diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,29439 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 7.0, + "eval_steps": 500, + "global_step": 13363, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0026191723415400735, + "grad_norm": 13.956555681658662, + "learning_rate": 1.1967090501121915e-07, + "loss": 0.6701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3421918749809265, + "step": 5, + "valid_targets_mean": 7214.4, + "valid_targets_min": 4847 + }, + { + "epoch": 0.005238344683080147, + "grad_norm": 13.325049387429804, + "learning_rate": 2.692595362752431e-07, + "loss": 0.6204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.31004971265792847, + "step": 10, + "valid_targets_mean": 6417.1, + "valid_targets_min": 4480 + }, + { + "epoch": 0.00785751702462022, + "grad_norm": 13.699640282734906, + "learning_rate": 4.1884816753926706e-07, + "loss": 0.6694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3358825445175171, + "step": 15, + "valid_targets_mean": 7605.4, + "valid_targets_min": 5236 + }, + { + "epoch": 0.010476689366160294, + "grad_norm": 14.345261335867612, + "learning_rate": 5.68436798803291e-07, + "loss": 0.6766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3078409731388092, + "step": 20, + "valid_targets_mean": 5827.9, + "valid_targets_min": 4727 + }, + { + "epoch": 0.013095861707700367, + "grad_norm": 12.223050412159346, + "learning_rate": 7.18025430067315e-07, + "loss": 0.6772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3556755781173706, + "step": 25, + "valid_targets_mean": 6448.2, + "valid_targets_min": 4211 + }, + { + "epoch": 0.01571503404924044, + "grad_norm": 12.058397944361914, + "learning_rate": 8.676140613313389e-07, + "loss": 0.6443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.3585449457168579, + "step": 30, + "valid_targets_mean": 6108.1, + "valid_targets_min": 4279 + }, + { + "epoch": 0.018334206390780514, + "grad_norm": 9.191973227621443, + "learning_rate": 1.0172026925953627e-06, + "loss": 0.6135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.29062384366989136, + "step": 35, + "valid_targets_mean": 5962.4, + "valid_targets_min": 4589 + }, + { + "epoch": 0.020953378732320588, + "grad_norm": 7.701044083780416, + "learning_rate": 1.1667913238593869e-06, + "loss": 0.6035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26141878962516785, + "step": 40, + "valid_targets_mean": 3503.2, + "valid_targets_min": 2015 + }, + { + "epoch": 0.02357255107386066, + "grad_norm": 5.558341416081123, + "learning_rate": 1.3163799551234108e-06, + "loss": 0.5326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.24188977479934692, + "step": 45, + "valid_targets_mean": 6102.5, + "valid_targets_min": 4778 + }, + { + "epoch": 0.026191723415400735, + "grad_norm": 5.766246542601014, + "learning_rate": 1.4659685863874346e-06, + "loss": 0.5062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2454686462879181, + "step": 50, + "valid_targets_mean": 5868.2, + "valid_targets_min": 4612 + }, + { + "epoch": 0.028810895756940808, + "grad_norm": 5.537624622653609, + "learning_rate": 1.6155572176514587e-06, + "loss": 0.5051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2567406892776489, + "step": 55, + "valid_targets_mean": 6105.8, + "valid_targets_min": 4491 + }, + { + "epoch": 0.03143006809848088, + "grad_norm": 2.4850562095283864, + "learning_rate": 1.7651458489154827e-06, + "loss": 0.4409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.202129065990448, + "step": 60, + "valid_targets_mean": 6962.0, + "valid_targets_min": 4754 + }, + { + "epoch": 0.03404924044002095, + "grad_norm": 2.1010325517127866, + "learning_rate": 1.9147344801795064e-06, + "loss": 0.4247, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2229827344417572, + "step": 65, + "valid_targets_mean": 6035.1, + "valid_targets_min": 4787 + }, + { + "epoch": 0.03666841278156103, + "grad_norm": 1.5883947054414373, + "learning_rate": 2.0643231114435305e-06, + "loss": 0.4126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20870044827461243, + "step": 70, + "valid_targets_mean": 6028.5, + "valid_targets_min": 5213 + }, + { + "epoch": 0.0392875851231011, + "grad_norm": 1.1403568665938775, + "learning_rate": 2.2139117427075543e-06, + "loss": 0.3932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16978001594543457, + "step": 75, + "valid_targets_mean": 6814.1, + "valid_targets_min": 4664 + }, + { + "epoch": 0.041906757464641176, + "grad_norm": 1.1107104529872416, + "learning_rate": 2.3635003739715784e-06, + "loss": 0.4041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1902724802494049, + "step": 80, + "valid_targets_mean": 6281.5, + "valid_targets_min": 4996 + }, + { + "epoch": 0.044525929806181246, + "grad_norm": 0.9970327627008816, + "learning_rate": 2.513089005235602e-06, + "loss": 0.3896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1846439391374588, + "step": 85, + "valid_targets_mean": 6628.9, + "valid_targets_min": 4582 + }, + { + "epoch": 0.04714510214772132, + "grad_norm": 0.8283098880343396, + "learning_rate": 2.6626776364996263e-06, + "loss": 0.3733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16836076974868774, + "step": 90, + "valid_targets_mean": 6536.6, + "valid_targets_min": 5205 + }, + { + "epoch": 0.04976427448926139, + "grad_norm": 0.7120863556714087, + "learning_rate": 2.8122662677636505e-06, + "loss": 0.3547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1779661625623703, + "step": 95, + "valid_targets_mean": 7031.0, + "valid_targets_min": 5098 + }, + { + "epoch": 0.05238344683080147, + "grad_norm": 0.6577396133155977, + "learning_rate": 2.961854899027674e-06, + "loss": 0.3581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19818326830863953, + "step": 100, + "valid_targets_mean": 6514.2, + "valid_targets_min": 5016 + }, + { + "epoch": 0.05500261917234154, + "grad_norm": 0.6701149357989862, + "learning_rate": 3.111443530291698e-06, + "loss": 0.3887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2214498221874237, + "step": 105, + "valid_targets_mean": 6330.0, + "valid_targets_min": 965 + }, + { + "epoch": 0.057621791513881616, + "grad_norm": 0.617747536640169, + "learning_rate": 3.261032161555722e-06, + "loss": 0.3436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17294740676879883, + "step": 110, + "valid_targets_mean": 6293.1, + "valid_targets_min": 5218 + }, + { + "epoch": 0.060240963855421686, + "grad_norm": 0.604689696950089, + "learning_rate": 3.410620792819746e-06, + "loss": 0.3404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15567070245742798, + "step": 115, + "valid_targets_mean": 6957.4, + "valid_targets_min": 4213 + }, + { + "epoch": 0.06286013619696176, + "grad_norm": 0.6091930710026149, + "learning_rate": 3.56020942408377e-06, + "loss": 0.359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18683934211730957, + "step": 120, + "valid_targets_mean": 5888.5, + "valid_targets_min": 4923 + }, + { + "epoch": 0.06547930853850184, + "grad_norm": 0.5360841206416338, + "learning_rate": 3.709798055347794e-06, + "loss": 0.3692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21735292673110962, + "step": 125, + "valid_targets_mean": 7822.9, + "valid_targets_min": 4993 + }, + { + "epoch": 0.0680984808800419, + "grad_norm": 0.6592603919460518, + "learning_rate": 3.859386686611818e-06, + "loss": 0.364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17987853288650513, + "step": 130, + "valid_targets_mean": 6098.6, + "valid_targets_min": 4872 + }, + { + "epoch": 0.07071765322158198, + "grad_norm": 0.549816808312439, + "learning_rate": 4.008975317875842e-06, + "loss": 0.3364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17024990916252136, + "step": 135, + "valid_targets_mean": 6892.2, + "valid_targets_min": 4706 + }, + { + "epoch": 0.07333682556312206, + "grad_norm": 0.6044484095758595, + "learning_rate": 4.158563949139865e-06, + "loss": 0.329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15200582146644592, + "step": 140, + "valid_targets_mean": 6352.0, + "valid_targets_min": 4889 + }, + { + "epoch": 0.07595599790466212, + "grad_norm": 0.5614827588776087, + "learning_rate": 4.30815258040389e-06, + "loss": 0.3276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15746107697486877, + "step": 145, + "valid_targets_mean": 5948.9, + "valid_targets_min": 3888 + }, + { + "epoch": 0.0785751702462022, + "grad_norm": 0.6144959270829504, + "learning_rate": 4.457741211667914e-06, + "loss": 0.3325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14844751358032227, + "step": 150, + "valid_targets_mean": 5846.2, + "valid_targets_min": 4051 + }, + { + "epoch": 0.08119434258774227, + "grad_norm": 0.463231581806026, + "learning_rate": 4.607329842931937e-06, + "loss": 0.3149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1530732959508896, + "step": 155, + "valid_targets_mean": 7319.2, + "valid_targets_min": 5230 + }, + { + "epoch": 0.08381351492928235, + "grad_norm": 0.511543151528542, + "learning_rate": 4.756918474195962e-06, + "loss": 0.3158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15434357523918152, + "step": 160, + "valid_targets_mean": 8235.1, + "valid_targets_min": 4997 + }, + { + "epoch": 0.08643268727082241, + "grad_norm": 0.5954325868937105, + "learning_rate": 4.906507105459986e-06, + "loss": 0.3194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1663242131471634, + "step": 165, + "valid_targets_mean": 6402.5, + "valid_targets_min": 5364 + }, + { + "epoch": 0.08905185961236249, + "grad_norm": 0.5310263639652525, + "learning_rate": 5.056095736724009e-06, + "loss": 0.3081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17375975847244263, + "step": 170, + "valid_targets_mean": 7117.4, + "valid_targets_min": 3391 + }, + { + "epoch": 0.09167103195390257, + "grad_norm": 0.47338939130952606, + "learning_rate": 5.205684367988033e-06, + "loss": 0.2708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14417192339897156, + "step": 175, + "valid_targets_mean": 7542.8, + "valid_targets_min": 4829 + }, + { + "epoch": 0.09429020429544265, + "grad_norm": 0.5118037970090422, + "learning_rate": 5.355272999252057e-06, + "loss": 0.3046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1398244947195053, + "step": 180, + "valid_targets_mean": 6284.4, + "valid_targets_min": 5180 + }, + { + "epoch": 0.09690937663698271, + "grad_norm": 0.5608690453066975, + "learning_rate": 5.504861630516081e-06, + "loss": 0.3103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16154824197292328, + "step": 185, + "valid_targets_mean": 5806.2, + "valid_targets_min": 5101 + }, + { + "epoch": 0.09952854897852279, + "grad_norm": 0.4700042301838497, + "learning_rate": 5.654450261780105e-06, + "loss": 0.305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14155127108097076, + "step": 190, + "valid_targets_mean": 6697.9, + "valid_targets_min": 4527 + }, + { + "epoch": 0.10214772132006286, + "grad_norm": 0.5082832852018448, + "learning_rate": 5.80403889304413e-06, + "loss": 0.2992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.167373389005661, + "step": 195, + "valid_targets_mean": 6410.4, + "valid_targets_min": 3082 + }, + { + "epoch": 0.10476689366160294, + "grad_norm": 0.4793610696928589, + "learning_rate": 5.9536275243081535e-06, + "loss": 0.2775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1430361568927765, + "step": 200, + "valid_targets_mean": 6847.1, + "valid_targets_min": 4685 + }, + { + "epoch": 0.107386066003143, + "grad_norm": 0.5508740298949809, + "learning_rate": 6.103216155572177e-06, + "loss": 0.2971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14911168813705444, + "step": 205, + "valid_targets_mean": 6194.0, + "valid_targets_min": 4644 + }, + { + "epoch": 0.11000523834468308, + "grad_norm": 0.4774747009353938, + "learning_rate": 6.252804786836201e-06, + "loss": 0.2967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13261504471302032, + "step": 210, + "valid_targets_mean": 6283.8, + "valid_targets_min": 4433 + }, + { + "epoch": 0.11262441068622316, + "grad_norm": 0.5298361375559186, + "learning_rate": 6.402393418100225e-06, + "loss": 0.2878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14252270758152008, + "step": 215, + "valid_targets_mean": 5917.2, + "valid_targets_min": 4857 + }, + { + "epoch": 0.11524358302776323, + "grad_norm": 0.5242449416796856, + "learning_rate": 6.5519820493642484e-06, + "loss": 0.297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17171205580234528, + "step": 220, + "valid_targets_mean": 6986.4, + "valid_targets_min": 5101 + }, + { + "epoch": 0.1178627553693033, + "grad_norm": 0.5889138833486921, + "learning_rate": 6.701570680628273e-06, + "loss": 0.2933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12677639722824097, + "step": 225, + "valid_targets_mean": 5299.5, + "valid_targets_min": 4442 + }, + { + "epoch": 0.12048192771084337, + "grad_norm": 0.536560813199085, + "learning_rate": 6.851159311892297e-06, + "loss": 0.2576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11181782186031342, + "step": 230, + "valid_targets_mean": 6133.9, + "valid_targets_min": 5130 + }, + { + "epoch": 0.12310110005238345, + "grad_norm": 0.5038819089276341, + "learning_rate": 7.0007479431563205e-06, + "loss": 0.2606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14731335639953613, + "step": 235, + "valid_targets_mean": 6477.4, + "valid_targets_min": 4162 + }, + { + "epoch": 0.12572027239392353, + "grad_norm": 0.5128667474576506, + "learning_rate": 7.150336574420344e-06, + "loss": 0.2669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12645837664604187, + "step": 240, + "valid_targets_mean": 6350.4, + "valid_targets_min": 3940 + }, + { + "epoch": 0.1283394447354636, + "grad_norm": 0.5657169458291228, + "learning_rate": 7.299925205684369e-06, + "loss": 0.276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15734316408634186, + "step": 245, + "valid_targets_mean": 6625.1, + "valid_targets_min": 4642 + }, + { + "epoch": 0.13095861707700368, + "grad_norm": 0.6659478810666077, + "learning_rate": 7.4495138369483925e-06, + "loss": 0.3107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1492205262184143, + "step": 250, + "valid_targets_mean": 4907.0, + "valid_targets_min": 4431 + }, + { + "epoch": 0.13357778941854373, + "grad_norm": 0.5155411725495067, + "learning_rate": 7.599102468212416e-06, + "loss": 0.3168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14482495188713074, + "step": 255, + "valid_targets_mean": 6465.0, + "valid_targets_min": 5333 + }, + { + "epoch": 0.1361969617600838, + "grad_norm": 0.5971339718848779, + "learning_rate": 7.748691099476442e-06, + "loss": 0.2668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1400701403617859, + "step": 260, + "valid_targets_mean": 5650.5, + "valid_targets_min": 4769 + }, + { + "epoch": 0.13881613410162388, + "grad_norm": 0.6737651722398803, + "learning_rate": 7.898279730740465e-06, + "loss": 0.2758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13320739567279816, + "step": 265, + "valid_targets_mean": 5540.0, + "valid_targets_min": 4281 + }, + { + "epoch": 0.14143530644316396, + "grad_norm": 0.5472662877185434, + "learning_rate": 8.047868362004489e-06, + "loss": 0.2782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18294978141784668, + "step": 270, + "valid_targets_mean": 6499.0, + "valid_targets_min": 4844 + }, + { + "epoch": 0.14405447878470404, + "grad_norm": 0.5032124689003608, + "learning_rate": 8.197456993268512e-06, + "loss": 0.2925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11865824460983276, + "step": 275, + "valid_targets_mean": 5736.5, + "valid_targets_min": 5227 + }, + { + "epoch": 0.14667365112624411, + "grad_norm": 0.4496234573162473, + "learning_rate": 8.347045624532537e-06, + "loss": 0.2662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10675523430109024, + "step": 280, + "valid_targets_mean": 5857.2, + "valid_targets_min": 2461 + }, + { + "epoch": 0.1492928234677842, + "grad_norm": 0.49627410303157343, + "learning_rate": 8.49663425579656e-06, + "loss": 0.2846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12157833576202393, + "step": 285, + "valid_targets_mean": 5793.4, + "valid_targets_min": 2815 + }, + { + "epoch": 0.15191199580932424, + "grad_norm": 0.4792114014293663, + "learning_rate": 8.646222887060584e-06, + "loss": 0.2725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.116509810090065, + "step": 290, + "valid_targets_mean": 6588.8, + "valid_targets_min": 4875 + }, + { + "epoch": 0.15453116815086432, + "grad_norm": 0.5084162945631525, + "learning_rate": 8.795811518324609e-06, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12259118258953094, + "step": 295, + "valid_targets_mean": 5984.8, + "valid_targets_min": 4901 + }, + { + "epoch": 0.1571503404924044, + "grad_norm": 0.45601474003248577, + "learning_rate": 8.945400149588632e-06, + "loss": 0.2663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12021372467279434, + "step": 300, + "valid_targets_mean": 6067.4, + "valid_targets_min": 3786 + }, + { + "epoch": 0.15976951283394447, + "grad_norm": 0.5087080401637396, + "learning_rate": 9.094988780852656e-06, + "loss": 0.3206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13261842727661133, + "step": 305, + "valid_targets_mean": 5852.5, + "valid_targets_min": 5057 + }, + { + "epoch": 0.16238868517548455, + "grad_norm": 0.5015109972208805, + "learning_rate": 9.244577412116679e-06, + "loss": 0.2539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14095407724380493, + "step": 310, + "valid_targets_mean": 6362.2, + "valid_targets_min": 3952 + }, + { + "epoch": 0.16500785751702463, + "grad_norm": 0.6047404492813938, + "learning_rate": 9.394166043380704e-06, + "loss": 0.2717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1349816620349884, + "step": 315, + "valid_targets_mean": 5360.2, + "valid_targets_min": 4267 + }, + { + "epoch": 0.1676270298585647, + "grad_norm": 0.5949556812864151, + "learning_rate": 9.543754674644728e-06, + "loss": 0.2794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1553954780101776, + "step": 320, + "valid_targets_mean": 6842.1, + "valid_targets_min": 4785 + }, + { + "epoch": 0.17024620220010478, + "grad_norm": 0.5341288285189475, + "learning_rate": 9.693343305908753e-06, + "loss": 0.245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11958827078342438, + "step": 325, + "valid_targets_mean": 6188.5, + "valid_targets_min": 5438 + }, + { + "epoch": 0.17286537454164483, + "grad_norm": 0.5388080593679015, + "learning_rate": 9.842931937172776e-06, + "loss": 0.2511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13774797320365906, + "step": 330, + "valid_targets_mean": 5848.6, + "valid_targets_min": 4734 + }, + { + "epoch": 0.1754845468831849, + "grad_norm": 0.485256593875416, + "learning_rate": 9.9925205684368e-06, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10812049359083176, + "step": 335, + "valid_targets_mean": 6278.2, + "valid_targets_min": 5636 + }, + { + "epoch": 0.17810371922472498, + "grad_norm": 0.493388223698597, + "learning_rate": 1.0142109199700823e-05, + "loss": 0.2599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1128229945898056, + "step": 340, + "valid_targets_mean": 6168.9, + "valid_targets_min": 4206 + }, + { + "epoch": 0.18072289156626506, + "grad_norm": 0.49568917581337374, + "learning_rate": 1.029169783096485e-05, + "loss": 0.25, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11211951076984406, + "step": 345, + "valid_targets_mean": 6226.5, + "valid_targets_min": 4535 + }, + { + "epoch": 0.18334206390780514, + "grad_norm": 0.4333049424875618, + "learning_rate": 1.0441286462228872e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12328267842531204, + "step": 350, + "valid_targets_mean": 7821.6, + "valid_targets_min": 5471 + }, + { + "epoch": 0.1859612362493452, + "grad_norm": 0.4907703671440544, + "learning_rate": 1.0590875093492897e-05, + "loss": 0.2583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10935990512371063, + "step": 355, + "valid_targets_mean": 6843.0, + "valid_targets_min": 5243 + }, + { + "epoch": 0.1885804085908853, + "grad_norm": 0.5624863264925903, + "learning_rate": 1.074046372475692e-05, + "loss": 0.2387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1467284858226776, + "step": 360, + "valid_targets_mean": 6658.6, + "valid_targets_min": 4906 + }, + { + "epoch": 0.19119958093242537, + "grad_norm": 0.4817414573035044, + "learning_rate": 1.0890052356020944e-05, + "loss": 0.2666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13065621256828308, + "step": 365, + "valid_targets_mean": 6799.5, + "valid_targets_min": 5168 + }, + { + "epoch": 0.19381875327396542, + "grad_norm": 0.7573628453485103, + "learning_rate": 1.1039640987284967e-05, + "loss": 0.2707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1387932151556015, + "step": 370, + "valid_targets_mean": 5376.2, + "valid_targets_min": 3130 + }, + { + "epoch": 0.1964379256155055, + "grad_norm": 1.7677671707709737, + "learning_rate": 1.1189229618548992e-05, + "loss": 0.3343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2786247730255127, + "step": 375, + "valid_targets_mean": 2226.6, + "valid_targets_min": 1111 + }, + { + "epoch": 0.19905709795704557, + "grad_norm": 1.3509938172860045, + "learning_rate": 1.1338818249813015e-05, + "loss": 0.3878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21497374773025513, + "step": 380, + "valid_targets_mean": 1648.0, + "valid_targets_min": 929 + }, + { + "epoch": 0.20167627029858565, + "grad_norm": 1.1220666241457584, + "learning_rate": 1.148840688107704e-05, + "loss": 0.3533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13884077966213226, + "step": 385, + "valid_targets_mean": 1430.0, + "valid_targets_min": 914 + }, + { + "epoch": 0.20429544264012572, + "grad_norm": 1.1543784447211902, + "learning_rate": 1.1637995512341062e-05, + "loss": 0.3368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14594638347625732, + "step": 390, + "valid_targets_mean": 1400.1, + "valid_targets_min": 684 + }, + { + "epoch": 0.2069146149816658, + "grad_norm": 0.9812725692414712, + "learning_rate": 1.1787584143605087e-05, + "loss": 0.3303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1464192420244217, + "step": 395, + "valid_targets_mean": 1884.6, + "valid_targets_min": 876 + }, + { + "epoch": 0.20953378732320588, + "grad_norm": 1.1161632290929437, + "learning_rate": 1.193717277486911e-05, + "loss": 0.3171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1372755467891693, + "step": 400, + "valid_targets_mean": 1341.0, + "valid_targets_min": 772 + }, + { + "epoch": 0.21215295966474593, + "grad_norm": 1.401680262070522, + "learning_rate": 1.2086761406133134e-05, + "loss": 0.323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13547280430793762, + "step": 405, + "valid_targets_mean": 1462.6, + "valid_targets_min": 655 + }, + { + "epoch": 0.214772132006286, + "grad_norm": 1.085784992314865, + "learning_rate": 1.2236350037397157e-05, + "loss": 0.2929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1361703872680664, + "step": 410, + "valid_targets_mean": 1407.8, + "valid_targets_min": 737 + }, + { + "epoch": 0.21739130434782608, + "grad_norm": 1.1045826898282667, + "learning_rate": 1.2385938668661183e-05, + "loss": 0.3113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1277555376291275, + "step": 415, + "valid_targets_mean": 1260.0, + "valid_targets_min": 759 + }, + { + "epoch": 0.22001047668936616, + "grad_norm": 1.0709020827113547, + "learning_rate": 1.2535527299925208e-05, + "loss": 0.2916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11939826607704163, + "step": 420, + "valid_targets_mean": 1251.4, + "valid_targets_min": 750 + }, + { + "epoch": 0.22262964903090623, + "grad_norm": 0.9472314691409904, + "learning_rate": 1.268511593118923e-05, + "loss": 0.2873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1769285947084427, + "step": 425, + "valid_targets_mean": 2022.1, + "valid_targets_min": 927 + }, + { + "epoch": 0.2252488213724463, + "grad_norm": 1.1122885229634296, + "learning_rate": 1.2834704562453255e-05, + "loss": 0.3009, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1492638885974884, + "step": 430, + "valid_targets_mean": 1364.9, + "valid_targets_min": 679 + }, + { + "epoch": 0.2278679937139864, + "grad_norm": 1.16549389381539, + "learning_rate": 1.2984293193717278e-05, + "loss": 0.2878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13810691237449646, + "step": 435, + "valid_targets_mean": 1375.4, + "valid_targets_min": 771 + }, + { + "epoch": 0.23048716605552647, + "grad_norm": 0.9649070587070621, + "learning_rate": 1.3133881824981303e-05, + "loss": 0.2747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.164086252450943, + "step": 440, + "valid_targets_mean": 1776.8, + "valid_targets_min": 878 + }, + { + "epoch": 0.23310633839706651, + "grad_norm": 0.9924225243646162, + "learning_rate": 1.3283470456245326e-05, + "loss": 0.269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12173327803611755, + "step": 445, + "valid_targets_mean": 1513.5, + "valid_targets_min": 1103 + }, + { + "epoch": 0.2357255107386066, + "grad_norm": 1.060335409806951, + "learning_rate": 