822144, + "step": 2135, + "valid_targets_mean": 1491.8, + "valid_targets_min": 867 + }, + { + "epoch": 1.1797133406835723, + "grad_norm": 0.7543931825563687, + "learning_rate": 3.943202075964229e-05, + "loss": 0.2309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12807095050811768, + "step": 2140, + "valid_targets_mean": 1994.0, + "valid_targets_min": 790 + }, + { + "epoch": 1.1824696802646086, + "grad_norm": 0.8958311708359955, + "learning_rate": 3.942549750876473e-05, + "loss": 0.2244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11700532585382462, + "step": 2145, + "valid_targets_mean": 1523.4, + "valid_targets_min": 896 + }, + { + "epoch": 1.185226019845645, + "grad_norm": 0.8278333029861888, + "learning_rate": 3.941893755743859e-05, + "loss": 0.2371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11836803704500198, + "step": 2150, + "valid_targets_mean": 1790.1, + "valid_targets_min": 1009 + }, + { + "epoch": 1.1879823594266814, + "grad_norm": 0.8109108244768911, + "learning_rate": 3.9412340918057525e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09498180449008942, + "step": 2155, + "valid_targets_mean": 1394.0, + "valid_targets_min": 717 + }, + { + "epoch": 1.1907386990077178, + "grad_norm": 1.074395029772338, + "learning_rate": 3.9405707603084525e-05, + "loss": 0.216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08787362277507782, + "step": 2160, + "valid_targets_mean": 1129.0, + "valid_targets_min": 659 + }, + { + "epoch": 1.1934950385887542, + "grad_norm": 0.866303182516471, + "learning_rate": 3.939903762505186e-05, + "loss": 0.2144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10428543388843536, + "step": 2165, + "valid_targets_mean": 1466.9, + "valid_targets_min": 642 + }, + { + "epoch": 1.1962513781697905, + "grad_norm": 0.7948980401918122, + "learning_rate": 3.9392330996561064e-05, + "loss": 0.2239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07902731746435165, + "step": 2170, + "valid_targets_mean": 1129.4, + "valid_targets_min": 765 + }, + { + "epoch": 1.199007717750827, + "grad_norm": 0.9151358934528343, + "learning_rate": 3.9385587730282936e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10825944691896439, + "step": 2175, + "valid_targets_mean": 1322.8, + "valid_targets_min": 631 + }, + { + "epoch": 1.2017640573318633, + "grad_norm": 0.893811105308566, + "learning_rate": 3.937880783895746e-05, + "loss": 0.2342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.132063090801239, + "step": 2180, + "valid_targets_mean": 1612.0, + "valid_targets_min": 623 + }, + { + "epoch": 1.2045203969128997, + "grad_norm": 0.7098557805972313, + "learning_rate": 3.9371991335393846e-05, + "loss": 0.2265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08403678238391876, + "step": 2185, + "valid_targets_mean": 1402.1, + "valid_targets_min": 940 + }, + { + "epoch": 1.207276736493936, + "grad_norm": 0.7334855255666238, + "learning_rate": 3.936513823247047e-05, + "loss": 0.2264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10120376944541931, + "step": 2190, + "valid_targets_mean": 1337.8, + "valid_targets_min": 776 + }, + { + "epoch": 1.2100330760749725, + "grad_norm": 0.8579608500424668, + "learning_rate": 3.935824854313483e-05, + "loss": 0.2243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13063853979110718, + "step": 2195, + "valid_targets_mean": 1525.5, + "valid_targets_min": 623 + }, + { + "epoch": 1.2127894156560088, + "grad_norm": 0.8214412080653641, + "learning_rate": 3.935132228040357e-05, + "loss": 0.2293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12357903271913528, + "step": 2200, + "valid_targets_mean": 1441.0, + "valid_targets_min": 693 + }, + { + "epoch": 1.2155457552370452, + "grad_norm": 0.8722666214417062, + "learning_rate": 3.934435945736244e-05, + "loss": 0.2239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06323891133069992, + "step": 2205, + "valid_targets_mean": 847.9, + "valid_targets_min": 611 + }, + { + "epoch": 1.2183020948180816, + "grad_norm": 0.8855851518400091, + "learning_rate": 3.933736008716624e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11711321771144867, + "step": 2210, + "valid_targets_mean": 1524.9, + "valid_targets_min": 793 + }, + { + "epoch": 1.221058434399118, + "grad_norm": 0.7770444739356825, + "learning_rate": 3.933032418303883e-05, + "loss": 0.2174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1193980798125267, + "step": 2215, + "valid_targets_mean": 1779.4, + "valid_targets_min": 563 + }, + { + "epoch": 1.2238147739801544, + "grad_norm": 0.7285164643071013, + "learning_rate": 3.932325175827309e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12358640879392624, + "step": 2220, + "valid_targets_mean": 1741.2, + "valid_targets_min": 1341 + }, + { + "epoch": 1.2265711135611908, + "grad_norm": 1.0107139722271623, + "learning_rate": 3.9316142826230905e-05, + "loss": 0.2182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13167104125022888, + "step": 2225, + "valid_targets_mean": 1875.6, + "valid_targets_min": 1107 + }, + { + "epoch": 1.2293274531422271, + "grad_norm": 0.8611514478298554, + "learning_rate": 3.9308997400343114e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10524449497461319, + "step": 2230, + "valid_targets_mean": 1419.5, + "valid_targets_min": 680 + }, + { + "epoch": 1.2320837927232635, + "grad_norm": 0.8038220045272352, + "learning_rate": 3.930181549410954e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14974023401737213, + "step": 2235, + "valid_targets_mean": 1968.2, + "valid_targets_min": 981 + }, + { + "epoch": 1.2348401323043, + "grad_norm": 0.8379378411989463, + "learning_rate": 3.9294597121098884e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10743536055088043, + "step": 2240, + "valid_targets_mean": 1568.9, + "valid_targets_min": 937 + }, + { + "epoch": 1.2375964718853363, + "grad_norm": 0.8606678345699176, + "learning_rate": 3.928734229494878e-05, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11842221021652222, + "step": 2245, + "valid_targets_mean": 1645.2, + "valid_targets_min": 804 + }, + { + "epoch": 1.2403528114663727, + "grad_norm": 1.1558822833631255, + "learning_rate": 3.928005102936571e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10608691722154617, + "step": 2250, + "valid_targets_mean": 1574.9, + "valid_targets_min": 1019 + }, + { + "epoch": 1.243109151047409, + "grad_norm": 0.8046589011699751, + "learning_rate": 3.9272723338125e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09760580956935883, + "step": 2255, + "valid_targets_mean": 1121.6, + "valid_targets_min": 882 + }, + { + "epoch": 1.2458654906284454, + "grad_norm": 0.7910321704623297, + "learning_rate": 3.926535923507081e-05, + "loss": 0.2107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11715759336948395, + "step": 2260, + "valid_targets_mean": 1818.1, + "valid_targets_min": 1168 + }, + { + "epoch": 1.2486218302094818, + "grad_norm": 0.9103774971620899, + "learning_rate": 3.92579587341161e-05, + "loss": 0.2043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10207825899124146, + "step": 2265, + "valid_targets_mean": 1417.8, + "valid_targets_min": 678 + }, + { + "epoch": 1.2513781697905182, + "grad_norm": 0.8667718330981383, + "learning_rate": 3.9250521849242555e-05, + "loss": 0.2189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08377937227487564, + "step": 2270, + "valid_targets_mean": 1051.0, + "valid_targets_min": 676 + }, + { + "epoch": 1.2541345093715546, + "grad_norm": 0.7522140246006966, + "learning_rate": 3.9243048594500647e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09457919001579285, + "step": 2275, + "valid_targets_mean": 1431.1, + "valid_targets_min": 794 + }, + { + "epoch": 1.256890848952591, + "grad_norm": 0.7833375041285963, + "learning_rate": 3.923553898400953e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11398820579051971, + "step": 2280, + "valid_targets_mean": 1486.9, + "valid_targets_min": 742 + }, + { + "epoch": 1.2596471885336273, + "grad_norm": 0.8820065732448717, + "learning_rate": 3.922799303195707e-05, + "loss": 0.2178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12904658913612366, + "step": 2285, + "valid_targets_mean": 1734.5, + "valid_targets_min": 741 + }, + { + "epoch": 1.2624035281146637, + "grad_norm": 1.0458300923190358, + "learning_rate": 3.922041075259975e-05, + "loss": 0.2259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12874142825603485, + "step": 2290, + "valid_targets_mean": 1712.5, + "valid_targets_min": 1157 + }, + { + "epoch": 1.2651598676957, + "grad_norm": 0.8414849879854118, + "learning_rate": 3.921279216026275e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14273688197135925, + "step": 2295, + "valid_targets_mean": 1723.8, + "valid_targets_min": 862 + }, + { + "epoch": 1.2679162072767365, + "grad_norm": 0.7507553233453157, + "learning_rate": 3.920513726933979e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.135962575674057, + "step": 2300, + "valid_targets_mean": 1633.1, + "valid_targets_min": 1019 + }, + { + "epoch": 1.2706725468577729, + "grad_norm": 0.7424030639177608, + "learning_rate": 3.919744609429321e-05, + "loss": 0.2179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10355131328105927, + "step": 2305, + "valid_targets_mean": 1376.1, + "valid_targets_min": 1058 + }, + { + "epoch": 1.2734288864388092, + "grad_norm": 0.8209384632431105, + "learning_rate": 3.918971864965389e-05, + "loss": 0.2331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11861588805913925, + "step": 2310, + "valid_targets_mean": 1576.5, + "valid_targets_min": 618 + }, + { + "epoch": 1.2761852260198456, + "grad_norm": 0.7350161808636799, + "learning_rate": 3.9181954950021236e-05, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1254390925168991, + "step": 2315, + "valid_targets_mean": 1615.8, + "valid_targets_min": 629 + }, + { + "epoch": 1.278941565600882, + "grad_norm": 0.7128873767515329, + "learning_rate": 3.917415501006315e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11184348165988922, + "step": 2320, + "valid_targets_mean": 1881.2, + "valid_targets_min": 1167 + }, + { + "epoch": 1.2816979051819184, + "grad_norm": 0.8157233475854566, + "learning_rate": 3.9166318844516e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1143147200345993, + "step": 2325, + "valid_targets_mean": 1367.2, + "valid_targets_min": 699 + }, + { + "epoch": 1.2844542447629548, + "grad_norm": 0.7620970021969637, + "learning_rate": 3.915844646818459e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08498851209878922, + "step": 2330, + "valid_targets_mean": 1174.9, + "valid_targets_min": 679 + }, + { + "epoch": 1.2872105843439912, + "grad_norm": 0.7280256874654399, + "learning_rate": 3.915053789594214e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10084791481494904, + "step": 2335, + "valid_targets_mean": 1556.2, + "valid_targets_min": 504 + }, + { + "epoch": 1.2899669239250275, + "grad_norm": 0.7144276374574214, + "learning_rate": 3.914259314273027e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12192879617214203, + "step": 2340, + "valid_targets_mean": 1885.0, + "valid_targets_min": 1331 + }, + { + "epoch": 1.292723263506064, + "grad_norm": 0.7785853389602345, + "learning_rate": 3.913461222355893e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07888998091220856, + "step": 2345, + "valid_targets_mean": 1306.1, + "valid_targets_min": 661 + }, + { + "epoch": 1.2954796030871003, + "grad_norm": 0.7819127909121658, + "learning_rate": 3.9126595153506424e-05, + "loss": 0.2186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08130539953708649, + "step": 2350, + "valid_targets_mean": 1330.1, + "valid_targets_min": 958 + }, + { + "epoch": 1.2982359426681367, + "grad_norm": 0.6764710714361292, + "learning_rate": 3.9118541947719334e-05, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08272341638803482, + "step": 2355, + "valid_targets_mean": 1465.0, + "valid_targets_min": 963 + }, + { + "epoch": 1.300992282249173, + "grad_norm": 0.8046149845532986, + "learning_rate": 3.911045262141252e-05, + "loss": 0.2144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10547973215579987, + "step": 2360, + "valid_targets_mean": 1428.0, + "valid_targets_min": 793 + }, + { + "epoch": 1.3037486218302095, + "grad_norm": 0.8113427563047173, + "learning_rate": 3.910232718986909e-05, + "loss": 0.2139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1371140480041504, + "step": 2365, + "valid_targets_mean": 1633.1, + "valid_targets_min": 917 + }, + { + "epoch": 1.3065049614112458, + "grad_norm": 0.732388969191063, + "learning_rate": 3.909416566844036e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08403348177671432, + "step": 2370, + "valid_targets_mean": 1246.5, + "valid_targets_min": 872 + }, + { + "epoch": 1.3092613009922822, + "grad_norm": 0.6634723729363554, + "learning_rate": 3.908596807254585e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11004636436700821, + "step": 2375, + "valid_targets_mean": 1903.6, + "valid_targets_min": 1220 + }, + { + "epoch": 1.3120176405733186, + "grad_norm": 0.804901126036983, + "learning_rate": 3.9077734417673186e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10086837410926819, + "step": 2380, + "valid_targets_mean": 1391.6, + "valid_targets_min": 712 + }, + { + "epoch": 1.314773980154355, + "grad_norm": 0.7774269977938469, + "learning_rate": 3.906946471937817e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11322802305221558, + "step": 2385, + "valid_targets_mean": 1295.0, + "valid_targets_min": 605 + }, + { + "epoch": 1.3175303197353914, + "grad_norm": 0.7528996526963784, + "learning_rate": 3.906115899328468e-05, + "loss": 0.2131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07025298476219177, + "step": 2390, + "valid_targets_mean": 1414.8, + "valid_targets_min": 746 + }, + { + "epoch": 1.3202866593164277, + "grad_norm": 0.8021651870846336, + "learning_rate": 3.905281725508466e-05, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08973024785518646, + "step": 2395, + "valid_targets_mean": 1059.4, + "valid_targets_min": 723 + }, + { + "epoch": 1.3230429988974641, + "grad_norm": 0.7880920406727848, + "learning_rate": 3.90444395205381e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11170729994773865, + "step": 2400, + "valid_targets_mean": 1634.2, + "valid_targets_min": 1327 + }, + { + "epoch": 1.3257993384785005, + "grad_norm": 0.7339879499153292, + "learning_rate": 3.9036025805472976e-05, + "loss": 0.2085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09265495836734772, + "step": 2405, + "valid_targets_mean": 1412.5, + "valid_targets_min": 902 + }, + { + "epoch": 1.328555678059537, + "grad_norm": 0.6993805718803109, + "learning_rate": 3.902757612578528e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0981290340423584, + "step": 2410, + "valid_targets_mean": 1558.8, + "valid_targets_min": 803 + }, + { + "epoch": 1.3313120176405733, + "grad_norm": 0.7697696259811568, + "learning_rate": 3.9019090497438916e-05, + "loss": 0.2107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10622967779636383, + "step": 2415, + "valid_targets_mean": 1441.2, + "valid_targets_min": 1079 + }, + { + "epoch": 1.3340683572216097, + "grad_norm": 0.9160934992360564, + "learning_rate": 3.901056893646572e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0996192991733551, + "step": 2420, + "valid_targets_mean": 1449.8, + "valid_targets_min": 1263 + }, + { + "epoch": 1.336824696802646, + "grad_norm": 0.6988782090900804, + "learning_rate": 3.900201145896541e-05, + "loss": 0.2199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10689469426870346, + "step": 2425, + "valid_targets_mean": 1487.4, + "valid_targets_min": 844 + }, + { + "epoch": 1.3395810363836824, + "grad_norm": 0.8728348074861788, + "learning_rate": 3.899341808110558e-05, + "loss": 0.2145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09317543357610703, + "step": 2430, + "valid_targets_mean": 1421.4, + "valid_targets_min": 667 + }, + { + "epoch": 1.3423373759647188, + "grad_norm": 0.768083822555348, + "learning_rate": 3.898478881912162e-05, + "loss": 0.2187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1281663179397583, + "step": 2435, + "valid_targets_mean": 1822.6, + "valid_targets_min": 1041 + }, + { + "epoch": 1.3450937155457552, + "grad_norm": 0.8058771796013698, + "learning_rate": 3.897612368931674e-05, + "loss": 0.2151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09885819256305695, + "step": 2440, + "valid_targets_mean": 1327.4, + "valid_targets_min": 777 + }, + { + "epoch": 1.3478500551267916, + "grad_norm": 0.7931039956517566, + "learning_rate": 3.896742270806189e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10787123441696167, + "step": 2445, + "valid_targets_mean": 1823.8, + "valid_targets_min": 1283 + }, + { + "epoch": 1.350606394707828, + "grad_norm": 0.8789894643215364, + "learning_rate": 3.895868589179579e-05, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07058829069137573, + "step": 2450, + "valid_targets_mean": 924.0, + "valid_targets_min": 620 + }, + { + "epoch": 1.3533627342888643, + "grad_norm": 0.7079883183099432, + "learning_rate": 3.894991325702483e-05, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09063678234815598, + "step": 2455, + "valid_targets_mean": 1480.2, + "valid_targets_min": 656 + }, + { + "epoch": 1.3561190738699007, + "grad_norm": 0.7065884437073642, + "learning_rate": 3.894110482032309e-05, + "loss": 0.2116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08403943479061127, + "step": 2460, + "valid_targets_mean": 1415.5, + "valid_targets_min": 697 + }, + { + "epoch": 1.358875413450937, + "grad_norm": 0.7355565737024445, + "learning_rate": 3.893226059833227e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10440299659967422, + "step": 2465, + "valid_targets_mean": 1655.8, + "valid_targets_min": 816 + }, + { + "epoch": 1.3616317530319735, + "grad_norm": 0.7288625329058336, + "learning_rate": 3.892338060776172e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09626175463199615, + "step": 2470, + "valid_targets_mean": 1553.0, + "valid_targets_min": 850 + }, + { + "epoch": 1.3643880926130099, + "grad_norm": 0.8105730934613236, + "learning_rate": 3.891446486538831e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10180351883172989, + "step": 2475, + "valid_targets_mean": 1353.9, + "valid_targets_min": 747 + }, + { + "epoch": 1.3671444321940462, + "grad_norm": 0.7961629713938433, + "learning_rate": 3.890551338805651e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09211812168359756, + "step": 2480, + "valid_targets_mean": 1420.9, + "valid_targets_min": 933 + }, + { + "epoch": 1.3699007717750826, + "grad_norm": 0.8368359240629544, + "learning_rate": 3.8896526192678265e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09910104423761368, + "step": 2485, + "valid_targets_mean": 1394.9, + "valid_targets_min": 827 + }, + { + "epoch": 1.372657111356119, + "grad_norm": 0.7619913878069878, + "learning_rate": 3.8887503296233034e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09077641367912292, + "step": 2490, + "valid_targets_mean": 1383.2, + "valid_targets_min": 909 + }, + { + "epoch": 1.3754134509371554, + "grad_norm": 0.7355952691470564, + "learning_rate": 3.8878444715767686e-05, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11392970383167267, + "step": 2495, + "valid_targets_mean": 1682.4, + "valid_targets_min": 1094 + }, + { + "epoch": 1.3781697905181918, + "grad_norm": 0.8682332405756354, + "learning_rate": 3.886935046839654e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09417766332626343, + "step": 2500, + "valid_targets_mean": 1118.4, + "valid_targets_min": 720 + }, + { + "epoch": 1.3809261300992282, + "grad_norm": 0.7880391452881526, + "learning_rate": 3.886022057130129e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10837307572364807, + "step": 2505, + "valid_targets_mean": 1538.2, + "valid_targets_min": 610 + }, + { + "epoch": 1.3836824696802645, + "grad_norm": 0.6902556632523349, + "learning_rate": 3.885105504173099e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11060640960931778, + "step": 2510, + "valid_targets_mean": 1570.1, + "valid_targets_min": 713 + }, + { + "epoch": 1.386438809261301, + "grad_norm": 0.7129236728974131, + "learning_rate": 3.8841853897002e-05, + "loss": 0.2119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10007058084011078, + "step": 2515, + "valid_targets_mean": 1537.8, + "valid_targets_min": 874 + }, + { + "epoch": 1.3891951488423373, + "grad_norm": 0.7749353167364372, + "learning_rate": 3.8832617154497974e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.105450838804245, + "step": 2520, + "valid_targets_mean": 1471.2, + "valid_targets_min": 765 + }, + { + "epoch": 1.3919514884233737, + "grad_norm": 0.7438210895927795, + "learning_rate": 3.882334483166983e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08946222066879272, + "step": 2525, + "valid_targets_mean": 1254.4, + "valid_targets_min": 816 + }, + { + "epoch": 1.39470782800441, + "grad_norm": 0.7620985365136406, + "learning_rate": 3.881403694603567e-05, + "loss": 0.2137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1053389385342598, + "step": 2530, + "valid_targets_mean": 1482.9, + "valid_targets_min": 1182 + }, + { + "epoch": 1.3974641675854464, + "grad_norm": 0.5736219861898945, + "learning_rate": 3.880469351518086e-05, + "loss": 0.3037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14309078454971313, + "step": 2535, + "valid_targets_mean": 4566.8, + "valid_targets_min": 702 + }, + { + "epoch": 1.4002205071664828, + "grad_norm": 0.5688041954114808, + "learning_rate": 3.8795314556757835e-05, + "loss": 0.3273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18974880874156952, + "step": 2540, + "valid_targets_mean": 5524.9, + "valid_targets_min": 1005 + }, + { + "epoch": 1.4029768467475192, + "grad_norm": 0.47958912842377466, + "learning_rate": 3.878590008848621e-05, + "loss": 0.3146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12795430421829224, + "step": 2545, + "valid_targets_mean": 4319.8, + "valid_targets_min": 2054 + }, + { + "epoch": 1.4057331863285556, + "grad_norm": 0.5487348352455858, + "learning_rate": 3.877645012815267e-05, + "loss": 0.3147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12353239208459854, + "step": 2550, + "valid_targets_mean": 3498.9, + "valid_targets_min": 1561 + }, + { + "epoch": 1.4084895259095922, + "grad_norm": 0.5498801109790631, + "learning_rate": 3.8766964693610945e-05, + "loss": 0.31, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.192888081073761, + "step": 2555, + "valid_targets_mean": 5020.6, + "valid_targets_min": 1970 + }, + { + "epoch": 1.4112458654906286, + "grad_norm": 0.4478413248834969, + "learning_rate": 3.875744380278181e-05, + "loss": 0.321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15529438853263855, + "step": 2560, + "valid_targets_mean": 6153.9, + "valid_targets_min": 1651 + }, + { + "epoch": 1.414002205071665, + "grad_norm": 0.5775198013743308, + "learning_rate": 3.8747887473653004e-05, + "loss": 0.3017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16269928216934204, + "step": 2565, + "valid_targets_mean": 7671.2, + "valid_targets_min": 2231 + }, + { + "epoch": 1.4167585446527013, + "grad_norm": 0.5135002430574198, + "learning_rate": 3.8738295724279226e-05, + "loss": 0.3089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10382284969091415, + "step": 2570, + "valid_targets_mean": 3124.8, + "valid_targets_min": 467 + }, + { + "epoch": 1.4195148842337377, + "grad_norm": 0.5660762732094216, + "learning_rate": 3.8728668572782103e-05, + "loss": 0.3143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17416036128997803, + "step": 2575, + "valid_targets_mean": 6102.0, + "valid_targets_min": 1748 + }, + { + "epoch": 1.422271223814774, + "grad_norm": 0.5112390667038244, + "learning_rate": 3.871900603735015e-05, + "loss": 0.3151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.146467387676239, + "step": 2580, + "valid_targets_mean": 4508.4, + "valid_targets_min": 1697 + }, + { + "epoch": 1.4250275633958105, + "grad_norm": 0.6114792483653783, + "learning_rate": 3.870930813623871e-05, + "loss": 0.3113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16385561227798462, + "step": 2585, + "valid_targets_mean": 3941.2, + "valid_targets_min": 1223 + }, + { + "epoch": 1.4277839029768469, + "grad_norm": 0.556845997377768, + "learning_rate": 3.8699574887769953e-05, + "loss": 0.308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15862058103084564, + "step": 2590, + "valid_targets_mean": 5561.2, + "valid_targets_min": 1552 + }, + { + "epoch": 1.4305402425578833, + "grad_norm": 0.5740254175216206, + "learning_rate": 3.8689806310332855e-05, + "loss": 0.3137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1706325113773346, + "step": 2595, + "valid_targets_mean": 4325.5, + "valid_targets_min": 1546 + }, + { + "epoch": 1.4332965821389196, + "grad_norm": 0.5795399866707149, + "learning_rate": 3.868000242238309e-05, + "loss": 0.3061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17380179464817047, + "step": 2600, + "valid_targets_mean": 4611.9, + "valid_targets_min": 1957 + }, + { + "epoch": 1.436052921719956, + "grad_norm": 0.5762937647772469, + "learning_rate": 3.867016324244308e-05, + "loss": 0.3179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12809300422668457, + "step": 2605, + "valid_targets_mean": 3394.4, + "valid_targets_min": 1538 + }, + { + "epoch": 1.4388092613009924, + "grad_norm": 0.6862011219223387, + "learning_rate": 3.8660288789101925e-05, + "loss": 0.3172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16662542521953583, + "step": 2610, + "valid_targets_mean": 3414.6, + "valid_targets_min": 1645 + }, + { + "epoch": 1.4415656008820288, + "grad_norm": 0.6528928417645603, + "learning_rate": 3.865037908101535e-05, + "loss": 0.3138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15131890773773193, + "step": 2615, + "valid_targets_mean": 3189.5, + "valid_targets_min": 935 + }, + { + "epoch": 1.4443219404630652, + "grad_norm": 0.7243424124151311, + "learning_rate": 3.864043413690569e-05, + "loss": 0.3238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1646781861782074, + "step": 2620, + "valid_targets_mean": 3584.8, + "valid_targets_min": 1269 + }, + { + "epoch": 1.4470782800441016, + "grad_norm": 0.7016778974731254, + "learning_rate": 3.8630453975561856e-05, + "loss": 0.3217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1248897835612297, + "step": 2625, + "valid_targets_mean": 2098.1, + "valid_targets_min": 1173 + }, + { + "epoch": 1.449834619625138, + "grad_norm": 0.627227584582156, + "learning_rate": 3.8620438615839286e-05, + "loss": 0.317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14370933175086975, + "step": 2630, + "valid_targets_mean": 4135.4, + "valid_targets_min": 1670 + }, + { + "epoch": 1.4525909592061743, + "grad_norm": 0.5900579596604182, + "learning_rate": 3.8610388076659944e-05, + "loss": 0.3174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15358346700668335, + "step": 2635, + "valid_targets_mean": 3372.5, + "valid_targets_min": 1329 + }, + { + "epoch": 1.4553472987872107, + "grad_norm": 0.6269679123440073, + "learning_rate": 3.8600302377012226e-05, + "loss": 0.3053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18681630492210388, + "step": 2640, + "valid_targets_mean": 3950.1, + "valid_targets_min": 2106 + }, + { + "epoch": 1.458103638368247, + "grad_norm": 0.4945944578280748, + "learning_rate": 3.8590181535950966e-05, + "loss": 0.2856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12697631120681763, + "step": 2645, + "valid_targets_mean": 4317.1, + "valid_targets_min": 2222 + }, + { + "epoch": 1.4608599779492835, + "grad_norm": 0.615609093221573, + "learning_rate": 3.858002557259741e-05, + "loss": 0.3134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1517053246498108, + "step": 2650, + "valid_targets_mean": 3463.2, + "valid_targets_min": 1565 + }, + { + "epoch": 1.4636163175303198, + "grad_norm": 0.6791954805528054, + "learning_rate": 3.856983450613912e-05, + "loss": 0.3112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15460029244422913, + "step": 2655, + "valid_targets_mean": 3481.2, + "valid_targets_min": 771 + }, + { + "epoch": 1.4663726571113562, + "grad_norm": 0.5485318567651271, + "learning_rate": 3.855960835583003e-05, + "loss": 0.3204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1500178873538971, + "step": 2660, + "valid_targets_mean": 4016.5, + "valid_targets_min": 1774 + }, + { + "epoch": 1.4691289966923926, + "grad_norm": 0.6900368304508916, + "learning_rate": 3.854934714099031e-05, + "loss": 0.297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12753799557685852, + "step": 2665, + "valid_targets_mean": 2433.4, + "valid_targets_min": 997 + }, + { + "epoch": 1.471885336273429, + "grad_norm": 0.591341423957613, + "learning_rate": 3.8539050881006404e-05, + "loss": 0.2966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14924734830856323, + "step": 2670, + "valid_targets_mean": 3968.9, + "valid_targets_min": 1741 + }, + { + "epoch": 1.4746416758544654, + "grad_norm": 0.6804904036454472, + "learning_rate": 3.852871959533096e-05, + "loss": 0.3131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1502523571252823, + "step": 2675, + "valid_targets_mean": 3747.4, + "valid_targets_min": 863 + }, + { + "epoch": 1.4773980154355018, + "grad_norm": 0.5928274742022687, + "learning_rate": 3.8518353303482796e-05, + "loss": 0.3042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14297261834144592, + "step": 2680, + "valid_targets_mean": 3097.6, + "valid_targets_min": 1410 + }, + { + "epoch": 1.4801543550165381, + "grad_norm": 0.5318907872351892, + "learning_rate": 3.850795202504688e-05, + "loss": 0.2927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1336861252784729, + "step": 2685, + "valid_targets_mean": 4840.6, + "valid_targets_min": 2110 + }, + { + "epoch": 1.4829106945975745, + "grad_norm": 0.6751929697272138, + "learning_rate": 3.8497515779674254e-05, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14627021551132202, + "step": 2690, + "valid_targets_mean": 3410.2, + "valid_targets_min": 1318 + }, + { + "epoch": 1.485667034178611, + "grad_norm": 0.5488878225395145, + "learning_rate": 3.848704458708205e-05, + "loss": 0.3091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15698304772377014, + "step": 2695, + "valid_targets_mean": 4229.5, + "valid_targets_min": 1876 + }, + { + "epoch": 1.4884233737596473, + "grad_norm": 0.6203494099712716, + "learning_rate": 3.847653846705342e-05, + "loss": 0.3007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17547138035297394, + "step": 2700, + "valid_targets_mean": 3783.2, + "valid_targets_min": 2605 + }, + { + "epoch": 1.4911797133406837, + "grad_norm": 0.6312784037199789, + "learning_rate": 3.846599743943749e-05, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13497483730316162, + "step": 2705, + "valid_targets_mean": 3068.0, + "valid_targets_min": 1446 + }, + { + "epoch": 1.49393605292172, + "grad_norm": 0.5745060982430574, + "learning_rate": 3.845542152414934e-05, + "loss": 0.3107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1503790318965912, + "step": 2710, + "valid_targets_mean": 3860.8, + "valid_targets_min": 1270 + }, + { + "epoch": 1.4966923925027564, + "grad_norm": 0.4716949528681372, + "learning_rate": 3.844481074116998e-05, + "loss": 0.3052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15905356407165527, + "step": 2715, + "valid_targets_mean": 5586.5, + "valid_targets_min": 2033 + }, + { + "epoch": 1.4994487320837928, + "grad_norm": 0.6032145188276715, + "learning_rate": 3.8434165110546287e-05, + "loss": 0.2961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17549791932106018, + "step": 2720, + "valid_targets_mean": 4080.6, + "valid_targets_min": 1924 + }, + { + "epoch": 1.5022050716648292, + "grad_norm": 0.5974442315076421, + "learning_rate": 3.842348465239096e-05, + "loss": 0.305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15511080622673035, + "step": 2725, + "valid_targets_mean": 3680.1, + "valid_targets_min": 2098 + }, + { + "epoch": 1.5049614112458656, + "grad_norm": 0.6428289124566768, + "learning_rate": 3.841276938688251e-05, + "loss": 0.3073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2035861313343048, + "step": 2730, + "valid_targets_mean": 4332.0, + "valid_targets_min": 1363 + }, + { + "epoch": 1.507717750826902, + "grad_norm": 0.6508020985181696, + "learning_rate": 3.8402019334265216e-05, + "loss": 0.2981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16090159118175507, + "step": 2735, + "valid_targets_mean": 3358.1, + "valid_targets_min": 1087 + }, + { + "epoch": 1.5104740904079383, + "grad_norm": 0.727551965594471, + "learning_rate": 3.839123451484907e-05, + "loss": 0.3085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1844637244939804, + "step": 2740, + "valid_targets_mean": 4630.6, + "valid_targets_min": 2483 + }, + { + "epoch": 1.5132304299889747, + "grad_norm": 0.5949844213110344, + "learning_rate": 3.838041494900975e-05, + "loss": 0.3019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1724122166633606, + "step": 2745, + "valid_targets_mean": 4551.2, + "valid_targets_min": 2316 + }, + { + "epoch": 1.515986769570011, + "grad_norm": 0.5674397692248111, + "learning_rate": 3.836956065718859e-05, + "loss": 0.295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14422392845153809, + "step": 2750, + "valid_targets_mean": 3923.0, + "valid_targets_min": 2022 + }, + { + "epoch": 1.5187431091510475, + "grad_norm": 0.6795214249005278, + "learning_rate": 3.8358671659892507e-05, + "loss": 0.2888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1066003292798996, + "step": 2755, + "valid_targets_mean": 2890.6, + "valid_targets_min": 1294 + }, + { + "epoch": 1.5214994487320839, + "grad_norm": 0.6866223733038214, + "learning_rate": 3.8347747977694014e-05, + "loss": 0.3102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17892253398895264, + "step": 2760, + "valid_targets_mean": 3638.6, + "valid_targets_min": 1700 + }, + { + "epoch": 1.5242557883131203, + "grad_norm": 0.6836811383818283, + "learning_rate": 3.8336789631231136e-05, + "loss": 0.2955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1282367706298828, + "step": 2765, + "valid_targets_mean": 2722.4, + "valid_targets_min": 1586 + }, + { + "epoch": 1.5270121278941566, + "grad_norm": 0.7157088895307722, + "learning_rate": 3.832579664120741e-05, + "loss": 0.3042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17222225666046143, + "step": 2770, + "valid_targets_mean": 3360.6, + "valid_targets_min": 2407 + }, + { + "epoch": 1.529768467475193, + "grad_norm": 0.7712015142662892, + "learning_rate": 3.83147690283918e-05, + "loss": 0.3072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1613660305738449, + "step": 2775, + "valid_targets_mean": 4063.0, + "valid_targets_min": 2290 + }, + { + "epoch": 1.5325248070562294, + "grad_norm": 0.5711228768985241, + "learning_rate": 3.830370681361869e-05, + "loss": 0.3039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1910039633512497, + "step": 2780, + "valid_targets_mean": 5192.1, + "valid_targets_min": 2189 + }, + { + "epoch": 1.5352811466372658, + "grad_norm": 0.640913984143447, + "learning_rate": 3.829261001778785e-05, + "loss": 0.2995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1917993277311325, + "step": 2785, + "valid_targets_mean": 4296.2, + "valid_targets_min": 2609 + }, + { + "epoch": 1.5380374862183022, + "grad_norm": 0.5795227575429781, + "learning_rate": 3.8281478661864365e-05, + "loss": 0.3116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21383626759052277, + "step": 2790, + "valid_targets_mean": 4534.8, + "valid_targets_min": 2011 + }, + { + "epoch": 1.5407938257993385, + "grad_norm": 0.5570378329551943, + "learning_rate": 3.8270312766878624e-05, + "loss": 0.2878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12305087596178055, + "step": 2795, + "valid_targets_mean": 3611.1, + "valid_targets_min": 1912 + }, + { + "epoch": 1.543550165380375, + "grad_norm": 0.5789291228641621, + "learning_rate": 3.825911235392628e-05, + "loss": 0.2872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1657719612121582, + "step": 2800, + "valid_targets_mean": 3801.9, + "valid_targets_min": 2266 + }, + { + "epoch": 1.5463065049614113, + "grad_norm": 0.5516814477622975, + "learning_rate": 3.824787744416818e-05, + "loss": 0.2868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13575157523155212, + "step": 2805, + "valid_targets_mean": 4454.2, + "valid_targets_min": 904 + }, + { + "epoch": 1.5490628445424477, + "grad_norm": 0.7191218969938455, + "learning_rate": 3.823660805883036e-05, + "loss": 0.2981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16610898077487946, + "step": 2810, + "valid_targets_mean": 3787.4, + "valid_targets_min": 1219 + }, + { + "epoch": 1.551819184123484, + "grad_norm": 0.5438910371478564, + "learning_rate": 3.822530421920399e-05, + "loss": 0.3001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14250554144382477, + "step": 2815, + "valid_targets_mean": 4048.5, + "valid_targets_min": 1658 + }, + { + "epoch": 1.5545755237045205, + "grad_norm": 0.609386158993546, + "learning_rate": 3.8213965946645324e-05, + "loss": 0.297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.146535262465477, + "step": 2820, + "valid_targets_mean": 3499.8, + "valid_targets_min": 1423 + }, + { + "epoch": 1.5573318632855568, + "grad_norm": 0.5205163748411508, + "learning_rate": 3.8202593262575674e-05, + "loss": 0.2941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1171484887599945, + "step": 2825, + "valid_targets_mean": 3917.6, + "valid_targets_min": 2320 + }, + { + "epoch": 1.5600882028665932, + "grad_norm": 0.5568293965599704, + "learning_rate": 3.819118618848137e-05, + "loss": 0.3057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1420169472694397, + "step": 2830, + "valid_targets_mean": 3650.2, + "valid_targets_min": 1188 + }, + { + "epoch": 1.5628445424476296, + "grad_norm": 0.5930379633669852, + "learning_rate": 3.8179744745913724e-05, + "loss": 0.3034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12819933891296387, + "step": 2835, + "valid_targets_mean": 3416.4, + "valid_targets_min": 618 + }, + { + "epoch": 1.565600882028666, + "grad_norm": 0.6752297492354238, + "learning_rate": 3.816826895648895e-05, + "loss": 0.2993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1388484686613083, + "step": 2840, + "valid_targets_mean": 2667.1, + "valid_targets_min": 1364 + }, + { + "epoch": 1.5683572216097024, + "grad_norm": 0.7580819200227801, + "learning_rate": 3.815675884188818e-05, + "loss": 0.2896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15080076456069946, + "step": 2845, + "valid_targets_mean": 3505.4, + "valid_targets_min": 1397 + }, + { + "epoch": 1.5711135611907387, + "grad_norm": 0.5312785137018929, + "learning_rate": 3.81452144238574e-05, + "loss": 0.2829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14913401007652283, + "step": 2850, + "valid_targets_mean": 4666.4, + "valid_targets_min": 1415 + }, + { + "epoch": 1.5738699007717751, + "grad_norm": 0.5839207247191694, + "learning_rate": 3.813363572420738e-05, + "loss": 0.2924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15702247619628906, + "step": 2855, + "valid_targets_mean": 3801.8, + "valid_targets_min": 621 + }, + { + "epoch": 1.5766262403528115, + "grad_norm": 0.54638327212491, + "learning_rate": 3.8122022764813674e-05, + "loss": 0.2807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11187419295310974, + "step": 2860, + "valid_targets_mean": 3552.0, + "valid_targets_min": 1290 + }, + { + "epoch": 1.579382579933848, + "grad_norm": 0.7456292637885612, + "learning_rate": 3.811037556761656e-05, + "loss": 0.2824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11813721805810928, + "step": 2865, + "valid_targets_mean": 2124.8, + "valid_targets_min": 642 + }, + { + "epoch": 1.5821389195148843, + "grad_norm": 0.7978485645960639, + "learning_rate": 3.8098694154621004e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15626396238803864, + "step": 2870, + "valid_targets_mean": 3453.9, + "valid_targets_min": 2109 + }, + { + "epoch": 1.5848952590959207, + "grad_norm": 0.6139186392819328, + "learning_rate": 3.808697854789661e-05, + "loss": 0.2964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21032336354255676, + "step": 2875, + "valid_targets_mean": 4593.1, + "valid_targets_min": 2786 + }, + { + "epoch": 1.587651598676957, + "grad_norm": 0.6093030407113463, + "learning_rate": 3.8075228769577586e-05, + "loss": 0.3102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12725956737995148, + "step": 2880, + "valid_targets_mean": 3278.0, + "valid_targets_min": 1510 + }, + { + "epoch": 1.5904079382579934, + "grad_norm": 0.6766297858090864, + "learning_rate": 3.806344484186271e-05, + "loss": 0.293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18584132194519043, + "step": 2885, + "valid_targets_mean": 3222.0, + "valid_targets_min": 987 + }, + { + "epoch": 1.5931642778390298, + "grad_norm": 0.644972052935103, + "learning_rate": 3.805162678701526e-05, + "loss": 0.2938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12861131131649017, + "step": 2890, + "valid_targets_mean": 3659.5, + "valid_targets_min": 1609 + }, + { + "epoch": 1.5959206174200662, + "grad_norm": 0.5489521465707476, + "learning_rate": 3.8039774627363e-05, + "loss": 0.2869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15786875784397125, + "step": 2895, + "valid_targets_mean": 4652.5, + "valid_targets_min": 1700 + }, + { + "epoch": 1.5986769570011026, + "grad_norm": 0.5015638190387303, + "learning_rate": 3.802788838529813e-05, + "loss": 0.2832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1281820386648178, + "step": 2900, + "valid_targets_mean": 5438.1, + "valid_targets_min": 1767 + }, + { + "epoch": 1.601433296582139, + "grad_norm": 0.5091299659514179, + "learning_rate": 3.801596808327724e-05, + "loss": 0.2921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1334393471479416, + "step": 2905, + "valid_targets_mean": 4504.2, + "valid_targets_min": 2381 + }, + { + "epoch": 1.6041896361631753, + "grad_norm": 0.6730298471759479, + "learning_rate": 3.800401374382127e-05, + "loss": 0.282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12320322543382645, + "step": 2910, + "valid_targets_mean": 2680.2, + "valid_targets_min": 1542 + }, + { + "epoch": 1.6069459757442117, + "grad_norm": 0.5662831434084198, + "learning_rate": 3.799202538951546e-05, + "loss": 0.2928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1066986694931984, + "step": 2915, + "valid_targets_mean": 3123.9, + "valid_targets_min": 1476 + }, + { + "epoch": 1.609702315325248, + "grad_norm": 0.605609471601668, + "learning_rate": 3.798000304300932e-05, + "loss": 0.289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1604762077331543, + "step": 2920, + "valid_targets_mean": 3869.0, + "valid_targets_min": 1711 + }, + { + "epoch": 1.6124586549062845, + "grad_norm": 0.5773857597616489, + "learning_rate": 3.796794672701658e-05, + "loss": 0.2964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13916608691215515, + "step": 2925, + "valid_targets_mean": 3236.9, + "valid_targets_min": 1669 + }, + { + "epoch": 1.6152149944873209, + "grad_norm": 0.6500150807118356, + "learning_rate": 3.795585646431516e-05, + "loss": 0.3028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14213910698890686, + "step": 2930, + "valid_targets_mean": 3207.0, + "valid_targets_min": 1318 + }, + { + "epoch": 1.6179713340683572, + "grad_norm": 0.7360212558185728, + "learning_rate": 3.794373227774708e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1755852997303009, + "step": 2935, + "valid_targets_mean": 2656.8, + "valid_targets_min": 1002 + }, + { + "epoch": 1.6207276736493936, + "grad_norm": 0.6624344166442204, + "learning_rate": 3.7931574190218504e-05, + "loss": 0.3146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11574605107307434, + "step": 2940, + "valid_targets_mean": 2533.6, + "valid_targets_min": 845 + }, + { + "epoch": 1.62348401323043, + "grad_norm": 0.5977239074468305, + "learning_rate": 3.791938222469959e-05, + "loss": 0.2874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1513993889093399, + "step": 2945, + "valid_targets_mean": 4455.6, + "valid_targets_min": 2611 + }, + { + "epoch": 1.6262403528114664, + "grad_norm": 0.5975932926178583, + "learning_rate": 3.790715640422455e-05, + "loss": 0.3065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13330146670341492, + "step": 2950, + "valid_targets_mean": 3460.1, + "valid_targets_min": 1241 + }, + { + "epoch": 1.6289966923925028, + "grad_norm": 0.8680567744773422, + "learning_rate": 3.789489675189152e-05, + "loss": 0.3108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1963823288679123, + "step": 2955, + "valid_targets_mean": 3751.0, + "valid_targets_min": 992 + }, + { + "epoch": 1.6317530319735392, + "grad_norm": 0.6150612728310105, + "learning_rate": 3.788260329086258e-05, + "loss": 0.2913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1437688171863556, + "step": 2960, + "valid_targets_mean": 3815.0, + "valid_targets_min": 1133 + }, + { + "epoch": 1.6345093715545755, + "grad_norm": 0.6475790327928744, + "learning_rate": 3.787027604436367e-05, + "loss": 0.2956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16920322179794312, + "step": 2965, + "valid_targets_mean": 4047.4, + "valid_targets_min": 2256 + }, + { + "epoch": 1.637265711135612, + "grad_norm": 0.5966153818002129, + "learning_rate": 3.785791503568457e-05, + "loss": 0.297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12129484117031097, + "step": 2970, + "valid_targets_mean": 3817.2, + "valid_targets_min": 1919 + }, + { + "epoch": 1.6400220507166483, + "grad_norm": 0.5925490923762734, + "learning_rate": 3.784552028817884e-05, + "loss": 0.3024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14577433466911316, + "step": 2975, + "valid_targets_mean": 3366.4, + "valid_targets_min": 2763 + }, + { + "epoch": 1.6427783902976847, + "grad_norm": 0.7024490895395094, + "learning_rate": 3.7833091825263785e-05, + "loss": 0.3008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15092559158802032, + "step": 2980, + "valid_targets_mean": 2599.4, + "valid_targets_min": 617 + }, + { + "epoch": 1.645534729878721, + "grad_norm": 0.557716095327691, + "learning_rate": 3.7820629670420406e-05, + "loss": 0.2761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13187824189662933, + "step": 2985, + "valid_targets_mean": 3771.2, + "valid_targets_min": 3147 + }, + { + "epoch": 1.6482910694597575, + "grad_norm": 0.6091434512913501, + "learning_rate": 3.780813384719336e-05, + "loss": 0.2826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14354920387268066, + "step": 2990, + "valid_targets_mean": 3780.9, + "valid_targets_min": 1844 + }, + { + "epoch": 1.6510474090407938, + "grad_norm": 0.5933314964776353, + "learning_rate": 3.779560437919091e-05, + "loss": 0.2886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13566184043884277, + "step": 2995, + "valid_targets_mean": 3895.0, + "valid_targets_min": 2330 + }, + { + "epoch": 1.6538037486218302, + "grad_norm": 0.755596442091081, + "learning_rate": 3.7783041290084904e-05, + "loss": 0.2943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14584577083587646, + "step": 3000, + "valid_targets_mean": 2856.8, + "valid_targets_min": 1066 + }, + { + "epoch": 1.6565600882028666, + "grad_norm": 0.5575730168103438, + "learning_rate": 3.777044460361067e-05, + "loss": 0.2837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12748906016349792, + "step": 3005, + "valid_targets_mean": 3496.4, + "valid_targets_min": 1713 + }, + { + "epoch": 1.659316427783903, + "grad_norm": 0.5935344661110408, + "learning_rate": 3.775781434356705e-05, + "loss": 0.2975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14406552910804749, + "step": 3010, + "valid_targets_mean": 3918.8, + "valid_targets_min": 981 + }, + { + "epoch": 1.6620727673649394, + "grad_norm": 0.5564009334396962, + "learning_rate": 3.77451505338163e-05, + "loss": 0.2869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15665720403194427, + "step": 3015, + "valid_targets_mean": 4680.2, + "valid_targets_min": 3325 + }, + { + "epoch": 1.6648291069459757, + "grad_norm": 0.7128652992855812, + "learning_rate": 3.773245319828405e-05, + "loss": 0.2935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1521500200033188, + "step": 3020, + "valid_targets_mean": 3138.9, + "valid_targets_min": 1353 + }, + { + "epoch": 1.6675854465270121, + "grad_norm": 0.6298853533529212, + "learning_rate": 3.771972236095929e-05, + "loss": 0.2866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12295828759670258, + "step": 3025, + "valid_targets_mean": 3986.6, + "valid_targets_min": 2273 + }, + { + "epoch": 1.6703417861080485, + "grad_norm": 0.5681727807988722, + "learning_rate": 3.77069580458943e-05, + "loss": 0.2783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11987817287445068, + "step": 3030, + "valid_targets_mean": 3204.4, + "valid_targets_min": 2106 + }, + { + "epoch": 1.673098125689085, + "grad_norm": 0.5593124677605196, + "learning_rate": 3.76941602772046e-05, + "loss": 0.272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1204424500465393, + "step": 3035, + "valid_targets_mean": 3788.4, + "valid_targets_min": 1687 + }, + { + "epoch": 1.6758544652701213, + "grad_norm": 0.6090954222256264, + "learning_rate": 3.768132907906893e-05, + "loss": 0.2807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15470297634601593, + "step": 3040, + "valid_targets_mean": 3502.8, + "valid_targets_min": 2246 + }, + { + "epoch": 1.6786108048511577, + "grad_norm": 0.6221114441731828, + "learning_rate": 3.766846447572916e-05, + "loss": 0.2818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13533233106136322, + "step": 3045, + "valid_targets_mean": 3663.8, + "valid_targets_min": 906 + }, + { + "epoch": 1.681367144432194, + "grad_norm": 0.5646025618422137, + "learning_rate": 3.765556649149031e-05, + "loss": 0.278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17447945475578308, + "step": 3050, + "valid_targets_mean": 4811.0, + "valid_targets_min": 1874 + }, + { + "epoch": 1.6841234840132304, + "grad_norm": 0.6284259610095806, + "learning_rate": 3.7642635150720426e-05, + "loss": 0.2728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14882586896419525, + "step": 3055, + "valid_targets_mean": 3815.6, + "valid_targets_min": 873 + }, + { + "epoch": 1.6868798235942668, + "grad_norm": 0.617505185432628, + "learning_rate": 3.7629670477850616e-05, + "loss": 0.2798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1285349726676941, + "step": 3060, + "valid_targets_mean": 3252.1, + "valid_targets_min": 1697 + }, + { + "epoch": 1.6896361631753032, + "grad_norm": 0.629538855679215, + "learning_rate": 3.761667249737491e-05, + "loss": 0.2795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11310988664627075, + "step": 3065, + "valid_targets_mean": 2395.0, + "valid_targets_min": 897 + }, + { + "epoch": 1.6923925027563396, + "grad_norm": 0.6057159770750449, + "learning_rate": 3.760364123385033e-05, + "loss": 0.302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13435395061969757, + "step": 3070, + "valid_targets_mean": 3104.0, + "valid_targets_min": 1695 + }, + { + "epoch": 1.695148842337376, + "grad_norm": 0.6362067887702987, + "learning_rate": 3.759057671189673e-05, + "loss": 0.287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13324424624443054, + "step": 3075, + "valid_targets_mean": 3211.9, + "valid_targets_min": 1712 + }, + { + "epoch": 1.6979051819184123, + "grad_norm": 0.5899731789724141, + "learning_rate": 3.757747895619681e-05, + "loss": 0.2833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12776611745357513, + "step": 3080, + "valid_targets_mean": 3500.5, + "valid_targets_min": 993 + }, + { + "epoch": 1.7006615214994487, + "grad_norm": 0.5721482363364679, + "learning_rate": 3.756434799149607e-05, + "loss": 0.2741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14713749289512634, + "step": 3085, + "valid_targets_mean": 4181.4, + "valid_targets_min": 1679 + }, + { + "epoch": 1.703417861080485, + "grad_norm": 0.5643597814145338, + "learning_rate": 3.7551183842602735e-05, + "loss": 0.284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12716394662857056, + "step": 3090, + "valid_targets_mean": 3779.9, + "valid_targets_min": 2250 + }, + { + "epoch": 1.7061742006615215, + "grad_norm": 0.5728216756393172, + "learning_rate": 3.7537986534387754e-05, + "loss": 0.2885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14222858846187592, + "step": 3095, + "valid_targets_mean": 4175.8, + "valid_targets_min": 2892 + }, + { + "epoch": 1.7089305402425579, + "grad_norm": 0.5567362516945203, + "learning_rate": 3.752475609178468e-05, + "loss": 0.2902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1489541232585907, + "step": 3100, + "valid_targets_mean": 4485.5, + "valid_targets_min": 2965 + }, + { + "epoch": 1.7116868798235942, + "grad_norm": 0.5852206769782148, + "learning_rate": 3.75114925397897e-05, + "loss": 0.2832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11441446840763092, + "step": 3105, + "valid_targets_mean": 3737.5, + "valid_targets_min": 1817 + }, + { + "epoch": 1.7144432194046306, + "grad_norm": 0.5738098922389069, + "learning_rate": 3.7498195903461556e-05, + "loss": 0.2861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16468864679336548, + "step": 3110, + "valid_targets_mean": 4726.0, + "valid_targets_min": 1727 + }, + { + "epoch": 1.717199558985667, + "grad_norm": 0.5743754217190098, + "learning_rate": 3.748486620792147e-05, + "loss": 0.2809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12456713616847992, + "step": 3115, + "valid_targets_mean": 3801.2, + "valid_targets_min": 1959 + }, + { + "epoch": 1.7199558985667034, + "grad_norm": 0.5652757525544154, + "learning_rate": 3.747150347835314e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10050415247678757, + "step": 3120, + "valid_targets_mean": 2283.4, + "valid_targets_min": 333 + }, + { + "epoch": 1.7227122381477398, + "grad_norm": 0.40391103080520785, + "learning_rate": 3.7458107740002676e-05, + "loss": 0.1328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060675811022520065, + "step": 3125, + "valid_targets_mean": 3613.2, + "valid_targets_min": 712 + }, + { + "epoch": 1.7254685777287762, + "grad_norm": 0.48786515623820126, + "learning_rate": 3.744467901817854e-05, + "loss": 0.1174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06299149245023727, + "step": 3130, + "valid_targets_mean": 2940.8, + "valid_targets_min": 732 + }, + { + "epoch": 1.7282249173098125, + "grad_norm": 0.48939094438235864, + "learning_rate": 3.743121733825152e-05, + "loss": 0.12, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0985046923160553, + "step": 3135, + "valid_targets_mean": 2727.5, + "valid_targets_min": 927 + }, + { + "epoch": 1.730981256890849, + "grad_norm": 0.4165521225330927, + "learning_rate": 3.741772272565468e-05, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047460176050662994, + "step": 3140, + "valid_targets_mean": 2835.4, + "valid_targets_min": 818 + }, + { + "epoch": 1.7337375964718853, + "grad_norm": 0.4547839365130038, + "learning_rate": 3.7404195205883274e-05, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06073397025465965, + "step": 3145, + "valid_targets_mean": 2910.0, + "valid_targets_min": 859 + }, + { + "epoch": 1.7364939360529217, + "grad_norm": 0.8367117366434351, + "learning_rate": 3.7390634804494757e-05, + "loss": 0.1666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06539306044578552, + "step": 3150, + "valid_targets_mean": 1040.4, + "valid_targets_min": 539 + }, + { + "epoch": 1.739250275633958, + "grad_norm": 0.4284423154508106, + "learning_rate": 3.737704154710868e-05, + "loss": 0.1294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05191371962428093, + "step": 3155, + "valid_targets_mean": 3001.5, + "valid_targets_min": 920 + }, + { + "epoch": 1.7420066152149944, + "grad_norm": 0.43505004882792514, + "learning_rate": 3.73634154594067e-05, + "loss": 0.1254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06267690658569336, + "step": 3160, + "valid_targets_mean": 3366.2, + "valid_targets_min": 2295 + }, + { + "epoch": 1.7447629547960308, + "grad_norm": 0.345551940932624, + "learning_rate": 3.734975656713248e-05, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042231932282447815, + "step": 3165, + "valid_targets_mean": 3784.6, + "valid_targets_min": 962 + }, + { + "epoch": 1.7475192943770672, + "grad_norm": 0.35565802561571236, + "learning_rate": 3.733606489609166e-05, + "loss": 0.0969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06565237045288086, + "step": 3170, + "valid_targets_mean": 4249.2, + "valid_targets_min": 2091 + }, + { + "epoch": 1.7502756339581036, + "grad_norm": 0.39440762611990454, + "learning_rate": 3.732234047215181e-05, + "loss": 0.1108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059343911707401276, + "step": 3175, + "valid_targets_mean": 4391.9, + "valid_targets_min": 3349 + }, + { + "epoch": 1.75303197353914, + "grad_norm": 0.37388295333660015, + "learning_rate": 3.730858332124239e-05, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05458410084247589, + "step": 3180, + "valid_targets_mean": 3944.4, + "valid_targets_min": 518 + }, + { + "epoch": 1.7557883131201764, + "grad_norm": 0.581074025014653, + "learning_rate": 3.729479346935468e-05, + "loss": 0.1637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06297165900468826, + "step": 3185, + "valid_targets_mean": 1885.0, + "valid_targets_min": 549 + }, + { + "epoch": 1.7585446527012127, + "grad_norm": 0.3938904876739415, + "learning_rate": 3.728097094254174e-05, + "loss": 0.1041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04946866258978844, + "step": 3190, + "valid_targets_mean": 3442.8, + "valid_targets_min": 2724 + }, + { + "epoch": 1.7613009922822491, + "grad_norm": 0.5074344818761156, + "learning_rate": 3.726711576691838e-05, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047059472650289536, + "step": 3195, + "valid_targets_mean": 1619.2, + "valid_targets_min": 518 + }, + { + "epoch": 1.7640573318632855, + "grad_norm": 0.3657549893418414, + "learning_rate": 3.725322796866106e-05, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031259194016456604, + "step": 3200, + "valid_targets_mean": 3329.6, + "valid_targets_min": 844 + }, + { + "epoch": 1.7668136714443219, + "grad_norm": 0.4524936302553174, + "learning_rate": 3.7239307574007916e-05, + "loss": 0.1161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057531069964170456, + "step": 3205, + "valid_targets_mean": 2772.8, + "valid_targets_min": 577 + }, + { + "epoch": 1.7695700110253583, + "grad_norm": 0.5198624897956776, + "learning_rate": 3.722535460925864e-05, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06530879437923431, + "step": 3210, + "valid_targets_mean": 2086.0, + "valid_targets_min": 851 + }, + { + "epoch": 1.7723263506063947, + "grad_norm": 0.45903917092488106, + "learning_rate": 3.721136910077446e-05, + "loss": 0.1296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07108600437641144, + "step": 3215, + "valid_targets_mean": 2864.6, + "valid_targets_min": 774 + }, + { + "epoch": 1.775082690187431, + "grad_norm": 0.4279435567928515, + "learning_rate": 3.71973510749781e-05, + "loss": 0.1268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051938027143478394, + "step": 3220, + "valid_targets_mean": 1743.6, + "valid_targets_min": 615 + }, + { + "epoch": 1.7778390297684674, + "grad_norm": 0.5571937279789996, + "learning_rate": 3.7183300558353704e-05, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12716928124427795, + "step": 3225, + "valid_targets_mean": 3786.1, + "valid_targets_min": 2819 + }, + { + "epoch": 1.7805953693495038, + "grad_norm": 0.32898101853695244, + "learning_rate": 3.716921757744682e-05, + "loss": 0.1071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04497702047228813, + "step": 3230, + "valid_targets_mean": 4070.4, + "valid_targets_min": 3732 + }, + { + "epoch": 1.7833517089305402, + "grad_norm": 0.4805542070531515, + "learning_rate": 3.715510215886431e-05, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0514492504298687, + "step": 3235, + "valid_targets_mean": 1204.5, + "valid_targets_min": 714 + }, + { + "epoch": 1.7861080485115766, + "grad_norm": 0.3853422978166455, + "learning_rate": 3.714095432927433e-05, + "loss": 0.1559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051514632999897, + "step": 3240, + "valid_targets_mean": 3416.2, + "valid_targets_min": 1812 + }, + { + "epoch": 1.788864388092613, + "grad_norm": 0.3828777485863772, + "learning_rate": 3.712677411540627e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05736031383275986, + "step": 3245, + "valid_targets_mean": 3800.5, + "valid_targets_min": 3178 + }, + { + "epoch": 1.7916207276736493, + "grad_norm": 0.7695359334914594, + "learning_rate": 3.711256154405071e-05, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14198487997055054, + "step": 3250, + "valid_targets_mean": 1691.4, + "valid_targets_min": 920 + }, + { + "epoch": 1.7943770672546857, + "grad_norm": 0.45531412123253867, + "learning_rate": 3.709831664205935e-05, + "loss": 0.3067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06461755186319351, + "step": 3255, + "valid_targets_mean": 3673.5, + "valid_targets_min": 2889 + }, + { + "epoch": 1.797133406835722, + "grad_norm": 0.4290328464077996, + "learning_rate": 3.708403943634499e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0556086041033268, + "step": 3260, + "valid_targets_mean": 3149.2, + "valid_targets_min": 833 + }, + { + "epoch": 1.7998897464167585, + "grad_norm": 0.5215972588923086, + "learning_rate": 3.706972995388143e-05, + "loss": 0.1287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10337137430906296, + "step": 3265, + "valid_targets_mean": 3340.4, + "valid_targets_min": 1803 + }, + { + "epoch": 1.8026460859977949, + "grad_norm": 0.4355056959738321, + "learning_rate": 3.705538822170348e-05, + "loss": 0.2234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07006622850894928, + "step": 3270, + "valid_targets_mean": 3548.2, + "valid_targets_min": 2782 + }, + { + "epoch": 1.8054024255788312, + "grad_norm": 0.26296474457890046, + "learning_rate": 3.704101426690686e-05, + "loss": 0.0935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04332941398024559, + "step": 3275, + "valid_targets_mean": 4605.8, + "valid_targets_min": 2664 + }, + { + "epoch": 1.8081587651598676, + "grad_norm": 0.33854207572996403, + "learning_rate": 3.702660811664819e-05, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028864098712801933, + "step": 3280, + "valid_targets_mean": 2750.2, + "valid_targets_min": 560 + }, + { + "epoch": 1.810915104740904, + "grad_norm": 0.380643648232678, + "learning_rate": 3.70121697981449e-05, + "loss": 0.1243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06073581799864769, + "step": 3285, + "valid_targets_mean": 4261.1, + "valid_targets_min": 1615 + }, + { + "epoch": 1.8136714443219404, + "grad_norm": 0.37611167333690815, + "learning_rate": 3.6997699338675184e-05, + "loss": 0.1085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04803231358528137, + "step": 3290, + "valid_targets_mean": 3580.5, + "valid_targets_min": 2172 + }, + { + "epoch": 1.8164277839029768, + "grad_norm": 0.4220463234687843, + "learning_rate": 3.698319676557799e-05, + "loss": 0.1056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05537541210651398, + "step": 3295, + "valid_targets_mean": 3065.4, + "valid_targets_min": 536 + }, + { + "epoch": 1.8191841234840131, + "grad_norm": 0.3905691293970137, + "learning_rate": 3.696866210625291e-05, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05592438578605652, + "step": 3300, + "valid_targets_mean": 3633.1, + "valid_targets_min": 2264 + }, + { + "epoch": 1.8219404630650495, + "grad_norm": 0.3717800431599445, + "learning_rate": 3.695409538816018e-05, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04645654559135437, + "step": 3305, + "valid_targets_mean": 2809.9, + "valid_targets_min": 820 + }, + { + "epoch": 1.824696802646086, + "grad_norm": 0.6033414221093523, + "learning_rate": 3.693949663882058e-05, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14735695719718933, + "step": 3310, + "valid_targets_mean": 2020.5, + "valid_targets_min": 864 + }, + { + "epoch": 1.8274531422271223, + "grad_norm": 0.36782891314375127, + "learning_rate": 3.692486588581543e-05, + "loss": 0.1037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05833785980939865, + "step": 3315, + "valid_targets_mean": 3609.9, + "valid_targets_min": 2719 + }, + { + "epoch": 1.8302094818081587, + "grad_norm": 0.637819829103743, + "learning_rate": 3.6910203156786496e-05, + "loss": 0.101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07648296654224396, + "step": 3320, + "valid_targets_mean": 1611.6, + "valid_targets_min": 785 + }, + { + "epoch": 1.832965821389195, + "grad_norm": 0.3387607560573485, + "learning_rate": 3.6895508479435964e-05, + "loss": 0.0983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04272644966840744, + "step": 3325, + "valid_targets_mean": 3711.8, + "valid_targets_min": 2353 + }, + { + "epoch": 1.8357221609702314, + "grad_norm": 0.4115837189708781, + "learning_rate": 3.688078188152639e-05, + "loss": 0.1002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04598747938871384, + "step": 3330, + "valid_targets_mean": 1996.5, + "valid_targets_min": 549 + }, + { + "epoch": 1.8384785005512678, + "grad_norm": 0.3722406544001807, + "learning_rate": 3.6866023390880605e-05, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05468544363975525, + "step": 3335, + "valid_targets_mean": 3332.5, + "valid_targets_min": 908 + }, + { + "epoch": 1.8412348401323042, + "grad_norm": 0.3965875366532395, + "learning_rate": 3.6851233035381746e-05, + "loss": 0.133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09218797087669373, + "step": 3340, + "valid_targets_mean": 3527.0, + "valid_targets_min": 2196 + }, + { + "epoch": 1.8439911797133406, + "grad_norm": 0.35633110324988476, + "learning_rate": 3.683641084297309e-05, + "loss": 0.1247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03563661873340607, + "step": 3345, + "valid_targets_mean": 1686.0, + "valid_targets_min": 598 + }, + { + "epoch": 1.846747519294377, + "grad_norm": 0.9366607524398746, + "learning_rate": 3.6821556841658116e-05, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12440724670886993, + "step": 3350, + "valid_targets_mean": 1330.2, + "valid_targets_min": 578 + }, + { + "epoch": 1.8495038588754134, + "grad_norm": 0.5648083443961956, + "learning_rate": 3.680667105950037e-05, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06604795902967453, + "step": 3355, + "valid_targets_mean": 1678.9, + "valid_targets_min": 695 + }, + { + "epoch": 1.8522601984564497, + "grad_norm": 0.5047526907510951, + "learning_rate": 3.6791753524623456e-05, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06543979793787003, + "step": 3360, + "valid_targets_mean": 1888.5, + "valid_targets_min": 691 + }, + { + "epoch": 1.8550165380374861, + "grad_norm": 0.3929515948503137, + "learning_rate": 3.6776804265210963e-05, + "loss": 0.1376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049328818917274475, + "step": 3365, + "valid_targets_mean": 3686.0, + "valid_targets_min": 2024 + }, + { + "epoch": 1.8577728776185225, + "grad_norm": 0.6675879390534764, + "learning_rate": 3.676182330950641e-05, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05995006859302521, + "step": 3370, + "valid_targets_mean": 1356.1, + "valid_targets_min": 523 + }, + { + "epoch": 1.8605292171995589, + "grad_norm": 0.33062837477646195, + "learning_rate": 3.67468106858132e-05, + "loss": 0.1114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040570877492427826, + "step": 3375, + "valid_targets_mean": 4120.2, + "valid_targets_min": 468 + }, + { + "epoch": 1.8632855567805953, + "grad_norm": 0.29126128737702156, + "learning_rate": 3.673176642249459e-05, + "loss": 0.1148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031423285603523254, + "step": 3380, + "valid_targets_mean": 5551.2, + "valid_targets_min": 4225 + }, + { + "epoch": 1.8660418963616316, + "grad_norm": 0.317363539696819, + "learning_rate": 3.671669054797358e-05, + "loss": 0.1038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04070987552404404, + "step": 3385, + "valid_targets_mean": 3806.8, + "valid_targets_min": 459 + }, + { + "epoch": 1.868798235942668, + "grad_norm": 0.32407912998685273, + "learning_rate": 3.670158309073291e-05, + "loss": 0.1019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04796344041824341, + "step": 3390, + "valid_targets_mean": 3498.2, + "valid_targets_min": 981 + }, + { + "epoch": 1.8715545755237044, + "grad_norm": 0.38711596204158405, + "learning_rate": 3.6686444079314995e-05, + "loss": 0.1125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06750103086233139, + "step": 3395, + "valid_targets_mean": 2464.6, + "valid_targets_min": 697 + }, + { + "epoch": 1.8743109151047408, + "grad_norm": 0.429025484050459, + "learning_rate": 3.667127354232185e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06696552038192749, + "step": 3400, + "valid_targets_mean": 3671.2, + "valid_targets_min": 2759 + }, + { + "epoch": 1.8770672546857772, + "grad_norm": 0.4202498363409056, + "learning_rate": 3.6656071508415064e-05, + "loss": 0.1208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060574326664209366, + "step": 3405, + "valid_targets_mean": 3702.6, + "valid_targets_min": 2901 + }, + { + "epoch": 1.8798235942668136, + "grad_norm": 0.3363721199935189, + "learning_rate": 3.6640838006315715e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04646093025803566, + "step": 3410, + "valid_targets_mean": 3455.4, + "valid_targets_min": 1099 + }, + { + "epoch": 1.88257993384785, + "grad_norm": 0.41928868227926547, + "learning_rate": 3.6625573064804366e-05, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06958258152008057, + "step": 3415, + "valid_targets_mean": 3808.8, + "valid_targets_min": 1874 + }, + { + "epoch": 1.8853362734288863, + "grad_norm": 0.3616294577365587, + "learning_rate": 3.661027671272094e-05, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04604456201195717, + "step": 3420, + "valid_targets_mean": 3013.8, + "valid_targets_min": 724 + }, + { + "epoch": 1.8880926130099227, + "grad_norm": 0.29221414796047357, + "learning_rate": 3.659494897896473e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04072284698486328, + "step": 3425, + "valid_targets_mean": 3806.5, + "valid_targets_min": 2769 + }, + { + "epoch": 1.890848952590959, + "grad_norm": 0.47262492335095596, + "learning_rate": 3.65795898924943e-05, + "loss": 0.123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04945040121674538, + "step": 3430, + "valid_targets_mean": 2427.1, + "valid_targets_min": 591 + }, + { + "epoch": 1.8936052921719955, + "grad_norm": 0.38380605220896585, + "learning_rate": 3.656419948232747e-05, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056629031896591187, + "step": 3435, + "valid_targets_mean": 3443.2, + "valid_targets_min": 600 + }, + { + "epoch": 1.8963616317530319, + "grad_norm": 0.6147016723534111, + "learning_rate": 3.654877777754123e-05, + "loss": 0.1304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06583196669816971, + "step": 3440, + "valid_targets_mean": 3542.1, + "valid_targets_min": 1648 + }, + { + "epoch": 1.8991179713340682, + "grad_norm": 0.4419831739309578, + "learning_rate": 3.653332480727168e-05, + "loss": 0.1186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06057174131274223, + "step": 3445, + "valid_targets_mean": 1957.9, + "valid_targets_min": 