1.343305908750935e-05, + "loss": 0.2827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15946930646896362, + "step": 450, + "valid_targets_mean": 1680.0, + "valid_targets_min": 821 + }, + { + "epoch": 0.23834468308014667, + "grad_norm": 0.9604012025511408, + "learning_rate": 1.3582647718773375e-05, + "loss": 0.2791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17137695848941803, + "step": 455, + "valid_targets_mean": 1960.6, + "valid_targets_min": 974 + }, + { + "epoch": 0.24096385542168675, + "grad_norm": 1.0761468003856411, + "learning_rate": 1.3732236350037398e-05, + "loss": 0.2849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13498663902282715, + "step": 460, + "valid_targets_mean": 1495.6, + "valid_targets_min": 869 + }, + { + "epoch": 0.24358302776322682, + "grad_norm": 1.007730248271164, + "learning_rate": 1.3881824981301422e-05, + "loss": 0.2814, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11989525705575943, + "step": 465, + "valid_targets_mean": 1296.5, + "valid_targets_min": 1127 + }, + { + "epoch": 0.2462022001047669, + "grad_norm": 1.013738571239438, + "learning_rate": 1.4031413612565445e-05, + "loss": 0.2808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.144391268491745, + "step": 470, + "valid_targets_mean": 1711.9, + "valid_targets_min": 709 + }, + { + "epoch": 0.24882137244630698, + "grad_norm": 1.0852102635913559, + "learning_rate": 1.418100224382947e-05, + "loss": 0.2707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12370958924293518, + "step": 475, + "valid_targets_mean": 1241.1, + "valid_targets_min": 720 + }, + { + "epoch": 0.25144054478784705, + "grad_norm": 1.0460067253548269, + "learning_rate": 1.4330590875093493e-05, + "loss": 0.2803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13514494895935059, + "step": 480, + "valid_targets_mean": 1343.6, + "valid_targets_min": 662 + }, + { + "epoch": 0.25405971712938713, + "grad_norm": 0.9489381493243331, + "learning_rate": 1.4480179506357519e-05, + "loss": 0.2848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10587510466575623, + "step": 485, + "valid_targets_mean": 1273.8, + "valid_targets_min": 859 + }, + { + "epoch": 0.2566788894709272, + "grad_norm": 0.9961585221733041, + "learning_rate": 1.4629768137621543e-05, + "loss": 0.2568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1385517567396164, + "step": 490, + "valid_targets_mean": 1475.9, + "valid_targets_min": 950 + }, + { + "epoch": 0.2592980618124673, + "grad_norm": 1.1072634145215436, + "learning_rate": 1.4779356768885566e-05, + "loss": 0.2685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13932949304580688, + "step": 495, + "valid_targets_mean": 1643.9, + "valid_targets_min": 686 + }, + { + "epoch": 0.26191723415400736, + "grad_norm": 0.9457831357787362, + "learning_rate": 1.4928945400149591e-05, + "loss": 0.2628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1233171820640564, + "step": 500, + "valid_targets_mean": 1486.2, + "valid_targets_min": 702 + }, + { + "epoch": 0.2645364064955474, + "grad_norm": 1.0712290060121283, + "learning_rate": 1.5078534031413614e-05, + "loss": 0.2588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10660302639007568, + "step": 505, + "valid_targets_mean": 1281.2, + "valid_targets_min": 779 + }, + { + "epoch": 0.26715557883708746, + "grad_norm": 1.3418306917175586, + "learning_rate": 1.5228122662677638e-05, + "loss": 0.2766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1407930552959442, + "step": 510, + "valid_targets_mean": 1427.4, + "valid_targets_min": 822 + }, + { + "epoch": 0.26977475117862754, + "grad_norm": 1.0589259309371697, + "learning_rate": 1.5377711293941663e-05, + "loss": 0.2509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09778021275997162, + "step": 515, + "valid_targets_mean": 1262.0, + "valid_targets_min": 868 + }, + { + "epoch": 0.2723939235201676, + "grad_norm": 1.1075285776320232, + "learning_rate": 1.5527299925205686e-05, + "loss": 0.2635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11571861058473587, + "step": 520, + "valid_targets_mean": 1303.9, + "valid_targets_min": 765 + }, + { + "epoch": 0.2750130958617077, + "grad_norm": 1.018588764343178, + "learning_rate": 1.567688855646971e-05, + "loss": 0.2679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13324031233787537, + "step": 525, + "valid_targets_mean": 1463.9, + "valid_targets_min": 793 + }, + { + "epoch": 0.27763226820324777, + "grad_norm": 1.0394181680078727, + "learning_rate": 1.582647718773373e-05, + "loss": 0.2687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17844894528388977, + "step": 530, + "valid_targets_mean": 1570.0, + "valid_targets_min": 857 + }, + { + "epoch": 0.28025144054478784, + "grad_norm": 0.981851121648432, + "learning_rate": 1.5976065818997758e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11975729465484619, + "step": 535, + "valid_targets_mean": 1509.6, + "valid_targets_min": 1219 + }, + { + "epoch": 0.2828706128863279, + "grad_norm": 1.0699996104108067, + "learning_rate": 1.612565445026178e-05, + "loss": 0.2585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11243554204702377, + "step": 540, + "valid_targets_mean": 1448.9, + "valid_targets_min": 660 + }, + { + "epoch": 0.285489785227868, + "grad_norm": 0.9522677850276425, + "learning_rate": 1.6275243081525804e-05, + "loss": 0.238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11272923648357391, + "step": 545, + "valid_targets_mean": 1586.9, + "valid_targets_min": 853 + }, + { + "epoch": 0.2881089575694081, + "grad_norm": 1.071239086885367, + "learning_rate": 1.6424831712789827e-05, + "loss": 0.257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13065925240516663, + "step": 550, + "valid_targets_mean": 1853.2, + "valid_targets_min": 890 + }, + { + "epoch": 0.29072812991094815, + "grad_norm": 0.9429979695434983, + "learning_rate": 1.6574420344053853e-05, + "loss": 0.2621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0913005992770195, + "step": 555, + "valid_targets_mean": 1249.4, + "valid_targets_min": 765 + }, + { + "epoch": 0.29334730225248823, + "grad_norm": 0.9290035481246481, + "learning_rate": 1.672400897531788e-05, + "loss": 0.2551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14317075908184052, + "step": 560, + "valid_targets_mean": 1580.0, + "valid_targets_min": 727 + }, + { + "epoch": 0.2959664745940283, + "grad_norm": 1.0058903370501906, + "learning_rate": 1.6873597606581902e-05, + "loss": 0.2515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10938141494989395, + "step": 565, + "valid_targets_mean": 1220.2, + "valid_targets_min": 722 + }, + { + "epoch": 0.2985856469355684, + "grad_norm": 1.1542713780035165, + "learning_rate": 1.7023186237845925e-05, + "loss": 0.2635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10995220392942429, + "step": 570, + "valid_targets_mean": 1463.0, + "valid_targets_min": 760 + }, + { + "epoch": 0.30120481927710846, + "grad_norm": 1.0440417663519497, + "learning_rate": 1.7172774869109948e-05, + "loss": 0.2444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10970985889434814, + "step": 575, + "valid_targets_mean": 1404.2, + "valid_targets_min": 1046 + }, + { + "epoch": 0.3038239916186485, + "grad_norm": 0.9600703541682362, + "learning_rate": 1.7322363500373974e-05, + "loss": 0.2663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1116146594285965, + "step": 580, + "valid_targets_mean": 1383.8, + "valid_targets_min": 692 + }, + { + "epoch": 0.30644316396018856, + "grad_norm": 0.9330677848729542, + "learning_rate": 1.7471952131637997e-05, + "loss": 0.2597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10352496802806854, + "step": 585, + "valid_targets_mean": 1496.4, + "valid_targets_min": 907 + }, + { + "epoch": 0.30906233630172864, + "grad_norm": 0.9514194457771882, + "learning_rate": 1.762154076290202e-05, + "loss": 0.2721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12670855224132538, + "step": 590, + "valid_targets_mean": 1602.6, + "valid_targets_min": 1214 + }, + { + "epoch": 0.3116815086432687, + "grad_norm": 0.91083422937553, + "learning_rate": 1.7771129394166046e-05, + "loss": 0.2478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10771577060222626, + "step": 595, + "valid_targets_mean": 1383.2, + "valid_targets_min": 822 + }, + { + "epoch": 0.3143006809848088, + "grad_norm": 1.0102441365733255, + "learning_rate": 1.792071802543007e-05, + "loss": 0.2477, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12448005378246307, + "step": 600, + "valid_targets_mean": 1601.6, + "valid_targets_min": 982 + }, + { + "epoch": 0.31691985332634887, + "grad_norm": 0.9213996885799169, + "learning_rate": 1.8070306656694092e-05, + "loss": 0.2588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09866012632846832, + "step": 605, + "valid_targets_mean": 1240.2, + "valid_targets_min": 713 + }, + { + "epoch": 0.31953902566788894, + "grad_norm": 1.2634906328961235, + "learning_rate": 1.8219895287958115e-05, + "loss": 0.2491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1237337589263916, + "step": 610, + "valid_targets_mean": 1525.5, + "valid_targets_min": 721 + }, + { + "epoch": 0.322158198009429, + "grad_norm": 0.9986611011316331, + "learning_rate": 1.836948391922214e-05, + "loss": 0.2429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10856451094150543, + "step": 615, + "valid_targets_mean": 1275.1, + "valid_targets_min": 741 + }, + { + "epoch": 0.3247773703509691, + "grad_norm": 0.9488586011093412, + "learning_rate": 1.8519072550486164e-05, + "loss": 0.2489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11882676184177399, + "step": 620, + "valid_targets_mean": 1554.5, + "valid_targets_min": 734 + }, + { + "epoch": 0.3273965426925092, + "grad_norm": 0.8383539242686089, + "learning_rate": 1.866866118175019e-05, + "loss": 0.2542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12229125201702118, + "step": 625, + "valid_targets_mean": 1809.5, + "valid_targets_min": 744 + }, + { + "epoch": 0.33001571503404925, + "grad_norm": 0.9247633651447293, + "learning_rate": 1.8818249813014213e-05, + "loss": 0.251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14622409641742706, + "step": 630, + "valid_targets_mean": 1836.4, + "valid_targets_min": 1295 + }, + { + "epoch": 0.3326348873755893, + "grad_norm": 0.9402880242528865, + "learning_rate": 1.8967838444278236e-05, + "loss": 0.2414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12124080955982208, + "step": 635, + "valid_targets_mean": 1427.8, + "valid_targets_min": 785 + }, + { + "epoch": 0.3352540597171294, + "grad_norm": 0.8314190621340193, + "learning_rate": 1.9117427075542262e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11245155334472656, + "step": 640, + "valid_targets_mean": 1787.8, + "valid_targets_min": 740 + }, + { + "epoch": 0.3378732320586695, + "grad_norm": 1.0286199098709077, + "learning_rate": 1.9267015706806285e-05, + "loss": 0.2431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09985199570655823, + "step": 645, + "valid_targets_mean": 1259.0, + "valid_targets_min": 711 + }, + { + "epoch": 0.34049240440020956, + "grad_norm": 0.9451930394161755, + "learning_rate": 1.9416604338070308e-05, + "loss": 0.2455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1365668624639511, + "step": 650, + "valid_targets_mean": 1798.9, + "valid_targets_min": 1122 + }, + { + "epoch": 0.3431115767417496, + "grad_norm": 1.0130274395862757, + "learning_rate": 1.956619296933433e-05, + "loss": 0.2428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1318720281124115, + "step": 655, + "valid_targets_mean": 1666.1, + "valid_targets_min": 875 + }, + { + "epoch": 0.34573074908328966, + "grad_norm": 0.962950145719036, + "learning_rate": 1.9715781600598357e-05, + "loss": 0.2285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.104540154337883, + "step": 660, + "valid_targets_mean": 1226.6, + "valid_targets_min": 567 + }, + { + "epoch": 0.34834992142482973, + "grad_norm": 0.9165994912418771, + "learning_rate": 1.986537023186238e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09978640079498291, + "step": 665, + "valid_targets_mean": 1456.9, + "valid_targets_min": 912 + }, + { + "epoch": 0.3509690937663698, + "grad_norm": 0.9159997313653303, + "learning_rate": 2.0014958863126406e-05, + "loss": 0.248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1311609148979187, + "step": 670, + "valid_targets_mean": 1766.4, + "valid_targets_min": 1277 + }, + { + "epoch": 0.3535882661079099, + "grad_norm": 0.9782152476683862, + "learning_rate": 2.016454749439043e-05, + "loss": 0.2309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10838239639997482, + "step": 675, + "valid_targets_mean": 1377.2, + "valid_targets_min": 773 + }, + { + "epoch": 0.35620743844944996, + "grad_norm": 0.9787824735333043, + "learning_rate": 2.0314136125654452e-05, + "loss": 0.2462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11166463792324066, + "step": 680, + "valid_targets_mean": 1212.6, + "valid_targets_min": 679 + }, + { + "epoch": 0.35882661079099004, + "grad_norm": 0.951410336343893, + "learning_rate": 2.046372475691848e-05, + "loss": 0.2412, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10627752542495728, + "step": 685, + "valid_targets_mean": 1523.2, + "valid_targets_min": 818 + }, + { + "epoch": 0.3614457831325301, + "grad_norm": 0.9785405331704724, + "learning_rate": 2.06133133881825e-05, + "loss": 0.2432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1076476126909256, + "step": 690, + "valid_targets_mean": 1375.5, + "valid_targets_min": 824 + }, + { + "epoch": 0.3640649554740702, + "grad_norm": 1.0134487896680853, + "learning_rate": 2.0762902019446524e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10936789214611053, + "step": 695, + "valid_targets_mean": 1494.4, + "valid_targets_min": 592 + }, + { + "epoch": 0.3666841278156103, + "grad_norm": 1.1151965488504085, + "learning_rate": 2.0912490650710547e-05, + "loss": 0.243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13315948843955994, + "step": 700, + "valid_targets_mean": 1652.6, + "valid_targets_min": 789 + }, + { + "epoch": 0.36930330015715035, + "grad_norm": 1.0270773736389838, + "learning_rate": 2.1062079281974573e-05, + "loss": 0.2453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10341659188270569, + "step": 705, + "valid_targets_mean": 1337.6, + "valid_targets_min": 622 + }, + { + "epoch": 0.3719224724986904, + "grad_norm": 1.2264238687733724, + "learning_rate": 2.1211667913238596e-05, + "loss": 0.2487, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11207322776317596, + "step": 710, + "valid_targets_mean": 1221.4, + "valid_targets_min": 625 + }, + { + "epoch": 0.3745416448402305, + "grad_norm": 0.9658206690648851, + "learning_rate": 2.136125654450262e-05, + "loss": 0.2531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10400964319705963, + "step": 715, + "valid_targets_mean": 1695.1, + "valid_targets_min": 888 + }, + { + "epoch": 0.3771608171817706, + "grad_norm": 0.9055245882414871, + "learning_rate": 2.1510845175766645e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13797059655189514, + "step": 720, + "valid_targets_mean": 1831.0, + "valid_targets_min": 635 + }, + { + "epoch": 0.37977998952331066, + "grad_norm": 1.0608985591659448, + "learning_rate": 2.1660433807030668e-05, + "loss": 0.2505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15078774094581604, + "step": 725, + "valid_targets_mean": 1643.6, + "valid_targets_min": 522 + }, + { + "epoch": 0.38239916186485073, + "grad_norm": 3.938139942253861, + "learning_rate": 2.181002243829469e-05, + "loss": 0.2352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10994037240743637, + "step": 730, + "valid_targets_mean": 1264.1, + "valid_targets_min": 697 + }, + { + "epoch": 0.38501833420639076, + "grad_norm": 0.9383536213597325, + "learning_rate": 2.1959611069558714e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13887813687324524, + "step": 735, + "valid_targets_mean": 1721.8, + "valid_targets_min": 868 + }, + { + "epoch": 0.38763750654793083, + "grad_norm": 1.0184466793175806, + "learning_rate": 2.210919970082274e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1270817518234253, + "step": 740, + "valid_targets_mean": 1489.8, + "valid_targets_min": 769 + }, + { + "epoch": 0.3902566788894709, + "grad_norm": 1.1638306909939866, + "learning_rate": 2.2258788332086763e-05, + "loss": 0.243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10942777246236801, + "step": 745, + "valid_targets_mean": 1321.8, + "valid_targets_min": 745 + }, + { + "epoch": 0.392875851231011, + "grad_norm": 1.228651382781768, + "learning_rate": 2.2408376963350786e-05, + "loss": 0.2359, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11496207118034363, + "step": 750, + "valid_targets_mean": 1332.2, + "valid_targets_min": 680 + }, + { + "epoch": 0.39549502357255106, + "grad_norm": 0.8578027015892162, + "learning_rate": 2.255796559461481e-05, + "loss": 0.2304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1369241178035736, + "step": 755, + "valid_targets_mean": 1908.4, + "valid_targets_min": 1347 + }, + { + "epoch": 0.39811419591409114, + "grad_norm": 1.032952570362794, + "learning_rate": 2.2707554225878835e-05, + "loss": 0.2468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13751846551895142, + "step": 760, + "valid_targets_mean": 1539.8, + "valid_targets_min": 851 + }, + { + "epoch": 0.4007333682556312, + "grad_norm": 0.9334342392521109, + "learning_rate": 2.2857142857142858e-05, + "loss": 0.2321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10404945909976959, + "step": 765, + "valid_targets_mean": 1462.5, + "valid_targets_min": 903 + }, + { + "epoch": 0.4033525405971713, + "grad_norm": 0.9702744762163089, + "learning_rate": 2.300673148840688e-05, + "loss": 0.2296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13138015568256378, + "step": 770, + "valid_targets_mean": 1629.5, + "valid_targets_min": 848 + }, + { + "epoch": 0.40597171293871137, + "grad_norm": 0.9193640198585356, + "learning_rate": 2.3156320119670904e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0928211361169815, + "step": 775, + "valid_targets_mean": 1626.6, + "valid_targets_min": 702 + }, + { + "epoch": 0.40859088528025145, + "grad_norm": 1.2326202460530613, + "learning_rate": 2.330590875093493e-05, + "loss": 0.2144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10401280224323273, + "step": 780, + "valid_targets_mean": 1578.0, + "valid_targets_min": 957 + }, + { + "epoch": 0.4112100576217915, + "grad_norm": 0.8744813914544552, + "learning_rate": 2.3455497382198953e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13442571461200714, + "step": 785, + "valid_targets_mean": 1799.2, + "valid_targets_min": 1377 + }, + { + "epoch": 0.4138292299633316, + "grad_norm": 0.8521216445027406, + "learning_rate": 2.3605086013462976e-05, + "loss": 0.2334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12470878660678864, + "step": 790, + "valid_targets_mean": 1747.6, + "valid_targets_min": 870 + }, + { + "epoch": 0.4164484023048717, + "grad_norm": 0.9878122126499209, + "learning_rate": 2.3754674644727006e-05, + "loss": 0.2371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12663641571998596, + "step": 795, + "valid_targets_mean": 1658.8, + "valid_targets_min": 822 + }, + { + "epoch": 0.41906757464641176, + "grad_norm": 0.9483420963619904, + "learning_rate": 2.390426327599103e-05, + "loss": 0.2397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10291539132595062, + "step": 800, + "valid_targets_mean": 1458.6, + "valid_targets_min": 873 + }, + { + "epoch": 0.42168674698795183, + "grad_norm": 1.1146280374790785, + "learning_rate": 2.405385190725505e-05, + "loss": 0.2298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10151064395904541, + "step": 805, + "valid_targets_mean": 1291.1, + "valid_targets_min": 591 + }, + { + "epoch": 0.42430591932949185, + "grad_norm": 2.012323624599317, + "learning_rate": 2.4203440538519078e-05, + "loss": 0.245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1058199554681778, + "step": 810, + "valid_targets_mean": 1378.5, + "valid_targets_min": 1066 + }, + { + "epoch": 0.42692509167103193, + "grad_norm": 0.7432971467787244, + "learning_rate": 2.43530291697831e-05, + "loss": 0.3427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17605727910995483, + "step": 815, + "valid_targets_mean": 5546.2, + "valid_targets_min": 1971 + }, + { + "epoch": 0.429544264012572, + "grad_norm": 0.736489891543121, + "learning_rate": 2.4502617801047123e-05, + "loss": 0.4077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2549963593482971, + "step": 820, + "valid_targets_mean": 5648.9, + "valid_targets_min": 1631 + }, + { + "epoch": 0.4321634363541121, + "grad_norm": 0.6488197312143061, + "learning_rate": 2.4652206432311146e-05, + "loss": 0.3955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.23194031417369843, + "step": 825, + "valid_targets_mean": 5411.4, + "valid_targets_min": 1509 + }, + { + "epoch": 0.43478260869565216, + "grad_norm": 0.8265034834707014, + "learning_rate": 2.4801795063575172e-05, + "loss": 0.3747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17416730523109436, + "step": 830, + "valid_targets_mean": 4960.2, + "valid_targets_min": 2736 + }, + { + "epoch": 0.43740178103719224, + "grad_norm": 0.6225699372813742, + "learning_rate": 2.4951383694839195e-05, + "loss": 0.3772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1611279547214508, + "step": 835, + "valid_targets_mean": 3726.8, + "valid_targets_min": 1439 + }, + { + "epoch": 0.4400209533787323, + "grad_norm": 0.6257398485834125, + "learning_rate": 2.5100972326103218e-05, + "loss": 0.3884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17108386754989624, + "step": 840, + "valid_targets_mean": 4269.4, + "valid_targets_min": 1838 + }, + { + "epoch": 0.4426401257202724, + "grad_norm": 0.5586100319766503, + "learning_rate": 2.525056095736724e-05, + "loss": 0.3578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22302839159965515, + "step": 845, + "valid_targets_mean": 5973.6, + "valid_targets_min": 2762 + }, + { + "epoch": 0.44525929806181247, + "grad_norm": 0.5631695158406826, + "learning_rate": 2.5400149588631267e-05, + "loss": 0.3681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1688145101070404, + "step": 850, + "valid_targets_mean": 3987.1, + "valid_targets_min": 1654 + }, + { + "epoch": 0.44787847040335255, + "grad_norm": 0.514248956356746, + "learning_rate": 2.554973821989529e-05, + "loss": 0.3677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16678676009178162, + "step": 855, + "valid_targets_mean": 5043.6, + "valid_targets_min": 2493 + }, + { + "epoch": 0.4504976427448926, + "grad_norm": 0.6066070080618734, + "learning_rate": 2.5699326851159313e-05, + "loss": 0.3797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1756962090730667, + "step": 860, + "valid_targets_mean": 4468.9, + "valid_targets_min": 874 + }, + { + "epoch": 0.4531168150864327, + "grad_norm": 0.5495071836081291, + "learning_rate": 2.584891548242334e-05, + "loss": 0.3636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1663469672203064, + "step": 865, + "valid_targets_mean": 5241.0, + "valid_targets_min": 953 + }, + { + "epoch": 0.4557359874279728, + "grad_norm": 0.5667214495528212, + "learning_rate": 2.5998504113687362e-05, + "loss": 0.37, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22836129367351532, + "step": 870, + "valid_targets_mean": 5277.0, + "valid_targets_min": 1265 + }, + { + "epoch": 0.45835515976951285, + "grad_norm": 0.6095241984355673, + "learning_rate": 2.6148092744951385e-05, + "loss": 0.3603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1914677917957306, + "step": 875, + "valid_targets_mean": 6127.8, + "valid_targets_min": 2566 + }, + { + "epoch": 0.46097433211105293, + "grad_norm": 0.5843240842276783, + "learning_rate": 2.6297681376215408e-05, + "loss": 0.3676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16033320128917694, + "step": 880, + "valid_targets_mean": 3446.9, + "valid_targets_min": 1759 + }, + { + "epoch": 0.463593504452593, + "grad_norm": 0.6730530597439103, + "learning_rate": 2.6447270007479434e-05, + "loss": 0.3718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18090486526489258, + "step": 885, + "valid_targets_mean": 3659.4, + "valid_targets_min": 2179 + }, + { + "epoch": 0.46621267679413303, + "grad_norm": 0.6481767994741637, + "learning_rate": 2.6596858638743457e-05, + "loss": 0.3731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19510102272033691, + "step": 890, + "valid_targets_mean": 4161.5, + "valid_targets_min": 1785 + }, + { + "epoch": 0.4688318491356731, + "grad_norm": 0.6380101616555064, + "learning_rate": 2.674644727000748e-05, + "loss": 0.3644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21468541026115417, + "step": 895, + "valid_targets_mean": 4467.9, + "valid_targets_min": 2345 + }, + { + "epoch": 0.4714510214772132, + "grad_norm": 0.6145570469817694, + "learning_rate": 2.6896035901271503e-05, + "loss": 0.3801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20558148622512817, + "step": 900, + "valid_targets_mean": 