457 + }, + { + "epoch": 1.9018743109151046, + "grad_norm": 0.7728955452678146, + "learning_rate": 3.651784060071401e-05, + "loss": 0.1297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06401053816080093, + "step": 3450, + "valid_targets_mean": 823.9, + "valid_targets_min": 612 + }, + { + "epoch": 1.904630650496141, + "grad_norm": 0.5281536829091233, + "learning_rate": 3.6502325187122425e-05, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09671596437692642, + "step": 3455, + "valid_targets_mean": 2383.2, + "valid_targets_min": 898 + }, + { + "epoch": 1.9073869900771774, + "grad_norm": 0.41695200410558086, + "learning_rate": 3.648677859581006e-05, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04814234375953674, + "step": 3460, + "valid_targets_mean": 3037.9, + "valid_targets_min": 1146 + }, + { + "epoch": 1.9101433296582138, + "grad_norm": 0.4325919859086952, + "learning_rate": 3.6471200856148984e-05, + "loss": 0.1149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07221969962120056, + "step": 3465, + "valid_targets_mean": 3039.9, + "valid_targets_min": 1511 + }, + { + "epoch": 1.9128996692392501, + "grad_norm": 0.6122624379365321, + "learning_rate": 3.6455591997570115e-05, + "loss": 0.1051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06000150740146637, + "step": 3470, + "valid_targets_mean": 1301.2, + "valid_targets_min": 628 + }, + { + "epoch": 1.9156560088202865, + "grad_norm": 0.5813826600687727, + "learning_rate": 3.643995204956315e-05, + "loss": 0.1268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07427273690700531, + "step": 3475, + "valid_targets_mean": 1488.0, + "valid_targets_min": 691 + }, + { + "epoch": 1.918412348401323, + "grad_norm": 0.3047816871244396, + "learning_rate": 3.6424281041676526e-05, + "loss": 0.1108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035443104803562164, + "step": 3480, + "valid_targets_mean": 3970.2, + "valid_targets_min": 3657 + }, + { + "epoch": 1.9211686879823593, + "grad_norm": 0.35709152492527946, + "learning_rate": 3.6408579003517347e-05, + "loss": 0.1007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06495866179466248, + "step": 3485, + "valid_targets_mean": 5056.2, + "valid_targets_min": 3831 + }, + { + "epoch": 1.9239250275633957, + "grad_norm": 0.3323477492095763, + "learning_rate": 3.639284596475138e-05, + "loss": 0.1106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0481954962015152, + "step": 3490, + "valid_targets_mean": 2402.1, + "valid_targets_min": 762 + }, + { + "epoch": 1.926681367144432, + "grad_norm": 0.3631017347271474, + "learning_rate": 3.637708195510293e-05, + "loss": 0.0975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05563568323850632, + "step": 3495, + "valid_targets_mean": 3574.1, + "valid_targets_min": 667 + }, + { + "epoch": 1.9294377067254684, + "grad_norm": 0.3716773307444722, + "learning_rate": 3.636128700435481e-05, + "loss": 0.0934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07049725949764252, + "step": 3500, + "valid_targets_mean": 5005.6, + "valid_targets_min": 4104 + }, + { + "epoch": 1.9321940463065048, + "grad_norm": 0.36015944752413936, + "learning_rate": 3.634546114234833e-05, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0472850427031517, + "step": 3505, + "valid_targets_mean": 3256.8, + "valid_targets_min": 784 + }, + { + "epoch": 1.9349503858875412, + "grad_norm": 1.001078791617526, + "learning_rate": 3.632960439898315e-05, + "loss": 0.1321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05685728043317795, + "step": 3510, + "valid_targets_mean": 3479.5, + "valid_targets_min": 1090 + }, + { + "epoch": 1.9377067254685776, + "grad_norm": 0.36975840012219524, + "learning_rate": 3.631371680421732e-05, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0647442489862442, + "step": 3515, + "valid_targets_mean": 3237.6, + "valid_targets_min": 704 + }, + { + "epoch": 1.940463065049614, + "grad_norm": 0.30640967759901694, + "learning_rate": 3.6297798388067126e-05, + "loss": 0.1098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04502519220113754, + "step": 3520, + "valid_targets_mean": 2824.0, + "valid_targets_min": 774 + }, + { + "epoch": 1.9432194046306503, + "grad_norm": 0.2972511206398028, + "learning_rate": 3.628184918060714e-05, + "loss": 0.136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04718878120183945, + "step": 3525, + "valid_targets_mean": 3551.1, + "valid_targets_min": 987 + }, + { + "epoch": 1.9459757442116867, + "grad_norm": 0.4006783277800931, + "learning_rate": 3.626586921197007e-05, + "loss": 0.1071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06285085529088974, + "step": 3530, + "valid_targets_mean": 3200.4, + "valid_targets_min": 1048 + }, + { + "epoch": 1.9487320837927231, + "grad_norm": 0.40847207591405554, + "learning_rate": 3.624985851234676e-05, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0557343065738678, + "step": 3535, + "valid_targets_mean": 3389.4, + "valid_targets_min": 1072 + }, + { + "epoch": 1.9514884233737595, + "grad_norm": 0.375823967992634, + "learning_rate": 3.6233817111986096e-05, + "loss": 0.1113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0841301828622818, + "step": 3540, + "valid_targets_mean": 3102.2, + "valid_targets_min": 906 + }, + { + "epoch": 1.9542447629547959, + "grad_norm": 0.5913416014358271, + "learning_rate": 3.621774504119498e-05, + "loss": 0.1683, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06845861673355103, + "step": 3545, + "valid_targets_mean": 1292.9, + "valid_targets_min": 570 + }, + { + "epoch": 1.9570011025358323, + "grad_norm": 0.43409841138694355, + "learning_rate": 3.620164233033826e-05, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07676853239536285, + "step": 3550, + "valid_targets_mean": 3471.0, + "valid_targets_min": 835 + }, + { + "epoch": 1.9597574421168686, + "grad_norm": 0.41345477106261025, + "learning_rate": 3.618550900983867e-05, + "loss": 0.1079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045319296419620514, + "step": 3555, + "valid_targets_mean": 2156.0, + "valid_targets_min": 635 + }, + { + "epoch": 1.962513781697905, + "grad_norm": 0.39556888020831904, + "learning_rate": 3.616934511017677e-05, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05361534282565117, + "step": 3560, + "valid_targets_mean": 2765.2, + "valid_targets_min": 1591 + }, + { + "epoch": 1.9652701212789414, + "grad_norm": 0.33669723484612357, + "learning_rate": 3.615315066189089e-05, + "loss": 0.1063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04625920578837395, + "step": 3565, + "valid_targets_mean": 3275.8, + "valid_targets_min": 800 + }, + { + "epoch": 1.9680264608599778, + "grad_norm": 0.3338254653315004, + "learning_rate": 3.6136925695577085e-05, + "loss": 0.0999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04942747950553894, + "step": 3570, + "valid_targets_mean": 3218.0, + "valid_targets_min": 958 + }, + { + "epoch": 1.9707828004410142, + "grad_norm": 0.5025088697659145, + "learning_rate": 3.612067024188907e-05, + "loss": 0.1371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06691572070121765, + "step": 3575, + "valid_targets_mean": 2122.1, + "valid_targets_min": 889 + }, + { + "epoch": 1.9735391400220506, + "grad_norm": 0.5397214267740374, + "learning_rate": 3.6104384331538144e-05, + "loss": 0.0963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055498458445072174, + "step": 3580, + "valid_targets_mean": 3309.6, + "valid_targets_min": 1024 + }, + { + "epoch": 1.976295479603087, + "grad_norm": 0.4725924877621237, + "learning_rate": 3.608806799529317e-05, + "loss": 0.1074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05428100377321243, + "step": 3585, + "valid_targets_mean": 2179.5, + "valid_targets_min": 786 + }, + { + "epoch": 1.9790518191841233, + "grad_norm": 0.43786733625492236, + "learning_rate": 3.607172126398046e-05, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06460972130298615, + "step": 3590, + "valid_targets_mean": 3064.8, + "valid_targets_min": 2573 + }, + { + "epoch": 1.98180815876516, + "grad_norm": 0.2821922690102509, + "learning_rate": 3.60553441684838e-05, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052218616008758545, + "step": 3595, + "valid_targets_mean": 3985.8, + "valid_targets_min": 2274 + }, + { + "epoch": 1.9845644983461963, + "grad_norm": 0.3558730721331269, + "learning_rate": 3.603893673974429e-05, + "loss": 0.0954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06763364374637604, + "step": 3600, + "valid_targets_mean": 4017.9, + "valid_targets_min": 2431 + }, + { + "epoch": 1.9873208379272327, + "grad_norm": 0.35787778170866685, + "learning_rate": 3.6022499008760374e-05, + "loss": 0.1053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06720007210969925, + "step": 3605, + "valid_targets_mean": 4359.2, + "valid_targets_min": 2555 + }, + { + "epoch": 1.990077177508269, + "grad_norm": 0.5421623576641638, + "learning_rate": 3.600603100658773e-05, + "loss": 0.1437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09993834793567657, + "step": 3610, + "valid_targets_mean": 2205.4, + "valid_targets_min": 1535 + }, + { + "epoch": 1.9928335170893055, + "grad_norm": 0.30151977037232075, + "learning_rate": 3.5989532764339254e-05, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031974755227565765, + "step": 3615, + "valid_targets_mean": 3608.8, + "valid_targets_min": 1097 + }, + { + "epoch": 1.9955898566703418, + "grad_norm": 0.3211355774905931, + "learning_rate": 3.5973004313184923e-05, + "loss": 0.1085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0405430942773819, + "step": 3620, + "valid_targets_mean": 3166.8, + "valid_targets_min": 840 + }, + { + "epoch": 1.9983461962513782, + "grad_norm": 0.3313604284007408, + "learning_rate": 3.5956445684351837e-05, + "loss": 0.1042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047243520617485046, + "step": 3625, + "valid_targets_mean": 3711.2, + "valid_targets_min": 2555 + }, + { + "epoch": 2.0011025358324144, + "grad_norm": 0.4417900552077343, + "learning_rate": 3.5939856909124085e-05, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10523481667041779, + "step": 3630, + "valid_targets_mean": 8514.0, + "valid_targets_min": 6788 + }, + { + "epoch": 2.0038588754134508, + "grad_norm": 0.43018383092552026, + "learning_rate": 3.5923238018842715e-05, + "loss": 0.2348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11579734086990356, + "step": 3635, + "valid_targets_mean": 6668.0, + "valid_targets_min": 5933 + }, + { + "epoch": 2.006615214994487, + "grad_norm": 0.44934129270932416, + "learning_rate": 3.590658904490568e-05, + "loss": 0.2225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10224385559558868, + "step": 3640, + "valid_targets_mean": 4107.0, + "valid_targets_min": 161 + }, + { + "epoch": 2.0093715545755235, + "grad_norm": 0.35776445753898584, + "learning_rate": 3.588991001876775e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09964029490947723, + "step": 3645, + "valid_targets_mean": 7946.0, + "valid_targets_min": 5596 + }, + { + "epoch": 2.01212789415656, + "grad_norm": 0.3831597223899965, + "learning_rate": 3.58732009719405e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11221307516098022, + "step": 3650, + "valid_targets_mean": 7392.8, + "valid_targets_min": 5629 + }, + { + "epoch": 2.0148842337375963, + "grad_norm": 0.3768326700781796, + "learning_rate": 3.5856461935992194e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10874032974243164, + "step": 3655, + "valid_targets_mean": 6879.6, + "valid_targets_min": 5113 + }, + { + "epoch": 2.0176405733186327, + "grad_norm": 0.4331225618487861, + "learning_rate": 3.583969294254779e-05, + "loss": 0.23, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11305272579193115, + "step": 3660, + "valid_targets_mean": 6919.5, + "valid_targets_min": 5080 + }, + { + "epoch": 2.020396912899669, + "grad_norm": 0.39667380863700974, + "learning_rate": 3.58228940232888e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09901660680770874, + "step": 3665, + "valid_targets_mean": 6377.0, + "valid_targets_min": 3939 + }, + { + "epoch": 2.0231532524807054, + "grad_norm": 0.36526035975548493, + "learning_rate": 3.580606520995331e-05, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10620422661304474, + "step": 3670, + "valid_targets_mean": 6392.2, + "valid_targets_min": 4355 + }, + { + "epoch": 2.025909592061742, + "grad_norm": 0.44472022035588415, + "learning_rate": 3.578920653433588e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11966025829315186, + "step": 3675, + "valid_targets_mean": 5196.5, + "valid_targets_min": 137 + }, + { + "epoch": 2.028665931642778, + "grad_norm": 0.3368227867537486, + "learning_rate": 3.5772318028287464e-05, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09297772496938705, + "step": 3680, + "valid_targets_mean": 7344.4, + "valid_targets_min": 5244 + }, + { + "epoch": 2.0314222712238146, + "grad_norm": 0.34217063335281195, + "learning_rate": 3.5755399723715396e-05, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08921800553798676, + "step": 3685, + "valid_targets_mean": 7921.0, + "valid_targets_min": 5311 + }, + { + "epoch": 2.034178610804851, + "grad_norm": 0.3559525933809306, + "learning_rate": 3.573845165258331e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1068873256444931, + "step": 3690, + "valid_targets_mean": 7931.2, + "valid_targets_min": 6015 + }, + { + "epoch": 2.0369349503858873, + "grad_norm": 0.3470199659928046, + "learning_rate": 3.572147384691107e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.094902902841568, + "step": 3695, + "valid_targets_mean": 7640.8, + "valid_targets_min": 5248 + }, + { + "epoch": 2.0396912899669237, + "grad_norm": 0.375195105439285, + "learning_rate": 3.5704466338774714e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0999273806810379, + "step": 3700, + "valid_targets_mean": 7249.6, + "valid_targets_min": 5217 + }, + { + "epoch": 2.04244762954796, + "grad_norm": 0.37681711667015055, + "learning_rate": 3.5687429160306405e-05, + "loss": 0.2082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11440923810005188, + "step": 3705, + "valid_targets_mean": 7619.9, + "valid_targets_min": 5094 + }, + { + "epoch": 2.0452039691289965, + "grad_norm": 0.3780238969966865, + "learning_rate": 3.567036234369435e-05, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09642066806554794, + "step": 3710, + "valid_targets_mean": 7732.8, + "valid_targets_min": 5842 + }, + { + "epoch": 2.047960308710033, + "grad_norm": 0.41923524002965057, + "learning_rate": 3.565326592118277e-05, + "loss": 0.2043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10410739481449127, + "step": 3715, + "valid_targets_mean": 7952.9, + "valid_targets_min": 5494 + }, + { + "epoch": 2.0507166482910693, + "grad_norm": 0.40058531027139305, + "learning_rate": 3.563613992507181e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10874912142753601, + "step": 3720, + "valid_targets_mean": 8001.0, + "valid_targets_min": 5454 + }, + { + "epoch": 2.0534729878721056, + "grad_norm": 0.5773795246834018, + "learning_rate": 3.5618984387717475e-05, + "loss": 0.2246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07925702631473541, + "step": 3725, + "valid_targets_mean": 2985.1, + "valid_targets_min": 1896 + }, + { + "epoch": 2.056229327453142, + "grad_norm": 0.377351160724215, + "learning_rate": 3.5601799341531605e-05, + "loss": 0.1836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08823058009147644, + "step": 3730, + "valid_targets_mean": 6714.6, + "valid_targets_min": 4917 + }, + { + "epoch": 2.0589856670341784, + "grad_norm": 0.38165768592624916, + "learning_rate": 3.558458481898178e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10995317250490189, + "step": 3735, + "valid_targets_mean": 7595.2, + "valid_targets_min": 5488 + }, + { + "epoch": 2.061742006615215, + "grad_norm": 0.4031546465553637, + "learning_rate": 3.556734085259127e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10904796421527863, + "step": 3740, + "valid_targets_mean": 7046.1, + "valid_targets_min": 5579 + }, + { + "epoch": 2.064498346196251, + "grad_norm": 0.40057008460804294, + "learning_rate": 3.555006747493896e-05, + "loss": 0.2092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10134632885456085, + "step": 3745, + "valid_targets_mean": 7108.0, + "valid_targets_min": 4963 + }, + { + "epoch": 2.0672546857772875, + "grad_norm": 0.39336912312667394, + "learning_rate": 3.553276471865934e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10373738408088684, + "step": 3750, + "valid_targets_mean": 6810.1, + "valid_targets_min": 5666 + }, + { + "epoch": 2.070011025358324, + "grad_norm": 0.3658505686501154, + "learning_rate": 3.551543261644236e-05, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09564688801765442, + "step": 3755, + "valid_targets_mean": 8068.5, + "valid_targets_min": 5524 + }, + { + "epoch": 2.0727673649393603, + "grad_norm": 0.40060763073847105, + "learning_rate": 3.549807120103343e-05, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08650697767734528, + "step": 3760, + "valid_targets_mean": 6553.8, + "valid_targets_min": 3621 + }, + { + "epoch": 2.075523704520397, + "grad_norm": 0.457879192281609, + "learning_rate": 3.548068050523336e-05, + "loss": 0.1826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09583864361047745, + "step": 3765, + "valid_targets_mean": 6762.5, + "valid_targets_min": 4968 + }, + { + "epoch": 2.078280044101433, + "grad_norm": 0.40713054730485654, + "learning_rate": 3.5463260561898246e-05, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09910881519317627, + "step": 3770, + "valid_targets_mean": 6143.9, + "valid_targets_min": 4917 + }, + { + "epoch": 2.08103638368247, + "grad_norm": 0.5585466642895837, + "learning_rate": 3.544581140393947e-05, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13889947533607483, + "step": 3775, + "valid_targets_mean": 4583.8, + "valid_targets_min": 215 + }, + { + "epoch": 2.083792723263506, + "grad_norm": 0.3679419716286392, + "learning_rate": 3.54283330643236e-05, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10516968369483948, + "step": 3780, + "valid_targets_mean": 7317.1, + "valid_targets_min": 4948 + }, + { + "epoch": 2.0865490628445427, + "grad_norm": 0.5021859270870359, + "learning_rate": 3.541082557607231e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1054849624633789, + "step": 3785, + "valid_targets_mean": 6437.6, + "valid_targets_min": 4510 + }, + { + "epoch": 2.0893054024255786, + "grad_norm": 0.4064916525253523, + "learning_rate": 3.53932889722624e-05, + "loss": 0.2148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11647052317857742, + "step": 3790, + "valid_targets_mean": 7707.9, + "valid_targets_min": 5796 + }, + { + "epoch": 2.0920617420066154, + "grad_norm": 0.4053900682027221, + "learning_rate": 3.5375723286025615e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09975050389766693, + "step": 3795, + "valid_targets_mean": 7900.9, + "valid_targets_min": 5643 + }, + { + "epoch": 2.0948180815876514, + "grad_norm": 0.3850444729387167, + "learning_rate": 3.53581285505487e-05, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11220157146453857, + "step": 3800, + "valid_targets_mean": 7627.2, + "valid_targets_min": 5742 + }, + { + "epoch": 2.097574421168688, + "grad_norm": 0.3594168476511988, + "learning_rate": 3.534050479907324e-05, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11910156905651093, + "step": 3805, + "valid_targets_mean": 8725.6, + "valid_targets_min": 5141 + }, + { + "epoch": 2.1003307607497246, + "grad_norm": 0.39780572395375674, + "learning_rate": 3.532285206489567e-05, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1142977774143219, + "step": 3810, + "valid_targets_mean": 6967.8, + "valid_targets_min": 6088 + }, + { + "epoch": 2.103087100330761, + "grad_norm": 0.34983760729932756, + "learning_rate": 3.5305170381367165e-05, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09630568325519562, + "step": 3815, + "valid_targets_mean": 7140.1, + "valid_targets_min": 4834 + }, + { + "epoch": 2.1058434399117973, + "grad_norm": 0.39351467643683563, + "learning_rate": 3.528745978189361e-05, + "loss": 0.2074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09595423936843872, + "step": 3820, + "valid_targets_mean": 6706.9, + "valid_targets_min": 5108 + }, + { + "epoch": 2.1085997794928337, + "grad_norm": 0.41173887295624034, + "learning_rate": 3.52697202999355e-05, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10367536544799805, + "step": 3825, + "valid_targets_mean": 5931.4, + "valid_targets_min": 4708 + }, + { + "epoch": 2.11135611907387, + "grad_norm": 0.3919925423909442, + "learning_rate": 3.52519519690079e-05, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09619426727294922, + "step": 3830, + "valid_targets_mean": 7145.6, + "valid_targets_min": 5280 + }, + { + "epoch": 2.1141124586549065, + "grad_norm": 0.6532048156368425, + "learning_rate": 3.5234154822680394e-05, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06856104731559753, + "step": 3835, + "valid_targets_mean": 1554.9, + "valid_targets_min": 592 + }, + { + "epoch": 2.116868798235943, + "grad_norm": 1.1864810148402238, + "learning_rate": 3.5216328894577e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11745339632034302, + "step": 3840, + "valid_targets_mean": 7069.5, + "valid_targets_min": 4765 + }, + { + "epoch": 2.1196251378169793, + "grad_norm": 0.3876035623303499, + "learning_rate": 3.5198474218376106e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10173526406288147, + "step": 3845, + "valid_targets_mean": 6193.4, + "valid_targets_min": 5587 + }, + { + "epoch": 2.1223814773980156, + "grad_norm": 0.42062890220045085, + "learning_rate": 3.518059082781043e-05, + "loss": 0.2041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1062692329287529, + "step": 3850, + "valid_targets_mean": 6612.2, + "valid_targets_min": 4894 + }, + { + "epoch": 2.125137816979052, + "grad_norm": 0.3865381102395448, + "learning_rate": 3.516267875666692e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09445300698280334, + "step": 3855, + "valid_targets_mean": 6605.9, + "valid_targets_min": 3795 + }, + { + "epoch": 2.1278941565600884, + "grad_norm": 0.3897109992677676, + "learning_rate": 3.514473803878672e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10271283239126205, + "step": 3860, + "valid_targets_mean": 7186.8, + "valid_targets_min": 4890 + }, + { + "epoch": 2.130650496141125, + "grad_norm": 0.4369678528592689, + "learning_rate": 3.51267687080651e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10411788523197174, + "step": 3865, + "valid_targets_mean": 6544.4, + "valid_targets_min": 4615 + }, + { + "epoch": 2.133406835722161, + "grad_norm": 0.3868620449792883, + "learning_rate": 3.5108770798451376e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11044663935899734, + "step": 3870, + "valid_targets_mean": 6339.2, + "valid_targets_min": 4798 + }, + { + "epoch": 2.1361631753031975, + "grad_norm": 0.5052632069251424, + "learning_rate": 3.5090744343948875e-05, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11794136464595795, + "step": 3875, + "valid_targets_mean": 4558.8, + "valid_targets_min": 148 + }, + { + "epoch": 2.138919514884234, + "grad_norm": 0.4365614227471576, + "learning_rate": 3.507268937861484e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11230622231960297, + "step": 3880, + "valid_targets_mean": 6807.9, + "valid_targets_min": 5164 + }, + { + "epoch": 2.1416758544652703, + "grad_norm": 0.4075717368051184, + "learning_rate": 3.505460593656039e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10206775367259979, + "step": 3885, + "valid_targets_mean": 7011.8, + "valid_targets_min": 4974 + }, + { + "epoch": 2.1444321940463067, + "grad_norm": 0.38856349793211015, + "learning_rate": 3.5036494051950414e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09245841205120087, + "step": 3890, + "valid_targets_mean": 5794.2, + "valid_targets_min": 4809 + }, + { + "epoch": 2.147188533627343, + "grad_norm": 0.3947688602801654, + "learning_rate": 3.5018353759003586e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09121550619602203, + "step": 3895, + "valid_targets_mean": 6401.8, + "valid_targets_min": 4842 + }, + { + "epoch": 2.1499448732083795, + "grad_norm": 0.397229689750155, + "learning_rate": 3.500018509199222e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10315051674842834, + "step": 3900, + "valid_targets_mean": 6182.1, + "valid_targets_min": 5116 + }, + { + "epoch": 2.152701212789416, + "grad_norm": 0.4173801967833107, + "learning_rate": 3.4981988085242243e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09169760346412659, + "step": 3905, + "valid_targets_mean": 6165.1, + "valid_targets_min": 5764 + }, + { + "epoch": 2.1554575523704522, + "grad_norm": 0.7569296089614398, + "learning_rate": 3.4963762773133126e-05, + "loss": 0.2252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08254774659872055, + "step": 3910, + "valid_targets_mean": 1428.2, + "valid_targets_min": 1062 + }, + { + "epoch": 2.1582138919514886, + "grad_norm": 0.7643262946287898, + "learning_rate": 3.494550919009782e-05, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09747211635112762, + "step": 3915, + "valid_targets_mean": 1663.0, + "valid_targets_min": 960 + }, + { + "epoch": 2.160970231532525, + "grad_norm": 0.8193420968815408, + "learning_rate": 3.4927227370622675e-05, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0964818000793457, + "step": 3920, + "valid_targets_mean": 1476.5, + "valid_targets_min": 933 + }, + { + "epoch": 2.1637265711135614, + "grad_norm": 0.7792236824224837, + "learning_rate": 3.49089173492474e-05, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09605573117733002, + "step": 3925, + "valid_targets_mean": 1478.1, + "valid_targets_min": 881 + }, + { + "epoch": 2.1664829106945978, + "grad_norm": 0.9915861163916563, + "learning_rate": 3.4890579160564985e-05, + "loss": 0.1996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11922421306371689, + "step": 3930, + "valid_targets_mean": 1640.6, + "valid_targets_min": 869 + }, + { + "epoch": 2.169239250275634, + "grad_norm": 0.7938020177510351, + "learning_rate": 3.487221283922164e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10822120308876038, + "step": 3935, + "valid_targets_mean": 1709.0, + "valid_targets_min": 1090 + }, + { + "epoch": 2.1719955898566705, + "grad_norm": 0.7684508316127598, + "learning_rate": 3.485381841991671e-05, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07659199833869934, + "step": 3940, + "valid_targets_mean": 1157.9, + "valid_targets_min": 761 + }, + { + "epoch": 2.174751929437707, + "grad_norm": 0.7246132518440405, + "learning_rate": 3.4835395937402636e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09514109790325165, + "step": 3945, + "valid_targets_mean": 1553.6, + "valid_targets_min": 1241 + }, + { + "epoch": 2.1775082690187433, + "grad_norm": 0.758700582326445, + "learning_rate": 3.481694542648488e-05, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0897405743598938, + "step": 3950, + "valid_targets_mean": 1538.1, + "valid_targets_min": 733 + }, + { + "epoch": 2.1802646085997797, + "grad_norm": 0.7381424144074343, + "learning_rate": 3.479846692202187e-05, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059913743287324905, + "step": 3955, + "valid_targets_mean": 1250.6, + "valid_targets_min": 835 + }, + { + "epoch": 2.183020948180816, + "grad_norm": 0.7399129343230754, + "learning_rate": 3.4779960458924876e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11009322851896286, + "step": 3960, + "valid_targets_mean": 1596.2, + "valid_targets_min": 1046 + }, + { + "epoch": 2.1857772877618524, + "grad_norm": 0.7924752025992841, + "learning_rate": 3.476142607215805e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09340207278728485, + "step": 3965, + "valid_targets_mean": 1646.5, + "valid_targets_min": 797 + }, + { + "epoch": 2.188533627342889, + "grad_norm": 0.7609616986609584, + "learning_rate": 3.474286379673826e-05, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0964922308921814, + "step": 3970, + "valid_targets_mean": 1448.9, + "valid_targets_min": 570 + }, + { + "epoch": 2.191289966923925, + "grad_norm": 0.7398133074067775, + "learning_rate": 3.472427366773508e-05, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06944597512483597, + "step": 3975, + "valid_targets_mean": 1164.1, + "valid_targets_min": 548 + }, + { + "epoch": 2.1940463065049616, + "grad_norm": 0.8582042351766288, + "learning_rate": 3.47056557202707e-05, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09755994379520416, + "step": 3980, + "valid_targets_mean": 1338.0, + "valid_targets_min": 682 + }, + { + "epoch": 2.196802646085998, + "grad_norm": 0.7389618213374148, + "learning_rate": 3.468700998951987e-05, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09869241714477539, + "step": 3985, + "valid_targets_mean": 1632.8, + "valid_targets_min": 896 + }, + { + "epoch": 2.1995589856670343, + "grad_norm": 0.7980056976235737, + "learning_rate": 3.466833651070983e-05, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12938037514686584, + "step": 3990, + "valid_targets_mean": 1806.5, + "valid_targets_min": 597 + }, + { + "epoch": 2.2023153252480707, + "grad_norm": 0.8297858123524207, + "learning_rate": 3.464963531912024e-05, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08366896957159042, + "step": 3995, + "valid_targets_mean": 1289.9, + "valid_targets_min": 706 + }, + { + "epoch": 2.205071664829107, + "grad_norm": 0.9174595516846432, + "learning_rate": 3.463090645008313e-05, + "loss": 0.1906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06147385388612747, + "step": 4000, + "valid_targets_mean": 1354.1, + "valid_targets_min": 754 + }, + { + "epoch": 2.2078280044101435, + "grad_norm": 0.71513722491629, + "learning_rate": 3.461214993898282e-05, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08852845430374146, + "step": 4005, + "valid_targets_mean": 1461.9, + "valid_targets_min": 819 + }, + { + "epoch": 2.21058434399118, + "grad_norm": 0.7378667566992704, + "learning_rate": 3.459336582125584e-05, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07387156039476395, + "step": 4010, + "valid_targets_mean": 1225.5, + "valid_targets_min": 684 + }, + { + "epoch": 2.2133406835722162, + "grad_norm": 0.8313652728590256, + "learning_rate": 3.4574554132390884e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09833618253469467, + "step": 4015, + "valid_targets_mean": 1282.2, + "valid_targets_min": 711 + }, + { + "epoch": 2.2160970231532526, + "grad_norm": 0.8017510024522516, + "learning_rate": 3.455571490792873e-05, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10198875516653061, + "step": 4020, + "valid_targets_mean": 1634.8, + "valid_targets_min": 1190 + }, + { + "epoch": 2.218853362734289, + "grad_norm": 0.7732885734598844, + "learning_rate": 3.45368481834622e-05, + "loss": 0.1832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09606744349002838, + "step": 4025, + "valid_targets_mean": 1615.9, + "valid_targets_min": 774 + }, + { + "epoch": 2.2216097023153254, + "grad_norm": 0.6718615003941802, + "learning_rate": 3.451795399463605e-05, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09657790511846542, + "step": 4030, + "valid_targets_mean": 1848.4, + "valid_targets_min": 734 + }, + { + "epoch": 2.224366041896362, + "grad_norm": 0.8009573169778992, + "learning_rate": 3.4499032377146926e-05, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09645441174507141, + "step": 4035, + "valid_targets_mean": 1570.6, + "valid_targets_min": 1011 + }, + { + "epoch": 2.227122381477398, + "grad_norm": 0.7916076096237811, + "learning_rate": 3.448008336674331e-05, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10767923295497894, + "step": 4040, + "valid_targets_mean": 1697.0, + "valid_targets_min": 803 + }, + { + "epoch": 2.2298787210584345, + "grad_norm": 0.7682343572471392, + "learning_rate": 3.446110699922541e-05, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08238881081342697, + "step": 4045, + "valid_targets_mean": 1307.1, + "valid_targets_min": 782 + }, + { + "epoch": 2.232635060639471, + "grad_norm": 0.7426469636049479, + "learning_rate": 3.4442103310445166e-05, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09898242354393005, + "step": 4050, + "valid_targets_mean": 1596.6, + "valid_targets_min": 646 + }, + { + "epoch": 2.2353914002205073, + "grad_norm": 0.7327785078919717, + "learning_rate": 3.442307233630608e-05, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13181385397911072, + "step": 4055, + "valid_targets_mean": 1817.5, + "valid_targets_min": 922 + }, + { + "epoch": 2.2381477398015437, + "grad_norm": 0.7695764798902701, + "learning_rate": 3.4404014112763236e-05, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10620490461587906, + "step": 4060, + "valid_targets_mean": 1822.2, + "valid_targets_min": 796 + }, + { + "epoch": 2.24090407938258, + "grad_norm": 0.7392151330565492, + "learning_rate": 3.43849286758232e-05, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09556275606155396, + "step": 4065, + "valid_targets_mean": 1542.4, + "valid_targets_min": 903 + }, + { + "epoch": 2.2436604189636165, + "grad_norm": 1.2072636697831685, + "learning_rate": 3.436581606154394e-05, + "loss": 0.1823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10454177111387253, + "step": 4070, + "valid_targets_mean": 1459.5, + "valid_targets_min": 780 + }, + { + "epoch": 2.246416758544653, + "grad_norm": 0.7049200026980261, + "learning_rate": 3.434667630603478e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0974837988615036, + "step": 4075, + "valid_targets_mean": 1581.1, + "valid_targets_min": 783 + }, + { + "epoch": 2.249173098125689, + "grad_norm": 0.7385765850674274, + "learning_rate": 3.4327509445456325e-05, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07353440672159195, + "step": 4080, + "valid_targets_mean": 1342.1, + "valid_targets_min": 662 + }, + { + "epoch": 2.2519294377067256, + "grad_norm": 0.7703430982396028, + "learning_rate": 3.430831551602038e-05, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08304440975189209, + "step": 4085, + "valid_targets_mean": 1269.1, + "valid_targets_min": 724 + }, + { + "epoch": 2.254685777287762, + "grad_norm": 0.7818880325666101, + "learning_rate": 3.4289094553989885e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10150696337223053, + "step": 4090, + "valid_targets_mean": 1510.1, + "valid_targets_min": 966 + }, + { + "epoch": 2.2574421168687984, + "grad_norm": 0.8183344143734358, + "learning_rate": 3.426984659567887e-05, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08161082863807678, + "step": 4095, + "valid_targets_mean": 1327.8, + "valid_targets_min": 773 + }, + { + "epoch": 2.2601984564498347, + "grad_norm": 0.7729170442786207, + "learning_rate": 3.425057167745236e-05, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07742751389741898, + "step": 4100, + "valid_targets_mean": 1353.2, + "valid_targets_min": 807 + }, + { + "epoch": 2.262954796030871, + "grad_norm": 0.7926863001292335, + "learning_rate": 3.42312698357263e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0726255476474762, + "step": 4105, + "valid_targets_mean": 1142.5, + "valid_targets_min": 782 + }, + { + "epoch": 2.2657111356119075, + "grad_norm": 0.6836604769577472, + "learning_rate": 3.421194110696754e-05, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10145572572946548, + "step": 4110, + "valid_targets_mean": 1849.9, + "valid_targets_min": 634 + }, + { + "epoch": 2.268467475192944, + "grad_norm": 0.7345703646672201, + "learning_rate": 3.419258552769369e-05, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09446150064468384, + "step": 4115, + "valid_targets_mean": 1599.6, + "valid_targets_min": 573 + }, + { + "epoch": 2.2712238147739803, + "grad_norm": 0.6506594613766818, + "learning_rate": 3.41732031344731e-05, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10397475212812424, + "step": 4120, + "valid_targets_mean": 1994.9, + "valid_targets_min": 1102 + }, + { + "epoch": 2.2739801543550167, + "grad_norm": 0.7029687233682509, + "learning_rate": 3.4153793963924794e-05, + "loss": 0.1913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06905566155910492, + "step": 4125, + "valid_targets_mean": 1239.2, + "valid_targets_min": 841 + }, + { + "epoch": 2.276736493936053, + "grad_norm": 0.7474508410797022, + "learning_rate": 3.413435805271836e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09163033962249756, + "step": 4130, + "valid_targets_mean": 1467.5, + "valid_targets_min": 956 + }, + { + "epoch": 2.2794928335170894, + "grad_norm": 0.8097085989459102, + "learning_rate": 3.411489543757394e-05, + "loss": 0.1889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09622394293546677, + "step": 4135, + "valid_targets_mean": 1300.9, + "valid_targets_min": 857 + }, + { + "epoch": 2.282249173098126, + "grad_norm": 0.6790936204515535, + "learning_rate": 3.409540615526209e-05, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0978386253118515, + "step": 4140, + "valid_targets_mean": 1389.8, + "valid_targets_min": 808 + }, + { + "epoch": 2.285005512679162, + "grad_norm": 0.73365560396833, + "learning_rate": 3.407589024260378e-05, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07963447272777557, + "step": 4145, + "valid_targets_mean": 1303.8, + "valid_targets_min": 677 + }, + { + "epoch": 2.2877618522601986, + "grad_norm": 0.8880167029376889, + "learning_rate": 3.405634773647027e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07855790108442307, + "step": 4150, + "valid_targets_mean": 1220.1, + "valid_targets_min": 727 + }, + { + "epoch": 2.290518191841235, + "grad_norm": 0.7237261237817447, + "learning_rate": 3.403677867378307e-05, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07669747620820999, + "step": 4155, + "valid_targets_mean": 1262.0, + "valid_targets_min": 726 + }, + { + "epoch": 2.2932745314222713, + "grad_norm": 0.8207230981715815, + "learning_rate": 3.4017183091513884e-05, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10847216099500656, + "step": 4160, + "valid_targets_mean": 1615.0, + "valid_targets_min": 669 + }, + { + "epoch": 2.2960308710033077, + "grad_norm": 0.6524441620880489, + "learning_rate": 3.399756102668448e-05, + "loss": 0.1841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1051015853881836, + "step": 4165, + "valid_targets_mean": 1806.5, + "valid_targets_min": 792 + }, + { + "epoch": 2.298787210584344, + "grad_norm": 0.7241234956872353, + "learning_rate": 3.3977912516366685e-05, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09316051006317139, + "step": 4170, + "valid_targets_mean": 1627.6, + "valid_targets_min": 736 + }, + { + "epoch": 2.3015435501653805, + "grad_norm": 0.7902467016458746, + "learning_rate": 3.395823759768229e-05, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1297021508216858, + "step": 4175, + "valid_targets_mean": 1822.9, + "valid_targets_min": 694 + }, + { + "epoch": 2.304299889746417, + "grad_norm": 0.8057193520368823, + "learning_rate": 3.393853630780297e-05, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11972413957118988, + "step": 4180, + "valid_targets_mean": 1848.9, + "valid_targets_min": 1437 + }, + { + "epoch": 2.3070562293274532, + "grad_norm": 0.7210446411616037, + "learning_rate": 3.391880868395022e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11316122114658356, + "step": 4185, + "valid_targets_mean": 1852.4, + "valid_targets_min": 809 + }, + { + "epoch": 2.3098125689084896, + "grad_norm": 0.7653675926336914, + "learning_rate": 3.38990547633953e-05, + "loss": 0.1801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08458778262138367, + "step": 4190, + "valid_targets_mean": 1375.5, + "valid_targets_min": 790 + }, + { + "epoch": 2.312568908489526, + "grad_norm": 0.8027386842758492, + "learning_rate": 3.3879274583459145e-05, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08519227802753448, + "step": 4195, + "valid_targets_mean": 1338.8, + "valid_targets_min": 750 + }, + { + "epoch": 2.3153252480705624, + "grad_norm": 0.7447303039543663, + "learning_rate": 3.3859468181512305e-05, + "loss": 0.1836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.112679623067379, + "step": 4200, + "valid_targets_mean": 1778.8, + "valid_targets_min": 826 + }, + { + "epoch": 2.3180815876515988, + "grad_norm": 0.7169650325683636, + "learning_rate": 3.3839635594974864e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08437786251306534, + "step": 4205, + "valid_targets_mean": 1600.8, + "valid_targets_min": 907 + }, + { + "epoch": 2.320837927232635, + "grad_norm": 0.711068937498706, + "learning_rate": 3.381977686131639e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09715353697538376, + "step": 4210, + "valid_targets_mean": 1474.5, + "valid_targets_min": 1093 + }, + { + "epoch": 2.3235942668136715, + "grad_norm": 0.7367881640693851, + "learning_rate": 3.379989201805582e-05, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0952945426106453, + "step": 4215, + "valid_targets_mean": 1577.2, + "valid_targets_min": 1119 + }, + { + "epoch": 2.326350606394708, + "grad_norm": 1.3646964698063697, + "learning_rate": 3.377998110276147e-05, + "loss": 0.1837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10611128062009811, + "step": 4220, + "valid_targets_mean": 1512.1, + "valid_targets_min": 700 + }, + { + "epoch": 2.3291069459757443, + "grad_norm": 0.6973103533185624, + "learning_rate": 3.376004415305086e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07343744486570358, + "step": 4225, + "valid_targets_mean": 1288.2, + "valid_targets_min": 810 + }, + { + "epoch": 2.3318632855567807, + "grad_norm": 0.7901620410617438, + "learning_rate": 3.374008120659073e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0885704755783081, + "step": 4230, + "valid_targets_mean": 1328.9, + "valid_targets_min": 635 + }, + { + "epoch": 2.334619625137817, + "grad_norm": 0.7675388100139938, + "learning_rate": 3.372009230109694e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11123967170715332, + "step": 4235, + "valid_targets_mean": 1542.5, + "valid_targets_min": 673 + }, + { + "epoch": 2.3373759647188534, + "grad_norm": 0.8427727790060149, + "learning_rate": 3.3700077474334354e-05, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10082757472991943, + "step": 4240, + "valid_targets_mean": 1215.6, + "valid_targets_min": 641 + }, + { + "epoch": 2.34013230429989, + "grad_norm": 0.7687538225255685, + "learning_rate": 3.3680036764116866e-05, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08242233842611313, + "step": 4245, + "valid_targets_mean": 1222.6, + "valid_targets_min": 701 + }, + { + "epoch": 2.342888643880926, + "grad_norm": 0.70949301323614, + "learning_rate": 3.365997020830722e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09950898587703705, + "step": 4250, + "valid_targets_mean": 1850.6, + "valid_targets_min": 1202 + }, + { + "epoch": 2.3456449834619626, + "grad_norm": 0.7318227059156271, + "learning_rate": 3.363987784481702e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07201792299747467, + "step": 4255, + "valid_targets_mean": 1265.5, + "valid_targets_min": 657 + }, + { + "epoch": 2.348401323042999, + "grad_norm": 0.7613006534173571, + "learning_rate": 3.361975971160662e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08599737286567688, + "step": 4260, + "valid_targets_mean": 1388.0, + "valid_targets_min": 672 + }, + { + "epoch": 2.3511576626240354, + "grad_norm": 0.7708648374274438, + "learning_rate": 3.359961584668505e-05, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11095890402793884, + "step": 4265, + "valid_targets_mean": 1813.6, + "valid_targets_min": 1019 + }, + { + "epoch": 2.3539140022050717, + "grad_norm": 0.7643167930863976, + "learning_rate": 3.357944628810998e-05, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0730147436261177, + "step": 4270, + "valid_targets_mean": 1114.6, + "valid_targets_min": 732 + }, + { + "epoch": 2.356670341786108, + "grad_norm": 0.7296246344567072, + "learning_rate": 3.35592510739876e-05, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06616927683353424, + "step": 4275, + "valid_targets_mean": 1194.5, + "valid_targets_min": 634 + }, + { + "epoch": 2.3594266813671445, + "grad_norm": 0.6412466958657433, + "learning_rate": 3.3539030242472574e-05, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0921182632446289, + "step": 4280, + "valid_targets_mean": 1756.5, + "valid_targets_min": 905 + }, + { + "epoch": 2.362183020948181, + "grad_norm": 0.7131943992221603, + "learning_rate": 3.351878383176797e-05, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10437525063753128, + "step": 4285, + "valid_targets_mean": 1554.9, + "valid_targets_min": 596 + }, + { + "epoch": 2.3649393605292173, + "grad_norm": 0.7772812024778029, + "learning_rate": 3.34985118801252e-05, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07284867763519287, + "step": 4290, + "valid_targets_mean": 1214.6, + "valid_targets_min": 687 + }, + { + "epoch": 2.3676957001102537, + "grad_norm": 0.7216085452538521, + "learning_rate": 3.347821442584389e-05, + "loss": 0.1914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10474824905395508, + "step": 4295, + "valid_targets_mean": 1955.1, + "valid_targets_min": 1199 + }, + { + "epoch": 2.37045203969129, + "grad_norm": 0.6988278047143789, + "learning_rate": 3.3457891507271876e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09360268712043762, + "step": 4300, + "valid_targets_mean": 1758.6, + "valid_targets_min": 1214 + }, + { + "epoch": 2.3732083792723264, + "grad_norm": 0.6625596677286694, + "learning_rate": 3.34375431628051e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0934302806854248, + "step": 4305, + "valid_targets_mean": 1711.5, + "valid_targets_min": 1122 + }, + { + "epoch": 2.375964718853363, + "grad_norm": 0.7701204530073166, + "learning_rate": 3.341716943088754e-05, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08988013863563538, + "step": 4310, + "valid_targets_mean": 1489.6, + "valid_targets_min": 850 + }, + { + "epoch": 2.378721058434399, + "grad_norm": 0.6858855671711186, + "learning_rate": 3.3396770350011134e-05, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07863776385784149, + "step": 4315, + "valid_targets_mean": 1572.4, + "valid_targets_min": 1318 + }, + { + "epoch": 2.3814773980154356, + "grad_norm": 0.8582902491969896, + "learning_rate": 3.3376345958715716e-05, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10398589074611664, + "step": 4320, + "valid_targets_mean": 1500.0, + "valid_targets_min": 1043 + }, + { + "epoch": 2.384233737596472, + "grad_norm": 0.6746621421487371, + "learning_rate": 3.335589629558894e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11182848364114761, + "step": 4325, + "valid_targets_mean": 1774.9, + "valid_targets_min": 920 + }, + { + "epoch": 2.3869900771775083, + "grad_norm": 0.659881692965294, + "learning_rate": 3.33354213992662e-05, + "loss": 0.1782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09315968304872513, + "step": 4330, + "valid_targets_mean": 1547.1, + "valid_targets_min": 847 + }, + { + "epoch": 2.3897464167585447, + "grad_norm": 0.6693494407503134, + "learning_rate": 3.3314921308430554e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09803806245326996, + "step": 4335, + "valid_targets_mean": 1769.4, + "valid_targets_min": 738 + }, + { + "epoch": 2.392502756339581, + "grad_norm": 0.7494741043458953, + "learning_rate": 3.32943960618127e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10205884277820587, + "step": 4340, + "valid_targets_mean": 1540.1, + "valid_targets_min": 594 + }, + { + "epoch": 2.3952590959206175, + "grad_norm": 0.6287092307830154, + "learning_rate": 3.32738456981908e-05, + "loss": 0.1848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09569642692804337, + "step": 4345, + "valid_targets_mean": 2228.6, + "valid_targets_min": 1393 + }, + { + "epoch": 2.398015435501654, + "grad_norm": 0.6038133037360167, + "learning_rate": 3.325327025639054e-05, + "loss": 0.2922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15292246639728546, + "step": 4350, + "valid_targets_mean": 3692.9, + "valid_targets_min": 770 + }, + { + "epoch": 2.4007717750826902, + "grad_norm": 0.5171589399439928, + "learning_rate": 3.3232669775284926e-05, + "loss": 0.281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1490662693977356, + "step": 4355, + "valid_targets_mean": 5237.4, + "valid_targets_min": 1799 + }, + { + "epoch": 2.4035281146637266, + "grad_norm": 0.5668616763886757, + "learning_rate": 3.321204429379429e-05, + "loss": 0.2728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14640048146247864, + "step": 4360, + "valid_targets_mean": 5888.0, + "valid_targets_min": 3015 + }, + { + "epoch": 2.406284454244763, + "grad_norm": 0.45446166035523816, + "learning_rate": 3.3191393850886225e-05, + "loss": 0.2705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1380573809146881, + "step": 4365, + "valid_targets_mean": 6029.1, + "valid_targets_min": 1681 + }, + { + "epoch": 2.4090407938257994, + "grad_norm": 0.5388708963900286, + "learning_rate": 3.317071848557544e-05, + "loss": 0.2788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11918790638446808, + "step": 4370, + "valid_targets_mean": 3299.6, + "valid_targets_min": 1374 + }, + { + "epoch": 2.4117971334068358, + "grad_norm": 0.5461549724542796, + "learning_rate": 3.315001823692376e-05, + "loss": 0.2814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13719627261161804, + "step": 4375, + "valid_targets_mean": 3858.8, + "valid_targets_min": 1951 + }, + { + "epoch": 2.414553472987872, + "grad_norm": 0.46288877993917416, + "learning_rate": 3.3129293144039995e-05, + "loss": 0.2687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13203312456607819, + "step": 4380, + "valid_targets_mean": 5250.8, + "valid_targets_min": 2042 + }, + { + "epoch": 2.4173098125689085, + "grad_norm": 0.47980382154300927, + "learning_rate": 3.310854324607993e-05, + "loss": 0.2746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12249109894037247, + "step": 4385, + "valid_targets_mean": 4937.4, + "valid_targets_min": 1831 + }, + { + "epoch": 2.420066152149945, + "grad_norm": 0.56513138979209, + "learning_rate": 3.308776858224619e-05, + "loss": 0.2834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14487086236476898, + "step": 4390, + "valid_targets_mean": 4049.4, + "valid_targets_min": 2077 + }, + { + "epoch": 2.4228224917309813, + "grad_norm": 0.5044598326602345, + "learning_rate": 3.3066969191788184e-05, + "loss": 0.2784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12669572234153748, + "step": 4395, + "valid_targets_mean": 4086.2, + "valid_targets_min": 2022 + }, + { + "epoch": 2.4255788313120177, + "grad_norm": 0.49915071080764417, + "learning_rate": 3.304614511400205e-05, + "loss": 0.2815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15158307552337646, + "step": 4400, + "valid_targets_mean": 4897.6, + "valid_targets_min": 1499 + }, + { + "epoch": 2.428335170893054, + "grad_norm": 0.4655200788909145, + "learning_rate": 3.3025296388230564e-05, + "loss": 0.2712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14626668393611908, + "step": 4405, + "valid_targets_mean": 4821.6, + "valid_targets_min": 1013 + }, + { + "epoch": 2.4310915104740904, + "grad_norm": 0.5140770139235108, + "learning_rate": 3.300442305386308e-05, + "loss": 0.2773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13385073840618134, + "step": 4410, + "valid_targets_mean": 4676.2, + "valid_targets_min": 1939 + }, + { + "epoch": 2.433847850055127, + "grad_norm": 0.5330693531731775, + "learning_rate": 3.2983525150335414e-05, + "loss": 0.2728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16179227828979492, + "step": 4415, + "valid_targets_mean": 4010.2, + "valid_targets_min": 1820 + }, + { + "epoch": 2.436604189636163, + "grad_norm": 0.5867758487020743, + "learning_rate": 3.2962602717129834e-05, + "loss": 0.2839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15667520463466644, + "step": 4420, + "valid_targets_mean": 3824.9, + "valid_targets_min": 1589 + }, + { + "epoch": 2.4393605292171996, + "grad_norm": 0.7259289966998842, + "learning_rate": 3.294165579377493e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12596291303634644, + "step": 4425, + "valid_targets_mean": 4046.5, + "valid_targets_min": 2668 + }, + { + "epoch": 2.442116868798236, + "grad_norm": 0.5655178148326797, + "learning_rate": 3.292068441984558e-05, + "loss": 0.2742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13360051810741425, + "step": 4430, + "valid_targets_mean": 3592.6, + "valid_targets_min": 1269 + }, + { + "epoch": 2.4448732083792724, + "grad_norm": 0.5986007223418688, + "learning_rate": 3.2899688634962846e-05, + "loss": 0.2919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14099165797233582, + "step": 4435, + "valid_targets_mean": 3423.4, + "valid_targets_min": 1695 + }, + { + "epoch": 2.4476295479603087, + "grad_norm": 0.5601984550455015, + "learning_rate": 3.287866847879389e-05, + "loss": 0.2837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13243792951107025, + "step": 4440, + "valid_targets_mean": 3522.2, + "valid_targets_min": 1128 + }, + { + "epoch": 2.450385887541345, + "grad_norm": 0.5692799976916502, + "learning_rate": 3.285762399105195e-05, + "loss": 0.2745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14252987504005432, + "step": 4445, + "valid_targets_mean": 3533.5, + "valid_targets_min": 1438 + }, + { + "epoch": 2.4531422271223815, + "grad_norm": 0.541196476897026, + "learning_rate": 3.283655521149623e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13198968768119812, + "step": 4450, + "valid_targets_mean": 3842.9, + "valid_targets_min": 1267 + }, + { + "epoch": 2.455898566703418, + "grad_norm": 0.5481393728055259, + "learning_rate": 3.281546217993181e-05, + "loss": 0.2672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11923009902238846, + "step": 4455, + "valid_targets_mean": 3671.9, + "valid_targets_min": 1026 + }, + { + "epoch": 2.4586549062844543, + "grad_norm": 0.6322023900518327, + "learning_rate": 3.27943449362096e-05, + "loss": 0.2579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1256844699382782, + "step": 4460, + "valid_targets_mean": 2866.2, + "valid_targets_min": 1403 + }, + { + "epoch": 2.4614112458654906, + "grad_norm": 0.5780936267571083, + "learning_rate": 3.277320352022624e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1536601185798645, + "step": 4465, + "valid_targets_mean": 3882.1, + "valid_targets_min": 1588 + }, + { + "epoch": 2.464167585446527, + "grad_norm": 0.6115464633401871, + "learning_rate": 3.275203797192406e-05, + "loss": 0.2796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12420637905597687, + "step": 4470, + "valid_targets_mean": 2827.6, + "valid_targets_min": 1276 + }, + { + "epoch": 2.4669239250275634, + "grad_norm": 0.46988748892006216, + "learning_rate": 3.2730848331290984e-05, + "loss": 0.2699, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11663773655891418, + "step": 4475, + "valid_targets_mean": 4232.4, + "valid_targets_min": 1631 + }, + { + "epoch": 2.4696802646086, + "grad_norm": 0.5363951505103595, + "learning_rate": 3.270963463836043e-05, + "loss": 0.2713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16379272937774658, + "step": 4480, + "valid_targets_mean": 5069.9, + "valid_targets_min": 3090 + }, + { + "epoch": 2.472436604189636, + "grad_norm": 0.5845283790053364, + "learning_rate": 3.268839693321128e-05, + "loss": 0.2615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1648246794939041, + "step": 4485, + "valid_targets_mean": 3732.4, + "valid_targets_min": 1802 + }, + { + "epoch": 2.4751929437706726, + "grad_norm": 0.616915331412931, + "learning_rate": 3.266713525596776e-05, + "loss": 0.2766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15771864354610443, + "step": 4490, + "valid_targets_mean": 3490.8, + "valid_targets_min": 1685 + }, + { + "epoch": 2.477949283351709, + "grad_norm": 0.5409078383170327, + "learning_rate": 3.2645849646799406e-05, + "loss": 0.26, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12113456428050995, + "step": 4495, + "valid_targets_mean": 3660.1, + "valid_targets_min": 1596 + }, + { + "epoch": 2.4807056229327453, + "grad_norm": 0.5464949761133283, + "learning_rate": 3.262454014592097e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1501331627368927, + "step": 4500, + "valid_targets_mean": 4306.4, + "valid_targets_min": 2251 + }, + { + "epoch": 2.4834619625137817, + "grad_norm": 0.6715668548656027, + "learning_rate": 3.260320679359232e-05, + "loss": 0.2675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1420055627822876, + "step": 4505, + "valid_targets_mean": 3107.1, + "valid_targets_min": 1458 + }, + { + "epoch": 2.486218302094818, + "grad_norm": 0.617218999381288, + "learning_rate": 3.258184963011842e-05, + "loss": 0.2745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13690271973609924, + "step": 4510, + "valid_targets_mean": 3664.6, + "valid_targets_min": 1162 + }, + { + "epoch": 2.4889746416758545, + "grad_norm": 0.6444646003633, + "learning_rate": 3.2560468695849174e-05, + "loss": 0.2646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12558194994926453, + "step": 4515, + "valid_targets_mean": 2834.2, + "valid_targets_min": 1480 + }, + { + "epoch": 2.491730981256891, + "grad_norm": 0.5920052003924651, + "learning_rate": 3.253906403117945e-05, + "loss": 0.2761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14179781079292297, + "step": 4520, + "valid_targets_mean": 3795.1, + "valid_targets_min": 1229 + }, + { + "epoch": 2.4944873208379272, + "grad_norm": 0.5447125928404647, + "learning_rate": 3.25176356765489e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1236436739563942, + "step": 4525, + "valid_targets_mean": 3559.6, + "valid_targets_min": 1276 + }, + { + "epoch": 2.4972436604189636, + "grad_norm": 0.5740625387536249, + "learning_rate": 3.249618367244197e-05, + "loss": 0.2668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12178811430931091, + "step": 4530, + "valid_targets_mean": 3479.6, + "valid_targets_min": 1357 + }, + { + "epoch": 2.5, + "grad_norm": 0.6330793528989025, + "learning_rate": 3.2474708059387775e-05, + "loss": 0.263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13628000020980835, + "step": 4535, + "valid_targets_mean": 2925.9, + "valid_targets_min": 924 + }, + { + "epoch": 2.5027563395810364, + "grad_norm": 0.6131241475713997, + "learning_rate": 3.245320887796001e-05, + "loss": 0.27, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14066705107688904, + "step": 4540, + "valid_targets_mean": 4030.5, + "valid_targets_min": 1674 + }, + { + "epoch": 2.5055126791620728, + "grad_norm": 0.5893019998965696, + "learning_rate": 3.2431686168776944e-05, + "loss": 0.2679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13394063711166382, + "step": 4545, + "valid_targets_mean": 3173.9, + "valid_targets_min": 1643 + }, + { + "epoch": 2.508269018743109, + "grad_norm": 0.5649991374171843, + "learning_rate": 3.241013997250126e-05, + "loss": 0.261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14290060102939606, + "step": 4550, + "valid_targets_mean": 4259.1, + "valid_targets_min": 1244 + }, + { + "epoch": 2.5110253583241455, + "grad_norm": 0.5720067895147269, + "learning_rate": 3.2388570329840025e-05, + "loss": 0.2769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12690326571464539, + "step": 4555, + "valid_targets_mean": 4065.8, + "valid_targets_min": 730 + }, + { + "epoch": 2.513781697905182, + "grad_norm": 0.6833903688725881, + "learning_rate": 3.236697728154461e-05, + "loss": 0.2659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11915001273155212, + "step": 4560, + "valid_targets_mean": 2945.9, + "valid_targets_min": 977 + }, + { + "epoch": 2.5165380374862183, + "grad_norm": 0.5674745685635874, + "learning_rate": 3.234536086841061e-05, + "loss": 0.2637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15082836151123047, + "step": 4565, + "valid_targets_mean": 5116.1, + "valid_targets_min": 2091 + }, + { + "epoch": 2.5192943770672547, + "grad_norm": 0.6052324113603591, + "learning_rate": 3.232372113127774e-05, + "loss": 0.2552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11858086287975311, + "step": 4570, + "valid_targets_mean": 3125.4, + "valid_targets_min": 1654 + }, + { + "epoch": 2.522050716648291, + "grad_norm": 0.5966094873588669, + "learning_rate": 3.230205811102979e-05, + "loss": 0.2773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13702893257141113, + "step": 4575, + "valid_targets_mean": 4081.6, + "valid_targets_min": 1432 + }, + { + "epoch": 2.5248070562293274, + "grad_norm": 0.5231823110840678, + "learning_rate": 3.228037184859456e-05, + "loss": 0.2582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12425928562879562, + "step": 4580, + "valid_targets_mean": 3868.6, + "valid_targets_min": 1261 + }, + { + "epoch": 2.527563395810364, + "grad_norm": 0.600016961508481, + "learning_rate": 3.2258662384943745e-05, + "loss": 0.2783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16004590690135956, + "step": 4585, + "valid_targets_mean": 4079.6, + "valid_targets_min": 2737 + }, + { + "epoch": 2.5303197353914, + "grad_norm": 0.5717247495307615, + "learning_rate": 3.2236929761092865e-05, + "loss": 0.2654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11346358060836792, + "step": 4590, + "valid_targets_mean": 3471.4, + "valid_targets_min": 1219 + }, + { + "epoch": 2.5330760749724366, + "grad_norm": 0.5945881001432054, + "learning_rate": 3.221517401810121e-05, + "loss": 0.2775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1457367092370987, + "step": 4595, + "valid_targets_mean": 4087.4, + "valid_targets_min": 2058 + }, + { + "epoch": 2.535832414553473, + "grad_norm": 0.6260953472928253, + "learning_rate": 3.219339519707173e-05, + "loss": 0.2606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12191076576709747, + "step": 4600, + "valid_targets_mean": 2736.6, + "valid_targets_min": 916 + }, + { + "epoch": 2.5385887541345094, + "grad_norm": 0.551667513055229, + "learning_rate": 3.217159333915101e-05, + "loss": 0.2751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1139068752527237, + "step": 4605, + "valid_targets_mean": 3695.1, + "valid_targets_min": 1302 + }, + { + "epoch": 2.5413450937155457, + "grad_norm": 0.5915628532903955, + "learning_rate": 3.214976848552913e-05, + "loss": 0.2596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16633102297782898, + "step": 4610, + "valid_targets_mean": 4246.1, + "valid_targets_min": 1500 + }, + { + "epoch": 2.544101433296582, + "grad_norm": 0.5324346425910389, + "learning_rate": 3.212792067743963e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14654788374900818, + "step": 4615, + "valid_targets_mean": 4837.6, + "valid_targets_min": 2803 + }, + { + "epoch": 2.5468577728776185, + "grad_norm": 0.5456017894719708, + "learning_rate": 3.21060499561594e-05, + "loss": 0.2468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1267225742340088, + "step": 4620, + "valid_targets_mean": 3349.0, + "valid_targets_min": 878 + }, + { + "epoch": 2.549614112458655, + "grad_norm": 0.540941170996631, + "learning_rate": 3.208415636300864e-05, + "loss": 0.274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11884927749633789, + "step": 4625, + "valid_targets_mean": 3753.9, + "valid_targets_min": 2662 + }, + { + "epoch": 2.5523704520396913, + "grad_norm": 0.5406464623406092, + "learning_rate": 3.2062239939350756e-05, + "loss": 0.2605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12344200909137726, + "step": 4630, + "valid_targets_mean": 3965.1, + "valid_targets_min": 1086 + }, + { + "epoch": 2.5551267916207276, + "grad_norm": 0.5265319307209029, + "learning_rate": 3.204030072659229e-05, + "loss": 0.2651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14444014430046082, + "step": 4635, + "valid_targets_mean": 5292.5, + "valid_targets_min": 1566 + }, + { + "epoch": 2.557883131201764, + "grad_norm": 0.788120370000757, + "learning_rate": 3.201833876618281e-05, + "loss": 0.2636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1548628807067871, + "step": 4640, + "valid_targets_mean": 4169.2, + "valid_targets_min": 2504 + }, + { + "epoch": 2.5606394707828004, + "grad_norm": 0.586897751499949, + "learning_rate": 3.199635409961491e-05, + "loss": 0.2716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16373735666275024, + "step": 4645, + "valid_targets_mean": 4600.9, + "valid_targets_min": 1548 + }, + { + "epoch": 2.563395810363837, + "grad_norm": 0.678554171902104, + "learning_rate": 3.197434676842404e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12485215067863464, + "step": 4650, + "valid_targets_mean": 3289.5, + "valid_targets_min": 1352 + }, + { + "epoch": 2.566152149944873, + "grad_norm": 0.5004671483321765, + "learning_rate": 3.19523168141885e-05, + "loss": 0.2585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13630211353302002, + "step": 4655, + "valid_targets_mean": 5279.6, + "valid_targets_min": 2339 + }, + { + "epoch": 2.5689084895259096, + "grad_norm": 0.55511293885383, + "learning_rate": 3.193026427852932e-05, + "loss": 0.2596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12733839452266693, + "step": 4660, + "valid_targets_mean": 4053.0, + "valid_targets_min": 2015 + }, + { + "epoch": 2.571664829106946, + "grad_norm": 0.5649307576352869, + "learning_rate": 3.190818920311018e-05, + "loss": 0.2541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15683680772781372, + "step": 4665, + "valid_targets_mean": 4836.5, + "valid_targets_min": 1731 + }, + { + "epoch": 2.5744211686879823, + "grad_norm": 0.6343475658899173, + "learning_rate": 3.1886091629637376e-05, + "loss": 0.2582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12634685635566711, + "step": 4670, + "valid_targets_mean": 3052.5, + "valid_targets_min": 1460 + }, + { + "epoch": 2.5771775082690187, + "grad_norm": 0.5565150806043635, + "learning_rate": 3.186397159985967e-05, + "loss": 0.2459, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1403944194316864, + "step": 4675, + "valid_targets_mean": 4621.9, + "valid_targets_min": 1282 + }, + { + "epoch": 2.579933847850055, + "grad_norm": 0.55961228100751, + "learning_rate": 3.1841829155568284e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10805290192365646, + "step": 4680, + "valid_targets_mean": 3373.4, + "valid_targets_min": 1421 + }, + { + "epoch": 2.5826901874310915, + "grad_norm": 0.5816464274694351, + "learning_rate": 3.181966433859676e-05, + "loss": 0.2558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15971043705940247, + "step": 4685, + "valid_targets_mean": 4483.5, + "valid_targets_min": 1764 + }, + { + "epoch": 2.585446527012128, + "grad_norm": 0.6337260375779412, + "learning_rate": 3.179747719082094e-05, + "loss": 0.2741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15532821416854858, + "step": 4690, + "valid_targets_mean": 3288.8, + "valid_targets_min": 1175 + }, + { + "epoch": 2.5882028665931642, + "grad_norm": 0.5008795483068977, + "learning_rate": 3.1775267754158824e-05, + "loss": 0.2637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1163165271282196, + "step": 4695, + "valid_targets_mean": 4656.6, + "valid_targets_min": 2546 + }, + { + "epoch": 2.5909592061742006, + "grad_norm": 0.6465628787800936, + "learning_rate": 3.1753036070570554e-05, + "loss": 0.2635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16101521253585815, + "step": 4700, + "valid_targets_mean": 3833.2, + "valid_targets_min": 1244 + }, + { + "epoch": 2.593715545755237, + "grad_norm": 0.5471742725760035, + "learning_rate": 3.1730782182058255e-05, + "loss": 0.2579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13099630177021027, + "step": 4705, + "valid_targets_mean": 4403.5, + "valid_targets_min": 1856 + }, + { + "epoch": 2.5964718853362734, + "grad_norm": 0.5639469711394286, + "learning_rate": 3.170850613066607e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14316251873970032, + "step": 4710, + "valid_targets_mean": 3836.6, + "valid_targets_min": 1948 + }, + { + "epoch": 2.5992282249173098, + "grad_norm": 0.5099023241399846, + "learning_rate": 3.168620795847994e-05, + "loss": 0.2494, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10834992676973343, + "step": 4715, + "valid_targets_mean": 3336.2, + "valid_targets_min": 1248 + }, + { + "epoch": 2.601984564498346, + "grad_norm": 0.484777881193385, + "learning_rate": 3.1663887707627666e-05, + "loss": 0.254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11737988889217377, + "step": 4720, + "valid_targets_mean": 4304.4, + "valid_targets_min": 1145 + }, + { + "epoch": 2.6047409040793825, + "grad_norm": 0.7025461341321813, + "learning_rate": 3.164154542027872e-05, + "loss": 0.2571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12096783518791199, + "step": 4725, + "valid_targets_mean": 3231.1, + "valid_targets_min": 1428 + }, + { + "epoch": 2.607497243660419, + "grad_norm": 0.563004275589238, + "learning_rate": 3.161918113864423e-05, + "loss": 0.2559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15011104941368103, + "step": 4730, + "valid_targets_mean": 4426.2, + "valid_targets_min": 1708 + }, + { + "epoch": 2.6102535832414553, + "grad_norm": 0.5835592325784728, + "learning_rate": 3.159679490497686e-05, + "loss": 0.2519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08973603695631027, + "step": 4735, + "valid_targets_mean": 2934.0, + "valid_targets_min": 1301 + }, + { + "epoch": 2.6130099228224917, + "grad_norm": 0.6161286659648397, + "learning_rate": 3.157438676157075e-05, + "loss": 0.2738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11196188628673553, + "step": 4740, + "valid_targets_mean": 3015.5, + "valid_targets_min": 1971 + }, + { + "epoch": 2.615766262403528, + "grad_norm": 0.6317403713727868, + "learning_rate": 3.1551956750761444e-05, + "loss": 0.266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16124911606311798, + "step": 4745, + "valid_targets_mean": 4887.2, + "valid_targets_min": 1054 + }, + { + "epoch": 2.6185226019845644, + "grad_norm": 0.5958437107453824, + "learning_rate": 3.15295049149258e-05, + "loss": 0.2658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14524579048156738, + "step": 4750, + "valid_targets_mean": 3527.2, + "valid_targets_min": 1612 + }, + { + "epoch": 2.621278941565601, + "grad_norm": 0.5606604578865675, + "learning_rate": 3.15070312964819e-05, + "loss": 0.2745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10916093736886978, + "step": 4755, + "valid_targets_mean": 3531.1, + "valid_targets_min": 721 + }, + { + "epoch": 2.624035281146637, + "grad_norm": 0.5472881016984428, + "learning_rate": 3.148453593788899e-05, + "loss": 0.2631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1299886405467987, + "step": 4760, + "valid_targets_mean": 3899.6, + "valid_targets_min": 1625 + }, + { + "epoch": 2.6267916207276736, + "grad_norm": 0.5928418992050666, + "learning_rate": 3.146201888164738e-05, + "loss": 0.272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13035714626312256, + "step": 4765, + "valid_targets_mean": 4175.0, + "valid_targets_min": 1381 + }, + { + "epoch": 2.62954796030871, + "grad_norm": 0.6311537935756215, + "learning_rate": 3.143948017029838e-05, + "loss": 0.2786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13156276941299438, + "step": 4770, + "valid_targets_mean": 3235.0, + "valid_targets_min": 998 + }, + { + "epoch": 2.6323042998897463, + "grad_norm": 0.589832861552052, + "learning_rate": 3.141691984642421e-05, + "loss": 0.2593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09890752285718918, + "step": 4775, + "valid_targets_mean": 2957.1, + "valid_targets_min": 1186 + }, + { + "epoch": 2.6350606394707827, + "grad_norm": 0.581641374034358, + "learning_rate": 3.139433795264791e-05, + "loss": 0.2637, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15560150146484375, + "step": 4780, + "valid_targets_mean": 3583.0, + "valid_targets_min": 1195 + }, + { + "epoch": 2.637816979051819, + "grad_norm": 0.6233194440014579, + "learning_rate": 3.137173453163332e-05, + "loss": 0.2629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12860727310180664, + "step": 4785, + "valid_targets_mean": 2696.1, + "valid_targets_min": 885 + }, + { + "epoch": 2.6405733186328555, + "grad_norm": 0.6091703770113168, + "learning_rate": 3.134910962608487e-05, + "loss": 0.2676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1277257204055786, + "step": 4790, + "valid_targets_mean": 3362.2, + "valid_targets_min": 814 + }, + { + "epoch": 2.643329658213892, + "grad_norm": 0.5849761596882086, + "learning_rate": 3.132646327874766e-05, + "loss": 0.2636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11339753121137619, + "step": 4795, + "valid_targets_mean": 3612.5, + "valid_targets_min": 2174 + }, + { + "epoch": 2.6460859977949283, + "grad_norm": 0.5049780402433115, + "learning_rate": 3.130379553240726e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14458245038986206, + "step": 4800, + "valid_targets_mean": 5472.4, + "valid_targets_min": 3025 + }, + { + "epoch": 2.6488423373759646, + "grad_norm": 0.6787730177624186, + "learning_rate": 3.1281106429889655e-05, + "loss": 0.2589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12500496208667755, + "step": 4805, + "valid_targets_mean": 3297.9, + "valid_targets_min": 1434 + }, + { + "epoch": 2.651598676957001, + "grad_norm": 0.6020441927164988, + "learning_rate": 3.125839601406123e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12655067443847656, + "step": 4810, + "valid_targets_mean": 3262.1, + "valid_targets_min": 1074 + }, + { + "epoch": 2.6543550165380374, + "grad_norm": 0.5830526247376141, + "learning_rate": 3.123566432782858e-05, + "loss": 0.2616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10224539786577225, + "step": 4815, + "valid_targets_mean": 3145.2, + "valid_targets_min": 980 + }, + { + "epoch": 2.657111356119074, + "grad_norm": 0.5589934119005219, + "learning_rate": 3.121291141413853e-05, + "loss": 0.2577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14394403994083405, + "step": 4820, + "valid_targets_mean": 4575.0, + "valid_targets_min": 2366 + }, + { + "epoch": 2.65986769570011, + "grad_norm": 0.5629357728318888, + "learning_rate": 3.1190137315977976e-05, + "loss": 0.264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13736607134342194, + "step": 4825, + "valid_targets_mean": 4999.9, + "valid_targets_min": 3139 + }, + { + "epoch": 2.6626240352811466, + "grad_norm": 0.6100645904027638, + "learning_rate": 3.116734207637386e-05, + "loss": 0.2524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1558312326669693, + "step": 4830, + "valid_targets_mean": 4034.4, + "valid_targets_min": 2138 + }, + { + "epoch": 2.665380374862183, + "grad_norm": 0.586640756195259, + "learning_rate": 3.114452573839306e-05, + "loss": 0.2569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12912359833717346, + "step": 4835, + "valid_targets_mean": 3610.9, + "valid_targets_min": 2014 + }, + { + "epoch": 2.6681367144432193, + "grad_norm": 0.6055511643911294, + "learning_rate": 3.112168834514231e-05, + "loss": 0.2589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14427797496318817, + "step": 4840, + "valid_targets_mean": 3795.6, + "valid_targets_min": 1084 + }, + { + "epoch": 2.6708930540242557, + "grad_norm": 0.5508958385952393, + "learning_rate": 3.1098829939768134e-05, + "loss": 0.247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1319529116153717, + "step": 4845, + "valid_targets_mean": 3684.4, + "valid_targets_min": 1618 + }, + { + "epoch": 2.673649393605292, + "grad_norm": 0.6306305690300144, + "learning_rate": 3.107595056545675e-05, + "loss": 0.2426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12024059891700745, + "step": 4850, + "valid_targets_mean": 3378.8, + "valid_targets_min": 1193 + }, + { + "epoch": 2.6764057331863285, + "grad_norm": 0.5245056714904476, + "learning_rate": 3.1053050265433975e-05, + "loss": 0.2507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13390839099884033, + "step": 4855, + "valid_targets_mean": 4732.5, + "valid_targets_min": 2092 + }, + { + "epoch": 2.679162072767365, + "grad_norm": 1.0689830061929217, + "learning_rate": 3.103012908296519e-05, + "loss": 0.2495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15070103108882904, + "step": 4860, + "valid_targets_mean": 4235.9, + "valid_targets_min": 2786 + }, + { + "epoch": 2.6819184123484012, + "grad_norm": 0.6234849673488102, + "learning_rate": 3.100718706135521e-05, + "loss": 0.2382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10436245799064636, + "step": 4865, + "valid_targets_mean": 4036.9, + "valid_targets_min": 1246 + }, + { + "epoch": 2.6846747519294376, + "grad_norm": 0.5466156662828254, + "learning_rate": 3.098422424394824e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10447341203689575, + "step": 4870, + "valid_targets_mean": 3434.8, + "valid_targets_min": 1484 + }, + { + "epoch": 2.687431091510474, + "grad_norm": 0.548267026292186, + "learning_rate": 3.096124067412774e-05, + "loss": 0.251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13372501730918884, + "step": 4875, + "valid_targets_mean": 4156.9, + "valid_targets_min": 1502 + }, + { + "epoch": 2.6901874310915104, + "grad_norm": 0.6779698081650369, + "learning_rate": 3.093823639531641e-05, + "loss": 0.2463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09966927766799927, + "step": 4880, + "valid_targets_mean": 3785.8, + "valid_targets_min": 1530 + }, + { + "epoch": 2.6929437706725468, + "grad_norm": 0.6147647884908257, + "learning_rate": 3.091521145097606e-05, + "loss": 0.2741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1413956582546234, + "step": 4885, + "valid_targets_mean": 3245.6, + "valid_targets_min": 1312 + }, + { + "epoch": 2.695700110253583, + "grad_norm": 0.6189998618898878, + "learning_rate": 3.0892165884607545e-05, + "loss": 0.2462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13669347763061523, + "step": 4890, + "valid_targets_mean": 5026.8, + "valid_targets_min": 2738 + }, + { + "epoch": 2.6984564498346195, + "grad_norm": 0.5830301463371578, + "learning_rate": 3.086909973975069e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13218913972377777, + "step": 4895, + "valid_targets_mean": 4030.1, + "valid_targets_min": 1759 + }, + { + "epoch": 2.701212789415656, + "grad_norm": 0.5547865666888152, + "learning_rate": 3.084601305998419e-05, + "loss": 0.2446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10884058475494385, + "step": 4900, + "valid_targets_mean": 3422.8, + "valid_targets_min": 2155 + }, + { + "epoch": 2.7039691289966923, + "grad_norm": 0.6601373078657142, + "learning_rate": 3.082290588892553e-05, + "loss": 0.2557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12264490872621536, + "step": 4905, + "valid_targets_mean": 3223.8, + "valid_targets_min": 2048 + }, + { + "epoch": 2.7067254685777287, + "grad_norm": 0.6329359557653016, + "learning_rate": 3.079977827023092e-05, + "loss": 0.2544, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10626572370529175, + "step": 4910, + "valid_targets_mean": 2658.6, + "valid_targets_min": 1233 + }, + { + "epoch": 2.709481808158765, + "grad_norm": 0.5486270264388196, + "learning_rate": 3.077663024759521e-05, + "loss": 0.2558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11485175788402557, + "step": 4915, + "valid_targets_mean": 3698.2, + "valid_targets_min": 1383 + }, + { + "epoch": 2.7122381477398014, + "grad_norm": 0.546072480281158, + "learning_rate": 3.075346186475178e-05, + "loss": 0.2549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12002241611480713, + "step": 4920, + "valid_targets_mean": 3766.5, + "valid_targets_min": 1800 + }, + { + "epoch": 2.714994487320838, + "grad_norm": 0.6190163107275517, + "learning_rate": 3.0730273165472486e-05, + "loss": 0.2558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12023594975471497, + "step": 4925, + "valid_targets_mean": 3114.6, + "valid_targets_min": 2327 + }, + { + "epoch": 2.717750826901874, + "grad_norm": 0.5912903813553242, + "learning_rate": 3.070706419356756e-05, + "loss": 0.2525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12503209710121155, + "step": 4930, + "valid_targets_mean": 3558.5, + "valid_targets_min": 1214 + }, + { + "epoch": 2.7205071664829106, + "grad_norm": 0.5520009584121827, + "learning_rate": 3.0683834992885535e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07634373754262924, + "step": 4935, + "valid_targets_mean": 2729.0, + "valid_targets_min": 543 + }, + { + "epoch": 2.723263506063947, + "grad_norm": 0.3640345900455235, + "learning_rate": 3.066058560731319e-05, + "loss": 0.1055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04107128456234932, + "step": 4940, + "valid_targets_mean": 3641.4, + "valid_targets_min": 2731 + }, + { + "epoch": 2.7260198456449833, + "grad_norm": 0.37989861538904884, + "learning_rate": 3.0637316080775396e-05, + "loss": 0.1003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03924212232232094, + "step": 4945, + "valid_targets_mean": 3748.1, + "valid_targets_min": 2927 + }, + { + "epoch": 2.7287761852260197, + "grad_norm": 0.4056159346875625, + "learning_rate": 3.0614026457235114e-05, + "loss": 0.1109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06826483458280563, + "step": 4950, + "valid_targets_mean": 3764.5, + "valid_targets_min": 2952 + }, + { + "epoch": 2.731532524807056, + "grad_norm": 0.4435564519986132, + "learning_rate": 3.059071678069324e-05, + "loss": 0.0957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0553722083568573, + "step": 4955, + "valid_targets_mean": 2691.2, + "valid_targets_min": 560 + }, + { + "epoch": 2.7342888643880925, + "grad_norm": 0.38937626483149435, + "learning_rate": 3.0567387095188586e-05, + "loss": 0.1022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04897605627775192, + "step": 4960, + "valid_targets_mean": 2852.8, + "valid_targets_min": 1002 + }, + { + "epoch": 2.737045203969129, + "grad_norm": 1.1672461281403435, + "learning_rate": 3.054403744479774e-05, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0984792560338974, + "step": 4965, + "valid_targets_mean": 1797.6, + "valid_targets_min": 924 + }, + { + "epoch": 2.7398015435501653, + "grad_norm": 0.38599389361591857, + "learning_rate": 3.0520667873635044e-05, + "loss": 0.101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04464899003505707, + "step": 4970, + "valid_targets_mean": 3289.8, + "valid_targets_min": 1037 + }, + { + "epoch": 2.7425578831312016, + "grad_norm": 0.5512739446980021, + "learning_rate": 3.0497278425852445e-05, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10020140558481216, + "step": 4975, + "valid_targets_mean": 2096.6, + "valid_targets_min": 883 + }, + { + "epoch": 2.745314222712238, + "grad_norm": 0.23800594476839598, + "learning_rate": 3.0473869145639453e-05, + "loss": 0.1225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031028637662529945, + "step": 4980, + "valid_targets_mean": 5168.5, + "valid_targets_min": 3488 + }, + { + "epoch": 2.7480705622932744, + "grad_norm": 0.3390202138850792, + "learning_rate": 3.0450440077223037e-05, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03113502636551857, + "step": 4985, + "valid_targets_mean": 3117.1, + "valid_targets_min": 764 + }, + { + "epoch": 2.750826901874311, + "grad_norm": 0.3661296600730058, + "learning_rate": 3.0426991264867586e-05, + "loss": 0.0986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042246147990226746, + "step": 4990, + "valid_targets_mean": 3702.8, + "valid_targets_min": 773 + }, + { + "epoch": 2.753583241455347, + "grad_norm": 0.44339318840406905, + "learning_rate": 3.0403522752874754e-05, + "loss": 0.0983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04989343136548996, + "step": 4995, + "valid_targets_mean": 2504.1, + "valid_targets_min": 492 + }, + { + "epoch": 2.7563395810363835, + "grad_norm": 0.386468436592715, + "learning_rate": 3.0380034585583425e-05, + "loss": 0.1422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04046399146318436, + "step": 5000, + "valid_targets_mean": 4191.1, + "valid_targets_min": 1961 + }, + { + "epoch": 2.75909592061742, + "grad_norm": 0.28390943322026696, + "learning_rate": 3.0356526807369624e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04103773087263107, + "step": 5005, + "valid_targets_mean": 3909.8, + "valid_targets_min": 3691 + }, + { + "epoch": 2.7618522601984563, + "grad_norm": 0.40635176243989063, + "learning_rate": 3.0332999462646414e-05, + "loss": 0.1084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05054667592048645, + "step": 5010, + "valid_targets_mean": 2946.5, + "valid_targets_min": 841 + }, + { + "epoch": 2.7646085997794927, + "grad_norm": 0.33092789973695735, + "learning_rate": 3.030945259586385e-05, + "loss": 0.0866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04302132502198219, + "step": 5015, + "valid_targets_mean": 3111.0, + "valid_targets_min": 863 + }, + { + "epoch": 2.767364939360529, + "grad_norm": 0.5811159992744044, + "learning_rate": 3.0285886251508842e-05, + "loss": 0.1064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05460406094789505, + "step": 5020, + "valid_targets_mean": 1650.6, + "valid_targets_min": 515 + }, + { + "epoch": 2.7701212789415655, + "grad_norm": 0.46164826884899124, + "learning_rate": 3.026230047410511e-05, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04569830745458603, + "step": 5025, + "valid_targets_mean": 2142.4, + "valid_targets_min": 589 + }, + { + "epoch": 2.772877618522602, + "grad_norm": 0.405269688140392, + "learning_rate": 3.0238695308213086e-05, + "loss": 0.1138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058004721999168396, + "step": 5030, + "valid_targets_mean": 3472.5, + "valid_targets_min": 2918 + }, + { + "epoch": 2.775633958103638, + "grad_norm": 0.39807671376940884, + "learning_rate": 3.0215070798429837e-05, + "loss": 0.116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0701775848865509, + "step": 5035, + "valid_targets_mean": 3046.4, + "valid_targets_min": 1341 + }, + { + "epoch": 2.7783902976846746, + "grad_norm": 0.3859183370496809, + "learning_rate": 3.0191426989388985e-05, + "loss": 0.1488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05624433979392052, + "step": 5040, + "valid_targets_mean": 3619.9, + "valid_targets_min": 2422 + }, + { + "epoch": 2.781146637265711, + "grad_norm": 0.37078235035499035, + "learning_rate": 3.0167763925760592e-05, + "loss": 0.0941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05189322680234909, + "step": 5045, + "valid_targets_mean": 3250.0, + "valid_targets_min": 669 + }, + { + "epoch": 2.7839029768467474, + "grad_norm": 0.6568799695665406, + "learning_rate": 3.0144081652251107e-05, + "loss": 0.1152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07461962848901749, + "step": 5050, + "valid_targets_mean": 1616.4, + "valid_targets_min": 702 + }, + { + "epoch": 2.7866593164277838, + "grad_norm": 0.687126583738047, + "learning_rate": 3.012038021360328e-05, + "loss": 0.1382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07858532667160034, + "step": 5055, + "valid_targets_mean": 1299.4, + "valid_targets_min": 323 + }, + { + "epoch": 2.78941565600882, + "grad_norm": 0.3650857890446623, + "learning_rate": 3.0096659654596064e-05, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05098271369934082, + "step": 5060, + "valid_targets_mean": 3785.8, + "valid_targets_min": 2827 + }, + { + "epoch": 2.7921719955898565, + "grad_norm": 0.4998569467108779, + "learning_rate": 3.0072920020044533e-05, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14985094964504242, + "step": 5065, + "valid_targets_mean": 3747.5, + "valid_targets_min": 3111 + }, + { + "epoch": 2.794928335170893, + "grad_norm": 0.39739527201441255, + "learning_rate": 3.00491613547998e-05, + "loss": 0.2575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0474465973675251, + "step": 5070, + "valid_targets_mean": 3522.1, + "valid_targets_min": 1932 + }, + { + "epoch": 2.7976846747519293, + "grad_norm": 0.5081821108270665, + "learning_rate": 3.002538370374894e-05, + "loss": 0.1192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0633089542388916, + "step": 5075, + "valid_targets_mean": 1954.6, + "valid_targets_min": 820 + }, + { + "epoch": 2.8004410143329657, + "grad_norm": 0.47480150848986713, + "learning_rate": 3.000158711181489e-05, + "loss": 0.1135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07328057289123535, + "step": 5080, + "valid_targets_mean": 2913.0, + "valid_targets_min": 965 + }, + { + "epoch": 2.803197353914002, + "grad_norm": 0.34331981367689635, + "learning_rate": 2.997777162395638e-05, + "loss": 0.1999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04325404390692711, + "step": 5085, + "valid_targets_mean": 4515.6, + "valid_targets_min": 2262 + }, + { + "epoch": 2.8059536934950384, + "grad_norm": 0.381459682822798, + "learning_rate": 2.9953937285167827e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07798004895448685, + "step": 5090, + "valid_targets_mean": 4605.8, + "valid_targets_min": 3872 + }, + { + "epoch": 2.808710033076075, + "grad_norm": 0.36373696378777826, + "learning_rate": 2.993008414047928e-05, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057075560092926025, + "step": 5095, + "valid_targets_mean": 3741.5, + "valid_targets_min": 513 + }, + { + "epoch": 2.811466372657111, + "grad_norm": 0.32273732125388893, + "learning_rate": 2.9906212234956302e-05, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03879999369382858, + "step": 5100, + "valid_targets_mean": 4117.2, + "valid_targets_min": 3262 + }, + { + "epoch": 2.8142227122381476, + "grad_norm": 0.32385108686663855, + "learning_rate": 2.9882321613699905e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047317683696746826, + "step": 5105, + "valid_targets_mean": 4868.8, + "valid_targets_min": 3361 + }, + { + "epoch": 2.8169790518191844, + "grad_norm": 0.36761506452524756, + "learning_rate": 2.9858412321846467e-05, + "loss": 0.0951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054517850279808044, + "step": 5110, + "valid_targets_mean": 3249.4, + "valid_targets_min": 1873 + }, + { + "epoch": 2.8197353914002203, + "grad_norm": 0.43066415988936546, + "learning_rate": 2.983448440456764e-05, + "loss": 0.0994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05636841803789139, + "step": 5115, + "valid_targets_mean": 2693.6, + "valid_targets_min": 2135 + }, + { + "epoch": 2.822491730981257, + "grad_norm": 0.3956846436526542, + "learning_rate": 2.9810537907070256e-05, + "loss": 0.0849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03473146632313728, + "step": 5120, + "valid_targets_mean": 2199.6, + "valid_targets_min": 597 + }, + { + "epoch": 2.825248070562293, + "grad_norm": 0.5431069971220153, + "learning_rate": 2.9786572874596254e-05, + "loss": 0.1543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0748295783996582, + "step": 5125, + "valid_targets_mean": 2131.1, + "valid_targets_min": 707 + }, + { + "epoch": 2.82800441014333, + "grad_norm": 0.35717472408188405, + "learning_rate": 2.976258935242259e-05, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036186207085847855, + "step": 5130, + "valid_targets_mean": 2893.8, + "valid_targets_min": 472 + }, + { + "epoch": 2.830760749724366, + "grad_norm": 0.42598012345771163, + "learning_rate": 2.9738587385861164e-05, + "loss": 0.092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045188747346401215, + "step": 5135, + "valid_targets_mean": 1483.1, + "valid_targets_min": 616 + }, + { + "epoch": 2.8335170893054027, + "grad_norm": 0.34597160842940883, + "learning_rate": 2.9714567020258712e-05, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042043812572956085, + "step": 5140, + "valid_targets_mean": 4241.9, + "valid_targets_min": 3128 + }, + { + "epoch": 2.8362734288864386, + "grad_norm": 0.4141295335365201, + "learning_rate": 2.9690528300996725e-05, + "loss": 0.1044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10820125043392181, + "step": 5145, + "valid_targets_mean": 2705.2, + "valid_targets_min": 795 + }, + { + "epoch": 2.8390297684674755, + "grad_norm": 0.31566044291046685, + "learning_rate": 2.9666471273491382e-05, + "loss": 0.1272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0464947372674942, + "step": 5150, + "valid_targets_mean": 3195.1, + "valid_targets_min": 729 + }, + { + "epoch": 2.8417861080485114, + "grad_norm": 0.5162767539605141, + "learning_rate": 2.9642395983193452e-05, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05300052464008331, + "step": 5155, + "valid_targets_mean": 1339.2, + "valid_targets_min": 688 + }, + { + "epoch": 2.844542447629548, + "grad_norm": 0.6559904073137092, + "learning_rate": 2.961830247558819e-05, + "loss": 0.1129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08550924062728882, + "step": 5160, + "valid_targets_mean": 1593.9, + "valid_targets_min": 664 + }, + { + "epoch": 2.847298787210584, + "grad_norm": 0.4813314042650658, + "learning_rate": 2.9594190796195295e-05, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05814025551080704, + "step": 5165, + "valid_targets_mean": 2042.6, + "valid_targets_min": 831 + }, + { + "epoch": 2.850055126791621, + "grad_norm": 0.35396934301585703, + "learning_rate": 2.957006099056878e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05300702154636383, + "step": 5170, + "valid_targets_mean": 3793.5, + "valid_targets_min": 2440 + }, + { + "epoch": 2.852811466372657, + "grad_norm": 0.37187670150277297, + "learning_rate": 2.9545913104296898e-05, + "loss": 0.1052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045011695474386215, + "step": 5175, + "valid_targets_mean": 3369.5, + "valid_targets_min": 1353 + }, + { + "epoch": 2.8555678059536937, + "grad_norm": 0.40549442051425133, + "learning_rate": 2.9521747183002086e-05, + "loss": 0.1234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0399019792675972, + "step": 5180, + "valid_targets_mean": 2820.6, + "valid_targets_min": 755 + }, + { + "epoch": 2.8583241455347297, + "grad_norm": 0.4301447055168382, + "learning_rate": 2.9497563272340834e-05, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047704875469207764, + "step": 5185, + "valid_targets_mean": 1759.4, + "valid_targets_min": 485 + }, + { + "epoch": 2.8610804851157665, + "grad_norm": 0.4193131589928958, + "learning_rate": 2.947336141800363e-05, + "loss": 0.0966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040927454829216, + "step": 5190, + "valid_targets_mean": 3318.9, + "valid_targets_min": 810 + }, + { + "epoch": 2.8638368246968025, + "grad_norm": 0.2587709928581089, + "learning_rate": 2.9449141665714846e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03734324499964714, + "step": 5195, + "valid_targets_mean": 5552.5, + "valid_targets_min": 5097 + }, + { + "epoch": 2.8665931642778393, + "grad_norm": 0.4253693397423741, + "learning_rate": 2.9424904061232687e-05, + "loss": 0.098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06116164103150368, + "step": 5200, + "valid_targets_mean": 3634.5, + "valid_targets_min": 884 + }, + { + "epoch": 2.869349503858875, + "grad_norm": 0.3478461006757537, + "learning_rate": 2.9400648650349094e-05, + "loss": 0.0914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044234372675418854, + "step": 5205, + "valid_targets_mean": 3571.2, + "valid_targets_min": 941 + }, + { + "epoch": 2.872105843439912, + "grad_norm": 0.448088188051257, + "learning_rate": 2.9376375478889613e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09306301176548004, + "step": 5210, + "valid_targets_mean": 2844.1, + "valid_targets_min": 808 + }, + { + "epoch": 2.874862183020948, + "grad_norm": 0.42252048997406416, + "learning_rate": 2.9352084592713378e-05, + "loss": 0.1031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0576554536819458, + "step": 5215, + "valid_targets_mean": 3228.9, + "valid_targets_min": 1814 + }, + { + "epoch": 2.877618522601985, + "grad_norm": 0.401161304764465, + "learning_rate": 2.9327776037712978e-05, + "loss": 0.11, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06287418305873871, + "step": 5220, + "valid_targets_mean": 3565.2, + "valid_targets_min": 2206 + }, + { + "epoch": 2.8803748621830207, + "grad_norm": 0.32837948855116306, + "learning_rate": 2.9303449859814385e-05, + "loss": 0.0915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042896874248981476, + "step": 5225, + "valid_targets_mean": 3906.1, + "valid_targets_min": 2943 + }, + { + "epoch": 2.8831312017640576, + "grad_norm": 0.4426399460506753, + "learning_rate": 2.927910610497687e-05, + "loss": 0.101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05767711251974106, + "step": 5230, + "valid_targets_mean": 2342.0, + "valid_targets_min": 885 + }, + { + "epoch": 2.8858875413450935, + "grad_norm": 0.3382191346485444, + "learning_rate": 2.925474481919291e-05, + "loss": 0.1048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04566001892089844, + "step": 5235, + "valid_targets_mean": 3424.8, + "valid_targets_min": 2309 + }, + { + "epoch": 2.8886438809261303, + "grad_norm": 0.35791504500881, + "learning_rate": 2.9230366048488097e-05, + "loss": 0.0827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043947815895080566, + "step": 5240, + "valid_targets_mean": 3467.0, + "valid_targets_min": 1034 + }, + { + "epoch": 2.8914002205071663, + "grad_norm": 0.35604338096082155, + "learning_rate": 2.9205969838921065e-05, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04667047783732414, + "step": 5245, + "valid_targets_mean": 2966.1, + "valid_targets_min": 531 + }, + { + "epoch": 2.894156560088203, + "grad_norm": 0.3249022525129365, + "learning_rate": 2.9181556236583385e-05, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043173838406801224, + "step": 5250, + "valid_targets_mean": 3222.6, + "valid_targets_min": 1003 + }, + { + "epoch": 2.896912899669239, + "grad_norm": 0.46702993825503897, + "learning_rate": 2.9157125287599508e-05, + "loss": 0.1234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03598783165216446, + "step": 5255, + "valid_targets_mean": 1531.1, + "valid_targets_min": 597 + }, + { + "epoch": 2.899669239250276, + "grad_norm": 0.4426919442037349, + "learning_rate": 2.9132677038126637e-05, + "loss": 0.1015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03478704392910004, + "step": 5260, + "valid_targets_mean": 1614.5, + "valid_targets_min": 665 + }, + { + "epoch": 2.902425578831312, + "grad_norm": 0.5832302559804213, + "learning_rate": 2.9108211534354672e-05, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05227617174386978, + "step": 5265, + "valid_targets_mean": 1205.4, + "valid_targets_min": 853 + }, + { + "epoch": 2.9051819184123486, + "grad_norm": 0.5104398026403176, + "learning_rate": 2.9083728822506103e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07641629874706268, + "step": 5270, + "valid_targets_mean": 1920.4, + "valid_targets_min": 897 + }, + { + "epoch": 2.9079382579933846, + "grad_norm": 0.5104466111670513, + "learning_rate": 2.9059228948835934e-05, + "loss": 0.1048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05660971999168396, + "step": 5275, + "valid_targets_mean": 2268.4, + "valid_targets_min": 740 + }, + { + "epoch": 2.9106945975744214, + "grad_norm": 0.3181320962410245, + "learning_rate": 2.9034711959631607e-05, + "loss": 0.0958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035852380096912384, + "step": 5280, + "valid_targets_mean": 3367.2, + "valid_targets_min": 2818 + }, + { + "epoch": 2.9134509371554573, + "grad_norm": 0.3610139868930466, + "learning_rate": 2.901017790121288e-05, + "loss": 0.0949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04274870082736015, + "step": 5285, + "valid_targets_mean": 3802.4, + "valid_targets_min": 2882 + }, + { + "epoch": 2.916207276736494, + "grad_norm": 0.5200744071970275, + "learning_rate": 2.898562681993177e-05, + "loss": 0.1255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056672580540180206, + "step": 5290, + "valid_targets_mean": 1496.0, + "valid_targets_min": 446 + }, + { + "epoch": 2.91896361631753, + "grad_norm": 0.36848532983107635, + "learning_rate": 2.8961058762172448e-05, + "loss": 0.0874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03905592858791351, + "step": 5295, + "valid_targets_mean": 2669.4, + "valid_targets_min": 681 + }, + { + "epoch": 2.921719955898567, + "grad_norm": 0.414344245644999, + "learning_rate": 2.8936473774351167e-05, + "loss": 0.0965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06977767497301102, + "step": 5300, + "valid_targets_mean": 3088.8, + "valid_targets_min": 799 + }, + { + "epoch": 2.924476295479603, + "grad_norm": 0.3599502269272189, + "learning_rate": 2.891187190291617e-05, + "loss": 0.096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04931969195604324, + "step": 5305, + "valid_targets_mean": 4061.5, + "valid_targets_min": 2470 + }, + { + "epoch": 2.9272326350606397, + "grad_norm": 0.3074581041516084, + "learning_rate": 2.888725319434758e-05, + "loss": 0.083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034211549907922745, + "step": 5310, + "valid_targets_mean": 3580.2, + "valid_targets_min": 861 + }, + { + "epoch": 2.9299889746416756, + "grad_norm": 0.38248383041814943, + "learning_rate": 2.886261769515734e-05, + "loss": 0.0915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06155550852417946, + "step": 5315, + "valid_targets_mean": 3982.2, + "valid_targets_min": 3024 + }, + { + "epoch": 2.9327453142227125, + "grad_norm": 0.32839870102574, + "learning_rate": 2.883796545188912e-05, + "loss": 0.099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060040079057216644, + "step": 5320, + "valid_targets_mean": 3709.5, + "valid_targets_min": 2069 + }, + { + "epoch": 2.9355016538037484, + "grad_norm": 0.3782007787916471, + "learning_rate": 2.8813296511118234e-05, + "loss": 0.1232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058224909007549286, + "step": 5325, + "valid_targets_mean": 3669.6, + "valid_targets_min": 2652 + }, + { + "epoch": 2.938257993384785, + "grad_norm": 0.39768188670822924, + "learning_rate": 2.8788610919451504e-05, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12294390797615051, + "step": 5330, + "valid_targets_mean": 3374.2, + "valid_targets_min": 958 + }, + { + "epoch": 2.941014332965821, + "grad_norm": 0.4294225012091141, + "learning_rate": 2.8763908723527256e-05, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044733114540576935, + "step": 5335, + "valid_targets_mean": 1488.0, + "valid_targets_min": 815 + }, + { + "epoch": 2.943770672546858, + "grad_norm": 0.3842503669948209, + "learning_rate": 2.8739189970015152e-05, + "loss": 0.1211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0487077534198761, + "step": 5340, + "valid_targets_mean": 3013.9, + "valid_targets_min": 723 + }, + { + "epoch": 2.946527012127894, + "grad_norm": 0.3804080110232099, + "learning_rate": 2.8714454705616157e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05617143213748932, + "step": 5345, + "valid_targets_mean": 3831.6, + "valid_targets_min": 3388 + }, + { + "epoch": 2.9492833517089307, + "grad_norm": 0.40763295119009785, + "learning_rate": 2.868970297706242e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04477081075310707, + "step": 5350, + "valid_targets_mean": 3597.4, + "valid_targets_min": 743 + }, + { + "epoch": 2.9520396912899667, + "grad_norm": 0.5453037953313432, + "learning_rate": 2.8664934831117205e-05, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10120311379432678, + "step": 5355, + "valid_targets_mean": 2573.4, + "valid_targets_min": 842 + }, + { + "epoch": 2.9547960308710035, + "grad_norm": 0.40043520750903233, + "learning_rate": 2.864015031457478e-05, + "loss": 0.1448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0959658995270729, + "step": 5360, + "valid_targets_mean": 2849.6, + "valid_targets_min": 1046 + }, + { + "epoch": 2.9575523704520394, + "grad_norm": 0.4299170001393988, + "learning_rate": 2.8615349474260345e-05, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04007365554571152, + "step": 5365, + "valid_targets_mean": 2458.2, + "valid_targets_min": 663 + }, + { + "epoch": 2.9603087100330763, + "grad_norm": 0.4911326900386594, + "learning_rate": 2.8590532357029955e-05, + "loss": 0.0993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05493290722370148, + "step": 5370, + "valid_targets_mean": 1659.5, + "valid_targets_min": 716 + }, + { + "epoch": 2.963065049614112, + "grad_norm": 0.40160141263588, + "learning_rate": 2.8565699009770393e-05, + "loss": 0.0955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05097433179616928, + "step": 5375, + "valid_targets_mean": 2279.1, + "valid_targets_min": 542 + }, + { + "epoch": 2.965821389195149, + "grad_norm": 0.39462546803693294, + "learning_rate": 2.8540849479399134e-05, + "loss": 0.0949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0521036721765995, + "step": 5380, + "valid_targets_mean": 3052.0, + "valid_targets_min": 1020 + }, + { + "epoch": 2.968577728776185, + "grad_norm": 0.3783390163125603, + "learning_rate": 2.8515983812864197e-05, + "loss": 0.0854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03995782136917114, + "step": 5385, + "valid_targets_mean": 1796.4, + "valid_targets_min": 635 + }, + { + "epoch": 2.971334068357222, + "grad_norm": 0.36406233328797316, + "learning_rate": 2.849110205714411e-05, + "loss": 0.1187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03996380418539047, + "step": 5390, + "valid_targets_mean": 2374.2, + "valid_targets_min": 794 + }, + { + "epoch": 2.9740904079382577, + "grad_norm": 0.4342625655777502, + "learning_rate": 2.8466204259247786e-05, + "loss": 0.0866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03301440179347992, + "step": 5395, + "valid_targets_mean": 1387.4, + "valid_targets_min": 704 + }, + { + "epoch": 2.9768467475192946, + "grad_norm": 0.369989436651256, + "learning_rate": 2.8441290466214453e-05, + "loss": 0.0961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035772405564785004, + "step": 5400, + "valid_targets_mean": 2657.4, + "valid_targets_min": 530 + }, + { + "epoch": 2.9796030871003305, + "grad_norm": 0.4030701376398804, + "learning_rate": 2.841636072511355e-05, + "loss": 0.1033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05784330889582634, + "step": 5405, + "valid_targets_mean": 3537.4, + "valid_targets_min": 2890 + }, + { + "epoch": 2.9823594266813673, + "grad_norm": 0.3540078370260115, + "learning_rate": 2.839141508304466e-05, + "loss": 0.0934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04086689278483391, + "step": 5410, + "valid_targets_mean": 3928.1, + "valid_targets_min": 516 + }, + { + "epoch": 2.9851157662624033, + "grad_norm": 0.2859847174207182, + "learning_rate": 2.8366453587137393e-05, + "loss": 0.0851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03838648647069931, + "step": 5415, + "valid_targets_mean": 3519.6, + "valid_targets_min": 1029 + }, + { + "epoch": 2.98787210584344, + "grad_norm": 0.2941599559877342, + "learning_rate": 2.8341476284551324e-05, + "loss": 0.0904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0462542399764061, + "step": 5420, + "valid_targets_mean": 3499.5, + "valid_targets_min": 605 + }, + { + "epoch": 2.990628445424476, + "grad_norm": 0.500277779118154, + "learning_rate": 2.8316483222475883e-05, + "loss": 0.1698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16218480467796326, + "step": 5425, + "valid_targets_mean": 2808.1, + "valid_targets_min": 1907 + }, + { + "epoch": 2.993384785005513, + "grad_norm": 0.3623718030741676, + "learning_rate": 2.829147444813028e-05, + "loss": 0.1152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045540012419223785, + "step": 5430, + "valid_targets_mean": 2939.9, + "valid_targets_min": 1235 + }, + { + "epoch": 2.996141124586549, + "grad_norm": 0.38830448586692146, + "learning_rate": 2.82664500087634e-05, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050593383610248566, + "step": 5435, + "valid_targets_mean": 2889.4, + "valid_targets_min": 717 + }, + { + "epoch": 2.9988974641675856, + "grad_norm": 0.38151525649832274, + "learning_rate": 2.8241409951653732e-05, + "loss": 0.0883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049526944756507874, + "step": 5440, + "valid_targets_mean": 2264.9, + "valid_targets_min": 1199 + }, + { + "epoch": 3.001653803748622, + "grad_norm": 0.5033979776168923, + "learning_rate": 2.8216354324109284e-05, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10315126180648804, + "step": 5445, + "valid_targets_mean": 5435.1, + "valid_targets_min": 3516 + }, + { + "epoch": 3.0044101433296584, + "grad_norm": 0.38074038527280785, + "learning_rate": 2.819128317346745e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10091400146484375, + "step": 5450, + "valid_targets_mean": 7654.2, + "valid_targets_min": 6354 + }, + { + "epoch": 3.0071664829106948, + "grad_norm": 0.3674300411298658, + "learning_rate": 2.8166196547094986e-05, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11628516018390656, + "step": 5455, + "valid_targets_mean": 7341.6, + "valid_targets_min": 5754 + }, + { + "epoch": 3.009922822491731, + "grad_norm": 0.3489454138483058, + "learning_rate": 2.8141094492387853e-05, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08703164756298065, + "step": 5460, + "valid_targets_mean": 7666.2, + "valid_targets_min": 5666 + }, + { + "epoch": 3.0126791620727675, + "grad_norm": 0.3838936773145169, + "learning_rate": 2.811597705677118e-05, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1008518785238266, + "step": 5465, + "valid_targets_mean": 7469.0, + "valid_targets_min": 5646 + }, + { + "epoch": 3.015435501653804, + "grad_norm": 0.4078061216525404, + "learning_rate": 2.8090844287699165e-05, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1155182272195816, + "step": 5470, + "valid_targets_mean": 7046.9, + "valid_targets_min": 5571 + }, + { + "epoch": 3.0181918412348403, + "grad_norm": 0.3931310787754795, + "learning_rate": 2.806569623265495e-05, + "loss": 0.2131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11119961738586426, + "step": 5475, + "valid_targets_mean": 7827.6, + "valid_targets_min": 5405 + }, + { + "epoch": 3.0209481808158767, + "grad_norm": 0.41136427398275194, + "learning_rate": 2.8040532939150577e-05, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10343065112829208, + "step": 5480, + "valid_targets_mean": 7023.5, + "valid_targets_min": 5664 + }, + { + "epoch": 3.023704520396913, + "grad_norm": 0.35729185420520077, + "learning_rate": 2.8015354454726863e-05, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0938064306974411, + "step": 5485, + "valid_targets_mean": 7069.8, + "valid_targets_min": 5103 + }, + { + "epoch": 3.0264608599779494, + "grad_norm": 0.40505908564712934, + "learning_rate": 2.7990160826953336e-05, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09251865744590759, + "step": 5490, + "valid_targets_mean": 6712.4, + "valid_targets_min": 5469 + }, + { + "epoch": 3.029217199558986, + "grad_norm": 0.3576632395785438, + "learning_rate": 2.796495210342813e-05, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09284640848636627, + "step": 5495, + "valid_targets_mean": 7289.2, + "valid_targets_min": 5513 + }, + { + "epoch": 3.031973539140022, + "grad_norm": 0.3789253014287821, + "learning_rate": 2.7939728331777892e-05, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09206780046224594, + "step": 5500, + "valid_targets_mean": 7022.1, + "valid_targets_min": 5691 + }, + { + "epoch": 3.0347298787210586, + "grad_norm": 0.38992569472716804, + "learning_rate": 2.7914489559657726e-05, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09995420277118683, + "step": 5505, + "valid_targets_mean": 6736.0, + "valid_targets_min": 5550 + }, + { + "epoch": 3.037486218302095, + "grad_norm": 0.3895498526194338, + "learning_rate": 2.788923583475104e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0875890851020813, + "step": 5510, + "valid_targets_mean": 6346.9, + "valid_targets_min": 5067 + }, + { + "epoch": 3.0402425578831314, + "grad_norm": 0.37934306373468946, + "learning_rate": 2.786396720476951e-05, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09450038522481918, + "step": 5515, + "valid_targets_mean": 6784.0, + "valid_targets_min": 5618 + }, + { + "epoch": 3.0429988974641677, + "grad_norm": 0.3902299301401398, + "learning_rate": 2.7838683717452988e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11106574535369873, + "step": 5520, + "valid_targets_mean": 7568.4, + "valid_targets_min": 5856 + }, + { + "epoch": 3.045755237045204, + "grad_norm": 0.3685698796904446, + "learning_rate": 2.7813385420569364e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08954589068889618, + "step": 5525, + "valid_targets_mean": 7198.5, + "valid_targets_min": 4793 + }, + { + "epoch": 3.0485115766262405, + "grad_norm": 0.39865771340237277, + "learning_rate": 2.778807236191452e-05, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09669599682092667, + "step": 5530, + "valid_targets_mean": 6680.5, + "valid_targets_min": 5350 + }, + { + "epoch": 3.051267916207277, + "grad_norm": 0.3948826064843428, + "learning_rate": 2.7762744589312238e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09900479018688202, + "step": 5535, + "valid_targets_mean": 6770.6, + "valid_targets_min": 5753 + }, + { + "epoch": 3.0540242557883133, + "grad_norm": 0.7892314720063306, + "learning_rate": 2.7737402150614098e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07253960520029068, + "step": 5540, + "valid_targets_mean": 1419.8, + "valid_targets_min": 289 + }, + { + "epoch": 3.0567805953693497, + "grad_norm": 0.4203206468022826, + "learning_rate": 2.7712045093699372e-05, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09826153516769409, + "step": 5545, + "valid_targets_mean": 6784.2, + "valid_targets_min": 5042 + }, + { + "epoch": 3.059536934950386, + "grad_norm": 0.3659100166037375, + "learning_rate": 2.768667346647496e-05, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09096711874008179, + "step": 5550, + "valid_targets_mean": 6774.9, + "valid_targets_min": 3875 + }, + { + "epoch": 3.0622932745314224, + "grad_norm": 0.40780749185553344, + "learning_rate": 2.766128731687529e-05, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09754148125648499, + "step": 5555, + "valid_targets_mean": 7047.6, + "valid_targets_min": 6134 + }, + { + "epoch": 3.065049614112459, + "grad_norm": 0.3845569113707035, + "learning_rate": 2.763588669286223e-05, + "loss": 0.1912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09984893351793289, + "step": 5560, + "valid_targets_mean": 6886.6, + "valid_targets_min": 5869 + }, + { + "epoch": 3.067805953693495, + "grad_norm": 0.37699120807626674, + "learning_rate": 2.761047164242499e-05, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08739814162254333, + "step": 5565, + "valid_targets_mean": 7559.8, + "valid_targets_min": 6145 + }, + { + "epoch": 3.0705622932745316, + "grad_norm": 0.3368430499912999, + "learning_rate": 2.758504221358003e-05, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08599332720041275, + "step": 5570, + "valid_targets_mean": 8920.5, + "valid_targets_min": 4780 + }, + { + "epoch": 3.073318632855568, + "grad_norm": 0.33328589042423834, + "learning_rate": 2.7559598454370995e-05, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07517801970243454, + "step": 5575, + "valid_targets_mean": 7393.4, + "valid_targets_min": 5116 + }, + { + "epoch": 3.0760749724366043, + "grad_norm": 0.324397830941256, + "learning_rate": 2.753414041286858e-05, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08365912735462189, + "step": 5580, + "valid_targets_mean": 8477.5, + "valid_targets_min": 4033 + }, + { + "epoch": 3.0788313120176407, + "grad_norm": 0.38647695364492934, + "learning_rate": 2.7508668137170473e-05, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09333128482103348, + "step": 5585, + "valid_targets_mean": 6736.6, + "valid_targets_min": 4646 + }, + { + "epoch": 3.081587651598677, + "grad_norm": 0.39668270438879594, + "learning_rate": 2.748318167540126e-05, + "loss": 0.1792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09583695232868195, + "step": 5590, + "valid_targets_mean": 7551.5, + "valid_targets_min": 5461 + }, + { + "epoch": 3.0843439911797135, + "grad_norm": 0.41212937199289223, + "learning_rate": 2.7457681075712326e-05, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11088399589061737, + "step": 5595, + "valid_targets_mean": 7091.0, + "valid_targets_min": 4879 + }, + { + "epoch": 3.08710033076075, + "grad_norm": 0.4135986279603793, + "learning_rate": 2.7432166386281768e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08932960033416748, + "step": 5600, + "valid_targets_mean": 6269.8, + "valid_targets_min": 4606 + }, + { + "epoch": 3.0898566703417862, + "grad_norm": 0.36978540884856576, + "learning_rate": 2.7406637655314288e-05, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08826185017824173, + "step": 5605, + "valid_targets_mean": 6818.4, + "valid_targets_min": 5525 + }, + { + "epoch": 3.0926130099228226, + "grad_norm": 0.377452632483711, + "learning_rate": 2.7381094931041134e-05, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09279502928256989, + "step": 5610, + "valid_targets_mean": 7365.4, + "valid_targets_min": 5000 + }, + { + "epoch": 3.095369349503859, + "grad_norm": 0.41866047338541146, + "learning_rate": 2.7355538261719994e-05, + "loss": 0.1908, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08730834722518921, + "step": 5615, + "valid_targets_mean": 6212.1, + "valid_targets_min": 3988 + }, + { + "epoch": 3.0981256890848954, + "grad_norm": 0.3550000990962687, + "learning_rate": 2.7329967695634885e-05, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09248140454292297, + "step": 5620, + "valid_targets_mean": 8396.9, + "valid_targets_min": 6131 + }, + { + "epoch": 3.1008820286659318, + "grad_norm": 0.3898180023777659, + "learning_rate": 2.7304383281096096e-05, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0959462821483612, + "step": 5625, + "valid_targets_mean": 7651.0, + "valid_targets_min": 4061 + }, + { + "epoch": 3.103638368246968, + "grad_norm": 0.36350133480674457, + "learning_rate": 2.7278785066440058e-05, + "loss": 0.1923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09393265098333359, + "step": 5630, + "valid_targets_mean": 7090.1, + "valid_targets_min": 5203 + }, + { + "epoch": 3.1063947078280045, + "grad_norm": 0.37869105546769055, + "learning_rate": 2.7253173100029297e-05, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09026610851287842, + "step": 5635, + "valid_targets_mean": 7001.0, + "valid_targets_min": 5362 + }, + { + "epoch": 3.109151047409041, + "grad_norm": 0.37446700108201875, + "learning_rate": 2.7227547430252315e-05, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09412816166877747, + "step": 5640, + "valid_targets_mean": 6359.2, + "valid_targets_min": 5236 + }, + { + "epoch": 3.1119073869900773, + "grad_norm": 0.5383267495526032, + "learning_rate": 2.7201908105523496e-05, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10232231020927429, + "step": 5645, + "valid_targets_mean": 5939.8, + "valid_targets_min": 4693 + }, + { + "epoch": 3.1146637265711137, + "grad_norm": 1.0672893061952489, + "learning_rate": 2.717625517428303e-05, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03676833584904671, + "step": 5650, + "valid_targets_mean": 208.6, + "valid_targets_min": 134 + }, + { + "epoch": 3.11742006615215, + "grad_norm": 0.487221478239995, + "learning_rate": 2.7150588684996803e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09813016653060913, + "step": 5655, + "valid_targets_mean": 6008.5, + "valid_targets_min": 4956 + }, + { + "epoch": 3.1201764057331864, + "grad_norm": 0.3728405770654858, + "learning_rate": 2.712490868615633e-05, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0817645788192749, + "step": 5660, + "valid_targets_mean": 6042.0, + "valid_targets_min": 4817 + }, + { + "epoch": 3.122932745314223, + "grad_norm": 0.43504931068149905, + "learning_rate": 2.7099215226278635e-05, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09250076860189438, + "step": 5665, + "valid_targets_mean": 6608.6, + "valid_targets_min": 5102 + }, + { + "epoch": 3.125689084895259, + "grad_norm": 0.46373381543197517, + "learning_rate": 2.707350835390619e-05, + "loss": 0.1837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09495482593774796, + "step": 5670, + "valid_targets_mean": 6670.1, + "valid_targets_min": 4869 + }, + { + "epoch": 3.1284454244762956, + "grad_norm": 0.38951643264124824, + "learning_rate": 2.7047788117606794e-05, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0834227129817009, + "step": 5675, + "valid_targets_mean": 6088.4, + "valid_targets_min": 4946 + }, + { + "epoch": 3.131201764057332, + "grad_norm": 0.38708165778516423, + "learning_rate": 2.7022054565973498e-05, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08991050720214844, + "step": 5680, + "valid_targets_mean": 6228.4, + "valid_targets_min": 4856 + }, + { + "epoch": 3.1339581036383684, + "grad_norm": 0.41282728509590705, + "learning_rate": 2.6996307747624514e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09633965790271759, + "step": 5685, + "valid_targets_mean": 5376.6, + "valid_targets_min": 4145 + }, + { + "epoch": 3.1367144432194047, + "grad_norm": 0.4253997799763961, + "learning_rate": 2.6970547711203114e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09415596723556519, + "step": 5690, + "valid_targets_mean": 6309.2, + "valid_targets_min": 5305 + }, + { + "epoch": 3.139470782800441, + "grad_norm": 0.3843324621966002, + "learning_rate": 2.6944774505377545e-05, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09042485058307648, + "step": 5695, + "valid_targets_mean": 6221.5, + "valid_targets_min": 4636 + }, + { + "epoch": 3.1422271223814775, + "grad_norm": 0.38358837073870367, + "learning_rate": 2.6918988178840932e-05, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09789557754993439, + "step": 5700, + "valid_targets_mean": 6831.5, + "valid_targets_min": 5668 + }, + { + "epoch": 3.144983461962514, + "grad_norm": 0.3774143586228477, + "learning_rate": 2.6893188780311182e-05, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09783899039030075, + "step": 5705, + "valid_targets_mean": 6437.1, + "valid_targets_min": 5004 + }, + { + "epoch": 3.1477398015435503, + "grad_norm": 0.38474350506471877, + "learning_rate": 2.686737635853093e-05, + "loss": 0.1793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08706305921077728, + "step": 5710, + "valid_targets_mean": 6205.8, + "valid_targets_min": 5341 + }, + { + "epoch": 3.1504961411245866, + "grad_norm": 0.4536050009353451, + "learning_rate": 2.6841550962267376e-05, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09977684915065765, + "step": 5715, + "valid_targets_mean": 6748.0, + "valid_targets_min": 5363 + }, + { + "epoch": 3.153252480705623, + "grad_norm": 0.4123783752164396, + "learning_rate": 2.681571264031225e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0862569659948349, + "step": 5720, + "valid_targets_mean": 5210.5, + "valid_targets_min": 791 + }, + { + "epoch": 3.1560088202866594, + "grad_norm": 0.733192466978557, + "learning_rate": 2.678986144148171e-05, + "loss": 0.2012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07823304831981659, + "step": 5725, + "valid_targets_mean": 1279.5, + "valid_targets_min": 746 + }, + { + "epoch": 3.158765159867696, + "grad_norm": 0.7315213644671683, + "learning_rate": 2.6763997414616226e-05, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09582959115505219, + "step": 5730, + "valid_targets_mean": 1635.8, + "valid_targets_min": 661 + }, + { + "epoch": 3.161521499448732, + "grad_norm": 0.7304714207358998, + "learning_rate": 2.673812060858052e-05, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0846966803073883, + "step": 5735, + "valid_targets_mean": 1420.2, + "valid_targets_min": 826 + }, + { + "epoch": 3.1642778390297686, + "grad_norm": 0.7309518234040029, + "learning_rate": 2.6712231072263453e-05, + "loss": 0.1672, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07674509286880493, + "step": 5740, + "valid_targets_mean": 1399.9, + "valid_targets_min": 742 + }, + { + "epoch": 3.167034178610805, + "grad_norm": 0.7266825735610742, + "learning_rate": 2.6686328854577923e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.069561667740345, + "step": 5745, + "valid_targets_mean": 1309.9, + "valid_targets_min": 718 + }, + { + "epoch": 3.1697905181918413, + "grad_norm": 0.6946700929212919, + "learning_rate": 2.6660414004460806e-05, + "loss": 0.1767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10234398394823074, + "step": 5750, + "valid_targets_mean": 1729.4, + "valid_targets_min": 764 + }, + { + "epoch": 3.1725468577728777, + "grad_norm": 0.679842590696938, + "learning_rate": 2.663448657087283e-05, + "loss": 0.1628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08368152379989624, + "step": 5755, + "valid_targets_mean": 1651.4, + "valid_targets_min": 814 + }, + { + "epoch": 3.175303197353914, + "grad_norm": 0.7571490294640401, + "learning_rate": 2.6608546602798512e-05, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06547502428293228, + "step": 5760, + "valid_targets_mean": 1192.0, + "valid_targets_min": 610 + }, + { + "epoch": 3.1780595369349505, + "grad_norm": 0.7215224440840499, + "learning_rate": 2.6582594149246035e-05, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09553395211696625, + "step": 5765, + "valid_targets_mean": 1879.0, + "valid_targets_min": 817 + }, + { + "epoch": 3.180815876515987, + "grad_norm": 0.6637574272511632, + "learning_rate": 2.6556629259247175e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0780840814113617, + "step": 5770, + "valid_targets_mean": 1500.4, + "valid_targets_min": 837 + }, + { + "epoch": 3.1835722160970232, + "grad_norm": 0.7588719906863618, + "learning_rate": 2.6530651981857217e-05, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11358550935983658, + "step": 5775, + "valid_targets_mean": 1661.8, + "valid_targets_min": 593 + }, + { + "epoch": 3.1863285556780596, + "grad_norm": 0.6924549271355628, + "learning_rate": 2.6504662366154826e-05, + "loss": 0.1693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08175567537546158, + "step": 5780, + "valid_targets_mean": 1537.8, + "valid_targets_min": 704 + }, + { + "epoch": 3.189084895259096, + "grad_norm": 0.706916332263908, + "learning_rate": 2.6478660461242e-05, + "loss": 0.1616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06291428953409195, + "step": 5785, + "valid_targets_mean": 1227.0, + "valid_targets_min": 722 + }, + { + "epoch": 3.1918412348401324, + "grad_norm": 0.7015056583205963, + "learning_rate": 2.6452646316243943e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07243005186319351, + "step": 5790, + "valid_targets_mean": 1364.5, + "valid_targets_min": 642 + }, + { + "epoch": 3.1945975744211688, + "grad_norm": 0.7503025055609516, + "learning_rate": 2.6426619980308988e-05, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08095967024564743, + "step": 5795, + "valid_targets_mean": 1358.5, + "valid_targets_min": 518 + }, + { + "epoch": 3.197353914002205, + "grad_norm": 0.7858358661442204, + "learning_rate": 2.640058150260849e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09902496635913849, + "step": 5800, + "valid_targets_mean": 1590.6, + "valid_targets_min": 728 + }, + { + "epoch": 3.2001102535832415, + "grad_norm": 0.7017508849978643, + "learning_rate": 2.6374530932336763e-05, + "loss": 0.1733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10237124562263489, + "step": 5805, + "valid_targets_mean": 1791.4, + "valid_targets_min": 820 + }, + { + "epoch": 3.202866593164278, + "grad_norm": 0.7269085216828398, + "learning_rate": 2.6348468318710952e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09065036475658417, + "step": 5810, + "valid_targets_mean": 1470.8, + "valid_targets_min": 818 + }, + { + "epoch": 3.2056229327453143, + "grad_norm": 0.7568569416348881, + "learning_rate": 2.6322393710970963e-05, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07012362033128738, + "step": 5815, + "valid_targets_mean": 1278.6, + "valid_targets_min": 625 + }, + { + "epoch": 3.2083792723263507, + "grad_norm": 0.7155667316054934, + "learning_rate": 2.6296307158379356e-05, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11300060153007507, + "step": 5820, + "valid_targets_mean": 2313.5, + "valid_targets_min": 790 + }, + { + "epoch": 3.211135611907387, + "grad_norm": 0.7956799282078975, + "learning_rate": 2.6270208710221253e-05, + "loss": 0.1758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11837980151176453, + "step": 5825, + "valid_targets_mean": 2231.9, + "valid_targets_min": 1335 + }, + { + "epoch": 3.2138919514884234, + "grad_norm": 0.6848857016842885, + "learning_rate": 2.6244098415804272e-05, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09672443568706512, + "step": 5830, + "valid_targets_mean": 1729.5, + "valid_targets_min": 719 + }, + { + "epoch": 3.21664829106946, + "grad_norm": 0.7093903704019959, + "learning_rate": 2.62179763244584e-05, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06476243585348129, + "step": 5835, + "valid_targets_mean": 1222.9, + "valid_targets_min": 697 + }, + { + "epoch": 3.219404630650496, + "grad_norm": 0.7918128816783675, + "learning_rate": 2.61918424855359e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09683743119239807, + "step": 5840, + "valid_targets_mean": 1521.1, + "valid_targets_min": 998 + }, + { + "epoch": 3.2221609702315326, + "grad_norm": 0.7006813480337333, + "learning_rate": 2.616569694841125e-05, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08557940274477005, + "step": 5845, + "valid_targets_mean": 1693.9, + "valid_targets_min": 629 + }, + { + "epoch": 3.224917309812569, + "grad_norm": 0.7174339823303884, + "learning_rate": 2.6139539762481007e-05, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08777754008769989, + "step": 5850, + "valid_targets_mean": 1630.2, + "valid_targets_min": 1078 + }, + { + "epoch": 3.2276736493936053, + "grad_norm": 0.6792389589202805, + "learning_rate": 2.6113370977163762e-05, + "loss": 0.1643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08455431461334229, + "step": 5855, + "valid_targets_mean": 1843.0, + "valid_targets_min": 1075 + }, + { + "epoch": 3.2304299889746417, + "grad_norm": 0.7187029897277646, + "learning_rate": 2.6087190641900005e-05, + "loss": 0.1714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09348919987678528, + "step": 5860, + "valid_targets_mean": 1848.2, + "valid_targets_min": 1317 + }, + { + "epoch": 3.233186328555678, + "grad_norm": 0.7525697077989609, + "learning_rate": 2.6060998806152045e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09482672810554504, + "step": 5865, + "valid_targets_mean": 1727.4, + "valid_targets_min": 957 + }, + { + "epoch": 3.2359426681367145, + "grad_norm": 0.7655107250010574, + "learning_rate": 2.6034795519403927e-05, + "loss": 0.1651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0670396089553833, + "step": 5870, + "valid_targets_mean": 1115.4, + "valid_targets_min": 643 + }, + { + "epoch": 3.238699007717751, + "grad_norm": 0.7572428073173217, + "learning_rate": 2.600858083116132e-05, + "loss": 0.1713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0872233659029007, + "step": 5875, + "valid_targets_mean": 1547.2, + "valid_targets_min": 1111 + }, + { + "epoch": 3.2414553472987873, + "grad_norm": 0.6944719049987913, + "learning_rate": 2.598235479095145e-05, + "loss": 0.1591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07313555479049683, + "step": 5880, + "valid_targets_mean": 1450.0, + "valid_targets_min": 705 + }, + { + "epoch": 3.2442116868798236, + "grad_norm": 0.749179544824971, + "learning_rate": 2.5956117448322972e-05, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08622872084379196, + "step": 5885, + "valid_targets_mean": 1539.0, + "valid_targets_min": 1288 + }, + { + "epoch": 3.24696802646086, + "grad_norm": 0.8826764169789537, + "learning_rate": 2.592986885284591e-05, + "loss": 0.1523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07555416971445084, + "step": 5890, + "valid_targets_mean": 1321.2, + "valid_targets_min": 753 + }, + { + "epoch": 3.2497243660418964, + "grad_norm": 0.748785801282293, + "learning_rate": 2.590360905411154e-05, + "loss": 0.1654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07820481061935425, + "step": 5895, + "valid_targets_mean": 1480.0, + "valid_targets_min": 817 + }, + { + "epoch": 3.252480705622933, + "grad_norm": 0.7103339688082797, + "learning_rate": 2.5877338101732293e-05, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06588625907897949, + "step": 5900, + "valid_targets_mean": 1293.9, + "valid_targets_min": 974 + }, + { + "epoch": 3.255237045203969, + "grad_norm": 0.8096086746003168, + "learning_rate": 2.5851056045341704e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08779530227184296, + "step": 5905, + "valid_targets_mean": 1529.2, + "valid_targets_min": 657 + }, + { + "epoch": 3.2579933847850056, + "grad_norm": 0.8001539827855592, + "learning_rate": 2.5824762934594256e-05, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10069283843040466, + "step": 5910, + "valid_targets_mean": 1678.0, + "valid_targets_min": 880 + }, + { + "epoch": 3.260749724366042, + "grad_norm": 0.7448801633900541, + "learning_rate": 2.5798458819165335e-05, + "loss": 0.167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09521423280239105, + "step": 5915, + "valid_targets_mean": 1609.4, + "valid_targets_min": 1050 + }, + { + "epoch": 3.2635060639470783, + "grad_norm": 0.7698277561659489, + "learning_rate": 2.57721437487511e-05, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07495944201946259, + "step": 5920, + "valid_targets_mean": 1449.0, + "valid_targets_min": 682 + }, + { + "epoch": 3.2662624035281147, + "grad_norm": 0.697169866938804, + "learning_rate": 2.5745817773068428e-05, + "loss": 0.1607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07287945598363876, + "step": 5925, + "valid_targets_mean": 1344.0, + "valid_targets_min": 703 + }, + { + "epoch": 3.269018743109151, + "grad_norm": 0.7650830799985997, + "learning_rate": 2.571948094185479e-05, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07763326168060303, + "step": 5930, + "valid_targets_mean": 1469.4, + "valid_targets_min": 934 + }, + { + "epoch": 3.2717750826901875, + "grad_norm": 0.71900197086927, + "learning_rate": 2.5693133304868164e-05, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12003777176141739, + "step": 5935, + "valid_targets_mean": 1829.1, + "valid_targets_min": 734 + }, + { + "epoch": 3.274531422271224, + "grad_norm": 0.7090467618771819, + "learning_rate": 2.5666774911886948e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10442422330379486, + "step": 5940, + "valid_targets_mean": 2010.8, + "valid_targets_min": 1437 + }, + { + "epoch": 3.2772877618522602, + "grad_norm": 0.7195143662671605, + "learning_rate": 2.5640405812709847e-05, + "loss": 0.1582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062431998550891876, + "step": 5945, + "valid_targets_mean": 1405.2, + "valid_targets_min": 551 + }, + { + "epoch": 3.2800441014332966, + "grad_norm": 0.8523578253341826, + "learning_rate": 2.5614026057155822e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10562093555927277, + "step": 5950, + "valid_targets_mean": 1817.6, + "valid_targets_min": 931 + }, + { + "epoch": 3.282800441014333, + "grad_norm": 0.7173010299720934, + "learning_rate": 2.558763569506394e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08226557821035385, + "step": 5955, + "valid_targets_mean": 1599.1, + "valid_targets_min": 904 + }, + { + "epoch": 3.2855567805953694, + "grad_norm": 0.7140061683748202, + "learning_rate": 2.556123477629332e-05, + "loss": 0.1587, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08809058368206024, + "step": 5960, + "valid_targets_mean": 1774.8, + "valid_targets_min": 1158 + }, + { + "epoch": 3.2883131201764058, + "grad_norm": 0.7276346256918413, + "learning_rate": 2.5534823350723022e-05, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07668167352676392, + "step": 5965, + "valid_targets_mean": 1229.2, + "valid_targets_min": 756 + }, + { + "epoch": 3.291069459757442, + "grad_norm": 0.7856179228039758, + "learning_rate": 2.550840146825195e-05, + "loss": 0.1686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0928514152765274, + "step": 5970, + "valid_targets_mean": 1536.9, + "valid_targets_min": 854 + }, + { + "epoch": 3.2938257993384785, + "grad_norm": 0.7086200779280016, + "learning_rate": 2.548196917879878e-05, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08730621635913849, + "step": 5975, + "valid_targets_mean": 1739.5, + "valid_targets_min": 1462 + }, + { + "epoch": 3.296582138919515, + "grad_norm": 0.6910874231366118, + "learning_rate": 2.5455526532301838e-05, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07231380045413971, + "step": 5980, + "valid_targets_mean": 1475.9, + "valid_targets_min": 944 + }, + { + "epoch": 3.2993384785005513, + "grad_norm": 0.7097962611162474, + "learning_rate": 2.5429073578719018e-05, + "loss": 0.1591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07361627370119095, + "step": 5985, + "valid_targets_mean": 1482.6, + "valid_targets_min": 1227 + }, + { + "epoch": 3.3020948180815877, + "grad_norm": 0.6952642521636526, + "learning_rate": 2.540261036802769e-05, + "loss": 0.1617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06603186577558517, + "step": 5990, + "valid_targets_mean": 1276.5, + "valid_targets_min": 740 + }, + { + "epoch": 3.304851157662624, + "grad_norm": 0.6884023440591046, + "learning_rate": 2.53761369502246e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09537939727306366, + "step": 5995, + "valid_targets_mean": 2024.8, + "valid_targets_min": 881 + }, + { + "epoch": 3.3076074972436604, + "grad_norm": 0.790700456508584, + "learning_rate": 2.5349653375325788e-05, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08138221502304077, + "step": 6000, + "valid_targets_mean": 1288.6, + "valid_targets_min": 636 + }, + { + "epoch": 3.310363836824697, + "grad_norm": 0.8189685549570664, + "learning_rate": 2.5323159693366472e-05, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05727199465036392, + "step": 6005, + "valid_targets_mean": 1103.9, + "valid_targets_min": 814 + }, + { + "epoch": 3.313120176405733, + "grad_norm": 0.7707590045061871, + "learning_rate": 2.529665595440096e-05, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07837671041488647, + "step": 6010, + "valid_targets_mean": 1392.9, + "valid_targets_min": 820 + }, + { + "epoch": 3.3158765159867696, + "grad_norm": 0.7808161724507438, + "learning_rate": 2.5270142208502586e-05, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07116658985614777, + "step": 6015, + "valid_targets_mean": 1392.4, + "valid_targets_min": 661 + }, + { + "epoch": 3.318632855567806, + "grad_norm": 0.759554226340249, + "learning_rate": 2.5243618505763562e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07573363184928894, + "step": 6020, + "valid_targets_mean": 1441.8, + "valid_targets_min": 728 + }, + { + "epoch": 3.3213891951488423, + "grad_norm": 0.760892434929881, + "learning_rate": 2.5217084896294926e-05, + "loss": 0.1559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07629093527793884, + "step": 6025, + "valid_targets_mean": 1413.5, + "valid_targets_min": 692 + }, + { + "epoch": 3.3241455347298787, + "grad_norm": 0.7417839215269948, + "learning_rate": 2.5190541430226432e-05, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0807802602648735, + "step": 6030, + "valid_targets_mean": 1323.5, + "valid_targets_min": 840 + }, + { + "epoch": 3.326901874310915, + "grad_norm": 0.7461133568092341, + "learning_rate": 2.516398815770645e-05, + "loss": 0.1638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07975639402866364, + "step": 6035, + "valid_targets_mean": 1488.4, + "valid_targets_min": 950 + }, + { + "epoch": 3.3296582138919515, + "grad_norm": 0.6933779109067502, + "learning_rate": 2.513742512890189e-05, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0891779363155365, + "step": 6040, + "valid_targets_mean": 1935.9, + "valid_targets_min": 878 + }, + { + "epoch": 3.332414553472988, + "grad_norm": 0.7822588388388589, + "learning_rate": 2.511085239399807e-05, + "loss": 0.1617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08235520124435425, + "step": 6045, + "valid_targets_mean": 1502.1, + "valid_targets_min": 836 + }, + { + "epoch": 3.3351708930540243, + "grad_norm": 0.7564187363485846, + "learning_rate": 2.5084270003198672e-05, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08628134429454803, + "step": 6050, + "valid_targets_mean": 1535.1, + "valid_targets_min": 1013 + }, + { + "epoch": 3.3379272326350606, + "grad_norm": 0.78653356880436, + "learning_rate": 2.5057678006725613e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07594513893127441, + "step": 6055, + "valid_targets_mean": 1346.4, + "valid_targets_min": 748 + }, + { + "epoch": 3.340683572216097, + "grad_norm": 0.7269972304462669, + "learning_rate": 2.503107645481895e-05, + "loss": 0.1609, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08174735307693481, + "step": 6060, + "valid_targets_mean": 1543.9, + "valid_targets_min": 729 + }, + { + "epoch": 3.3434399117971334, + "grad_norm": 0.6530745308096902, + "learning_rate": 2.5004465397736794e-05, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07158280909061432, + "step": 6065, + "valid_targets_mean": 1541.9, + "valid_targets_min": 1219 + }, + { + "epoch": 3.34619625137817, + "grad_norm": 0.6802513403943059, + "learning_rate": 2.4977844885755228e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08008342981338501, + "step": 6070, + "valid_targets_mean": 1632.8, + "valid_targets_min": 916 + }, + { + "epoch": 3.348952590959206, + "grad_norm": 0.7143614111292231, + "learning_rate": 2.4951214969168177e-05, + "loss": 0.1624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07691249996423721, + "step": 6075, + "valid_targets_mean": 1349.9, + "valid_targets_min": 646 + }, + { + "epoch": 3.3517089305402425, + "grad_norm": 0.7660824462440491, + "learning_rate": 2.492457569828736e-05, + "loss": 0.1574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08363286405801773, + "step": 6080, + "valid_targets_mean": 1383.0, + "valid_targets_min": 906 + }, + { + "epoch": 3.354465270121279, + "grad_norm": 0.702909112057972, + "learning_rate": 2.4897927123442135e-05, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0768410712480545, + "step": 6085, + "valid_targets_mean": 1577.5, + "valid_targets_min": 848 + }, + { + "epoch": 3.3572216097023153, + "grad_norm": 0.7984701713809915, + "learning_rate": 2.4871269294979473e-05, + "loss": 0.1652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07541509717702866, + "step": 6090, + "valid_targets_mean": 1183.5, + "valid_targets_min": 692 + }, + { + "epoch": 3.3599779492833517, + "grad_norm": 0.774072093641246, + "learning_rate": 2.4844602263263805e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07216143608093262, + "step": 6095, + "valid_targets_mean": 1373.9, + "valid_targets_min": 821 + }, + { + "epoch": 3.362734288864388, + "grad_norm": 0.6981318114824472, + "learning_rate": 2.481792607867696e-05, + "loss": 0.1519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06920501589775085, + "step": 6100, + "valid_targets_mean": 1428.1, + "valid_targets_min": 1013 + }, + { + "epoch": 