4057.2, + "valid_targets_min": 2767 + }, + { + "epoch": 0.47407019381875326, + "grad_norm": 0.6664558223626487, + "learning_rate": 2.704562453253553e-05, + "loss": 0.3807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19282910227775574, + "step": 905, + "valid_targets_mean": 3450.2, + "valid_targets_min": 1205 + }, + { + "epoch": 0.47668936616029334, + "grad_norm": 0.6039929888544922, + "learning_rate": 2.7195213163799552e-05, + "loss": 0.3695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17805667221546173, + "step": 910, + "valid_targets_mean": 3777.1, + "valid_targets_min": 2004 + }, + { + "epoch": 0.4793085385018334, + "grad_norm": 0.5844771100466323, + "learning_rate": 2.7344801795063575e-05, + "loss": 0.3747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19960585236549377, + "step": 915, + "valid_targets_mean": 4273.0, + "valid_targets_min": 1469 + }, + { + "epoch": 0.4819277108433735, + "grad_norm": 0.7048003467473493, + "learning_rate": 2.7494390426327598e-05, + "loss": 0.3535, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22104474902153015, + "step": 920, + "valid_targets_mean": 4196.8, + "valid_targets_min": 1889 + }, + { + "epoch": 0.48454688318491357, + "grad_norm": 0.6089146024786963, + "learning_rate": 2.7643979057591624e-05, + "loss": 0.3473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18689574301242828, + "step": 925, + "valid_targets_mean": 4319.2, + "valid_targets_min": 2046 + }, + { + "epoch": 0.48716605552645365, + "grad_norm": 0.6322409136376674, + "learning_rate": 2.7793567688855647e-05, + "loss": 0.3514, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21273566782474518, + "step": 930, + "valid_targets_mean": 4580.1, + "valid_targets_min": 3373 + }, + { + "epoch": 0.4897852278679937, + "grad_norm": 0.6533652551169314, + "learning_rate": 2.7943156320119677e-05, + "loss": 0.3667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18637657165527344, + "step": 935, + "valid_targets_mean": 3786.0, + "valid_targets_min": 1747 + }, + { + "epoch": 0.4924044002095338, + "grad_norm": 0.6330535063927669, + "learning_rate": 2.80927449513837e-05, + "loss": 0.3654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16132210195064545, + "step": 940, + "valid_targets_mean": 3339.9, + "valid_targets_min": 1166 + }, + { + "epoch": 0.4950235725510739, + "grad_norm": 0.6028755929755526, + "learning_rate": 2.8242333582647723e-05, + "loss": 0.347, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17320868372917175, + "step": 945, + "valid_targets_mean": 4114.1, + "valid_targets_min": 1537 + }, + { + "epoch": 0.49764274489261395, + "grad_norm": 0.5635506423644553, + "learning_rate": 2.8391922213911745e-05, + "loss": 0.3466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1570136845111847, + "step": 950, + "valid_targets_mean": 4550.4, + "valid_targets_min": 3119 + }, + { + "epoch": 0.500261917234154, + "grad_norm": 0.6102420623699707, + "learning_rate": 2.8541510845175772e-05, + "loss": 0.3605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16992977261543274, + "step": 955, + "valid_targets_mean": 3810.8, + "valid_targets_min": 2562 + }, + { + "epoch": 0.5028810895756941, + "grad_norm": 0.6357613744648668, + "learning_rate": 2.8691099476439795e-05, + "loss": 0.3547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1734837293624878, + "step": 960, + "valid_targets_mean": 4045.5, + "valid_targets_min": 2659 + }, + { + "epoch": 0.5055002619172342, + "grad_norm": 0.5672018736295554, + "learning_rate": 2.8840688107703817e-05, + "loss": 0.3433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15104591846466064, + "step": 965, + "valid_targets_mean": 4064.8, + "valid_targets_min": 1349 + }, + { + "epoch": 0.5081194342587743, + "grad_norm": 0.6992361441845018, + "learning_rate": 2.899027673896784e-05, + "loss": 0.3505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1616719663143158, + "step": 970, + "valid_targets_mean": 3371.4, + "valid_targets_min": 2474 + }, + { + "epoch": 0.5107386066003143, + "grad_norm": 0.7201340905893235, + "learning_rate": 2.9139865370231867e-05, + "loss": 0.3617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14932557940483093, + "step": 975, + "valid_targets_mean": 2590.5, + "valid_targets_min": 738 + }, + { + "epoch": 0.5133577789418544, + "grad_norm": 0.6245515596001965, + "learning_rate": 2.928945400149589e-05, + "loss": 0.3374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16441687941551208, + "step": 980, + "valid_targets_mean": 3570.5, + "valid_targets_min": 2033 + }, + { + "epoch": 0.5159769512833945, + "grad_norm": 0.7745866845987275, + "learning_rate": 2.9439042632759912e-05, + "loss": 0.3657, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19962342083454132, + "step": 985, + "valid_targets_mean": 2933.1, + "valid_targets_min": 1346 + }, + { + "epoch": 0.5185961236249346, + "grad_norm": 0.7723187310234857, + "learning_rate": 2.958863126402394e-05, + "loss": 0.3572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19229379296302795, + "step": 990, + "valid_targets_mean": 2991.1, + "valid_targets_min": 1821 + }, + { + "epoch": 0.5212152959664746, + "grad_norm": 0.6524215127574097, + "learning_rate": 2.973821989528796e-05, + "loss": 0.3493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16409315168857574, + "step": 995, + "valid_targets_mean": 3400.1, + "valid_targets_min": 2535 + }, + { + "epoch": 0.5238344683080147, + "grad_norm": 0.6783160520443756, + "learning_rate": 2.9887808526551984e-05, + "loss": 0.3417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16737578809261322, + "step": 1000, + "valid_targets_mean": 3411.5, + "valid_targets_min": 2352 + }, + { + "epoch": 0.5264536406495547, + "grad_norm": 0.6658252423421862, + "learning_rate": 3.0037397157816007e-05, + "loss": 0.3471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1571807563304901, + "step": 1005, + "valid_targets_mean": 3319.8, + "valid_targets_min": 1153 + }, + { + "epoch": 0.5290728129910948, + "grad_norm": 0.6364765835409558, + "learning_rate": 3.0186985789080034e-05, + "loss": 0.3423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17302633821964264, + "step": 1010, + "valid_targets_mean": 3718.0, + "valid_targets_min": 2072 + }, + { + "epoch": 0.5316919853326348, + "grad_norm": 0.6884342277791285, + "learning_rate": 3.0336574420344056e-05, + "loss": 0.354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16124552488327026, + "step": 1015, + "valid_targets_mean": 3451.0, + "valid_targets_min": 922 + }, + { + "epoch": 0.5343111576741749, + "grad_norm": 0.6080791470208546, + "learning_rate": 3.048616305160808e-05, + "loss": 0.3488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13839368522167206, + "step": 1020, + "valid_targets_mean": 3381.6, + "valid_targets_min": 1417 + }, + { + "epoch": 0.536930330015715, + "grad_norm": 0.7724471346673448, + "learning_rate": 3.06357516828721e-05, + "loss": 0.3453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15144342184066772, + "step": 1025, + "valid_targets_mean": 4260.6, + "valid_targets_min": 1844 + }, + { + "epoch": 0.5395495023572551, + "grad_norm": 0.6653787979206836, + "learning_rate": 3.078534031413613e-05, + "loss": 0.345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17715337872505188, + "step": 1030, + "valid_targets_mean": 3525.2, + "valid_targets_min": 1261 + }, + { + "epoch": 0.5421686746987951, + "grad_norm": 0.8187554902465186, + "learning_rate": 3.093492894540015e-05, + "loss": 0.339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19987159967422485, + "step": 1035, + "valid_targets_mean": 4548.6, + "valid_targets_min": 2077 + }, + { + "epoch": 0.5447878470403352, + "grad_norm": 0.7446105257307885, + "learning_rate": 3.1084517576664174e-05, + "loss": 0.3451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18285146355628967, + "step": 1040, + "valid_targets_mean": 4140.9, + "valid_targets_min": 2724 + }, + { + "epoch": 0.5474070193818753, + "grad_norm": 0.5540130673432141, + "learning_rate": 3.12341062079282e-05, + "loss": 0.3342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22029636800289154, + "step": 1045, + "valid_targets_mean": 5291.4, + "valid_targets_min": 2292 + }, + { + "epoch": 0.5500261917234154, + "grad_norm": 0.6998413435998067, + "learning_rate": 3.138369483919222e-05, + "loss": 0.3555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17263412475585938, + "step": 1050, + "valid_targets_mean": 2965.1, + "valid_targets_min": 1377 + }, + { + "epoch": 0.5526453640649555, + "grad_norm": 0.751588014137714, + "learning_rate": 3.1533283470456246e-05, + "loss": 0.3565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1517869234085083, + "step": 1055, + "valid_targets_mean": 3099.5, + "valid_targets_min": 1350 + }, + { + "epoch": 0.5552645364064955, + "grad_norm": 0.653395490289795, + "learning_rate": 3.168287210172027e-05, + "loss": 0.3429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1674426794052124, + "step": 1060, + "valid_targets_mean": 3031.9, + "valid_targets_min": 654 + }, + { + "epoch": 0.5578837087480356, + "grad_norm": 0.6248244115873748, + "learning_rate": 3.183246073298429e-05, + "loss": 0.3418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13438619673252106, + "step": 1065, + "valid_targets_mean": 3463.8, + "valid_targets_min": 1706 + }, + { + "epoch": 0.5605028810895757, + "grad_norm": 0.6142513284269528, + "learning_rate": 3.198204936424832e-05, + "loss": 0.3473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18299362063407898, + "step": 1070, + "valid_targets_mean": 4105.5, + "valid_targets_min": 1739 + }, + { + "epoch": 0.5631220534311158, + "grad_norm": 0.6783748138280209, + "learning_rate": 3.2131637995512345e-05, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14706772565841675, + "step": 1075, + "valid_targets_mean": 2944.4, + "valid_targets_min": 1302 + }, + { + "epoch": 0.5657412257726558, + "grad_norm": 0.5856718631309122, + "learning_rate": 3.228122662677637e-05, + "loss": 0.3213, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14660827815532684, + "step": 1080, + "valid_targets_mean": 3531.5, + "valid_targets_min": 2501 + }, + { + "epoch": 0.5683603981141959, + "grad_norm": 0.5660701879503212, + "learning_rate": 3.243081525804039e-05, + "loss": 0.3385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17245368659496307, + "step": 1085, + "valid_targets_mean": 4673.8, + "valid_targets_min": 3232 + }, + { + "epoch": 0.570979570455736, + "grad_norm": 0.5778778954293626, + "learning_rate": 3.258040388930442e-05, + "loss": 0.321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16617146134376526, + "step": 1090, + "valid_targets_mean": 4465.1, + "valid_targets_min": 1101 + }, + { + "epoch": 0.5735987427972761, + "grad_norm": 0.6102601396381382, + "learning_rate": 3.272999252056844e-05, + "loss": 0.3554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18329699337482452, + "step": 1095, + "valid_targets_mean": 4113.2, + "valid_targets_min": 2079 + }, + { + "epoch": 0.5762179151388162, + "grad_norm": 0.6056214083000495, + "learning_rate": 3.287958115183246e-05, + "loss": 0.3333, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15843001008033752, + "step": 1100, + "valid_targets_mean": 3891.4, + "valid_targets_min": 2138 + }, + { + "epoch": 0.5788370874803562, + "grad_norm": 0.5811583185884692, + "learning_rate": 3.302916978309649e-05, + "loss": 0.3427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16537189483642578, + "step": 1105, + "valid_targets_mean": 3470.2, + "valid_targets_min": 1234 + }, + { + "epoch": 0.5814562598218963, + "grad_norm": 0.586801732996604, + "learning_rate": 3.3178758414360515e-05, + "loss": 0.3442, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1705385148525238, + "step": 1110, + "valid_targets_mean": 4038.5, + "valid_targets_min": 1822 + }, + { + "epoch": 0.5840754321634364, + "grad_norm": 0.5588528221657763, + "learning_rate": 3.3328347045624535e-05, + "loss": 0.3507, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17157317698001862, + "step": 1115, + "valid_targets_mean": 4214.6, + "valid_targets_min": 1651 + }, + { + "epoch": 0.5866946045049765, + "grad_norm": 0.6545258613780315, + "learning_rate": 3.347793567688856e-05, + "loss": 0.3402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19783565402030945, + "step": 1120, + "valid_targets_mean": 3447.9, + "valid_targets_min": 1434 + }, + { + "epoch": 0.5893137768465165, + "grad_norm": 0.942670632466867, + "learning_rate": 3.362752430815259e-05, + "loss": 0.3291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18441694974899292, + "step": 1125, + "valid_targets_mean": 4753.4, + "valid_targets_min": 2725 + }, + { + "epoch": 0.5919329491880566, + "grad_norm": 0.5631877716881122, + "learning_rate": 3.3777112939416607e-05, + "loss": 0.3285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1637195646762848, + "step": 1130, + "valid_targets_mean": 4005.8, + "valid_targets_min": 1966 + }, + { + "epoch": 0.5945521215295967, + "grad_norm": 0.6044559273166187, + "learning_rate": 3.392670157068063e-05, + "loss": 0.3312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.161268949508667, + "step": 1135, + "valid_targets_mean": 3669.0, + "valid_targets_min": 1407 + }, + { + "epoch": 0.5971712938711368, + "grad_norm": 0.6167603070867331, + "learning_rate": 3.407629020194465e-05, + "loss": 0.3295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17486314475536346, + "step": 1140, + "valid_targets_mean": 4330.5, + "valid_targets_min": 1444 + }, + { + "epoch": 0.5997904662126768, + "grad_norm": 0.6963410223830024, + "learning_rate": 3.422587883320868e-05, + "loss": 0.3153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1478385031223297, + "step": 1145, + "valid_targets_mean": 3077.8, + "valid_targets_min": 1504 + }, + { + "epoch": 0.6024096385542169, + "grad_norm": 0.6106269305667551, + "learning_rate": 3.4375467464472705e-05, + "loss": 0.3221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13165739178657532, + "step": 1150, + "valid_targets_mean": 3051.6, + "valid_targets_min": 858 + }, + { + "epoch": 0.6050288108957569, + "grad_norm": 0.6531606382877673, + "learning_rate": 3.4525056095736724e-05, + "loss": 0.3368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14299984276294708, + "step": 1155, + "valid_targets_mean": 3231.8, + "valid_targets_min": 1288 + }, + { + "epoch": 0.607647983237297, + "grad_norm": 0.563240839310093, + "learning_rate": 3.467464472700075e-05, + "loss": 0.3571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16313305497169495, + "step": 1160, + "valid_targets_mean": 4053.0, + "valid_targets_min": 1537 + }, + { + "epoch": 0.610267155578837, + "grad_norm": 0.6215447242963815, + "learning_rate": 3.482423335826478e-05, + "loss": 0.3305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16837164759635925, + "step": 1165, + "valid_targets_mean": 3917.1, + "valid_targets_min": 2060 + }, + { + "epoch": 0.6128863279203771, + "grad_norm": 0.6025581951770803, + "learning_rate": 3.4973821989528796e-05, + "loss": 0.343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1821030080318451, + "step": 1170, + "valid_targets_mean": 4482.6, + "valid_targets_min": 2515 + }, + { + "epoch": 0.6155055002619172, + "grad_norm": 0.6555276706566531, + "learning_rate": 3.512341062079282e-05, + "loss": 0.3226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1646737903356552, + "step": 1175, + "valid_targets_mean": 3847.2, + "valid_targets_min": 1744 + }, + { + "epoch": 0.6181246726034573, + "grad_norm": 0.5629187679610791, + "learning_rate": 3.527299925205684e-05, + "loss": 0.3344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15804052352905273, + "step": 1180, + "valid_targets_mean": 5009.9, + "valid_targets_min": 2062 + }, + { + "epoch": 0.6207438449449973, + "grad_norm": 0.611184152769321, + "learning_rate": 3.542258788332087e-05, + "loss": 0.33, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1545385718345642, + "step": 1185, + "valid_targets_mean": 3550.4, + "valid_targets_min": 1181 + }, + { + "epoch": 0.6233630172865374, + "grad_norm": 0.6455571552414104, + "learning_rate": 3.5572176514584895e-05, + "loss": 0.3145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17990398406982422, + "step": 1190, + "valid_targets_mean": 4010.6, + "valid_targets_min": 1937 + }, + { + "epoch": 0.6259821896280775, + "grad_norm": 0.6614684169758552, + "learning_rate": 3.5721765145848914e-05, + "loss": 0.3432, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20303750038146973, + "step": 1195, + "valid_targets_mean": 3624.9, + "valid_targets_min": 1649 + }, + { + "epoch": 0.6286013619696176, + "grad_norm": 0.713814732975456, + "learning_rate": 3.587135377711294e-05, + "loss": 0.3249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15035101771354675, + "step": 1200, + "valid_targets_mean": 2808.1, + "valid_targets_min": 1197 + }, + { + "epoch": 0.6312205343111577, + "grad_norm": 0.7735397086520968, + "learning_rate": 3.602094240837697e-05, + "loss": 0.3391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17234613001346588, + "step": 1205, + "valid_targets_mean": 3251.4, + "valid_targets_min": 1824 + }, + { + "epoch": 0.6338397066526977, + "grad_norm": 0.7180036086194709, + "learning_rate": 3.6170531039640986e-05, + "loss": 0.3446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1728033721446991, + "step": 1210, + "valid_targets_mean": 3145.6, + "valid_targets_min": 538 + }, + { + "epoch": 0.6364588789942378, + "grad_norm": 0.5782279885877064, + "learning_rate": 3.632011967090502e-05, + "loss": 0.3318, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14776058495044708, + "step": 1215, + "valid_targets_mean": 3597.1, + "valid_targets_min": 537 + }, + { + "epoch": 0.6390780513357779, + "grad_norm": 0.6536712336711337, + "learning_rate": 3.646970830216904e-05, + "loss": 0.3641, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16705811023712158, + "step": 1220, + "valid_targets_mean": 3386.5, + "valid_targets_min": 2169 + }, + { + "epoch": 0.641697223677318, + "grad_norm": 0.6149361223387535, + "learning_rate": 3.6619296933433065e-05, + "loss": 0.3334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17140933871269226, + "step": 1225, + "valid_targets_mean": 3687.6, + "valid_targets_min": 1625 + }, + { + "epoch": 0.644316396018858, + "grad_norm": 0.6334010080889872, + "learning_rate": 3.6768885564697085e-05, + "loss": 0.339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14733102917671204, + "step": 1230, + "valid_targets_mean": 3807.0, + "valid_targets_min": 2642 + }, + { + "epoch": 0.6469355683603981, + "grad_norm": 0.6264494105359202, + "learning_rate": 3.691847419596111e-05, + "loss": 0.3463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1885928511619568, + "step": 1235, + "valid_targets_mean": 4289.9, + "valid_targets_min": 2935 + }, + { + "epoch": 0.6495547407019382, + "grad_norm": 0.5146385837367431, + "learning_rate": 3.706806282722514e-05, + "loss": 0.3404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1753353625535965, + "step": 1240, + "valid_targets_mean": 5189.9, + "valid_targets_min": 2652 + }, + { + "epoch": 0.6521739130434783, + "grad_norm": 0.6237914260971164, + "learning_rate": 3.7217651458489157e-05, + "loss": 0.3414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22259917855262756, + "step": 1245, + "valid_targets_mean": 4117.2, + "valid_targets_min": 648 + }, + { + "epoch": 0.6547930853850183, + "grad_norm": 0.6248272882179953, + "learning_rate": 3.736724008975318e-05, + "loss": 0.3413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14603295922279358, + "step": 1250, + "valid_targets_mean": 3492.8, + "valid_targets_min": 637 + }, + { + "epoch": 0.6574122577265584, + "grad_norm": 0.6546303777337685, + "learning_rate": 3.751682872101721e-05, + "loss": 0.3405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1761663556098938, + "step": 1255, + "valid_targets_mean": 3412.5, + "valid_targets_min": 719 + }, + { + "epoch": 0.6600314300680985, + "grad_norm": 0.5422017898171986, + "learning_rate": 3.766641735228123e-05, + "loss": 0.3355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15139931440353394, + "step": 1260, + "valid_targets_mean": 4227.4, + "valid_targets_min": 2182 + }, + { + "epoch": 0.6626506024096386, + "grad_norm": 0.5982919827512452, + "learning_rate": 3.7816005983545255e-05, + "loss": 0.3259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15435075759887695, + "step": 1265, + "valid_targets_mean": 3211.1, + "valid_targets_min": 1372 + }, + { + "epoch": 0.6652697747511787, + "grad_norm": 0.6156087252532637, + "learning_rate": 3.796559461480928e-05, + "loss": 0.3185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19151005148887634, + "step": 1270, + "valid_targets_mean": 4490.9, + "valid_targets_min": 2594 + }, + { + "epoch": 0.6678889470927187, + "grad_norm": 0.5841833074288615, + "learning_rate": 3.81151832460733e-05, + "loss": 0.3305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14895989000797272, + "step": 1275, + "valid_targets_mean": 3054.6, + "valid_targets_min": 530 + }, + { + "epoch": 0.6705081194342588, + "grad_norm": 0.6451901742570407, + "learning_rate": 3.826477187733733e-05, + "loss": 0.3319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.2149169147014618, + "step": 1280, + "valid_targets_mean": 4110.2, + "valid_targets_min": 611 + }, + { + "epoch": 0.6731272917757989, + "grad_norm": 0.6256493557818691, + "learning_rate": 3.8414360508601346e-05, + "loss": 0.3296, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17664754390716553, + "step": 1285, + "valid_targets_mean": 3804.2, + "valid_targets_min": 1987 + }, + { + "epoch": 0.675746464117339, + "grad_norm": 0.6593530624440215, + "learning_rate": 3.856394913986537e-05, + "loss": 0.3351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1696583330631256, + "step": 1290, + "valid_targets_mean": 3203.2, + "valid_targets_min": 1362 + }, + { + "epoch": 0.678365636458879, + "grad_norm": 0.6326237919700105, + "learning_rate": 3.87135377711294e-05, + "loss": 0.3263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15649345517158508, + "step": 1295, + "valid_targets_mean": 3245.1, + "valid_targets_min": 852 + }, + { + "epoch": 0.6809848088004191, + "grad_norm": 0.6769003859295334, + "learning_rate": 3.886312640239342e-05, + "loss": 0.3366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1369558721780777, + "step": 1300, + "valid_targets_mean": 2537.4, + "valid_targets_min": 1195 + }, + { + "epoch": 0.6836039811419592, + "grad_norm": 0.6354502978564346, + "learning_rate": 3.9012715033657445e-05, + "loss": 0.3289, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17351612448692322, + "step": 1305, + "valid_targets_mean": 4383.0, + "valid_targets_min": 1397 + }, + { + "epoch": 0.6862231534834992, + "grad_norm": 0.6049639267393389, + "learning_rate": 3.916230366492147e-05, + "loss": 0.3183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.128768652677536, + "step": 1310, + "valid_targets_mean": 3421.0, + "valid_targets_min": 1975 + }, + { + "epoch": 0.6888423258250392, + "grad_norm": 0.6201635179402464, + "learning_rate": 3.931189229618549e-05, + "loss": 0.3111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1679595410823822, + "step": 1315, + "valid_targets_mean": 4100.6, + "valid_targets_min": 2564 + }, + { + "epoch": 0.6914614981665793, + "grad_norm": 0.5154340087310141, + "learning_rate": 3.946148092744952e-05, + "loss": 0.3163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14673280715942383, + "step": 1320, + "valid_targets_mean": 4986.9, + "valid_targets_min": 2595 + }, + { + "epoch": 0.6940806705081194, + "grad_norm": 0.6067751988903565, + "learning_rate": 3.9611069558713536e-05, + "loss": 0.3264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18214502930641174, + "step": 1325, + "valid_targets_mean": 3992.9, + "valid_targets_min": 2845 + }, + { + "epoch": 0.6966998428496595, + "grad_norm": 0.736277071204803, + "learning_rate": 3.976065818997756e-05, + "loss": 0.3142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12144848704338074, + "step": 1330, + "valid_targets_mean": 2390.1, + "valid_targets_min": 993 + }, + { + "epoch": 0.6993190151911995, + "grad_norm": 0.605888039510588, + "learning_rate": 3.991024682124159e-05, + "loss": 0.3142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13026097416877747, + "step": 1335, + "valid_targets_mean": 3729.1, + "valid_targets_min": 1575 + }, + { + "epoch": 0.7019381875327396, + "grad_norm": 0.6182757712058891, + "learning_rate": 3.999999727028487e-05, + "loss": 0.3256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1745564490556717, + "step": 1340, + "valid_targets_mean": 3469.0, + "valid_targets_min": 1360 + }, + { + "epoch": 0.7045573598742797, + "grad_norm": 0.6020657450587164, + "learning_rate": 3.999996656099815e-05, + "loss": 0.3143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15214449167251587, + "step": 1345, + "valid_targets_mean": 3555.5, + "valid_targets_min": 2278 + }, + { + "epoch": 0.7071765322158198, + "grad_norm": 0.5937485102189005, + "learning_rate": 3.999990173033336e-05, + "loss": 0.3444, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18724879622459412, + "step": 