3.3654906284454245, + "grad_norm": 0.7120012739030731, + "learning_rate": 2.479124079161805e-05, + "loss": 0.1543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07236483693122864, + "step": 6105, + "valid_targets_mean": 1420.9, + "valid_targets_min": 781 + }, + { + "epoch": 3.368246968026461, + "grad_norm": 0.7916478244099848, + "learning_rate": 2.4764546452503394e-05, + "loss": 0.1695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08564971387386322, + "step": 6110, + "valid_targets_mean": 1255.1, + "valid_targets_min": 548 + }, + { + "epoch": 3.3710033076074972, + "grad_norm": 0.7134462760274762, + "learning_rate": 2.4737843111766407e-05, + "loss": 0.154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08115824311971664, + "step": 6115, + "valid_targets_mean": 1411.0, + "valid_targets_min": 694 + }, + { + "epoch": 3.3737596471885336, + "grad_norm": 0.6331464051239999, + "learning_rate": 2.471113081985752e-05, + "loss": 0.1608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09116631746292114, + "step": 6120, + "valid_targets_mean": 1973.8, + "valid_targets_min": 1030 + }, + { + "epoch": 3.37651598676957, + "grad_norm": 0.6327927321240675, + "learning_rate": 2.4684409627244064e-05, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06312400102615356, + "step": 6125, + "valid_targets_mean": 1580.2, + "valid_targets_min": 682 + }, + { + "epoch": 3.3792723263506064, + "grad_norm": 0.8016502164107208, + "learning_rate": 2.4657679584410187e-05, + "loss": 0.152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07806111872196198, + "step": 6130, + "valid_targets_mean": 1425.4, + "valid_targets_min": 990 + }, + { + "epoch": 3.3820286659316428, + "grad_norm": 0.7011020573728186, + "learning_rate": 2.4630940741856762e-05, + "loss": 0.159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05860481411218643, + "step": 6135, + "valid_targets_mean": 1299.6, + "valid_targets_min": 639 + }, + { + "epoch": 3.384785005512679, + "grad_norm": 0.6678036776366516, + "learning_rate": 2.4604193150101287e-05, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06740453839302063, + "step": 6140, + "valid_targets_mean": 1458.8, + "valid_targets_min": 721 + }, + { + "epoch": 3.3875413450937155, + "grad_norm": 0.7355268013986647, + "learning_rate": 2.4577436859677795e-05, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0741904228925705, + "step": 6145, + "valid_targets_mean": 1329.8, + "valid_targets_min": 874 + }, + { + "epoch": 3.390297684674752, + "grad_norm": 0.6484073897094829, + "learning_rate": 2.455067192113674e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07578721642494202, + "step": 6150, + "valid_targets_mean": 1559.9, + "valid_targets_min": 686 + }, + { + "epoch": 3.3930540242557883, + "grad_norm": 0.7523815162229731, + "learning_rate": 2.4523898385044924e-05, + "loss": 0.1622, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08505912125110626, + "step": 6155, + "valid_targets_mean": 1392.5, + "valid_targets_min": 713 + }, + { + "epoch": 3.3958103638368247, + "grad_norm": 0.779287921821761, + "learning_rate": 2.449711630198539e-05, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14290107786655426, + "step": 6160, + "valid_targets_mean": 2341.5, + "valid_targets_min": 920 + }, + { + "epoch": 3.398566703417861, + "grad_norm": 0.5697396852762711, + "learning_rate": 2.447032572255733e-05, + "loss": 0.272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12825366854667664, + "step": 6165, + "valid_targets_mean": 5153.4, + "valid_targets_min": 746 + }, + { + "epoch": 3.4013230429988974, + "grad_norm": 0.4571967434597234, + "learning_rate": 2.4443526697375987e-05, + "loss": 0.2508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12569063901901245, + "step": 6170, + "valid_targets_mean": 5458.5, + "valid_targets_min": 1846 + }, + { + "epoch": 3.404079382579934, + "grad_norm": 0.483509878257088, + "learning_rate": 2.441671927707256e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09417054057121277, + "step": 6175, + "valid_targets_mean": 4888.1, + "valid_targets_min": 1321 + }, + { + "epoch": 3.40683572216097, + "grad_norm": 0.48065856521204625, + "learning_rate": 2.4389903512294115e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12299534678459167, + "step": 6180, + "valid_targets_mean": 5733.0, + "valid_targets_min": 2373 + }, + { + "epoch": 3.4095920617420066, + "grad_norm": 0.5412923320496345, + "learning_rate": 2.4363079453703467e-05, + "loss": 0.2555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11524488031864166, + "step": 6185, + "valid_targets_mean": 4224.2, + "valid_targets_min": 2075 + }, + { + "epoch": 3.412348401323043, + "grad_norm": 0.49594981181436676, + "learning_rate": 2.4336247151979128e-05, + "loss": 0.2487, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12700672447681427, + "step": 6190, + "valid_targets_mean": 5458.2, + "valid_targets_min": 1991 + }, + { + "epoch": 3.4151047409040793, + "grad_norm": 0.5512218677248406, + "learning_rate": 2.4309406657815146e-05, + "loss": 0.2449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10996672511100769, + "step": 6195, + "valid_targets_mean": 3290.8, + "valid_targets_min": 1748 + }, + { + "epoch": 3.4178610804851157, + "grad_norm": 0.4917679833924096, + "learning_rate": 2.4282558021921086e-05, + "loss": 0.2472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10907179862260818, + "step": 6200, + "valid_targets_mean": 4719.1, + "valid_targets_min": 1170 + }, + { + "epoch": 3.420617420066152, + "grad_norm": 0.5010042927435682, + "learning_rate": 2.425570129502187e-05, + "loss": 0.2568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11089029908180237, + "step": 6205, + "valid_targets_mean": 4463.9, + "valid_targets_min": 1241 + }, + { + "epoch": 3.4233737596471885, + "grad_norm": 0.5001753229589382, + "learning_rate": 2.4228836527857705e-05, + "loss": 0.2501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1325031965970993, + "step": 6210, + "valid_targets_mean": 4915.0, + "valid_targets_min": 2747 + }, + { + "epoch": 3.426130099228225, + "grad_norm": 0.4614057875369472, + "learning_rate": 2.4201963771184008e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1145433560013771, + "step": 6215, + "valid_targets_mean": 5663.9, + "valid_targets_min": 2144 + }, + { + "epoch": 3.4288864388092613, + "grad_norm": 0.5250211856090421, + "learning_rate": 2.4175083075771283e-05, + "loss": 0.2448, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1191571056842804, + "step": 6220, + "valid_targets_mean": 5095.8, + "valid_targets_min": 2587 + }, + { + "epoch": 3.4316427783902976, + "grad_norm": 0.4861769648670513, + "learning_rate": 2.4148194492405013e-05, + "loss": 0.2511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12507018446922302, + "step": 6225, + "valid_targets_mean": 5127.1, + "valid_targets_min": 1366 + }, + { + "epoch": 3.434399117971334, + "grad_norm": 0.5788839732892925, + "learning_rate": 2.4121298071885613e-05, + "loss": 0.2445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1129520982503891, + "step": 6230, + "valid_targets_mean": 3455.5, + "valid_targets_min": 1721 + }, + { + "epoch": 3.4371554575523704, + "grad_norm": 0.5988542572999631, + "learning_rate": 2.409439386502828e-05, + "loss": 0.255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11578713357448578, + "step": 6235, + "valid_targets_mean": 3213.5, + "valid_targets_min": 746 + }, + { + "epoch": 3.439911797133407, + "grad_norm": 0.5654342435496106, + "learning_rate": 2.4067481922662945e-05, + "loss": 0.2427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11922165751457214, + "step": 6240, + "valid_targets_mean": 3643.8, + "valid_targets_min": 1538 + }, + { + "epoch": 3.442668136714443, + "grad_norm": 0.5565748149285735, + "learning_rate": 2.4040562295634137e-05, + "loss": 0.2451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10118311643600464, + "step": 6245, + "valid_targets_mean": 3245.9, + "valid_targets_min": 1451 + }, + { + "epoch": 3.4454244762954795, + "grad_norm": 0.5626299985711787, + "learning_rate": 2.4013635034800897e-05, + "loss": 0.2577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13005909323692322, + "step": 6250, + "valid_targets_mean": 4339.1, + "valid_targets_min": 2550 + }, + { + "epoch": 3.448180815876516, + "grad_norm": 0.5737630530329318, + "learning_rate": 2.398670019103672e-05, + "loss": 0.2541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10685109347105026, + "step": 6255, + "valid_targets_mean": 3526.1, + "valid_targets_min": 1323 + }, + { + "epoch": 3.4509371554575523, + "grad_norm": 0.5636515318753583, + "learning_rate": 2.395975781522938e-05, + "loss": 0.248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14333531260490417, + "step": 6260, + "valid_targets_mean": 4593.9, + "valid_targets_min": 1921 + }, + { + "epoch": 3.4536934950385887, + "grad_norm": 0.5554860330118253, + "learning_rate": 2.3932807958280938e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10563274472951889, + "step": 6265, + "valid_targets_mean": 4119.6, + "valid_targets_min": 2004 + }, + { + "epoch": 3.456449834619625, + "grad_norm": 0.5645325677538005, + "learning_rate": 2.390585067110754e-05, + "loss": 0.2427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.100934237241745, + "step": 6270, + "valid_targets_mean": 3539.5, + "valid_targets_min": 2377 + }, + { + "epoch": 3.4592061742006615, + "grad_norm": 0.6264637512750715, + "learning_rate": 2.38788860046394e-05, + "loss": 0.2242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1300964504480362, + "step": 6275, + "valid_targets_mean": 3529.1, + "valid_targets_min": 2183 + }, + { + "epoch": 3.461962513781698, + "grad_norm": 0.5814102418120312, + "learning_rate": 2.3851914009820645e-05, + "loss": 0.2498, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14582836627960205, + "step": 6280, + "valid_targets_mean": 4312.6, + "valid_targets_min": 3031 + }, + { + "epoch": 3.464718853362734, + "grad_norm": 0.5445121140722001, + "learning_rate": 2.3824934737609283e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09232787787914276, + "step": 6285, + "valid_targets_mean": 3192.2, + "valid_targets_min": 1475 + }, + { + "epoch": 3.4674751929437706, + "grad_norm": 0.5256659879023111, + "learning_rate": 2.3797948238977044e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10246536135673523, + "step": 6290, + "valid_targets_mean": 3245.2, + "valid_targets_min": 648 + }, + { + "epoch": 3.470231532524807, + "grad_norm": 0.5638651238671871, + "learning_rate": 2.3770954564909326e-05, + "loss": 0.2439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1300908923149109, + "step": 6295, + "valid_targets_mean": 3821.4, + "valid_targets_min": 1233 + }, + { + "epoch": 3.4729878721058434, + "grad_norm": 0.580301099473287, + "learning_rate": 2.3743953766405067e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11024269461631775, + "step": 6300, + "valid_targets_mean": 3471.6, + "valid_targets_min": 2831 + }, + { + "epoch": 3.4757442116868797, + "grad_norm": 0.568266464549775, + "learning_rate": 2.3716945894476692e-05, + "loss": 0.2436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11894211918115616, + "step": 6305, + "valid_targets_mean": 3616.6, + "valid_targets_min": 1781 + }, + { + "epoch": 3.478500551267916, + "grad_norm": 0.6026284913719204, + "learning_rate": 2.3689931000149962e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12074629962444305, + "step": 6310, + "valid_targets_mean": 3264.4, + "valid_targets_min": 1108 + }, + { + "epoch": 3.4812568908489525, + "grad_norm": 0.610631994714555, + "learning_rate": 2.3662909134463917e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08928288519382477, + "step": 6315, + "valid_targets_mean": 2588.5, + "valid_targets_min": 1913 + }, + { + "epoch": 3.484013230429989, + "grad_norm": 0.5869577094809124, + "learning_rate": 2.363588034847077e-05, + "loss": 0.2351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12638024985790253, + "step": 6320, + "valid_targets_mean": 3787.1, + "valid_targets_min": 1828 + }, + { + "epoch": 3.4867695700110253, + "grad_norm": 0.5850943258769616, + "learning_rate": 2.3608844693235807e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11785274744033813, + "step": 6325, + "valid_targets_mean": 4211.4, + "valid_targets_min": 1589 + }, + { + "epoch": 3.4895259095920617, + "grad_norm": 0.5928166571950234, + "learning_rate": 2.358180221983729e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11187279224395752, + "step": 6330, + "valid_targets_mean": 3294.1, + "valid_targets_min": 1328 + }, + { + "epoch": 3.492282249173098, + "grad_norm": 0.621951075817021, + "learning_rate": 2.3554752979366367e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09094385802745819, + "step": 6335, + "valid_targets_mean": 2840.2, + "valid_targets_min": 1248 + }, + { + "epoch": 3.4950385887541344, + "grad_norm": 0.6305593204351143, + "learning_rate": 2.352769702292696e-05, + "loss": 0.2481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10779999196529388, + "step": 6340, + "valid_targets_mean": 3823.9, + "valid_targets_min": 1727 + }, + { + "epoch": 3.497794928335171, + "grad_norm": 0.5490727484182603, + "learning_rate": 2.3500634401635695e-05, + "loss": 0.2343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12658058106899261, + "step": 6345, + "valid_targets_mean": 4394.8, + "valid_targets_min": 1965 + }, + { + "epoch": 3.500551267916207, + "grad_norm": 0.5719755504942184, + "learning_rate": 2.3473565166621778e-05, + "loss": 0.2337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10791321843862534, + "step": 6350, + "valid_targets_mean": 3348.1, + "valid_targets_min": 1184 + }, + { + "epoch": 3.5033076074972436, + "grad_norm": 0.5516093559332497, + "learning_rate": 2.3446489369026912e-05, + "loss": 0.2361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11206869035959244, + "step": 6355, + "valid_targets_mean": 4168.8, + "valid_targets_min": 2490 + }, + { + "epoch": 3.50606394707828, + "grad_norm": 0.5850716090256856, + "learning_rate": 2.3419407060005207e-05, + "loss": 0.2441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12256944924592972, + "step": 6360, + "valid_targets_mean": 3463.2, + "valid_targets_min": 2500 + }, + { + "epoch": 3.5088202866593163, + "grad_norm": 1.0311635976309454, + "learning_rate": 2.339231829072307e-05, + "loss": 0.234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15043023228645325, + "step": 6365, + "valid_targets_mean": 3757.6, + "valid_targets_min": 1648 + }, + { + "epoch": 3.5115766262403527, + "grad_norm": 0.5800743069828463, + "learning_rate": 2.33652231123591e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11990758776664734, + "step": 6370, + "valid_targets_mean": 3793.1, + "valid_targets_min": 1580 + }, + { + "epoch": 3.514332965821389, + "grad_norm": 0.6288569424790974, + "learning_rate": 2.3338121576104028e-05, + "loss": 0.2371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1268543004989624, + "step": 6375, + "valid_targets_mean": 3714.8, + "valid_targets_min": 1577 + }, + { + "epoch": 3.5170893054024255, + "grad_norm": 0.5056659366450683, + "learning_rate": 2.3311013733160585e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10863229632377625, + "step": 6380, + "valid_targets_mean": 4516.8, + "valid_targets_min": 2095 + }, + { + "epoch": 3.519845644983462, + "grad_norm": 0.5616937059437629, + "learning_rate": 2.3283899634743414e-05, + "loss": 0.2334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10985977202653885, + "step": 6385, + "valid_targets_mean": 3676.5, + "valid_targets_min": 1673 + }, + { + "epoch": 3.5226019845644982, + "grad_norm": 0.5342582158894116, + "learning_rate": 2.3256779332078986e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13937783241271973, + "step": 6390, + "valid_targets_mean": 4783.1, + "valid_targets_min": 2494 + }, + { + "epoch": 3.5253583241455346, + "grad_norm": 0.5942178316901913, + "learning_rate": 2.322965287640548e-05, + "loss": 0.2323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10439326614141464, + "step": 6395, + "valid_targets_mean": 3445.6, + "valid_targets_min": 742 + }, + { + "epoch": 3.528114663726571, + "grad_norm": 0.6510831817104914, + "learning_rate": 2.3202520318972708e-05, + "loss": 0.2455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.127878338098526, + "step": 6400, + "valid_targets_mean": 4115.5, + "valid_targets_min": 1766 + }, + { + "epoch": 3.5308710033076074, + "grad_norm": 0.5755934478069096, + "learning_rate": 2.317538171104202e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12001635134220123, + "step": 6405, + "valid_targets_mean": 4264.9, + "valid_targets_min": 1504 + }, + { + "epoch": 3.5336273428886438, + "grad_norm": 0.5970400953842206, + "learning_rate": 2.314823710388618e-05, + "loss": 0.2489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13674861192703247, + "step": 6410, + "valid_targets_mean": 3899.8, + "valid_targets_min": 1170 + }, + { + "epoch": 3.53638368246968, + "grad_norm": 0.5912084258532354, + "learning_rate": 2.312108654878929e-05, + "loss": 0.2305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11348285526037216, + "step": 6415, + "valid_targets_mean": 2840.8, + "valid_targets_min": 1955 + }, + { + "epoch": 3.5391400220507165, + "grad_norm": 0.5612912359706568, + "learning_rate": 2.3093930097046697e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11426002532243729, + "step": 6420, + "valid_targets_mean": 3381.4, + "valid_targets_min": 1221 + }, + { + "epoch": 3.541896361631753, + "grad_norm": 0.5260385817149287, + "learning_rate": 2.3066767799964885e-05, + "loss": 0.2269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09829437732696533, + "step": 6425, + "valid_targets_mean": 4147.0, + "valid_targets_min": 2756 + }, + { + "epoch": 3.5446527012127893, + "grad_norm": 0.5676782673026454, + "learning_rate": 2.303959970886138e-05, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1137348935008049, + "step": 6430, + "valid_targets_mean": 4371.0, + "valid_targets_min": 2526 + }, + { + "epoch": 3.5474090407938257, + "grad_norm": 0.6159915231656697, + "learning_rate": 2.3012425875064657e-05, + "loss": 0.2236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13987073302268982, + "step": 6435, + "valid_targets_mean": 4382.9, + "valid_targets_min": 464 + }, + { + "epoch": 3.550165380374862, + "grad_norm": 0.5694200096555087, + "learning_rate": 2.298524634991403e-05, + "loss": 0.242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11723729223012924, + "step": 6440, + "valid_targets_mean": 3763.0, + "valid_targets_min": 1003 + }, + { + "epoch": 3.5529217199558984, + "grad_norm": 0.57946939733429, + "learning_rate": 2.2958061184759586e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12732458114624023, + "step": 6445, + "valid_targets_mean": 3726.9, + "valid_targets_min": 1632 + }, + { + "epoch": 3.555678059536935, + "grad_norm": 0.5752295487414472, + "learning_rate": 2.2930870430962055e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11474553495645523, + "step": 6450, + "valid_targets_mean": 3750.4, + "valid_targets_min": 1969 + }, + { + "epoch": 3.558434399117971, + "grad_norm": 0.5464253969709971, + "learning_rate": 2.2903674139892717e-05, + "loss": 0.2334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11070410907268524, + "step": 6455, + "valid_targets_mean": 3843.2, + "valid_targets_min": 1460 + }, + { + "epoch": 3.5611907386990076, + "grad_norm": 0.5660088546952129, + "learning_rate": 2.2876472362933333e-05, + "loss": 0.2429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11207298189401627, + "step": 6460, + "valid_targets_mean": 4060.6, + "valid_targets_min": 1172 + }, + { + "epoch": 3.563947078280044, + "grad_norm": 0.5357918643071977, + "learning_rate": 2.2849265151476016e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1279747039079666, + "step": 6465, + "valid_targets_mean": 4138.0, + "valid_targets_min": 1818 + }, + { + "epoch": 3.5667034178610804, + "grad_norm": 0.5551478855817771, + "learning_rate": 2.2822052556923145e-05, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11141138523817062, + "step": 6470, + "valid_targets_mean": 3926.0, + "valid_targets_min": 1535 + }, + { + "epoch": 3.5694597574421167, + "grad_norm": 0.5706296277594384, + "learning_rate": 2.279483463068728e-05, + "loss": 0.2323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14532016217708588, + "step": 6475, + "valid_targets_mean": 5372.1, + "valid_targets_min": 1814 + }, + { + "epoch": 3.572216097023153, + "grad_norm": 0.6388298681291077, + "learning_rate": 2.2767611424191043e-05, + "loss": 0.2249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12820503115653992, + "step": 6480, + "valid_targets_mean": 2874.2, + "valid_targets_min": 1073 + }, + { + "epoch": 3.5749724366041895, + "grad_norm": 0.5884020498116125, + "learning_rate": 2.274038298886704e-05, + "loss": 0.225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11537346243858337, + "step": 6485, + "valid_targets_mean": 3723.6, + "valid_targets_min": 2420 + }, + { + "epoch": 3.577728776185226, + "grad_norm": 0.6160288831284202, + "learning_rate": 2.271314937615775e-05, + "loss": 0.2167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12494195997714996, + "step": 6490, + "valid_targets_mean": 4068.2, + "valid_targets_min": 1079 + }, + { + "epoch": 3.5804851157662623, + "grad_norm": 0.5969402661376738, + "learning_rate": 2.2685910637515427e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09781438857316971, + "step": 6495, + "valid_targets_mean": 3211.1, + "valid_targets_min": 1143 + }, + { + "epoch": 3.5832414553472987, + "grad_norm": 0.5673717023479458, + "learning_rate": 2.2658666824402037e-05, + "loss": 0.2268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09495288133621216, + "step": 6500, + "valid_targets_mean": 3013.6, + "valid_targets_min": 1853 + }, + { + "epoch": 3.585997794928335, + "grad_norm": 0.6478393216834487, + "learning_rate": 2.26314179882891e-05, + "loss": 0.2532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11126609146595001, + "step": 6505, + "valid_targets_mean": 2993.5, + "valid_targets_min": 1410 + }, + { + "epoch": 3.5887541345093714, + "grad_norm": 0.6118625371750637, + "learning_rate": 2.2604164180657648e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10047833621501923, + "step": 6510, + "valid_targets_mean": 2561.1, + "valid_targets_min": 1566 + }, + { + "epoch": 3.591510474090408, + "grad_norm": 0.5991158339299018, + "learning_rate": 2.2576905452998093e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10771271586418152, + "step": 6515, + "valid_targets_mean": 3800.0, + "valid_targets_min": 2318 + }, + { + "epoch": 3.594266813671444, + "grad_norm": 0.5632061933717047, + "learning_rate": 2.2549641856810144e-05, + "loss": 0.2245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09061384201049805, + "step": 6520, + "valid_targets_mean": 2651.1, + "valid_targets_min": 1255 + }, + { + "epoch": 3.5970231532524806, + "grad_norm": 0.6062276982167984, + "learning_rate": 2.252237344360272e-05, + "loss": 0.2269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1338488757610321, + "step": 6525, + "valid_targets_mean": 4281.9, + "valid_targets_min": 2258 + }, + { + "epoch": 3.599779492833517, + "grad_norm": 0.5723569998454953, + "learning_rate": 2.249510026489383e-05, + "loss": 0.2223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10074644535779953, + "step": 6530, + "valid_targets_mean": 3365.5, + "valid_targets_min": 1281 + }, + { + "epoch": 3.6025358324145533, + "grad_norm": 0.583069107349578, + "learning_rate": 2.2467822372210487e-05, + "loss": 0.2265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11017134040594101, + "step": 6535, + "valid_targets_mean": 3460.4, + "valid_targets_min": 1445 + }, + { + "epoch": 3.6052921719955897, + "grad_norm": 0.5997827767595241, + "learning_rate": 2.244053981708861e-05, + "loss": 0.2312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14697425067424774, + "step": 6540, + "valid_targets_mean": 4763.4, + "valid_targets_min": 3705 + }, + { + "epoch": 3.608048511576626, + "grad_norm": 0.5724852814886714, + "learning_rate": 2.241325265107293e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12684254348278046, + "step": 6545, + "valid_targets_mean": 4469.2, + "valid_targets_min": 2717 + }, + { + "epoch": 3.6108048511576625, + "grad_norm": 0.5819739657680111, + "learning_rate": 2.238596092571689e-05, + "loss": 0.2214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11992213129997253, + "step": 6550, + "valid_targets_mean": 3554.1, + "valid_targets_min": 1508 + }, + { + "epoch": 3.613561190738699, + "grad_norm": 0.6155329089217028, + "learning_rate": 2.2358664692582542e-05, + "loss": 0.2411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11543053388595581, + "step": 6555, + "valid_targets_mean": 4410.9, + "valid_targets_min": 3504 + }, + { + "epoch": 3.6163175303197352, + "grad_norm": 0.5430589118916701, + "learning_rate": 2.2331364003240456e-05, + "loss": 0.236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11416298896074295, + "step": 6560, + "valid_targets_mean": 4097.1, + "valid_targets_min": 2368 + }, + { + "epoch": 3.6190738699007716, + "grad_norm": 0.6745967207821026, + "learning_rate": 2.230405890926962e-05, + "loss": 0.2435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.153903067111969, + "step": 6565, + "valid_targets_mean": 4886.9, + "valid_targets_min": 3446 + }, + { + "epoch": 3.621830209481808, + "grad_norm": 0.6295722900776561, + "learning_rate": 2.2276749462257355e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11935634166002274, + "step": 6570, + "valid_targets_mean": 3707.2, + "valid_targets_min": 1135 + }, + { + "epoch": 3.6245865490628444, + "grad_norm": 0.6960135865711051, + "learning_rate": 2.2249435713799188e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13569754362106323, + "step": 6575, + "valid_targets_mean": 4419.4, + "valid_targets_min": 1449 + }, + { + "epoch": 3.6273428886438808, + "grad_norm": 0.6153291833982304, + "learning_rate": 2.2222117715498788e-05, + "loss": 0.2419, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1153145432472229, + "step": 6580, + "valid_targets_mean": 3437.9, + "valid_targets_min": 1400 + }, + { + "epoch": 3.630099228224917, + "grad_norm": 0.5536115293673116, + "learning_rate": 2.2194795518967847e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09976896643638611, + "step": 6585, + "valid_targets_mean": 3713.0, + "valid_targets_min": 1433 + }, + { + "epoch": 3.6328555678059535, + "grad_norm": 0.5822886348031091, + "learning_rate": 2.2167469175825974e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09710104763507843, + "step": 6590, + "valid_targets_mean": 3649.5, + "valid_targets_min": 614 + }, + { + "epoch": 3.63561190738699, + "grad_norm": 0.6033443495566788, + "learning_rate": 2.214013873770065e-05, + "loss": 0.2321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10931587219238281, + "step": 6595, + "valid_targets_mean": 3474.1, + "valid_targets_min": 2041 + }, + { + "epoch": 3.6383682469680263, + "grad_norm": 0.6593281959545368, + "learning_rate": 2.211280425622705e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11606459319591522, + "step": 6600, + "valid_targets_mean": 3439.5, + "valid_targets_min": 1333 + }, + { + "epoch": 3.6411245865490627, + "grad_norm": 0.555232369512389, + "learning_rate": 2.2085465783048022e-05, + "loss": 0.2312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11951564252376556, + "step": 6605, + "valid_targets_mean": 4520.8, + "valid_targets_min": 1325 + }, + { + "epoch": 3.643880926130099, + "grad_norm": 0.5142357211199826, + "learning_rate": 2.2058123369813932e-05, + "loss": 0.2307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.095572829246521, + "step": 6610, + "valid_targets_mean": 3994.5, + "valid_targets_min": 2754 + }, + { + "epoch": 3.646637265711136, + "grad_norm": 0.5909716209504419, + "learning_rate": 2.203077706818259e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10437170416116714, + "step": 6615, + "valid_targets_mean": 3253.2, + "valid_targets_min": 1991 + }, + { + "epoch": 3.649393605292172, + "grad_norm": 0.5960506966229246, + "learning_rate": 2.2003426929819184e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13167747855186462, + "step": 6620, + "valid_targets_mean": 3927.8, + "valid_targets_min": 1076 + }, + { + "epoch": 3.6521499448732087, + "grad_norm": 0.5563087201040284, + "learning_rate": 2.1976073006396113e-05, + "loss": 0.2135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10953681170940399, + "step": 6625, + "valid_targets_mean": 4188.8, + "valid_targets_min": 1686 + }, + { + "epoch": 3.6549062844542446, + "grad_norm": 0.6049126686215717, + "learning_rate": 2.1948715349592944e-05, + "loss": 0.2374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11234384775161743, + "step": 6630, + "valid_targets_mean": 4221.0, + "valid_targets_min": 2185 + }, + { + "epoch": 3.6576626240352814, + "grad_norm": 0.5714383130657831, + "learning_rate": 2.1921354011096302e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08993025124073029, + "step": 6635, + "valid_targets_mean": 3487.9, + "valid_targets_min": 2024 + }, + { + "epoch": 3.6604189636163174, + "grad_norm": 0.5893815812183495, + "learning_rate": 2.1893989042599744e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11648046970367432, + "step": 6640, + "valid_targets_mean": 3816.6, + "valid_targets_min": 1777 + }, + { + "epoch": 3.663175303197354, + "grad_norm": 0.591313083919461, + "learning_rate": 2.1866620495803722e-05, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09757280349731445, + "step": 6645, + "valid_targets_mean": 3629.8, + "valid_targets_min": 1965 + }, + { + "epoch": 3.66593164277839, + "grad_norm": 0.5862533079324999, + "learning_rate": 2.183924842241542e-05, + "loss": 0.2355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1414259970188141, + "step": 6650, + "valid_targets_mean": 5056.9, + "valid_targets_min": 2587 + }, + { + "epoch": 3.668687982359427, + "grad_norm": 0.5529113658759379, + "learning_rate": 2.1811872874148688e-05, + "loss": 0.2274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11376511305570602, + "step": 6655, + "valid_targets_mean": 4733.9, + "valid_targets_min": 2762 + }, + { + "epoch": 3.671444321940463, + "grad_norm": 0.6533628097965359, + "learning_rate": 2.178449390272396e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10488781332969666, + "step": 6660, + "valid_targets_mean": 3052.5, + "valid_targets_min": 1035 + }, + { + "epoch": 3.6742006615214997, + "grad_norm": 0.5955189107506295, + "learning_rate": 2.175711155986811e-05, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1356402337551117, + "step": 6665, + "valid_targets_mean": 4380.9, + "valid_targets_min": 1876 + }, + { + "epoch": 3.6769570011025356, + "grad_norm": 0.5750831601745393, + "learning_rate": 2.172972589731441e-05, + "loss": 0.2199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11031937599182129, + "step": 6670, + "valid_targets_mean": 3833.4, + "valid_targets_min": 2487 + }, + { + "epoch": 3.6797133406835725, + "grad_norm": 0.5736787857202076, + "learning_rate": 2.170233696680238e-05, + "loss": 0.2204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10239791125059128, + "step": 6675, + "valid_targets_mean": 3927.4, + "valid_targets_min": 1360 + }, + { + "epoch": 3.6824696802646084, + "grad_norm": 0.5605299726379632, + "learning_rate": 2.167494482007772e-05, + "loss": 0.2139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11839331686496735, + "step": 6680, + "valid_targets_mean": 4537.6, + "valid_targets_min": 1496 + }, + { + "epoch": 3.6852260198456452, + "grad_norm": 0.6240189501052203, + "learning_rate": 2.164754950889221e-05, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12150789052248001, + "step": 6685, + "valid_targets_mean": 3578.1, + "valid_targets_min": 2059 + }, + { + "epoch": 3.687982359426681, + "grad_norm": 0.5577352453762487, + "learning_rate": 2.1620151085003625e-05, + "loss": 0.216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09704630821943283, + "step": 6690, + "valid_targets_mean": 3722.0, + "valid_targets_min": 1152 + }, + { + "epoch": 3.690738699007718, + "grad_norm": 0.570421100584476, + "learning_rate": 2.1592749600175594e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09543738514184952, + "step": 6695, + "valid_targets_mean": 3562.4, + "valid_targets_min": 1598 + }, + { + "epoch": 3.693495038588754, + "grad_norm": 0.5642343027896894, + "learning_rate": 2.1565345106177533e-05, + "loss": 0.239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10487445443868637, + "step": 6700, + "valid_targets_mean": 4045.2, + "valid_targets_min": 2267 + }, + { + "epoch": 3.6962513781697908, + "grad_norm": 0.6228344692663372, + "learning_rate": 2.1537937654784556e-05, + "loss": 0.2164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10362084209918976, + "step": 6705, + "valid_targets_mean": 3061.8, + "valid_targets_min": 1674 + }, + { + "epoch": 3.6990077177508267, + "grad_norm": 0.587927595737517, + "learning_rate": 2.1510527297777357e-05, + "loss": 0.2199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10690215975046158, + "step": 6710, + "valid_targets_mean": 3347.2, + "valid_targets_min": 1972 + }, + { + "epoch": 3.7017640573318635, + "grad_norm": 0.6100572661333825, + "learning_rate": 2.148311408694212e-05, + "loss": 0.219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0939115509390831, + "step": 6715, + "valid_targets_mean": 3000.5, + "valid_targets_min": 1664 + }, + { + "epoch": 3.7045203969128995, + "grad_norm": 0.5211476827334464, + "learning_rate": 2.1455698074070422e-05, + "loss": 0.2201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09135039895772934, + "step": 6720, + "valid_targets_mean": 4733.4, + "valid_targets_min": 3218 + }, + { + "epoch": 3.7072767364939363, + "grad_norm": 0.5722861174933341, + "learning_rate": 2.1428279310959136e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12320278584957123, + "step": 6725, + "valid_targets_mean": 3917.2, + "valid_targets_min": 2418 + }, + { + "epoch": 3.7100330760749722, + "grad_norm": 0.5853041712023183, + "learning_rate": 2.1400857849410324e-05, + "loss": 0.2254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12183592468500137, + "step": 6730, + "valid_targets_mean": 4429.4, + "valid_targets_min": 2100 + }, + { + "epoch": 3.712789415656009, + "grad_norm": 0.5902085133163445, + "learning_rate": 2.1373433741231144e-05, + "loss": 0.2262, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11789324879646301, + "step": 6735, + "valid_targets_mean": 3558.9, + "valid_targets_min": 1365 + }, + { + "epoch": 3.715545755237045, + "grad_norm": 0.5986150830367052, + "learning_rate": 2.1346007038233774e-05, + "loss": 0.2281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10624006390571594, + "step": 6740, + "valid_targets_mean": 3915.9, + "valid_targets_min": 1572 + }, + { + "epoch": 3.718302094818082, + "grad_norm": 0.616090422753814, + "learning_rate": 2.131857779223528e-05, + "loss": 0.2294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1315944641828537, + "step": 6745, + "valid_targets_mean": 4033.1, + "valid_targets_min": 876 + }, + { + "epoch": 3.7210584343991178, + "grad_norm": 0.414324442598967, + "learning_rate": 2.129114605505752e-05, + "loss": 0.1283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04831632599234581, + "step": 6750, + "valid_targets_mean": 2751.5, + "valid_targets_min": 458 + }, + { + "epoch": 3.7238147739801546, + "grad_norm": 0.3862184914109146, + "learning_rate": 2.126371187852708e-05, + "loss": 0.0899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042404964566230774, + "step": 6755, + "valid_targets_mean": 3367.2, + "valid_targets_min": 2802 + }, + { + "epoch": 3.7265711135611905, + "grad_norm": 0.4712309835577553, + "learning_rate": 2.1236275314475144e-05, + "loss": 0.086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04963439702987671, + "step": 6760, + "valid_targets_mean": 3343.4, + "valid_targets_min": 2068 + }, + { + "epoch": 3.7293274531422274, + "grad_norm": 0.4728078895721381, + "learning_rate": 2.120883641473742e-05, + "loss": 0.1056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05058549344539642, + "step": 6765, + "valid_targets_mean": 3412.8, + "valid_targets_min": 2425 + }, + { + "epoch": 3.7320837927232633, + "grad_norm": 0.4760670795382871, + "learning_rate": 2.1181395231154004e-05, + "loss": 0.0849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04970414564013481, + "step": 6770, + "valid_targets_mean": 2778.2, + "valid_targets_min": 1198 + }, + { + "epoch": 3.7348401323043, + "grad_norm": 0.683037825758872, + "learning_rate": 2.1153951815569326e-05, + "loss": 0.099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060338228940963745, + "step": 6775, + "valid_targets_mean": 1061.6, + "valid_targets_min": 807 + }, + { + "epoch": 3.737596471885336, + "grad_norm": 0.48898702383509246, + "learning_rate": 2.112650621983203e-05, + "loss": 0.1406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057330094277858734, + "step": 6780, + "valid_targets_mean": 3322.6, + "valid_targets_min": 1955 + }, + { + "epoch": 3.740352811466373, + "grad_norm": 0.45151676972595, + "learning_rate": 2.1099058495794874e-05, + "loss": 0.0928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06137385219335556, + "step": 6785, + "valid_targets_mean": 3079.1, + "valid_targets_min": 781 + }, + { + "epoch": 3.743109151047409, + "grad_norm": 0.38036919556273396, + "learning_rate": 2.1071608695314644e-05, + "loss": 0.1061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035455670207738876, + "step": 6790, + "valid_targets_mean": 3501.0, + "valid_targets_min": 799 + }, + { + "epoch": 3.7458654906284456, + "grad_norm": 0.3247132143560577, + "learning_rate": 2.104415687025204e-05, + "loss": 0.1106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040393758565187454, + "step": 6795, + "valid_targets_mean": 4797.0, + "valid_targets_min": 2802 + }, + { + "epoch": 3.7486218302094816, + "grad_norm": 0.34093370195689393, + "learning_rate": 2.1016703072471593e-05, + "loss": 0.0833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028735142201185226, + "step": 6800, + "valid_targets_mean": 2412.6, + "valid_targets_min": 574 + }, + { + "epoch": 3.7513781697905184, + "grad_norm": 0.36728757589918004, + "learning_rate": 2.0989247353841557e-05, + "loss": 0.0909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04940414801239967, + "step": 6805, + "valid_targets_mean": 3956.6, + "valid_targets_min": 2642 + }, + { + "epoch": 3.7541345093715544, + "grad_norm": 0.5815467694179637, + "learning_rate": 2.0961789766233824e-05, + "loss": 0.0924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05976349487900734, + "step": 6810, + "valid_targets_mean": 1665.5, + "valid_targets_min": 593 + }, + { + "epoch": 3.756890848952591, + "grad_norm": 0.3921125789436971, + "learning_rate": 2.0934330361523812e-05, + "loss": 0.1247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038540199398994446, + "step": 6815, + "valid_targets_mean": 3217.5, + "valid_targets_min": 1353 + }, + { + "epoch": 3.759647188533627, + "grad_norm": 0.5139575483954977, + "learning_rate": 2.0906869191590365e-05, + "loss": 0.0879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059729330241680145, + "step": 6820, + "valid_targets_mean": 2078.4, + "valid_targets_min": 818 + }, + { + "epoch": 3.762403528114664, + "grad_norm": 0.38767276342852597, + "learning_rate": 2.0879406308315668e-05, + "loss": 0.0855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034458715468645096, + "step": 6825, + "valid_targets_mean": 1870.9, + "valid_targets_min": 910 + }, + { + "epoch": 3.7651598676957, + "grad_norm": 0.3569330841427939, + "learning_rate": 2.0851941763585147e-05, + "loss": 0.0781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04207883030176163, + "step": 6830, + "valid_targets_mean": 3722.6, + "valid_targets_min": 2840 + }, + { + "epoch": 3.7679162072767367, + "grad_norm": 0.4580190533369842, + "learning_rate": 2.082447560928736e-05, + "loss": 0.0982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04615724831819534, + "step": 6835, + "valid_targets_mean": 2762.1, + "valid_targets_min": 471 + }, + { + "epoch": 3.7706725468577726, + "grad_norm": 0.4014583481098832, + "learning_rate": 2.079700789731391e-05, + "loss": 0.0988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057879846543073654, + "step": 6840, + "valid_targets_mean": 3063.6, + "valid_targets_min": 807 + }, + { + "epoch": 3.7734288864388095, + "grad_norm": 0.42447060160386446, + "learning_rate": 2.0769538679559343e-05, + "loss": 0.102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04705766588449478, + "step": 6845, + "valid_targets_mean": 2927.5, + "valid_targets_min": 474 + }, + { + "epoch": 3.7761852260198454, + "grad_norm": 0.4250658633394308, + "learning_rate": 2.074206800792105e-05, + "loss": 0.0993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04357674717903137, + "step": 6850, + "valid_targets_mean": 3334.0, + "valid_targets_min": 640 + }, + { + "epoch": 3.7789415656008822, + "grad_norm": 0.3587857459429799, + "learning_rate": 2.071459593429916e-05, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03554356470704079, + "step": 6855, + "valid_targets_mean": 3534.8, + "valid_targets_min": 902 + }, + { + "epoch": 3.781697905181918, + "grad_norm": 0.4310288965297041, + "learning_rate": 2.0687122510596462e-05, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04486100375652313, + "step": 6860, + "valid_targets_mean": 3005.8, + "valid_targets_min": 1837 + }, + { + "epoch": 3.784454244762955, + "grad_norm": 0.7011618122828553, + "learning_rate": 2.06596477887183e-05, + "loss": 0.1161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08555816859006882, + "step": 6865, + "valid_targets_mean": 1328.5, + "valid_targets_min": 756 + }, + { + "epoch": 3.787210584343991, + "grad_norm": 0.6050972906416856, + "learning_rate": 2.0632171820572454e-05, + "loss": 0.1219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08734875172376633, + "step": 6870, + "valid_targets_mean": 2216.6, + "valid_targets_min": 599 + }, + { + "epoch": 3.7899669239250278, + "grad_norm": 0.3933270715015242, + "learning_rate": 2.0604694658069065e-05, + "loss": 0.0904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03355976939201355, + "step": 6875, + "valid_targets_mean": 3032.1, + "valid_targets_min": 895 + }, + { + "epoch": 3.7927232635060637, + "grad_norm": 0.6860216616283776, + "learning_rate": 2.057721635312054e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17418301105499268, + "step": 6880, + "valid_targets_mean": 2477.9, + "valid_targets_min": 878 + }, + { + "epoch": 3.7954796030871005, + "grad_norm": 0.42544277832058824, + "learning_rate": 2.054973695764143e-05, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04193788021802902, + "step": 6885, + "valid_targets_mean": 2695.6, + "valid_targets_min": 756 + }, + { + "epoch": 3.7982359426681365, + "grad_norm": 0.38080711279470264, + "learning_rate": 2.0522256523548362e-05, + "loss": 0.1049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04206139221787453, + "step": 6890, + "valid_targets_mean": 3336.9, + "valid_targets_min": 1968 + }, + { + "epoch": 3.8009922822491733, + "grad_norm": 0.8238633013342236, + "learning_rate": 2.049477510275991e-05, + "loss": 0.1641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21490833163261414, + "step": 6895, + "valid_targets_mean": 2741.2, + "valid_targets_min": 1198 + }, + { + "epoch": 3.8037486218302092, + "grad_norm": 0.3087102194571276, + "learning_rate": 2.0467292747196516e-05, + "loss": 0.119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033327631652355194, + "step": 6900, + "valid_targets_mean": 4575.6, + "valid_targets_min": 1700 + }, + { + "epoch": 3.806504961411246, + "grad_norm": 0.34292312145775455, + "learning_rate": 2.04398095087804e-05, + "loss": 0.0896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03276168182492256, + "step": 6905, + "valid_targets_mean": 3423.9, + "valid_targets_min": 789 + }, + { + "epoch": 3.809261300992282, + "grad_norm": 0.397821481589785, + "learning_rate": 2.041232543943543e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05220165103673935, + "step": 6910, + "valid_targets_mean": 3709.8, + "valid_targets_min": 995 + }, + { + "epoch": 3.812017640573319, + "grad_norm": 0.33771039240599576, + "learning_rate": 2.038484059108707e-05, + "loss": 0.0878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03350666165351868, + "step": 6915, + "valid_targets_mean": 2735.9, + "valid_targets_min": 488 + }, + { + "epoch": 3.8147739801543548, + "grad_norm": 0.46607308929641567, + "learning_rate": 2.0357355015662228e-05, + "loss": 0.0899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05797896534204483, + "step": 6920, + "valid_targets_mean": 3915.0, + "valid_targets_min": 2437 + }, + { + "epoch": 3.8175303197353916, + "grad_norm": 0.4854273018021652, + "learning_rate": 2.0329868765089197e-05, + "loss": 0.0854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034125491976737976, + "step": 6925, + "valid_targets_mean": 1575.5, + "valid_targets_min": 727 + }, + { + "epoch": 3.8202866593164275, + "grad_norm": 0.4313894739359839, + "learning_rate": 2.0302381891297553e-05, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03989235684275627, + "step": 6930, + "valid_targets_mean": 3361.6, + "valid_targets_min": 1770 + }, + { + "epoch": 3.8230429988974644, + "grad_norm": 0.37672863974906423, + "learning_rate": 2.0274894446218042e-05, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047423094511032104, + "step": 6935, + "valid_targets_mean": 3829.2, + "valid_targets_min": 3085 + }, + { + "epoch": 3.8257993384785003, + "grad_norm": 0.34525451883459607, + "learning_rate": 2.024740648178249e-05, + "loss": 0.1358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031176216900348663, + "step": 6940, + "valid_targets_mean": 3627.6, + "valid_targets_min": 2924 + }, + { + "epoch": 3.828555678059537, + "grad_norm": 0.3807962571559121, + "learning_rate": 2.0219918049923707e-05, + "loss": 0.0693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03367726877331734, + "step": 6945, + "valid_targets_mean": 3040.9, + "valid_targets_min": 722 + }, + { + "epoch": 3.831312017640573, + "grad_norm": 0.362222477876988, + "learning_rate": 2.0192429202575372e-05, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025427933782339096, + "step": 6950, + "valid_targets_mean": 3535.1, + "valid_targets_min": 2535 + }, + { + "epoch": 3.83406835722161, + "grad_norm": 0.3933163530735051, + "learning_rate": 2.0164939991671976e-05, + "loss": 0.0802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03857443481683731, + "step": 6955, + "valid_targets_mean": 3227.9, + "valid_targets_min": 948 + }, + { + "epoch": 3.836824696802646, + "grad_norm": 0.5045075614569309, + "learning_rate": 2.0137450469148674e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03957320749759674, + "step": 6960, + "valid_targets_mean": 1027.5, + "valid_targets_min": 490 + }, + { + "epoch": 3.8395810363836826, + "grad_norm": 0.350060004521901, + "learning_rate": 2.0109960686941223e-05, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044740233570337296, + "step": 6965, + "valid_targets_mean": 3668.2, + "valid_targets_min": 2800 + }, + { + "epoch": 3.8423373759647186, + "grad_norm": 0.508687107184572, + "learning_rate": 2.0082470696985856e-05, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0697401612997055, + "step": 6970, + "valid_targets_mean": 1860.5, + "valid_targets_min": 799 + }, + { + "epoch": 3.8450937155457554, + "grad_norm": 0.45151481176135644, + "learning_rate": 2.005498055121921e-05, + "loss": 0.0955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04484109953045845, + "step": 6975, + "valid_targets_mean": 2671.0, + "valid_targets_min": 524 + }, + { + "epoch": 3.8478500551267913, + "grad_norm": 0.4266515391374727, + "learning_rate": 2.0027490301578217e-05, + "loss": 0.11, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05730786547064781, + "step": 6980, + "valid_targets_mean": 3287.6, + "valid_targets_min": 711 + }, + { + "epoch": 3.850606394707828, + "grad_norm": 0.3571717333434041, + "learning_rate": 2e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03596620261669159, + "step": 6985, + "valid_targets_mean": 2955.9, + "valid_targets_min": 2664 + }, + { + "epoch": 3.853362734288864, + "grad_norm": 0.38899247867427805, + "learning_rate": 1.997250969842179e-05, + "loss": 0.0943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039237797260284424, + "step": 6990, + "valid_targets_mean": 3090.8, + "valid_targets_min": 890 + }, + { + "epoch": 3.856119073869901, + "grad_norm": 0.5376850108567875, + "learning_rate": 1.9945019448780798e-05, + "loss": 0.1119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03905249759554863, + "step": 6995, + "valid_targets_mean": 3790.5, + "valid_targets_min": 2393 + }, + { + "epoch": 3.8588754134509373, + "grad_norm": 0.42559583597887907, + "learning_rate": 1.9917529303014148e-05, + "loss": 0.0985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05210528522729874, + "step": 7000, + "valid_targets_mean": 3415.4, + "valid_targets_min": 1169 + }, + { + "epoch": 3.8616317530319737, + "grad_norm": 0.4184131887946897, + "learning_rate": 1.9890039313058784e-05, + "loss": 0.0914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05213485658168793, + "step": 7005, + "valid_targets_mean": 5094.2, + "valid_targets_min": 3695 + }, + { + "epoch": 3.86438809261301, + "grad_norm": 0.4767031580190544, + "learning_rate": 1.986254953085133e-05, + "loss": 0.083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05157960578799248, + "step": 7010, + "valid_targets_mean": 2507.0, + "valid_targets_min": 596 + }, + { + "epoch": 3.8671444321940465, + "grad_norm": 0.28475739975483927, + "learning_rate": 1.9835060008328028e-05, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028322117403149605, + "step": 7015, + "valid_targets_mean": 4531.6, + "valid_targets_min": 3804 + }, + { + "epoch": 3.869900771775083, + "grad_norm": 0.3679026108000909, + "learning_rate": 1.9807570797424634e-05, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03763202950358391, + "step": 7020, + "valid_targets_mean": 3736.5, + "valid_targets_min": 2862 + }, + { + "epoch": 3.8726571113561192, + "grad_norm": 0.4716463014444333, + "learning_rate": 1.97800819500763e-05, + "loss": 0.1042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04827326536178589, + "step": 7025, + "valid_targets_mean": 1109.2, + "valid_targets_min": 757 + }, + { + "epoch": 3.8754134509371556, + "grad_norm": 0.4046375234366424, + "learning_rate": 1.9752593518217514e-05, + "loss": 0.0892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0461716502904892, + "step": 7030, + "valid_targets_mean": 3880.9, + "valid_targets_min": 2443 + }, + { + "epoch": 3.878169790518192, + "grad_norm": 0.6962279813451587, + "learning_rate": 1.972510555378196e-05, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05995872616767883, + "step": 7035, + "valid_targets_mean": 1380.8, + "valid_targets_min": 785 + }, + { + "epoch": 3.8809261300992284, + "grad_norm": 0.4904255001645391, + "learning_rate": 1.969761810870245e-05, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040304332971572876, + "step": 7040, + "valid_targets_mean": 3651.2, + "valid_targets_min": 905 + }, + { + "epoch": 3.8836824696802648, + "grad_norm": 0.4315504525477013, + "learning_rate": 1.967013123491081e-05, + "loss": 0.0928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05931123346090317, + "step": 7045, + "valid_targets_mean": 3003.0, + "valid_targets_min": 657 + }, + { + "epoch": 3.886438809261301, + "grad_norm": 0.41179740077341453, + "learning_rate": 1.964264498433778e-05, + "loss": 0.0914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04805886372923851, + "step": 7050, + "valid_targets_mean": 3722.6, + "valid_targets_min": 2768 + }, + { + "epoch": 3.8891951488423375, + "grad_norm": 0.5116515176485033, + "learning_rate": 1.9615159408912937e-05, + "loss": 0.0789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051784344017505646, + "step": 7055, + "valid_targets_mean": 1402.1, + "valid_targets_min": 627 + }, + { + "epoch": 3.891951488423374, + "grad_norm": 0.40256457016144126, + "learning_rate": 1.9587674560564573e-05, + "loss": 0.0904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04227212816476822, + "step": 7060, + "valid_targets_mean": 2333.2, + "valid_targets_min": 827 + }, + { + "epoch": 3.8947078280044103, + "grad_norm": 0.4560331702324507, + "learning_rate": 1.956019049121961e-05, + "loss": 0.0917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048387594521045685, + "step": 7065, + "valid_targets_mean": 2010.9, + "valid_targets_min": 940 + }, + { + "epoch": 3.8974641675854467, + "grad_norm": 0.38796608139964395, + "learning_rate": 1.9532707252803488e-05, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046201881021261215, + "step": 7070, + "valid_targets_mean": 3348.4, + "valid_targets_min": 1672 + }, + { + "epoch": 3.900220507166483, + "grad_norm": 0.39492981744489114, + "learning_rate": 1.9505224897240097e-05, + "loss": 0.0881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035610131919384, + "step": 7075, + "valid_targets_mean": 2466.5, + "valid_targets_min": 898 + }, + { + "epoch": 3.9029768467475194, + "grad_norm": 0.48203504134803477, + "learning_rate": 1.9477743476451644e-05, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04065325856208801, + "step": 7080, + "valid_targets_mean": 2059.2, + "valid_targets_min": 759 + }, + { + "epoch": 3.905733186328556, + "grad_norm": 0.5976112539211857, + "learning_rate": 1.9450263042358574e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07330667972564697, + "step": 7085, + "valid_targets_mean": 1602.0, + "valid_targets_min": 521 + }, + { + "epoch": 3.908489525909592, + "grad_norm": 0.4678125471701663, + "learning_rate": 1.942278364687947e-05, + "loss": 0.0925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05721612274646759, + "step": 7090, + "valid_targets_mean": 3223.2, + "valid_targets_min": 1032 + }, + { + "epoch": 3.9112458654906286, + "grad_norm": 0.2893128195514858, + "learning_rate": 1.939530534193094e-05, + "loss": 0.0769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029451623558998108, + "step": 7095, + "valid_targets_mean": 3639.2, + "valid_targets_min": 2741 + }, + { + "epoch": 3.914002205071665, + "grad_norm": 0.3583179807976411, + "learning_rate": 1.9367828179427553e-05, + "loss": 0.0853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03309059515595436, + "step": 7100, + "valid_targets_mean": 2828.0, + "valid_targets_min": 1692 + }, + { + "epoch": 3.9167585446527013, + "grad_norm": 0.4034751978357684, + "learning_rate": 1.9340352211281707e-05, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044495489448308945, + "step": 7105, + "valid_targets_mean": 3017.1, + "valid_targets_min": 935 + }, + { + "epoch": 3.9195148842337377, + "grad_norm": 0.36342445354253194, + "learning_rate": 1.931287748940354e-05, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026262428611516953, + "step": 7110, + "valid_targets_mean": 3058.1, + "valid_targets_min": 739 + }, + { + "epoch": 3.922271223814774, + "grad_norm": 0.3315314255841373, + "learning_rate": 1.9285404065700847e-05, + "loss": 0.089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03300041705369949, + "step": 7115, + "valid_targets_mean": 5307.2, + "valid_targets_min": 3282 + }, + { + "epoch": 3.9250275633958105, + "grad_norm": 0.2945808863420148, + "learning_rate": 1.9257931992078956e-05, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035855960100889206, + "step": 7120, + "valid_targets_mean": 4050.1, + "valid_targets_min": 995 + }, + { + "epoch": 3.927783902976847, + "grad_norm": 0.2676175820687097, + "learning_rate": 1.923046132044066e-05, + "loss": 0.0707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02531379647552967, + "step": 7125, + "valid_targets_mean": 4380.8, + "valid_targets_min": 949 + }, + { + "epoch": 3.9305402425578833, + "grad_norm": 0.4351478890471173, + "learning_rate": 1.9202992102686096e-05, + "loss": 0.0946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0610286146402359, + "step": 7130, + "valid_targets_mean": 4149.6, + "valid_targets_min": 2724 + }, + { + "epoch": 3.9332965821389196, + "grad_norm": 0.33404938015819025, + "learning_rate": 1.9175524390712645e-05, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029404640197753906, + "step": 7135, + "valid_targets_mean": 3625.5, + "valid_targets_min": 424 + }, + { + "epoch": 3.936052921719956, + "grad_norm": 0.3497354711125289, + "learning_rate": 1.914805823641486e-05, + "loss": 0.1127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03452690318226814, + "step": 7140, + "valid_targets_mean": 3225.6, + "valid_targets_min": 531 + }, + { + "epoch": 3.9388092613009924, + "grad_norm": 0.3767757076419721, + "learning_rate": 1.9120593691684335e-05, + "loss": 0.1096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04007373005151749, + "step": 7145, + "valid_targets_mean": 2986.6, + "valid_targets_min": 855 + }, + { + "epoch": 3.941565600882029, + "grad_norm": 0.36031982668497287, + "learning_rate": 1.909313080840964e-05, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03912299498915672, + "step": 7150, + "valid_targets_mean": 3038.5, + "valid_targets_min": 722 + }, + { + "epoch": 3.944321940463065, + "grad_norm": 0.3572096481427015, + "learning_rate": 1.9065669638476195e-05, + "loss": 0.1107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03597118705511093, + "step": 7155, + "valid_targets_mean": 3720.9, + "valid_targets_min": 1251 + }, + { + "epoch": 3.9470782800441016, + "grad_norm": 0.3840590371294833, + "learning_rate": 1.903821023376618e-05, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05131332203745842, + "step": 7160, + "valid_targets_mean": 3339.5, + "valid_targets_min": 1382 + }, + { + "epoch": 3.949834619625138, + "grad_norm": 0.4809661863006892, + "learning_rate": 1.9010752646158447e-05, + "loss": 0.0935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035056039690971375, + "step": 7165, + "valid_targets_mean": 1042.4, + "valid_targets_min": 595 + }, + { + "epoch": 3.9525909592061743, + "grad_norm": 0.5256532250763575, + "learning_rate": 1.8983296927528413e-05, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061255961656570435, + "step": 7170, + "valid_targets_mean": 1326.4, + "valid_targets_min": 520 + }, + { + "epoch": 3.9553472987872107, + "grad_norm": 0.38240782558906855, + "learning_rate": 1.8955843129747965e-05, + "loss": 0.1233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03645171597599983, + "step": 7175, + "valid_targets_mean": 3413.5, + "valid_targets_min": 867 + }, + { + "epoch": 3.958103638368247, + "grad_norm": 0.3449069822910184, + "learning_rate": 1.8928391304685363e-05, + "loss": 0.0912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03258441016077995, + "step": 7180, + "valid_targets_mean": 3746.5, + "valid_targets_min": 3009 + }, + { + "epoch": 3.9608599779492835, + "grad_norm": 0.4095996924560781, + "learning_rate": 1.8900941504205133e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046572767198085785, + "step": 7185, + "valid_targets_mean": 3094.6, + "valid_targets_min": 676 + }, + { + "epoch": 3.96361631753032, + "grad_norm": 0.44510480950335474, + "learning_rate": 1.8873493780167972e-05, + "loss": 0.0864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03528179973363876, + "step": 7190, + "valid_targets_mean": 1820.2, + "valid_targets_min": 616 + }, + { + "epoch": 3.9663726571113562, + "grad_norm": 0.4432821023610876, + "learning_rate": 1.8846048184430677e-05, + "loss": 0.0844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041720904409885406, + "step": 7195, + "valid_targets_mean": 2719.1, + "valid_targets_min": 541 + }, + { + "epoch": 3.9691289966923926, + "grad_norm": 0.6744602570357758, + "learning_rate": 1.8818604768846003e-05, + "loss": 0.0798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053586315363645554, + "step": 7200, + "valid_targets_mean": 992.2, + "valid_targets_min": 758 + }, + { + "epoch": 3.971885336273429, + "grad_norm": 0.39987231360523523, + "learning_rate": 1.8791163585262588e-05, + "loss": 0.0977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04073847830295563, + "step": 7205, + "valid_targets_mean": 2946.1, + "valid_targets_min": 1101 + }, + { + "epoch": 3.9746416758544654, + "grad_norm": 0.45762121894288915, + "learning_rate": 1.876372468552486e-05, + "loss": 0.0805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0455935075879097, + "step": 7210, + "valid_targets_mean": 3203.0, + "valid_targets_min": 738 + }, + { + "epoch": 3.9773980154355018, + "grad_norm": 0.36643098775640687, + "learning_rate": 1.8736288121472927e-05, + "loss": 0.0827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046014755964279175, + "step": 7215, + "valid_targets_mean": 3274.0, + "valid_targets_min": 1006 + }, + { + "epoch": 3.980154355016538, + "grad_norm": 0.3872223634022616, + "learning_rate": 1.8708853944942486e-05, + "loss": 0.0955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04911606013774872, + "step": 7220, + "valid_targets_mean": 4115.1, + "valid_targets_min": 2773 + }, + { + "epoch": 3.9829106945975745, + "grad_norm": 0.248167145241782, + "learning_rate": 1.868142220776473e-05, + "loss": 0.0762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02858860045671463, + "step": 7225, + "valid_targets_mean": 5720.0, + "valid_targets_min": 3677 + }, + { + "epoch": 3.985667034178611, + "grad_norm": 0.44521265968197765, + "learning_rate": 1.865399296176623e-05, + "loss": 0.0812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04739736393094063, + "step": 7230, + "valid_targets_mean": 2422.5, + "valid_targets_min": 675 + }, + { + "epoch": 3.9884233737596473, + "grad_norm": 0.34625208309687805, + "learning_rate": 1.862656625876886e-05, + "loss": 0.0799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03521757200360298, + "step": 7235, + "valid_targets_mean": 3946.6, + "valid_targets_min": 795 + }, + { + "epoch": 3.9911797133406837, + "grad_norm": 0.5007830804370347, + "learning_rate": 1.8599142150589682e-05, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08738721907138824, + "step": 7240, + "valid_targets_mean": 3114.8, + "valid_targets_min": 1060 + }, + { + "epoch": 3.99393605292172, + "grad_norm": 0.33418732986029803, + "learning_rate": 1.8571720689040867e-05, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03781001269817352, + "step": 7245, + "valid_targets_mean": 3099.6, + "valid_targets_min": 985 + }, + { + "epoch": 3.9966923925027564, + "grad_norm": 0.3938563102203653, + "learning_rate": 1.854430192592958e-05, + "loss": 0.0907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05036578327417374, + "step": 7250, + "valid_targets_mean": 2988.1, + "valid_targets_min": 878 + }, + { + "epoch": 3.999448732083793, + "grad_norm": 0.613885962070175, + "learning_rate": 1.8516885913057886e-05, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12135718017816544, + "step": 7255, + "valid_targets_mean": 2265.0, + "valid_targets_min": 1367 + }, + { + "epoch": 4.002205071664829, + "grad_norm": 0.5868862312247577, + "learning_rate": 1.848947270222265e-05, + "loss": 0.1778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15168806910514832, + "step": 7260, + "valid_targets_mean": 8029.9, + "valid_targets_min": 6142 + }, + { + "epoch": 4.004961411245866, + "grad_norm": 0.4109924837700702, + "learning_rate": 1.8462062345215447e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09119362384080887, + "step": 7265, + "valid_targets_mean": 6664.0, + "valid_targets_min": 5470 + }, + { + "epoch": 4.0077177508269015, + "grad_norm": 0.39875415415587445, + "learning_rate": 1.8434654893822474e-05, + "loss": 0.1832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09609955549240112, + "step": 7270, + "valid_targets_mean": 7126.9, + "valid_targets_min": 5368 + }, + { + "epoch": 4.010474090407938, + "grad_norm": 0.36362375422212684, + "learning_rate": 1.8407250399824416e-05, + "loss": 0.1775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08652584254741669, + "step": 7275, + "valid_targets_mean": 7634.9, + "valid_targets_min": 5614 + }, + { + "epoch": 4.013230429988974, + "grad_norm": 0.34560739753872555, + "learning_rate": 1.837984891499638e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08456164598464966, + "step": 7280, + "valid_targets_mean": 8630.8, + "valid_targets_min": 6362 + }, + { + "epoch": 4.015986769570011, + "grad_norm": 0.38349760331462707, + "learning_rate": 1.8352450491107793e-05, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0923943817615509, + "step": 7285, + "valid_targets_mean": 7383.6, + "valid_targets_min": 5844 + }, + { + "epoch": 4.018743109151047, + "grad_norm": 0.3831180124623771, + "learning_rate": 1.8325055179922285e-05, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10637330263853073, + "step": 7290, + "valid_targets_mean": 8017.0, + "valid_targets_min": 5283 + }, + { + "epoch": 4.021499448732084, + "grad_norm": 0.36140345611440117, + "learning_rate": 1.829766303319763e-05, + "loss": 0.1821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07794812321662903, + "step": 7295, + "valid_targets_mean": 7051.1, + "valid_targets_min": 4135 + }, + { + "epoch": 4.02425578831312, + "grad_norm": 0.3975724014126985, + "learning_rate": 1.82702741026856e-05, + "loss": 0.1776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07743655145168304, + "step": 7300, + "valid_targets_mean": 5370.4, + "valid_targets_min": 3919 + }, + { + "epoch": 4.027012127894157, + "grad_norm": 0.378949447700742, + "learning_rate": 1.8242888440131894e-05, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07727095484733582, + "step": 7305, + "valid_targets_mean": 6063.6, + "valid_targets_min": 4863 + }, + { + "epoch": 4.029768467475193, + "grad_norm": 0.4183455868707706, + "learning_rate": 1.8215506097276047e-05, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08854006230831146, + "step": 7310, + "valid_targets_mean": 7001.2, + "valid_targets_min": 6020 + }, + { + "epoch": 4.032524807056229, + "grad_norm": 0.3821166590899154, + "learning_rate": 1.8188127125851315e-05, + "loss": 0.1724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0858040452003479, + "step": 7315, + "valid_targets_mean": 6181.9, + "valid_targets_min": 4669 + }, + { + "epoch": 4.035281146637265, + "grad_norm": 0.3671118003691682, + "learning_rate": 1.8160751577584587e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07912043482065201, + "step": 7320, + "valid_targets_mean": 6250.2, + "valid_targets_min": 4900 + }, + { + "epoch": 4.038037486218302, + "grad_norm": 0.3739615773572713, + "learning_rate": 1.8133379504196288e-05, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07736102491617203, + "step": 7325, + "valid_targets_mean": 7156.4, + "valid_targets_min": 5323 + }, + { + "epoch": 4.040793825799338, + "grad_norm": 0.39173118085355657, + "learning_rate": 1.8106010957400263e-05, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08898690342903137, + "step": 7330, + "valid_targets_mean": 7215.8, + "valid_targets_min": 4720 + }, + { + "epoch": 4.043550165380375, + "grad_norm": 0.3739519594359209, + "learning_rate": 1.807864598890371e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0858980268239975, + "step": 7335, + "valid_targets_mean": 6550.8, + "valid_targets_min": 5739 + }, + { + "epoch": 4.046306504961411, + "grad_norm": 0.36759555722743276, + "learning_rate": 1.805128465040706e-05, + "loss": 0.1712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08817552775144577, + "step": 7340, + "valid_targets_mean": 6726.8, + "valid_targets_min": 4540 + }, + { + "epoch": 4.049062844542448, + "grad_norm": 0.47682025576408227, + "learning_rate": 1.8023926993603894e-05, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09653215110301971, + "step": 7345, + "valid_targets_mean": 4573.4, + "valid_targets_min": 1039 + }, + { + "epoch": 4.051819184123484, + "grad_norm": 0.43815482834179914, + "learning_rate": 1.7996573070180823e-05, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08916212618350983, + "step": 7350, + "valid_targets_mean": 6430.0, + "valid_targets_min": 5412 + }, + { + "epoch": 4.0545755237045205, + "grad_norm": 0.6288613719590671, + "learning_rate": 1.7969222931817412e-05, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02580735832452774, + "step": 7355, + "valid_targets_mean": 583.1, + "valid_targets_min": 158 + }, + { + "epoch": 4.057331863285556, + "grad_norm": 0.3871618239534102, + "learning_rate": 1.7941876630186078e-05, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09024959802627563, + "step": 7360, + "valid_targets_mean": 8230.9, + "valid_targets_min": 5061 + }, + { + "epoch": 4.060088202866593, + "grad_norm": 0.40947017137915137, + "learning_rate": 1.7914534216951988e-05, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09496884793043137, + "step": 7365, + "valid_targets_mean": 7238.9, + "valid_targets_min": 5112 + }, + { + "epoch": 4.062844542447629, + "grad_norm": 0.3924518153907183, + "learning_rate": 1.7887195743772953e-05, + "loss": 0.1714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07767713069915771, + "step": 7370, + "valid_targets_mean": 6322.8, + "valid_targets_min": 4867 + }, + { + "epoch": 4.065600882028666, + "grad_norm": 0.36532342620783315, + "learning_rate": 1.785986126229936e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08819594979286194, + "step": 7375, + "valid_targets_mean": 7887.0, + "valid_targets_min": 6004 + }, + { + "epoch": 4.068357221609702, + "grad_norm": 0.34929758034000724, + "learning_rate": 1.783253082417403e-05, + "loss": 0.1712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07817757874727249, + "step": 7380, + "valid_targets_mean": 7090.2, + "valid_targets_min": 4917 + }, + { + "epoch": 4.071113561190739, + "grad_norm": 0.44519482222591406, + "learning_rate": 1.780520448103216e-05, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08096392452716827, + "step": 7385, + "valid_targets_mean": 7441.9, + "valid_targets_min": 4202 + }, + { + "epoch": 4.073869900771775, + "grad_norm": 0.3791984089359034, + "learning_rate": 1.7777882284501215e-05, + "loss": 0.1523, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07987791299819946, + "step": 7390, + "valid_targets_mean": 7901.8, + "valid_targets_min": 5060 + }, + { + "epoch": 4.0766262403528115, + "grad_norm": 0.3505124878557592, + "learning_rate": 1.7750564286200816e-05, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07280844449996948, + "step": 7395, + "valid_targets_mean": 8557.6, + "valid_targets_min": 6331 + }, + { + "epoch": 4.0793825799338475, + "grad_norm": 0.43166906872600114, + "learning_rate": 1.772325053774265e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08392757177352905, + "step": 7400, + "valid_targets_mean": 4976.4, + "valid_targets_min": 2847 + }, + { + "epoch": 4.082138919514884, + "grad_norm": 0.42442475646061834, + "learning_rate": 1.7695941090730385e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09423413127660751, + "step": 7405, + "valid_targets_mean": 7129.1, + "valid_targets_min": 5331 + }, + { + "epoch": 4.08489525909592, + "grad_norm": 0.4274780936696062, + "learning_rate": 1.7668635996759547e-05, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09242204576730728, + "step": 7410, + "valid_targets_mean": 6864.0, + "valid_targets_min": 4697 + }, + { + "epoch": 4.087651598676957, + "grad_norm": 0.40341253388301845, + "learning_rate": 1.7641335307417465e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08589634299278259, + "step": 7415, + "valid_targets_mean": 6631.9, + "valid_targets_min": 4623 + }, + { + "epoch": 4.090407938257993, + "grad_norm": 0.3979400664333842, + "learning_rate": 1.7614039074283117e-05, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08108409494161606, + "step": 7420, + "valid_targets_mean": 6330.4, + "valid_targets_min": 5424 + }, + { + "epoch": 4.09316427783903, + "grad_norm": 0.3816674694896844, + "learning_rate": 1.7586747348927075e-05, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08219548314809799, + "step": 7425, + "valid_targets_mean": 6814.8, + "valid_targets_min": 4964 + }, + { + "epoch": 4.095920617420066, + "grad_norm": 0.4231061512230252, + "learning_rate": 1.7559460182911396e-05, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09714823961257935, + "step": 7430, + "valid_targets_mean": 6915.8, + "valid_targets_min": 5321 + }, + { + "epoch": 4.098676957001103, + "grad_norm": 0.388301821556405, + "learning_rate": 1.7532177627789516e-05, + "loss": 0.1727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08823524415493011, + "step": 7435, + "valid_targets_mean": 6464.5, + "valid_targets_min": 5185 + }, + { + "epoch": 4.1014332965821385, + "grad_norm": 0.40708498261459614, + "learning_rate": 1.7504899735106173e-05, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08980473130941391, + "step": 7440, + "valid_targets_mean": 7048.6, + "valid_targets_min": 4913 + }, + { + "epoch": 4.104189636163175, + "grad_norm": 0.39151103334745724, + "learning_rate": 1.7477626556397283e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09863029420375824, + "step": 7445, + "valid_targets_mean": 7469.2, + "valid_targets_min": 5426 + }, + { + "epoch": 4.106945975744211, + "grad_norm": 0.42991690816691247, + "learning_rate": 1.745035814318986e-05, + "loss": 0.1801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09425477683544159, + "step": 7450, + "valid_targets_mean": 6819.4, + "valid_targets_min": 5286 + }, + { + "epoch": 4.109702315325248, + "grad_norm": 0.377702277998009, + "learning_rate": 1.7423094547001913e-05, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09287574142217636, + "step": 7455, + "valid_targets_mean": 7107.5, + "valid_targets_min": 5201 + }, + { + "epoch": 4.112458654906284, + "grad_norm": 0.3962025114437257, + "learning_rate": 1.7395835819342355e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08525830507278442, + "step": 7460, + "valid_targets_mean": 5642.6, + "valid_targets_min": 4838 + }, + { + "epoch": 4.115214994487321, + "grad_norm": 0.7952582233400199, + "learning_rate": 1.7368582011710904e-05, + "loss": 0.1441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14508584141731262, + "step": 7465, + "valid_targets_mean": 2054.1, + "valid_targets_min": 134 + }, + { + "epoch": 4.117971334068357, + "grad_norm": 0.42725856807649715, + "learning_rate": 1.734133317559797e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08205440640449524, + "step": 7470, + "valid_targets_mean": 6686.4, + "valid_targets_min": 5256 + }, + { + "epoch": 4.120727673649394, + "grad_norm": 0.40079896544849697, + "learning_rate": 1.731408936248458e-05, + "loss": 0.172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09299857169389725, + "step": 7475, + "valid_targets_mean": 6216.6, + "valid_targets_min": 5542 + }, + { + "epoch": 4.12348401323043, + "grad_norm": 0.3807103892673187, + "learning_rate": 1.7286850623842258e-05, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08683011680841446, + "step": 7480, + "valid_targets_mean": 6486.1, + "valid_targets_min": 5531 + }, + { + "epoch": 4.126240352811466, + "grad_norm": 0.3993026258895936, + "learning_rate": 1.7259617011132967e-05, + "loss": 0.173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08637335896492004, + "step": 7485, + "valid_targets_mean": 5797.4, + "valid_targets_min": 3911 + }, + { + "epoch": 4.128996692392502, + "grad_norm": 0.41450772354665877, + "learning_rate": 1.7232388575808964e-05, + "loss": 0.1713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08757546544075012, + "step": 7490, + "valid_targets_mean": 6357.5, + "valid_targets_min": 5195 + }, + { + "epoch": 4.131753031973539, + "grad_norm": 0.4197591964186917, + "learning_rate": 1.7205165369312728e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08856083452701569, + "step": 7495, + "valid_targets_mean": 5808.9, + "valid_targets_min": 4959 + }, + { + "epoch": 4.134509371554575, + "grad_norm": 0.5207986919707032, + "learning_rate": 1.7177947443076862e-05, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05734080448746681, + "step": 7500, + "valid_targets_mean": 2591.5, + "valid_targets_min": 842 + }, + { + "epoch": 4.137265711135612, + "grad_norm": 0.5069242556926572, + "learning_rate": 1.715073484852399e-05, + "loss": 0.1454, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08351719379425049, + "step": 7505, + "valid_targets_mean": 6006.9, + "valid_targets_min": 4769 + }, + { + "epoch": 4.140022050716648, + "grad_norm": 0.4157285900767651, + "learning_rate": 1.7123527637066673e-05, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08392911404371262, + "step": 7510, + "valid_targets_mean": 6764.2, + "valid_targets_min": 5046 + }, + { + "epoch": 4.142778390297685, + "grad_norm": 0.4400235395374115, + "learning_rate": 1.7096325860107286e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09232664108276367, + "step": 7515, + "valid_targets_mean": 6055.8, + "valid_targets_min": 5182 + }, + { + "epoch": 4.145534729878721, + "grad_norm": 0.3961299494278715, + "learning_rate": 1.7069129569037955e-05, + "loss": 0.1706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08667701482772827, + "step": 7520, + "valid_targets_mean": 7101.6, + "valid_targets_min": 5366 + }, + { + "epoch": 4.1482910694597575, + "grad_norm": 0.38607457185746913, + "learning_rate": 1.704193881524042e-05, + "loss": 0.1674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08817920088768005, + "step": 7525, + "valid_targets_mean": 7111.9, + "valid_targets_min": 5192 + }, + { + "epoch": 4.151047409040794, + "grad_norm": 0.42018368638225184, + "learning_rate": 1.7014753650085972e-05, + "loss": 0.1733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0860932469367981, + "step": 7530, + "valid_targets_mean": 5509.9, + "valid_targets_min": 4790 + }, + { + "epoch": 4.15380374862183, + "grad_norm": 0.8687851283530922, + "learning_rate": 1.6987574124935353e-05, + "loss": 0.1719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09845466911792755, + "step": 7535, + "valid_targets_mean": 1581.9, + "valid_targets_min": 864 + }, + { + "epoch": 4.156560088202866, + "grad_norm": 1.1835269279121519, + "learning_rate": 1.6960400291138625e-05, + "loss": 0.1751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07912611961364746, + "step": 7540, + "valid_targets_mean": 1333.2, + "valid_targets_min": 632 + }, + { + "epoch": 4.159316427783903, + "grad_norm": 0.9600808335987302, + "learning_rate": 1.693323220003512e-05, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08147436380386353, + "step": 7545, + "valid_targets_mean": 1602.9, + "valid_targets_min": 875 + }, + { + "epoch": 4.16207276736494, + "grad_norm": 0.9590594501141085, + "learning_rate": 1.690606990295331e-05, + "loss": 0.1582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08258432149887085, + "step": 7550, + "valid_targets_mean": 1340.4, + "valid_targets_min": 701 + }, + { + "epoch": 4.164829106945976, + "grad_norm": 0.7869306415627356, + "learning_rate": 1.6878913451210715e-05, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09345021098852158, + "step": 7555, + "valid_targets_mean": 1792.9, + "valid_targets_min": 898 + }, + { + "epoch": 4.167585446527012, + "grad_norm": 0.7691832039337874, + "learning_rate": 1.6851762896113827e-05, + "loss": 0.1535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07955098152160645, + "step": 7560, + "valid_targets_mean": 1433.1, + "valid_targets_min": 1070 + }, + { + "epoch": 4.1703417861080485, + "grad_norm": 0.7420800723203675, + "learning_rate": 1.682461828895799e-05, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07861204445362091, + "step": 7565, + "valid_targets_mean": 1589.5, + "valid_targets_min": 970 + }, + { + "epoch": 4.173098125689085, + "grad_norm": 0.7406691807803736, + "learning_rate": 1.67974796810273e-05, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06828086078166962, + "step": 7570, + "valid_targets_mean": 1270.5, + "valid_targets_min": 648 + }, + { + "epoch": 4.175854465270121, + "grad_norm": 0.7666410227020131, + "learning_rate": 1.6770347123594527e-05, + "loss": 0.1623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0879657045006752, + "step": 7575, + "valid_targets_mean": 1674.5, + "valid_targets_min": 727 + }, + { + "epoch": 4.178610804851157, + "grad_norm": 0.7237221769620675, + "learning_rate": 1.674322066792102e-05, + "loss": 0.1427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059045881032943726, + "step": 7580, + "valid_targets_mean": 1406.0, + "valid_targets_min": 770 + }, + { + "epoch": 4.181367144432194, + "grad_norm": 0.7775472477501587, + "learning_rate": 1.6716100365256593e-05, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06578196585178375, + "step": 7585, + "valid_targets_mean": 1381.2, + "valid_targets_min": 572 + }, + { + "epoch": 4.184123484013231, + "grad_norm": 0.7651294531083268, + "learning_rate": 1.668898626683942e-05, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09009277820587158, + "step": 7590, + "valid_targets_mean": 1580.4, + "valid_targets_min": 633 + }, + { + "epoch": 4.186879823594267, + "grad_norm": 0.7599234522742094, + "learning_rate": 1.6661878423895975e-05, + "loss": 0.1481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07442671060562134, + "step": 7595, + "valid_targets_mean": 1478.8, + "valid_targets_min": 923 + }, + { + "epoch": 4.189636163175303, + "grad_norm": 1.0250055249006012, + "learning_rate": 1.6634776887640902e-05, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0690312534570694, + "step": 7600, + "valid_targets_mean": 1401.0, + "valid_targets_min": 740 + }, + { + "epoch": 4.19239250275634, + "grad_norm": 0.8022798063569384, + "learning_rate": 1.6607681709276938e-05, + "loss": 0.1415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07062997668981552, + "step": 7605, + "valid_targets_mean": 1588.9, + "valid_targets_min": 870 + }, + { + "epoch": 4.195148842337376, + "grad_norm": 0.7418310309592586, + "learning_rate": 1.6580592939994796e-05, + "loss": 0.1422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10092377662658691, + "step": 7610, + "valid_targets_mean": 2008.4, + "valid_targets_min": 1042 + }, + { + "epoch": 4.197905181918412, + "grad_norm": 0.7421471654056028, + "learning_rate": 1.655351063097309e-05, + "loss": 0.1517, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06037698686122894, + "step": 7615, + "valid_targets_mean": 1209.0, + "valid_targets_min": 870 + }, + { + "epoch": 4.200661521499449, + "grad_norm": 0.7547123771721225, + "learning_rate": 1.6526434833378232e-05, + "loss": 0.1515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07588573545217514, + "step": 7620, + "valid_targets_mean": 1535.5, + "valid_targets_min": 740 + }, + { + "epoch": 4.203417861080485, + "grad_norm": 0.6618484725825743, + "learning_rate": 1.649936559836431e-05, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07391175627708435, + "step": 7625, + "valid_targets_mean": 1642.9, + "valid_targets_min": 906 + }, + { + "epoch": 4.206174200661522, + "grad_norm": 0.7761842422523347, + "learning_rate": 1.6472302977073044e-05, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07683582603931427, + "step": 7630, + "valid_targets_mean": 1570.5, + "valid_targets_min": 814 + }, + { + "epoch": 4.208930540242558, + "grad_norm": 0.7151940273240022, + "learning_rate": 1.644524702063364e-05, + "loss": 0.1497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06618554145097733, + "step": 7635, + "valid_targets_mean": 1722.1, + "valid_targets_min": 832 + }, + { + "epoch": 4.211686879823595, + "grad_norm": 0.6731778804725842, + "learning_rate": 1.6418197780162718e-05, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05793902650475502, + "step": 7640, + "valid_targets_mean": 1453.5, + "valid_targets_min": 891 + }, + { + "epoch": 4.214443219404631, + "grad_norm": 0.8459947331395259, + "learning_rate": 1.63911553067642e-05, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09052550047636032, + "step": 7645, + "valid_targets_mean": 1584.4, + "valid_targets_min": 734 + }, + { + "epoch": 4.2171995589856675, + "grad_norm": 0.773772095751057, + "learning_rate": 1.6364119651529232e-05, + "loss": 0.136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06478758156299591, + "step": 7650, + "valid_targets_mean": 1495.2, + "valid_targets_min": 737 + }, + { + "epoch": 4.219955898566703, + "grad_norm": 0.7783194889456755, + "learning_rate": 1.633709086553609e-05, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0804729163646698, + "step": 7655, + "valid_targets_mean": 1737.0, + "valid_targets_min": 872 + }, + { + "epoch": 4.22271223814774, + "grad_norm": 0.7951170994000503, + "learning_rate": 1.6310068999850045e-05, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07339446246623993, + "step": 7660, + "valid_targets_mean": 1431.6, + "valid_targets_min": 805 + }, + { + "epoch": 4.225468577728776, + "grad_norm": 0.7286221019824687, + "learning_rate": 1.6283054105523315e-05, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08249394595623016, + "step": 7665, + "valid_targets_mean": 2060.2, + "valid_targets_min": 1253 + }, + { + "epoch": 4.228224917309813, + "grad_norm": 0.7524542998277257, + "learning_rate": 1.6256046233594937e-05, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08828938007354736, + "step": 7670, + "valid_targets_mean": 2063.9, + "valid_targets_min": 1264 + }, + { + "epoch": 4.230981256890849, + "grad_norm": 0.7322072344702071, + "learning_rate": 1.622904543509068e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05442691594362259, + "step": 7675, + "valid_targets_mean": 1036.5, + "valid_targets_min": 613 + }, + { + "epoch": 4.233737596471886, + "grad_norm": 0.6988933353326047, + "learning_rate": 1.620205176102296e-05, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06954734772443771, + "step": 7680, + "valid_targets_mean": 1674.4, + "valid_targets_min": 768 + }, + { + "epoch": 4.236493936052922, + "grad_norm": 0.7917272139128514, + "learning_rate": 1.6175065262390724e-05, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06278039515018463, + "step": 7685, + "valid_targets_mean": 1249.8, + "valid_targets_min": 714 + }, + { + "epoch": 4.2392502756339585, + "grad_norm": 0.7671666239918831, + "learning_rate": 1.614808599017936e-05, + "loss": 0.1517, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053422339260578156, + "step": 7690, + "valid_targets_mean": 1114.8, + "valid_targets_min": 800 + }, + { + "epoch": 4.2420066152149944, + "grad_norm": 0.7296560336332643, + "learning_rate": 1.612111399536061e-05, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08131042122840881, + "step": 7695, + "valid_targets_mean": 1780.9, + "valid_targets_min": 933 + }, + { + "epoch": 4.244762954796031, + "grad_norm": 0.8281371003408712, + "learning_rate": 1.609414932889246e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06340353190898895, + "step": 7700, + "valid_targets_mean": 1113.6, + "valid_targets_min": 746 + }, + { + "epoch": 4.247519294377067, + "grad_norm": 0.8068755901718376, + "learning_rate": 1.6067192041719065e-05, + "loss": 0.13, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0721794068813324, + "step": 7705, + "valid_targets_mean": 1265.0, + "valid_targets_min": 814 + }, + { + "epoch": 4.250275633958104, + "grad_norm": 0.7615283452421395, + "learning_rate": 1.604024218477062e-05, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08977312594652176, + "step": 7710, + "valid_targets_mean": 1652.1, + "valid_targets_min": 635 + }, + { + "epoch": 4.25303197353914, + "grad_norm": 0.7764656654587005, + "learning_rate": 1.601329980896329e-05, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05557342618703842, + "step": 7715, + "valid_targets_mean": 1231.2, + "valid_targets_min": 739 + }, + { + "epoch": 4.255788313120177, + "grad_norm": 0.7189732497862915, + "learning_rate": 1.59863649651991e-05, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08693213015794754, + "step": 7720, + "valid_targets_mean": 1682.4, + "valid_targets_min": 831 + }, + { + "epoch": 4.258544652701213, + "grad_norm": 0.8200966361830749, + "learning_rate": 1.5959437704365866e-05, + "loss": 0.1441, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07333334535360336, + "step": 7725, + "valid_targets_mean": 1571.6, + "valid_targets_min": 1015 + }, + { + "epoch": 4.26130099228225, + "grad_norm": 0.6981616778500295, + "learning_rate": 1.5932518077337055e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0621609166264534, + "step": 7730, + "valid_targets_mean": 1276.8, + "valid_targets_min": 711 + }, + { + "epoch": 4.2640573318632855, + "grad_norm": 0.8143791126836486, + "learning_rate": 1.590560613497172e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046325407922267914, + "step": 7735, + "valid_targets_mean": 1147.5, + "valid_targets_min": 733 + }, + { + "epoch": 4.266813671444322, + "grad_norm": 0.830457143589052, + "learning_rate": 1.587870192811439e-05, + "loss": 0.1437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09101492911577225, + "step": 7740, + "valid_targets_mean": 1604.0, + "valid_targets_min": 953 + }, + { + "epoch": 4.269570011025358, + "grad_norm": 0.7218066528719973, + "learning_rate": 1.5851805507594987e-05, + "loss": 0.1491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.075267493724823, + "step": 7745, + "valid_targets_mean": 1784.0, + "valid_targets_min": 1011 + }, + { + "epoch": 4.272326350606395, + "grad_norm": 0.7782490066373582, + "learning_rate": 1.582491692422872e-05, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08508297801017761, + "step": 7750, + "valid_targets_mean": 1729.0, + "valid_targets_min": 1143 + }, + { + "epoch": 4.275082690187431, + "grad_norm": 0.7783264408942374, + "learning_rate": 1.5798036228815988e-05, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06350135058164597, + "step": 7755, + "valid_targets_mean": 1483.9, + "valid_targets_min": 699 + }, + { + "epoch": 4.277839029768468, + "grad_norm": 0.7652960817419282, + "learning_rate": 1.57711634721423e-05, + "loss": 0.1337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0602324903011322, + "step": 7760, + "valid_targets_mean": 1198.4, + "valid_targets_min": 680 + }, + { + "epoch": 4.280595369349504, + "grad_norm": 0.8220790399186273, + "learning_rate": 1.5744298704978135e-05, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07067703455686569, + "step": 7765, + "valid_targets_mean": 1365.2, + "valid_targets_min": 687 + }, + { + "epoch": 4.283351708930541, + "grad_norm": 0.8258287963284947, + "learning_rate": 1.5717441978078914e-05, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056421615183353424, + "step": 7770, + "valid_targets_mean": 1142.9, + "valid_targets_min": 835 + }, + { + "epoch": 4.286108048511577, + "grad_norm": 0.7807207365038205, + "learning_rate": 1.5690593342184854e-05, + "loss": 0.1333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06151098012924194, + "step": 7775, + "valid_targets_mean": 1316.9, + "valid_targets_min": 829 + }, + { + "epoch": 4.288864388092613, + "grad_norm": 0.7632943083470607, + "learning_rate": 1.5663752848020875e-05, + "loss": 0.1359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07813870906829834, + "step": 7780, + "valid_targets_mean": 1739.2, + "valid_targets_min": 1138 + }, + { + "epoch": 4.291620727673649, + "grad_norm": 0.7292445057454993, + "learning_rate": 1.5636920546296533e-05, + "loss": 0.1428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06419949233531952, + "step": 7785, + "valid_targets_mean": 1592.2, + "valid_targets_min": 1174 + }, + { + "epoch": 4.294377067254686, + "grad_norm": 0.7779092142839422, + "learning_rate": 1.561009648770589e-05, + "loss": 0.1433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07135245949029922, + "step": 7790, + "valid_targets_mean": 1543.1, + "valid_targets_min": 554 + }, + { + "epoch": 4.297133406835722, + "grad_norm": 0.7553552027395628, + "learning_rate": 1.5583280722927437e-05, + "loss": 0.1373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06733568757772446, + "step": 7795, + "valid_targets_mean": 1430.0, + "valid_targets_min": 843 + }, + { + "epoch": 4.299889746416759, + "grad_norm": 0.6430602286567954, + "learning_rate": 1.5556473302624017e-05, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06489567458629608, + "step": 7800, + "valid_targets_mean": 1622.9, + "valid_targets_min": 849 + }, + { + "epoch": 4.302646085997795, + "grad_norm": 0.7617245786655169, + "learning_rate": 1.5529674277442674e-05, + "loss": 0.1422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06259912252426147, + "step": 7805, + "valid_targets_mean": 1429.0, + "valid_targets_min": 612 + }, + { + "epoch": 4.305402425578832, + "grad_norm": 0.7902842698833721, + "learning_rate": 1.5502883698014614e-05, + "loss": 0.1406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05701011046767235, + "step": 7810, + "valid_targets_mean": 1283.0, + "valid_targets_min": 782 + }, + { + "epoch": 4.308158765159868, + "grad_norm": 0.8208516039839285, + "learning_rate": 1.547610161495508e-05, + "loss": 0.1377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0798402726650238, + "step": 7815, + "valid_targets_mean": 1457.0, + "valid_targets_min": 926 + }, + { + "epoch": 4.3109151047409044, + "grad_norm": 0.7448401874013599, + "learning_rate": 1.5449328078863265e-05, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05994705110788345, + "step": 7820, + "valid_targets_mean": 1399.6, + "valid_targets_min": 822 + }, + { + "epoch": 4.31367144432194, + "grad_norm": 0.673369674275143, + "learning_rate": 1.5422563140322208e-05, + "loss": 0.1272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0629093199968338, + "step": 7825, + "valid_targets_mean": 1464.6, + "valid_targets_min": 670 + }, + { + "epoch": 4.316427783902977, + "grad_norm": 0.7635635522549837, + "learning_rate": 1.5395806849898713e-05, + "loss": 0.1409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0685824602842331, + "step": 7830, + "valid_targets_mean": 1504.8, + "valid_targets_min": 647 + }, + { + "epoch": 4.319184123484013, + "grad_norm": 0.7410832191020527, + "learning_rate": 1.536905925814324e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05398734658956528, + "step": 7835, + "valid_targets_mean": 1276.9, + "valid_targets_min": 791 + }, + { + "epoch": 4.32194046306505, + "grad_norm": 0.8167589271569274, + "learning_rate": 1.5342320415589816e-05, + "loss": 0.1363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06219908222556114, + "step": 7840, + "valid_targets_mean": 1291.4, + "valid_targets_min": 746 + }, + { + "epoch": 4.324696802646086, + "grad_norm": 0.7713857057219649, + "learning_rate": 1.531559037275594e-05, + "loss": 0.1389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05645304545760155, + "step": 7845, + "valid_targets_mean": 1409.1, + "valid_targets_min": 892 + }, + { + "epoch": 4.327453142227123, + "grad_norm": 0.775758746633171, + "learning_rate": 1.5288869180142482e-05, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0788961797952652, + "step": 7850, + "valid_targets_mean": 1581.9, + "valid_targets_min": 914 + }, + { + "epoch": 4.330209481808159, + "grad_norm": 0.7795825150988547, + "learning_rate": 1.5262156888233596e-05, + "loss": 0.1403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0639851838350296, + "step": 7855, + "valid_targets_mean": 1647.6, + "valid_targets_min": 636 + }, + { + "epoch": 4.3329658213891955, + "grad_norm": 0.7662229688435073, + "learning_rate": 1.523545354749661e-05, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0698130875825882, + "step": 7860, + "valid_targets_mean": 1523.5, + "valid_targets_min": 641 + }, + { + "epoch": 4.335722160970231, + "grad_norm": 0.8045013586486651, + "learning_rate": 1.5208759208381952e-05, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07863293588161469, + "step": 7865, + "valid_targets_mean": 1629.4, + "valid_targets_min": 1020 + }, + { + "epoch": 4.338478500551268, + "grad_norm": 0.8575189065781754, + "learning_rate": 1.5182073921323044e-05, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06408463418483734, + "step": 7870, + "valid_targets_mean": 1163.4, + "valid_targets_min": 729 + }, + { + "epoch": 4.341234840132304, + "grad_norm": 0.8611042786467799, + "learning_rate": 1.5155397736736197e-05, + "loss": 0.1467, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09173031151294708, + "step": 7875, + "valid_targets_mean": 1613.2, + "valid_targets_min": 1001 + }, + { + "epoch": 4.343991179713341, + "grad_norm": 0.747893583474277, + "learning_rate": 1.5128730705020528e-05, + "loss": 0.146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07269272208213806, + "step": 7880, + "valid_targets_mean": 1567.0, + "valid_targets_min": 1175 + }, + { + "epoch": 4.346747519294377, + "grad_norm": 0.831330564944061, + "learning_rate": 1.5102072876557867e-05, + "loss": 0.141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08520862460136414, + "step": 7885, + "valid_targets_mean": 1609.0, + "valid_targets_min": 892 + }, + { + "epoch": 4.349503858875414, + "grad_norm": 0.8012109367793205, + "learning_rate": 1.5075424301712645e-05, + "loss": 0.1434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06527456641197205, + "step": 7890, + "valid_targets_mean": 1425.2, + "valid_targets_min": 801 + }, + { + "epoch": 4.35226019845645, + "grad_norm": 0.744440500847799, + "learning_rate": 1.5048785030831821e-05, + "loss": 0.1301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06488320231437683, + "step": 7895, + "valid_targets_mean": 1462.4, + "valid_targets_min": 564 + }, + { + "epoch": 4.355016538037487, + "grad_norm": 0.7556394306605797, + "learning_rate": 1.5022155114244773e-05, + "loss": 0.1325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06646358966827393, + "step": 7900, + "valid_targets_mean": 1749.1, + "valid_targets_min": 651 + }, + { + "epoch": 4.3577728776185225, + "grad_norm": 0.7992767284291226, + "learning_rate": 1.499553460226321e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05667203292250633, + "step": 7905, + "valid_targets_mean": 1161.0, + "valid_targets_min": 716 + }, + { + "epoch": 4.360529217199559, + "grad_norm": 0.7679838837418798, + "learning_rate": 1.4968923545181055e-05, + "loss": 0.1363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06611639261245728, + "step": 7910, + "valid_targets_mean": 1425.6, + "valid_targets_min": 744 + }, + { + "epoch": 4.363285556780595, + "grad_norm": 0.7137153777651363, + "learning_rate": 1.4942321993274387e-05, + "loss": 0.1297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06086380407214165, + "step": 7915, + "valid_targets_mean": 1391.1, + "valid_targets_min": 1218 + }, + { + "epoch": 4.366041896361632, + "grad_norm": 0.8555803588101032, + "learning_rate": 1.4915729996801326e-05, + "loss": 0.137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0789387971162796, + "step": 7920, + "valid_targets_mean": 1724.5, + "valid_targets_min": 793 + }, + { + "epoch": 4.368798235942668, + "grad_norm": 0.7553838511315474, + "learning_rate": 1.4889147606001932e-05, + "loss": 0.1476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06265434622764587, + "step": 7925, + "valid_targets_mean": 1366.8, + "valid_targets_min": 846 + }, + { + "epoch": 4.371554575523705, + "grad_norm": 0.7765298774038851, + "learning_rate": 1.4862574871098118e-05, + "loss": 0.1331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07874997705221176, + "step": 7930, + "valid_targets_mean": 1772.1, + "valid_targets_min": 979 + }, + { + "epoch": 4.374310915104741, + "grad_norm": 0.7190471976156436, + "learning_rate": 1.4836011842293554e-05, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07440187782049179, + "step": 7935, + "valid_targets_mean": 1778.9, + "valid_targets_min": 1093 + }, + { + "epoch": 4.377067254685778, + "grad_norm": 0.70228274545568, + "learning_rate": 1.4809458569773569e-05, + "loss": 0.1198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06422741711139679, + "step": 7940, + "valid_targets_mean": 1510.8, + "valid_targets_min": 595 + }, + { + "epoch": 4.379823594266814, + "grad_norm": 0.7957599423275012, + "learning_rate": 1.4782915103705076e-05, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06474331021308899, + "step": 7945, + "valid_targets_mean": 1330.4, + "valid_targets_min": 868 + }, + { + "epoch": 4.38257993384785, + "grad_norm": 0.788320648270686, + "learning_rate": 1.4756381494236443e-05, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07642141729593277, + "step": 7950, + "valid_targets_mean": 1504.2, + "valid_targets_min": 914 + }, + { + "epoch": 4.385336273428886, + "grad_norm": 0.7150705510679891, + "learning_rate": 1.4729857791497417e-05, + "loss": 0.1352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07674597203731537, + "step": 7955, + "valid_targets_mean": 1972.9, + "valid_targets_min": 1080 + }, + { + "epoch": 4.388092613009923, + "grad_norm": 0.7407042074999919, + "learning_rate": 1.4703344045599037e-05, + "loss": 0.1354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07219728827476501, + "step": 7960, + "valid_targets_mean": 1519.6, + "valid_targets_min": 658 + }, + { + "epoch": 4.390848952590959, + "grad_norm": 0.7689301207598833, + "learning_rate": 1.467684030663353e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06139424443244934, + "step": 7965, + "valid_targets_mean": 1362.0, + "valid_targets_min": 1018 + }, + { + "epoch": 4.393605292171996, + "grad_norm": 0.7612109330028191, + "learning_rate": 1.4650346624674212e-05, + "loss": 0.1416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07849573343992233, + "step": 7970, + "valid_targets_mean": 1623.0, + "valid_targets_min": 1041 + }, + { + "epoch": 4.396361631753032, + "grad_norm": 1.1049541715892977, + "learning_rate": 1.46238630497754e-05, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17924700677394867, + "step": 7975, + "valid_targets_mean": 5392.9, + "valid_targets_min": 2506 + }, + { + "epoch": 4.399117971334069, + "grad_norm": 0.6916999688183976, + "learning_rate": 1.4597389631972314e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12939326465129852, + "step": 7980, + "valid_targets_mean": 5558.8, + "valid_targets_min": 533 + }, + { + "epoch": 4.401874310915105, + "grad_norm": 0.5628851376652243, + "learning_rate": 1.4570926421280982e-05, + "loss": 0.2316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11944962292909622, + "step": 7985, + "valid_targets_mean": 4660.4, + "valid_targets_min": 1752 + }, + { + "epoch": 4.404630650496141, + "grad_norm": 0.6023019446157618, + "learning_rate": 1.4544473467698165e-05, + "loss": 0.2281, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12325648963451385, + "step": 7990, + "valid_targets_mean": 4029.9, + "valid_targets_min": 1201 + }, + { + "epoch": 4.407386990077177, + "grad_norm": 0.5134212943688999, + "learning_rate": 1.4518030821201223e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10634707659482956, + "step": 7995, + "valid_targets_mean": 4551.0, + "valid_targets_min": 2259 + }, + { + "epoch": 4.410143329658214, + "grad_norm": 0.5680461579956786, + "learning_rate": 1.4491598531748053e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1437857747077942, + "step": 8000, + "valid_targets_mean": 5568.8, + "valid_targets_min": 2171 + }, + { + "epoch": 4.41289966923925, + "grad_norm": 0.5114863709172676, + "learning_rate": 1.4465176649276984e-05, + "loss": 0.2213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12304528057575226, + "step": 8005, + "valid_targets_mean": 7026.1, + "valid_targets_min": 4500 + }, + { + "epoch": 4.415656008820287, + "grad_norm": 0.5175236624901945, + "learning_rate": 1.443876522370668e-05, + "loss": 0.228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12371765822172165, + "step": 8010, + "valid_targets_mean": 5688.4, + "valid_targets_min": 1855 + }, + { + "epoch": 4.418412348401323, + "grad_norm": 0.595110581139693, + "learning_rate": 1.441236430493606e-05, + "loss": 0.2235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12023981660604477, + "step": 8015, + "valid_targets_mean": 4019.0, + "valid_targets_min": 1024 + }, + { + "epoch": 4.42116868798236, + "grad_norm": 0.48272194396140256, + "learning_rate": 1.4385973942844178e-05, + "loss": 0.2354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10892777144908905, + "step": 8020, + "valid_targets_mean": 4988.4, + "valid_targets_min": 2569 + }, + { + "epoch": 4.423925027563396, + "grad_norm": 0.5866044738354743, + "learning_rate": 1.4359594187290151e-05, + "loss": 0.2305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12635213136672974, + "step": 8025, + "valid_targets_mean": 3892.2, + "valid_targets_min": 802 + }, + { + "epoch": 4.4266813671444325, + "grad_norm": 0.5080733385988445, + "learning_rate": 1.4333225088113058e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11392886191606522, + "step": 8030, + "valid_targets_mean": 5692.4, + "valid_targets_min": 1597 + }, + { + "epoch": 4.429437706725468, + "grad_norm": 0.5638797047566162, + "learning_rate": 1.4306866695131837e-05, + "loss": 0.2318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11581313610076904, + "step": 8035, + "valid_targets_mean": 3506.8, + "valid_targets_min": 1650 + }, + { + "epoch": 4.432194046306505, + "grad_norm": 0.5257032305505889, + "learning_rate": 1.4280519058145212e-05, + "loss": 0.2247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11211000382900238, + "step": 8040, + "valid_targets_mean": 4337.4, + "valid_targets_min": 1385 + }, + { + "epoch": 4.434950385887541, + "grad_norm": 0.5917967382910359, + "learning_rate": 1.4254182226931574e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1194198727607727, + "step": 8045, + "valid_targets_mean": 3932.2, + "valid_targets_min": 2549 + }, + { + "epoch": 4.437706725468578, + "grad_norm": 0.5506758239826691, + "learning_rate": 1.4227856251248904e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11180636286735535, + "step": 8050, + "valid_targets_mean": 4973.0, + "valid_targets_min": 2052 + }, + { + "epoch": 4.440463065049614, + "grad_norm": 0.5978816946123382, + "learning_rate": 1.4201541180834673e-05, + "loss": 0.2185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1058175265789032, + "step": 8055, + "valid_targets_mean": 4029.6, + "valid_targets_min": 1210 + }, + { + "epoch": 4.443219404630651, + "grad_norm": 0.5847247985018135, + "learning_rate": 1.4175237065405745e-05, + "loss": 0.223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12738429009914398, + "step": 8060, + "valid_targets_mean": 4094.8, + "valid_targets_min": 1064 + }, + { + "epoch": 4.445975744211687, + "grad_norm": 0.6363327709736324, + "learning_rate": 1.4148943954658299e-05, + "loss": 0.2341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13722604513168335, + "step": 8065, + "valid_targets_mean": 3760.1, + "valid_targets_min": 1678 + }, + { + "epoch": 4.448732083792724, + "grad_norm": 0.5975335919614989, + "learning_rate": 1.4122661898267706e-05, + "loss": 0.2255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11916285008192062, + "step": 8070, + "valid_targets_mean": 3665.2, + "valid_targets_min": 671 + }, + { + "epoch": 4.4514884233737595, + "grad_norm": 0.6588762396754851, + "learning_rate": 1.4096390945888467e-05, + "loss": 0.2226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10171541571617126, + "step": 8075, + "valid_targets_mean": 3325.0, + "valid_targets_min": 1864 + }, + { + "epoch": 4.454244762954796, + "grad_norm": 0.5882888268363246, + "learning_rate": 1.407013114715409e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11612389981746674, + "step": 8080, + "valid_targets_mean": 3751.9, + "valid_targets_min": 1532 + }, + { + "epoch": 4.457001102535832, + "grad_norm": 0.592960720901953, + "learning_rate": 1.4043882551677028e-05, + "loss": 0.2141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09083391726016998, + "step": 8085, + "valid_targets_mean": 3773.9, + "valid_targets_min": 1051 + }, + { + "epoch": 4.459757442116869, + "grad_norm": 0.6120321382231287, + "learning_rate": 1.4017645209048554e-05, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11319591104984283, + "step": 8090, + "valid_targets_mean": 3679.9, + "valid_targets_min": 1345 + }, + { + "epoch": 4.462513781697905, + "grad_norm": 0.6011645556656778, + "learning_rate": 1.3991419168838683e-05, + "loss": 0.2228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10772179812192917, + "step": 8095, + "valid_targets_mean": 4461.5, + "valid_targets_min": 2231 + }, + { + "epoch": 4.465270121278942, + "grad_norm": 0.6241083165495994, + "learning_rate": 1.396520448059608e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10733223706483841, + "step": 8100, + "valid_targets_mean": 3884.6, + "valid_targets_min": 2127 + }, + { + "epoch": 4.468026460859978, + "grad_norm": 0.6597069890947413, + "learning_rate": 1.3939001193847955e-05, + "loss": 0.2085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10251784324645996, + "step": 8105, + "valid_targets_mean": 3261.0, + "valid_targets_min": 1599 + }, + { + "epoch": 4.470782800441015, + "grad_norm": 0.6029877045943878, + "learning_rate": 1.3912809358099995e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08413989841938019, + "step": 8110, + "valid_targets_mean": 2629.1, + "valid_targets_min": 908 + }, + { + "epoch": 4.4735391400220506, + "grad_norm": 0.6458529558889584, + "learning_rate": 1.3886629022836238e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13481256365776062, + "step": 8115, + "valid_targets_mean": 3459.9, + "valid_targets_min": 1264 + }, + { + "epoch": 4.476295479603087, + "grad_norm": 0.5612816186459422, + "learning_rate": 1.3860460237518993e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09364003688097, + "step": 8120, + "valid_targets_mean": 2603.2, + "valid_targets_min": 926 + }, + { + "epoch": 4.479051819184123, + "grad_norm": 0.6457152297072929, + "learning_rate": 1.3834303051588757e-05, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10920501500368118, + "step": 8125, + "valid_targets_mean": 3409.6, + "valid_targets_min": 1200 + }, + { + "epoch": 4.48180815876516, + "grad_norm": 0.6319247596238564, + "learning_rate": 1.3808157514464102e-05, + "loss": 0.2083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11417315900325775, + "step": 8130, + "valid_targets_mean": 3607.5, + "valid_targets_min": 2019 + }, + { + "epoch": 4.484564498346196, + "grad_norm": 0.6157957550323329, + "learning_rate": 1.3782023675541606e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1320439726114273, + "step": 8135, + "valid_targets_mean": 4661.0, + "valid_targets_min": 2908 + }, + { + "epoch": 4.487320837927233, + "grad_norm": 0.6057250661596274, + "learning_rate": 1.3755901584195728e-05, + "loss": 0.213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08959028124809265, + "step": 8140, + "valid_targets_mean": 3215.2, + "valid_targets_min": 1110 + }, + { + "epoch": 4.490077177508269, + "grad_norm": 0.6172221291863343, + "learning_rate": 1.372979128977875e-05, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10946552455425262, + "step": 8145, + "valid_targets_mean": 3720.9, + "valid_targets_min": 513 + }, + { + "epoch": 4.492833517089306, + "grad_norm": 0.6027850590344314, + "learning_rate": 1.3703692841620653e-05, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1165655106306076, + "step": 8150, + "valid_targets_mean": 4162.0, + "valid_targets_min": 2558 + }, + { + "epoch": 4.495589856670342, + "grad_norm": 0.6343158264117856, + "learning_rate": 1.367760628902904e-05, + "loss": 0.224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10658375918865204, + "step": 8155, + "valid_targets_mean": 3291.6, + "valid_targets_min": 1673 + }, + { + "epoch": 4.498346196251378, + "grad_norm": 0.5948958402146832, + "learning_rate": 1.3651531681289048e-05, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09390570223331451, + "step": 8160, + "valid_targets_mean": 3464.6, + "valid_targets_min": 1860 + }, + { + "epoch": 4.501102535832414, + "grad_norm": 0.5392245147527028, + "learning_rate": 1.3625469067663237e-05, + "loss": 0.2107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10819543153047562, + "step": 8165, + "valid_targets_mean": 5273.0, + "valid_targets_min": 2495 + }, + { + "epoch": 4.503858875413451, + "grad_norm": 0.6153351670855557, + "learning_rate": 1.3599418497391512e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12458068877458572, + "step": 8170, + "valid_targets_mean": 3913.6, + "valid_targets_min": 1945 + }, + { + "epoch": 4.506615214994487, + "grad_norm": 0.5966301638879071, + "learning_rate": 1.3573380019691019e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10793530941009521, + "step": 8175, + "valid_targets_mean": 4705.5, + "valid_targets_min": 1908 + }, + { + "epoch": 4.509371554575524, + "grad_norm": 0.6032336124756742, + "learning_rate": 1.3547353683756056e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11116038262844086, + "step": 8180, + "valid_targets_mean": 3445.9, + "valid_targets_min": 1757 + }, + { + "epoch": 4.51212789415656, + "grad_norm": 0.5884018667624914, + "learning_rate": 1.3521339538758e-05, + "loss": 0.2104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09658658504486084, + "step": 8185, + "valid_targets_mean": 4265.4, + "valid_targets_min": 2439 + }, + { + "epoch": 4.514884233737597, + "grad_norm": 0.5936648135341931, + "learning_rate": 1.3495337633845175e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10096180438995361, + "step": 8190, + "valid_targets_mean": 3829.6, + "valid_targets_min": 2482 + }, + { + "epoch": 4.517640573318633, + "grad_norm": 0.6114466208689194, + "learning_rate": 1.3469348018142786e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10825655609369278, + "step": 8195, + "valid_targets_mean": 3815.4, + "valid_targets_min": 1833 + }, + { + "epoch": 4.5203969128996695, + "grad_norm": 0.6625003583468229, + "learning_rate": 1.3443370740752823e-05, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09280657768249512, + "step": 8200, + "valid_targets_mean": 2736.6, + "valid_targets_min": 1182 + }, + { + "epoch": 4.523153252480705, + "grad_norm": 0.5491792011609379, + "learning_rate": 1.3417405850753968e-05, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10392202436923981, + "step": 8205, + "valid_targets_mean": 4636.2, + "valid_targets_min": 1869 + }, + { + "epoch": 4.525909592061742, + "grad_norm": 0.5899303503674906, + "learning_rate": 1.3391453397201493e-05, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1181587353348732, + "step": 8210, + "valid_targets_mean": 4108.8, + "valid_targets_min": 2498 + }, + { + "epoch": 4.528665931642778, + "grad_norm": 0.6371395811427715, + "learning_rate": 1.3365513429127172e-05, + "loss": 0.2235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09593787044286728, + "step": 8215, + "valid_targets_mean": 3190.2, + "valid_targets_min": 988 + }, + { + "epoch": 4.531422271223815, + "grad_norm": 0.6417258081191761, + "learning_rate": 1.33395859955392e-05, + "loss": 0.2129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10807552188634872, + "step": 8220, + "valid_targets_mean": 3140.5, + "valid_targets_min": 1361 + }, + { + "epoch": 4.534178610804851, + "grad_norm": 0.5382294712727117, + "learning_rate": 1.3313671145422077e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09786634147167206, + "step": 8225, + "valid_targets_mean": 4282.2, + "valid_targets_min": 2252 + }, + { + "epoch": 4.536934950385888, + "grad_norm": 2.389591222792242, + "learning_rate": 1.328776892773655e-05, + "loss": 0.2152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10175527632236481, + "step": 8230, + "valid_targets_mean": 3130.8, + "valid_targets_min": 1436 + }, + { + "epoch": 4.539691289966924, + "grad_norm": 0.5834715160334898, + "learning_rate": 1.3261879391419478e-05, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09242591261863708, + "step": 8235, + "valid_targets_mean": 3714.9, + "valid_targets_min": 2200 + }, + { + "epoch": 4.5424476295479606, + "grad_norm": 0.5989519208322802, + "learning_rate": 1.3236002585383774e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10470782965421677, + "step": 8240, + "valid_targets_mean": 4340.8, + "valid_targets_min": 1272 + }, + { + "epoch": 4.5452039691289965, + "grad_norm": 0.5832260062955171, + "learning_rate": 1.3210138558518295e-05, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10514132678508759, + "step": 8245, + "valid_targets_mean": 3423.1, + "valid_targets_min": 1319 + }, + { + "epoch": 4.547960308710033, + "grad_norm": 0.5648716780991663, + "learning_rate": 1.3184287359687751e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09550131112337112, + "step": 8250, + "valid_targets_mean": 3725.6, + "valid_targets_min": 1194 + }, + { + "epoch": 4.550716648291069, + "grad_norm": 0.6205504666754108, + "learning_rate": 1.3158449037732628e-05, + "loss": 0.2263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09223237633705139, + "step": 8255, + "valid_targets_mean": 2583.6, + "valid_targets_min": 1453 + }, + { + "epoch": 4.553472987872106, + "grad_norm": 0.5326882446724851, + "learning_rate": 1.3132623641469074e-05, + "loss": 0.2019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08561265468597412, + "step": 8260, + "valid_targets_mean": 5286.5, + "valid_targets_min": 3139 + }, + { + "epoch": 4.556229327453142, + "grad_norm": 0.5943428319443794, + "learning_rate": 1.3106811219688815e-05, + "loss": 0.2178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09145857393741608, + "step": 8265, + "valid_targets_mean": 3116.1, + "valid_targets_min": 1256 + }, + { + "epoch": 4.558985667034179, + "grad_norm": 0.6157424129193856, + "learning_rate": 1.3081011821159075e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1326896846294403, + "step": 8270, + "valid_targets_mean": 4477.9, + "valid_targets_min": 1608 + }, + { + "epoch": 4.561742006615215, + "grad_norm": 0.6186926821889323, + "learning_rate": 1.3055225494622459e-05, + "loss": 0.2204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10684438049793243, + "step": 8275, + "valid_targets_mean": 3772.4, + "valid_targets_min": 1611 + }, + { + "epoch": 4.564498346196252, + "grad_norm": 0.5974510558649161, + "learning_rate": 1.3029452288796886e-05, + "loss": 0.2112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11817403882741928, + "step": 8280, + "valid_targets_mean": 4032.5, + "valid_targets_min": 2202 + }, + { + "epoch": 4.5672546857772875, + "grad_norm": 0.6013613169937765, + "learning_rate": 1.3003692252375486e-05, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1226746216416359, + "step": 8285, + "valid_targets_mean": 4019.0, + "valid_targets_min": 2582 + }, + { + "epoch": 4.570011025358324, + "grad_norm": 0.6339462976006086, + "learning_rate": 1.2977945434026504e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11144387722015381, + "step": 8290, + "valid_targets_mean": 3423.1, + "valid_targets_min": 2314 + }, + { + "epoch": 4.57276736493936, + "grad_norm": 0.6034908424438389, + "learning_rate": 1.2952211882393211e-05, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10312724858522415, + "step": 8295, + "valid_targets_mean": 3991.5, + "valid_targets_min": 2343 + }, + { + "epoch": 4.575523704520397, + "grad_norm": 0.5913979136300587, + "learning_rate": 1.292649164609381e-05, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08746472001075745, + "step": 8300, + "valid_targets_mean": 2864.9, + "valid_targets_min": 1026 + }, + { + "epoch": 4.578280044101433, + "grad_norm": 0.6711444176081476, + "learning_rate": 1.2900784773721365e-05, + "loss": 0.1991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11053238809108734, + "step": 8305, + "valid_targets_mean": 4022.8, + "valid_targets_min": 1042 + }, + { + "epoch": 4.58103638368247, + "grad_norm": 0.6032206746494746, + "learning_rate": 1.2875091313843674e-05, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09786505997180939, + "step": 8310, + "valid_targets_mean": 3428.8, + "valid_targets_min": 1274 + }, + { + "epoch": 4.583792723263506, + "grad_norm": 0.6582304751458176, + "learning_rate": 1.2849411315003199e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11556897312402725, + "step": 8315, + "valid_targets_mean": 4167.0, + "valid_targets_min": 914 + }, + { + "epoch": 4.586549062844543, + "grad_norm": 0.6006504912083689, + "learning_rate": 1.2823744825716974e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09089648723602295, + "step": 8320, + "valid_targets_mean": 3927.2, + "valid_targets_min": 2093 + }, + { + "epoch": 4.589305402425579, + "grad_norm": 0.6140588885835073, + "learning_rate": 1.2798091894476503e-05, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12398666888475418, + "step": 8325, + "valid_targets_mean": 3882.8, + "valid_targets_min": 1683 + }, + { + "epoch": 4.592061742006615, + "grad_norm": 0.5730243007746728, + "learning_rate": 1.2772452569747685e-05, + "loss": 0.2088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1055247113108635, + "step": 8330, + "valid_targets_mean": 4466.9, + "valid_targets_min": 2253 + }, + { + "epoch": 4.594818081587651, + "grad_norm": 0.5894996166126563, + "learning_rate": 1.2746826899970706e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0980580598115921, + "step": 8335, + "valid_targets_mean": 3929.2, + "valid_targets_min": 1789 + }, + { + "epoch": 4.597574421168688, + "grad_norm": 0.6222354898980191, + "learning_rate": 1.2721214933559947e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10830824077129364, + "step": 8340, + "valid_targets_mean": 4174.0, + "valid_targets_min": 1586 + }, + { + "epoch": 4.600330760749724, + "grad_norm": 0.6903389709208698, + "learning_rate": 1.269561671890391e-05, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11992015689611435, + "step": 8345, + "valid_targets_mean": 3746.6, + "valid_targets_min": 2033 + }, + { + "epoch": 4.603087100330761, + "grad_norm": 0.5547779797641604, + "learning_rate": 1.2670032304365116e-05, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09813828766345978, + "step": 8350, + "valid_targets_mean": 3945.5, + "valid_targets_min": 1616 + }, + { + "epoch": 4.605843439911797, + "grad_norm": 0.591769383251808, + "learning_rate": 1.264446173828001e-05, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0865781381726265, + "step": 8355, + "valid_targets_mean": 3534.4, + "valid_targets_min": 2156 + }, + { + "epoch": 4.608599779492834, + "grad_norm": 0.5970706736431347, + "learning_rate": 1.2618905068958867e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10986503958702087, + "step": 8360, + "valid_targets_mean": 3635.5, + "valid_targets_min": 1750 + }, + { + "epoch": 4.61135611907387, + "grad_norm": 0.6819024126782567, + "learning_rate": 1.2593362344685717e-05, + "loss": 0.2044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12637746334075928, + "step": 8365, + "valid_targets_mean": 3780.1, + "valid_targets_min": 2126 + }, + { + "epoch": 4.6141124586549065, + "grad_norm": 0.6488901826687372, + "learning_rate": 1.2567833613718237e-05, + "loss": 0.2182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11596396565437317, + "step": 8370, + "valid_targets_mean": 3756.1, + "valid_targets_min": 1834 + }, + { + "epoch": 4.616868798235942, + "grad_norm": 1.5207110577744747, + "learning_rate": 1.2542318924287678e-05, + "loss": 0.2094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12189093232154846, + "step": 8375, + "valid_targets_mean": 4082.5, + "valid_targets_min": 1571 + }, + { + "epoch": 4.619625137816979, + "grad_norm": 0.6707620099448164, + "learning_rate": 1.2516818324598742e-05, + "loss": 0.2197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1132071241736412, + "step": 8380, + "valid_targets_mean": 3371.9, + "valid_targets_min": 1487 + }, + { + "epoch": 4.622381477398015, + "grad_norm": 0.6265803159006925, + "learning_rate": 1.2491331862829532e-05, + "loss": 0.2084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08453784883022308, + "step": 8385, + "valid_targets_mean": 3824.5, + "valid_targets_min": 1970 + }, + { + "epoch": 4.625137816979052, + "grad_norm": 0.5949086707334191, + "learning_rate": 1.2465859587131428e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10151436924934387, + "step": 8390, + "valid_targets_mean": 3956.8, + "valid_targets_min": 2492 + }, + { + "epoch": 4.627894156560088, + "grad_norm": 0.6286607759865586, + "learning_rate": 1.2440401545629009e-05, + "loss": 0.2137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0926315113902092, + "step": 8395, + "valid_targets_mean": 3557.9, + "valid_targets_min": 1675 + }, + { + "epoch": 4.630650496141125, + "grad_norm": 0.6173564990953846, + "learning_rate": 1.2414957786419967e-05, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09937307238578796, + "step": 8400, + "valid_targets_mean": 3468.0, + "valid_targets_min": 2437 + }, + { + "epoch": 4.633406835722161, + "grad_norm": 0.6474437467019091, + "learning_rate": 1.2389528357575013e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10454729199409485, + "step": 8405, + "valid_targets_mean": 3597.0, + "valid_targets_min": 1969 + }, + { + "epoch": 4.6361631753031975, + "grad_norm": 0.6080896499145835, + "learning_rate": 1.2364113307137773e-05, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09152059257030487, + "step": 8410, + "valid_targets_mean": 3417.6, + "valid_targets_min": 1005 + }, + { + "epoch": 4.6389195148842335, + "grad_norm": 0.637351409121502, + "learning_rate": 1.2338712683124717e-05, + "loss": 0.2095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08515492081642151, + "step": 8415, + "valid_targets_mean": 3225.5, + "valid_targets_min": 1207 + }, + { + "epoch": 4.64167585446527, + "grad_norm": 0.6334649300590226, + "learning_rate": 1.231332653352505e-05, + "loss": 0.2122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11460841447114944, + "step": 8420, + "valid_targets_mean": 4136.9, + "valid_targets_min": 1257 + }, + { + "epoch": 4.644432194046306, + "grad_norm": 0.6333541830023642, + "learning_rate": 1.2287954906300638e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10046584159135818, + "step": 8425, + "valid_targets_mean": 3068.1, + "valid_targets_min": 1357 + }, + { + "epoch": 4.647188533627343, + "grad_norm": 0.5593221660645111, + "learning_rate": 1.2262597849385908e-05, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0848647952079773, + "step": 8430, + "valid_targets_mean": 4579.2, + "valid_targets_min": 1410 + }, + { + "epoch": 4.649944873208379, + "grad_norm": 0.5788611963723003, + "learning_rate": 1.2237255410687766e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09438008069992065, + "step": 8435, + "valid_targets_mean": 3438.5, + "valid_targets_min": 1390 + }, + { + "epoch": 4.652701212789416, + "grad_norm": 0.6500151427677997, + "learning_rate": 1.2211927638085487e-05, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0868106335401535, + "step": 8440, + "valid_targets_mean": 3037.1, + "valid_targets_min": 790 + }, + { + "epoch": 4.655457552370452, + "grad_norm": 0.6268808763994755, + "learning_rate": 1.218661457943065e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10260500013828278, + "step": 8445, + "valid_targets_mean": 3695.4, + "valid_targets_min": 1901 + }, + { + "epoch": 4.658213891951489, + "grad_norm": 0.6501256553707975, + "learning_rate": 1.2161316282547024e-05, + "loss": 0.2046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09765063226222992, + "step": 8450, + "valid_targets_mean": 2810.9, + "valid_targets_min": 1460 + }, + { + "epoch": 4.6609702315325245, + "grad_norm": 0.5620920389781153, + "learning_rate": 1.2136032795230492e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11658498644828796, + "step": 8455, + "valid_targets_mean": 5414.1, + "valid_targets_min": 2900 + }, + { + "epoch": 4.663726571113561, + "grad_norm": 0.7107115576824473, + "learning_rate": 1.211076416524897e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11848993599414825, + "step": 8460, + "valid_targets_mean": 3211.0, + "valid_targets_min": 1768 + }, + { + "epoch": 4.666482910694597, + "grad_norm": 0.5768741729277652, + "learning_rate": 1.2085510440342282e-05, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09341521561145782, + "step": 8465, + "valid_targets_mean": 4090.2, + "valid_targets_min": 1931 + }, + { + "epoch": 4.669239250275634, + "grad_norm": 0.5350209358955261, + "learning_rate": 1.2060271668222115e-05, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09028655290603638, + "step": 8470, + "valid_targets_mean": 5752.6, + "valid_targets_min": 1233 + }, + { + "epoch": 4.67199558985667, + "grad_norm": 0.5435579642793948, + "learning_rate": 1.2035047896571884e-05, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09956096857786179, + "step": 8475, + "valid_targets_mean": 5342.2, + "valid_targets_min": 1741 + }, + { + "epoch": 4.674751929437707, + "grad_norm": 0.5782258720367679, + "learning_rate": 1.2009839173046673e-05, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10321082174777985, + "step": 8480, + "valid_targets_mean": 3760.9, + "valid_targets_min": 856 + }, + { + "epoch": 4.677508269018743, + "grad_norm": 0.5833422457244973, + "learning_rate": 1.1984645545273145e-05, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1067807748913765, + "step": 8485, + "valid_targets_mean": 3722.1, + "valid_targets_min": 1062 + }, + { + "epoch": 4.68026460859978, + "grad_norm": 0.5795746931526324, + "learning_rate": 1.1959467060849432e-05, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09334102272987366, + "step": 8490, + "valid_targets_mean": 3910.2, + "valid_targets_min": 974 + }, + { + "epoch": 4.683020948180816, + "grad_norm": 0.6393433568486007, + "learning_rate": 1.1934303767345057e-05, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0970325618982315, + "step": 8495, + "valid_targets_mean": 3008.9, + "valid_targets_min": 1447 + }, + { + "epoch": 4.685777287761852, + "grad_norm": 0.6122726895219656, + "learning_rate": 1.1909155712300844e-05, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10893280804157257, + "step": 8500, + "valid_targets_mean": 4858.2, + "valid_targets_min": 3617 + }, + { + "epoch": 4.688533627342888, + "grad_norm": 0.6715385574528783, + "learning_rate": 1.1884022943228824e-05, + "loss": 0.1934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10608501732349396, + "step": 8505, + "valid_targets_mean": 4037.0, + "valid_targets_min": 1515 + }, + { + "epoch": 4.691289966923925, + "grad_norm": 0.7060001056259848, + "learning_rate": 1.1858905507612156e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1053161695599556, + "step": 8510, + "valid_targets_mean": 2778.0, + "valid_targets_min": 1349 + }, + { + "epoch": 4.694046306504961, + "grad_norm": 0.653020597518514, + "learning_rate": 1.1833803452905025e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10597722232341766, + "step": 8515, + "valid_targets_mean": 3199.2, + "valid_targets_min": 1906 + }, + { + "epoch": 4.696802646085998, + "grad_norm": 0.6596756094926586, + "learning_rate": 1.1808716826532556e-05, + "loss": 0.195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10283436626195908, + "step": 8520, + "valid_targets_mean": 3461.5, + "valid_targets_min": 2076 + }, + { + "epoch": 4.699558985667034, + "grad_norm": 0.6570669536065473, + "learning_rate": 1.1783645675890725e-05, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0806070938706398, + "step": 8525, + "valid_targets_mean": 3071.8, + "valid_targets_min": 1021 + }, + { + "epoch": 4.702315325248071, + "grad_norm": 0.6685112334989192, + "learning_rate": 1.175859004834627e-05, + "loss": 0.2028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09340055286884308, + "step": 8530, + "valid_targets_mean": 3088.9, + "valid_targets_min": 654 + }, + { + "epoch": 4.705071664829107, + "grad_norm": 0.636133079794076, + "learning_rate": 1.1733549991236608e-05, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11072291433811188, + "step": 8535, + "valid_targets_mean": 4179.1, + "valid_targets_min": 1605 + }, + { + "epoch": 4.7078280044101435, + "grad_norm": 0.7591576775878808, + "learning_rate": 1.1708525551869729e-05, + "loss": 0.2075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10244138538837433, + "step": 8540, + "valid_targets_mean": 2737.0, + "valid_targets_min": 815 + }, + { + "epoch": 4.710584343991179, + "grad_norm": 0.6340986023202737, + "learning_rate": 1.1683516777524127e-05, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09207886457443237, + "step": 8545, + "valid_targets_mean": 3384.8, + "valid_targets_min": 1878 + }, + { + "epoch": 4.713340683572216, + "grad_norm": 0.6042345910383736, + "learning_rate": 1.1658523715448683e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10194215178489685, + "step": 8550, + "valid_targets_mean": 3888.4, + "valid_targets_min": 1793 + }, + { + "epoch": 4.716097023153252, + "grad_norm": 0.6086468425454878, + "learning_rate": 1.1633546412862612e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08884093165397644, + "step": 8555, + "valid_targets_mean": 3238.4, + "valid_targets_min": 1217 + }, + { + "epoch": 4.718853362734289, + "grad_norm": 0.6305657267699502, + "learning_rate": 1.1608584916955348e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07489611953496933, + "step": 8560, + "valid_targets_mean": 1873.8, + "valid_targets_min": 634 + }, + { + "epoch": 4.721609702315325, + "grad_norm": 0.3960409486902877, + "learning_rate": 1.1583639274886461e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032717444002628326, + "step": 8565, + "valid_targets_mean": 2854.6, + "valid_targets_min": 834 + }, + { + "epoch": 4.724366041896362, + "grad_norm": 0.4229676264607288, + "learning_rate": 1.155870953378556e-05, + "loss": 0.0812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03650623559951782, + "step": 8570, + "valid_targets_mean": 3432.9, + "valid_targets_min": 2790 + }, + { + "epoch": 4.727122381477398, + "grad_norm": 0.4014140109882132, + "learning_rate": 1.1533795740752228e-05, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04279500991106033, + "step": 8575, + "valid_targets_mean": 3537.2, + "valid_targets_min": 2133 + }, + { + "epoch": 4.7298787210584345, + "grad_norm": 0.38927267540992194, + "learning_rate": 1.1508897942855902e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03486490994691849, + "step": 8580, + "valid_targets_mean": 2455.5, + "valid_targets_min": 526 + }, + { + "epoch": 4.7326350606394705, + "grad_norm": 0.40346362143159514, + "learning_rate": 1.1484016187135815e-05, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0423397459089756, + "step": 8585, + "valid_targets_mean": 2771.6, + "valid_targets_min": 890 + }, + { + "epoch": 4.735391400220507, + "grad_norm": 1.0845228176695407, + "learning_rate": 1.1459150520600873e-05, + "loss": 0.1031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07804582267999649, + "step": 8590, + "valid_targets_mean": 1142.2, + "valid_targets_min": 559 + }, + { + "epoch": 4.738147739801543, + "grad_norm": 0.4414562542099629, + "learning_rate": 1.143430099022961e-05, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031130073592066765, + "step": 8595, + "valid_targets_mean": 2681.6, + "valid_targets_min": 647 + }, + { + "epoch": 4.74090407938258, + "grad_norm": 0.45355340561649854, + "learning_rate": 1.1409467642970053e-05, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04014985263347626, + "step": 8600, + "valid_targets_mean": 3005.8, + "valid_targets_min": 947 + }, + { + "epoch": 4.743660418963616, + "grad_norm": 0.5501128805203366, + "learning_rate": 1.1384650525739662e-05, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09858731925487518, + "step": 8605, + "valid_targets_mean": 1957.4, + "valid_targets_min": 859 + }, + { + "epoch": 4.746416758544653, + "grad_norm": 0.3101496803474559, + "learning_rate": 1.1359849685425232e-05, + "loss": 0.082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04386935383081436, + "step": 8610, + "valid_targets_mean": 5439.9, + "valid_targets_min": 636 + }, + { + "epoch": 4.749173098125689, + "grad_norm": 0.38916172557805273, + "learning_rate": 1.1335065168882807e-05, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04180110991001129, + "step": 8615, + "valid_targets_mean": 2588.5, + "valid_targets_min": 826 + }, + { + "epoch": 4.751929437706726, + "grad_norm": 0.31645373016216244, + "learning_rate": 1.1310297022937589e-05, + "loss": 0.0797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036886654794216156, + "step": 8620, + "valid_targets_mean": 3753.1, + "valid_targets_min": 1096 + }, + { + "epoch": 4.7546857772877615, + "grad_norm": 0.516088174549427, + "learning_rate": 1.1285545294383855e-05, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047880880534648895, + "step": 8625, + "valid_targets_mean": 2879.6, + "valid_targets_min": 688 + }, + { + "epoch": 4.757442116868798, + "grad_norm": 0.363385659685484, + "learning_rate": 1.1260810029984861e-05, + "loss": 0.102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03266862779855728, + "step": 8630, + "valid_targets_mean": 3137.2, + "valid_targets_min": 553 + }, + { + "epoch": 4.760198456449834, + "grad_norm": 0.4454930414928354, + "learning_rate": 1.1236091276472751e-05, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042061131447553635, + "step": 8635, + "valid_targets_mean": 2416.8, + "valid_targets_min": 566 + }, + { + "epoch": 4.762954796030871, + "grad_norm": 0.3092593547844018, + "learning_rate": 1.1211389080548499e-05, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03263853117823601, + "step": 8640, + "valid_targets_mean": 4076.0, + "valid_targets_min": 3337 + }, + { + "epoch": 4.765711135611907, + "grad_norm": 0.4461504512476077, + "learning_rate": 1.1186703488881773e-05, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05571845918893814, + "step": 8645, + "valid_targets_mean": 1685.1, + "valid_targets_min": 622 + }, + { + "epoch": 4.768467475192944, + "grad_norm": 0.5764761034253417, + "learning_rate": 1.116203454811088e-05, + "loss": 0.0913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053428564220666885, + "step": 8650, + "valid_targets_mean": 2427.1, + "valid_targets_min": 696 + }, + { + "epoch": 4.77122381477398, + "grad_norm": 0.5775920685367084, + "learning_rate": 1.1137382304842665e-05, + "loss": 0.0881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055976223200559616, + "step": 8655, + "valid_targets_mean": 3013.5, + "valid_targets_min": 1922 + }, + { + "epoch": 4.773980154355017, + "grad_norm": 0.4302314668779673, + "learning_rate": 1.1112746805652432e-05, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028733866289258003, + "step": 8660, + "valid_targets_mean": 2437.6, + "valid_targets_min": 802 + }, + { + "epoch": 4.776736493936053, + "grad_norm": 0.48206612242702457, + "learning_rate": 1.1088128097083841e-05, + "loss": 0.0951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056801795959472656, + "step": 8665, + "valid_targets_mean": 2044.0, + "valid_targets_min": 957 + }, + { + "epoch": 4.779492833517089, + "grad_norm": 0.3773758156088737, + "learning_rate": 1.106352622564884e-05, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04158919304609299, + "step": 8670, + "valid_targets_mean": 2513.2, + "valid_targets_min": 641 + }, + { + "epoch": 4.782249173098125, + "grad_norm": 0.38156482186795, + "learning_rate": 1.1038941237827565e-05, + "loss": 0.0767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04479191452264786, + "step": 8675, + "valid_targets_mean": 2685.9, + "valid_targets_min": 1257 + }, + { + "epoch": 4.785005512679162, + "grad_norm": 0.5906076703730488, + "learning_rate": 1.1014373180068246e-05, + "loss": 0.1172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08775226771831512, + "step": 8680, + "valid_targets_mean": 2143.6, + "valid_targets_min": 700 + }, + { + "epoch": 4.787761852260198, + "grad_norm": 0.4687614137407911, + "learning_rate": 1.0989822098787126e-05, + "loss": 0.0986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05650041997432709, + "step": 8685, + "valid_targets_mean": 2739.1, + "valid_targets_min": 610 + }, + { + "epoch": 4.790518191841235, + "grad_norm": 0.4898591682748263, + "learning_rate": 1.09652880403684e-05, + "loss": 0.0826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05731917917728424, + "step": 8690, + "valid_targets_mean": 2732.4, + "valid_targets_min": 893 + }, + { + "epoch": 4.793274531422271, + "grad_norm": 0.6644616610205819, + "learning_rate": 1.0940771051164072e-05, + "loss": 0.2482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2619195282459259, + "step": 8695, + "valid_targets_mean": 2854.8, + "valid_targets_min": 1171 + }, + { + "epoch": 4.796030871003308, + "grad_norm": 0.48615249497429847, + "learning_rate": 1.0916271177493907e-05, + "loss": 0.1208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039592280983924866, + "step": 8700, + "valid_targets_mean": 2121.0, + "valid_targets_min": 720 + }, + { + "epoch": 4.798787210584344, + "grad_norm": 0.39807034314583306, + "learning_rate": 1.0891788465645338e-05, + "loss": 0.0971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051629189401865005, + "step": 8705, + "valid_targets_mean": 3264.9, + "valid_targets_min": 615 + }, + { + "epoch": 4.8015435501653805, + "grad_norm": 0.5142471667528368, + "learning_rate": 1.0867322961873373e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09411557763814926, + "step": 8710, + "valid_targets_mean": 3200.5, + "valid_targets_min": 1368 + }, + { + "epoch": 4.804299889746416, + "grad_norm": 0.40304419181411166, + "learning_rate": 1.0842874712400499e-05, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031656742095947266, + "step": 8715, + "valid_targets_mean": 2314.5, + "valid_targets_min": 830 + }, + { + "epoch": 4.807056229327453, + "grad_norm": 0.39074545393919596, + "learning_rate": 1.0818443763416622e-05, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0460924431681633, + "step": 8720, + "valid_targets_mean": 5013.2, + "valid_targets_min": 2671 + }, + { + "epoch": 4.809812568908489, + "grad_norm": 0.3414185194877373, + "learning_rate": 1.0794030161078948e-05, + "loss": 0.0877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031554799526929855, + "step": 8725, + "valid_targets_mean": 3405.1, + "valid_targets_min": 997 + }, + { + "epoch": 4.812568908489526, + "grad_norm": 0.4370892527431256, + "learning_rate": 1.0769633951511915e-05, + "loss": 0.0789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05111114680767059, + "step": 8730, + "valid_targets_mean": 3456.9, + "valid_targets_min": 873 + }, + { + "epoch": 4.815325248070562, + "grad_norm": 0.44020069463205397, + "learning_rate": 1.0745255180807094e-05, + "loss": 0.0805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043057918548583984, + "step": 8735, + "valid_targets_mean": 3722.5, + "valid_targets_min": 672 + }, + { + "epoch": 4.818081587651599, + "grad_norm": 0.3425934277776865, + "learning_rate": 1.0720893895023134e-05, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031090885400772095, + "step": 8740, + "valid_targets_mean": 3212.6, + "valid_targets_min": 1769 + }, + { + "epoch": 4.820837927232635, + "grad_norm": 0.4079788777323649, + "learning_rate": 1.069655014018562e-05, + "loss": 0.081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03715807572007179, + "step": 8745, + "valid_targets_mean": 3638.0, + "valid_targets_min": 2540 + }, + { + "epoch": 4.8235942668136715, + "grad_norm": 0.42223066937232373, + "learning_rate": 1.067222396228703e-05, + "loss": 0.0704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05115380883216858, + "step": 