1350, + "valid_targets_mean": 3812.2, + "valid_targets_min": 1454 + }, + { + "epoch": 0.7097957045573599, + "grad_norm": 0.6990820857287618, + "learning_rate": 3.99998027784011e-05, + "loss": 0.331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1770421415567398, + "step": 1355, + "valid_targets_mean": 3796.6, + "valid_targets_min": 2344 + }, + { + "epoch": 0.7124148768988999, + "grad_norm": 0.6826191974023101, + "learning_rate": 3.999966970537021e-05, + "loss": 0.3307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1755611002445221, + "step": 1360, + "valid_targets_mean": 3324.1, + "valid_targets_min": 1278 + }, + { + "epoch": 0.71503404924044, + "grad_norm": 0.6906743990968118, + "learning_rate": 3.999950251146769e-05, + "loss": 0.3142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17887212336063385, + "step": 1365, + "valid_targets_mean": 3381.4, + "valid_targets_min": 1858 + }, + { + "epoch": 0.7176532215819801, + "grad_norm": 0.634504099352907, + "learning_rate": 3.9999301196978804e-05, + "loss": 0.3248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12715640664100647, + "step": 1370, + "valid_targets_mean": 2566.0, + "valid_targets_min": 420 + }, + { + "epoch": 0.7202723939235202, + "grad_norm": 0.6483388411131212, + "learning_rate": 3.9999065762247e-05, + "loss": 0.3193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14489418268203735, + "step": 1375, + "valid_targets_mean": 2649.4, + "valid_targets_min": 1039 + }, + { + "epoch": 0.7228915662650602, + "grad_norm": 0.5650683344632254, + "learning_rate": 3.999879620767397e-05, + "loss": 0.3327, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14652296900749207, + "step": 1380, + "valid_targets_mean": 4475.1, + "valid_targets_min": 1737 + }, + { + "epoch": 0.7255107386066003, + "grad_norm": 0.5759056581523377, + "learning_rate": 3.999849253371956e-05, + "loss": 0.3244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19368822872638702, + "step": 1385, + "valid_targets_mean": 4150.8, + "valid_targets_min": 809 + }, + { + "epoch": 0.7281299109481404, + "grad_norm": 0.5804812639690735, + "learning_rate": 3.999815474090188e-05, + "loss": 0.3224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1640108823776245, + "step": 1390, + "valid_targets_mean": 3864.1, + "valid_targets_min": 1780 + }, + { + "epoch": 0.7307490832896805, + "grad_norm": 0.5970569123002243, + "learning_rate": 3.9997782829797225e-05, + "loss": 0.3308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1548389345407486, + "step": 1395, + "valid_targets_mean": 4058.0, + "valid_targets_min": 2185 + }, + { + "epoch": 0.7333682556312205, + "grad_norm": 0.7407058719501103, + "learning_rate": 3.99973768010401e-05, + "loss": 0.2871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09251627326011658, + "step": 1400, + "valid_targets_mean": 2736.0, + "valid_targets_min": 876 + }, + { + "epoch": 0.7359874279727606, + "grad_norm": 0.7123366834918636, + "learning_rate": 3.9996936655323216e-05, + "loss": 0.1992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07504437118768692, + "step": 1405, + "valid_targets_mean": 1472.5, + "valid_targets_min": 324 + }, + { + "epoch": 0.7386066003143007, + "grad_norm": 0.5281139362837265, + "learning_rate": 3.9996462393397505e-05, + "loss": 0.1527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08766627311706543, + "step": 1410, + "valid_targets_mean": 2699.9, + "valid_targets_min": 1064 + }, + { + "epoch": 0.7412257726558408, + "grad_norm": 0.6575655245607706, + "learning_rate": 3.999595401607208e-05, + "loss": 0.1451, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07486118376255035, + "step": 1415, + "valid_targets_mean": 3137.0, + "valid_targets_min": 710 + }, + { + "epoch": 0.7438449449973809, + "grad_norm": 0.45553199170653613, + "learning_rate": 3.9995411524214275e-05, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050425510853528976, + "step": 1420, + "valid_targets_mean": 3472.9, + "valid_targets_min": 1223 + }, + { + "epoch": 0.7464641173389209, + "grad_norm": 0.5005426582691173, + "learning_rate": 3.999483491874962e-05, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0859721302986145, + "step": 1425, + "valid_targets_mean": 3326.0, + "valid_targets_min": 1431 + }, + { + "epoch": 0.749083289680461, + "grad_norm": 0.836202950371586, + "learning_rate": 3.999422420066184e-05, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08828853815793991, + "step": 1430, + "valid_targets_mean": 1161.2, + "valid_targets_min": 559 + }, + { + "epoch": 0.7517024620220011, + "grad_norm": 0.4532628184562695, + "learning_rate": 3.9993579370992874e-05, + "loss": 0.1879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07064370810985565, + "step": 1435, + "valid_targets_mean": 3598.2, + "valid_targets_min": 1623 + }, + { + "epoch": 0.7543216343635412, + "grad_norm": 0.5599430799169327, + "learning_rate": 3.9992900430842836e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07889188826084137, + "step": 1440, + "valid_targets_mean": 3178.2, + "valid_targets_min": 846 + }, + { + "epoch": 0.7569408067050812, + "grad_norm": 0.631261836422846, + "learning_rate": 3.999218738137005e-05, + "loss": 0.2056, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1353520154953003, + "step": 1445, + "valid_targets_mean": 3622.6, + "valid_targets_min": 1949 + }, + { + "epoch": 0.7595599790466213, + "grad_norm": 0.403486573367227, + "learning_rate": 3.999144022379103e-05, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06411699950695038, + "step": 1450, + "valid_targets_mean": 3344.9, + "valid_targets_min": 764 + }, + { + "epoch": 0.7621791513881614, + "grad_norm": 0.5947859741427595, + "learning_rate": 3.9990658959380485e-05, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07560296356678009, + "step": 1455, + "valid_targets_mean": 2391.6, + "valid_targets_min": 776 + }, + { + "epoch": 0.7647983237297015, + "grad_norm": 0.45931456788085523, + "learning_rate": 3.998984358947131e-05, + "loss": 0.1401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07437349855899811, + "step": 1460, + "valid_targets_mean": 2818.5, + "valid_targets_min": 754 + }, + { + "epoch": 0.7674174960712414, + "grad_norm": 0.5832161301591761, + "learning_rate": 3.998899411545457e-05, + "loss": 0.1678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08421052992343903, + "step": 1465, + "valid_targets_mean": 1646.6, + "valid_targets_min": 635 + }, + { + "epoch": 0.7700366684127815, + "grad_norm": 0.44065082758692464, + "learning_rate": 3.998811053877955e-05, + "loss": 0.1546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0719112902879715, + "step": 1470, + "valid_targets_mean": 3374.8, + "valid_targets_min": 1862 + }, + { + "epoch": 0.7726558407543216, + "grad_norm": 0.4942334789102453, + "learning_rate": 3.998719286095368e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08540884405374527, + "step": 1475, + "valid_targets_mean": 2837.4, + "valid_targets_min": 574 + }, + { + "epoch": 0.7752750130958617, + "grad_norm": 0.5154986893426464, + "learning_rate": 3.998624108354258e-05, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06067972257733345, + "step": 1480, + "valid_targets_mean": 2618.4, + "valid_targets_min": 523 + }, + { + "epoch": 0.7778941854374017, + "grad_norm": 0.5313616629936678, + "learning_rate": 3.9985255208170065e-05, + "loss": 0.1329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07875044643878937, + "step": 1485, + "valid_targets_mean": 3548.9, + "valid_targets_min": 2189 + }, + { + "epoch": 0.7805133577789418, + "grad_norm": 0.47868406767999894, + "learning_rate": 3.99842352365181e-05, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05243462324142456, + "step": 1490, + "valid_targets_mean": 3564.6, + "valid_targets_min": 827 + }, + { + "epoch": 0.7831325301204819, + "grad_norm": 0.5326962248807613, + "learning_rate": 3.9983181170326835e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07231694459915161, + "step": 1495, + "valid_targets_mean": 1936.6, + "valid_targets_min": 597 + }, + { + "epoch": 0.785751702462022, + "grad_norm": 0.5442698357718908, + "learning_rate": 3.998209301139458e-05, + "loss": 0.1543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06426198780536652, + "step": 1500, + "valid_targets_mean": 1747.5, + "valid_targets_min": 701 + }, + { + "epoch": 0.788370874803562, + "grad_norm": 0.583516724247055, + "learning_rate": 3.998097076157781e-05, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16074779629707336, + "step": 1505, + "valid_targets_mean": 3329.6, + "valid_targets_min": 1652 + }, + { + "epoch": 0.7909900471451021, + "grad_norm": 0.4507185683421478, + "learning_rate": 3.997981442279117e-05, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06859997659921646, + "step": 1510, + "valid_targets_mean": 3117.1, + "valid_targets_min": 786 + }, + { + "epoch": 0.7936092194866422, + "grad_norm": 0.5916821039913717, + "learning_rate": 3.997862399700744e-05, + "loss": 0.1332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08142507076263428, + "step": 1515, + "valid_targets_mean": 2574.4, + "valid_targets_min": 971 + }, + { + "epoch": 0.7962283918281823, + "grad_norm": 0.4409848210686347, + "learning_rate": 3.99773994862576e-05, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07521601766347885, + "step": 1520, + "valid_targets_mean": 3320.0, + "valid_targets_min": 2610 + }, + { + "epoch": 0.7988475641697224, + "grad_norm": 0.46801534693314467, + "learning_rate": 3.997614089263073e-05, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05124581232666969, + "step": 1525, + "valid_targets_mean": 2580.6, + "valid_targets_min": 753 + }, + { + "epoch": 0.8014667365112624, + "grad_norm": 0.589071157570724, + "learning_rate": 3.997484821827409e-05, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06667555123567581, + "step": 1530, + "valid_targets_mean": 1537.1, + "valid_targets_min": 619 + }, + { + "epoch": 0.8040859088528025, + "grad_norm": 0.625596827000952, + "learning_rate": 3.9973521465393075e-05, + "loss": 0.3689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1508081555366516, + "step": 1535, + "valid_targets_mean": 2271.2, + "valid_targets_min": 987 + }, + { + "epoch": 0.8067050811943426, + "grad_norm": 0.722759028705856, + "learning_rate": 3.997216063625121e-05, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07766801863908768, + "step": 1540, + "valid_targets_mean": 1351.5, + "valid_targets_min": 329 + }, + { + "epoch": 0.8093242535358827, + "grad_norm": 0.5791938823616255, + "learning_rate": 3.997076573317019e-05, + "loss": 0.1449, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06811608374118805, + "step": 1545, + "valid_targets_mean": 1553.9, + "valid_targets_min": 608 + }, + { + "epoch": 0.8119434258774227, + "grad_norm": 0.5013381487806157, + "learning_rate": 3.9969336758529804e-05, + "loss": 0.2543, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09672108292579651, + "step": 1550, + "valid_targets_mean": 3645.6, + "valid_targets_min": 873 + }, + { + "epoch": 0.8145625982189628, + "grad_norm": 0.30114899022781444, + "learning_rate": 3.996787371476799e-05, + "loss": 0.1233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03411861136555672, + "step": 1555, + "valid_targets_mean": 4720.6, + "valid_targets_min": 3529 + }, + { + "epoch": 0.8171817705605029, + "grad_norm": 0.41717017083414426, + "learning_rate": 3.9966376604380795e-05, + "loss": 0.1418, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07612346112728119, + "step": 1560, + "valid_targets_mean": 4437.6, + "valid_targets_min": 2336 + }, + { + "epoch": 0.819800942902043, + "grad_norm": 0.3836993047409372, + "learning_rate": 3.9964845429922405e-05, + "loss": 0.1382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06162077188491821, + "step": 1565, + "valid_targets_mean": 3533.1, + "valid_targets_min": 1226 + }, + { + "epoch": 0.822420115243583, + "grad_norm": 0.4629626197898817, + "learning_rate": 3.996328019400512e-05, + "loss": 0.1345, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06373080611228943, + "step": 1570, + "valid_targets_mean": 3299.2, + "valid_targets_min": 798 + }, + { + "epoch": 0.8250392875851231, + "grad_norm": 0.46005708860005584, + "learning_rate": 3.996168089929934e-05, + "loss": 0.1272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05273229628801346, + "step": 1575, + "valid_targets_mean": 1836.9, + "valid_targets_min": 723 + }, + { + "epoch": 0.8276584599266632, + "grad_norm": 0.4416803937849114, + "learning_rate": 3.996004754853356e-05, + "loss": 0.1292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0633687749505043, + "step": 1580, + "valid_targets_mean": 3316.0, + "valid_targets_min": 898 + }, + { + "epoch": 0.8302776322682033, + "grad_norm": 0.3494395850300165, + "learning_rate": 3.995838014449442e-05, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047255970537662506, + "step": 1585, + "valid_targets_mean": 4027.2, + "valid_targets_min": 3278 + }, + { + "epoch": 0.8328968046097434, + "grad_norm": 0.5032857844666191, + "learning_rate": 3.9956678690026625e-05, + "loss": 0.1326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060806356370449066, + "step": 1590, + "valid_targets_mean": 2052.8, + "valid_targets_min": 739 + }, + { + "epoch": 0.8355159769512834, + "grad_norm": 0.5507266894343988, + "learning_rate": 3.995494318803297e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07169269770383835, + "step": 1595, + "valid_targets_mean": 1574.9, + "valid_targets_min": 538 + }, + { + "epoch": 0.8381351492928235, + "grad_norm": 0.44337679255506685, + "learning_rate": 3.995317364147434e-05, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05059821903705597, + "step": 1600, + "valid_targets_mean": 1956.9, + "valid_targets_min": 654 + }, + { + "epoch": 0.8407543216343636, + "grad_norm": 0.5190606342693176, + "learning_rate": 3.9951370053369726e-05, + "loss": 0.1303, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.075295589864254, + "step": 1605, + "valid_targets_mean": 2735.4, + "valid_targets_min": 930 + }, + { + "epoch": 0.8433734939759037, + "grad_norm": 0.4420454078324289, + "learning_rate": 3.994953242679617e-05, + "loss": 0.1167, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06831593811511993, + "step": 1610, + "valid_targets_mean": 2965.0, + "valid_targets_min": 684 + }, + { + "epoch": 0.8459926663174437, + "grad_norm": 0.42420575167866587, + "learning_rate": 3.994766076488879e-05, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06846761703491211, + "step": 1615, + "valid_targets_mean": 3344.6, + "valid_targets_min": 999 + }, + { + "epoch": 0.8486118386589837, + "grad_norm": 0.5619185352725445, + "learning_rate": 3.994575507084078e-05, + "loss": 0.1425, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07901346683502197, + "step": 1620, + "valid_targets_mean": 2445.4, + "valid_targets_min": 696 + }, + { + "epoch": 0.8512310110005238, + "grad_norm": 0.424238841861574, + "learning_rate": 3.994381534790339e-05, + "loss": 0.1597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0722447857260704, + "step": 1625, + "valid_targets_mean": 3514.9, + "valid_targets_min": 1068 + }, + { + "epoch": 0.8538501833420639, + "grad_norm": 0.4946218871411475, + "learning_rate": 3.994184159938593e-05, + "loss": 0.1416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06080213561654091, + "step": 1630, + "valid_targets_mean": 1602.8, + "valid_targets_min": 536 + }, + { + "epoch": 0.8564693556836039, + "grad_norm": 0.5162399116055593, + "learning_rate": 3.993983382865576e-05, + "loss": 0.1982, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14086511731147766, + "step": 1635, + "valid_targets_mean": 3568.8, + "valid_targets_min": 1507 + }, + { + "epoch": 0.859088528025144, + "grad_norm": 0.4858013396621461, + "learning_rate": 3.993779203913827e-05, + "loss": 0.1435, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06756429374217987, + "step": 1640, + "valid_targets_mean": 3292.2, + "valid_targets_min": 2342 + }, + { + "epoch": 0.8617077003666841, + "grad_norm": 0.673701008442784, + "learning_rate": 3.99357162343169e-05, + "loss": 0.1734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08246868848800659, + "step": 1645, + "valid_targets_mean": 1567.5, + "valid_targets_min": 687 + }, + { + "epoch": 0.8643268727082242, + "grad_norm": 0.5951739003317199, + "learning_rate": 3.993360641773313e-05, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04255099594593048, + "step": 1650, + "valid_targets_mean": 1145.4, + "valid_targets_min": 449 + }, + { + "epoch": 0.8669460450497642, + "grad_norm": 0.41812700829285937, + "learning_rate": 3.993146259298646e-05, + "loss": 0.1394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0566326379776001, + "step": 1655, + "valid_targets_mean": 3191.0, + "valid_targets_min": 658 + }, + { + "epoch": 0.8695652173913043, + "grad_norm": 0.6849894413028811, + "learning_rate": 3.99292847637344e-05, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08343459665775299, + "step": 1660, + "valid_targets_mean": 1924.8, + "valid_targets_min": 577 + }, + { + "epoch": 0.8721843897328444, + "grad_norm": 0.38500370302634074, + "learning_rate": 3.9927072933692484e-05, + "loss": 0.114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054320476949214935, + "step": 1665, + "valid_targets_mean": 3992.5, + "valid_targets_min": 726 + }, + { + "epoch": 0.8748035620743845, + "grad_norm": 0.3869820206184452, + "learning_rate": 3.992482710663426e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06619054079055786, + "step": 1670, + "valid_targets_mean": 3626.4, + "valid_targets_min": 2252 + }, + { + "epoch": 0.8774227344159246, + "grad_norm": 0.4803448578377591, + "learning_rate": 3.992254728639127e-05, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05170255899429321, + "step": 1675, + "valid_targets_mean": 1556.2, + "valid_targets_min": 531 + }, + { + "epoch": 0.8800419067574646, + "grad_norm": 0.47279735808467427, + "learning_rate": 3.9920233476853035e-05, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06663917750120163, + "step": 1680, + "valid_targets_mean": 2410.5, + "valid_targets_min": 744 + }, + { + "epoch": 0.8826610790990047, + "grad_norm": 0.4398733484239651, + "learning_rate": 3.99178856819671e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07559102773666382, + "step": 1685, + "valid_targets_mean": 3795.1, + "valid_targets_min": 3136 + }, + { + "epoch": 0.8852802514405448, + "grad_norm": 0.41518957825090896, + "learning_rate": 3.991550390573897e-05, + "loss": 0.1233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053255148231983185, + "step": 1690, + "valid_targets_mean": 3324.4, + "valid_targets_min": 2366 + }, + { + "epoch": 0.8878994237820849, + "grad_norm": 0.3936672608404722, + "learning_rate": 3.9913088152232105e-05, + "loss": 0.1211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046618618071079254, + "step": 1695, + "valid_targets_mean": 1668.1, + "valid_targets_min": 669 + }, + { + "epoch": 0.8905185961236249, + "grad_norm": 0.4180997315955265, + "learning_rate": 3.9910638425567975e-05, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07795260846614838, + "step": 1700, + "valid_targets_mean": 3529.8, + "valid_targets_min": 1785 + }, + { + "epoch": 0.893137768465165, + "grad_norm": 0.31810424902249995, + "learning_rate": 3.990815472992598e-05, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05735762417316437, + "step": 1705, + "valid_targets_mean": 4120.1, + "valid_targets_min": 3862 + }, + { + "epoch": 0.8957569408067051, + "grad_norm": 0.46414341973173484, + "learning_rate": 3.9905637069543486e-05, + "loss": 0.1325, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06045958027243614, + "step": 1710, + "valid_targets_mean": 3062.2, + "valid_targets_min": 539 + }, + { + "epoch": 0.8983761131482452, + "grad_norm": 0.4285826224244272, + "learning_rate": 3.9903085448715805e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05149327218532562, + "step": 1715, + "valid_targets_mean": 2781.5, + "valid_targets_min": 1016 + }, + { + "epoch": 0.9009952854897852, + "grad_norm": 0.7086771263878007, + "learning_rate": 3.990049987179618e-05, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08912264555692673, + "step": 1720, + "valid_targets_mean": 1102.9, + "valid_targets_min": 888 + }, + { + "epoch": 0.9036144578313253, + "grad_norm": 0.5211685848036772, + "learning_rate": 3.98978803431958e-05, + "loss": 0.1366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056305788457393646, + "step": 1725, + "valid_targets_mean": 2602.5, + "valid_targets_min": 697 + }, + { + "epoch": 0.9062336301728654, + "grad_norm": 0.6769341120512076, + "learning_rate": 3.989522686738375e-05, + "loss": 0.138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08414820581674576, + "step": 1730, + "valid_targets_mean": 1592.9, + "valid_targets_min": 605 + }, + { + "epoch": 0.9088528025144055, + "grad_norm": 0.42413037335519205, + "learning_rate": 3.989253944888707e-05, + "loss": 0.1342, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05431302636861801, + "step": 1735, + "valid_targets_mean": 2899.1, + "valid_targets_min": 1117 + }, + { + "epoch": 0.9114719748559456, + "grad_norm": 0.44287853146204226, + "learning_rate": 3.988981809229067e-05, + "loss": 0.1606, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05965159833431244, + "step": 1740, + "valid_targets_mean": 3376.6, + "valid_targets_min": 1388 + }, + { + "epoch": 0.9140911471974856, + "grad_norm": 0.41766281857592913, + "learning_rate": 3.98870628022374e-05, + "loss": 0.1285, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05729830265045166, + "step": 1745, + "valid_targets_mean": 3243.4, + "valid_targets_min": 524 + }, + { + "epoch": 0.9167103195390257, + "grad_norm": 0.4519527973846729, + "learning_rate": 3.988427358342797e-05, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05896112322807312, + "step": 1750, + "valid_targets_mean": 2468.0, + "valid_targets_min": 970 + }, + { + "epoch": 0.9193294918805658, + "grad_norm": 0.4828208300260953, + "learning_rate": 3.9881450440620994e-05, + "loss": 0.1431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06769686937332153, + "step": 1755, + "valid_targets_mean": 2959.6, + "valid_targets_min": 877 + }, + { + "epoch": 0.9219486642221059, + "grad_norm": 0.33134772248567895, + "learning_rate": 3.987859337863295e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05763865262269974, + "step": 1760, + "valid_targets_mean": 3886.5, + "valid_targets_min": 3251 + }, + { + "epoch": 0.9245678365636459, + "grad_norm": 0.5679750958469983, + "learning_rate": 3.98757024023382e-05, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06938252598047256, + "step": 1765, + "valid_targets_mean": 2695.6, + "valid_targets_min": 564 + }, + { + "epoch": 0.927187008905186, + "grad_norm": 0.34980574520971525, + "learning_rate": 3.9872777516668935e-05, + "loss": 0.1231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0683695524930954, + "step": 1770, + "valid_targets_mean": 4404.8, + "valid_targets_min": 2555 + }, + { + "epoch": 0.929806181246726, + "grad_norm": 8.577386726490248, + "learning_rate": 3.9869818726615245e-05, + "loss": 0.1137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05616206303238869, + "step": 1775, + "valid_targets_mean": 4410.5, + "valid_targets_min": 3261 + }, + { + "epoch": 0.9324253535882661, + "grad_norm": 0.40794995313949345, + "learning_rate": 3.9866826037225025e-05, + "loss": 0.1055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04890760779380798, + "step": 1780, + "valid_targets_mean": 1768.5, + "valid_targets_min": 618 + }, + { + "epoch": 0.9350445259298061, + "grad_norm": 0.35150606054737477, + "learning_rate": 3.9863799453604026e-05, + "loss": 0.1334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05490623414516449, + "step": 1785, + "valid_targets_mean": 3985.8, + "valid_targets_min": 3213 + }, + { + "epoch": 0.9376636982713462, + "grad_norm": 0.39216639802133396, + "learning_rate": 3.986073898091581e-05, + "loss": 0.1292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11858995258808136, + "step": 1790, + "valid_targets_mean": 3149.1, + "valid_targets_min": 1539 + }, + { + "epoch": 0.9402828706128863, + "grad_norm": 0.42554580466377134, + "learning_rate": 3.985764462438176e-05, + "loss": 0.1402, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045684341341257095, + "step": 1795, + "valid_targets_mean": 2094.8, + "valid_targets_min": 509 + }, + { + "epoch": 0.9429020429544264, + "grad_norm": 0.3395099467564252, + "learning_rate": 3.985451638928108e-05, + "loss": 0.1295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03697893023490906, + "step": 1800, + "valid_targets_mean": 2441.4, + "valid_targets_min": 633 + }, + { + "epoch": 0.9455212152959664, + "grad_norm": 0.41696492484282005, + "learning_rate": 3.9851354280950756e-05, + "loss": 0.1565, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061639487743377686, + "step": 1805, + "valid_targets_mean": 2740.9, + "valid_targets_min": 737 + }, + { + "epoch": 0.9481403876375065, + "grad_norm": 0.37835978178507046, + "learning_rate": 