8750, + "valid_targets_mean": 3400.1, + "valid_targets_min": 1240 + }, + { + "epoch": 4.8263506063947075, + "grad_norm": 0.39239995607476846, + "learning_rate": 1.0647915407286632e-05, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024672508239746094, + "step": 8755, + "valid_targets_mean": 3303.6, + "valid_targets_min": 623 + }, + { + "epoch": 4.829106945975744, + "grad_norm": 0.4440530317230788, + "learning_rate": 1.0623624521110397e-05, + "loss": 0.0664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03301651030778885, + "step": 8760, + "valid_targets_mean": 3239.2, + "valid_targets_min": 589 + }, + { + "epoch": 4.83186328555678, + "grad_norm": 0.37946300456563165, + "learning_rate": 1.059935134965092e-05, + "loss": 0.0715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03843479976058006, + "step": 8765, + "valid_targets_mean": 3886.8, + "valid_targets_min": 3337 + }, + { + "epoch": 4.834619625137817, + "grad_norm": 0.4427802010399419, + "learning_rate": 1.057509593876732e-05, + "loss": 0.0698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0213426873087883, + "step": 8770, + "valid_targets_mean": 1702.9, + "valid_targets_min": 526 + }, + { + "epoch": 4.837375964718853, + "grad_norm": 0.7019541627174667, + "learning_rate": 1.0550858334285165e-05, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09355956315994263, + "step": 8775, + "valid_targets_mean": 1594.4, + "valid_targets_min": 195 + }, + { + "epoch": 4.84013230429989, + "grad_norm": 0.371311500509889, + "learning_rate": 1.0526638581996378e-05, + "loss": 0.0757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04627921059727669, + "step": 8780, + "valid_targets_mean": 3802.2, + "valid_targets_min": 3050 + }, + { + "epoch": 4.842888643880926, + "grad_norm": 0.492342580408381, + "learning_rate": 1.0502436727659169e-05, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05462776869535446, + "step": 8785, + "valid_targets_mean": 2269.0, + "valid_targets_min": 964 + }, + { + "epoch": 4.845644983461963, + "grad_norm": 0.43409629210872025, + "learning_rate": 1.0478252816997918e-05, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038070887327194214, + "step": 8790, + "valid_targets_mean": 3338.1, + "valid_targets_min": 1030 + }, + { + "epoch": 4.8484013230429985, + "grad_norm": 0.6521654596774289, + "learning_rate": 1.0454086895703108e-05, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06396842002868652, + "step": 8795, + "valid_targets_mean": 1621.4, + "valid_targets_min": 916 + }, + { + "epoch": 4.851157662624035, + "grad_norm": 0.5657821829989775, + "learning_rate": 1.0429939009431229e-05, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04818355664610863, + "step": 8800, + "valid_targets_mean": 2154.6, + "valid_targets_min": 1039 + }, + { + "epoch": 4.853914002205071, + "grad_norm": 0.6134895497727506, + "learning_rate": 1.0405809203804713e-05, + "loss": 0.103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08407645672559738, + "step": 8805, + "valid_targets_mean": 1777.0, + "valid_targets_min": 1044 + }, + { + "epoch": 4.856670341786108, + "grad_norm": 0.4526358521094904, + "learning_rate": 1.0381697524411817e-05, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03344227373600006, + "step": 8810, + "valid_targets_mean": 2840.4, + "valid_targets_min": 707 + }, + { + "epoch": 4.859426681367144, + "grad_norm": 0.377244376618932, + "learning_rate": 1.0357604016806562e-05, + "loss": 0.0858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037456318736076355, + "step": 8815, + "valid_targets_mean": 3473.6, + "valid_targets_min": 2850 + }, + { + "epoch": 4.862183020948181, + "grad_norm": 0.3986553868787883, + "learning_rate": 1.0333528726508628e-05, + "loss": 0.0854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030658364295959473, + "step": 8820, + "valid_targets_mean": 2471.0, + "valid_targets_min": 749 + }, + { + "epoch": 4.864939360529217, + "grad_norm": 0.34839709833996324, + "learning_rate": 1.0309471699003288e-05, + "loss": 0.073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03640341758728027, + "step": 8825, + "valid_targets_mean": 4595.0, + "valid_targets_min": 1052 + }, + { + "epoch": 4.867695700110254, + "grad_norm": 0.3350071543881154, + "learning_rate": 1.0285432979741295e-05, + "loss": 0.0697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0364806205034256, + "step": 8830, + "valid_targets_mean": 4013.5, + "valid_targets_min": 2540 + }, + { + "epoch": 4.87045203969129, + "grad_norm": 0.32835726379997304, + "learning_rate": 1.0261412614138839e-05, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03303632140159607, + "step": 8835, + "valid_targets_mean": 3955.2, + "valid_targets_min": 1859 + }, + { + "epoch": 4.873208379272326, + "grad_norm": 0.41533243207518855, + "learning_rate": 1.0237410647577415e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03483493626117706, + "step": 8840, + "valid_targets_mean": 3165.2, + "valid_targets_min": 893 + }, + { + "epoch": 4.875964718853362, + "grad_norm": 0.4579605250098644, + "learning_rate": 1.0213427125403756e-05, + "loss": 0.0873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04111481085419655, + "step": 8845, + "valid_targets_mean": 2416.5, + "valid_targets_min": 753 + }, + { + "epoch": 4.878721058434399, + "grad_norm": 0.31000907382656384, + "learning_rate": 1.0189462092929756e-05, + "loss": 0.0885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02988453023135662, + "step": 8850, + "valid_targets_mean": 4199.1, + "valid_targets_min": 3934 + }, + { + "epoch": 4.881477398015435, + "grad_norm": 0.45664858125465757, + "learning_rate": 1.016551559543237e-05, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04417853057384491, + "step": 8855, + "valid_targets_mean": 3616.2, + "valid_targets_min": 1576 + }, + { + "epoch": 4.884233737596472, + "grad_norm": 0.45851912077024354, + "learning_rate": 1.0141587678153542e-05, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03523758798837662, + "step": 8860, + "valid_targets_mean": 2826.6, + "valid_targets_min": 907 + }, + { + "epoch": 4.886990077177508, + "grad_norm": 0.3562384556328529, + "learning_rate": 1.0117678386300107e-05, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028954891487956047, + "step": 8865, + "valid_targets_mean": 3433.5, + "valid_targets_min": 980 + }, + { + "epoch": 4.889746416758545, + "grad_norm": 0.6021445730992876, + "learning_rate": 1.0093787765043715e-05, + "loss": 0.0814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045668262988328934, + "step": 8870, + "valid_targets_mean": 1216.8, + "valid_targets_min": 572 + }, + { + "epoch": 4.892502756339581, + "grad_norm": 0.5270598363121801, + "learning_rate": 1.0069915859520726e-05, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056421004235744476, + "step": 8875, + "valid_targets_mean": 2862.2, + "valid_targets_min": 1020 + }, + { + "epoch": 4.8952590959206175, + "grad_norm": 0.5561336669132625, + "learning_rate": 1.0046062714832176e-05, + "loss": 0.0815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04982057213783264, + "step": 8880, + "valid_targets_mean": 1887.5, + "valid_targets_min": 680 + }, + { + "epoch": 4.898015435501653, + "grad_norm": 0.48865643599642994, + "learning_rate": 1.0022228376043626e-05, + "loss": 0.0954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03968169540166855, + "step": 8885, + "valid_targets_mean": 2533.8, + "valid_targets_min": 1015 + }, + { + "epoch": 4.90077177508269, + "grad_norm": 0.6013834525796368, + "learning_rate": 9.998412888185117e-06, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046440668404102325, + "step": 8890, + "valid_targets_mean": 1521.4, + "valid_targets_min": 705 + }, + { + "epoch": 4.903528114663727, + "grad_norm": 0.5098852369864809, + "learning_rate": 9.974616296251067e-06, + "loss": 0.0897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02790037915110588, + "step": 8895, + "valid_targets_mean": 2497.8, + "valid_targets_min": 622 + }, + { + "epoch": 4.906284454244763, + "grad_norm": 0.7151602488002191, + "learning_rate": 9.950838645200209e-06, + "loss": 0.1004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03871447592973709, + "step": 8900, + "valid_targets_mean": 3972.1, + "valid_targets_min": 3568 + }, + { + "epoch": 4.909040793825799, + "grad_norm": 0.39571760355334784, + "learning_rate": 9.927079979955479e-06, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033239733427762985, + "step": 8905, + "valid_targets_mean": 2801.8, + "valid_targets_min": 814 + }, + { + "epoch": 4.911797133406836, + "grad_norm": 0.4352383513981877, + "learning_rate": 9.903340345403947e-06, + "loss": 0.0701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04350413382053375, + "step": 8910, + "valid_targets_mean": 3067.6, + "valid_targets_min": 874 + }, + { + "epoch": 4.914553472987873, + "grad_norm": 0.5151989601549686, + "learning_rate": 9.87961978639673e-06, + "loss": 0.0868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08120858669281006, + "step": 8915, + "valid_targets_mean": 2946.4, + "valid_targets_min": 1420 + }, + { + "epoch": 4.9173098125689085, + "grad_norm": 0.409822077829259, + "learning_rate": 9.855918347748905e-06, + "loss": 0.0947, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03193500265479088, + "step": 8920, + "valid_targets_mean": 3060.4, + "valid_targets_min": 1109 + }, + { + "epoch": 4.9200661521499445, + "grad_norm": 0.38161724210716536, + "learning_rate": 9.832236074239415e-06, + "loss": 0.0627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029528655111789703, + "step": 8925, + "valid_targets_mean": 2645.6, + "valid_targets_min": 684 + }, + { + "epoch": 4.922822491730981, + "grad_norm": 0.4368429009728695, + "learning_rate": 9.80857301061102e-06, + "loss": 0.0837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0372949056327343, + "step": 8930, + "valid_targets_mean": 2308.4, + "valid_targets_min": 671 + }, + { + "epoch": 4.925578831312018, + "grad_norm": 0.3802886979476411, + "learning_rate": 9.784929201570166e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03584164381027222, + "step": 8935, + "valid_targets_mean": 4689.0, + "valid_targets_min": 3801 + }, + { + "epoch": 4.928335170893054, + "grad_norm": 0.7243981581575706, + "learning_rate": 9.761304691786921e-06, + "loss": 0.0638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0321965366601944, + "step": 8940, + "valid_targets_mean": 3696.0, + "valid_targets_min": 1243 + }, + { + "epoch": 4.93109151047409, + "grad_norm": 0.3869712002813597, + "learning_rate": 9.737699525894902e-06, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03750772029161453, + "step": 8945, + "valid_targets_mean": 4104.5, + "valid_targets_min": 3322 + }, + { + "epoch": 4.933847850055127, + "grad_norm": 0.5405955971979123, + "learning_rate": 9.714113748491167e-06, + "loss": 0.0729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04424634575843811, + "step": 8950, + "valid_targets_mean": 1621.6, + "valid_targets_min": 680 + }, + { + "epoch": 4.936604189636164, + "grad_norm": 0.44478406626226896, + "learning_rate": 9.69054740413616e-06, + "loss": 0.1045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03872369974851608, + "step": 8955, + "valid_targets_mean": 2706.6, + "valid_targets_min": 564 + }, + { + "epoch": 4.9393605292172, + "grad_norm": 0.2596914651882368, + "learning_rate": 9.667000537353593e-06, + "loss": 0.0887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0175639558583498, + "step": 8960, + "valid_targets_mean": 3608.0, + "valid_targets_min": 929 + }, + { + "epoch": 4.9421168687982355, + "grad_norm": 0.601644718933038, + "learning_rate": 9.643473192630391e-06, + "loss": 0.0941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0850646048784256, + "step": 8965, + "valid_targets_mean": 3499.9, + "valid_targets_min": 3049 + }, + { + "epoch": 4.944873208379272, + "grad_norm": 0.4078391559121084, + "learning_rate": 9.61996541441659e-06, + "loss": 0.0741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038350313901901245, + "step": 8970, + "valid_targets_mean": 3800.0, + "valid_targets_min": 2619 + }, + { + "epoch": 4.947629547960309, + "grad_norm": 0.6938495281219366, + "learning_rate": 9.596477247125253e-06, + "loss": 0.0821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03929572552442551, + "step": 8975, + "valid_targets_mean": 870.1, + "valid_targets_min": 693 + }, + { + "epoch": 4.950385887541345, + "grad_norm": 0.39524057121534517, + "learning_rate": 9.57300873513242e-06, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03986988961696625, + "step": 8980, + "valid_targets_mean": 3939.4, + "valid_targets_min": 3589 + }, + { + "epoch": 4.953142227122381, + "grad_norm": 0.7907756338585721, + "learning_rate": 9.549559922776968e-06, + "loss": 0.1044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05116159841418266, + "step": 8985, + "valid_targets_mean": 1234.0, + "valid_targets_min": 543 + }, + { + "epoch": 4.955898566703418, + "grad_norm": 0.4972681406067724, + "learning_rate": 9.52613085436056e-06, + "loss": 0.1074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055099137127399445, + "step": 8990, + "valid_targets_mean": 2036.9, + "valid_targets_min": 503 + }, + { + "epoch": 4.958654906284455, + "grad_norm": 0.5088097504284836, + "learning_rate": 9.502721574147567e-06, + "loss": 0.0819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051327429711818695, + "step": 8995, + "valid_targets_mean": 2206.4, + "valid_targets_min": 625 + }, + { + "epoch": 4.961411245865491, + "grad_norm": 0.44771221033208825, + "learning_rate": 9.479332126364966e-06, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03366886079311371, + "step": 9000, + "valid_targets_mean": 3586.4, + "valid_targets_min": 2822 + }, + { + "epoch": 4.964167585446527, + "grad_norm": 0.4253178763105831, + "learning_rate": 9.455962555202267e-06, + "loss": 0.078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04645800590515137, + "step": 9005, + "valid_targets_mean": 3573.2, + "valid_targets_min": 2987 + }, + { + "epoch": 4.966923925027563, + "grad_norm": 0.3877519729673687, + "learning_rate": 9.432612904811429e-06, + "loss": 0.0715, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03358805179595947, + "step": 9010, + "valid_targets_mean": 2631.4, + "valid_targets_min": 856 + }, + { + "epoch": 4.9696802646086, + "grad_norm": 0.951704700604764, + "learning_rate": 9.409283219306773e-06, + "loss": 0.0788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04917289316654205, + "step": 9015, + "valid_targets_mean": 835.6, + "valid_targets_min": 606 + }, + { + "epoch": 4.972436604189636, + "grad_norm": 0.42310018483075795, + "learning_rate": 9.385973542764895e-06, + "loss": 0.0804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029722090810537338, + "step": 9020, + "valid_targets_mean": 2728.6, + "valid_targets_min": 751 + }, + { + "epoch": 4.975192943770672, + "grad_norm": 0.6162903734018601, + "learning_rate": 9.362683919224606e-06, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04949551820755005, + "step": 9025, + "valid_targets_mean": 2848.4, + "valid_targets_min": 864 + }, + { + "epoch": 4.977949283351709, + "grad_norm": 0.45421711690722083, + "learning_rate": 9.339414392686816e-06, + "loss": 0.0726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04611853137612343, + "step": 9030, + "valid_targets_mean": 3805.9, + "valid_targets_min": 2900 + }, + { + "epoch": 4.980705622932746, + "grad_norm": 0.35126328472846435, + "learning_rate": 9.316165007114469e-06, + "loss": 0.0822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030206523835659027, + "step": 9035, + "valid_targets_mean": 3559.4, + "valid_targets_min": 485 + }, + { + "epoch": 4.983461962513782, + "grad_norm": 0.32117835914739795, + "learning_rate": 9.29293580643245e-06, + "loss": 0.0688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03324657678604126, + "step": 9040, + "valid_targets_mean": 4905.2, + "valid_targets_min": 2551 + }, + { + "epoch": 4.986218302094818, + "grad_norm": 0.356018024707087, + "learning_rate": 9.269726834527524e-06, + "loss": 0.0746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03918033093214035, + "step": 9045, + "valid_targets_mean": 4418.6, + "valid_targets_min": 3365 + }, + { + "epoch": 4.9889746416758545, + "grad_norm": 0.45491030577409736, + "learning_rate": 9.246538135248227e-06, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04992976784706116, + "step": 9050, + "valid_targets_mean": 2728.2, + "valid_targets_min": 872 + }, + { + "epoch": 4.991730981256891, + "grad_norm": 0.36482499515159567, + "learning_rate": 9.223369752404794e-06, + "loss": 0.1624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03252779319882393, + "step": 9055, + "valid_targets_mean": 2450.1, + "valid_targets_min": 1015 + }, + { + "epoch": 4.994487320837927, + "grad_norm": 0.3885238275639326, + "learning_rate": 9.200221729769084e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051455143839120865, + "step": 9060, + "valid_targets_mean": 2359.6, + "valid_targets_min": 1151 + }, + { + "epoch": 4.997243660418963, + "grad_norm": 0.3713347931163772, + "learning_rate": 9.17709411107448e-06, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030409283936023712, + "step": 9065, + "valid_targets_mean": 3571.2, + "valid_targets_min": 2948 + }, + { + "epoch": 5.0, + "grad_norm": 0.4164155870214026, + "learning_rate": 9.153986940015817e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04649411886930466, + "step": 9070, + "valid_targets_mean": 4470.5, + "valid_targets_min": 1273 + }, + { + "epoch": 5.002756339581037, + "grad_norm": 0.6287425883886955, + "learning_rate": 9.130900260249315e-06, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08950003981590271, + "step": 9075, + "valid_targets_mean": 7389.0, + "valid_targets_min": 5627 + }, + { + "epoch": 5.005512679162073, + "grad_norm": 0.6998852500558269, + "learning_rate": 9.10783411539246e-06, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048513367772102356, + "step": 9080, + "valid_targets_mean": 1294.8, + "valid_targets_min": 422 + }, + { + "epoch": 5.00826901874311, + "grad_norm": 0.4104240898647683, + "learning_rate": 9.084788549023951e-06, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09272446483373642, + "step": 9085, + "valid_targets_mean": 8539.9, + "valid_targets_min": 6761 + }, + { + "epoch": 5.0110253583241455, + "grad_norm": 0.37918682711353363, + "learning_rate": 9.061763604683603e-06, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07931452989578247, + "step": 9090, + "valid_targets_mean": 7579.0, + "valid_targets_min": 6174 + }, + { + "epoch": 5.013781697905182, + "grad_norm": 0.3748668169238068, + "learning_rate": 9.03875932587227e-06, + "loss": 0.1705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07690298557281494, + "step": 9095, + "valid_targets_mean": 7328.2, + "valid_targets_min": 4839 + }, + { + "epoch": 5.016538037486218, + "grad_norm": 0.4076336052030063, + "learning_rate": 9.015775756051774e-06, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09488382190465927, + "step": 9100, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5034 + }, + { + "epoch": 5.019294377067255, + "grad_norm": 0.4122191427884472, + "learning_rate": 8.992812938644798e-06, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09417833387851715, + "step": 9105, + "valid_targets_mean": 7248.2, + "valid_targets_min": 5567 + }, + { + "epoch": 5.022050716648291, + "grad_norm": 0.42596032191763006, + "learning_rate": 8.969870917034821e-06, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07725115120410919, + "step": 9110, + "valid_targets_mean": 6234.0, + "valid_targets_min": 5603 + }, + { + "epoch": 5.024807056229328, + "grad_norm": 0.9729271060323995, + "learning_rate": 8.946949734566037e-06, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024551725015044212, + "step": 9115, + "valid_targets_mean": 602.0, + "valid_targets_min": 152 + }, + { + "epoch": 5.027563395810364, + "grad_norm": 0.40937482559396116, + "learning_rate": 8.92404943454326e-06, + "loss": 0.1559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07909681648015976, + "step": 9120, + "valid_targets_mean": 7299.9, + "valid_targets_min": 6158 + }, + { + "epoch": 5.030319735391401, + "grad_norm": 0.38949963074016547, + "learning_rate": 8.901170060231873e-06, + "loss": 0.1716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07978375256061554, + "step": 9125, + "valid_targets_mean": 6140.0, + "valid_targets_min": 4648 + }, + { + "epoch": 5.033076074972437, + "grad_norm": 0.45893332442750145, + "learning_rate": 8.878311654857697e-06, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0854397788643837, + "step": 9130, + "valid_targets_mean": 6603.0, + "valid_targets_min": 4882 + }, + { + "epoch": 5.035832414553473, + "grad_norm": 0.39855491962266754, + "learning_rate": 8.855474261606948e-06, + "loss": 0.1661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08396764099597931, + "step": 9135, + "valid_targets_mean": 8015.0, + "valid_targets_min": 6344 + }, + { + "epoch": 5.038588754134509, + "grad_norm": 0.3806106696340218, + "learning_rate": 8.832657923626149e-06, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07367017865180969, + "step": 9140, + "valid_targets_mean": 6305.2, + "valid_targets_min": 4622 + }, + { + "epoch": 5.041345093715546, + "grad_norm": 0.39208050855672094, + "learning_rate": 8.809862684022032e-06, + "loss": 0.1657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08695493638515472, + "step": 9145, + "valid_targets_mean": 7890.8, + "valid_targets_min": 5485 + }, + { + "epoch": 5.044101433296582, + "grad_norm": 0.39427439403751846, + "learning_rate": 8.787088585861481e-06, + "loss": 0.1688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07824088633060455, + "step": 9150, + "valid_targets_mean": 7016.6, + "valid_targets_min": 4960 + }, + { + "epoch": 5.046857772877619, + "grad_norm": 0.38585251024142053, + "learning_rate": 8.764335672171428e-06, + "loss": 0.1635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07859860360622406, + "step": 9155, + "valid_targets_mean": 6527.1, + "valid_targets_min": 5399 + }, + { + "epoch": 5.049614112458655, + "grad_norm": 0.4472651926332101, + "learning_rate": 8.74160398593878e-06, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07984521985054016, + "step": 9160, + "valid_targets_mean": 4995.0, + "valid_targets_min": 618 + }, + { + "epoch": 5.052370452039692, + "grad_norm": 0.4010731893344359, + "learning_rate": 8.718893570110343e-06, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09042403846979141, + "step": 9165, + "valid_targets_mean": 7219.4, + "valid_targets_min": 5764 + }, + { + "epoch": 5.055126791620728, + "grad_norm": 0.4221741715364257, + "learning_rate": 8.696204467592749e-06, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08408285677433014, + "step": 9170, + "valid_targets_mean": 7998.5, + "valid_targets_min": 5902 + }, + { + "epoch": 5.0578831312017645, + "grad_norm": 0.38492676627560474, + "learning_rate": 8.673536721252342e-06, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06708844006061554, + "step": 9175, + "valid_targets_mean": 7377.4, + "valid_targets_min": 5015 + }, + { + "epoch": 5.0606394707828, + "grad_norm": 0.42731291868073384, + "learning_rate": 8.65089037391513e-06, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06805749237537384, + "step": 9180, + "valid_targets_mean": 6794.0, + "valid_targets_min": 5283 + }, + { + "epoch": 5.063395810363837, + "grad_norm": 0.42107009467436873, + "learning_rate": 8.62826546836669e-06, + "loss": 0.1678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09243755787611008, + "step": 9185, + "valid_targets_mean": 6656.4, + "valid_targets_min": 4364 + }, + { + "epoch": 5.066152149944873, + "grad_norm": 0.39099846063715665, + "learning_rate": 8.605662047352088e-06, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07883159816265106, + "step": 9190, + "valid_targets_mean": 7178.4, + "valid_targets_min": 4448 + }, + { + "epoch": 5.06890848952591, + "grad_norm": 0.42036741758207996, + "learning_rate": 8.583080153575797e-06, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09880538284778595, + "step": 9195, + "valid_targets_mean": 7619.8, + "valid_targets_min": 5737 + }, + { + "epoch": 5.071664829106946, + "grad_norm": 0.3569958124626343, + "learning_rate": 8.560519829701627e-06, + "loss": 0.1478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06978779286146164, + "step": 9200, + "valid_targets_mean": 8754.4, + "valid_targets_min": 5752 + }, + { + "epoch": 5.074421168687983, + "grad_norm": 0.410890765224846, + "learning_rate": 8.537981118352627e-06, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07648485898971558, + "step": 9205, + "valid_targets_mean": 6341.0, + "valid_targets_min": 4543 + }, + { + "epoch": 5.077177508269019, + "grad_norm": 0.3820297123964515, + "learning_rate": 8.515464062111017e-06, + "loss": 0.1427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08091546595096588, + "step": 9210, + "valid_targets_mean": 7181.9, + "valid_targets_min": 4563 + }, + { + "epoch": 5.0799338478500555, + "grad_norm": 0.7149506762309057, + "learning_rate": 8.492968703518099e-06, + "loss": 0.1695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05435410887002945, + "step": 9215, + "valid_targets_mean": 1104.4, + "valid_targets_min": 161 + }, + { + "epoch": 5.0826901874310915, + "grad_norm": 0.3973114591805267, + "learning_rate": 8.470495085074201e-06, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07462739944458008, + "step": 9220, + "valid_targets_mean": 7018.2, + "valid_targets_min": 5424 + }, + { + "epoch": 5.085446527012128, + "grad_norm": 0.4407160906264473, + "learning_rate": 8.448043249238557e-06, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08889006078243256, + "step": 9225, + "valid_targets_mean": 7738.4, + "valid_targets_min": 5329 + }, + { + "epoch": 5.088202866593164, + "grad_norm": 0.4214047746076292, + "learning_rate": 8.425613238429256e-06, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08301016688346863, + "step": 9230, + "valid_targets_mean": 6426.6, + "valid_targets_min": 5560 + }, + { + "epoch": 5.090959206174201, + "grad_norm": 0.42586582071614065, + "learning_rate": 8.403205095023148e-06, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0843120813369751, + "step": 9235, + "valid_targets_mean": 6169.9, + "valid_targets_min": 3446 + }, + { + "epoch": 5.093715545755237, + "grad_norm": 0.38745787451698316, + "learning_rate": 8.380818861355775e-06, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08617784082889557, + "step": 9240, + "valid_targets_mean": 7204.6, + "valid_targets_min": 5381 + }, + { + "epoch": 5.096471885336274, + "grad_norm": 0.3932567920875333, + "learning_rate": 8.358454579721283e-06, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08688873052597046, + "step": 9245, + "valid_targets_mean": 7600.4, + "valid_targets_min": 5241 + }, + { + "epoch": 5.09922822491731, + "grad_norm": 0.4211848740604058, + "learning_rate": 8.33611229237234e-06, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08955967426300049, + "step": 9250, + "valid_targets_mean": 7098.5, + "valid_targets_min": 5607 + }, + { + "epoch": 5.101984564498347, + "grad_norm": 0.4169208862183568, + "learning_rate": 8.313792041520066e-06, + "loss": 0.1684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09009672701358795, + "step": 9255, + "valid_targets_mean": 6955.1, + "valid_targets_min": 5530 + }, + { + "epoch": 5.1047409040793825, + "grad_norm": 0.40853941218782985, + "learning_rate": 8.29149386933394e-06, + "loss": 0.1702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08663903176784515, + "step": 9260, + "valid_targets_mean": 6748.1, + "valid_targets_min": 5364 + }, + { + "epoch": 5.107497243660419, + "grad_norm": 1.0085192867837203, + "learning_rate": 8.269217817941747e-06, + "loss": 0.171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08319495618343353, + "step": 9265, + "valid_targets_mean": 6273.9, + "valid_targets_min": 5438 + }, + { + "epoch": 5.110253583241455, + "grad_norm": 0.398994340214317, + "learning_rate": 8.246963929429454e-06, + "loss": 0.1656, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07883229106664658, + "step": 9270, + "valid_targets_mean": 6419.5, + "valid_targets_min": 4879 + }, + { + "epoch": 5.113009922822492, + "grad_norm": 0.40872431230124207, + "learning_rate": 8.224732245841176e-06, + "loss": 0.1651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07858918607234955, + "step": 9275, + "valid_targets_mean": 6483.4, + "valid_targets_min": 4309 + }, + { + "epoch": 5.115766262403528, + "grad_norm": 0.44931413582277746, + "learning_rate": 8.202522809179065e-06, + "loss": 0.1361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09789764881134033, + "step": 9280, + "valid_targets_mean": 6825.0, + "valid_targets_min": 5205 + }, + { + "epoch": 5.118522601984565, + "grad_norm": 0.4261767353297368, + "learning_rate": 8.180335661403242e-06, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07687218487262726, + "step": 9285, + "valid_targets_mean": 6729.4, + "valid_targets_min": 5200 + }, + { + "epoch": 5.121278941565601, + "grad_norm": 0.42791158962696185, + "learning_rate": 8.158170844431725e-06, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07875870913267136, + "step": 9290, + "valid_targets_mean": 6094.0, + "valid_targets_min": 5302 + }, + { + "epoch": 5.124035281146638, + "grad_norm": 0.4349113589069798, + "learning_rate": 8.136028400140337e-06, + "loss": 0.1689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.089628204703331, + "step": 9295, + "valid_targets_mean": 7140.0, + "valid_targets_min": 5270 + }, + { + "epoch": 5.126791620727674, + "grad_norm": 0.4112470548192313, + "learning_rate": 8.113908370362633e-06, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07766247540712357, + "step": 9300, + "valid_targets_mean": 7425.0, + "valid_targets_min": 5281 + }, + { + "epoch": 5.12954796030871, + "grad_norm": 0.42698476846123384, + "learning_rate": 8.091810796889825e-06, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1046593189239502, + "step": 9305, + "valid_targets_mean": 7063.4, + "valid_targets_min": 5386 + }, + { + "epoch": 5.132304299889746, + "grad_norm": 0.40725767199057267, + "learning_rate": 8.069735721470685e-06, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08565770089626312, + "step": 9310, + "valid_targets_mean": 6827.0, + "valid_targets_min": 4547 + }, + { + "epoch": 5.135060639470783, + "grad_norm": 0.5281566697647033, + "learning_rate": 8.047683185811501e-06, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03952178359031677, + "step": 9315, + "valid_targets_mean": 2047.0, + "valid_targets_min": 165 + }, + { + "epoch": 5.137816979051819, + "grad_norm": 0.40716497002843677, + "learning_rate": 8.025653231575963e-06, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07930710911750793, + "step": 9320, + "valid_targets_mean": 6754.1, + "valid_targets_min": 5196 + }, + { + "epoch": 5.140573318632856, + "grad_norm": 0.4602549745462907, + "learning_rate": 8.003645900385097e-06, + "loss": 0.1632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08084987103939056, + "step": 9325, + "valid_targets_mean": 5985.6, + "valid_targets_min": 5028 + }, + { + "epoch": 5.143329658213892, + "grad_norm": 0.45066138842252673, + "learning_rate": 7.981661233817196e-06, + "loss": 0.1667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07569888979196548, + "step": 9330, + "valid_targets_mean": 5963.9, + "valid_targets_min": 5237 + }, + { + "epoch": 5.146085997794929, + "grad_norm": 0.42270945808651167, + "learning_rate": 7.959699273407722e-06, + "loss": 0.1618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08487412333488464, + "step": 9335, + "valid_targets_mean": 6319.0, + "valid_targets_min": 5412 + }, + { + "epoch": 5.148842337375965, + "grad_norm": 0.3885456254357696, + "learning_rate": 7.937760060649247e-06, + "loss": 0.1538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06720639765262604, + "step": 9340, + "valid_targets_mean": 7125.8, + "valid_targets_min": 5080 + }, + { + "epoch": 5.1515986769570015, + "grad_norm": 0.43215969918966934, + "learning_rate": 7.915843636991362e-06, + "loss": 0.1652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08058339357376099, + "step": 9345, + "valid_targets_mean": 6123.5, + "valid_targets_min": 4521 + }, + { + "epoch": 5.154355016538037, + "grad_norm": 0.9018565976174738, + "learning_rate": 7.893950043840605e-06, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09741668403148651, + "step": 9350, + "valid_targets_mean": 1886.0, + "valid_targets_min": 1111 + }, + { + "epoch": 5.157111356119074, + "grad_norm": 0.7499131228345041, + "learning_rate": 7.87207932256038e-06, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08249892294406891, + "step": 9355, + "valid_targets_mean": 1497.6, + "valid_targets_min": 893 + }, + { + "epoch": 5.15986769570011, + "grad_norm": 0.7327968663798908, + "learning_rate": 7.85023151447087e-06, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05139123275876045, + "step": 9360, + "valid_targets_mean": 1356.0, + "valid_targets_min": 914 + }, + { + "epoch": 5.162624035281147, + "grad_norm": 0.7384973869810015, + "learning_rate": 7.82840666084899e-06, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05653153359889984, + "step": 9365, + "valid_targets_mean": 1324.5, + "valid_targets_min": 680 + }, + { + "epoch": 5.165380374862183, + "grad_norm": 0.7684729201583548, + "learning_rate": 7.806604802928272e-06, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06960443407297134, + "step": 9370, + "valid_targets_mean": 1931.8, + "valid_targets_min": 876 + }, + { + "epoch": 5.16813671444322, + "grad_norm": 0.8002815473659572, + "learning_rate": 7.7848259818988e-06, + "loss": 0.1354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059808921068906784, + "step": 9375, + "valid_targets_mean": 1292.2, + "valid_targets_min": 734 + }, + { + "epoch": 5.170893054024256, + "grad_norm": 0.8142804099571869, + "learning_rate": 7.76307023890714e-06, + "loss": 0.1465, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061800092458724976, + "step": 9380, + "valid_targets_mean": 1469.6, + "valid_targets_min": 655 + }, + { + "epoch": 5.1736493936052925, + "grad_norm": 0.7880174085894266, + "learning_rate": 7.741337615056259e-06, + "loss": 0.124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07316965609788895, + "step": 9385, + "valid_targets_mean": 1548.4, + "valid_targets_min": 737 + }, + { + "epoch": 5.1764057331863285, + "grad_norm": 0.7554715162856147, + "learning_rate": 7.71962815140544e-06, + "loss": 0.1439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05990245193243027, + "step": 9390, + "valid_targets_mean": 1316.5, + "valid_targets_min": 759 + }, + { + "epoch": 5.179162072767365, + "grad_norm": 0.7788337047540141, + "learning_rate": 7.697941888970213e-06, + "loss": 0.1329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05191852152347565, + "step": 9395, + "valid_targets_mean": 1249.6, + "valid_targets_min": 753 + }, + { + "epoch": 5.181918412348401, + "grad_norm": 0.8115848116513763, + "learning_rate": 7.67627886872227e-06, + "loss": 0.1331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08080168068408966, + "step": 9400, + "valid_targets_mean": 1833.0, + "valid_targets_min": 927 + }, + { + "epoch": 5.184674751929438, + "grad_norm": 0.9097219143930516, + "learning_rate": 7.654639131589394e-06, + "loss": 0.1346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06380439549684525, + "step": 9405, + "valid_targets_mean": 1338.4, + "valid_targets_min": 679 + }, + { + "epoch": 5.187431091510474, + "grad_norm": 0.8103692789446969, + "learning_rate": 7.633022718455387e-06, + "loss": 0.1322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06401893496513367, + "step": 9410, + "valid_targets_mean": 1244.6, + "valid_targets_min": 771 + }, + { + "epoch": 5.190187431091511, + "grad_norm": 0.6927793357742599, + "learning_rate": 7.611429670159976e-06, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06003626435995102, + "step": 9415, + "valid_targets_mean": 1732.5, + "valid_targets_min": 986 + }, + { + "epoch": 5.192943770672547, + "grad_norm": 0.7648129870782827, + "learning_rate": 7.589860027498746e-06, + "loss": 0.1268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05825617164373398, + "step": 9420, + "valid_targets_mean": 1389.9, + "valid_targets_min": 1103 + }, + { + "epoch": 5.195700110253584, + "grad_norm": 0.8502471672161547, + "learning_rate": 7.568313831223062e-06, + "loss": 0.1299, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07196826487779617, + "step": 9425, + "valid_targets_mean": 1593.2, + "valid_targets_min": 675 + }, + { + "epoch": 5.1984564498346195, + "grad_norm": 0.7816535379543088, + "learning_rate": 7.546791122039994e-06, + "loss": 0.1341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07823461294174194, + "step": 9430, + "valid_targets_mean": 1926.6, + "valid_targets_min": 974 + }, + { + "epoch": 5.201212789415656, + "grad_norm": 0.7872308755079952, + "learning_rate": 7.525291940612236e-06, + "loss": 0.1337, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07676415145397186, + "step": 9435, + "valid_targets_mean": 1953.1, + "valid_targets_min": 869 + }, + { + "epoch": 5.203969128996692, + "grad_norm": 0.7511524967452499, + "learning_rate": 7.503816327558036e-06, + "loss": 0.1319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056879736483097076, + "step": 9440, + "valid_targets_mean": 1327.2, + "valid_targets_min": 1023 + }, + { + "epoch": 5.206725468577729, + "grad_norm": 0.7938464060275623, + "learning_rate": 7.482364323451106e-06, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06877030432224274, + "step": 9445, + "valid_targets_mean": 1717.9, + "valid_targets_min": 709 + }, + { + "epoch": 5.209481808158765, + "grad_norm": 0.8523239569695993, + "learning_rate": 7.46093596882056e-06, + "loss": 0.132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05594945698976517, + "step": 9450, + "valid_targets_mean": 1139.9, + "valid_targets_min": 658 + }, + { + "epoch": 5.212238147739802, + "grad_norm": 0.7755995319783854, + "learning_rate": 7.439531304150824e-06, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06204564496874809, + "step": 9455, + "valid_targets_mean": 1444.8, + "valid_targets_min": 662 + }, + { + "epoch": 5.214994487320838, + "grad_norm": 0.7826643651062807, + "learning_rate": 7.4181503698815845e-06, + "loss": 0.1322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04769410565495491, + "step": 9460, + "valid_targets_mean": 1242.2, + "valid_targets_min": 859 + }, + { + "epoch": 5.217750826901875, + "grad_norm": 0.7984073819857292, + "learning_rate": 7.396793206407677e-06, + "loss": 0.1173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06961159408092499, + "step": 9465, + "valid_targets_mean": 1569.2, + "valid_targets_min": 950 + }, + { + "epoch": 5.220507166482911, + "grad_norm": 0.8484478612665184, + "learning_rate": 7.375459854079034e-06, + "loss": 0.1268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06427760422229767, + "step": 9470, + "valid_targets_mean": 1539.8, + "valid_targets_min": 686 + }, + { + "epoch": 5.223263506063947, + "grad_norm": 0.8992372306751923, + "learning_rate": 7.354150353200596e-06, + "loss": 0.1267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07229681313037872, + "step": 9475, + "valid_targets_mean": 1610.6, + "valid_targets_min": 702 + }, + { + "epoch": 5.226019845644983, + "grad_norm": 0.8100493476165674, + "learning_rate": 7.332864744032246e-06, + "loss": 0.125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047055214643478394, + "step": 9480, + "valid_targets_mean": 1190.4, + "valid_targets_min": 779 + }, + { + "epoch": 5.22877618522602, + "grad_norm": 0.8213389670041202, + "learning_rate": 7.31160306678873e-06, + "loss": 0.1377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06914189457893372, + "step": 9485, + "valid_targets_mean": 1467.1, + "valid_targets_min": 822 + }, + { + "epoch": 5.231532524807056, + "grad_norm": 0.7666197544810105, + "learning_rate": 7.290365361639577e-06, + "loss": 0.1244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05807756632566452, + "step": 9490, + "valid_targets_mean": 1350.5, + "valid_targets_min": 868 + }, + { + "epoch": 5.234288864388093, + "grad_norm": 0.7934533810916561, + "learning_rate": 7.269151668709024e-06, + "loss": 0.1287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051963068544864655, + "step": 9495, + "valid_targets_mean": 1283.5, + "valid_targets_min": 781 + }, + { + "epoch": 5.237045203969129, + "grad_norm": 0.7839123574096966, + "learning_rate": 7.247962028075946e-06, + "loss": 0.1274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057137809693813324, + "step": 9500, + "valid_targets_mean": 1303.2, + "valid_targets_min": 793 + }, + { + "epoch": 5.239801543550166, + "grad_norm": 0.7055718909869786, + "learning_rate": 7.226796479773763e-06, + "loss": 0.1285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05104418098926544, + "step": 9505, + "valid_targets_mean": 1309.1, + "valid_targets_min": 806 + }, + { + "epoch": 5.242557883131202, + "grad_norm": 0.7684302890873762, + "learning_rate": 7.205655063790408e-06, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05304491147398949, + "step": 9510, + "valid_targets_mean": 1397.6, + "valid_targets_min": 803 + }, + { + "epoch": 5.2453142227122385, + "grad_norm": 0.7575385001407081, + "learning_rate": 7.184537820068194e-06, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047933585941791534, + "step": 9515, + "valid_targets_mean": 1187.8, + "valid_targets_min": 660 + }, + { + "epoch": 5.248070562293274, + "grad_norm": 0.7877380846822787, + "learning_rate": 7.1634447885037705e-06, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0574641227722168, + "step": 9520, + "valid_targets_mean": 1547.0, + "valid_targets_min": 853 + }, + { + "epoch": 5.250826901874311, + "grad_norm": 0.8140009933518288, + "learning_rate": 7.142376008948049e-06, + "loss": 0.1287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06659401953220367, + "step": 9525, + "valid_targets_mean": 1704.2, + "valid_targets_min": 719 + }, + { + "epoch": 5.253583241455347, + "grad_norm": 0.8393174756650995, + "learning_rate": 7.121331521206114e-06, + "loss": 0.1267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05018569529056549, + "step": 9530, + "valid_targets_mean": 1255.0, + "valid_targets_min": 765 + }, + { + "epoch": 5.256339581036384, + "grad_norm": 0.8436909420771066, + "learning_rate": 7.1003113650371645e-06, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06812391430139542, + "step": 9535, + "valid_targets_mean": 1550.8, + "valid_targets_min": 875 + }, + { + "epoch": 5.25909592061742, + "grad_norm": 0.8023155922514585, + "learning_rate": 7.079315580154425e-06, + "loss": 0.1231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06176289543509483, + "step": 9540, + "valid_targets_mean": 1362.0, + "valid_targets_min": 739 + }, + { + "epoch": 5.261852260198457, + "grad_norm": 0.8373287969456584, + "learning_rate": 7.058344206225074e-06, + "loss": 0.1266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07029099017381668, + "step": 9545, + "valid_targets_mean": 1666.8, + "valid_targets_min": 760 + }, + { + "epoch": 5.264608599779493, + "grad_norm": 0.8263575630126925, + "learning_rate": 7.037397282870171e-06, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05301699414849281, + "step": 9550, + "valid_targets_mean": 1319.6, + "valid_targets_min": 1046 + }, + { + "epoch": 5.2673649393605295, + "grad_norm": 0.8124859450912627, + "learning_rate": 7.016474849664592e-06, + "loss": 0.1297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07182811200618744, + "step": 9555, + "valid_targets_mean": 1727.9, + "valid_targets_min": 851 + }, + { + "epoch": 5.2701212789415655, + "grad_norm": 0.8430306929427734, + "learning_rate": 6.995576946136928e-06, + "loss": 0.1316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06796213239431381, + "step": 9560, + "valid_targets_mean": 1643.2, + "valid_targets_min": 907 + }, + { + "epoch": 5.272877618522602, + "grad_norm": 0.8025127198870289, + "learning_rate": 6.974703611769438e-06, + "loss": 0.1325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05780463665723801, + "step": 9565, + "valid_targets_mean": 1478.5, + "valid_targets_min": 764 + }, + { + "epoch": 5.275633958103638, + "grad_norm": 0.7548687183603374, + "learning_rate": 6.9538548859979574e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056000128388404846, + "step": 9570, + "valid_targets_mean": 1522.6, + "valid_targets_min": 822 + }, + { + "epoch": 5.278390297684675, + "grad_norm": 0.75095694135792, + "learning_rate": 6.933030808211825e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06004592776298523, + "step": 9575, + "valid_targets_mean": 1589.0, + "valid_targets_min": 982 + }, + { + "epoch": 5.281146637265711, + "grad_norm": 0.7910918648717721, + "learning_rate": 6.912231417753818e-06, + "loss": 0.1294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05975862219929695, + "step": 9580, + "valid_targets_mean": 1574.8, + "valid_targets_min": 816 + }, + { + "epoch": 5.283902976846748, + "grad_norm": 0.8165534203240029, + "learning_rate": 6.891456753920076e-06, + "loss": 0.1233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06046292185783386, + "step": 9585, + "valid_targets_mean": 1422.0, + "valid_targets_min": 721 + }, + { + "epoch": 5.286659316427784, + "grad_norm": 0.7698997925577203, + "learning_rate": 6.8707068559600095e-06, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047070130705833435, + "step": 9590, + "valid_targets_mean": 1150.6, + "valid_targets_min": 741 + }, + { + "epoch": 5.289415656008821, + "grad_norm": 0.8290051921331925, + "learning_rate": 6.849981763076255e-06, + "loss": 0.1257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07121714949607849, + "step": 9595, + "valid_targets_mean": 1684.2, + "valid_targets_min": 734 + }, + { + "epoch": 5.2921719955898565, + "grad_norm": 0.807620342556852, + "learning_rate": 6.829281514424564e-06, + "loss": 0.1248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07628461718559265, + "step": 9600, + "valid_targets_mean": 1937.1, + "valid_targets_min": 744 + }, + { + "epoch": 5.294928335170893, + "grad_norm": 0.8278163482528178, + "learning_rate": 6.808606149113779e-06, + "loss": 0.1259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0714419037103653, + "step": 9605, + "valid_targets_mean": 1842.2, + "valid_targets_min": 620 + }, + { + "epoch": 5.297684674751929, + "grad_norm": 0.8542000554724274, + "learning_rate": 6.787955706205709e-06, + "loss": 0.1203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0732768326997757, + "step": 9610, + "valid_targets_mean": 1637.6, + "valid_targets_min": 785 + }, + { + "epoch": 5.300441014332966, + "grad_norm": 0.7806526243104633, + "learning_rate": 6.76733022471508e-06, + "loss": 0.1251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06764659285545349, + "step": 9615, + "valid_targets_mean": 1969.9, + "valid_targets_min": 740 + }, + { + "epoch": 5.303197353914002, + "grad_norm": 0.7918525054808784, + "learning_rate": 6.746729743609466e-06, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05738300085067749, + "step": 9620, + "valid_targets_mean": 1432.1, + "valid_targets_min": 711 + }, + { + "epoch": 5.305953693495039, + "grad_norm": 0.7394644284433025, + "learning_rate": 6.726154301809198e-06, + "loss": 0.126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06533952057361603, + "step": 9625, + "valid_targets_mean": 1695.6, + "valid_targets_min": 1122 + }, + { + "epoch": 5.308710033076075, + "grad_norm": 0.7611829118418365, + "learning_rate": 6.705603938187308e-06, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06340274214744568, + "step": 9630, + "valid_targets_mean": 1524.5, + "valid_targets_min": 722 + }, + { + "epoch": 5.311466372657112, + "grad_norm": 0.773600264537803, + "learning_rate": 6.685078691569446e-06, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05463189259171486, + "step": 9635, + "valid_targets_mean": 1289.2, + "valid_targets_min": 661 + }, + { + "epoch": 5.314222712238148, + "grad_norm": 0.7916134603878063, + "learning_rate": 6.664578600733811e-06, + "loss": 0.1138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057815637439489365, + "step": 9640, + "valid_targets_mean": 1519.5, + "valid_targets_min": 1011 + }, + { + "epoch": 5.316979051819184, + "grad_norm": 0.8080825358672397, + "learning_rate": 6.64410370441106e-06, + "loss": 0.1255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06752176582813263, + "step": 9645, + "valid_targets_mean": 1731.4, + "valid_targets_min": 1277 + }, + { + "epoch": 5.31973539140022, + "grad_norm": 0.7823391035618867, + "learning_rate": 6.623654041284282e-06, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06353241950273514, + "step": 9650, + "valid_targets_mean": 1486.1, + "valid_targets_min": 807 + }, + { + "epoch": 5.322491730981257, + "grad_norm": 0.79478414422437, + "learning_rate": 6.603229649988867e-06, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06282871961593628, + "step": 9655, + "valid_targets_mean": 1549.9, + "valid_targets_min": 679 + }, + { + "epoch": 5.325248070562293, + "grad_norm": 0.7812568522406627, + "learning_rate": 6.5828305691124615e-06, + "loss": 0.1243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06186152994632721, + "step": 9660, + "valid_targets_mean": 1656.0, + "valid_targets_min": 1294 + }, + { + "epoch": 5.32800441014333, + "grad_norm": 0.7746952631071106, + "learning_rate": 6.562456837194903e-06, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05604814738035202, + "step": 9665, + "valid_targets_mean": 1368.8, + "valid_targets_min": 790 + }, + { + "epoch": 5.330760749724366, + "grad_norm": 0.8230595894872133, + "learning_rate": 6.54210849272813e-06, + "loss": 0.1244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06772781908512115, + "step": 9670, + "valid_targets_mean": 1765.8, + "valid_targets_min": 592 + }, + { + "epoch": 5.333517089305403, + "grad_norm": 0.8252468527478534, + "learning_rate": 6.521785574156119e-06, + "loss": 0.1221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06634710729122162, + "step": 9675, + "valid_targets_mean": 1586.8, + "valid_targets_min": 662 + }, + { + "epoch": 5.336273428886439, + "grad_norm": 0.8507707069321597, + "learning_rate": 6.501488119874811e-06, + "loss": 0.1214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055437665432691574, + "step": 9680, + "valid_targets_mean": 1336.9, + "valid_targets_min": 622 + }, + { + "epoch": 5.3390297684674755, + "grad_norm": 0.8087573145343762, + "learning_rate": 6.481216168232034e-06, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039332639425992966, + "step": 9685, + "valid_targets_mean": 1097.1, + "valid_targets_min": 625 + }, + { + "epoch": 5.341786108048511, + "grad_norm": 0.8906990345748933, + "learning_rate": 6.460969757527438e-06, + "loss": 0.1324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05329204350709915, + "step": 9690, + "valid_targets_mean": 1484.9, + "valid_targets_min": 888 + }, + { + "epoch": 5.344542447629548, + "grad_norm": 0.7915478863249309, + "learning_rate": 6.440748926012408e-06, + "loss": 0.1283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06572801619768143, + "step": 9695, + "valid_targets_mean": 1757.0, + "valid_targets_min": 635 + }, + { + "epoch": 5.347298787210584, + "grad_norm": 0.8607139677924318, + "learning_rate": 6.420553711890025e-06, + "loss": 0.1234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06778993457555771, + "step": 9700, + "valid_targets_mean": 1476.1, + "valid_targets_min": 522 + }, + { + "epoch": 5.350055126791621, + "grad_norm": 0.7823867615059517, + "learning_rate": 6.400384153314953e-06, + "loss": 0.1225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05170275270938873, + "step": 9705, + "valid_targets_mean": 1230.9, + "valid_targets_min": 697 + }, + { + "epoch": 5.352811466372657, + "grad_norm": 0.8027074331024887, + "learning_rate": 6.380240288393389e-06, + "loss": 0.1161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06518241763114929, + "step": 9710, + "valid_targets_mean": 1671.0, + "valid_targets_min": 1010 + }, + { + "epoch": 5.355567805953694, + "grad_norm": 0.8555211917903962, + "learning_rate": 6.3601221551829865e-06, + "loss": 0.1168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05680500343441963, + "step": 9715, + "valid_targets_mean": 1411.0, + "valid_targets_min": 769 + }, + { + "epoch": 5.35832414553473, + "grad_norm": 0.9138357410605112, + "learning_rate": 6.340029791692788e-06, + "loss": 0.1223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0561920590698719, + "step": 9720, + "valid_targets_mean": 1298.6, + "valid_targets_min": 745 + }, + { + "epoch": 5.3610804851157665, + "grad_norm": 0.8611241376137627, + "learning_rate": 6.319963235883142e-06, + "loss": 0.1171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046066250652074814, + "step": 9725, + "valid_targets_mean": 1118.9, + "valid_targets_min": 680 + }, + { + "epoch": 5.3638368246968025, + "grad_norm": 0.7679940513986213, + "learning_rate": 6.299922525665649e-06, + "loss": 0.1172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07572624087333679, + "step": 9730, + "valid_targets_mean": 1905.9, + "valid_targets_min": 1347 + }, + { + "epoch": 5.366593164277839, + "grad_norm": 0.9328546037708514, + "learning_rate": 6.279907698903071e-06, + "loss": 0.1272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08783890306949615, + "step": 9735, + "valid_targets_mean": 1636.0, + "valid_targets_min": 851 + }, + { + "epoch": 5.369349503858875, + "grad_norm": 0.7907928342089855, + "learning_rate": 6.259918793409276e-06, + "loss": 0.1187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05180340260267258, + "step": 9740, + "valid_targets_mean": 1457.2, + "valid_targets_min": 903 + }, + { + "epoch": 5.372105843439912, + "grad_norm": 0.7913517804895548, + "learning_rate": 6.239955846949144e-06, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.064766526222229, + "step": 9745, + "valid_targets_mean": 1674.4, + "valid_targets_min": 1240 + }, + { + "epoch": 5.374862183020948, + "grad_norm": 0.603939585120721, + "learning_rate": 6.220018897238538e-06, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04988217353820801, + "step": 9750, + "valid_targets_mean": 1697.1, + "valid_targets_min": 917 + }, + { + "epoch": 5.377618522601985, + "grad_norm": 0.7602698467050504, + "learning_rate": 6.200107981944181e-06, + "loss": 0.1077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05516316741704941, + "step": 9755, + "valid_targets_mean": 1580.9, + "valid_targets_min": 930 + }, + { + "epoch": 5.380374862183021, + "grad_norm": 0.805270295484969, + "learning_rate": 6.180223138683619e-06, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0635179802775383, + "step": 9760, + "valid_targets_mean": 1655.1, + "valid_targets_min": 1227 + }, + { + "epoch": 5.383131201764058, + "grad_norm": 0.773687447860335, + "learning_rate": 6.160364405025141e-06, + "loss": 0.1168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0589483305811882, + "step": 9765, + "valid_targets_mean": 1565.8, + "valid_targets_min": 870 + }, + { + "epoch": 5.3858875413450935, + "grad_norm": 0.820844449971196, + "learning_rate": 6.1405318184877e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05996646732091904, + "step": 9770, + "valid_targets_mean": 1564.6, + "valid_targets_min": 822 + }, + { + "epoch": 5.38864388092613, + "grad_norm": 0.874022439709675, + "learning_rate": 6.120725416540861e-06, + "loss": 0.1247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05863456428050995, + "step": 9775, + "valid_targets_mean": 1540.5, + "valid_targets_min": 1214 + }, + { + "epoch": 5.391400220507166, + "grad_norm": 0.8036684699714407, + "learning_rate": 6.1009452366047075e-06, + "loss": 0.1152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053471602499485016, + "step": 9780, + "valid_targets_mean": 1268.6, + "valid_targets_min": 591 + }, + { + "epoch": 5.394156560088203, + "grad_norm": 0.7789202958124845, + "learning_rate": 6.08119131604979e-06, + "loss": 0.1208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047239162027835846, + "step": 9785, + "valid_targets_mean": 1232.5, + "valid_targets_min": 646 + }, + { + "epoch": 5.396912899669239, + "grad_norm": 1.0507379474173126, + "learning_rate": 6.061463692197036e-06, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1185937225818634, + "step": 9790, + "valid_targets_mean": 5097.2, + "valid_targets_min": 1971 + }, + { + "epoch": 5.399669239250276, + "grad_norm": 0.8315968460285696, + "learning_rate": 6.0417624023177145e-06, + "loss": 0.227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11832074820995331, + "step": 9795, + "valid_targets_mean": 5217.0, + "valid_targets_min": 1631 + }, + { + "epoch": 5.402425578831312, + "grad_norm": 0.7009891617479235, + "learning_rate": 6.022087483633319e-06, + "loss": 0.2254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14191794395446777, + "step": 9800, + "valid_targets_mean": 5966.0, + "valid_targets_min": 3294 + }, + { + "epoch": 5.405181918412349, + "grad_norm": 0.6621724430742648, + "learning_rate": 6.0024389733155275e-06, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08743992447853088, + "step": 9805, + "valid_targets_mean": 4246.1, + "valid_targets_min": 2736 + }, + { + "epoch": 5.407938257993385, + "grad_norm": 0.5930701423613242, + "learning_rate": 5.982816908486122e-06, + "loss": 0.2126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09182260930538177, + "step": 9810, + "valid_targets_mean": 3896.2, + "valid_targets_min": 1439 + }, + { + "epoch": 5.410694597574421, + "grad_norm": 0.5403422969332776, + "learning_rate": 5.963221326216931e-06, + "loss": 0.2223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08809615671634674, + "step": 9815, + "valid_targets_mean": 4481.8, + "valid_targets_min": 1512 + }, + { + "epoch": 5.413450937155457, + "grad_norm": 0.5418526822152258, + "learning_rate": 5.9436522635297375e-06, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11375907063484192, + "step": 9820, + "valid_targets_mean": 5127.9, + "valid_targets_min": 2405 + }, + { + "epoch": 5.416207276736494, + "grad_norm": 0.5401129374600557, + "learning_rate": 5.924109757396228e-06, + "loss": 0.2143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08411034941673279, + "step": 9825, + "valid_targets_mean": 3253.8, + "valid_targets_min": 1654 + }, + { + "epoch": 5.41896361631753, + "grad_norm": 0.4743654977926586, + "learning_rate": 5.904593844737916e-06, + "loss": 0.2102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10529467463493347, + "step": 9830, + "valid_targets_mean": 5750.2, + "valid_targets_min": 4644 + }, + { + "epoch": 5.421719955898567, + "grad_norm": 0.588592175506997, + "learning_rate": 5.885104562426067e-06, + "loss": 0.2229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10333730280399323, + "step": 9835, + "valid_targets_mean": 4141.6, + "valid_targets_min": 1544 + }, + { + "epoch": 5.424476295479603, + "grad_norm": 0.49659060490763135, + "learning_rate": 5.865641947281638e-06, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1037105917930603, + "step": 9840, + "valid_targets_mean": 5869.8, + "valid_targets_min": 953 + }, + { + "epoch": 5.42723263506064, + "grad_norm": 0.8386097329973943, + "learning_rate": 5.84620603607521e-06, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11549592763185501, + "step": 9845, + "valid_targets_mean": 4973.6, + "valid_targets_min": 1265 + }, + { + "epoch": 5.429988974641676, + "grad_norm": 0.5217705449199311, + "learning_rate": 5.8267968655269026e-06, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1002914234995842, + "step": 9850, + "valid_targets_mean": 5359.9, + "valid_targets_min": 1982 + }, + { + "epoch": 5.4327453142227125, + "grad_norm": 0.589688576391544, + "learning_rate": 5.807414472306319e-06, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0966816395521164, + "step": 9855, + "valid_targets_mean": 3892.5, + "valid_targets_min": 1759 + }, + { + "epoch": 5.435501653803748, + "grad_norm": 0.6206310541817519, + "learning_rate": 5.788058893032469e-06, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.098000168800354, + "step": 9860, + "valid_targets_mean": 3185.8, + "valid_targets_min": 1451 + }, + { + "epoch": 5.438257993384785, + "grad_norm": 0.5943009462250293, + "learning_rate": 5.7687301642737035e-06, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10360642522573471, + "step": 9865, + "valid_targets_mean": 4283.2, + "valid_targets_min": 1861 + }, + { + "epoch": 5.441014332965821, + "grad_norm": 0.6042865049391751, + "learning_rate": 5.749428322547652e-06, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10924914479255676, + "step": 9870, + "valid_targets_mean": 4340.8, + "valid_targets_min": 2345 + }, + { + "epoch": 5.443770672546858, + "grad_norm": 0.5906988839380525, + "learning_rate": 5.730153404321137e-06, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10864311456680298, + "step": 9875, + "valid_targets_mean": 3923.9, + "valid_targets_min": 2767 + }, + { + "epoch": 5.446527012127894, + "grad_norm": 0.6307548886025672, + "learning_rate": 5.710905446010122e-06, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10840160399675369, + "step": 9880, + "valid_targets_mean": 3651.8, + "valid_targets_min": 1205 + }, + { + "epoch": 5.449283351708931, + "grad_norm": 0.5602544010700703, + "learning_rate": 5.691684483979629e-06, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08745454996824265, + "step": 9885, + "valid_targets_mean": 3835.4, + "valid_targets_min": 2004 + }, + { + "epoch": 5.452039691289967, + "grad_norm": 0.6174622010627452, + "learning_rate": 5.672490554543673e-06, + "loss": 0.2112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10086788237094879, + "step": 9890, + "valid_targets_mean": 3778.0, + "valid_targets_min": 1020 + }, + { + "epoch": 5.4547960308710035, + "grad_norm": 0.6069304982942756, + "learning_rate": 5.653323693965218e-06, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11614396423101425, + "step": 9895, + "valid_targets_mean": 4307.8, + "valid_targets_min": 2308 + }, + { + "epoch": 5.4575523704520394, + "grad_norm": 0.5914961832606614, + "learning_rate": 5.634183938456061e-06, + "loss": 0.1931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10859249532222748, + "step": 9900, + "valid_targets_mean": 4553.9, + "valid_targets_min": 2046 + }, + { + "epoch": 5.460308710033076, + "grad_norm": 0.6552635340295919, + "learning_rate": 5.615071324176806e-06, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13640441000461578, + "step": 9905, + "valid_targets_mean": 4660.2, + "valid_targets_min": 3373 + }, + { + "epoch": 5.463065049614112, + "grad_norm": 0.6382454114467285, + "learning_rate": 5.59598588723677e-06, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1038421168923378, + "step": 9910, + "valid_targets_mean": 3727.4, + "valid_targets_min": 1747 + }, + { + "epoch": 5.465821389195149, + "grad_norm": 0.6447242092563158, + "learning_rate": 5.576927663693929e-06, + "loss": 0.2063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0920390635728836, + "step": 9915, + "valid_targets_mean": 3326.1, + "valid_targets_min": 1570 + }, + { + "epoch": 5.468577728776185, + "grad_norm": 0.7095974615698987, + "learning_rate": 5.557896689554843e-06, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09703696519136429, + "step": 9920, + "valid_targets_mean": 4600.8, + "valid_targets_min": 1537 + }, + { + "epoch": 5.471334068357222, + "grad_norm": 0.5561511181197509, + "learning_rate": 5.53889300077459e-06, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08840534836053848, + "step": 9925, + "valid_targets_mean": 4328.9, + "valid_targets_min": 1934 + }, + { + "epoch": 5.474090407938258, + "grad_norm": 0.6286163788963307, + "learning_rate": 5.519916633256699e-06, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08585743606090546, + "step": 9930, + "valid_targets_mean": 3494.0, + "valid_targets_min": 2562 + }, + { + "epoch": 5.476846747519295, + "grad_norm": 0.6030072491982261, + "learning_rate": 5.500967622853073e-06, + "loss": 0.197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0996018722653389, + "step": 9935, + "valid_targets_mean": 4076.8, + "valid_targets_min": 2659 + }, + { + "epoch": 5.4796030871003305, + "grad_norm": 0.583501668686857, + "learning_rate": 5.482046005363952e-06, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09014305472373962, + "step": 9940, + "valid_targets_mean": 4040.6, + "valid_targets_min": 1349 + }, + { + "epoch": 5.482359426681367, + "grad_norm": 0.6497993389334279, + "learning_rate": 5.463151816537802e-06, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07902663946151733, + "step": 9945, + "valid_targets_mean": 3000.6, + "valid_targets_min": 1567 + }, + { + "epoch": 5.485115766262403, + "grad_norm": 0.6525644782466111, + "learning_rate": 5.4442850920712704e-06, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07981274276971817, + "step": 9950, + "valid_targets_mean": 2628.4, + "valid_targets_min": 1500 + }, + { + "epoch": 5.48787210584344, + "grad_norm": 0.5703031024632731, + "learning_rate": 5.4254458676091245e-06, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0834328681230545, + "step": 9955, + "valid_targets_mean": 3413.6, + "valid_targets_min": 2033 + }, + { + "epoch": 5.490628445424476, + "grad_norm": 0.7559294667305987, + "learning_rate": 5.406634178744168e-06, + "loss": 0.2008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11895156651735306, + "step": 9960, + "valid_targets_mean": 2852.0, + "valid_targets_min": 1346 + }, + { + "epoch": 5.493384785005513, + "grad_norm": 0.7653482121060263, + "learning_rate": 5.387850061017186e-06, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11395329236984253, + "step": 9965, + "valid_targets_mean": 2981.6, + "valid_targets_min": 1821 + }, + { + "epoch": 5.496141124586549, + "grad_norm": 0.6341511533831207, + "learning_rate": 5.369093549916873e-06, + "loss": 0.1985, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10190219432115555, + "step": 9970, + "valid_targets_mean": 3951.2, + "valid_targets_min": 2535 + }, + { + "epoch": 5.498897464167586, + "grad_norm": 0.6467711788572762, + "learning_rate": 5.350364680879769e-06, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09147093445062637, + "step": 9975, + "valid_targets_mean": 3048.8, + "valid_targets_min": 2080 + }, + { + "epoch": 5.501653803748622, + "grad_norm": 0.6361136811823609, + "learning_rate": 5.331663489290184e-06, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08800049871206284, + "step": 9980, + "valid_targets_mean": 2975.6, + "valid_targets_min": 1153 + }, + { + "epoch": 5.504410143329658, + "grad_norm": 0.6658660891069653, + "learning_rate": 5.312990010480137e-06, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09004563838243484, + "step": 9985, + "valid_targets_mean": 3207.6, + "valid_targets_min": 1858 + }, + { + "epoch": 5.507166482910694, + "grad_norm": 0.6242026891229103, + "learning_rate": 5.294344279729307e-06, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08953320980072021, + "step": 9990, + "valid_targets_mean": 3643.6, + "valid_targets_min": 2197 + }, + { + "epoch": 5.509922822491731, + "grad_norm": 0.6010863785746396, + "learning_rate": 5.2757263322649235e-06, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07524748146533966, + "step": 9995, + "valid_targets_mean": 3769.4, + "valid_targets_min": 1394 + }, + { + "epoch": 5.512679162072767, + "grad_norm": 0.6049589607622924, + "learning_rate": 5.257136203261741e-06, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10305860638618469, + "step": 10000, + "valid_targets_mean": 4348.6, + "valid_targets_min": 1844 + }, + { + "epoch": 5.515435501653804, + "grad_norm": 0.6349132857139035, + "learning_rate": 5.2385739278419525e-06, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11376864463090897, + "step": 10005, + "valid_targets_mean": 4341.1, + "valid_targets_min": 1734 + }, + { + "epoch": 5.51819184123484, + "grad_norm": 0.6245854024604846, + "learning_rate": 5.220039541075126e-06, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10989496111869812, + "step": 10010, + "valid_targets_mean": 4252.6, + "valid_targets_min": 2077 + }, + { + "epoch": 5.520948180815877, + "grad_norm": 0.6084501863832185, + "learning_rate": 5.201533077978141e-06, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09706464409828186, + "step": 10015, + "valid_targets_mean": 4177.9, + "valid_targets_min": 2349 + }, + { + "epoch": 5.523704520396913, + "grad_norm": 0.5957760393786873, + "learning_rate": 5.18305457351512e-06, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10871130228042603, + "step": 10020, + "valid_targets_mean": 4160.2, + "valid_targets_min": 1607 + }, + { + "epoch": 5.5264608599779494, + "grad_norm": 0.6891760628236288, + "learning_rate": 5.164604062597369e-06, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10562652349472046, + "step": 10025, + "valid_targets_mean": 3273.0, + "valid_targets_min": 1377 + }, + { + "epoch": 5.529217199558985, + "grad_norm": 0.5927323420815415, + "learning_rate": 5.1461815800832935e-06, + "loss": 0.1993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08917251229286194, + "step": 10030, + "valid_targets_mean": 3304.5, + "valid_targets_min": 1902 + }, + { + "epoch": 5.531973539140022, + "grad_norm": 0.6290782264224402, + "learning_rate": 5.127787160778366e-06, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08902949094772339, + "step": 10035, + "valid_targets_mean": 3324.1, + "valid_targets_min": 654 + }, + { + "epoch": 5.534729878721058, + "grad_norm": 0.6087601063392238, + "learning_rate": 5.109420839435018e-06, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0987095832824707, + "step": 10040, + "valid_targets_mean": 3907.8, + "valid_targets_min": 2607 + }, + { + "epoch": 5.537486218302095, + "grad_norm": 0.6495787932389602, + "learning_rate": 5.0910826507526055e-06, + "loss": 0.2009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10356637835502625, + "step": 10045, + "valid_targets_mean": 4467.2, + "valid_targets_min": 1739 + }, + { + "epoch": 5.540242557883131, + "grad_norm": 0.6332721124327321, + "learning_rate": 5.072772629377334e-06, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08762010931968689, + "step": 10050, + "valid_targets_mean": 3224.1, + "valid_targets_min": 1140 + }, + { + "epoch": 5.542998897464168, + "grad_norm": 0.5981593904254959, + "learning_rate": 5.054490809902188e-06, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09178424626588821, + "step": 10055, + "valid_targets_mean": 3543.4, + "valid_targets_min": 2501 + }, + { + "epoch": 5.545755237045204, + "grad_norm": 0.6038060146033585, + "learning_rate": 5.03623722686688e-06, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09875612705945969, + "step": 10060, + "valid_targets_mean": 4364.1, + "valid_targets_min": 1298 + }, + { + "epoch": 5.5485115766262405, + "grad_norm": 0.6045263843081549, + "learning_rate": 5.018011914757763e-06, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10252553224563599, + "step": 10065, + "valid_targets_mean": 4377.9, + "valid_targets_min": 1101 + }, + { + "epoch": 5.551267916207276, + "grad_norm": 0.6272165499554502, + "learning_rate": 4.999814908007787e-06, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10709838569164276, + "step": 10070, + "valid_targets_mean": 3894.2, + "valid_targets_min": 2079 + }, + { + "epoch": 5.554024255788313, + "grad_norm": 0.6139901237409496, + "learning_rate": 4.981646240996421e-06, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07953338325023651, + "step": 10075, + "valid_targets_mean": 3054.4, + "valid_targets_min": 1743 + }, + { + "epoch": 5.556780595369349, + "grad_norm": 0.6020981891703987, + "learning_rate": 4.963505948049587e-06, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10315129160881042, + "step": 10080, + "valid_targets_mean": 3688.0, + "valid_targets_min": 1234 + }, + { + "epoch": 5.559536934950386, + "grad_norm": 0.6370789803998861, + "learning_rate": 4.945394063439621e-06, + "loss": 0.2027, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10798792541027069, + "step": 10085, + "valid_targets_mean": 3790.0, + "valid_targets_min": 1822 + }, + { + "epoch": 5.562293274531422, + "grad_norm": 0.5984341921163994, + "learning_rate": 4.927310621385162e-06, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10043574124574661, + "step": 10090, + "valid_targets_mean": 4106.9, + "valid_targets_min": 1651 + }, + { + "epoch": 5.565049614112459, + "grad_norm": 0.627961965722196, + "learning_rate": 4.909255656051126e-06, + "loss": 0.1959, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10550257563591003, + "step": 10095, + "valid_targets_mean": 3287.2, + "valid_targets_min": 1374 + }, + { + "epoch": 5.567805953693495, + "grad_norm": 0.6438540876589002, + "learning_rate": 4.891229201548626e-06, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10774396359920502, + "step": 10100, + "valid_targets_mean": 4399.2, + "valid_targets_min": 2725 + }, + { + "epoch": 5.570562293274532, + "grad_norm": 0.5750160525633067, + "learning_rate": 4.873231291934905e-06, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09933438152074814, + "step": 10105, + "valid_targets_mean": 4269.0, + "valid_targets_min": 1966 + }, + { + "epoch": 5.5733186328555675, + "grad_norm": 0.6652096470446953, + "learning_rate": 4.855261961213285e-06, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10257354378700256, + "step": 10110, + "valid_targets_mean": 3716.2, + "valid_targets_min": 1561 + }, + { + "epoch": 5.576074972436604, + "grad_norm": 0.6735605132864844, + "learning_rate": 4.8373212433330884e-06, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09631796181201935, + "step": 10115, + "valid_targets_mean": 3750.1, + "valid_targets_min": 1444 + }, + { + "epoch": 5.57883131201764, + "grad_norm": 0.6066847540985302, + "learning_rate": 4.819409172189578e-06, + "loss": 0.1759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07779453694820404, + "step": 10120, + "valid_targets_mean": 3042.1, + "valid_targets_min": 1504 + }, + { + "epoch": 5.581587651598677, + "grad_norm": 0.6749894825877648, + "learning_rate": 4.801525781623899e-06, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08670322597026825, + "step": 10125, + "valid_targets_mean": 3153.0, + "valid_targets_min": 858 + }, + { + "epoch": 5.584343991179713, + "grad_norm": 0.6995451251700214, + "learning_rate": 4.7836711054230025e-06, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10984198749065399, + "step": 10130, + "valid_targets_mean": 4063.5, + "valid_targets_min": 1288 + }, + { + "epoch": 5.58710033076075, + "grad_norm": 0.6257963705252517, + "learning_rate": 4.765845177319608e-06, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09115734696388245, + "step": 10135, + "valid_targets_mean": 3914.9, + "valid_targets_min": 1537 + }, + { + "epoch": 5.589856670341786, + "grad_norm": 0.7683695541713659, + "learning_rate": 4.748048030992105e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10501518845558167, + "step": 10140, + "valid_targets_mean": 4236.5, + "valid_targets_min": 2972 + }, + { + "epoch": 5.592613009922823, + "grad_norm": 0.6278859810045703, + "learning_rate": 4.730279700064508e-06, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10661099851131439, + "step": 10145, + "valid_targets_mean": 4411.0, + "valid_targets_min": 2013 + }, + { + "epoch": 5.595369349503859, + "grad_norm": 0.6754796976041855, + "learning_rate": 4.712540218106394e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09621165692806244, + "step": 10150, + "valid_targets_mean": 3548.4, + "valid_targets_min": 1744 + }, + { + "epoch": 5.598125689084895, + "grad_norm": 0.6162052605607824, + "learning_rate": 4.694829618632834e-06, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08420822024345398, + "step": 10155, + "valid_targets_mean": 4129.6, + "valid_targets_min": 1914 + }, + { + "epoch": 5.600882028665931, + "grad_norm": 0.6318665565348551, + "learning_rate": 4.677147935104334e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08037639409303665, + "step": 10160, + "valid_targets_mean": 3826.4, + "valid_targets_min": 2414 + }, + { + "epoch": 5.603638368246968, + "grad_norm": 0.5970848903914144, + "learning_rate": 4.659495200926767e-06, + "loss": 0.1789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10262730717658997, + "step": 10165, + "valid_targets_mean": 4195.0, + "valid_targets_min": 1937 + }, + { + "epoch": 5.606394707828004, + "grad_norm": 0.6678809451102694, + "learning_rate": 4.6418714494513115e-06, + "loss": 0.1969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10332827270030975, + "step": 10170, + "valid_targets_mean": 3154.9, + "valid_targets_min": 1649 + }, + { + "epoch": 5.609151047409041, + "grad_norm": 0.6940472276841765, + "learning_rate": 4.624276713974389e-06, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08273456245660782, + "step": 10175, + "valid_targets_mean": 3088.5, + "valid_targets_min": 1463 + }, + { + "epoch": 5.611907386990077, + "grad_norm": 0.6769764789260587, + "learning_rate": 4.60671102773761e-06, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09613484144210815, + "step": 10180, + "valid_targets_mean": 3241.4, + "valid_targets_min": 1426 + }, + { + "epoch": 5.614663726571114, + "grad_norm": 0.6538109446073231, + "learning_rate": 4.589174423927692e-06, + "loss": 0.1971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07858619093894958, + "step": 10185, + "valid_targets_mean": 2768.9, + "valid_targets_min": 538 + }, + { + "epoch": 5.61742006615215, + "grad_norm": 0.5916816399195927, + "learning_rate": 4.571666935676411e-06, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10007632523775101, + "step": 10190, + "valid_targets_mean": 4483.5, + "valid_targets_min": 1646 + }, + { + "epoch": 5.620176405733186, + "grad_norm": 0.6985917124019474, + "learning_rate": 4.554188596060532e-06, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11314556747674942, + "step": 10195, + "valid_targets_mean": 3761.9, + "valid_targets_min": 2169 + }, + { + "epoch": 5.622932745314222, + "grad_norm": 0.6592775799908733, + "learning_rate": 4.536739438101756e-06, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08677846938371658, + "step": 10200, + "valid_targets_mean": 3168.4, + "valid_targets_min": 1625 + }, + { + "epoch": 5.625689084895259, + "grad_norm": 0.5909519263780183, + "learning_rate": 4.519319494766645e-06, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08142441511154175, + "step": 10205, + "valid_targets_mean": 3449.0, + "valid_targets_min": 1997 + }, + { + "epoch": 5.628445424476295, + "grad_norm": 0.644508943897817, + "learning_rate": 4.501928798966573e-06, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12072750180959702, + "step": 10210, + "valid_targets_mean": 4406.4, + "valid_targets_min": 2949 + }, + { + "epoch": 5.631201764057332, + "grad_norm": 0.5682574986444868, + "learning_rate": 4.484567383557648e-06, + "loss": 0.1969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10797111690044403, + "step": 10215, + "valid_targets_mean": 5358.5, + "valid_targets_min": 2652 + }, + { + "epoch": 5.633958103638368, + "grad_norm": 0.6200893162019737, + "learning_rate": 4.467235281340667e-06, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11172810941934586, + "step": 10220, + "valid_targets_mean": 4314.5, + "valid_targets_min": 1381 + }, + { + "epoch": 5.636714443219405, + "grad_norm": 0.6526045461312353, + "learning_rate": 4.449932525061036e-06, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08064618706703186, + "step": 10225, + "valid_targets_mean": 3395.9, + "valid_targets_min": 637 + }, + { + "epoch": 5.639470782800441, + "grad_norm": 0.633542606851885, + "learning_rate": 4.432659147408736e-06, + "loss": 0.1914, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08411476761102676, + "step": 10230, + "valid_targets_mean": 3256.4, + "valid_targets_min": 719 + }, + { + "epoch": 5.6422271223814775, + "grad_norm": 0.6571960765011223, + "learning_rate": 4.415415181018223e-06, + "loss": 0.2001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0926264151930809, + "step": 10235, + "valid_targets_mean": 4725.1, + "valid_targets_min": 3210 + }, + { + "epoch": 5.644983461962513, + "grad_norm": 0.6531519713058611, + "learning_rate": 4.398200658468397e-06, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10036902129650116, + "step": 10240, + "valid_targets_mean": 3775.4, + "valid_targets_min": 1372 + }, + { + "epoch": 5.64773980154355, + "grad_norm": 0.6864031494979287, + "learning_rate": 4.381015612282529e-06, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11498215794563293, + "step": 10245, + "valid_targets_mean": 4266.1, + "valid_targets_min": 2768 + }, + { + "epoch": 5.650496141124586, + "grad_norm": 0.6183062975014961, + "learning_rate": 4.363860074928197e-06, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07273196429014206, + "step": 10250, + "valid_targets_mean": 2540.5, + "valid_targets_min": 530 + }, + { + "epoch": 5.653252480705623, + "grad_norm": 0.6509859773886898, + "learning_rate": 4.346734078817234e-06, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09752316027879715, + "step": 10255, + "valid_targets_mean": 3903.2, + "valid_targets_min": 1841 + }, + { + "epoch": 5.656008820286659, + "grad_norm": 0.6760959967792044, + "learning_rate": 4.329637656305654e-06, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08936239778995514, + "step": 10260, + "valid_targets_mean": 3476.8, + "valid_targets_min": 1728 + }, + { + "epoch": 5.658765159867696, + "grad_norm": 0.6937770648664427, + "learning_rate": 4.312570839693606e-06, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10721161961555481, + "step": 10265, + "valid_targets_mean": 3712.2, + "valid_targets_min": 1362 + }, + { + "epoch": 5.661521499448732, + "grad_norm": 0.5961941561340756, + "learning_rate": 4.295533661225295e-06, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09604410827159882, + "step": 10270, + "valid_targets_mean": 3693.8, + "valid_targets_min": 852 + }, + { + "epoch": 5.664277839029769, + "grad_norm": 0.670812061856129, + "learning_rate": 4.2785261530889335e-06, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08651874959468842, + "step": 10275, + "valid_targets_mean": 2822.6, + "valid_targets_min": 1195 + }, + { + "epoch": 5.6670341786108045, + "grad_norm": 0.684228045806374, + "learning_rate": 4.261548347416691e-06, + "loss": 0.1915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11524487286806107, + "step": 10280, + "valid_targets_mean": 4747.8, + "valid_targets_min": 1397 + }, + { + "epoch": 5.669790518191841, + "grad_norm": 0.6584354317437892, + "learning_rate": 4.244600276284607e-06, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07702432572841644, + "step": 10285, + "valid_targets_mean": 3487.6, + "valid_targets_min": 1975 + }, + { + "epoch": 5.672546857772877, + "grad_norm": 0.6577668019945223, + "learning_rate": 4.227681971712543e-06, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10127304494380951, + "step": 10290, + "valid_targets_mean": 4387.6, + "valid_targets_min": 2564 + }, + { + "epoch": 5.675303197353914, + "grad_norm": 0.6038932600257564, + "learning_rate": 4.210793465664127e-06, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08997679501771927, + "step": 10295, + "valid_targets_mean": 6404.1, + "valid_targets_min": 2595 + }, + { + "epoch": 5.67805953693495, + "grad_norm": 0.6340728642466201, + "learning_rate": 4.19393479004669e-06, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0969192385673523, + "step": 10300, + "valid_targets_mean": 4013.2, + "valid_targets_min": 2868 + }, + { + "epoch": 5.680815876515987, + "grad_norm": 0.6352834259074204, + "learning_rate": 4.177105976711204e-06, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0784451961517334, + "step": 10305, + "valid_targets_mean": 2886.0, + "valid_targets_min": 993 + }, + { + "epoch": 5.683572216097023, + "grad_norm": 0.6174227436616805, + "learning_rate": 4.160307057452218e-06, + "loss": 0.177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08509683609008789, + "step": 10310, + "valid_targets_mean": 3805.1, + "valid_targets_min": 1575 + }, + { + "epoch": 5.68632855567806, + "grad_norm": 0.7310484344311533, + "learning_rate": 4.143538064007808e-06, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1045471727848053, + "step": 10315, + "valid_targets_mean": 3722.0, + "valid_targets_min": 1360 + }, + { + "epoch": 5.6890848952590956, + "grad_norm": 0.6614805158137332, + "learning_rate": 4.126799028059505e-06, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08636932820081711, + "step": 10320, + "valid_targets_mean": 3669.4, + "valid_targets_min": 2414 + }, + { + "epoch": 5.691841234840132, + "grad_norm": 0.6814735485601177, + "learning_rate": 4.110089981232253e-06, + "loss": 0.1979, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10713367164134979, + "step": 10325, + "valid_targets_mean": 4327.4, + "valid_targets_min": 1454 + }, + { + "epoch": 5.694597574421168, + "grad_norm": 0.6762380233075796, + "learning_rate": 4.093410955094328e-06, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09172564744949341, + "step": 10330, + "valid_targets_mean": 3619.9, + "valid_targets_min": 1720 + }, + { + "epoch": 5.697353914002205, + "grad_norm": 0.7054553177167843, + "learning_rate": 4.076761981157289e-06, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09728621691465378, + "step": 10335, + "valid_targets_mean": 3105.5, + "valid_targets_min": 1278 + }, + { + "epoch": 5.700110253583241, + "grad_norm": 0.7358296804953542, + "learning_rate": 4.060143090875921e-06, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10228489339351654, + "step": 10340, + "valid_targets_mean": 3147.8, + "valid_targets_min": 1858 + }, + { + "epoch": 5.702866593164278, + "grad_norm": 0.7033761411645387, + "learning_rate": 4.043554315648166e-06, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0767245665192604, + "step": 10345, + "valid_targets_mean": 2984.5, + "valid_targets_min": 628 + }, + { + "epoch": 5.705622932745314, + "grad_norm": 0.7777532999857604, + "learning_rate": 4.026995686815081e-06, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11184249818325043, + "step": 10350, + "valid_targets_mean": 2602.6, + "valid_targets_min": 1039 + }, + { + "epoch": 5.708379272326351, + "grad_norm": 0.662738358493995, + "learning_rate": 4.010467235660753e-06, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09488224983215332, + "step": 10355, + "valid_targets_mean": 4317.8, + "valid_targets_min": 2530 + }, + { + "epoch": 5.711135611907387, + "grad_norm": 0.6535778162551846, + "learning_rate": 3.9939689934122694e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1062779575586319, + "step": 10360, + "valid_targets_mean": 3711.0, + "valid_targets_min": 809 + }, + { + "epoch": 5.713891951488423, + "grad_norm": 0.6290465394111918, + "learning_rate": 3.977500991239631e-06, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07900494337081909, + "step": 10365, + "valid_targets_mean": 3450.5, + "valid_targets_min": 1780 + }, + { + "epoch": 5.716648291069459, + "grad_norm": 0.6270154818490707, + "learning_rate": 3.961063260255713e-06, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10349388420581818, + "step": 10370, + "valid_targets_mean": 4356.6, + "valid_targets_min": 2185 + }, + { + "epoch": 5.719404630650496, + "grad_norm": 0.47396140391197167, + "learning_rate": 3.944655831516204e-06, + "loss": 0.158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03962010145187378, + "step": 10375, + "valid_targets_mean": 2424.2, + "valid_targets_min": 876 + }, + { + "epoch": 5.722160970231532, + "grad_norm": 0.49928617026945793, + "learning_rate": 3.928278736019538e-06, + "loss": 0.0967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03295748680830002, + "step": 10380, + "valid_targets_mean": 1550.6, + "valid_targets_min": 669 + }, + { + "epoch": 5.724917309812569, + "grad_norm": 0.3818493149241301, + "learning_rate": 3.911932004706836e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04298469051718712, + "step": 10385, + "valid_targets_mean": 3182.0, + "valid_targets_min": 2264 + }, + { + "epoch": 5.727673649393605, + "grad_norm": 0.40126875847844024, + "learning_rate": 3.895615668461856e-06, + "loss": 0.0734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035019777715206146, + "step": 10390, + "valid_targets_mean": 3035.2, + "valid_targets_min": 710 + }, + { + "epoch": 5.730429988974642, + "grad_norm": 0.34014406527186064, + "learning_rate": 3.8793297581109325e-06, + "loss": 0.0856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02653319388628006, + "step": 10395, + "valid_targets_mean": 3213.8, + "valid_targets_min": 1223 + }, + { + "epoch": 5.733186328555679, + "grad_norm": 0.39827135599529034, + "learning_rate": 3.863074304422918e-06, + "loss": 0.0734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04135286435484886, + "step": 10400, + "valid_targets_mean": 3550.9, + "valid_targets_min": 2558 + }, + { + "epoch": 5.7359426681367145, + "grad_norm": 0.8013837458076982, + "learning_rate": 3.846849338109116e-06, + "loss": 0.1072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0530775785446167, + "step": 10405, + "valid_targets_mean": 966.1, + "valid_targets_min": 559 + }, + { + "epoch": 5.73869900771775, + "grad_norm": 0.38798643056627274, + "learning_rate": 3.830654889823238e-06, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0347786545753479, + "step": 10410, + "valid_targets_mean": 3475.8, + "valid_targets_min": 1013 + }, + { + "epoch": 5.741455347298787, + "grad_norm": 0.4229030435857725, + "learning_rate": 3.8144909901613303e-06, + "loss": 0.0839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04271835833787918, + "step": 10415, + "valid_targets_mean": 3533.5, + "valid_targets_min": 3049 + }, + { + "epoch": 5.744211686879824, + "grad_norm": 0.47904371774936433, + "learning_rate": 3.7983576696617364e-06, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06912767887115479, + "step": 10420, + "valid_targets_mean": 3382.9, + "valid_targets_min": 1949 + }, + { + "epoch": 5.74696802646086, + "grad_norm": 0.5195740482535015, + "learning_rate": 3.7822549588050185e-06, + "loss": 0.0632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037995751947164536, + "step": 10425, + "valid_targets_mean": 3985.5, + "valid_targets_min": 929 + }, + { + "epoch": 5.749724366041896, + "grad_norm": 0.4544039449543665, + "learning_rate": 3.766182888013905e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038354936987161636, + "step": 10430, + "valid_targets_mean": 2270.4, + "valid_targets_min": 447 + }, + { + "epoch": 5.752480705622933, + "grad_norm": 0.36411888288356975, + "learning_rate": 3.7501414876532425e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035912804305553436, + "step": 10435, + "valid_targets_mean": 3307.2, + "valid_targets_min": 754 + }, + { + "epoch": 5.75523704520397, + "grad_norm": 0.5243676010249338, + "learning_rate": 3.7341307880299306e-06, + "loss": 0.0954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05137958377599716, + "step": 10440, + "valid_targets_mean": 1430.6, + "valid_targets_min": 604 + }, + { + "epoch": 5.7579933847850056, + "grad_norm": 0.782637140257056, + "learning_rate": 3.718150819392863e-06, + "loss": 0.0909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03883840888738632, + "step": 10445, + "valid_targets_mean": 3397.1, + "valid_targets_min": 1862 + }, + { + "epoch": 5.7607497243660415, + "grad_norm": 0.41050384352533514, + "learning_rate": 3.7022016119328763e-06, + "loss": 0.0735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0337601974606514, + "step": 10450, + "valid_targets_mean": 2834.5, + "valid_targets_min": 574 + }, + { + "epoch": 5.763506063947078, + "grad_norm": 0.52992988598071, + "learning_rate": 3.686283195782692e-06, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0328754298388958, + "step": 10455, + "valid_targets_mean": 2592.2, + "valid_targets_min": 523 + }, + { + "epoch": 5.766262403528115, + "grad_norm": 0.5044352988814064, + "learning_rate": 3.670395601016854e-06, + "loss": 0.074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04858379065990448, + "step": 10460, + "valid_targets_mean": 3489.2, + "valid_targets_min": 2189 + }, + { + "epoch": 5.769018743109151, + "grad_norm": 0.4086852133030197, + "learning_rate": 3.654538857651675e-06, + "loss": 0.0778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03138654679059982, + "step": 10465, + "valid_targets_mean": 3051.6, + "valid_targets_min": 812 + }, + { + "epoch": 5.771775082690187, + "grad_norm": 0.4439806688714205, + "learning_rate": 3.638712995645188e-06, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03418845310807228, + "step": 10470, + "valid_targets_mean": 1836.4, + "valid_targets_min": 597 + }, + { + "epoch": 5.774531422271224, + "grad_norm": 0.52065747161447, + "learning_rate": 3.622918044897077e-06, + "loss": 0.084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04114387184381485, + "step": 10475, + "valid_targets_mean": 2217.6, + "valid_targets_min": 718 + }, + { + "epoch": 5.777287761852261, + "grad_norm": 0.5450144231896091, + "learning_rate": 3.6071540352486234e-06, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09723679721355438, + "step": 10480, + "valid_targets_mean": 3171.4, + "valid_targets_min": 1297 + }, + { + "epoch": 5.780044101433297, + "grad_norm": 0.5720372232515836, + "learning_rate": 3.591420996482655e-06, + "loss": 0.0935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035124771296978, + "step": 10485, + "valid_targets_mean": 2718.0, + "valid_targets_min": 786 + }, + { + "epoch": 5.7828004410143325, + "grad_norm": 0.6646733475534774, + "learning_rate": 3.5757189583234817e-06, + "loss": 0.0748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05701075866818428, + "step": 10490, + "valid_targets_mean": 1962.9, + "valid_targets_min": 954 + }, + { + "epoch": 5.785556780595369, + "grad_norm": 0.4254487726224609, + "learning_rate": 3.5600479504368533e-06, + "loss": 0.1042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04113711044192314, + "step": 10495, + "valid_targets_mean": 3074.2, + "valid_targets_min": 1304 + }, + { + "epoch": 5.788313120176406, + "grad_norm": 0.4322228819952317, + "learning_rate": 3.5444080024298844e-06, + "loss": 0.0931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028675921261310577, + "step": 10500, + "valid_targets_mean": 3154.2, + "valid_targets_min": 794 + }, + { + "epoch": 5.791069459757442, + "grad_norm": 0.5579548105836135, + "learning_rate": 3.528799143851016e-06, + "loss": 0.08, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04868166521191597, + "step": 10505, + "valid_targets_mean": 2348.6, + "valid_targets_min": 717 + }, + { + "epoch": 5.793825799338478, + "grad_norm": 0.6005067867593166, + "learning_rate": 3.5132214041899482e-06, + "loss": 0.261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10796661674976349, + "step": 10510, + "valid_targets_mean": 2661.4, + "valid_targets_min": 987 + }, + { + "epoch": 5.796582138919515, + "grad_norm": 0.6129227945094345, + "learning_rate": 3.497674812877583e-06, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07499954849481583, + "step": 10515, + "valid_targets_mean": 1800.9, + "valid_targets_min": 757 + }, + { + "epoch": 5.799338478500552, + "grad_norm": 0.4647056116709096, + "learning_rate": 3.4821593992859893e-06, + "loss": 0.0822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03271179646253586, + "step": 10520, + "valid_targets_mean": 1464.6, + "valid_targets_min": 673 + }, + { + "epoch": 5.802094818081588, + "grad_norm": 0.5488906969864392, + "learning_rate": 3.4666751927283217e-06, + "loss": 0.1727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044862501323223114, + "step": 10525, + "valid_targets_mean": 2554.9, + "valid_targets_min": 666 + }, + { + "epoch": 5.804851157662624, + "grad_norm": 0.3082985148123354, + "learning_rate": 3.4512222224587742e-06, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0241299606859684, + "step": 10530, + "valid_targets_mean": 5195.2, + "valid_targets_min": 3529 + }, + { + "epoch": 5.80760749724366, + "grad_norm": 0.475354415581038, + "learning_rate": 3.4358005176725317e-06, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0503893718123436, + "step": 10535, + "valid_targets_mean": 4099.0, + "valid_targets_min": 2336 + }, + { + "epoch": 5.810363836824697, + "grad_norm": 0.46636772468680376, + "learning_rate": 3.4204101075057026e-06, + "loss": 0.0789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041669245809316635, + "step": 10540, + "valid_targets_mean": 3430.0, + "valid_targets_min": 1226 + }, + { + "epoch": 5.813120176405733, + "grad_norm": 0.39990917263557746, + "learning_rate": 3.4050510210352793e-06, + "loss": 0.0717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04218015819787979, + "step": 10545, + "valid_targets_mean": 4274.5, + "valid_targets_min": 3275 + }, + { + "epoch": 5.815876515986769, + "grad_norm": 0.4288928553933431, + "learning_rate": 3.3897232872790675e-06, + "loss": 0.0727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034764934331178665, + "step": 10550, + "valid_targets_mean": 2150.2, + "valid_targets_min": 723 + }, + { + "epoch": 5.818632855567806, + "grad_norm": 0.4503506215542405, + "learning_rate": 3.374426935195643e-06, + "loss": 0.069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04112943634390831, + "step": 10555, + "valid_targets_mean": 3443.2, + "valid_targets_min": 2234 + }, + { + "epoch": 5.821389195148843, + "grad_norm": 0.3414228981484224, + "learning_rate": 3.3591619936842832e-06, + "loss": 0.0692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025745894759893417, + "step": 10560, + "valid_targets_mean": 3956.0, + "valid_targets_min": 3226 + }, + { + "epoch": 5.824145534729879, + "grad_norm": 0.4778303228444589, + "learning_rate": 3.34392849158494e-06, + "loss": 0.07, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034124862402677536, + "step": 10565, + "valid_targets_mean": 2450.4, + "valid_targets_min": 739 + }, + { + "epoch": 5.826901874310915, + "grad_norm": 0.45357164089359436, + "learning_rate": 3.328726457678153e-06, + "loss": 0.1074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032301925122737885, + "step": 10570, + "valid_targets_mean": 1556.5, + "valid_targets_min": 606 + }, + { + "epoch": 5.8296582138919515, + "grad_norm": 0.32331244622963384, + "learning_rate": 3.3135559206850075e-06, + "loss": 0.0574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02269374206662178, + "step": 10575, + "valid_targets_mean": 1918.8, + "valid_targets_min": 585 + }, + { + "epoch": 5.832414553472988, + "grad_norm": 0.4395243378707259, + "learning_rate": 3.2984169092670927e-06, + "loss": 0.0713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03863196074962616, + "step": 10580, + "valid_targets_mean": 2679.6, + "valid_targets_min": 582 + }, + { + "epoch": 5.835170893054024, + "grad_norm": 0.4553946569443369, + "learning_rate": 3.2833094520264265e-06, + "loss": 0.0644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040224380791187286, + "step": 10585, + "valid_targets_mean": 3722.2, + "valid_targets_min": 3260 + }, + { + "epoch": 5.83792723263506, + "grad_norm": 0.45075204371193855, + "learning_rate": 3.2682335775054174e-06, + "loss": 0.1051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044177889823913574, + "step": 10590, + "valid_targets_mean": 3233.8, + "valid_targets_min": 999 + }, + { + "epoch": 5.840683572216097, + "grad_norm": 0.6223339161321185, + "learning_rate": 3.253189314186802e-06, + "loss": 0.09, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06959833204746246, + "step": 10595, + "valid_targets_mean": 2054.0, + "valid_targets_min": 696 + }, + { + "epoch": 5.843439911797134, + "grad_norm": 0.462088389340766, + "learning_rate": 3.2381766904935997e-06, + "loss": 0.0892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035623885691165924, + "step": 10600, + "valid_targets_mean": 2716.5, + "valid_targets_min": 635 + }, + { + "epoch": 5.84619625137817, + "grad_norm": 0.45772597743846927, + "learning_rate": 3.223195734789044e-06, + "loss": 0.0754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0330030582845211, + "step": 10605, + "valid_targets_mean": 1854.2, + "valid_targets_min": 816 + }, + { + "epoch": 5.848952590959206, + "grad_norm": 0.5851854287680309, + "learning_rate": 3.208246475376542e-06, + "loss": 0.1201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1414634883403778, + "step": 10610, + "valid_targets_mean": 3260.4, + "valid_targets_min": 1183 + }, + { + "epoch": 5.8517089305402425, + "grad_norm": 0.4581770528330838, + "learning_rate": 3.193328940499629e-06, + "loss": 0.0812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03929039463400841, + "step": 10615, + "valid_targets_mean": 3414.1, + "valid_targets_min": 2342 + }, + { + "epoch": 5.854465270121279, + "grad_norm": 0.6212271972631411, + "learning_rate": 3.178443158341886e-06, + "loss": 0.0994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047269005328416824, + "step": 10620, + "valid_targets_mean": 1638.2, + "valid_targets_min": 687 + }, + { + "epoch": 5.857221609702315, + "grad_norm": 0.4843253980233887, + "learning_rate": 3.1635891570269115e-06, + "loss": 0.0692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026598358526825905, + "step": 10625, + "valid_targets_mean": 1557.5, + "valid_targets_min": 449 + }, + { + "epoch": 5.859977949283351, + "grad_norm": 0.4117415869905507, + "learning_rate": 3.148766964618262e-06, + "loss": 0.0782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02659780904650688, + "step": 10630, + "valid_targets_mean": 2998.2, + "valid_targets_min": 658 + }, + { + "epoch": 5.862734288864388, + "grad_norm": 0.6211470847078804, + "learning_rate": 3.1339766091193937e-06, + "loss": 0.0833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03644643351435661, + "step": 10635, + "valid_targets_mean": 1299.1, + "valid_targets_min": 577 + }, + { + "epoch": 5.865490628445425, + "grad_norm": 0.3742818990817659, + "learning_rate": 3.119218118473617e-06, + "loss": 0.0677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04009568691253662, + "step": 10640, + "valid_targets_mean": 4539.9, + "valid_targets_min": 2883 + }, + { + "epoch": 5.868246968026461, + "grad_norm": 0.4127794494480315, + "learning_rate": 3.104491520564039e-06, + "loss": 0.0641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03809334710240364, + "step": 10645, + "valid_targets_mean": 3626.6, + "valid_targets_min": 2252 + }, + { + "epoch": 5.871003307607497, + "grad_norm": 0.468588606185368, + "learning_rate": 3.0897968432135105e-06, + "loss": 0.0728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03301672264933586, + "step": 10650, + "valid_targets_mean": 2025.5, + "valid_targets_min": 654 + }, + { + "epoch": 5.873759647188534, + "grad_norm": 0.45996577459086574, + "learning_rate": 3.0751341141845794e-06, + "loss": 0.0864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03782917559146881, + "step": 10655, + "valid_targets_mean": 3139.9, + "valid_targets_min": 744 + }, + { + "epoch": 5.87651598676957, + "grad_norm": 0.47626108012312407, + "learning_rate": 3.060503361179423e-06, + "loss": 0.0852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04639463499188423, + "step": 10660, + "valid_targets_mean": 3662.5, + "valid_targets_min": 3014 + }, + { + "epoch": 5.879272326350606, + "grad_norm": 0.31508333951272044, + "learning_rate": 3.045904611839827e-06, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026540221646428108, + "step": 10665, + "valid_targets_mean": 3219.4, + "valid_targets_min": 2366 + }, + { + "epoch": 5.882028665931642, + "grad_norm": 0.37380776232375756, + "learning_rate": 3.031337893747095e-06, + "loss": 0.0697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032113216817379, + "step": 10670, + "valid_targets_mean": 1961.0, + "valid_targets_min": 669 + }, + { + "epoch": 5.884785005512679, + "grad_norm": 0.519294973535354, + "learning_rate": 3.0168032344220165e-06, + "loss": 0.0854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04676566272974014, + "step": 10675, + "valid_targets_mean": 3525.8, + "valid_targets_min": 1785 + }, + { + "epoch": 5.887541345093716, + "grad_norm": 0.35542776956541156, + "learning_rate": 3.0023006613248217e-06, + "loss": 0.0608, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027141498401761055, + "step": 10680, + "valid_targets_mean": 4065.1, + "valid_targets_min": 3684 + }, + { + "epoch": 5.890297684674752, + "grad_norm": 0.39361203345197576, + "learning_rate": 2.9878302018551087e-06, + "loss": 0.0764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03101731650531292, + "step": 10685, + "valid_targets_mean": 3246.6, + "valid_targets_min": 595 + }, + { + "epoch": 5.893054024255788, + "grad_norm": 0.4656733580315256, + "learning_rate": 2.9733918833518127e-06, + "loss": 0.0749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03168364241719246, + "step": 10690, + "valid_targets_mean": 2653.8, + "valid_targets_min": 1016 + }, + { + "epoch": 5.895810363836825, + "grad_norm": 0.6810123389432357, + "learning_rate": 2.9589857330931404e-06, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07483795285224915, + "step": 10695, + "valid_targets_mean": 1386.5, + "valid_targets_min": 908 + }, + { + "epoch": 5.8985667034178615, + "grad_norm": 0.5476665759149487, + "learning_rate": 2.9446117782965266e-06, + "loss": 0.0805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04098716378211975, + "step": 10700, + "valid_targets_mean": 2152.5, + "valid_targets_min": 697 + }, + { + "epoch": 5.901323042998897, + "grad_norm": 0.7325926756580193, + "learning_rate": 2.9302700461185707e-06, + "loss": 0.0759, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05114099755883217, + "step": 10705, + "valid_targets_mean": 1274.8, + "valid_targets_min": 736 + }, + { + "epoch": 5.904079382579933, + "grad_norm": 0.3893932862306938, + "learning_rate": 2.915960563655014e-06, + "loss": 0.075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02533455565571785, + "step": 10710, + "valid_targets_mean": 2531.1, + "valid_targets_min": 804 + }, + { + "epoch": 5.90683572216097, + "grad_norm": 0.5157892256174564, + "learning_rate": 2.9016833579406455e-06, + "loss": 0.0948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03244553506374359, + "step": 10715, + "valid_targets_mean": 2974.6, + "valid_targets_min": 1018 + }, + { + "epoch": 5.909592061742007, + "grad_norm": 0.5291242211768539, + "learning_rate": 2.8874384559492897e-06, + "loss": 0.0756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036275848746299744, + "step": 10720, + "valid_targets_mean": 3626.9, + "valid_targets_min": 1632 + }, + { + "epoch": 5.912348401323043, + "grad_norm": 0.46007449567376857, + "learning_rate": 2.8732258845937313e-06, + "loss": 0.0647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036086555570364, + "step": 10725, + "valid_targets_mean": 3119.6, + "valid_targets_min": 1006 + }, + { + "epoch": 5.915104740904079, + "grad_norm": 0.49494227818997394, + "learning_rate": 2.8590456707256776e-06, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036067597568035126, + "step": 10730, + "valid_targets_mean": 2694.8, + "valid_targets_min": 903 + }, + { + "epoch": 5.917861080485116, + "grad_norm": 0.4361898898134223, + "learning_rate": 2.8448978411356987e-06, + "loss": 0.0809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03062455728650093, + "step": 10735, + "valid_targets_mean": 3786.0, + "valid_targets_min": 3052 + }, + { + "epoch": 5.9206174200661525, + "grad_norm": 0.5801711918249567, + "learning_rate": 2.830782422553189e-06, + "loss": 0.0628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0501018688082695, + "step": 10740, + "valid_targets_mean": 2972.0, + "valid_targets_min": 564 + }, + { + "epoch": 5.9233737596471885, + "grad_norm": 0.430281661837875, + "learning_rate": 2.8166994416463023e-06, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04174640774726868, + "step": 10745, + "valid_targets_mean": 3655.0, + "valid_targets_min": 2555 + }, + { + "epoch": 5.926130099228224, + "grad_norm": 0.33135624198130686, + "learning_rate": 2.802648925021907e-06, + "loss": 0.0644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029390770941972733, + "step": 10750, + "valid_targets_mean": 4186.8, + "valid_targets_min": 2991 + }, + { + "epoch": 5.928886438809261, + "grad_norm": 0.4190685779158301, + "learning_rate": 2.788630899225542e-06, + "loss": 0.0597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026183726266026497, + "step": 10755, + "valid_targets_mean": 2357.9, + "valid_targets_min": 618 + }, + { + "epoch": 5.931642778390298, + "grad_norm": 0.34821432580100314, + "learning_rate": 2.7746453907413616e-06, + "loss": 0.0821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028266333043575287, + "step": 10760, + "valid_targets_mean": 3658.1, + "valid_targets_min": 2944 + }, + { + "epoch": 5.934399117971334, + "grad_norm": 0.35412187793649497, + "learning_rate": 2.760692425992084e-06, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06042296811938286, + "step": 10765, + "valid_targets_mean": 3401.8, + "valid_targets_min": 1539 + }, + { + "epoch": 5.93715545755237, + "grad_norm": 0.37023890832379075, + "learning_rate": 2.7467720313389402e-06, + "loss": 0.0916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023135410621762276, + "step": 10770, + "valid_targets_mean": 2312.4, + "valid_targets_min": 509 + }, + { + "epoch": 5.939911797133407, + "grad_norm": 0.35734590719698195, + "learning_rate": 2.73288423308163e-06, + "loss": 0.0818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022928431630134583, + "step": 10775, + "valid_targets_mean": 2753.0, + "valid_targets_min": 637 + }, + { + "epoch": 5.942668136714444, + "grad_norm": 0.4498698850757854, + "learning_rate": 2.7190290574582647e-06, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03963606059551239, + "step": 10780, + "valid_targets_mean": 2755.8, + "valid_targets_min": 737 + }, + { + "epoch": 5.9454244762954795, + "grad_norm": 0.3742151508297915, + "learning_rate": 2.705206530645328e-06, + "loss": 0.0636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03264743834733963, + "step": 10785, + "valid_targets_mean": 3081.4, + "valid_targets_min": 492 + }, + { + "epoch": 5.948180815876516, + "grad_norm": 0.5552071167802782, + "learning_rate": 2.6914166787576167e-06, + "loss": 0.0799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03361296281218529, + "step": 10790, + "valid_targets_mean": 1433.2, + "valid_targets_min": 697 + }, + { + "epoch": 5.950937155457552, + "grad_norm": 0.38912999076346116, + "learning_rate": 2.6776595278481955e-06, + "loss": 0.0666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027643222361803055, + "step": 10795, + "valid_targets_mean": 3534.8, + "valid_targets_min": 923 + }, + { + "epoch": 5.953693495038589, + "grad_norm": 0.6422775856450526, + "learning_rate": 2.6639351039083437e-06, + "loss": 0.1004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03875361382961273, + "step": 10800, + "valid_targets_mean": 1486.1, + "valid_targets_min": 531 + }, + { + "epoch": 5.956449834619625, + "grad_norm": 0.5082077080751465, + "learning_rate": 2.650243432867521e-06, + "loss": 0.1016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06048717722296715, + "step": 10805, + "valid_targets_mean": 2096.5, + "valid_targets_min": 575 + }, + { + "epoch": 5.959206174200662, + "grad_norm": 0.5431253457177914, + "learning_rate": 2.636584540593299e-06, + "loss": 0.0727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03834383934736252, + "step": 10810, + "valid_targets_mean": 3725.2, + "valid_targets_min": 2784 + }, + { + "epoch": 5.961962513781698, + "grad_norm": 0.40857318303089063, + "learning_rate": 2.62295845289132e-06, + "loss": 0.071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03316022455692291, + "step": 10815, + "valid_targets_mean": 3427.1, + "valid_targets_min": 2525 + }, + { + "epoch": 5.964718853362735, + "grad_norm": 0.43782311193309503, + "learning_rate": 2.6093651955052492e-06, + "loss": 0.0702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02762661688029766, + "step": 10820, + "valid_targets_mean": 2502.5, + "valid_targets_min": 970 + }, + { + "epoch": 5.967475192943771, + "grad_norm": 0.467290764179899, + "learning_rate": 2.5958047941167296e-06, + "loss": 0.0646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029610853642225266, + "step": 10825, + "valid_targets_mean": 1527.6, + "valid_targets_min": 729 + }, + { + "epoch": 5.970231532524807, + "grad_norm": 0.9491450816047334, + "learning_rate": 2.582277274345324e-06, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05592062696814537, + "step": 10830, + "valid_targets_mean": 1267.9, + "valid_targets_min": 637 + }, + { + "epoch": 5.972987872105843, + "grad_norm": 0.3693740896375291, + "learning_rate": 2.568782661748479e-06, + "loss": 0.0629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030214782804250717, + "step": 10835, + "valid_targets_mean": 3062.4, + "valid_targets_min": 893 + }, + { + "epoch": 5.97574421168688, + "grad_norm": 0.3750817575896887, + "learning_rate": 2.5553209818214632e-06, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0300207007676363, + "step": 10840, + "valid_targets_mean": 3553.1, + "valid_targets_min": 2404 + }, + { + "epoch": 5.978500551267916, + "grad_norm": 0.4775427828474954, + "learning_rate": 2.541892259997332e-06, + "loss": 0.0667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0314633809030056, + "step": 10845, + "valid_targets_mean": 3326.0, + "valid_targets_min": 986 + }, + { + "epoch": 5.981256890848953, + "grad_norm": 0.38222858668811793, + "learning_rate": 2.5284965216468617e-06, + "loss": 0.0774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034078434109687805, + "step": 10850, + "valid_targets_mean": 3475.9, + "valid_targets_min": 1078 + }, + { + "epoch": 5.984013230429989, + "grad_norm": 0.38984319229779846, + "learning_rate": 2.5151337920785323e-06, + "loss": 0.0642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03615422174334526, + "step": 10855, + "valid_targets_mean": 4152.8, + "valid_targets_min": 761 + }, + { + "epoch": 5.986769570011026, + "grad_norm": 0.39373634052282147, + "learning_rate": 2.501804096538447e-06, + "loss": 0.0678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030123712494969368, + "step": 10860, + "valid_targets_mean": 1554.6, + "valid_targets_min": 507 + }, + { + "epoch": 5.989525909592062, + "grad_norm": 0.4862494712008259, + "learning_rate": 2.4885074602102987e-06, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07518617808818817, + "step": 10865, + "valid_targets_mean": 2711.8, + "valid_targets_min": 780 + }, + { + "epoch": 5.9922822491730985, + "grad_norm": 0.39358147993441794, + "learning_rate": 2.4752439082153235e-06, + "loss": 0.1423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03519464656710625, + "step": 10870, + "valid_targets_mean": 3892.5, + "valid_targets_min": 2347 + }, + { + "epoch": 5.995038588754134, + "grad_norm": 0.4556652792382218, + "learning_rate": 2.462013465612254e-06, + "loss": 0.0687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038769692182540894, + "step": 10875, + "valid_targets_mean": 3665.6, + "valid_targets_min": 2473 + }, + { + "epoch": 5.997794928335171, + "grad_norm": 0.3932445719890084, + "learning_rate": 2.4488161573972667e-06, + "loss": 0.0676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03287797421216965, + "step": 10880, + "valid_targets_mean": 3330.6, + "valid_targets_min": 2245 + }, + { + "epoch": 6.000551267916207, + "grad_norm": 0.830195377169508, + "learning_rate": 2.435652008503939e-06, + "loss": 0.1245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0959482491016388, + "step": 10885, + "valid_targets_mean": 8215.2, + "valid_targets_min": 6163 + }, + { + "epoch": 6.003307607497244, + "grad_norm": 0.8133146104544636, + "learning_rate": 2.422521043803199e-06, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09750667214393616, + "step": 10890, + "valid_targets_mean": 8594.4, + "valid_targets_min": 6006 + }, + { + "epoch": 6.00606394707828, + "grad_norm": 0.7262189391442316, + "learning_rate": 2.4094232881032787e-06, + "loss": 0.1725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07513216137886047, + "step": 10895, + "valid_targets_mean": 3578.4, + "valid_targets_min": 289 + }, + { + "epoch": 6.008820286659317, + "grad_norm": 0.5899106170991115, + "learning_rate": 2.3963587661496713e-06, + "loss": 0.1749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08812828361988068, + "step": 10900, + "valid_targets_mean": 7019.9, + "valid_targets_min": 4985 + }, + { + "epoch": 6.011576626240353, + "grad_norm": 0.5402804107885898, + "learning_rate": 2.383327502625088e-06, + "loss": 0.164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08568492531776428, + "step": 10905, + "valid_targets_mean": 6956.6, + "valid_targets_min": 1457 + }, + { + "epoch": 6.0143329658213895, + "grad_norm": 0.4409357892192358, + "learning_rate": 2.3703295221493926e-06, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07944756746292114, + "step": 10910, + "valid_targets_mean": 7361.6, + "valid_targets_min": 4961 + }, + { + "epoch": 6.0170893054024255, + "grad_norm": 0.5183577531625619, + "learning_rate": 2.3573648492795754e-06, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1063915491104126, + "step": 10915, + "valid_targets_mean": 7058.5, + "valid_targets_min": 5985 + }, + { + "epoch": 6.019845644983462, + "grad_norm": 0.4587100343786981, + "learning_rate": 2.3444335085096957e-06, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09453105926513672, + "step": 10920, + "valid_targets_mean": 7664.2, + "valid_targets_min": 5169 + }, + { + "epoch": 6.022601984564498, + "grad_norm": 0.4458086002395623, + "learning_rate": 2.33153552427084e-06, + "loss": 0.1653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08388198912143707, + "step": 10925, + "valid_targets_mean": 7111.8, + "valid_targets_min": 5348 + }, + { + "epoch": 6.025358324145535, + "grad_norm": 0.4758807635342571, + "learning_rate": 2.318670920931074e-06, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07096244394779205, + "step": 10930, + "valid_targets_mean": 4238.9, + "valid_targets_min": 152 + }, + { + "epoch": 6.028114663726571, + "grad_norm": 0.44108016407191514, + "learning_rate": 2.3058397227954e-06, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08425255119800568, + "step": 10935, + "valid_targets_mean": 7199.4, + "valid_targets_min": 4258 + }, + { + "epoch": 6.030871003307608, + "grad_norm": 0.41204185707095686, + "learning_rate": 2.2930419541057015e-06, + "loss": 0.1652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07964490354061127, + "step": 10940, + "valid_targets_mean": 7243.4, + "valid_targets_min": 5831 + }, + { + "epoch": 6.033627342888644, + "grad_norm": 0.4212251371380489, + "learning_rate": 2.2802776390407087e-06, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0816851258277893, + "step": 10945, + "valid_targets_mean": 6950.6, + "valid_targets_min": 3862 + }, + { + "epoch": 6.036383682469681, + "grad_norm": 0.4210129715198048, + "learning_rate": 2.2675468017159518e-06, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08783965557813644, + "step": 10950, + "valid_targets_mean": 6991.9, + "valid_targets_min": 5696 + }, + { + "epoch": 6.0391400220507165, + "grad_norm": 0.4117346962660137, + "learning_rate": 2.2548494661837063e-06, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08715663850307465, + "step": 10955, + "valid_targets_mean": 6963.5, + "valid_targets_min": 4947 + }, + { + "epoch": 6.041896361631753, + "grad_norm": 0.4106401356507897, + "learning_rate": 2.2421856564329515e-06, + "loss": 0.159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07666948437690735, + "step": 10960, + "valid_targets_mean": 6436.9, + "valid_targets_min": 4988 + }, + { + "epoch": 6.044652701212789, + "grad_norm": 0.4166690542501751, + "learning_rate": 2.2295553963893313e-06, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08009791374206543, + "step": 10965, + "valid_targets_mean": 7518.0, + "valid_targets_min": 5476 + }, + { + "epoch": 6.047409040793826, + "grad_norm": 0.39219544178488075, + "learning_rate": 2.2169587099151002e-06, + "loss": 0.1574, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07475851476192474, + "step": 10970, + "valid_targets_mean": 7422.6, + "valid_targets_min": 5226 + }, + { + "epoch": 6.050165380374862, + "grad_norm": 0.4434492996382606, + "learning_rate": 2.204395620809088e-06, + "loss": 0.1678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10266125202178955, + "step": 10975, + "valid_targets_mean": 7080.0, + "valid_targets_min": 3753 + }, + { + "epoch": 6.052921719955899, + "grad_norm": 0.525855317845387, + "learning_rate": 2.191866152806643e-06, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09228961169719696, + "step": 10980, + "valid_targets_mean": 6064.6, + "valid_targets_min": 5161 + }, + { + "epoch": 6.055678059536935, + "grad_norm": 0.4014723573421417, + "learning_rate": 2.1793703295795998e-06, + "loss": 0.1286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07551388442516327, + "step": 10985, + "valid_targets_mean": 7388.1, + "valid_targets_min": 5795 + }, + { + "epoch": 6.058434399117972, + "grad_norm": 0.3887769604288449, + "learning_rate": 2.1669081747362196e-06, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07398205250501633, + "step": 10990, + "valid_targets_mean": 7696.0, + "valid_targets_min": 4658 + }, + { + "epoch": 6.061190738699008, + "grad_norm": 0.4275598497392737, + "learning_rate": 2.1544797118211604e-06, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08565613627433777, + "step": 10995, + "valid_targets_mean": 7799.1, + "valid_targets_min": 5508 + }, + { + "epoch": 6.063947078280044, + "grad_norm": 0.4290255463156266, + "learning_rate": 2.1420849643154294e-06, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0804474726319313, + "step": 11000, + "valid_targets_mean": 7341.6, + "valid_targets_min": 6454 + }, + { + "epoch": 6.06670341786108, + "grad_norm": 0.4184145592373292, + "learning_rate": 2.129723955636329e-06, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07007165253162384, + "step": 11005, + "valid_targets_mean": 6818.0, + "valid_targets_min": 4392 + }, + { + "epoch": 6.069459757442117, + "grad_norm": 0.3967271965545618, + "learning_rate": 2.1173967091374227e-06, + "loss": 0.1595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06977850943803787, + "step": 11010, + "valid_targets_mean": 7857.6, + "valid_targets_min": 5394 + }, + { + "epoch": 6.072216097023153, + "grad_norm": 0.3672241948690627, + "learning_rate": 2.105103248108482e-06, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06653592735528946, + "step": 11015, + "valid_targets_mean": 8314.1, + "valid_targets_min": 4837 + }, + { + "epoch": 6.07497243660419, + "grad_norm": 0.45203137877013944, + "learning_rate": 2.0928435957754557e-06, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0598309263586998, + "step": 11020, + "valid_targets_mean": 7368.2, + "valid_targets_min": 6027 + }, + { + "epoch": 6.077728776185226, + "grad_norm": 0.4296056605405978, + "learning_rate": 2.080617775300413e-06, + "loss": 0.1477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07937491685152054, + "step": 11025, + "valid_targets_mean": 6435.0, + "valid_targets_min": 4238 + }, + { + "epoch": 6.080485115766263, + "grad_norm": 1.1647982363615113, + "learning_rate": 2.0684258097815047e-06, + "loss": 0.1445, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03918331116437912, + "step": 11030, + "valid_targets_mean": 247.8, + "valid_targets_min": 137 + }, + { + "epoch": 6.083241455347299, + "grad_norm": 0.43232933716386873, + "learning_rate": 2.0562677222529204e-06, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07743510603904724, + "step": 11035, + "valid_targets_mean": 6807.0, + "valid_targets_min": 5165 + }, + { + "epoch": 6.0859977949283355, + "grad_norm": 0.4922268940033915, + "learning_rate": 2.0441435356848436e-06, + "loss": 0.1713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08986328542232513, + "step": 11040, + "valid_targets_mean": 6989.9, + "valid_targets_min": 5104 + }, + { + "epoch": 6.088754134509371, + "grad_norm": 0.4332023914977946, + "learning_rate": 2.0320532729834187e-06, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09126178920269012, + "step": 11045, + "valid_targets_mean": 6692.2, + "valid_targets_min": 5068 + }, + { + "epoch": 6.091510474090408, + "grad_norm": 0.4455544005511616, + "learning_rate": 2.019996956990682e-06, + "loss": 0.1649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10053694248199463, + "step": 11050, + "valid_targets_mean": 7419.6, + "valid_targets_min": 5578 + }, + { + "epoch": 6.094266813671444, + "grad_norm": 0.4041684554099856, + "learning_rate": 2.0079746104845445e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08033193647861481, + "step": 11055, + "valid_targets_mean": 7536.5, + "valid_targets_min": 5075 + }, + { + "epoch": 6.097023153252481, + "grad_norm": 0.3825112080286794, + "learning_rate": 1.9959862561787347e-06, + "loss": 0.1667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07547086477279663, + "step": 11060, + "valid_targets_mean": 7221.5, + "valid_targets_min": 5841 + }, + { + "epoch": 6.099779492833517, + "grad_norm": 0.376584035986751, + "learning_rate": 1.984031916722762e-06, + "loss": 0.159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07793419808149338, + "step": 11065, + "valid_targets_mean": 7287.5, + "valid_targets_min": 5151 + }, + { + "epoch": 6.102535832414554, + "grad_norm": 0.43210016871733903, + "learning_rate": 1.972111614701873e-06, + "loss": 0.168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07853943109512329, + "step": 11070, + "valid_targets_mean": 6738.4, + "valid_targets_min": 4042 + }, + { + "epoch": 6.10529217199559, + "grad_norm": 0.423806944954139, + "learning_rate": 1.9602253726370056e-06, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0892292708158493, + "step": 11075, + "valid_targets_mean": 6331.0, + "valid_targets_min": 5496 + }, + { + "epoch": 6.1080485115766265, + "grad_norm": 0.4228518370065503, + "learning_rate": 1.9483732129847466e-06, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07852067053318024, + "step": 11080, + "valid_targets_mean": 6271.1, + "valid_targets_min": 4695 + }, + { + "epoch": 6.1108048511576625, + "grad_norm": 0.44326117512479235, + "learning_rate": 1.9365551581372964e-06, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08527281880378723, + "step": 11085, + "valid_targets_mean": 5790.8, + "valid_targets_min": 5021 + }, + { + "epoch": 6.113561190738699, + "grad_norm": 0.5132187312759089, + "learning_rate": 1.924771230422413e-06, + "loss": 0.1538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05910894647240639, + "step": 11090, + "valid_targets_mean": 3386.8, + "valid_targets_min": 2279 + }, + { + "epoch": 6.116317530319735, + "grad_norm": 0.40593024349806645, + "learning_rate": 1.913021452103392e-06, + "loss": 0.1327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07802097499370575, + "step": 11095, + "valid_targets_mean": 6240.5, + "valid_targets_min": 5093 + }, + { + "epoch": 6.119073869900772, + "grad_norm": 0.4328355594194063, + "learning_rate": 1.901305845378998e-06, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08134867250919342, + "step": 11100, + "valid_targets_mean": 6329.5, + "valid_targets_min": 4589 + }, + { + "epoch": 6.121830209481808, + "grad_norm": 0.5478010382702108, + "learning_rate": 1.8896244323834412e-06, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0863918587565422, + "step": 11105, + "valid_targets_mean": 7589.2, + "valid_targets_min": 4963 + }, + { + "epoch": 6.124586549062845, + "grad_norm": 0.41837864319592266, + "learning_rate": 1.877977235186328e-06, + "loss": 0.1602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06824932247400284, + "step": 11110, + "valid_targets_mean": 5577.9, + "valid_targets_min": 4458 + }, + { + "epoch": 6.127342888643881, + "grad_norm": 0.4354799547881119, + "learning_rate": 1.8663642757926227e-06, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07742257416248322, + "step": 11115, + "valid_targets_mean": 5432.5, + "valid_targets_min": 4433 + }, + { + "epoch": 6.130099228224918, + "grad_norm": 0.4260076374309989, + "learning_rate": 1.8547855761426014e-06, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08262777328491211, + "step": 11120, + "valid_targets_mean": 6361.8, + "valid_targets_min": 5201 + }, + { + "epoch": 6.1328555678059535, + "grad_norm": 0.4817785098496001, + "learning_rate": 1.843241158111817e-06, + "loss": 0.1631, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07751452177762985, + "step": 11125, + "valid_targets_mean": 5824.4, + "valid_targets_min": 5129 + }, + { + "epoch": 6.13561190738699, + "grad_norm": 0.7422683280526852, + "learning_rate": 1.8317310435110535e-06, + "loss": 0.125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07361220568418503, + "step": 11130, + "valid_targets_mean": 1637.6, + "valid_targets_min": 137 + }, + { + "epoch": 6.138368246968026, + "grad_norm": 0.4527709092654473, + "learning_rate": 1.820255254086285e-06, + "loss": 0.1579, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08575209975242615, + "step": 11135, + "valid_targets_mean": 6474.6, + "valid_targets_min": 4751 + }, + { + "epoch": 6.141124586549063, + "grad_norm": 0.4498974351955801, + "learning_rate": 1.8088138115186304e-06, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08654439449310303, + "step": 11140, + "valid_targets_mean": 7186.2, + "valid_targets_min": 5358 + }, + { + "epoch": 6.143880926130099, + "grad_norm": 0.4195681701520791, + "learning_rate": 1.7974067374243298e-06, + "loss": 0.1613, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08009402453899384, + "step": 11145, + "valid_targets_mean": 7126.8, + "valid_targets_min": 4078 + }, + { + "epoch": 6.146637265711136, + "grad_norm": 0.4101540268518501, + "learning_rate": 1.786034053354684e-06, + "loss": 0.1535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07620936632156372, + "step": 11150, + "valid_targets_mean": 7022.1, + "valid_targets_min": 4940 + }, + { + "epoch": 6.149393605292172, + "grad_norm": 0.4310229834129872, + "learning_rate": 1.7746957807960141e-06, + "loss": 0.1511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08204454183578491, + "step": 11155, + "valid_targets_mean": 5900.6, + "valid_targets_min": 4779 + }, + { + "epoch": 6.152149944873209, + "grad_norm": 0.4413530873645671, + "learning_rate": 1.763391941169641e-06, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08073413372039795, + "step": 11160, + "valid_targets_mean": 6290.1, + "valid_targets_min": 4620 + }, + { + "epoch": 6.154906284454245, + "grad_norm": 0.8647522980533363, + "learning_rate": 1.7521225558318212e-06, + "loss": 0.1561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060017623007297516, + "step": 11165, + "valid_targets_mean": 1162.2, + "valid_targets_min": 788 + }, + { + "epoch": 6.157662624035281, + "grad_norm": 0.8428718269588573, + "learning_rate": 1.7408876460737234e-06, + "loss": 0.1462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07018640637397766, + "step": 11170, + "valid_targets_mean": 1500.6, + "valid_targets_min": 957 + }, + { + "epoch": 6.160418963616317, + "grad_norm": 0.9843772465985785, + "learning_rate": 1.7296872331213776e-06, + "loss": 0.132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06101137027144432, + "step": 11175, + "valid_targets_mean": 1371.2, + "valid_targets_min": 796 + }, + { + "epoch": 6.163175303197354, + "grad_norm": 0.7290490581833566, + "learning_rate": 1.7185213381356414e-06, + "loss": 0.1306, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0667063444852829, + "step": 11180, + "valid_targets_mean": 1433.6, + "valid_targets_min": 697 + }, + { + "epoch": 6.16593164277839, + "grad_norm": 0.7333017064197753, + "learning_rate": 1.7073899822121532e-06, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06299866735935211, + "step": 11185, + "valid_targets_mean": 1578.9, + "valid_targets_min": 638 + }, + { + "epoch": 6.168687982359427, + "grad_norm": 0.692243485224548, + "learning_rate": 1.6962931863813103e-06, + "loss": 0.1288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06333142518997192, + "step": 11190, + "valid_targets_mean": 1746.4, + "valid_targets_min": 727 + }, + { + "epoch": 6.171444321940463, + "grad_norm": 0.7564065220714383, + "learning_rate": 1.6852309716082027e-06, + "loss": 0.1352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05198981985449791, + "step": 11195, + "valid_targets_mean": 1309.9, + "valid_targets_min": 761 + }, + { + "epoch": 6.1742006615215, + "grad_norm": 0.8015578168538194, + "learning_rate": 1.6742033587925921e-06, + "loss": 0.1172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06624957919120789, + "step": 11200, + "valid_targets_mean": 1368.6, + "valid_targets_min": 714 + }, + { + "epoch": 6.176957001102536, + "grad_norm": 0.8072139924247778, + "learning_rate": 1.6632103687688617e-06, + "loss": 0.1374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06270375102758408, + "step": 11205, + "valid_targets_mean": 1491.8, + "valid_targets_min": 867 + }, + { + "epoch": 6.1797133406835725, + "grad_norm": 0.9756376503572808, + "learning_rate": 1.652252022305989e-06, + "loss": 0.1303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07472027838230133, + "step": 11210, + "valid_targets_mean": 1994.0, + "valid_targets_min": 790 + }, + { + "epoch": 6.182469680264608, + "grad_norm": 0.8182950138151031, + "learning_rate": 1.6413283401074975e-06, + "loss": 0.1208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06229248270392418, + "step": 11215, + "valid_targets_mean": 1523.4, + "valid_targets_min": 896 + }, + { + "epoch": 6.185226019845645, + "grad_norm": 0.7689633075454018, + "learning_rate": 1.6304393428114185e-06, + "loss": 0.1269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0641670972108841, + "step": 11220, + "valid_targets_mean": 1790.1, + "valid_targets_min": 1009 + }, + { + "epoch": 6.187982359426681, + "grad_norm": 0.8063159109133723, + "learning_rate": 1.6195850509902534e-06, + "loss": 0.1223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05097575485706329, + "step": 11225, + "valid_targets_mean": 1394.0, + "valid_targets_min": 717 + }, + { + "epoch": 6.190738699007718, + "grad_norm": 0.8109341656891335, + "learning_rate": 1.6087654851509339e-06, + "loss": 0.115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049293868243694305, + "step": 11230, + "valid_targets_mean": 1129.0, + "valid_targets_min": 659 + }, + { + "epoch": 6.193495038588754, + "grad_norm": 0.7435360133120247, + "learning_rate": 1.597980665734784e-06, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055080730468034744, + "step": 11235, + "valid_targets_mean": 1466.9, + "valid_targets_min": 642 + }, + { + "epoch": 6.196251378169791, + "grad_norm": 0.806227602546308, + "learning_rate": 1.5872306131174896e-06, + "loss": 0.1212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044107306748628616, + "step": 11240, + "valid_targets_mean": 1129.4, + "valid_targets_min": 765 + }, + { + "epoch": 6.199007717750827, + "grad_norm": 0.8600515992336581, + "learning_rate": 1.5765153476090444e-06, + "loss": 0.1295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05448248237371445, + "step": 11245, + "valid_targets_mean": 1322.8, + "valid_targets_min": 631 + }, + { + "epoch": 6.2017640573318635, + "grad_norm": 0.905876939662869, + "learning_rate": 1.5658348894537167e-06, + "loss": 0.1289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07411297410726547, + "step": 11250, + "valid_targets_mean": 1612.0, + "valid_targets_min": 623 + }, + { + "epoch": 6.2045203969128995, + "grad_norm": 0.7705606174979731, + "learning_rate": 1.5551892588300188e-06, + "loss": 0.1199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0481376014649868, + "step": 11255, + "valid_targets_mean": 1402.1, + "valid_targets_min": 940 + }, + { + "epoch": 6.207276736493936, + "grad_norm": 0.7783931904091883, + "learning_rate": 1.5445784758506599e-06, + "loss": 0.1254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05487378314137459, + "step": 11260, + "valid_targets_mean": 1337.8, + "valid_targets_min": 776 + }, + { + "epoch": 6.210033076074972, + "grad_norm": 0.8396327760072656, + "learning_rate": 1.5340025605625152e-06, + "loss": 0.1255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06241946294903755, + "step": 11265, + "valid_targets_mean": 1525.5, + "valid_targets_min": 623 + }, + { + "epoch": 6.212789415656009, + "grad_norm": 0.8126166104009247, + "learning_rate": 1.523461532946584e-06, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05872516334056854, + "step": 11270, + "valid_targets_mean": 1441.0, + "valid_targets_min": 693 + }, + { + "epoch": 6.215545755237045, + "grad_norm": 0.8099972165873763, + "learning_rate": 1.5129554129179514e-06, + "loss": 0.1206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033097706735134125, + "step": 11275, + "valid_targets_mean": 847.9, + "valid_targets_min": 611 + }, + { + "epoch": 6.218302094818082, + "grad_norm": 0.8705002155685101, + "learning_rate": 1.5024842203257506e-06, + "loss": 0.1142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059609122574329376, + "step": 11280, + "valid_targets_mean": 1524.9, + "valid_targets_min": 793 + }, + { + "epoch": 6.221058434399118, + "grad_norm": 0.796321986370449, + "learning_rate": 1.4920479749531302e-06, + "loss": 0.117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06296596676111221, + "step": 11285, + "valid_targets_mean": 1779.4, + "valid_targets_min": 563 + }, + { + "epoch": 6.223814773980155, + "grad_norm": 0.7419043121482826, + "learning_rate": 1.4816466965172094e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0608222521841526, + "step": 11290, + "valid_targets_mean": 1741.2, + "valid_targets_min": 1341 + }, + { + "epoch": 6.2265711135611905, + "grad_norm": 0.8343950181518147, + "learning_rate": 1.4712804046690466e-06, + "loss": 0.1226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07092001289129257, + "step": 11295, + "valid_targets_mean": 1875.6, + "valid_targets_min": 1107 + }, + { + "epoch": 6.229327453142227, + "grad_norm": 0.8234078709861874, + "learning_rate": 1.460949118993602e-06, + "loss": 0.1282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05934404209256172, + "step": 11300, + "valid_targets_mean": 1419.5, + "valid_targets_min": 680 + }, + { + "epoch": 6.232083792723263, + "grad_norm": 0.8006137237492997, + "learning_rate": 1.4506528590096957e-06, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07914485037326813, + "step": 11305, + "valid_targets_mean": 1968.2, + "valid_targets_min": 981 + }, + { + "epoch": 6.2348401323043, + "grad_norm": 0.808595061838774, + "learning_rate": 1.4403916441699762e-06, + "loss": 0.1213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061145734041929245, + "step": 11310, + "valid_targets_mean": 1568.9, + "valid_targets_min": 937 + }, + { + "epoch": 6.237596471885336, + "grad_norm": 0.8197257472391454, + "learning_rate": 1.43016549386088e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06587523967027664, + "step": 11315, + "valid_targets_mean": 1645.2, + "valid_targets_min": 804 + }, + { + "epoch": 6.240352811466373, + "grad_norm": 0.8778401607429439, + "learning_rate": 1.419974427402595e-06, + "loss": 0.1223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06121107190847397, + "step": 11320, + "valid_targets_mean": 1574.9, + "valid_targets_min": 1019 + }, + { + "epoch": 6.243109151047409, + "grad_norm": 0.8339039107316731, + "learning_rate": 1.4098184640490331e-06, + "loss": 0.1165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04983334243297577, + "step": 11325, + "valid_targets_mean": 1121.6, + "valid_targets_min": 882 + }, + { + "epoch": 6.245865490628446, + "grad_norm": 0.7365616682191019, + "learning_rate": 1.3996976229877768e-06, + "loss": 0.1138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062467753887176514, + "step": 11330, + "valid_targets_mean": 1818.1, + "valid_targets_min": 1168 + }, + { + "epoch": 6.248621830209482, + "grad_norm": 1.00283856730802, + "learning_rate": 1.3896119233400595e-06, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05539535731077194, + "step": 11335, + "valid_targets_mean": 1417.8, + "valid_targets_min": 678 + }, + { + "epoch": 6.251378169790518, + "grad_norm": 0.864215690963083, + "learning_rate": 1.3795613841607148e-06, + "loss": 0.1202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04719821736216545, + "step": 11340, + "valid_targets_mean": 1051.0, + "valid_targets_min": 676 + }, + { + "epoch": 6.254134509371554, + "grad_norm": 0.7991873585034648, + "learning_rate": 1.369546024438153e-06, + "loss": 0.1174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05476917326450348, + "step": 11345, + "valid_targets_mean": 1431.1, + "valid_targets_min": 794 + }, + { + "epoch": 6.256890848952591, + "grad_norm": 0.7927606118515688, + "learning_rate": 1.3595658630943186e-06, + "loss": 0.117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058689527213573456, + "step": 11350, + "valid_targets_mean": 1486.9, + "valid_targets_min": 742 + }, + { + "epoch": 6.259647188533627, + "grad_norm": 0.815032932802139, + "learning_rate": 1.34962091898466e-06, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07233321666717529, + "step": 11355, + "valid_targets_mean": 1734.5, + "valid_targets_min": 741 + }, + { + "epoch": 6.262403528114664, + "grad_norm": 0.869467871297026, + "learning_rate": 1.3397112108980826e-06, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07222054153680801, + "step": 11360, + "valid_targets_mean": 1712.5, + "valid_targets_min": 1157 + }, + { + "epoch": 6.2651598676957, + "grad_norm": 0.8154092380441742, + "learning_rate": 1.329836757556926e-06, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07199612259864807, + "step": 11365, + "valid_targets_mean": 1723.8, + "valid_targets_min": 862 + }, + { + "epoch": 6.267916207276737, + "grad_norm": 0.8450208580130824, + "learning_rate": 1.3199975776169183e-06, + "loss": 0.1221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06796437501907349, + "step": 11370, + "valid_targets_mean": 1633.1, + "valid_targets_min": 1019 + }, + { + "epoch": 6.270672546857773, + "grad_norm": 0.8125471897172998, + "learning_rate": 1.3101936896671562e-06, + "loss": 0.1201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050617776811122894, + "step": 11375, + "valid_targets_mean": 1376.1, + "valid_targets_min": 1058 + }, + { + "epoch": 6.2734288864388095, + "grad_norm": 0.7980783358360874, + "learning_rate": 1.3004251122300481e-06, + "loss": 0.1266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06025683134794235, + "step": 11380, + "valid_targets_mean": 1576.5, + "valid_targets_min": 618 + }, + { + "epoch": 6.276185226019845, + "grad_norm": 0.7958116711660125, + "learning_rate": 1.2906918637612975e-06, + "loss": 0.1106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06442537903785706, + "step": 11385, + "valid_targets_mean": 1615.8, + "valid_targets_min": 629 + }, + { + "epoch": 6.278941565600882, + "grad_norm": 0.9065968640834524, + "learning_rate": 1.2809939626498547e-06, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06441132724285126, + "step": 