3.9848158304785576e-05, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05642569810152054, + "step": 1810, + "valid_targets_mean": 2726.9, + "valid_targets_min": 492 + }, + { + "epoch": 0.9507595599790466, + "grad_norm": 0.6170717113022811, + "learning_rate": 3.984492846623811e-05, + "loss": 0.1426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08761419355869293, + "step": 1815, + "valid_targets_mean": 1607.9, + "valid_targets_min": 833 + }, + { + "epoch": 0.9533787323205867, + "grad_norm": 0.35260124248682106, + "learning_rate": 3.984166477081868e-05, + "loss": 0.1195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03975562006235123, + "step": 1820, + "valid_targets_mean": 3502.6, + "valid_targets_min": 923 + }, + { + "epoch": 0.9559979046621268, + "grad_norm": 0.7136251455406558, + "learning_rate": 3.983836722409539e-05, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08471337705850601, + "step": 1825, + "valid_targets_mean": 1553.9, + "valid_targets_min": 531 + }, + { + "epoch": 0.9586170770036668, + "grad_norm": 0.5314775492940201, + "learning_rate": 3.98350358316941e-05, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10734546184539795, + "step": 1830, + "valid_targets_mean": 1825.6, + "valid_targets_min": 575 + }, + { + "epoch": 0.9612362493452069, + "grad_norm": 0.4723329435469871, + "learning_rate": 3.9831670599298394e-05, + "loss": 0.1266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06796310842037201, + "step": 1835, + "valid_targets_mean": 3701.9, + "valid_targets_min": 2784 + }, + { + "epoch": 0.963855421686747, + "grad_norm": 0.3811253923538834, + "learning_rate": 3.9828271532649595e-05, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056041419506073, + "step": 1840, + "valid_targets_mean": 3575.8, + "valid_targets_min": 2525 + }, + { + "epoch": 0.9664745940282871, + "grad_norm": 0.36473666342325856, + "learning_rate": 3.9824838637546774e-05, + "loss": 0.1221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04743999242782593, + "step": 1845, + "valid_targets_mean": 3081.0, + "valid_targets_min": 970 + }, + { + "epoch": 0.9690937663698271, + "grad_norm": 0.4765982713916448, + "learning_rate": 3.982137191984668e-05, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06084107235074043, + "step": 1850, + "valid_targets_mean": 1963.5, + "valid_targets_min": 817 + }, + { + "epoch": 0.9717129387113672, + "grad_norm": 0.7699227534656983, + "learning_rate": 3.98178713854638e-05, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10203574597835541, + "step": 1855, + "valid_targets_mean": 1270.9, + "valid_targets_min": 637 + }, + { + "epoch": 0.9743321110529073, + "grad_norm": 0.4125237530623026, + "learning_rate": 3.981433704037027e-05, + "loss": 0.1109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05152922868728638, + "step": 1860, + "valid_targets_mean": 2594.1, + "valid_targets_min": 893 + }, + { + "epoch": 0.9769512833944474, + "grad_norm": 0.35709532219621265, + "learning_rate": 3.981076889059596e-05, + "loss": 0.1159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05419781059026718, + "step": 1865, + "valid_targets_mean": 3222.2, + "valid_targets_min": 1606 + }, + { + "epoch": 0.9795704557359874, + "grad_norm": 0.4098281114646845, + "learning_rate": 3.980716694222838e-05, + "loss": 0.1233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060739316046237946, + "step": 1870, + "valid_targets_mean": 3023.8, + "valid_targets_min": 678 + }, + { + "epoch": 0.9821896280775275, + "grad_norm": 0.30947930389770656, + "learning_rate": 3.9803531201412716e-05, + "loss": 0.1288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054963093250989914, + "step": 1875, + "valid_targets_mean": 4263.8, + "valid_targets_min": 1078 + }, + { + "epoch": 0.9848088004190676, + "grad_norm": 0.3052314115483043, + "learning_rate": 3.9799861674351794e-05, + "loss": 0.1095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06365899741649628, + "step": 1880, + "valid_targets_mean": 4150.0, + "valid_targets_min": 761 + }, + { + "epoch": 0.9874279727606077, + "grad_norm": 0.6745207808339296, + "learning_rate": 3.9796158367306095e-05, + "loss": 0.1237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060658685863018036, + "step": 1885, + "valid_targets_mean": 1368.4, + "valid_targets_min": 538 + }, + { + "epoch": 0.9900471451021478, + "grad_norm": 0.5940379932039093, + "learning_rate": 3.979242128659373e-05, + "loss": 0.14, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11413190513849258, + "step": 1890, + "valid_targets_mean": 2630.4, + "valid_targets_min": 780 + }, + { + "epoch": 0.9926663174436878, + "grad_norm": 0.35844590553528477, + "learning_rate": 3.9788650438590425e-05, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05106251686811447, + "step": 1895, + "valid_targets_mean": 3353.9, + "valid_targets_min": 2347 + }, + { + "epoch": 0.9952854897852279, + "grad_norm": 0.41657080961783366, + "learning_rate": 3.9784845829729516e-05, + "loss": 0.1203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0722593367099762, + "step": 1900, + "valid_targets_mean": 3880.2, + "valid_targets_min": 2473 + }, + { + "epoch": 0.997904662126768, + "grad_norm": 0.3831783479997989, + "learning_rate": 3.978100746650194e-05, + "loss": 0.1188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0609828382730484, + "step": 1905, + "valid_targets_mean": 3556.4, + "valid_targets_min": 2245 + }, + { + "epoch": 1.000523834468308, + "grad_norm": 0.5891550548515241, + "learning_rate": 3.977713535545622e-05, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13301968574523926, + "step": 1910, + "valid_targets_mean": 6935.0, + "valid_targets_min": 5295 + }, + { + "epoch": 1.0031430068098481, + "grad_norm": 0.46446767759878943, + "learning_rate": 3.9773229503198455e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14454826712608337, + "step": 1915, + "valid_targets_mean": 7672.9, + "valid_targets_min": 5066 + }, + { + "epoch": 1.0057621791513882, + "grad_norm": 0.42191403284017054, + "learning_rate": 3.976928991639231e-05, + "loss": 0.2377, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12194779515266418, + "step": 1920, + "valid_targets_mean": 7347.6, + "valid_targets_min": 4923 + }, + { + "epoch": 1.0083813514929283, + "grad_norm": 0.4918589611572868, + "learning_rate": 3.9765316601759e-05, + "loss": 0.2358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11181892454624176, + "step": 1925, + "valid_targets_mean": 5446.8, + "valid_targets_min": 3922 + }, + { + "epoch": 1.0110005238344684, + "grad_norm": 0.44184099396100585, + "learning_rate": 3.976130956607729e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11097045242786407, + "step": 1930, + "valid_targets_mean": 6924.9, + "valid_targets_min": 4797 + }, + { + "epoch": 1.0136196961760084, + "grad_norm": 0.4398772947733048, + "learning_rate": 3.9757268816183464e-05, + "loss": 0.2497, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12383102625608444, + "step": 1935, + "valid_targets_mean": 6443.5, + "valid_targets_min": 3853 + }, + { + "epoch": 1.0162388685175485, + "grad_norm": 0.4811811807472896, + "learning_rate": 3.975319435897134e-05, + "loss": 0.2405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11831942200660706, + "step": 1940, + "valid_targets_mean": 5446.5, + "valid_targets_min": 4487 + }, + { + "epoch": 1.0188580408590886, + "grad_norm": 0.44206430025195753, + "learning_rate": 3.9749086201392224e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11301036924123764, + "step": 1945, + "valid_targets_mean": 6210.2, + "valid_targets_min": 4137 + }, + { + "epoch": 1.0214772132006287, + "grad_norm": 0.5585132229832563, + "learning_rate": 3.9744944350454935e-05, + "loss": 0.2594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.20359641313552856, + "step": 1950, + "valid_targets_mean": 5404.4, + "valid_targets_min": 698 + }, + { + "epoch": 1.0240963855421688, + "grad_norm": 0.3992929084889272, + "learning_rate": 3.9740768813225765e-05, + "loss": 0.2194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12981699407100677, + "step": 1955, + "valid_targets_mean": 7197.6, + "valid_targets_min": 4499 + }, + { + "epoch": 1.0267155578837088, + "grad_norm": 0.4944221686072624, + "learning_rate": 3.973655959682847e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1508607268333435, + "step": 1960, + "valid_targets_mean": 5538.4, + "valid_targets_min": 4291 + }, + { + "epoch": 1.029334730225249, + "grad_norm": 0.462007695468172, + "learning_rate": 3.9732316708444306e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0942046195268631, + "step": 1965, + "valid_targets_mean": 5531.4, + "valid_targets_min": 4050 + }, + { + "epoch": 1.031953902566789, + "grad_norm": 0.46605941489900976, + "learning_rate": 3.9728040155311906e-05, + "loss": 0.2301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12895697355270386, + "step": 1970, + "valid_targets_mean": 7526.6, + "valid_targets_min": 5196 + }, + { + "epoch": 1.034573074908329, + "grad_norm": 0.4004042730515384, + "learning_rate": 3.972372994472741e-05, + "loss": 0.2108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10380735993385315, + "step": 1975, + "valid_targets_mean": 6248.9, + "valid_targets_min": 4749 + }, + { + "epoch": 1.0371922472498691, + "grad_norm": 0.44585808850678105, + "learning_rate": 3.971938608404432e-05, + "loss": 0.2105, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05970005691051483, + "step": 1980, + "valid_targets_mean": 3248.5, + "valid_targets_min": 1908 + }, + { + "epoch": 1.0398114195914092, + "grad_norm": 0.42195346065799316, + "learning_rate": 3.9715008580673595e-05, + "loss": 0.208, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10527674853801727, + "step": 1985, + "valid_targets_mean": 6225.6, + "valid_targets_min": 4808 + }, + { + "epoch": 1.0424305919329493, + "grad_norm": 0.4040996548809383, + "learning_rate": 3.9710597442083555e-05, + "loss": 0.223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10849954187870026, + "step": 1990, + "valid_targets_mean": 5383.0, + "valid_targets_min": 3774 + }, + { + "epoch": 1.0450497642744894, + "grad_norm": 0.42728059622720166, + "learning_rate": 3.970615267579993e-05, + "loss": 0.217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11274106800556183, + "step": 1995, + "valid_targets_mean": 6291.4, + "valid_targets_min": 5335 + }, + { + "epoch": 1.0476689366160294, + "grad_norm": 0.35910712967050745, + "learning_rate": 3.97016742894058e-05, + "loss": 0.1986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11301387846469879, + "step": 2000, + "valid_targets_mean": 7749.1, + "valid_targets_min": 6205 + }, + { + "epoch": 1.0502881089575693, + "grad_norm": 0.41013968417644747, + "learning_rate": 3.969716229054162e-05, + "loss": 0.2061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10402145236730576, + "step": 2005, + "valid_targets_mean": 6882.1, + "valid_targets_min": 4428 + }, + { + "epoch": 1.0529072812991094, + "grad_norm": 0.4342526960220498, + "learning_rate": 3.969261668690518e-05, + "loss": 0.2275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11215817928314209, + "step": 2010, + "valid_targets_mean": 5751.8, + "valid_targets_min": 4458 + }, + { + "epoch": 1.0555264536406495, + "grad_norm": 0.43164225122665206, + "learning_rate": 3.9688037486251615e-05, + "loss": 0.2305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09638996422290802, + "step": 2015, + "valid_targets_mean": 6003.9, + "valid_targets_min": 4458 + }, + { + "epoch": 1.0581456259821895, + "grad_norm": 0.4327578038511442, + "learning_rate": 3.9683424696393355e-05, + "loss": 0.2091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09885101765394211, + "step": 2020, + "valid_targets_mean": 6033.8, + "valid_targets_min": 4066 + }, + { + "epoch": 1.0607647983237296, + "grad_norm": 0.42312508515196295, + "learning_rate": 3.9678778325200164e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11809338629245758, + "step": 2025, + "valid_targets_mean": 6601.0, + "valid_targets_min": 4708 + }, + { + "epoch": 1.0633839706652697, + "grad_norm": 0.4404501074623691, + "learning_rate": 3.967409838059908e-05, + "loss": 0.233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12121831625699997, + "step": 2030, + "valid_targets_mean": 6092.1, + "valid_targets_min": 4789 + }, + { + "epoch": 1.0660031430068098, + "grad_norm": 0.4205967733069136, + "learning_rate": 3.966938487057442e-05, + "loss": 0.2526, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1221914291381836, + "step": 2035, + "valid_targets_mean": 6474.6, + "valid_targets_min": 4224 + }, + { + "epoch": 1.0686223153483498, + "grad_norm": 0.42061079539530033, + "learning_rate": 3.9664637803167775e-05, + "loss": 0.2413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11798917502164841, + "step": 2040, + "valid_targets_mean": 6312.4, + "valid_targets_min": 4617 + }, + { + "epoch": 1.07124148768989, + "grad_norm": 0.4649904202566063, + "learning_rate": 3.965985718647798e-05, + "loss": 0.2184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11309303343296051, + "step": 2045, + "valid_targets_mean": 6333.8, + "valid_targets_min": 4783 + }, + { + "epoch": 1.07386066003143, + "grad_norm": 0.38758470719743, + "learning_rate": 3.965504302866112e-05, + "loss": 0.2201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11203166097402573, + "step": 2050, + "valid_targets_mean": 7380.1, + "valid_targets_min": 6020 + }, + { + "epoch": 1.07647983237297, + "grad_norm": 0.42401428004549396, + "learning_rate": 3.965019533793048e-05, + "loss": 0.2248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.116042360663414, + "step": 2055, + "valid_targets_mean": 6009.0, + "valid_targets_min": 3049 + }, + { + "epoch": 1.0790990047145101, + "grad_norm": 0.4319950319481241, + "learning_rate": 3.964531412255657e-05, + "loss": 0.2212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10162168741226196, + "step": 2060, + "valid_targets_mean": 5973.1, + "valid_targets_min": 4822 + }, + { + "epoch": 1.0817181770560502, + "grad_norm": 0.4354363871830838, + "learning_rate": 3.96403993908671e-05, + "loss": 0.2196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10495162010192871, + "step": 2065, + "valid_targets_mean": 5754.9, + "valid_targets_min": 2910 + }, + { + "epoch": 1.0843373493975903, + "grad_norm": 0.4438673073108337, + "learning_rate": 3.963545115124695e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1243264451622963, + "step": 2070, + "valid_targets_mean": 6414.6, + "valid_targets_min": 4791 + }, + { + "epoch": 1.0869565217391304, + "grad_norm": 0.44515527066321914, + "learning_rate": 3.963046941213818e-05, + "loss": 0.221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11217040568590164, + "step": 2075, + "valid_targets_mean": 5675.9, + "valid_targets_min": 3983 + }, + { + "epoch": 1.0895756940806705, + "grad_norm": 0.3615051018614547, + "learning_rate": 3.962545418203999e-05, + "loss": 0.2012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08994972705841064, + "step": 2080, + "valid_targets_mean": 6975.5, + "valid_targets_min": 4383 + }, + { + "epoch": 1.0921948664222105, + "grad_norm": 0.36746219695139787, + "learning_rate": 3.9620405469508714e-05, + "loss": 0.1933, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10471417009830475, + "step": 2085, + "valid_targets_mean": 6417.1, + "valid_targets_min": 3182 + }, + { + "epoch": 1.0948140387637506, + "grad_norm": 0.4092513667330847, + "learning_rate": 3.9615323283157825e-05, + "loss": 0.2264, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11968555301427841, + "step": 2090, + "valid_targets_mean": 6218.6, + "valid_targets_min": 5077 + }, + { + "epoch": 1.0974332111052907, + "grad_norm": 0.9486185166907464, + "learning_rate": 3.96102076316579e-05, + "loss": 0.2244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11007987707853317, + "step": 2095, + "valid_targets_mean": 4774.9, + "valid_targets_min": 3663 + }, + { + "epoch": 1.1000523834468308, + "grad_norm": 0.44314330051118417, + "learning_rate": 3.96050585237366e-05, + "loss": 0.2126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11296605318784714, + "step": 2100, + "valid_targets_mean": 6316.8, + "valid_targets_min": 4105 + }, + { + "epoch": 1.1026715557883708, + "grad_norm": 0.3813886000772806, + "learning_rate": 3.959987596817868e-05, + "loss": 0.2134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09754054248332977, + "step": 2105, + "valid_targets_mean": 6528.9, + "valid_targets_min": 4537 + }, + { + "epoch": 1.105290728129911, + "grad_norm": 0.40503950573119174, + "learning_rate": 3.9594659973825956e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09919708967208862, + "step": 2110, + "valid_targets_mean": 6416.8, + "valid_targets_min": 2494 + }, + { + "epoch": 1.107909900471451, + "grad_norm": 0.4258719750750446, + "learning_rate": 3.958941054957728e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11352545022964478, + "step": 2115, + "valid_targets_mean": 6070.9, + "valid_targets_min": 4921 + }, + { + "epoch": 1.110529072812991, + "grad_norm": 0.4521245042661743, + "learning_rate": 3.958412770438854e-05, + "loss": 0.2191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11638234555721283, + "step": 2120, + "valid_targets_mean": 6701.4, + "valid_targets_min": 5285 + }, + { + "epoch": 1.1131482451545311, + "grad_norm": 0.3783860004319722, + "learning_rate": 3.957881144727266e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09974069893360138, + "step": 2125, + "valid_targets_mean": 6273.9, + "valid_targets_min": 4464 + }, + { + "epoch": 1.1157674174960712, + "grad_norm": 0.40825091884734466, + "learning_rate": 3.957346178729956e-05, + "loss": 0.2249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0967276468873024, + "step": 2130, + "valid_targets_mean": 5554.8, + "valid_targets_min": 4439 + }, + { + "epoch": 1.1183865898376113, + "grad_norm": 0.3719768450575819, + "learning_rate": 3.956807873359613e-05, + "loss": 0.2007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09004443138837814, + "step": 2135, + "valid_targets_mean": 6043.0, + "valid_targets_min": 4686 + }, + { + "epoch": 1.1210057621791514, + "grad_norm": 0.39907344585700716, + "learning_rate": 3.9562662295346256e-05, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09112632274627686, + "step": 2140, + "valid_targets_mean": 5912.5, + "valid_targets_min": 3381 + }, + { + "epoch": 1.1236249345206915, + "grad_norm": 0.39115142502878664, + "learning_rate": 3.955721248179076e-05, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09805980324745178, + "step": 2145, + "valid_targets_mean": 6495.5, + "valid_targets_min": 5309 + }, + { + "epoch": 1.1262441068622315, + "grad_norm": 0.4027184257805789, + "learning_rate": 3.9551729302227436e-05, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1089298278093338, + "step": 2150, + "valid_targets_mean": 6483.9, + "valid_targets_min": 5161 + }, + { + "epoch": 1.1288632792037716, + "grad_norm": 0.4070778974419733, + "learning_rate": 3.9546212766010956e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10415033251047134, + "step": 2155, + "valid_targets_mean": 5958.9, + "valid_targets_min": 4666 + }, + { + "epoch": 1.1314824515453117, + "grad_norm": 0.6243340872466699, + "learning_rate": 3.9540662882552944e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1264190673828125, + "step": 2160, + "valid_targets_mean": 1691.5, + "valid_targets_min": 1440 + }, + { + "epoch": 1.1341016238868518, + "grad_norm": 0.4193938439336817, + "learning_rate": 3.95350796613219e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09369084239006042, + "step": 2165, + "valid_targets_mean": 5299.1, + "valid_targets_min": 4547 + }, + { + "epoch": 1.1367207962283918, + "grad_norm": 0.4762993785513843, + "learning_rate": 3.9529463111843186e-05, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10690684616565704, + "step": 2170, + "valid_targets_mean": 6003.5, + "valid_targets_min": 5031 + }, + { + "epoch": 1.139339968569932, + "grad_norm": 0.3728037785180108, + "learning_rate": 3.952381324369905e-05, + "loss": 0.2131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09777495265007019, + "step": 2175, + "valid_targets_mean": 8169.6, + "valid_targets_min": 5487 + }, + { + "epoch": 1.141959140911472, + "grad_norm": 0.39460835746169454, + "learning_rate": 3.951813006652856e-05, + "loss": 0.2154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10081788897514343, + "step": 2180, + "valid_targets_mean": 6474.9, + "valid_targets_min": 4253 + }, + { + "epoch": 1.144578313253012, + "grad_norm": 0.41274989886935637, + "learning_rate": 3.951241359002764e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12249577790498734, + "step": 2185, + "valid_targets_mean": 7021.0, + "valid_targets_min": 4990 + }, + { + "epoch": 1.1471974855945521, + "grad_norm": 1.829767434939365, + "learning_rate": 3.950666382394901e-05, + "loss": 0.2097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12657414376735687, + "step": 2190, + "valid_targets_mean": 6356.0, + "valid_targets_min": 4385 + }, + { + "epoch": 1.1498166579360922, + "grad_norm": 0.3857172215805059, + "learning_rate": 3.9500880778102163e-05, + "loss": 0.2162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11194617301225662, + "step": 2195, + "valid_targets_mean": 6815.1, + "valid_targets_min": 5164 + }, + { + "epoch": 1.1524358302776323, + "grad_norm": 0.38697334942397965, + "learning_rate": 3.949506446235341e-05, + "loss": 0.2048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08038259297609329, + "step": 2200, + "valid_targets_mean": 5588.5, + "valid_targets_min": 4098 + }, + { + "epoch": 1.1550550026191724, + "grad_norm": 0.39555401703490717, + "learning_rate": 3.9489214886625794e-05, + "loss": 0.1953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10800637304782867, + "step": 2205, + "valid_targets_mean": 6167.6, + "valid_targets_min": 4801 + }, + { + "epoch": 1.1576741749607125, + "grad_norm": 0.499162137918683, + "learning_rate": 3.948333206089912e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22356413304805756, + "step": 2210, + "valid_targets_mean": 6224.9, + "valid_targets_min": 4772 + }, + { + "epoch": 1.1602933473022525, + "grad_norm": 0.4242115823938771, + "learning_rate": 3.947741599520989e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11829758435487747, + "step": 2215, + "valid_targets_mean": 6549.2, + "valid_targets_min": 5025 + }, + { + "epoch": 1.1629125196437926, + "grad_norm": 0.47155716508802664, + "learning_rate": 3.947146669965136e-05, + "loss": 0.1976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11615519225597382, + "step": 2220, + "valid_targets_mean": 6680.8, + "valid_targets_min": 4818 + }, + { + "epoch": 1.1655316919853327, + "grad_norm": 0.41664141568414825, + "learning_rate": 3.9465484184373445e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10588368773460388, + "step": 2225, + "valid_targets_mean": 5835.1, + "valid_targets_min": 4111 + }, + { + "epoch": 1.1681508643268728, + "grad_norm": 0.44815553437875244, + "learning_rate": 3.945946845958274e-05, + "loss": 0.2148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10050582140684128, + "step": 2230, + "valid_targets_mean": 7029.9, + "valid_targets_min": 2275 + }, + { + "epoch": 1.1707700366684128, + "grad_norm": 0.4073485288669727, + "learning_rate": 3.945341953554251e-05, + "loss": 0.1904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09106802940368652, + "step": 2235, + "valid_targets_mean": 5621.2, + "valid_targets_min": 4826 + }, + { + "epoch": 1.173389209009953, + "grad_norm": 1.0254904680087145, + "learning_rate": 3.944733742257266e-05, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09302625060081482, + "step": 2240, + "valid_targets_mean": 6858.8, + "valid_targets_min": 4710 + }, + { + "epoch": 1.176008381351493, + "grad_norm": 0.36589838556139503, + "learning_rate": 3.9441222131049696e-05, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13208436965942383, + "step": 2245, + "valid_targets_mean": 7012.5, + "valid_targets_min": 5421 + }, + { + "epoch": 1.178627553693033, + "grad_norm": 0.3657121520363973, + "learning_rate": 3.943507367140676e-05, + "loss": 0.1978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09454962611198425, + "step": 2250, + "valid_targets_mean": 6967.6, + "valid_targets_min": 4949 + }, + { + "epoch": 1.1812467260345731, + "grad_norm": 0.47667241204286165, + "learning_rate": 3.9428892054133546e-05, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1025569960474968, + "step": 2255, + "valid_targets_mean": 6348.5, + "valid_targets_min": 4831 + }, + { + "epoch": 1.1838658983761132, + "grad_norm": 0.38318781651180284, + "learning_rate": 3.942267728977635e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12559780478477478, + "step": 2260, + "valid_targets_mean": 6597.2, + "valid_targets_min": 4018 + }, + { + "epoch": 1.1864850707176533, + "grad_norm": 0.35818556666895773, + "learning_rate": 3.941642938893801e-05, + "loss": 0.1942, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07480041682720184, + "step": 2265, + "valid_targets_mean": 6420.9, + "valid_targets_min": 5034 + }, + { + "epoch": 1.1891042430591934, + "grad_norm": 