11390, + "valid_targets_mean": 1881.2, + "valid_targets_min": 1167 + }, + { + "epoch": 6.281697905181918, + "grad_norm": 0.8571108393207725, + "learning_rate": 1.2713314272178967e-06, + "loss": 0.1156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059110451489686966, + "step": 11395, + "valid_targets_mean": 1367.2, + "valid_targets_min": 699 + }, + { + "epoch": 6.284454244762955, + "grad_norm": 0.8145578262457542, + "learning_rate": 1.2617042757207787e-06, + "loss": 0.1117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04505478963255882, + "step": 11400, + "valid_targets_mean": 1174.9, + "valid_targets_min": 679 + }, + { + "epoch": 6.287210584343991, + "grad_norm": 0.8161818387207244, + "learning_rate": 1.2521125263470046e-06, + "loss": 0.1109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05662080645561218, + "step": 11405, + "valid_targets_mean": 1556.2, + "valid_targets_min": 504 + }, + { + "epoch": 6.289966923925028, + "grad_norm": 0.7706216392973066, + "learning_rate": 1.2425561972181987e-06, + "loss": 0.1179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06375948339700699, + "step": 11410, + "valid_targets_mean": 1885.0, + "valid_targets_min": 1331 + }, + { + "epoch": 6.292723263506064, + "grad_norm": 0.7737313560877468, + "learning_rate": 1.2330353063890588e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04245840013027191, + "step": 11415, + "valid_targets_mean": 1306.1, + "valid_targets_min": 661 + }, + { + "epoch": 6.2954796030871005, + "grad_norm": 0.8499825513795952, + "learning_rate": 1.223549871847336e-06, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045838773250579834, + "step": 11420, + "valid_targets_mean": 1330.1, + "valid_targets_min": 958 + }, + { + "epoch": 6.2982359426681365, + "grad_norm": 0.7490464900071796, + "learning_rate": 1.2140999115137953e-06, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04533727467060089, + "step": 11425, + "valid_targets_mean": 1465.0, + "valid_targets_min": 963 + }, + { + "epoch": 6.300992282249173, + "grad_norm": 0.7857984182275605, + "learning_rate": 1.2046854432421706e-06, + "loss": 0.1164, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052597470581531525, + "step": 11430, + "valid_targets_mean": 1428.0, + "valid_targets_min": 793 + }, + { + "epoch": 6.303748621830209, + "grad_norm": 0.8157166262834775, + "learning_rate": 1.1953064848191497e-06, + "loss": 0.1107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07008466124534607, + "step": 11435, + "valid_targets_mean": 1633.1, + "valid_targets_min": 917 + }, + { + "epoch": 6.306504961411246, + "grad_norm": 0.7746147964512519, + "learning_rate": 1.1859630539643297e-06, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04509121552109718, + "step": 11440, + "valid_targets_mean": 1246.5, + "valid_targets_min": 872 + }, + { + "epoch": 6.309261300992282, + "grad_norm": 0.8114033657925088, + "learning_rate": 1.176655168330183e-06, + "loss": 0.1139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055706821382045746, + "step": 11445, + "valid_targets_mean": 1903.6, + "valid_targets_min": 1220 + }, + { + "epoch": 6.312017640573319, + "grad_norm": 0.8508110766480135, + "learning_rate": 1.167382845502032e-06, + "loss": 0.1038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04836345463991165, + "step": 11450, + "valid_targets_mean": 1391.6, + "valid_targets_min": 712 + }, + { + "epoch": 6.314773980154355, + "grad_norm": 0.8614379308570799, + "learning_rate": 1.1581461029980078e-06, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05467426776885986, + "step": 11455, + "valid_targets_mean": 1295.0, + "valid_targets_min": 605 + }, + { + "epoch": 6.317530319735392, + "grad_norm": 0.7951520143456746, + "learning_rate": 1.148944958269016e-06, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04246007651090622, + "step": 11460, + "valid_targets_mean": 1414.8, + "valid_targets_min": 746 + }, + { + "epoch": 6.3202866593164275, + "grad_norm": 0.880499829430079, + "learning_rate": 1.1397794286987107e-06, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04724094271659851, + "step": 11465, + "valid_targets_mean": 1059.4, + "valid_targets_min": 723 + }, + { + "epoch": 6.323042998897464, + "grad_norm": 0.8042663795376459, + "learning_rate": 1.1306495316034648e-06, + "loss": 0.1138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06463780999183655, + "step": 11470, + "valid_targets_mean": 1634.2, + "valid_targets_min": 1327 + }, + { + "epoch": 6.3257993384785, + "grad_norm": 0.8036592945550639, + "learning_rate": 1.1215552842323208e-06, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05047212168574333, + "step": 11475, + "valid_targets_mean": 1412.5, + "valid_targets_min": 902 + }, + { + "epoch": 6.328555678059537, + "grad_norm": 0.8848343415055617, + "learning_rate": 1.1124967037669764e-06, + "loss": 0.1183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060192905366420746, + "step": 11480, + "valid_targets_mean": 1558.8, + "valid_targets_min": 803 + }, + { + "epoch": 6.331312017640573, + "grad_norm": 0.8527304800084307, + "learning_rate": 1.1034738073217356e-06, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060911376029253006, + "step": 11485, + "valid_targets_mean": 1441.2, + "valid_targets_min": 1079 + }, + { + "epoch": 6.33406835722161, + "grad_norm": 0.8376984088524029, + "learning_rate": 1.0944866119434948e-06, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050710279494524, + "step": 11490, + "valid_targets_mean": 1449.8, + "valid_targets_min": 1263 + }, + { + "epoch": 6.336824696802646, + "grad_norm": 0.7918658468398081, + "learning_rate": 1.0855351346116927e-06, + "loss": 0.1149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05404765158891678, + "step": 11495, + "valid_targets_mean": 1487.4, + "valid_targets_min": 844 + }, + { + "epoch": 6.339581036383683, + "grad_norm": 0.8902244112548787, + "learning_rate": 1.0766193922382895e-06, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05585566163063049, + "step": 11500, + "valid_targets_mean": 1421.4, + "valid_targets_min": 667 + }, + { + "epoch": 6.342337375964719, + "grad_norm": 0.8746592691087922, + "learning_rate": 1.0677394016677333e-06, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06913750618696213, + "step": 11505, + "valid_targets_mean": 1822.6, + "valid_targets_min": 1041 + }, + { + "epoch": 6.345093715545755, + "grad_norm": 0.8679530687007504, + "learning_rate": 1.058895179676922e-06, + "loss": 0.1181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05002432316541672, + "step": 11510, + "valid_targets_mean": 1327.4, + "valid_targets_min": 777 + }, + { + "epoch": 6.347850055126791, + "grad_norm": 0.8517977479285103, + "learning_rate": 1.050086742975176e-06, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05987849086523056, + "step": 11515, + "valid_targets_mean": 1823.8, + "valid_targets_min": 1283 + }, + { + "epoch": 6.350606394707828, + "grad_norm": 0.8679782093825655, + "learning_rate": 1.0413141082042167e-06, + "loss": 0.1096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04239334166049957, + "step": 11520, + "valid_targets_mean": 924.0, + "valid_targets_min": 620 + }, + { + "epoch": 6.353362734288864, + "grad_norm": 0.8421116531992919, + "learning_rate": 1.0325772919381127e-06, + "loss": 0.1096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04813992977142334, + "step": 11525, + "valid_targets_mean": 1480.2, + "valid_targets_min": 656 + }, + { + "epoch": 6.356119073869901, + "grad_norm": 0.7867504560684611, + "learning_rate": 1.0238763106832694e-06, + "loss": 0.1109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04450366646051407, + "step": 11530, + "valid_targets_mean": 1415.5, + "valid_targets_min": 697 + }, + { + "epoch": 6.358875413450937, + "grad_norm": 0.8257402352934446, + "learning_rate": 1.015211180878386e-06, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05235455930233002, + "step": 11535, + "valid_targets_mean": 1655.8, + "valid_targets_min": 816 + }, + { + "epoch": 6.361631753031974, + "grad_norm": 0.7904646171039388, + "learning_rate": 1.0065819188944247e-06, + "loss": 0.1068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050029825419187546, + "step": 11540, + "valid_targets_mean": 1553.0, + "valid_targets_min": 850 + }, + { + "epoch": 6.36438809261301, + "grad_norm": 0.7785063829536303, + "learning_rate": 9.97988541034589e-07, + "loss": 0.1075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05332258716225624, + "step": 11545, + "valid_targets_mean": 1353.9, + "valid_targets_min": 747 + }, + { + "epoch": 6.3671444321940465, + "grad_norm": 0.804306327515265, + "learning_rate": 9.894310635342852e-07, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05002604424953461, + "step": 11550, + "valid_targets_mean": 1420.9, + "valid_targets_min": 933 + }, + { + "epoch": 6.369900771775082, + "grad_norm": 0.7999483552600012, + "learning_rate": 9.809095025610894e-07, + "loss": 0.1095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054485779255628586, + "step": 11555, + "valid_targets_mean": 1394.9, + "valid_targets_min": 827 + }, + { + "epoch": 6.372657111356119, + "grad_norm": 0.7447782305743578, + "learning_rate": 9.724238742147207e-07, + "loss": 0.1069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045727457851171494, + "step": 11560, + "valid_targets_mean": 1383.2, + "valid_targets_min": 909 + }, + { + "epoch": 6.375413450937155, + "grad_norm": 0.7465949142427039, + "learning_rate": 9.63974194527022e-07, + "loss": 0.1094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056954339146614075, + "step": 11565, + "valid_targets_mean": 1682.4, + "valid_targets_min": 1094 + }, + { + "epoch": 6.378169790518192, + "grad_norm": 0.8323547363920327, + "learning_rate": 9.555604794619056e-07, + "loss": 0.0997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04833999276161194, + "step": 11570, + "valid_targets_mean": 1118.4, + "valid_targets_min": 720 + }, + { + "epoch": 6.380926130099228, + "grad_norm": 0.7864842987645517, + "learning_rate": 9.471827449153425e-07, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05248040705919266, + "step": 11575, + "valid_targets_mean": 1538.2, + "valid_targets_min": 610 + }, + { + "epoch": 6.383682469680265, + "grad_norm": 0.8114004826810696, + "learning_rate": 9.38841006715323e-07, + "loss": 0.1072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05489123985171318, + "step": 11580, + "valid_targets_mean": 1570.1, + "valid_targets_min": 713 + }, + { + "epoch": 6.386438809261301, + "grad_norm": 0.9408922185453135, + "learning_rate": 9.305352806218338e-07, + "loss": 0.1133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052733976393938065, + "step": 11585, + "valid_targets_mean": 1537.8, + "valid_targets_min": 874 + }, + { + "epoch": 6.3891951488423375, + "grad_norm": 0.8195507785162902, + "learning_rate": 9.222655823268201e-07, + "loss": 0.1113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05540091171860695, + "step": 11590, + "valid_targets_mean": 1471.2, + "valid_targets_min": 765 + }, + { + "epoch": 6.3919514884233735, + "grad_norm": 0.8525489645311881, + "learning_rate": 9.140319274541621e-07, + "loss": 0.1071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04638671875, + "step": 11595, + "valid_targets_mean": 1254.4, + "valid_targets_min": 816 + }, + { + "epoch": 6.39470782800441, + "grad_norm": 0.8562929947712857, + "learning_rate": 9.058343315596407e-07, + "loss": 0.1113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05270727723836899, + "step": 11600, + "valid_targets_mean": 1482.9, + "valid_targets_min": 1182 + }, + { + "epoch": 6.397464167585446, + "grad_norm": 0.9305168606878684, + "learning_rate": 8.976728101309162e-07, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1052725613117218, + "step": 11605, + "valid_targets_mean": 4566.8, + "valid_targets_min": 702 + }, + { + "epoch": 6.400220507166483, + "grad_norm": 0.9050701634359898, + "learning_rate": 8.895473785874875e-07, + "loss": 0.218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12730266153812408, + "step": 11610, + "valid_targets_mean": 5524.9, + "valid_targets_min": 1005 + }, + { + "epoch": 6.402976846747519, + "grad_norm": 0.7549160566539754, + "learning_rate": 8.814580522806748e-07, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08531876653432846, + "step": 11615, + "valid_targets_mean": 4319.8, + "valid_targets_min": 2054 + }, + { + "epoch": 6.405733186328556, + "grad_norm": 0.79933558771932, + "learning_rate": 8.734048464935841e-07, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08440594375133514, + "step": 11620, + "valid_targets_mean": 3498.9, + "valid_targets_min": 1561 + }, + { + "epoch": 6.408489525909592, + "grad_norm": 0.8811177435615725, + "learning_rate": 8.653877764410734e-07, + "loss": 0.2116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13121287524700165, + "step": 11625, + "valid_targets_mean": 5020.6, + "valid_targets_min": 1970 + }, + { + "epoch": 6.411245865490629, + "grad_norm": 0.7792290840369537, + "learning_rate": 8.574068572697359e-07, + "loss": 0.2148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10423257946968079, + "step": 11630, + "valid_targets_mean": 6153.9, + "valid_targets_min": 1651 + }, + { + "epoch": 6.4140022050716645, + "grad_norm": 0.712141084157552, + "learning_rate": 8.494621040578632e-07, + "loss": 0.2038, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11638592183589935, + "step": 11635, + "valid_targets_mean": 7671.2, + "valid_targets_min": 2231 + }, + { + "epoch": 6.416758544652701, + "grad_norm": 0.641451631886228, + "learning_rate": 8.415535318154178e-07, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06950785219669342, + "step": 11640, + "valid_targets_mean": 3124.8, + "valid_targets_min": 467 + }, + { + "epoch": 6.419514884233737, + "grad_norm": 0.6517980607409072, + "learning_rate": 8.336811554840074e-07, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12310626357793808, + "step": 11645, + "valid_targets_mean": 6102.0, + "valid_targets_min": 1748 + }, + { + "epoch": 6.422271223814774, + "grad_norm": 0.6370036458774444, + "learning_rate": 8.258449899368526e-07, + "loss": 0.2118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09951814264059067, + "step": 11650, + "valid_targets_mean": 4508.4, + "valid_targets_min": 1697 + }, + { + "epoch": 6.42502756339581, + "grad_norm": 0.6476157349689953, + "learning_rate": 8.180450499787662e-07, + "loss": 0.2074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10266970098018646, + "step": 11655, + "valid_targets_mean": 3941.2, + "valid_targets_min": 1223 + }, + { + "epoch": 6.427783902976847, + "grad_norm": 0.5885880843710047, + "learning_rate": 8.102813503461115e-07, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11135601252317429, + "step": 11660, + "valid_targets_mean": 5561.2, + "valid_targets_min": 1552 + }, + { + "epoch": 6.430540242557883, + "grad_norm": 0.6128895688649686, + "learning_rate": 8.025539057067954e-07, + "loss": 0.2098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11152774840593338, + "step": 11665, + "valid_targets_mean": 4325.5, + "valid_targets_min": 1546 + }, + { + "epoch": 6.43329658213892, + "grad_norm": 0.6363585350291446, + "learning_rate": 7.948627306602174e-07, + "loss": 0.2041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10572242736816406, + "step": 11670, + "valid_targets_mean": 4611.9, + "valid_targets_min": 1957 + }, + { + "epoch": 6.436052921719956, + "grad_norm": 0.658974193047082, + "learning_rate": 7.872078397372607e-07, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08210612833499908, + "step": 11675, + "valid_targets_mean": 3394.4, + "valid_targets_min": 1538 + }, + { + "epoch": 6.438809261300992, + "grad_norm": 0.6822219984672709, + "learning_rate": 7.795892474002497e-07, + "loss": 0.2047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10255004465579987, + "step": 11680, + "valid_targets_mean": 3414.6, + "valid_targets_min": 1645 + }, + { + "epoch": 6.441565600882028, + "grad_norm": 0.6613252074594733, + "learning_rate": 7.720069680429399e-07, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09575942158699036, + "step": 11685, + "valid_targets_mean": 3189.5, + "valid_targets_min": 935 + }, + { + "epoch": 6.444321940463065, + "grad_norm": 0.6825531289690463, + "learning_rate": 7.644610159904719e-07, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11005303263664246, + "step": 11690, + "valid_targets_mean": 3584.8, + "valid_targets_min": 1269 + }, + { + "epoch": 6.447078280044101, + "grad_norm": 0.6911161920587015, + "learning_rate": 7.569514054993576e-07, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08074985444545746, + "step": 11695, + "valid_targets_mean": 2098.1, + "valid_targets_min": 1173 + }, + { + "epoch": 6.449834619625138, + "grad_norm": 0.5992233213010436, + "learning_rate": 7.494781507574478e-07, + "loss": 0.2039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09788631647825241, + "step": 11700, + "valid_targets_mean": 4135.4, + "valid_targets_min": 1670 + }, + { + "epoch": 6.452590959206174, + "grad_norm": 0.6476826767935878, + "learning_rate": 7.42041265883906e-07, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09313251078128815, + "step": 11705, + "valid_targets_mean": 3372.5, + "valid_targets_min": 1329 + }, + { + "epoch": 6.455347298787211, + "grad_norm": 0.6690212210535087, + "learning_rate": 7.346407649291887e-07, + "loss": 0.1885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11533982306718826, + "step": 11710, + "valid_targets_mean": 3950.1, + "valid_targets_min": 2106 + }, + { + "epoch": 6.458103638368247, + "grad_norm": 0.5428345072692039, + "learning_rate": 7.272766618750049e-07, + "loss": 0.1799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07974278926849365, + "step": 11715, + "valid_targets_mean": 4317.1, + "valid_targets_min": 2222 + }, + { + "epoch": 6.4608599779492835, + "grad_norm": 0.6437426939235511, + "learning_rate": 7.19948970634301e-07, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09425188601016998, + "step": 11720, + "valid_targets_mean": 3463.2, + "valid_targets_min": 1565 + }, + { + "epoch": 6.463616317530319, + "grad_norm": 0.6303728388122336, + "learning_rate": 7.126577050512274e-07, + "loss": 0.1994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09752606600522995, + "step": 11725, + "valid_targets_mean": 3481.2, + "valid_targets_min": 771 + }, + { + "epoch": 6.466372657111356, + "grad_norm": 0.6263019374144929, + "learning_rate": 7.05402878901118e-07, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09662850946187973, + "step": 11730, + "valid_targets_mean": 4016.5, + "valid_targets_min": 1774 + }, + { + "epoch": 6.469128996692392, + "grad_norm": 0.7097550828307055, + "learning_rate": 6.981845058904646e-07, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07941223680973053, + "step": 11735, + "valid_targets_mean": 2433.4, + "valid_targets_min": 997 + }, + { + "epoch": 6.471885336273429, + "grad_norm": 0.6240273597535071, + "learning_rate": 6.91002599656887e-07, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09605465829372406, + "step": 11740, + "valid_targets_mean": 3968.9, + "valid_targets_min": 1741 + }, + { + "epoch": 6.474641675854465, + "grad_norm": 0.6577447812263911, + "learning_rate": 6.838571737691024e-07, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.092373788356781, + "step": 11745, + "valid_targets_mean": 3747.4, + "valid_targets_min": 863 + }, + { + "epoch": 6.477398015435502, + "grad_norm": 0.6403877567872455, + "learning_rate": 6.767482417269145e-07, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0892789214849472, + "step": 11750, + "valid_targets_mean": 3097.6, + "valid_targets_min": 1410 + }, + { + "epoch": 6.480154355016538, + "grad_norm": 0.5723115702324024, + "learning_rate": 6.69675816961175e-07, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08157838135957718, + "step": 11755, + "valid_targets_mean": 4840.6, + "valid_targets_min": 2110 + }, + { + "epoch": 6.4829106945975745, + "grad_norm": 0.6399883801070787, + "learning_rate": 6.626399128337647e-07, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08690197765827179, + "step": 11760, + "valid_targets_mean": 3410.2, + "valid_targets_min": 1318 + }, + { + "epoch": 6.4856670341786105, + "grad_norm": 0.6137870029608242, + "learning_rate": 6.556405426375634e-07, + "loss": 0.1926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0992293730378151, + "step": 11765, + "valid_targets_mean": 4229.5, + "valid_targets_min": 1876 + }, + { + "epoch": 6.488423373759647, + "grad_norm": 0.652098258055984, + "learning_rate": 6.486777195964334e-07, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11016365140676498, + "step": 11770, + "valid_targets_mean": 3783.2, + "valid_targets_min": 2605 + }, + { + "epoch": 6.491179713340683, + "grad_norm": 0.6543726399398024, + "learning_rate": 6.417514568651783e-07, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08489050716161728, + "step": 11775, + "valid_targets_mean": 3068.0, + "valid_targets_min": 1446 + }, + { + "epoch": 6.49393605292172, + "grad_norm": 0.6259584717340286, + "learning_rate": 6.348617675295399e-07, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09272638708353043, + "step": 11780, + "valid_targets_mean": 3860.8, + "valid_targets_min": 1270 + }, + { + "epoch": 6.496692392502756, + "grad_norm": 0.6104313174751909, + "learning_rate": 6.280086646061567e-07, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10172820091247559, + "step": 11785, + "valid_targets_mean": 5586.5, + "valid_targets_min": 2033 + }, + { + "epoch": 6.499448732083793, + "grad_norm": 0.6471017908914637, + "learning_rate": 6.211921610425431e-07, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10383600741624832, + "step": 11790, + "valid_targets_mean": 4080.6, + "valid_targets_min": 1924 + }, + { + "epoch": 6.502205071664829, + "grad_norm": 0.6607444201491905, + "learning_rate": 6.144122697170707e-07, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10039277374744415, + "step": 11795, + "valid_targets_mean": 3680.1, + "valid_targets_min": 2098 + }, + { + "epoch": 6.504961411245866, + "grad_norm": 0.6787475574680036, + "learning_rate": 6.076690034389355e-07, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12542612850666046, + "step": 11800, + "valid_targets_mean": 4332.0, + "valid_targets_min": 1363 + }, + { + "epoch": 6.5077177508269015, + "grad_norm": 0.6671721253401424, + "learning_rate": 6.009623749481442e-07, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10061569511890411, + "step": 11805, + "valid_targets_mean": 3358.1, + "valid_targets_min": 1087 + }, + { + "epoch": 6.510474090407938, + "grad_norm": 0.6392777686836943, + "learning_rate": 5.94292396915479e-07, + "loss": 0.1967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11996272206306458, + "step": 11810, + "valid_targets_mean": 4630.6, + "valid_targets_min": 2483 + }, + { + "epoch": 6.513230429988974, + "grad_norm": 0.6273099620525072, + "learning_rate": 5.876590819424777e-07, + "loss": 0.1952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10927227884531021, + "step": 11815, + "valid_targets_mean": 4551.2, + "valid_targets_min": 2316 + }, + { + "epoch": 6.515986769570011, + "grad_norm": 0.6277344843799346, + "learning_rate": 5.810624425614175e-07, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09606456011533737, + "step": 11820, + "valid_targets_mean": 3923.0, + "valid_targets_min": 2022 + }, + { + "epoch": 6.518743109151047, + "grad_norm": 0.5944519669905324, + "learning_rate": 5.745024912352759e-07, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06713511049747467, + "step": 11825, + "valid_targets_mean": 2890.6, + "valid_targets_min": 1294 + }, + { + "epoch": 6.521499448732084, + "grad_norm": 0.6804157498189497, + "learning_rate": 5.679792403577234e-07, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11154188960790634, + "step": 11830, + "valid_targets_mean": 3638.6, + "valid_targets_min": 1700 + }, + { + "epoch": 6.52425578831312, + "grad_norm": 0.7039121745005009, + "learning_rate": 5.614927022530881e-07, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08469922095537186, + "step": 11835, + "valid_targets_mean": 2722.4, + "valid_targets_min": 1586 + }, + { + "epoch": 6.527012127894157, + "grad_norm": 0.7050538625744206, + "learning_rate": 5.550428891763382e-07, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10084112733602524, + "step": 11840, + "valid_targets_mean": 3360.6, + "valid_targets_min": 2407 + }, + { + "epoch": 6.529768467475193, + "grad_norm": 0.6899260322327609, + "learning_rate": 5.486298133130574e-07, + "loss": 0.1971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1089468002319336, + "step": 11845, + "valid_targets_mean": 4063.0, + "valid_targets_min": 2290 + }, + { + "epoch": 6.532524807056229, + "grad_norm": 0.6679450013372239, + "learning_rate": 5.422534867794227e-07, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12761497497558594, + "step": 11850, + "valid_targets_mean": 5192.1, + "valid_targets_min": 2189 + }, + { + "epoch": 6.535281146637265, + "grad_norm": 0.7864234449580422, + "learning_rate": 5.359139216221843e-07, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12240655720233917, + "step": 11855, + "valid_targets_mean": 4296.2, + "valid_targets_min": 2609 + }, + { + "epoch": 6.538037486218302, + "grad_norm": 0.652728073897788, + "learning_rate": 5.296111298186324e-07, + "loss": 0.1996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13656824827194214, + "step": 11860, + "valid_targets_mean": 4534.8, + "valid_targets_min": 2011 + }, + { + "epoch": 6.540793825799338, + "grad_norm": 0.6010888611381239, + "learning_rate": 5.23345123276584e-07, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07805797457695007, + "step": 11865, + "valid_targets_mean": 3611.1, + "valid_targets_min": 1912 + }, + { + "epoch": 6.543550165380375, + "grad_norm": 0.6458113207046382, + "learning_rate": 5.171159138343607e-07, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10506947338581085, + "step": 11870, + "valid_targets_mean": 3801.9, + "valid_targets_min": 2266 + }, + { + "epoch": 6.546306504961411, + "grad_norm": 0.5885751080290865, + "learning_rate": 5.109235132607593e-07, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0811060220003128, + "step": 11875, + "valid_targets_mean": 4454.2, + "valid_targets_min": 904 + }, + { + "epoch": 6.549062844542448, + "grad_norm": 0.8615159461660636, + "learning_rate": 5.047679332550392e-07, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11189670860767365, + "step": 11880, + "valid_targets_mean": 3787.4, + "valid_targets_min": 1219 + }, + { + "epoch": 6.551819184123484, + "grad_norm": 0.6078396403730886, + "learning_rate": 4.986491854468933e-07, + "loss": 0.1906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09138184785842896, + "step": 11885, + "valid_targets_mean": 4048.5, + "valid_targets_min": 1658 + }, + { + "epoch": 6.5545755237045205, + "grad_norm": 0.7163765441718282, + "learning_rate": 4.925672813964211e-07, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09492910653352737, + "step": 11890, + "valid_targets_mean": 3499.8, + "valid_targets_min": 1423 + }, + { + "epoch": 6.557331863285556, + "grad_norm": 0.7841943103034212, + "learning_rate": 4.865222325941244e-07, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07614389061927795, + "step": 11895, + "valid_targets_mean": 3917.6, + "valid_targets_min": 2320 + }, + { + "epoch": 6.560088202866593, + "grad_norm": 0.6307155457073156, + "learning_rate": 4.805140504608629e-07, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08870051801204681, + "step": 11900, + "valid_targets_mean": 3650.2, + "valid_targets_min": 1188 + }, + { + "epoch": 6.56284454244763, + "grad_norm": 0.6326838345080782, + "learning_rate": 4.7454274634785914e-07, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08016106486320496, + "step": 11905, + "valid_targets_mean": 3416.4, + "valid_targets_min": 618 + }, + { + "epoch": 6.565600882028666, + "grad_norm": 0.6660671080083779, + "learning_rate": 4.686083315366463e-07, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08724120259284973, + "step": 11910, + "valid_targets_mean": 2667.1, + "valid_targets_min": 1364 + }, + { + "epoch": 6.568357221609702, + "grad_norm": 0.6637149908070656, + "learning_rate": 4.627108172390737e-07, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10071809589862823, + "step": 11915, + "valid_targets_mean": 3505.4, + "valid_targets_min": 1397 + }, + { + "epoch": 6.571113561190739, + "grad_norm": 0.6153937842192099, + "learning_rate": 4.5685021459727305e-07, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09327477216720581, + "step": 11920, + "valid_targets_mean": 4666.4, + "valid_targets_min": 1415 + }, + { + "epoch": 6.573869900771776, + "grad_norm": 0.6876774195753756, + "learning_rate": 4.510265346836318e-07, + "loss": 0.1793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09770476818084717, + "step": 11925, + "valid_targets_mean": 3801.8, + "valid_targets_min": 621 + }, + { + "epoch": 6.5766262403528115, + "grad_norm": 0.6112481277954771, + "learning_rate": 4.4523978850079087e-07, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07001069188117981, + "step": 11930, + "valid_targets_mean": 3552.0, + "valid_targets_min": 1290 + }, + { + "epoch": 6.5793825799338475, + "grad_norm": 0.717713627011567, + "learning_rate": 4.3948998698160496e-07, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07267051935195923, + "step": 11935, + "valid_targets_mean": 2124.8, + "valid_targets_min": 642 + }, + { + "epoch": 6.582138919514884, + "grad_norm": 0.6934722765572611, + "learning_rate": 4.3377714098913115e-07, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10135942697525024, + "step": 11940, + "valid_targets_mean": 3453.9, + "valid_targets_min": 2109 + }, + { + "epoch": 6.584895259095921, + "grad_norm": 0.7200257034511055, + "learning_rate": 4.281012613166069e-07, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1362164467573166, + "step": 11945, + "valid_targets_mean": 4593.1, + "valid_targets_min": 2786 + }, + { + "epoch": 6.587651598676957, + "grad_norm": 0.6461775341405339, + "learning_rate": 4.2246235868742993e-07, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07973875105381012, + "step": 11950, + "valid_targets_mean": 3278.0, + "valid_targets_min": 1510 + }, + { + "epoch": 6.590407938257993, + "grad_norm": 0.7639907401855286, + "learning_rate": 4.1686044375513824e-07, + "loss": 0.1828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11009377241134644, + "step": 11955, + "valid_targets_mean": 3222.0, + "valid_targets_min": 987 + }, + { + "epoch": 6.59316427783903, + "grad_norm": 0.625782376379307, + "learning_rate": 4.1129552710338806e-07, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08447177708148956, + "step": 11960, + "valid_targets_mean": 3659.5, + "valid_targets_min": 1609 + }, + { + "epoch": 6.595920617420067, + "grad_norm": 0.6004315816828475, + "learning_rate": 4.0576761924593364e-07, + "loss": 0.1837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09928400069475174, + "step": 11965, + "valid_targets_mean": 4652.5, + "valid_targets_min": 1700 + }, + { + "epoch": 6.598676957001103, + "grad_norm": 0.5567896944584304, + "learning_rate": 4.00276730626612e-07, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07947394251823425, + "step": 11970, + "valid_targets_mean": 5438.1, + "valid_targets_min": 1767 + }, + { + "epoch": 6.6014332965821385, + "grad_norm": 0.5447775435107003, + "learning_rate": 3.948228716193181e-07, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08383291214704514, + "step": 11975, + "valid_targets_mean": 4504.2, + "valid_targets_min": 2381 + }, + { + "epoch": 6.604189636163175, + "grad_norm": 1.0861953182638164, + "learning_rate": 3.8940605252798525e-07, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08314919471740723, + "step": 11980, + "valid_targets_mean": 2680.2, + "valid_targets_min": 1542 + }, + { + "epoch": 6.606945975744212, + "grad_norm": 0.608168751958949, + "learning_rate": 3.840262835865716e-07, + "loss": 0.1849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06814450025558472, + "step": 11985, + "valid_targets_mean": 3123.9, + "valid_targets_min": 1476 + }, + { + "epoch": 6.609702315325248, + "grad_norm": 0.6588681544651486, + "learning_rate": 3.7868357495903563e-07, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09766653925180435, + "step": 11990, + "valid_targets_mean": 3869.0, + "valid_targets_min": 1711 + }, + { + "epoch": 6.612458654906284, + "grad_norm": 0.6449505854371408, + "learning_rate": 3.7337793673931423e-07, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08892601728439331, + "step": 11995, + "valid_targets_mean": 3236.9, + "valid_targets_min": 1669 + }, + { + "epoch": 6.615214994487321, + "grad_norm": 0.6940211170258311, + "learning_rate": 3.681093789513135e-07, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08802533149719238, + "step": 12000, + "valid_targets_mean": 3207.0, + "valid_targets_min": 1318 + }, + { + "epoch": 6.617971334068358, + "grad_norm": 0.7377315976734184, + "learning_rate": 3.6287791154887785e-07, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10279898345470428, + "step": 12005, + "valid_targets_mean": 2656.8, + "valid_targets_min": 1002 + }, + { + "epoch": 6.620727673649394, + "grad_norm": 0.6576498307706411, + "learning_rate": 3.5768354441578333e-07, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07606662064790726, + "step": 12010, + "valid_targets_mean": 2533.6, + "valid_targets_min": 845 + }, + { + "epoch": 6.62348401323043, + "grad_norm": 0.6007548428321595, + "learning_rate": 3.525262873657065e-07, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09908254444599152, + "step": 12015, + "valid_targets_mean": 4455.6, + "valid_targets_min": 2611 + }, + { + "epoch": 6.626240352811466, + "grad_norm": 0.645918010609413, + "learning_rate": 3.474061501422155e-07, + "loss": 0.2057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08578713238239288, + "step": 12020, + "valid_targets_mean": 3460.1, + "valid_targets_min": 1241 + }, + { + "epoch": 6.628996692392503, + "grad_norm": 0.7404407713923428, + "learning_rate": 3.423231424187501e-07, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1273348480463028, + "step": 12025, + "valid_targets_mean": 3751.0, + "valid_targets_min": 992 + }, + { + "epoch": 6.631753031973539, + "grad_norm": 0.6813452579774011, + "learning_rate": 3.3727727379859523e-07, + "loss": 0.1862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09462852776050568, + "step": 12030, + "valid_targets_mean": 3815.0, + "valid_targets_min": 1133 + }, + { + "epoch": 6.634509371554575, + "grad_norm": 0.759355013891321, + "learning_rate": 3.32268553814874e-07, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10586322844028473, + "step": 12035, + "valid_targets_mean": 4047.4, + "valid_targets_min": 2256 + }, + { + "epoch": 6.637265711135612, + "grad_norm": 0.6437410020278483, + "learning_rate": 3.2729699193052357e-07, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07768144458532333, + "step": 12040, + "valid_targets_mean": 3817.2, + "valid_targets_min": 1919 + }, + { + "epoch": 6.640022050716649, + "grad_norm": 0.6378691418699467, + "learning_rate": 3.223625975382838e-07, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09436723589897156, + "step": 12045, + "valid_targets_mean": 3366.4, + "valid_targets_min": 2763 + }, + { + "epoch": 6.642778390297685, + "grad_norm": 0.8229721273939434, + "learning_rate": 3.174653799606642e-07, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09201578795909882, + "step": 12050, + "valid_targets_mean": 2599.4, + "valid_targets_min": 617 + }, + { + "epoch": 6.645534729878721, + "grad_norm": 0.6116592035991468, + "learning_rate": 3.126053484499436e-07, + "loss": 0.1743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08392857015132904, + "step": 12055, + "valid_targets_mean": 3771.2, + "valid_targets_min": 3147 + }, + { + "epoch": 6.6482910694597575, + "grad_norm": 0.6311222317328303, + "learning_rate": 3.0778251218814394e-07, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09503331780433655, + "step": 12060, + "valid_targets_mean": 3780.9, + "valid_targets_min": 1844 + }, + { + "epoch": 6.651047409040794, + "grad_norm": 0.6615474428388356, + "learning_rate": 3.0299688028701204e-07, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0888574868440628, + "step": 12065, + "valid_targets_mean": 3895.0, + "valid_targets_min": 2330 + }, + { + "epoch": 6.65380374862183, + "grad_norm": 0.8125445144881208, + "learning_rate": 2.9824846178801105e-07, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09517012536525726, + "step": 12070, + "valid_targets_mean": 2856.8, + "valid_targets_min": 1066 + }, + { + "epoch": 6.656560088202866, + "grad_norm": 0.6072604772385034, + "learning_rate": 2.93537265662287e-07, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08542303740978241, + "step": 12075, + "valid_targets_mean": 3496.4, + "valid_targets_min": 1713 + }, + { + "epoch": 6.659316427783903, + "grad_norm": 0.6565747565340673, + "learning_rate": 2.888633008106734e-07, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09407410770654678, + "step": 12080, + "valid_targets_mean": 3918.8, + "valid_targets_min": 981 + }, + { + "epoch": 6.66207276736494, + "grad_norm": 0.6151075556070734, + "learning_rate": 2.842265760636531e-07, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10053390264511108, + "step": 12085, + "valid_targets_mean": 4680.2, + "valid_targets_min": 3325 + }, + { + "epoch": 6.664829106945976, + "grad_norm": 0.7158219670707938, + "learning_rate": 2.796271001813566e-07, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09116075932979584, + "step": 12090, + "valid_targets_mean": 3138.9, + "valid_targets_min": 1353 + }, + { + "epoch": 6.667585446527012, + "grad_norm": 0.5980259409095454, + "learning_rate": 2.75064881853544e-07, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08184714615345001, + "step": 12095, + "valid_targets_mean": 3986.6, + "valid_targets_min": 2273 + }, + { + "epoch": 6.6703417861080485, + "grad_norm": 0.641159749756951, + "learning_rate": 2.705399296995759e-07, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07826073467731476, + "step": 12100, + "valid_targets_mean": 3204.4, + "valid_targets_min": 2106 + }, + { + "epoch": 6.673098125689085, + "grad_norm": 0.6180556493285716, + "learning_rate": 2.660522522684117e-07, + "loss": 0.1757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07708348333835602, + "step": 12105, + "valid_targets_mean": 3788.4, + "valid_targets_min": 1687 + }, + { + "epoch": 6.675854465270121, + "grad_norm": 0.6584530982640199, + "learning_rate": 2.6160185803858926e-07, + "loss": 0.1826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09951594471931458, + "step": 12110, + "valid_targets_mean": 3502.8, + "valid_targets_min": 2246 + }, + { + "epoch": 6.678610804851157, + "grad_norm": 0.6503866882600573, + "learning_rate": 2.5718875541820285e-07, + "loss": 0.1801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0868140384554863, + "step": 12115, + "valid_targets_mean": 3663.8, + "valid_targets_min": 906 + }, + { + "epoch": 6.681367144432194, + "grad_norm": 0.6410739734105947, + "learning_rate": 2.5281295274489413e-07, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11007958650588989, + "step": 12120, + "valid_targets_mean": 4811.0, + "valid_targets_min": 1874 + }, + { + "epoch": 6.684123484013231, + "grad_norm": 0.6146106562511152, + "learning_rate": 2.484744582858389e-07, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09254806488752365, + "step": 12125, + "valid_targets_mean": 3815.6, + "valid_targets_min": 873 + }, + { + "epoch": 6.686879823594267, + "grad_norm": 0.700806610116043, + "learning_rate": 2.4417328023771835e-07, + "loss": 0.1738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08219181001186371, + "step": 12130, + "valid_targets_mean": 3252.1, + "valid_targets_min": 1697 + }, + { + "epoch": 6.689636163175303, + "grad_norm": 0.6815593635686432, + "learning_rate": 2.399094267267188e-07, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07025375217199326, + "step": 12135, + "valid_targets_mean": 2395.0, + "valid_targets_min": 897 + }, + { + "epoch": 6.69239250275634, + "grad_norm": 0.7153914659956025, + "learning_rate": 2.3568290580850307e-07, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08613619953393936, + "step": 12140, + "valid_targets_mean": 3104.0, + "valid_targets_min": 1695 + }, + { + "epoch": 6.695148842337376, + "grad_norm": 0.6679613653144995, + "learning_rate": 2.3149372546821258e-07, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07888029515743256, + "step": 12145, + "valid_targets_mean": 3211.9, + "valid_targets_min": 1712 + }, + { + "epoch": 6.697905181918412, + "grad_norm": 0.6001086973185641, + "learning_rate": 2.273418936204297e-07, + "loss": 0.1761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0760924220085144, + "step": 12150, + "valid_targets_mean": 3500.5, + "valid_targets_min": 993 + }, + { + "epoch": 6.700661521499448, + "grad_norm": 0.5960386040650999, + "learning_rate": 2.2322741810918202e-07, + "loss": 0.1724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09013932943344116, + "step": 12155, + "valid_targets_mean": 4181.4, + "valid_targets_min": 1679 + }, + { + "epoch": 6.703417861080485, + "grad_norm": 0.589655002536354, + "learning_rate": 2.1915030670791594e-07, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07699194550514221, + "step": 12160, + "valid_targets_mean": 3779.9, + "valid_targets_min": 2250 + }, + { + "epoch": 6.706174200661522, + "grad_norm": 0.6818963002304603, + "learning_rate": 2.1511056711948751e-07, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09373496472835541, + "step": 12165, + "valid_targets_mean": 4175.8, + "valid_targets_min": 2892 + }, + { + "epoch": 6.708930540242558, + "grad_norm": 0.6161690530056709, + "learning_rate": 2.1110820697614943e-07, + "loss": 0.1848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09949900209903717, + "step": 12170, + "valid_targets_mean": 4485.5, + "valid_targets_min": 2965 + }, + { + "epoch": 6.711686879823594, + "grad_norm": 0.6187759568333919, + "learning_rate": 2.071432338395285e-07, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07552440464496613, + "step": 12175, + "valid_targets_mean": 3737.5, + "valid_targets_min": 1817 + }, + { + "epoch": 6.714443219404631, + "grad_norm": 0.6446058631185863, + "learning_rate": 2.0321565520061925e-07, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10836657881736755, + "step": 12180, + "valid_targets_mean": 4726.0, + "valid_targets_min": 1727 + }, + { + "epoch": 6.7171995589856675, + "grad_norm": 0.6026129299417062, + "learning_rate": 1.9932547847976823e-07, + "loss": 0.1799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07878053188323975, + "step": 12185, + "valid_targets_mean": 3801.2, + "valid_targets_min": 1959 + }, + { + "epoch": 6.719955898566703, + "grad_norm": 0.5220463406109639, + "learning_rate": 1.9547271102665854e-07, + "loss": 0.1432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06594249606132507, + "step": 12190, + "valid_targets_mean": 2283.4, + "valid_targets_min": 333 + }, + { + "epoch": 6.722712238147739, + "grad_norm": 0.3909883436457965, + "learning_rate": 1.91657360120292e-07, + "loss": 0.0847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03644857555627823, + "step": 12195, + "valid_targets_mean": 3613.2, + "valid_targets_min": 712 + }, + { + "epoch": 6.725468577728776, + "grad_norm": 0.45788162218901496, + "learning_rate": 1.878794329689848e-07, + "loss": 0.0758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03990302234888077, + "step": 12200, + "valid_targets_mean": 2940.8, + "valid_targets_min": 732 + }, + { + "epoch": 6.728224917309813, + "grad_norm": 0.5158179063068205, + "learning_rate": 1.8413893671034744e-07, + "loss": 0.0811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07193579524755478, + "step": 12205, + "valid_targets_mean": 2727.5, + "valid_targets_min": 927 + }, + { + "epoch": 6.730981256890849, + "grad_norm": 0.3715249011622323, + "learning_rate": 1.804358784112692e-07, + "loss": 0.0728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029560724273324013, + "step": 12210, + "valid_targets_mean": 2835.4, + "valid_targets_min": 818 + }, + { + "epoch": 6.733737596471885, + "grad_norm": 0.4807673845042068, + "learning_rate": 1.7677026506791372e-07, + "loss": 0.078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0408896878361702, + "step": 12215, + "valid_targets_mean": 2910.0, + "valid_targets_min": 859 + }, + { + "epoch": 6.736493936052922, + "grad_norm": 0.85234481558148, + "learning_rate": 1.731421036056946e-07, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04766116291284561, + "step": 12220, + "valid_targets_mean": 1040.4, + "valid_targets_min": 539 + }, + { + "epoch": 6.7392502756339585, + "grad_norm": 0.4262097081362721, + "learning_rate": 1.6955140087927314e-07, + "loss": 0.0894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03587142378091812, + "step": 12225, + "valid_targets_mean": 3001.5, + "valid_targets_min": 920 + }, + { + "epoch": 6.7420066152149944, + "grad_norm": 0.471952747799622, + "learning_rate": 1.6599816367253385e-07, + "loss": 0.0836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04335939884185791, + "step": 12230, + "valid_targets_mean": 3366.2, + "valid_targets_min": 2295 + }, + { + "epoch": 6.74476295479603, + "grad_norm": 0.37799639424165166, + "learning_rate": 1.6248239869858463e-07, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028279876336455345, + "step": 12235, + "valid_targets_mean": 3784.6, + "valid_targets_min": 962 + }, + { + "epoch": 6.747519294377067, + "grad_norm": 0.38844987721167556, + "learning_rate": 1.5900411259973214e-07, + "loss": 0.0635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04187430068850517, + "step": 12240, + "valid_targets_mean": 4249.2, + "valid_targets_min": 2091 + }, + { + "epoch": 6.750275633958104, + "grad_norm": 0.41882727469572417, + "learning_rate": 1.5556331194747755e-07, + "loss": 0.0733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040693528950214386, + "step": 12245, + "valid_targets_mean": 4391.9, + "valid_targets_min": 3349 + }, + { + "epoch": 6.75303197353914, + "grad_norm": 0.414359538743385, + "learning_rate": 1.5216000324250079e-07, + "loss": 0.0732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035535454750061035, + "step": 12250, + "valid_targets_mean": 3944.4, + "valid_targets_min": 518 + }, + { + "epoch": 6.755788313120176, + "grad_norm": 0.5792107724122278, + "learning_rate": 1.4879419291464747e-07, + "loss": 0.1156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04074002057313919, + "step": 12255, + "valid_targets_mean": 1885.0, + "valid_targets_min": 549 + }, + { + "epoch": 6.758544652701213, + "grad_norm": 0.3973515999310904, + "learning_rate": 1.4546588732291532e-07, + "loss": 0.069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0340508371591568, + "step": 12260, + "valid_targets_mean": 3442.8, + "valid_targets_min": 2724 + }, + { + "epoch": 6.76130099228225, + "grad_norm": 0.5270437874198519, + "learning_rate": 1.4217509275545217e-07, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03382771834731102, + "step": 12265, + "valid_targets_mean": 1619.2, + "valid_targets_min": 518 + }, + { + "epoch": 6.7640573318632855, + "grad_norm": 0.3785315563525352, + "learning_rate": 1.3892181542952688e-07, + "loss": 0.0652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022429881617426872, + "step": 12270, + "valid_targets_mean": 3329.6, + "valid_targets_min": 844 + }, + { + "epoch": 6.766813671444321, + "grad_norm": 0.4597048065066275, + "learning_rate": 1.357060614915362e-07, + "loss": 0.0771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0368209183216095, + "step": 12275, + "valid_targets_mean": 2772.8, + "valid_targets_min": 577 + }, + { + "epoch": 6.769570011025358, + "grad_norm": 0.47195402946743803, + "learning_rate": 1.3252783701698026e-07, + "loss": 0.0786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040549248456954956, + "step": 12280, + "valid_targets_mean": 2086.0, + "valid_targets_min": 851 + }, + { + "epoch": 6.772326350606395, + "grad_norm": 0.5298467049399819, + "learning_rate": 1.2938714801044916e-07, + "loss": 0.0851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04280668869614601, + "step": 12285, + "valid_targets_mean": 2864.6, + "valid_targets_min": 774 + }, + { + "epoch": 6.775082690187431, + "grad_norm": 0.5046119365311154, + "learning_rate": 1.2628400040562983e-07, + "loss": 0.0819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03481695055961609, + "step": 12290, + "valid_targets_mean": 1743.6, + "valid_targets_min": 615 + }, + { + "epoch": 6.777839029768468, + "grad_norm": 0.5696418703541161, + "learning_rate": 1.2321840006527475e-07, + "loss": 0.12, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08726068586111069, + "step": 12295, + "valid_targets_mean": 3786.1, + "valid_targets_min": 2819 + }, + { + "epoch": 6.780595369349504, + "grad_norm": 0.32172191151049767, + "learning_rate": 1.2019035278119983e-07, + "loss": 0.0706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02877993881702423, + "step": 12300, + "valid_targets_mean": 4070.4, + "valid_targets_min": 3732 + }, + { + "epoch": 6.783351708930541, + "grad_norm": 0.45127270391519925, + "learning_rate": 1.1719986427427111e-07, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035286448895931244, + "step": 12305, + "valid_targets_mean": 1204.5, + "valid_targets_min": 714 + }, + { + "epoch": 6.786108048511577, + "grad_norm": 0.4170250230468887, + "learning_rate": 1.1424694019439575e-07, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03490734100341797, + "step": 12310, + "valid_targets_mean": 3416.2, + "valid_targets_min": 1812 + }, + { + "epoch": 6.788864388092613, + "grad_norm": 0.39673114740919957, + "learning_rate": 1.1133158612051331e-07, + "loss": 0.092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03723118454217911, + "step": 12315, + "valid_targets_mean": 3800.5, + "valid_targets_min": 3178 + }, + { + "epoch": 6.791620727673649, + "grad_norm": 0.7306074211597423, + "learning_rate": 1.084538075605801e-07, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1054002195596695, + "step": 12320, + "valid_targets_mean": 1691.4, + "valid_targets_min": 920 + }, + { + "epoch": 6.794377067254686, + "grad_norm": 0.4612553003193309, + "learning_rate": 1.0561360995156034e-07, + "loss": 0.2472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04180268943309784, + "step": 12325, + "valid_targets_mean": 3673.5, + "valid_targets_min": 2889 + }, + { + "epoch": 6.797133406835722, + "grad_norm": 0.43227465363070633, + "learning_rate": 1.0281099865941946e-07, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03677086532115936, + "step": 12330, + "valid_targets_mean": 3149.2, + "valid_targets_min": 833 + }, + { + "epoch": 6.799889746416759, + "grad_norm": 0.5124894833700212, + "learning_rate": 1.0004597897911084e-07, + "loss": 0.0842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060411401093006134, + "step": 12335, + "valid_targets_mean": 3340.4, + "valid_targets_min": 1803 + }, + { + "epoch": 6.802646085997795, + "grad_norm": 0.4544472226261938, + "learning_rate": 9.731855613456687e-08, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04826788231730461, + "step": 12340, + "valid_targets_mean": 3548.2, + "valid_targets_min": 2782 + }, + { + "epoch": 6.805402425578832, + "grad_norm": 0.3092707113465959, + "learning_rate": 9.462873527868343e-08, + "loss": 0.063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029714608564972878, + "step": 12345, + "valid_targets_mean": 4605.8, + "valid_targets_min": 2664 + }, + { + "epoch": 6.808158765159868, + "grad_norm": 0.3866299172219348, + "learning_rate": 9.197652149332215e-08, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.020423540845513344, + "step": 12350, + "valid_targets_mean": 2750.2, + "valid_targets_min": 560 + }, + { + "epoch": 6.8109151047409044, + "grad_norm": 0.42083414839603134, + "learning_rate": 8.936191978929254e-08, + "loss": 0.0824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039483144879341125, + "step": 12355, + "valid_targets_mean": 4261.1, + "valid_targets_min": 1615 + }, + { + "epoch": 6.81367144432194, + "grad_norm": 0.4039241318188518, + "learning_rate": 8.678493510633656e-08, + "loss": 0.0701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032757408916950226, + "step": 12360, + "valid_targets_mean": 3580.5, + "valid_targets_min": 2172 + }, + { + "epoch": 6.816427783902977, + "grad_norm": 0.4122039614742872, + "learning_rate": 8.424557231313746e-08, + "loss": 0.0687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03359784558415413, + "step": 12365, + "valid_targets_mean": 3065.4, + "valid_targets_min": 536 + }, + { + "epoch": 6.819184123484013, + "grad_norm": 0.4569222227881019, + "learning_rate": 8.174383620729087e-08, + "loss": 0.0702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03456124663352966, + "step": 12370, + "valid_targets_mean": 3633.1, + "valid_targets_min": 2264 + }, + { + "epoch": 6.82194046306505, + "grad_norm": 0.3931431426941363, + "learning_rate": 7.927973151531155e-08, + "loss": 0.0674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028268206864595413, + "step": 12375, + "valid_targets_mean": 2809.9, + "valid_targets_min": 820 + }, + { + "epoch": 6.824696802646086, + "grad_norm": 0.7223468191997556, + "learning_rate": 7.685326289261108e-08, + "loss": 0.1088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10635796189308167, + "step": 12380, + "valid_targets_mean": 2020.5, + "valid_targets_min": 864 + }, + { + "epoch": 6.827453142227123, + "grad_norm": 0.41388616680797247, + "learning_rate": 7.446443492350242e-08, + "loss": 0.0663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03578038513660431, + "step": 12385, + "valid_targets_mean": 3609.9, + "valid_targets_min": 2719 + }, + { + "epoch": 6.830209481808159, + "grad_norm": 0.6497017787282025, + "learning_rate": 7.211325212117759e-08, + "loss": 0.0639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047782279551029205, + "step": 12390, + "valid_targets_mean": 1611.6, + "valid_targets_min": 785 + }, + { + "epoch": 6.8329658213891955, + "grad_norm": 0.351740590926108, + "learning_rate": 6.979971892770998e-08, + "loss": 0.0638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02668915130198002, + "step": 12395, + "valid_targets_mean": 3711.8, + "valid_targets_min": 2353 + }, + { + "epoch": 6.835722160970231, + "grad_norm": 0.3908225232173894, + "learning_rate": 6.752383971404097e-08, + "loss": 0.0647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027315063402056694, + "step": 12400, + "valid_targets_mean": 1996.5, + "valid_targets_min": 549 + }, + { + "epoch": 6.838478500551268, + "grad_norm": 0.42117055427360395, + "learning_rate": 6.528561877997108e-08, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034233927726745605, + "step": 12405, + "valid_targets_mean": 3332.5, + "valid_targets_min": 908 + }, + { + "epoch": 6.841234840132304, + "grad_norm": 0.4325717714330446, + "learning_rate": 6.308506035415552e-08, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06830979883670807, + "step": 12410, + "valid_targets_mean": 3527.0, + "valid_targets_min": 2196 + }, + { + "epoch": 6.843991179713341, + "grad_norm": 0.37220981377896306, + "learning_rate": 6.092216859409306e-08, + "loss": 0.0787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023197397589683533, + "step": 12415, + "valid_targets_mean": 1686.0, + "valid_targets_min": 598 + }, + { + "epoch": 6.846747519294377, + "grad_norm": 0.9411783637766447, + "learning_rate": 5.879694758611942e-08, + "loss": 0.0911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07831624150276184, + "step": 12420, + "valid_targets_mean": 1330.2, + "valid_targets_min": 578 + }, + { + "epoch": 6.849503858875414, + "grad_norm": 0.6080331147119308, + "learning_rate": 5.6709401345396154e-08, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04399195313453674, + "step": 12425, + "valid_targets_mean": 1678.9, + "valid_targets_min": 695 + }, + { + "epoch": 6.85226019845645, + "grad_norm": 0.5020349777580884, + "learning_rate": 5.4659533815912826e-08, + "loss": 0.0798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04514855146408081, + "step": 12430, + "valid_targets_mean": 1888.5, + "valid_targets_min": 691 + }, + { + "epoch": 6.855016538037487, + "grad_norm": 0.408588204066445, + "learning_rate": 5.264734887046707e-08, + "loss": 0.0921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030762769281864166, + "step": 12435, + "valid_targets_mean": 3686.0, + "valid_targets_min": 2024 + }, + { + "epoch": 6.8577728776185225, + "grad_norm": 0.638486488939466, + "learning_rate": 5.06728503106646e-08, + "loss": 0.0745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039804793894290924, + "step": 12440, + "valid_targets_mean": 1356.1, + "valid_targets_min": 523 + }, + { + "epoch": 6.860529217199559, + "grad_norm": 0.33776795442332186, + "learning_rate": 4.8736041866912496e-08, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027880966663360596, + "step": 12445, + "valid_targets_mean": 4120.2, + "valid_targets_min": 468 + }, + { + "epoch": 6.863285556780595, + "grad_norm": 0.3441553133279704, + "learning_rate": 4.683692719840594e-08, + "loss": 0.0788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02116471529006958, + "step": 12450, + "valid_targets_mean": 5551.2, + "valid_targets_min": 4225 + }, + { + "epoch": 6.866041896361632, + "grad_norm": 0.3577275991179054, + "learning_rate": 4.497550989313038e-08, + "loss": 0.0694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02522752620279789, + "step": 12455, + "valid_targets_mean": 3806.8, + "valid_targets_min": 459 + }, + { + "epoch": 6.868798235942668, + "grad_norm": 0.38212762522758753, + "learning_rate": 4.315179346784826e-08, + "loss": 0.0674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031144842505455017, + "step": 12460, + "valid_targets_mean": 3498.2, + "valid_targets_min": 981 + }, + { + "epoch": 6.871554575523705, + "grad_norm": 0.4531027225251501, + "learning_rate": 4.136578136809233e-08, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048623986542224884, + "step": 12465, + "valid_targets_mean": 2464.6, + "valid_targets_min": 697 + }, + { + "epoch": 6.874310915104741, + "grad_norm": 0.429242838228943, + "learning_rate": 3.96174769681612e-08, + "loss": 0.0841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04650985077023506, + "step": 12470, + "valid_targets_mean": 3671.2, + "valid_targets_min": 2759 + }, + { + "epoch": 6.877067254685778, + "grad_norm": 0.41519654001572204, + "learning_rate": 3.790688357111494e-08, + "loss": 0.0847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04367494583129883, + "step": 12475, + "valid_targets_mean": 3702.6, + "valid_targets_min": 2901 + }, + { + "epoch": 6.879823594266814, + "grad_norm": 0.39307067497367426, + "learning_rate": 3.62340044087639e-08, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031008891761302948, + "step": 12480, + "valid_targets_mean": 3455.4, + "valid_targets_min": 1099 + }, + { + "epoch": 6.88257993384785, + "grad_norm": 0.4372383919832764, + "learning_rate": 3.45988426416688e-08, + "loss": 0.0724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045628637075424194, + "step": 12485, + "valid_targets_mean": 3808.8, + "valid_targets_min": 1874 + }, + { + "epoch": 6.885336273428886, + "grad_norm": 0.39235922671681506, + "learning_rate": 3.3001401359125106e-08, + "loss": 0.0803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028775421902537346, + "step": 12490, + "valid_targets_mean": 3013.8, + "valid_targets_min": 724 + }, + { + "epoch": 6.888092613009923, + "grad_norm": 0.43766053297123725, + "learning_rate": 3.1441683579167545e-08, + "loss": 0.0594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02661021798849106, + "step": 12495, + "valid_targets_mean": 3806.5, + "valid_targets_min": 2769 + }, + { + "epoch": 6.890848952590959, + "grad_norm": 0.4621432857025419, + "learning_rate": 2.991969224856117e-08, + "loss": 0.0793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030213432386517525, + "step": 12500, + "valid_targets_mean": 2427.1, + "valid_targets_min": 591 + }, + { + "epoch": 6.893605292171996, + "grad_norm": 0.41857253315042975, + "learning_rate": 2.8435430242792496e-08, + "loss": 0.0745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03696204721927643, + "step": 12505, + "valid_targets_mean": 3443.2, + "valid_targets_min": 600 + }, + { + "epoch": 6.896361631753032, + "grad_norm": 0.4553247570479275, + "learning_rate": 2.6988900366067274e-08, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04457489773631096, + "step": 12510, + "valid_targets_mean": 3542.1, + "valid_targets_min": 1648 + }, + { + "epoch": 6.899117971334069, + "grad_norm": 0.5289416795035029, + "learning_rate": 2.558010535130162e-08, + "loss": 0.0775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03840632736682892, + "step": 12515, + "valid_targets_mean": 1957.9, + "valid_targets_min": 457 + }, + { + "epoch": 6.901874310915105, + "grad_norm": 0.7813040797291291, + "learning_rate": 2.4209047860122015e-08, + "loss": 0.0837, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04574095085263252, + "step": 12520, + "valid_targets_mean": 823.9, + "valid_targets_min": 612 + }, + { + "epoch": 6.904630650496141, + "grad_norm": 0.548787914532329, + "learning_rate": 2.2875730482856405e-08, + "loss": 0.078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06787295639514923, + "step": 12525, + "valid_targets_mean": 2383.2, + "valid_targets_min": 898 + }, + { + "epoch": 6.907386990077177, + "grad_norm": 0.4151890452596018, + "learning_rate": 2.1580155738534226e-08, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03062249720096588, + "step": 12530, + "valid_targets_mean": 3037.9, + "valid_targets_min": 1146 + }, + { + "epoch": 6.910143329658214, + "grad_norm": 0.46379196205177065, + "learning_rate": 2.0322326074875277e-08, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04577287659049034, + "step": 12535, + "valid_targets_mean": 3039.9, + "valid_targets_min": 1511 + }, + { + "epoch": 6.91289966923925, + "grad_norm": 0.5915621699417132, + "learning_rate": 1.9102243868285298e-08, + "loss": 0.0678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041524745523929596, + "step": 12540, + "valid_targets_mean": 1301.2, + "valid_targets_min": 628 + }, + { + "epoch": 6.915656008820287, + "grad_norm": 0.5884241052387209, + "learning_rate": 1.791991142385596e-08, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04933217167854309, + "step": 12545, + "valid_targets_mean": 1488.0, + "valid_targets_min": 691 + }, + { + "epoch": 6.918412348401323, + "grad_norm": 0.3591364905189163, + "learning_rate": 1.677533097536266e-08, + "loss": 0.0737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022982992231845856, + "step": 12550, + "valid_targets_mean": 3970.2, + "valid_targets_min": 3657 + }, + { + "epoch": 6.92116868798236, + "grad_norm": 0.3995600810057674, + "learning_rate": 1.5668504685251163e-08, + "loss": 0.0655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04318730905652046, + "step": 12555, + "valid_targets_mean": 5056.2, + "valid_targets_min": 3831 + }, + { + "epoch": 6.923925027563396, + "grad_norm": 0.3466962612074655, + "learning_rate": 1.4599434644639866e-08, + "loss": 0.0736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034520477056503296, + "step": 12560, + "valid_targets_mean": 2402.1, + "valid_targets_min": 762 + }, + { + "epoch": 6.9266813671444325, + "grad_norm": 0.39196700173136256, + "learning_rate": 1.3568122873313106e-08, + "loss": 0.0638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035335905849933624, + "step": 12565, + "valid_targets_mean": 3574.1, + "valid_targets_min": 667 + }, + { + "epoch": 6.929437706725468, + "grad_norm": 0.3941968702092846, + "learning_rate": 1.2574571319723395e-08, + "loss": 0.0624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04730827361345291, + "step": 12570, + "valid_targets_mean": 5005.6, + "valid_targets_min": 4104 + }, + { + "epoch": 6.932194046306505, + "grad_norm": 0.36864835416408176, + "learning_rate": 1.161878186098031e-08, + "loss": 0.0811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031641606241464615, + "step": 12575, + "valid_targets_mean": 3256.8, + "valid_targets_min": 784 + }, + { + "epoch": 6.934950385887541, + "grad_norm": 0.5155955099025141, + "learning_rate": 1.070075630284828e-08, + "loss": 0.0926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04099475219845772, + "step": 12580, + "valid_targets_mean": 3479.5, + "valid_targets_min": 1090 + }, + { + "epoch": 6.937706725468578, + "grad_norm": 0.43212592117045895, + "learning_rate": 9.820496379748801e-09, + "loss": 0.0729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04355751723051071, + "step": 12585, + "valid_targets_mean": 3237.6, + "valid_targets_min": 704 + }, + { + "epoch": 6.940463065049614, + "grad_norm": 0.3648365943596451, + "learning_rate": 8.978003754744891e-09, + "loss": 0.0763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02757037803530693, + "step": 12590, + "valid_targets_mean": 2824.0, + "valid_targets_min": 774 + }, + { + "epoch": 6.943219404630651, + "grad_norm": 0.3449380864952986, + "learning_rate": 8.17328001955886e-09, + "loss": 0.0936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02720784582197666, + "step": 12595, + "valid_targets_mean": 3551.1, + "valid_targets_min": 987 + }, + { + "epoch": 6.945975744211687, + "grad_norm": 0.4595368629250328, + "learning_rate": 7.406326694543442e-09, + "loss": 0.0706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04205995798110962, + "step": 12600, + "valid_targets_mean": 3200.4, + "valid_targets_min": 1048 + }, + { + "epoch": 6.948732083792724, + "grad_norm": 0.41137258628883905, + "learning_rate": 6.677145228699555e-09, + "loss": 0.0756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03384694457054138, + "step": 12605, + "valid_targets_mean": 3389.4, + "valid_targets_min": 1072 + }, + { + "epoch": 6.9514884233737595, + "grad_norm": 0.41647050906199456, + "learning_rate": 5.985736999665204e-09, + "loss": 0.0719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056980643421411514, + "step": 12610, + "valid_targets_mean": 3102.2, + "valid_targets_min": 906 + }, + { + "epoch": 6.954244762954796, + "grad_norm": 0.6310660771680378, + "learning_rate": 5.332103313713255e-09, + "loss": 0.1123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04794767126441002, + "step": 12615, + "valid_targets_mean": 1292.9, + "valid_targets_min": 570 + }, + { + "epoch": 6.957001102535832, + "grad_norm": 0.4823392176892498, + "learning_rate": 4.716245405747e-09, + "loss": 0.0887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04831778258085251, + "step": 12620, + "valid_targets_mean": 3471.0, + "valid_targets_min": 835 + }, + { + "epoch": 6.959757442116869, + "grad_norm": 0.44602579507617335, + "learning_rate": 4.138164439304593e-09, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02972240373492241, + "step": 12625, + "valid_targets_mean": 2156.0, + "valid_targets_min": 635 + }, + { + "epoch": 6.962513781697905, + "grad_norm": 0.3816195294655178, + "learning_rate": 3.5978615065479504e-09, + "loss": 0.0708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035399068146944046, + "step": 12630, + "valid_targets_mean": 2765.2, + "valid_targets_min": 1591 + }, + { + "epoch": 6.965270121278942, + "grad_norm": 0.40241035428757116, + "learning_rate": 3.0953376282694126e-09, + "loss": 0.0694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029072264209389687, + "step": 12635, + "valid_targets_mean": 3275.8, + "valid_targets_min": 800 + }, + { + "epoch": 6.968026460859978, + "grad_norm": 0.37866771160960916, + "learning_rate": 2.6305937538828595e-09, + "loss": 0.063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030400212854146957, + "step": 12640, + "valid_targets_mean": 3218.0, + "valid_targets_min": 958 + }, + { + "epoch": 6.970782800441015, + "grad_norm": 0.5636136931934452, + "learning_rate": 2.203630761423714e-09, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037963055074214935, + "step": 12645, + "valid_targets_mean": 2122.1, + "valid_targets_min": 889 + }, + { + "epoch": 6.9735391400220506, + "grad_norm": 0.4261677790560192, + "learning_rate": 1.81444945755338e-09, + "loss": 0.0634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03908313438296318, + "step": 12650, + "valid_targets_mean": 3309.6, + "valid_targets_min": 1024 + }, + { + "epoch": 6.976295479603087, + "grad_norm": 0.4624991509553099, + "learning_rate": 1.4630505775459213e-09, + "loss": 0.0688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03498661145567894, + "step": 12655, + "valid_targets_mean": 2179.5, + "valid_targets_min": 786 + }, + { + "epoch": 6.979051819184123, + "grad_norm": 0.4628864992941429, + "learning_rate": 1.1494347852991639e-09, + "loss": 0.0694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04407598078250885, + "step": 12660, + "valid_targets_mean": 3064.8, + "valid_targets_min": 2573 + }, + { + "epoch": 6.98180815876516, + "grad_norm": 0.349595869506714, + "learning_rate": 8.736026733213721e-10, + "loss": 0.0728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03491140156984329, + "step": 12665, + "valid_targets_mean": 3985.8, + "valid_targets_min": 2274 + }, + { + "epoch": 6.984564498346196, + "grad_norm": 0.5251134439727938, + "learning_rate": 6.355547627445724e-10, + "loss": 0.0626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039491817355155945, + "step": 12670, + "valid_targets_mean": 4017.9, + "valid_targets_min": 2431 + }, + { + "epoch": 6.987320837927233, + "grad_norm": 0.3942397708615575, + "learning_rate": 4.352915033090099e-10, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04371756315231323, + "step": 12675, + "valid_targets_mean": 4359.2, + "valid_targets_min": 2555 + }, + { + "epoch": 6.990077177508269, + "grad_norm": 0.57213961619608, + "learning_rate": 2.728132733720301e-10, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07508163154125214, + "step": 12680, + "valid_targets_mean": 2205.4, + "valid_targets_min": 1535 + }, + { + "epoch": 6.992833517089306, + "grad_norm": 0.32932520037608015, + "learning_rate": 1.4812037989919702e-10, + "loss": 0.1203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.01998429372906685, + "step": 12685, + "valid_targets_mean": 3608.8, + "valid_targets_min": 1097 + }, + { + "epoch": 6.995589856670342, + "grad_norm": 0.36955539605268856, + "learning_rate": 6.121305847317516e-11, + "loss": 0.0698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026766585186123848, + "step": 12690, + "valid_targets_mean": 3166.8, + "valid_targets_min": 840 + }, + { + "epoch": 6.998346196251378, + "grad_norm": 0.3456831196901275, + "learning_rate": 1.2091473289288502e-11, + "loss": 0.0662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030314182862639427, + "step": 12695, + "valid_targets_mean": 3711.2, + "valid_targets_min": 2555 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04305639863014221, + "step": 12698, + "total_flos": 3.8836436423755694e+18, + "train_loss": 0.06260271916794653, + "train_runtime": 120516.8314, + "train_samples_per_second": 1.686, + "train_steps_per_second": 0.105, + "valid_targets_mean": 4470.5, + "valid_targets_min": 1273 + } + ], + "logging_steps": 5, + "max_steps": 12698, + "num_input_tokens_seen": 0, + "num_train_epochs": 7, + "save_steps": 200, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": true + }, + "attributes": {} + } + }, + "total_flos": 3.8836436423755694e+18, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}