0.4001919066756894, + "learning_rate": 3.9410148362277884e-05, + "loss": 0.2014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11709417402744293, + "step": 2270, + "valid_targets_mean": 6962.4, + "valid_targets_min": 5442 + }, + { + "epoch": 1.1917234154007335, + "grad_norm": 0.4383966402980479, + "learning_rate": 3.940383422051185e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10852479934692383, + "step": 2275, + "valid_targets_mean": 5610.9, + "valid_targets_min": 4609 + }, + { + "epoch": 1.1943425877422735, + "grad_norm": 0.4795485906550113, + "learning_rate": 3.93974869744123e-05, + "loss": 0.2193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10586725175380707, + "step": 2280, + "valid_targets_mean": 5038.4, + "valid_targets_min": 2301 + }, + { + "epoch": 1.1969617600838136, + "grad_norm": 1.1071549594383883, + "learning_rate": 3.939110663480809e-05, + "loss": 0.2522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1425037831068039, + "step": 2285, + "valid_targets_mean": 1241.9, + "valid_targets_min": 788 + }, + { + "epoch": 1.1995809324253537, + "grad_norm": 0.8711537080048511, + "learning_rate": 3.9384693212584524e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12247404456138611, + "step": 2290, + "valid_targets_mean": 1518.8, + "valid_targets_min": 1244 + }, + { + "epoch": 1.2022001047668938, + "grad_norm": 0.9006651247635944, + "learning_rate": 3.9378246718683374e-05, + "loss": 0.2348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12721437215805054, + "step": 2295, + "valid_targets_mean": 1552.4, + "valid_targets_min": 796 + }, + { + "epoch": 1.2048192771084336, + "grad_norm": 0.9196865688823059, + "learning_rate": 3.937176716410281e-05, + "loss": 0.2339, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12445694953203201, + "step": 2300, + "valid_targets_mean": 1342.1, + "valid_targets_min": 697 + }, + { + "epoch": 1.207438449449974, + "grad_norm": 0.7768722340329917, + "learning_rate": 3.9365254559897426e-05, + "loss": 0.2355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13839882612228394, + "step": 2305, + "valid_targets_mean": 1741.4, + "valid_targets_min": 638 + }, + { + "epoch": 1.2100576217915138, + "grad_norm": 0.780544703244386, + "learning_rate": 3.9358708917178194e-05, + "loss": 0.2275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11841660737991333, + "step": 2310, + "valid_targets_mean": 1658.0, + "valid_targets_min": 727 + }, + { + "epoch": 1.212676794133054, + "grad_norm": 0.8410815579947879, + "learning_rate": 3.9352130247112444e-05, + "loss": 0.2385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1197308674454689, + "step": 2315, + "valid_targets_mean": 1523.4, + "valid_targets_min": 1025 + }, + { + "epoch": 1.215295966474594, + "grad_norm": 0.8544834744098359, + "learning_rate": 3.934551856092386e-05, + "loss": 0.2191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12641999125480652, + "step": 2320, + "valid_targets_mean": 1389.6, + "valid_targets_min": 714 + }, + { + "epoch": 1.2179151388161342, + "grad_norm": 0.7884743833205939, + "learning_rate": 3.933887386989245e-05, + "loss": 0.2385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10710620135068893, + "step": 2325, + "valid_targets_mean": 1423.6, + "valid_targets_min": 944 + }, + { + "epoch": 1.220534311157674, + "grad_norm": 0.7590907706954583, + "learning_rate": 3.933219618535454e-05, + "loss": 0.2271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10507871210575104, + "step": 2330, + "valid_targets_mean": 1938.8, + "valid_targets_min": 790 + }, + { + "epoch": 1.2231534834992142, + "grad_norm": 0.9138808275753042, + "learning_rate": 3.932548551870273e-05, + "loss": 0.221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12357285618782043, + "step": 2335, + "valid_targets_mean": 1524.5, + "valid_targets_min": 896 + }, + { + "epoch": 1.2257726558407542, + "grad_norm": 1.2596109128541106, + "learning_rate": 3.9318741881385906e-05, + "loss": 0.2349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10876993834972382, + "step": 2340, + "valid_targets_mean": 1766.2, + "valid_targets_min": 1027 + }, + { + "epoch": 1.2283918281822943, + "grad_norm": 0.7510154769224328, + "learning_rate": 3.9311965284909205e-05, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09838436543941498, + "step": 2345, + "valid_targets_mean": 1471.9, + "valid_targets_min": 763 + }, + { + "epoch": 1.2310110005238344, + "grad_norm": 2.267425830861198, + "learning_rate": 3.930515574083397e-05, + "loss": 0.2122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08276930451393127, + "step": 2350, + "valid_targets_mean": 998.5, + "valid_targets_min": 659 + }, + { + "epoch": 1.2336301728653745, + "grad_norm": 0.7922986210199073, + "learning_rate": 3.929831326077779e-05, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1182938739657402, + "step": 2355, + "valid_targets_mean": 1568.8, + "valid_targets_min": 744 + }, + { + "epoch": 1.2362493452069145, + "grad_norm": 0.9815552033577466, + "learning_rate": 3.929143785641441e-05, + "loss": 0.2225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10616190731525421, + "step": 2360, + "valid_targets_mean": 1345.0, + "valid_targets_min": 874 + }, + { + "epoch": 1.2388685175484546, + "grad_norm": 1.279790543123393, + "learning_rate": 3.9284529539473765e-05, + "loss": 0.2298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1088862493634224, + "step": 2365, + "valid_targets_mean": 1335.8, + "valid_targets_min": 798 + }, + { + "epoch": 1.2414876898899947, + "grad_norm": 0.8607389281618566, + "learning_rate": 3.9277588321741954e-05, + "loss": 0.2314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13361486792564392, + "step": 2370, + "valid_targets_mean": 1635.2, + "valid_targets_min": 802 + }, + { + "epoch": 1.2441068622315348, + "grad_norm": 0.7566096499371763, + "learning_rate": 3.927061421506118e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1237526461482048, + "step": 2375, + "valid_targets_mean": 2001.0, + "valid_targets_min": 940 + }, + { + "epoch": 1.2467260345730748, + "grad_norm": 0.7158055169250003, + "learning_rate": 3.926360723132977e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09132258594036102, + "step": 2380, + "valid_targets_mean": 1358.5, + "valid_targets_min": 776 + }, + { + "epoch": 1.249345206914615, + "grad_norm": 0.9097030122163224, + "learning_rate": 3.9256567382502145e-05, + "loss": 0.2232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12393206357955933, + "step": 2385, + "valid_targets_mean": 1450.4, + "valid_targets_min": 640 + }, + { + "epoch": 1.251964379256155, + "grad_norm": 0.7886022583621429, + "learning_rate": 3.9249494680588786e-05, + "loss": 0.2279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12118816375732422, + "step": 2390, + "valid_targets_mean": 1593.1, + "valid_targets_min": 693 + }, + { + "epoch": 1.254583551597695, + "grad_norm": 0.8425170265940868, + "learning_rate": 3.924238913765624e-05, + "loss": 0.2222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08324533700942993, + "step": 2395, + "valid_targets_mean": 1005.0, + "valid_targets_min": 634 + }, + { + "epoch": 1.2572027239392352, + "grad_norm": 0.9097550639139973, + "learning_rate": 3.9235250765827075e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10276619344949722, + "step": 2400, + "valid_targets_mean": 1430.8, + "valid_targets_min": 793 + }, + { + "epoch": 1.2598218962807752, + "grad_norm": 0.7024450800412536, + "learning_rate": 3.922807957727985e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11671581864356995, + "step": 2405, + "valid_targets_mean": 1673.6, + "valid_targets_min": 563 + }, + { + "epoch": 1.2624410686223153, + "grad_norm": 0.711707962512294, + "learning_rate": 3.9220875584249135e-05, + "loss": 0.2166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1091507077217102, + "step": 2410, + "valid_targets_mean": 1586.8, + "valid_targets_min": 837 + }, + { + "epoch": 1.2650602409638554, + "grad_norm": 0.7611513612472521, + "learning_rate": 3.921363879902546e-05, + "loss": 0.2143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09916854649782181, + "step": 2415, + "valid_targets_mean": 1620.4, + "valid_targets_min": 1107 + }, + { + "epoch": 1.2676794133053955, + "grad_norm": 0.8148682204004281, + "learning_rate": 3.9206369233955304e-05, + "loss": 0.2273, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08057921379804611, + "step": 2420, + "valid_targets_mean": 1290.5, + "valid_targets_min": 680 + }, + { + "epoch": 1.2702985856469355, + "grad_norm": 0.7301626870270491, + "learning_rate": 3.919906690144106e-05, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14655739068984985, + "step": 2425, + "valid_targets_mean": 1927.4, + "valid_targets_min": 933 + }, + { + "epoch": 1.2729177579884756, + "grad_norm": 0.7856678919046975, + "learning_rate": 3.919173181394103e-05, + "loss": 0.2151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11562366783618927, + "step": 2430, + "valid_targets_mean": 1654.8, + "valid_targets_min": 631 + }, + { + "epoch": 1.2755369303300157, + "grad_norm": 0.7329935506191725, + "learning_rate": 3.9184363983969405e-05, + "loss": 0.2188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09420134127140045, + "step": 2435, + "valid_targets_mean": 1519.0, + "valid_targets_min": 804 + }, + { + "epoch": 1.2781561026715558, + "grad_norm": 0.7747224556703134, + "learning_rate": 3.917696342409623e-05, + "loss": 0.2219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11276663839817047, + "step": 2440, + "valid_targets_mean": 1539.5, + "valid_targets_min": 1019 + }, + { + "epoch": 1.2807752750130958, + "grad_norm": 0.7935045402639614, + "learning_rate": 3.9169530146947396e-05, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09454011172056198, + "step": 2445, + "valid_targets_mean": 1112.8, + "valid_targets_min": 856 + }, + { + "epoch": 1.283394447354636, + "grad_norm": 0.7227099872788371, + "learning_rate": 3.916206416520459e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13150903582572937, + "step": 2450, + "valid_targets_mean": 1914.2, + "valid_targets_min": 1168 + }, + { + "epoch": 1.286013619696176, + "grad_norm": 0.8178906209511275, + "learning_rate": 3.915456549160533e-05, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11146867275238037, + "step": 2455, + "valid_targets_mean": 1613.0, + "valid_targets_min": 678 + }, + { + "epoch": 1.288632792037716, + "grad_norm": 0.8549594119877713, + "learning_rate": 3.914703413894289e-05, + "loss": 0.2209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1310146152973175, + "step": 2460, + "valid_targets_mean": 1470.0, + "valid_targets_min": 676 + }, + { + "epoch": 1.2912519643792562, + "grad_norm": 0.7390440723394032, + "learning_rate": 3.9139470120066295e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0856282189488411, + "step": 2465, + "valid_targets_mean": 1185.5, + "valid_targets_min": 794 + }, + { + "epoch": 1.2938711367207962, + "grad_norm": 0.7678281913449002, + "learning_rate": 3.9131873447880296e-05, + "loss": 0.2158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09801091253757477, + "step": 2470, + "valid_targets_mean": 1259.0, + "valid_targets_min": 742 + }, + { + "epoch": 1.2964903090623363, + "grad_norm": 1.0128307309401963, + "learning_rate": 3.912424413534537e-05, + "loss": 0.2113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11717940866947174, + "step": 2475, + "valid_targets_mean": 1792.5, + "valid_targets_min": 741 + }, + { + "epoch": 1.2991094814038764, + "grad_norm": 0.826081955666918, + "learning_rate": 3.9116582195477676e-05, + "loss": 0.2238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.109868124127388, + "step": 2480, + "valid_targets_mean": 1599.6, + "valid_targets_min": 1157 + }, + { + "epoch": 1.3017286537454165, + "grad_norm": 0.7623072638570032, + "learning_rate": 3.910888764134902e-05, + "loss": 0.2033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1621159017086029, + "step": 2485, + "valid_targets_mean": 1989.1, + "valid_targets_min": 862 + }, + { + "epoch": 1.3043478260869565, + "grad_norm": 0.7566942984282442, + "learning_rate": 3.9101160486086874e-05, + "loss": 0.2217, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12045988440513611, + "step": 2490, + "valid_targets_mean": 1428.1, + "valid_targets_min": 884 + }, + { + "epoch": 1.3069669984284966, + "grad_norm": 0.7372353230362683, + "learning_rate": 3.909340074287431e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12013622373342514, + "step": 2495, + "valid_targets_mean": 1570.4, + "valid_targets_min": 1058 + }, + { + "epoch": 1.3095861707700367, + "grad_norm": 0.7653242692253024, + "learning_rate": 3.908560842495002e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11029034107923508, + "step": 2500, + "valid_targets_mean": 1509.6, + "valid_targets_min": 618 + }, + { + "epoch": 1.3122053431115768, + "grad_norm": 0.7104124909195312, + "learning_rate": 3.9077783545608244e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1064552366733551, + "step": 2505, + "valid_targets_mean": 1528.2, + "valid_targets_min": 744 + }, + { + "epoch": 1.3148245154531168, + "grad_norm": 0.7403256903213563, + "learning_rate": 3.906992611819879e-05, + "loss": 0.216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10778047144412994, + "step": 2510, + "valid_targets_mean": 1736.6, + "valid_targets_min": 1071 + }, + { + "epoch": 1.317443687794657, + "grad_norm": 0.856641442215035, + "learning_rate": 3.9062036156127e-05, + "loss": 0.2147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09905392676591873, + "step": 2515, + "valid_targets_mean": 1120.4, + "valid_targets_min": 677 + }, + { + "epoch": 1.320062860136197, + "grad_norm": 0.7784754508322069, + "learning_rate": 3.905411367285371e-05, + "loss": 0.2036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0852539986371994, + "step": 2520, + "valid_targets_mean": 1139.6, + "valid_targets_min": 679 + }, + { + "epoch": 1.322682032477737, + "grad_norm": 0.7647903613204383, + "learning_rate": 3.9046158681895245e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10324622690677643, + "step": 2525, + "valid_targets_mean": 1509.2, + "valid_targets_min": 504 + }, + { + "epoch": 1.3253012048192772, + "grad_norm": 0.7046976894735208, + "learning_rate": 3.90381711968234e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09556830674409866, + "step": 2530, + "valid_targets_mean": 1787.6, + "valid_targets_min": 1347 + }, + { + "epoch": 1.3279203771608172, + "grad_norm": 0.7693468941655391, + "learning_rate": 3.903015123126538e-05, + "loss": 0.2144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0962136909365654, + "step": 2535, + "valid_targets_mean": 1404.1, + "valid_targets_min": 768 + }, + { + "epoch": 1.3305395495023573, + "grad_norm": 0.7368098559006363, + "learning_rate": 3.9022098798903846e-05, + "loss": 0.2151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08603163808584213, + "step": 2540, + "valid_targets_mean": 1428.6, + "valid_targets_min": 1184 + }, + { + "epoch": 1.3331587218438974, + "grad_norm": 0.6549608076966834, + "learning_rate": 3.90140139134768e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08998078852891922, + "step": 2545, + "valid_targets_mean": 1489.2, + "valid_targets_min": 963 + }, + { + "epoch": 1.3357778941854375, + "grad_norm": 0.8080328815245538, + "learning_rate": 3.900589658877765e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11275771260261536, + "step": 2550, + "valid_targets_mean": 1493.4, + "valid_targets_min": 931 + }, + { + "epoch": 1.3383970665269775, + "grad_norm": 0.8368597903329557, + "learning_rate": 3.899774683865513e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1276080310344696, + "step": 2555, + "valid_targets_mean": 1552.0, + "valid_targets_min": 861 + }, + { + "epoch": 1.3410162388685176, + "grad_norm": 0.7348901053832353, + "learning_rate": 3.898956467701331e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09074932336807251, + "step": 2560, + "valid_targets_mean": 1285.2, + "valid_targets_min": 797 + }, + { + "epoch": 1.3436354112100577, + "grad_norm": 0.649394121209089, + "learning_rate": 3.8981350117811525e-05, + "loss": 0.2136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1057843267917633, + "step": 2565, + "valid_targets_mean": 1849.1, + "valid_targets_min": 1220 + }, + { + "epoch": 1.3462545835515978, + "grad_norm": 0.8070613798353503, + "learning_rate": 3.897310317506441e-05, + "loss": 0.1885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09375861287117004, + "step": 2570, + "valid_targets_mean": 1488.0, + "valid_targets_min": 712 + }, + { + "epoch": 1.3488737558931378, + "grad_norm": 0.7574271077544024, + "learning_rate": 3.8964823862841855e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11036104708909988, + "step": 2575, + "valid_targets_mean": 1415.0, + "valid_targets_min": 605 + }, + { + "epoch": 1.351492928234678, + "grad_norm": 0.6396674537277155, + "learning_rate": 3.8956512195268936e-05, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07955646514892578, + "step": 2580, + "valid_targets_mean": 1473.4, + "valid_targets_min": 746 + }, + { + "epoch": 1.3541121005762178, + "grad_norm": 0.8063620884542072, + "learning_rate": 3.894816818652598e-05, + "loss": 0.1953, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08809489011764526, + "step": 2585, + "valid_targets_mean": 1135.9, + "valid_targets_min": 723 + }, + { + "epoch": 1.356731272917758, + "grad_norm": 0.7002179582641632, + "learning_rate": 3.893979185084844e-05, + "loss": 0.212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10178270936012268, + "step": 2590, + "valid_targets_mean": 1624.0, + "valid_targets_min": 855 + }, + { + "epoch": 1.359350445259298, + "grad_norm": 0.7212980586239315, + "learning_rate": 3.893138320252698e-05, + "loss": 0.2072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09820833802223206, + "step": 2595, + "valid_targets_mean": 1534.1, + "valid_targets_min": 1094 + }, + { + "epoch": 1.3619696176008382, + "grad_norm": 0.6742047855391174, + "learning_rate": 3.8922942255907316e-05, + "loss": 0.2109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10583124309778214, + "step": 2600, + "valid_targets_mean": 1657.8, + "valid_targets_min": 803 + }, + { + "epoch": 1.364588789942378, + "grad_norm": 0.7608559559573908, + "learning_rate": 3.891446902539033e-05, + "loss": 0.2112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10718117654323578, + "step": 2605, + "valid_targets_mean": 1427.2, + "valid_targets_min": 740 + }, + { + "epoch": 1.3672079622839184, + "grad_norm": 0.754692382140475, + "learning_rate": 3.8905963525431975e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09368541836738586, + "step": 2610, + "valid_targets_mean": 1393.8, + "valid_targets_min": 1180 + }, + { + "epoch": 1.3698271346254582, + "grad_norm": 0.6647269540981037, + "learning_rate": 3.889742577054321e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1116383969783783, + "step": 2615, + "valid_targets_mean": 1632.5, + "valid_targets_min": 1276 + }, + { + "epoch": 1.3724463069669985, + "grad_norm": 0.7832902164275183, + "learning_rate": 3.888885577529009e-05, + "loss": 0.2127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10254848003387451, + "step": 2620, + "valid_targets_mean": 1496.6, + "valid_targets_min": 812 + }, + { + "epoch": 1.3750654793085384, + "grad_norm": 0.7120527153104592, + "learning_rate": 3.888025355429362e-05, + "loss": 0.216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13120970129966736, + "step": 2625, + "valid_targets_mean": 1708.0, + "valid_targets_min": 1041 + }, + { + "epoch": 1.3776846516500787, + "grad_norm": 0.7866373151345482, + "learning_rate": 3.8871619122229816e-05, + "loss": 0.2123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10362714529037476, + "step": 2630, + "valid_targets_mean": 1354.1, + "valid_targets_min": 777 + }, + { + "epoch": 1.3803038239916186, + "grad_norm": 0.7562407163473897, + "learning_rate": 3.886295249382964e-05, + "loss": 0.2124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09727167338132858, + "step": 2635, + "valid_targets_mean": 1665.0, + "valid_targets_min": 1283 + }, + { + "epoch": 1.3829229963331588, + "grad_norm": 0.7463958312608147, + "learning_rate": 3.885425368387898e-05, + "loss": 0.2017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08751638978719711, + "step": 2640, + "valid_targets_mean": 1153.0, + "valid_targets_min": 620 + }, + { + "epoch": 1.3855421686746987, + "grad_norm": 0.6951198861554263, + "learning_rate": 3.8845522707218626e-05, + "loss": 0.21, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0867944061756134, + "step": 2645, + "valid_targets_mean": 1386.5, + "valid_targets_min": 656 + }, + { + "epoch": 1.388161341016239, + "grad_norm": 0.7109993087184034, + "learning_rate": 3.8836759578744255e-05, + "loss": 0.2125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09662780165672302, + "step": 2650, + "valid_targets_mean": 1611.2, + "valid_targets_min": 984 + }, + { + "epoch": 1.3907805133577789, + "grad_norm": 0.6386859379838086, + "learning_rate": 3.882796431340639e-05, + "loss": 0.201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09424383193254471, + "step": 2655, + "valid_targets_mean": 1801.8, + "valid_targets_min": 682 + }, + { + "epoch": 1.393399685699319, + "grad_norm": 0.6664677477090022, + "learning_rate": 3.881913692621039e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08829391002655029, + "step": 2660, + "valid_targets_mean": 1508.9, + "valid_targets_min": 850 + }, + { + "epoch": 1.396018858040859, + "grad_norm": 0.7376237184286827, + "learning_rate": 3.8810277432216404e-05, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10379143059253693, + "step": 2665, + "valid_targets_mean": 1573.5, + "valid_targets_min": 747 + }, + { + "epoch": 1.398638030382399, + "grad_norm": 0.7023603515132003, + "learning_rate": 3.880138584653937e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08785136044025421, + "step": 2670, + "valid_targets_mean": 1439.5, + "valid_targets_min": 933 + }, + { + "epoch": 1.4012572027239392, + "grad_norm": 0.7556951096862679, + "learning_rate": 3.879246218434896e-05, + "loss": 0.2045, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10422730445861816, + "step": 2675, + "valid_targets_mean": 1379.0, + "valid_targets_min": 799 + }, + { + "epoch": 1.4038763750654792, + "grad_norm": 0.6754023756160091, + "learning_rate": 3.8783506460869605e-05, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08867405354976654, + "step": 2680, + "valid_targets_mean": 1371.1, + "valid_targets_min": 1074 + }, + { + "epoch": 1.4064955474070193, + "grad_norm": 0.6756199407984119, + "learning_rate": 3.877451869138039e-05, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10129301995038986, + "step": 2685, + "valid_targets_mean": 1555.4, + "valid_targets_min": 741 + }, + { + "epoch": 1.4091147197485594, + "grad_norm": 0.7554397099830699, + "learning_rate": 3.876549889121511e-05, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10054688900709152, + "step": 2690, + "valid_targets_mean": 1237.8, + "valid_targets_min": 720 + }, + { + "epoch": 1.4117338920900995, + "grad_norm": 0.9321691670258012, + "learning_rate": 3.875644707576218e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10564795136451721, + "step": 2695, + "valid_targets_mean": 1458.6, + "valid_targets_min": 610 + }, + { + "epoch": 1.4143530644316396, + "grad_norm": 0.6583706953082205, + "learning_rate": 3.8747363260464644e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1348811686038971, + "step": 2700, + "valid_targets_mean": 1887.2, + "valid_targets_min": 713 + }, + { + "epoch": 1.4169722367731796, + "grad_norm": 0.6799342532777773, + "learning_rate": 3.873824746082016e-05, + "loss": 0.2083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10016189515590668, + "step": 2705, + "valid_targets_mean": 1466.2, + "valid_targets_min": 874 + }, + { + "epoch": 1.4195914091147197, + "grad_norm": 0.6864022857822357, + "learning_rate": 3.872909969238092e-05, + "loss": 0.2032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10612594336271286, + "step": 2710, + "valid_targets_mean": 1510.6, + "valid_targets_min": 765 + }, + { + "epoch": 1.4222105814562598, + "grad_norm": 0.7032653557997202, + "learning_rate": 3.871991997075368e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09451894462108612, + "step": 2715, + "valid_targets_mean": 1297.1, + "valid_targets_min": 589 + }, + { + "epoch": 1.4248297537977999, + "grad_norm": 0.681842309516101, + "learning_rate": 3.87107083115997e-05, + "loss": 0.215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1271815448999405, + "step": 2720, + "valid_targets_mean": 1671.8, + "valid_targets_min": 1230 + }, + { + "epoch": 1.42744892613934, + "grad_norm": 0.723773119779326, + "learning_rate": 3.8701464730634734e-05, + "loss": 0.302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16962957382202148, + "step": 2725, + "valid_targets_mean": 4792.6, + "valid_targets_min": 702 + }, + { + "epoch": 1.43006809848088, + "grad_norm": 0.539319266583007, + "learning_rate": 3.8692189243629e-05, + "loss": 0.3238, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17272919416427612, + "step": 2730, + "valid_targets_mean": 5322.8, + "valid_targets_min": 1005 + }, + { + "epoch": 1.43268727082242, + "grad_norm": 0.4881542042305075, + "learning_rate": 3.868288186640714e-05, + "loss": 0.3099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11419947445392609, + "step": 2735, + "valid_targets_mean": 3584.6, + "valid_targets_min": 2054 + }, + { + "epoch": 1.4353064431639602, + "grad_norm": 0.563625820979566, + "learning_rate": 3.86735426148482e-05, + "loss": 0.3114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1368965059518814, + "step": 2740, + "valid_targets_mean": 3685.6, + "valid_targets_min": 1751 + }, + { + "epoch": 1.4379256155055002, + "grad_norm": 0.46302839447549154, + "learning_rate": 3.8664171504885625e-05, + "loss": 0.3034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1875678449869156, + "step": 2745, + "valid_targets_mean": 5944.2, + "valid_targets_min": 1970 + }, + { + "epoch": 1.4405447878470403, + "grad_norm": 0.442681042930155, + "learning_rate": 3.86547685525072e-05, + "loss": 0.3197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15010172128677368, + "step": 2750, + "valid_targets_mean": 5916.0, + "valid_targets_min": 920 + }, + { + "epoch": 1.4431639601885804, + "grad_norm": 0.4691950139701078, + "learning_rate": 3.864533377375503e-05, + "loss": 0.2992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14268054068088531, + "step": 2755, + "valid_targets_mean": 6475.4, + "valid_targets_min": 2231 + }, + { + "epoch": 1.4457831325301205, + "grad_norm": 0.5045779150420908, + "learning_rate": 3.8635867184725516e-05, + "loss": 0.3024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11577323079109192, + "step": 2760, + "valid_targets_mean": 4043.4, + "valid_targets_min": 467 + }, + { + "epoch": 1.4484023048716606, + "grad_norm": 0.474244353293432, + "learning_rate": 3.862636880156936e-05, + "loss": 0.3143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16245116293430328, + "step": 2765, + "valid_targets_mean": 5549.8, + "valid_targets_min": 3440 + }, + { + "epoch": 1.4510214772132006, + "grad_norm": 0.5156206839232728, + "learning_rate": 3.8616838640491474e-05, + "loss": 0.313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1551964282989502, + "step": 2770, + "valid_targets_mean": 4736.9, + "valid_targets_min": 1697 + }, + { + "epoch": 1.4536406495547407, + "grad_norm": 0.5532630078425921, + "learning_rate": 3.860727671775099e-05, + "loss": 0.3077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18160893023014069, + "step": 2775, + "valid_targets_mean": 4838.2, + "valid_targets_min": 1223 + }, + { + "epoch": 1.4562598218962808, + "grad_norm": 0.5049358800424307, + "learning_rate": 3.859768304966125e-05, + "loss": 0.3057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15451818704605103, + "step": 2780, + "valid_targets_mean": 5466.9, + "valid_targets_min": 1552 + }, + { + "epoch": 1.4588789942378209, + "grad_norm": 0.4990178646017903, + "learning_rate": 3.8588057652589724e-05, + "loss": 0.3111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17876386642456055, + "step": 2785, + "valid_targets_mean": 5166.0, + "valid_targets_min": 1546 + }, + { + "epoch": 1.461498166579361, + "grad_norm": 0.5267191732540031, + "learning_rate": 3.857840054295804e-05, + "loss": 0.3022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1500575840473175, + "step": 2790, + "valid_targets_mean": 3848.2, + "valid_targets_min": 1957 + }, + { + "epoch": 1.464117338920901, + "grad_norm": 0.5625065784003082, + "learning_rate": 3.8568711737241925e-05, + "loss": 0.3175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14559558033943176, + "step": 2795, + "valid_targets_mean": 3487.5, + "valid_targets_min": 1538 + }, + { + "epoch": 1.466736511262441, + "grad_norm": 0.619436426859455, + "learning_rate": 3.855899125197116e-05, + "loss": 0.3113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15097582340240479, + "step": 2800, + "valid_targets_mean": 3397.2, + "valid_targets_min": 1645 + }, + { + "epoch": 1.4693556836039812, + "grad_norm": 0.6088634021002824, + "learning_rate": 3.8549239103729606e-05, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1317932903766632, + "step": 2805, + "valid_targets_mean": 2859.4, + "valid_targets_min": 935 + }, + { + "epoch": 1.4719748559455212, + "grad_norm": 0.6156222622020914, + "learning_rate": 3.8539455309155116e-05, + "loss": 0.3226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16073517501354218, + "step": 2810, + "valid_targets_mean": 3490.6, + "valid_targets_min": 1269 + }, + { + "epoch": 1.4745940282870613, + "grad_norm": 0.6147357458325968, + "learning_rate": 3.852963988493955e-05, + "loss": 0.3178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14917299151420593, + "step": 2815, + "valid_targets_mean": 2860.0, + "valid_targets_min": 1173 + }, + { + "epoch": 1.4772132006286014, + "grad_norm": 0.545960724105271, + "learning_rate": 3.851979284782873e-05, + "loss": 0.313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13690051436424255, + "step": 2820, + "valid_targets_mean": 3480.9, + "valid_targets_min": 1392 + }, + { + "epoch": 1.4798323729701415, + "grad_norm": 0.7551418345180866, + "learning_rate": 3.850991421462241e-05, + "loss": 0.3146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17518110573291779, + "step": 2825, + "valid_targets_mean": 3878.6, + "valid_targets_min": 1883 + }, + { + "epoch": 1.4824515453116816, + "grad_norm": 0.5855841827996562, + "learning_rate": 3.850000400217424e-05, + "loss": 0.3053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.176192969083786, + "step": 2830, + "valid_targets_mean": 3721.4, + "valid_targets_min": 2106 + }, + { + "epoch": 1.4850707176532216, + "grad_norm": 0.4804187507244692, + "learning_rate": 3.849006222739175e-05, + "loss": 0.2823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10946094989776611, + "step": 2835, + "valid_targets_mean": 3868.5, + "valid_targets_min": 2222 + }, + { + "epoch": 1.4876898899947617, + "grad_norm": 0.6022056043484764, + "learning_rate": 3.848008890723633e-05, + "loss": 0.3101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1461934596300125, + "step": 2840, + "valid_targets_mean": 3010.1, + "valid_targets_min": 958 + }, + { + "epoch": 1.4903090623363018, + "grad_norm": 0.5365693351133142, + "learning_rate": 3.8470084058723174e-05, + "loss": 0.3068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1488455981016159, + "step": 2845, + "valid_targets_mean": 3789.1, + "valid_targets_min": 771 + }, + { + "epoch": 1.4929282346778419, + "grad_norm": 0.4975267755885652, + "learning_rate": 3.846004769892128e-05, + "loss": 0.3129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10034919530153275, + "step": 2850, + "valid_targets_mean": 3578.9, + "valid_targets_min": 1774 + }, + { + "epoch": 1.495547407019382, + "grad_norm": 0.6261290783584289, + "learning_rate": 3.844997984495338e-05, + "loss": 0.3008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1437961608171463, + "step": 2855, + "valid_targets_mean": 2545.8, + "valid_targets_min": 997 + }, + { + "epoch": 1.498166579360922, + "grad_norm": 0.5617655886164745, + "learning_rate": 3.843988051399598e-05, + "loss": 0.2952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15878400206565857, + "step": 2860, + "valid_targets_mean": 3667.1, + "valid_targets_min": 1741 + }, + { + "epoch": 1.500785751702462, + "grad_norm": 0.5907520274999801, + "learning_rate": 3.842974972327925e-05, + "loss": 0.3095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1543917953968048, + "step": 2865, + "valid_targets_mean": 3606.1, + "valid_targets_min": 863 + }, + { + "epoch": 1.503404924044002, + "grad_norm": 0.5480379676249457, + "learning_rate": 3.841958749008704e-05, + "loss": 0.3018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1570451855659485, + "step": 2870, + "valid_targets_mean": 3876.9, + "valid_targets_min": 1854 + }, + { + "epoch": 1.5060240963855422, + "grad_norm": 0.4999441685313448, + "learning_rate": 3.8409393831756857e-05, + "loss": 0.2924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1467728465795517, + "step": 2875, + "valid_targets_mean": 4763.2, + "valid_targets_min": 2040 + }, + { + "epoch": 1.508643268727082, + "grad_norm": 0.6046460788215894, + "learning_rate": 3.839916876567981e-05, + "loss": 0.3043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14319008588790894, + "step": 2880, + "valid_targets_mean": 3521.9, + "valid_targets_min": 1036 + }, + { + "epoch": 1.5112624410686224, + "grad_norm": 0.5262218875177235, + "learning_rate": 3.838891230930061e-05, + "loss": 0.3052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15236154198646545, + "step": 2885, + "valid_targets_mean": 4304.8, + "valid_targets_min": 1279 + }, + { + "epoch": 1.5138816134101623, + "grad_norm": 0.5989480295635137, + "learning_rate": 3.837862448011748e-05, + "loss": 0.2976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1572301983833313, + "step": 2890, + "valid_targets_mean": 3600.0, + "valid_targets_min": 1798 + }, + { + "epoch": 1.5165007857517026, + "grad_norm": 0.6956951178853307, + "learning_rate": 3.836830529568222e-05, + "loss": 0.3109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13771530985832214, + "step": 2895, + "valid_targets_mean": 3067.2, + "valid_targets_min": 1446 + }, + { + "epoch": 1.5191199580932424, + "grad_norm": 0.5219029869297505, + "learning_rate": 3.835795477360008e-05, + "loss": 0.3034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15468549728393555, + "step": 2900, + "valid_targets_mean": 4183.2, + "valid_targets_min": 1270 + }, + { + "epoch": 1.5217391304347827, + "grad_norm": 0.49674103390662777, + "learning_rate": 3.834757293152981e-05, + "loss": 0.3036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15922504663467407, + "step": 2905, + "valid_targets_mean": 5485.8, + "valid_targets_min": 1216 + }, + { + "epoch": 1.5243583027763226, + "grad_norm": 0.5890978574942454, + "learning_rate": 3.833715978718357e-05, + "loss": 0.2926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17861109972000122, + "step": 2910, + "valid_targets_mean": 3994.0, + "valid_targets_min": 2028 + }, + { + "epoch": 1.5269774751178629, + "grad_norm": 0.6083581092573603, + "learning_rate": 3.832671535832693e-05, + "loss": 0.3044, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17515309154987335, + "step": 2915, + "valid_targets_mean": 3700.6, + "valid_targets_min": 2329 + }, + { + "epoch": 1.5295966474594027, + "grad_norm": 0.6116613838205515, + "learning_rate": 3.831623966277884e-05, + "loss": 0.3022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.19010843336582184, + "step": 2920, + "valid_targets_mean": 4075.2, + "valid_targets_min": 1363 + }, + { + "epoch": 1.532215819800943, + "grad_norm": 0.5493174110776619, + "learning_rate": 3.830573271841157e-05, + "loss": 0.293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13848401606082916, + "step": 2925, + "valid_targets_mean": 3560.8, + "valid_targets_min": 1087 + }, + { + "epoch": 1.5348349921424829, + "grad_norm": 0.5750474544151101, + "learning_rate": 3.829519454315075e-05, + "loss": 0.3092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1505936086177826, + "step": 2930, + "valid_targets_mean": 3860.1, + "valid_targets_min": 2483 + }, + { + "epoch": 1.5374541644840232, + "grad_norm": 0.5484449082367139, + "learning_rate": 3.828462515497524e-05, + "loss": 0.301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1770758330821991, + "step": 2935, + "valid_targets_mean": 4583.2, + "valid_targets_min": 2316 + }, + { + "epoch": 1.540073336825563, + "grad_norm": 0.5100701550282972, + "learning_rate": 3.8274024571917194e-05, + "loss": 0.294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14798647165298462, + "step": 2940, + "valid_targets_mean": 4198.9, + "valid_targets_min": 2022 + }, + { + "epoch": 1.5426925091671033, + "grad_norm": 1.0422589684046593, + "learning_rate": 3.826339281206195e-05, + "loss": 0.2848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11863376200199127, + "step": 2945, + "valid_targets_mean": 3228.9, + "valid_targets_min": 1294 + }, + { + "epoch": 1.5453116815086432, + "grad_norm": 0.6340815399612624, + "learning_rate": 3.8252729893548066e-05, + "loss": 0.31, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17853164672851562, + "step": 2950, + "valid_targets_mean": 3846.8, + "valid_targets_min": 2716 + }, + { + "epoch": 1.5479308538501835, + "grad_norm": 0.6326595080955764, + "learning_rate": 3.824203583456723e-05, + "loss": 0.2909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14683550596237183, + "step": 2955, + "valid_targets_mean": 3211.6, + "valid_targets_min": 2060 + }, + { + "epoch": 1.5505500261917233, + "grad_norm": 0.6255563894990052, + "learning_rate": 3.8231310653364293e-05, + "loss": 0.3049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17071311175823212, + "step": 2960, + "valid_targets_mean": 3320.4, + "valid_targets_min": 2407 + }, + { + "epoch": 1.5531691985332636, + "grad_norm": 0.5570154177700162, + "learning_rate": 3.822055436823717e-05, + "loss": 0.3053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15477371215820312, + "step": 2965, + "valid_targets_mean": 4165.1, + "valid_targets_min": 1533 + }, + { + "epoch": 1.5557883708748035, + "grad_norm": 0.5287746380068383, + "learning_rate": 3.8209766997536867e-05, + "loss": 0.3069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1815161108970642, + "step": 2970, + "valid_targets_mean": 5201.2, + "valid_targets_min": 2189 + }, + { + "epoch": 1.5584075432163438, + "grad_norm": 0.5624454709812003, + "learning_rate": 3.81989485596674e-05, + "loss": 0.2951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18315951526165009, + "step": 2975, + "valid_targets_mean": 4193.1, + "valid_targets_min": 1328 + }, + { + "epoch": 1.5610267155578836, + "grad_norm": 0.5769043545634327, + "learning_rate": 3.81880990730858e-05, + "loss": 0.3101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.21385367214679718, + "step": 2980, + "valid_targets_mean": 4500.4, + "valid_targets_min": 1708 + }, + { + "epoch": 1.563645887899424, + "grad_norm": 0.5585266847887569, + "learning_rate": 3.817721855630209e-05, + "loss": 0.2863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11502377688884735, + "step": 2985, + "valid_targets_mean": 3494.0, + "valid_targets_min": 1912 + }, + { + "epoch": 1.5662650602409638, + "grad_norm": 0.6180984525796841, + "learning_rate": 3.816630702787919e-05, + "loss": 0.2879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17739161849021912, + "step": 2990, + "valid_targets_mean": 4145.2, + "valid_targets_min": 2606 + }, + { + "epoch": 1.568884232582504, + "grad_norm": 0.5273350847958702, + "learning_rate": 3.815536450643296e-05, + "loss": 0.2842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1415066123008728, + "step": 2995, + "valid_targets_mean": 4189.0, + "valid_targets_min": 1378 + }, + { + "epoch": 1.571503404924044, + "grad_norm": 0.5768743256668842, + "learning_rate": 3.814439101063212e-05, + "loss": 0.2955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.171949565410614, + "step": 3000, + "valid_targets_mean": 3983.4, + "valid_targets_min": 1219 + }, + { + "epoch": 1.574122577265584, + "grad_norm": 0.5516576810524875, + "learning_rate": 3.8133386559198255e-05, + "loss": 0.2995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15621724724769592, + "step": 3005, + "valid_targets_mean": 3997.0, + "valid_targets_min": 2631 + }, + { + "epoch": 1.576741749607124, + "grad_norm": 0.584823736207226, + "learning_rate": 3.812235117090574e-05, + "loss": 0.2946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18098807334899902, + "step": 3010, + "valid_targets_mean": 4167.4, + "valid_targets_min": 1423 + }, + { + "epoch": 1.5793609219486642, + "grad_norm": 0.524222828668809, + "learning_rate": 3.8111284864581734e-05, + "loss": 0.2931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11262401938438416, + "step": 3015, + "valid_targets_mean": 3381.9, + "valid_targets_min": 2320 + }, + { + "epoch": 1.5819800942902043, + "grad_norm": 0.5787836885402088, + "learning_rate": 3.8100187659106155e-05, + "loss": 0.3024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15317469835281372, + "step": 3020, + "valid_targets_mean": 3174.9, + "valid_targets_min": 1188 + }, + { + "epoch": 1.5845992666317443, + "grad_norm": 0.5330078543949884, + "learning_rate": 3.808905957341164e-05, + "loss": 0.3012, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11782100796699524, + "step": 3025, + "valid_targets_mean": 3378.6, + "valid_targets_min": 618 + }, + { + "epoch": 1.5872184389732844, + "grad_norm": 0.6410441192601284, + "learning_rate": 3.807790062648349e-05, + "loss": 0.2986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15787401795387268, + "step": 3030, + "valid_targets_mean": 2988.1, + "valid_targets_min": 1364 + }, + { + "epoch": 1.5898376113148245, + "grad_norm": 0.6020007715891702, + "learning_rate": 3.806671083735967e-05, + "loss": 0.2902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12509912252426147, + "step": 3035, + "valid_targets_mean": 3199.1, + "valid_targets_min": 1397 + }, + { + "epoch": 1.5924567836563646, + "grad_norm": 0.49479023303795283, + "learning_rate": 3.805549022513077e-05, + "loss": 0.2806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16238363087177277, + "step": 3040, + "valid_targets_mean": 5348.0, + "valid_targets_min": 1778 + }, + { + "epoch": 1.5950759559979046, + "grad_norm": 0.5659948319245839, + "learning_rate": 3.804423880893995e-05, + "loss": 0.2902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14986366033554077, + "step": 3045, + "valid_targets_mean": 3444.9, + "valid_targets_min": 621 + }, + { + "epoch": 1.5976951283394447, + "grad_norm": 0.5411459678777185, + "learning_rate": 3.803295660798294e-05, + "loss": 0.2773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12254117429256439, + "step": 3050, + "valid_targets_mean": 3872.9, + "valid_targets_min": 1290 + }, + { + "epoch": 1.6003143006809848, + "grad_norm": 0.645741556185479, + "learning_rate": 3.802164364150799e-05, + "loss": 0.2822, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1370348036289215, + "step": 3055, + "valid_targets_mean": 2546.6, + "valid_targets_min": 642 + }, + { + "epoch": 1.6029334730225249, + "grad_norm": 0.6458131486074098, + "learning_rate": 3.801029992881582e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14946508407592773, + "step": 3060, + "valid_targets_mean": 3428.2, + "valid_targets_min": 1978 + }, + { + "epoch": 1.605552645364065, + "grad_norm": 0.6045302449215129, + "learning_rate": 3.799892548925963e-05, + "loss": 0.2963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1920478641986847, + "step": 3065, + "valid_targets_mean": 4160.0, + "valid_targets_min": 2642 + }, + { + "epoch": 1.608171817705605, + "grad_norm": 0.5849900786759074, + "learning_rate": 3.798752034224502e-05, + "loss": 0.3059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11567557603120804, + "step": 3070, + "valid_targets_mean": 2937.1, + "valid_targets_min": 1510 + }, + { + "epoch": 1.610790990047145, + "grad_norm": 0.6662578081675633, + "learning_rate": 3.7976084507229993e-05, + "loss": 0.2911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15177395939826965, + "step": 3075, + "valid_targets_mean": 2744.2, + "valid_targets_min": 987 + }, + { + "epoch": 1.6134101623886852, + "grad_norm": 0.5689753936048727, + "learning_rate": 3.79646180037249e-05, + "loss": 0.2925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13514591753482819, + "step": 3080, + "valid_targets_mean": 3638.4, + "valid_targets_min": 1428 + }, + { + "epoch": 1.6160293347302253, + "grad_norm": 0.5478681478403845, + "learning_rate": 3.7953120851292434e-05, + "loss": 0.2875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16971135139465332, + "step": 3085, + "valid_targets_mean": 4731.5, + "valid_targets_min": 1700 + }, + { + "epoch": 1.6186485070717653, + "grad_norm": 0.45081441832834396, + "learning_rate": 3.794159306954754e-05, + "loss": 0.2803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13726529479026794, + "step": 3090, + "valid_targets_mean": 5699.6, + "valid_targets_min": 1767 + }, + { + "epoch": 1.6212676794133054, + "grad_norm": 0.4721150801653287, + "learning_rate": 3.793003467815746e-05, + "loss": 0.2903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12897364795207977, + "step": 3095, + "valid_targets_mean": 4101.5, + "valid_targets_min": 2684 + }, + { + "epoch": 1.6238868517548455, + "grad_norm": 0.6446904616112241, + "learning_rate": 3.791844569684163e-05, + "loss": 0.2818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13025550544261932, + "step": 3100, + "valid_targets_mean": 2761.2, + "valid_targets_min": 1542 + }, + { + "epoch": 1.6265060240963856, + "grad_norm": 0.577713579041728, + "learning_rate": 3.790682614537168e-05, + "loss": 0.2893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11840159446001053, + "step": 3105, + "valid_targets_mean": 2765.8, + "valid_targets_min": 1476 + }, + { + "epoch": 1.6291251964379256, + "grad_norm": 0.5340383486053628, + "learning_rate": 3.7895176043571406e-05, + "loss": 0.2859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.137783020734787, + "step": 3110, + "valid_targets_mean": 3884.5, + "valid_targets_min": 1711 + }, + { + "epoch": 1.6317443687794657, + "grad_norm": 0.5930009819293217, + "learning_rate": 3.7883495411316716e-05, + "loss": 0.2978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14755618572235107, + "step": 3115, + "valid_targets_mean": 3235.6, + "valid_targets_min": 1669 + }, + { + "epoch": 1.6343635411210058, + "grad_norm": 0.6278827317988277, + "learning_rate": 3.787178426853561e-05, + "loss": 0.3014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14962062239646912, + "step": 3120, + "valid_targets_mean": 3388.5, + "valid_targets_min": 1624 + }, + { + "epoch": 1.6369827134625459, + "grad_norm": 0.6206369838121335, + "learning_rate": 3.786004263520812e-05, + "loss": 0.2934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1683076173067093, + "step": 3125, + "valid_targets_mean": 3380.9, + "valid_targets_min": 1676 + }, + { + "epoch": 1.639601885804086, + "grad_norm": 0.5776621725110782, + "learning_rate": 3.784827053136634e-05, + "loss": 0.3115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12504719197750092, + "step": 3130, + "valid_targets_mean": 3148.4, + "valid_targets_min": 1500 + }, + { + "epoch": 1.642221058145626, + "grad_norm": 0.5389410503186015, + "learning_rate": 3.783646797709431e-05, + "loss": 0.2848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1391826570034027, + "step": 3135, + "valid_targets_mean": 4343.5, + "valid_targets_min": 2556 + }, + { + "epoch": 1.644840230487166, + "grad_norm": 0.6290315607215606, + "learning_rate": 3.782463499252804e-05, + "loss": 0.3051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12004496157169342, + "step": 3140, + "valid_targets_mean": 3299.5, + "valid_targets_min": 1241 + }, + { + "epoch": 1.6474594028287062, + "grad_norm": 0.6007650816168316, + "learning_rate": 3.781277159785544e-05, + "loss": 0.3091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1847737729549408, + "step": 3145, + "valid_targets_mean": 3547.0, + "valid_targets_min": 992 + }, + { + "epoch": 1.6500785751702463, + "grad_norm": 0.5950666854019176, + "learning_rate": 3.780087781331632e-05, + "loss": 0.2893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15870344638824463, + "step": 3150, + "valid_targets_mean": 4328.6, + "valid_targets_min": 1133 + }, + { + "epoch": 1.6526977475117863, + "grad_norm": 0.6140209516606467, + "learning_rate": 3.7788953659202346e-05, + "loss": 0.293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15737590193748474, + "step": 3155, + "valid_targets_mean": 4202.6, + "valid_targets_min": 2369 + }, + { + "epoch": 1.6553169198533264, + "grad_norm": 0.5420678882929577, + "learning_rate": 3.7776999155856954e-05, + "loss": 0.2973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12105736136436462, + "step": 3160, + "valid_targets_mean": 3563.0, + "valid_targets_min": 1919 + }, + { + "epoch": 1.6579360921948663, + "grad_norm": 0.7035656048111731, + "learning_rate": 3.776501432367539e-05, + "loss": 0.3028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16013208031654358, + "step": 3165, + "valid_targets_mean": 3975.5, + "valid_targets_min": 2902 + }, + { + "epoch": 1.6605552645364066, + "grad_norm": 0.6348615081088321, + "learning_rate": 3.775299918310466e-05, + "loss": 0.2989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18218766152858734, + "step": 3170, + "valid_targets_mean": 2975.5, + "valid_targets_min": 617 + }, + { + "epoch": 1.6631744368779464, + "grad_norm": 0.5318648254326336, + "learning_rate": 3.774095375464342e-05, + "loss": 0.2711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1221066266298294, + "step": 3175, + "valid_targets_mean": 3290.4, + "valid_targets_min": 1349 + }, + { + "epoch": 1.6657936092194867, + "grad_norm": 0.5747596870174381, + "learning_rate": 3.772887805884207e-05, + "loss": 0.2827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1485045850276947, + "step": 3180, + "valid_targets_mean": 3459.0, + "valid_targets_min": 1769 + }, + { + "epoch": 1.6684127815610266, + "grad_norm": 0.5233998690171558, + "learning_rate": 3.7716772116302585e-05, + "loss": 0.2844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13598954677581787, + "step": 3185, + "valid_targets_mean": 4186.1, + "valid_targets_min": 2097 + }, + { + "epoch": 1.6710319539025669, + "grad_norm": 0.616188943535829, + "learning_rate": 3.770463594767859e-05, + "loss": 0.293, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15629859268665314, + "step": 3190, + "valid_targets_mean": 3304.0, + "valid_targets_min": 1066 + }, + { + "epoch": 1.6736511262441067, + "grad_norm": 0.5318378638319025, + "learning_rate": 3.769246957367525e-05, + "loss": 0.2821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13819821178913116, + "step": 3195, + "valid_targets_mean": 4457.9, + "valid_targets_min": 2635 + }, + { + "epoch": 1.676270298585647, + "grad_norm": 0.5410474043453303, + "learning_rate": 3.768027301504928e-05, + "loss": 0.2996, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1514570415019989, + "step": 3200, + "valid_targets_mean": 4296.2, + "valid_targets_min": 981 + }, + { + "epoch": 1.6788894709271869, + "grad_norm": 0.551074735027614, + "learning_rate": 3.7668046292608885e-05, + "loss": 0.2801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13885951042175293, + "step": 3205, + "valid_targets_mean": 4020.1, + "valid_targets_min": 2131 + }, + { + "epoch": 1.6815086432687272, + "grad_norm": 0.6126016088397215, + "learning_rate": 3.765578942721372e-05, + "loss": 0.2913, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14835964143276215, + "step": 3210, + "valid_targets_mean": 3541.0, + "valid_targets_min": 1353 + }, + { + "epoch": 1.684127815610267, + "grad_norm": 0.5674725021470198, + "learning_rate": 3.764350243977488e-05, + "loss": 0.2891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12189812958240509, + "step": 3215, + "valid_targets_mean": 3362.0, + "valid_targets_min": 1509 + }, + { + "epoch": 1.6867469879518073, + "grad_norm": 1.3220120776860627, + "learning_rate": 3.763118535125486e-05, + "loss": 0.2782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13082823157310486, + "step": 3220, + "valid_targets_mean": 3918.9, + "valid_targets_min": 2128 + }, + { + "epoch": 1.6893661602933472, + "grad_norm": 0.530743665739903, + "learning_rate": 3.761883818266748e-05, + "loss": 0.2721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10776387155056, + "step": 3225, + "valid_targets_mean": 2919.0, + "valid_targets_min": 1687 + }, + { + "epoch": 1.6919853326348875, + "grad_norm": 0.5760318207108317, + "learning_rate": 3.76064609550779e-05, + "loss": 0.2779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15961535274982452, + "step": 3230, + "valid_targets_mean": 3696.9, + "valid_targets_min": 1803 + }, + { + "epoch": 1.6946045049764273, + "grad_norm": 0.5319780268531669, + "learning_rate": 3.759405368960257e-05, + "loss": 0.2798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13486801087856293, + "step": 3235, + "valid_targets_mean": 4102.0, + "valid_targets_min": 2263 + }, + { + "epoch": 1.6972236773179676, + "grad_norm": 0.5433311734238312, + "learning_rate": 3.7581616407409156e-05, + "loss": 0.2795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17059211432933807, + "step": 3240, + "valid_targets_mean": 4621.4, + "valid_targets_min": 2232 + }, + { + "epoch": 1.6998428496595075, + "grad_norm": 0.569450624970359, + "learning_rate": 3.756914912971657e-05, + "loss": 0.2721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16468310356140137, + "step": 3245, + "valid_targets_mean": 4051.6, + "valid_targets_min": 873 + }, + { + "epoch": 1.7024620220010478, + "grad_norm": 0.545893326548132, + "learning_rate": 3.755665187779487e-05, + "loss": 0.2782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12401072680950165, + "step": 3250, + "valid_targets_mean": 3340.1, + "valid_targets_min": 1942 + }, + { + "epoch": 1.7050811943425876, + "grad_norm": 0.6187773613425988, + "learning_rate": 3.754412467296528e-05, + "loss": 0.2795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1403244137763977, + "step": 3255, + "valid_targets_mean": 3179.4, + "valid_targets_min": 897 + }, + { + "epoch": 1.707700366684128, + "grad_norm": 0.5786590117920204, + "learning_rate": 3.7531567536600096e-05, + "loss": 0.2989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14639341831207275, + "step": 3260, + "valid_targets_mean": 3492.2, + "valid_targets_min": 1695 + }, + { + "epoch": 1.7103195390256678, + "grad_norm": 0.5724373880151019, + "learning_rate": 3.75189804901227e-05, + "loss": 0.283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13018471002578735, + "step": 3265, + "valid_targets_mean": 4043.9, + "valid_targets_min": 1712 + }, + { + "epoch": 1.712938711367208, + "grad_norm": 0.5879671818155303, + "learning_rate": 3.75063635550075e-05, + "loss": 0.282, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1221366822719574, + "step": 3270, + "valid_targets_mean": 2987.9, + "valid_targets_min": 993 + }, + { + "epoch": 1.715557883708748, + "grad_norm": 0.536463698971678, + "learning_rate": 3.7493716752779885e-05, + "loss": 0.2743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14443989098072052, + "step": 3275, + "valid_targets_mean": 3854.8, + "valid_targets_min": 1679 + }, + { + "epoch": 1.7181770560502883, + "grad_norm": 0.5502694031466201, + "learning_rate": 3.748104010501621e-05, + "loss": 0.2843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1322992444038391, + "step": 3280, + "valid_targets_mean": 3938.8, + "valid_targets_min": 2345 + }, + { + "epoch": 1.720796228391828, + "grad_norm": 0.5777659554255037, + "learning_rate": 3.746833363334374e-05, + "loss": 0.284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14358791708946228, + "step": 3285, + "valid_targets_mean": 4507.9, + "valid_targets_min": 3068 + }, + { + "epoch": 1.7234154007333684, + "grad_norm": 0.5081594817829422, + "learning_rate": 3.745559735944064e-05, + "loss": 0.2901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13418079912662506, + "step": 3290, + "valid_targets_mean": 4080.9, + "valid_targets_min": 2965 + }, + { + "epoch": 1.7260345730749083, + "grad_norm": 0.5359743198856676, + "learning_rate": 3.74428313050359e-05, + "loss": 0.2831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11728323251008987, + "step": 3295, + "valid_targets_mean": 2993.5, + "valid_targets_min": 1817 + }, + { + "epoch": 1.7286537454164486, + "grad_norm": 0.5718175688484074, + "learning_rate": 3.743003549190933e-05, + "loss": 0.2824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13692590594291687, + "step": 3300, + "valid_targets_mean": 3795.2, + "valid_targets_min": 1727 + }, + { + "epoch": 1.7312729177579884, + "grad_norm": 0.5052357839163294, + "learning_rate": 3.741720994189149e-05, + "loss": 0.2813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14718163013458252, + "step": 3305, + "valid_targets_mean": 4095.9, + "valid_targets_min": 2491 + }, + { + "epoch": 1.7338920900995285, + "grad_norm": 0.5431824715638952, + "learning_rate": 3.740435467686371e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.129964679479599, + "step": 3310, + "valid_targets_mean": 2046.1, + "valid_targets_min": 333 + }, + { + "epoch": 1.7365112624410686, + "grad_norm": 0.38886822924642506, + "learning_rate": 3.739146971875798e-05, + "loss": 0.1266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06550989300012589, + "step": 3315, + "valid_targets_mean": 3455.9, + "valid_targets_min": 712 + }, + { + "epoch": 1.7391304347826086, + "grad_norm": 0.4613746369250542, + "learning_rate": 3.737855508955696e-05, + "loss": 0.1149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05240120738744736, + "step": 3320, + "valid_targets_mean": 2452.6, + "valid_targets_min": 732 + }, + { + "epoch": 1.7417496071241487, + "grad_norm": 0.46980718129166754, + "learning_rate": 3.736561081129395e-05, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05329611524939537, + "step": 3325, + "valid_targets_mean": 2228.9, + "valid_targets_min": 803 + }, + { + "epoch": 1.7443687794656888, + "grad_norm": 0.38579936721427543, + "learning_rate": 3.7352636906052804e-05, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057813260704278946, + "step": 3330, + "valid_targets_mean": 3038.6, + "valid_targets_min": 818 + }, + { + "epoch": 1.7469879518072289, + "grad_norm": 0.4440048994825392, + "learning_rate": 3.733963339596795e-05, + "loss": 0.1176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061383575201034546, + "step": 3335, + "valid_targets_mean": 3049.4, + "valid_targets_min": 1524 + }, + { + "epoch": 1.749607124148769, + "grad_norm": 0.7821465168985577, + "learning_rate": 3.7326600303224286e-05, + "loss": 0.1689, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07493439316749573, + "step": 3340, + "valid_targets_mean": 1256.1, + "valid_targets_min": 539 + }, + { + "epoch": 1.752226296490309, + "grad_norm": 0.38372791747215335, + "learning_rate": 3.731353765005723e-05, + "loss": 0.1288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05535704642534256, + "step": 3345, + "valid_targets_mean": 3597.0, + "valid_targets_min": 2920 + }, + { + "epoch": 1.754845468831849, + "grad_norm": 0.4258897015104855, + "learning_rate": 3.730044545875258e-05, + "loss": 0.1266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061235859990119934, + "step": 3350, + "valid_targets_mean": 2737.9, + "valid_targets_min": 1048 + }, + { + "epoch": 1.7574646411733892, + "grad_norm": 0.308156419319461, + "learning_rate": 3.728732375164657e-05, + "loss": 0.1566, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041672930121421814, + "step": 3355, + "valid_targets_mean": 4340.1, + "valid_targets_min": 962 + }, + { + "epoch": 1.7600838135149293, + "grad_norm": 0.3261537227691458, + "learning_rate": 3.727417255112577e-05, + "loss": 0.0984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06446874141693115, + "step": 3360, + "valid_targets_mean": 4283.1, + "valid_targets_min": 2091 + }, + { + "epoch": 1.7627029858564693, + "grad_norm": 0.3443297808058669, + "learning_rate": 3.7260991879627077e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05160898715257645, + "step": 3365, + "valid_targets_mean": 4320.9, + "valid_targets_min": 3349 + }, + { + "epoch": 1.7653221581980094, + "grad_norm": 0.3172835679744185, + "learning_rate": 3.724778175963765e-05, + "loss": 0.1116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058801665902137756, + "step": 3370, + "valid_targets_mean": 3895.9, + "valid_targets_min": 518 + }, + { + "epoch": 1.7679413305395495, + "grad_norm": 0.48470033792505424, + "learning_rate": 3.723454221369491e-05, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06509764492511749, + "step": 3375, + "valid_targets_mean": 2219.6, + "valid_targets_min": 549 + }, + { + "epoch": 1.7705605028810896, + "grad_norm": 0.3512724448729901, + "learning_rate": 3.7221273264386476e-05, + "loss": 0.1018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042328815907239914, + "step": 3380, + "valid_targets_mean": 3700.2, + "valid_targets_min": 2724 + }, + { + "epoch": 1.7731796752226296, + "grad_norm": 0.5048034197543899, + "learning_rate": 3.720797493435012e-05, + "loss": 0.1211, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0569320023059845, + "step": 3385, + "valid_targets_mean": 1672.5, + "valid_targets_min": 775 + }, + { + "epoch": 1.7757988475641697, + "grad_norm": 0.3537581019031032, + "learning_rate": 3.7194647246273745e-05, + "loss": 0.0989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029500192031264305, + "step": 3390, + "valid_targets_mean": 3546.2, + "valid_targets_min": 3183 + }, + { + "epoch": 1.7784180199057098, + "grad_norm": 0.4297425358645938, + "learning_rate": 3.718129022289536e-05, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05278308689594269, + "step": 3395, + "valid_targets_mean": 2473.2, + "valid_targets_min": 577 + }, + { + "epoch": 1.7810371922472499, + "grad_norm": 0.5804454614756202, + "learning_rate": 3.7167903887002984e-05, + "loss": 0.1294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08498586714267731, + "step": 3400, + "valid_targets_mean": 1597.0, + "valid_targets_min": 557 + }, + { + "epoch": 1.78365636458879, + "grad_norm": 0.4056400944051235, + "learning_rate": 3.715448826143468e-05, + "loss": 0.1239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06798717379570007, + "step": 3405, + "valid_targets_mean": 3019.1, + "valid_targets_min": 774 + }, + { + "epoch": 1.78627553693033, + "grad_norm": 0.42579563237315016, + "learning_rate": 3.7141043369078466e-05, + "loss": 0.1265, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05185361206531525, + "step": 3410, + "valid_targets_mean": 1825.8, + "valid_targets_min": 680 + }, + { + "epoch": 1.78889470927187, + "grad_norm": 0.4926192326486754, + "learning_rate": 3.712756923287229e-05, + "loss": 0.172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08931027352809906, + "step": 3415, + "valid_targets_mean": 3686.9, + "valid_targets_min": 2819 + }, + { + "epoch": 1.7915138816134102, + "grad_norm": 0.3298804977093151, + "learning_rate": 3.711406587580399e-05, + "loss": 0.1055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04466564953327179, + "step": 3420, + "valid_targets_mean": 4137.8, + "valid_targets_min": 3927 + }, + { + "epoch": 1.7941330539549503, + "grad_norm": 0.5364313257035771, + "learning_rate": 3.710053332091128e-05, + "loss": 0.1189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07550039887428284, + "step": 3425, + "valid_targets_mean": 1376.1, + "valid_targets_min": 714 + }, + { + "epoch": 1.7967522262964903, + "grad_norm": 0.38939869223910956, + "learning_rate": 3.708697159128165e-05, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05331094563007355, + "step": 3430, + "valid_targets_mean": 3432.0, + "valid_targets_min": 754 + }, + { + "epoch": 1.7993713986380304, + "grad_norm": 0.34492267299130824, + "learning_rate": 3.7073380710052394e-05, + "loss": 0.1331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057714078575372696, + "step": 3435, + "valid_targets_mean": 3855.2, + "valid_targets_min": 3178 + }, + { + "epoch": 1.8019905709795705, + "grad_norm": 0.688314337162085, + "learning_rate": 3.705976070041053e-05, + "loss": 0.1568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14388839900493622, + "step": 3440, + "valid_targets_mean": 1566.6, + "valid_targets_min": 920 + }, + { + "epoch": 1.8046097433211106, + "grad_norm": 0.3999222272300496, + "learning_rate": 3.704611158559277e-05, + "loss": 0.3032, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05768517404794693, + "step": 3445, + "valid_targets_mean": 3478.0, + "valid_targets_min": 2314 + }, + { + "epoch": 1.8072289156626506, + "grad_norm": 0.41655719319383894, + "learning_rate": 3.703243338888548e-05, + "loss": 0.1299, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04701980575919151, + "step": 3450, + "valid_targets_mean": 2561.1, + "valid_targets_min": 833 + }, + { + "epoch": 1.8098480880041907, + "grad_norm": 0.44420679075797576, + "learning_rate": 3.701872613362466e-05, + "loss": 0.1253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08168505877256393, + "step": 3455, + "valid_targets_mean": 3509.2, + "valid_targets_min": 1803 + }, + { + "epoch": 1.8124672603457308, + "grad_norm": 0.38214922483382247, + "learning_rate": 3.7004989843195846e-05, + "loss": 0.2256, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07963943481445312, + "step": 3460, + "valid_targets_mean": 4291.4, + "valid_targets_min": 3089 + }, + { + "epoch": 1.8150864326872709, + "grad_norm": 0.24607854741282936, + "learning_rate": 3.6991224541034155e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04560163989663124, + "step": 3465, + "valid_targets_mean": 4843.2, + "valid_targets_min": 2664 + }, + { + "epoch": 1.8177056050288107, + "grad_norm": 0.324276655854062, + "learning_rate": 3.697743025062418e-05, + "loss": 0.1258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.036305949091911316, + "step": 3470, + "valid_targets_mean": 3042.1, + "valid_targets_min": 560 + }, + { + "epoch": 1.820324777370351, + "grad_norm": 0.40090895577438646, + "learning_rate": 3.696360699549996e-05, + "loss": 0.1245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06240606680512428, + "step": 3475, + "valid_targets_mean": 3863.6, + "valid_targets_min": 1615 + }, + { + "epoch": 1.8229439497118909, + "grad_norm": 0.3488949369663456, + "learning_rate": 3.694975479924496e-05, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038241446018218994, + "step": 3480, + "valid_targets_mean": 3312.5, + "valid_targets_min": 2172 + }, + { + "epoch": 1.8255631220534312, + "grad_norm": 0.3766757998217491, + "learning_rate": 3.693587368549203e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047901373356580734, + "step": 3485, + "valid_targets_mean": 3059.9, + "valid_targets_min": 536 + }, + { + "epoch": 1.828182294394971, + "grad_norm": 0.37147694411809196, + "learning_rate": 3.6921963677923336e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05512933433055878, + "step": 3490, + "valid_targets_mean": 3881.0, + "valid_targets_min": 3295 + }, + { + "epoch": 1.8308014667365113, + "grad_norm": 0.33762380384722646, + "learning_rate": 3.6908024800270364e-05, + "loss": 0.1015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04481356963515282, + "step": 3495, + "valid_targets_mean": 3287.2, + "valid_targets_min": 820 + }, + { + "epoch": 1.8334206390780512, + "grad_norm": 0.5869901247545575, + "learning_rate": 3.689405707631383e-05, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1565512716770172, + "step": 3500, + "valid_targets_mean": 2566.6, + "valid_targets_min": 1046 + }, + { + "epoch": 1.8360398114195915, + "grad_norm": 0.33929803396633845, + "learning_rate": 3.688006052988366e-05, + "loss": 0.1003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05045881122350693, + "step": 3505, + "valid_targets_mean": 3590.1, + "valid_targets_min": 2719 + }, + { + "epoch": 1.8386589837611313, + "grad_norm": 0.6334874474935439, + "learning_rate": 3.6866035184858996e-05, + "loss": 0.1039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0817365050315857, + "step": 3510, + "valid_targets_mean": 1638.1, + "valid_targets_min": 699 + }, + { + "epoch": 1.8412781561026716, + "grad_norm": 0.32348764595186325, + "learning_rate": 3.685198106516806e-05, + "loss": 0.0967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04379964619874954, + "step": 3515, + "valid_targets_mean": 3621.6, + "valid_targets_min": 2353 + }, + { + "epoch": 1.8438973284442115, + "grad_norm": 0.37904997502218807, + "learning_rate": 3.683789819478819e-05, + "loss": 0.0998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050226498395204544, + "step": 3520, + "valid_targets_mean": 2328.1, + "valid_targets_min": 549 + }, + { + "epoch": 1.8465165007857518, + "grad_norm": 0.3630310295655135, + "learning_rate": 3.682378659774579e-05, + "loss": 0.1628, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04445043206214905, + "step": 3525, + "valid_targets_mean": 3330.4, + "valid_targets_min": 908 + }, + { + "epoch": 1.8491356731272917, + "grad_norm": 0.3676589648855832, + "learning_rate": 3.680964629811624e-05, + "loss": 0.131, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08851621299982071, + "step": 3530, + "valid_targets_mean": 3458.4, + "valid_targets_min": 2196 + }, + { + "epoch": 1.851754845468832, + "grad_norm": 0.35139370366929573, + "learning_rate": 3.6795477320023915e-05, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04188999533653259, + "step": 3535, + "valid_targets_mean": 1252.1, + "valid_targets_min": 598 + }, + { + "epoch": 1.8543740178103718, + "grad_norm": 0.8478295274873181, + "learning_rate": 3.6781279687642105e-05, + "loss": 0.1427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12599429488182068, + "step": 3540, + "valid_targets_mean": 1353.1, + "valid_targets_min": 578 + }, + { + "epoch": 1.856993190151912, + "grad_norm": 0.6051185229462231, + "learning_rate": 3.676705342519299e-05, + "loss": 0.1666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09067556262016296, + "step": 3545, + "valid_targets_mean": 1974.2, + "valid_targets_min": 695 + }, + { + "epoch": 1.859612362493452, + "grad_norm": 0.5104635760844917, + "learning_rate": 3.675279855694759e-05, + "loss": 0.1141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04893461614847183, + "step": 3550, + "valid_targets_mean": 1724.5, + "valid_targets_min": 691 + }, + { + "epoch": 1.8622315348349923, + "grad_norm": 0.34490442267552013, + "learning_rate": 3.673851510722574e-05, + "loss": 0.1393, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045739125460386276, + "step": 3555, + "valid_targets_mean": 3842.6, + "valid_targets_min": 2024 + }, + { + "epoch": 1.8648507071765321, + "grad_norm": 0.5984571579132926, + "learning_rate": 3.672420310039603e-05, + "loss": 0.1227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07412593066692352, + "step": 3560, + "valid_targets_mean": 1651.0, + "valid_targets_min": 523 + }, + { + "epoch": 1.8674698795180724, + "grad_norm": 0.2994686102233471, + "learning_rate": 3.670986256087577e-05, + "loss": 0.1081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04276078939437866, + "step": 3565, + "valid_targets_mean": 3532.5, + "valid_targets_min": 468 + }, + { + "epoch": 1.8700890518596123, + "grad_norm": 0.2552751521928846, + "learning_rate": 3.669549351313094e-05, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028061550110578537, + "step": 3570, + "valid_targets_mean": 5298.0, + "valid_targets_min": 1195 + }, + { + "epoch": 1.8727082242011526, + "grad_norm": 0.31663609850096974, + "learning_rate": 3.668109598167618e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03680366650223732, + "step": 3575, + "valid_targets_mean": 3562.6, + "valid_targets_min": 459 + }, + { + "epoch": 1.8753273965426924, + "grad_norm": 0.30586408337736504, + "learning_rate": 3.666666999107471e-05, + "loss": 0.1011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04620771110057831, + "step": 3580, + "valid_targets_mean": 3548.0, + "valid_targets_min": 981 + }, + { + "epoch": 1.8779465688842327, + "grad_norm": 0.4010603474435427, + "learning_rate": 3.6652215565938296e-05, + "loss": 0.1135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06624223291873932, + "step": 3585, + "valid_targets_mean": 2130.1, + "valid_targets_min": 697 + }, + { + "epoch": 1.8805657412257726, + "grad_norm": 0.3958079218923452, + "learning_rate": 3.663773273092723e-05, + "loss": 0.1239, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05781649425625801, + "step": 3590, + "valid_targets_mean": 3661.6, + "valid_targets_min": 2759 + }, + { + "epoch": 1.8831849135673129, + "grad_norm": 0.3832291655416344, + "learning_rate": 3.662322151075027e-05, + "loss": 0.1198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05941053852438927, + "step": 3595, + "valid_targets_mean": 3909.5, + "valid_targets_min": 3391 + }, + { + "epoch": 1.8858040859088527, + "grad_norm": 0.30687924417394397, + "learning_rate": 3.660868193016459e-05, + "loss": 0.1037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04502261057496071, + "step": 3600, + "valid_targets_mean": 3319.9, + "valid_targets_min": 1099 + }, + { + "epoch": 1.888423258250393, + "grad_norm": 0.37220060003799615, + "learning_rate": 3.6594114013975773e-05, + "loss": 0.107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05653055012226105, + "step": 3605, + "valid_targets_mean": 3717.0, + "valid_targets_min": 1874 + }, + { + "epoch": 1.8910424305919329, + "grad_norm": 0.3348285137824071, + "learning_rate": 3.6579517787037716e-05, + "loss": 0.1243, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051292359828948975, + "step": 3610, + "valid_targets_mean": 3263.5, + "valid_targets_min": 724 + }, + { + "epoch": 1.8936616029334732, + "grad_norm": 0.3160799556583055, + "learning_rate": 3.656489327425264e-05, + "loss": 0.092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041962191462516785, + "step": 3615, + "valid_targets_mean": 3707.5, + "valid_targets_min": 2769 + }, + { + "epoch": 1.896280775275013, + "grad_norm": 0.4033536459967635, + "learning_rate": 3.6550240500571e-05, + "loss": 0.1206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050984740257263184, + "step": 3620, + "valid_targets_mean": 2815.8, + "valid_targets_min": 645 + }, + { + "epoch": 1.8988999476165531, + "grad_norm": 0.34253687803544475, + "learning_rate": 3.653555949099149e-05, + "loss": 0.1132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04667484015226364, + "step": 3625, + "valid_targets_mean": 3488.8, + "valid_targets_min": 600 + }, + { + "epoch": 1.9015191199580932, + "grad_norm": 0.36815609359149515, + "learning_rate": 3.652085027056095e-05, + "loss": 0.1329, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05535770207643509, + "step": 3630, + "valid_targets_mean": 3373.4, + "valid_targets_min": 1648 + }, + { + "epoch": 1.9041382922996333, + "grad_norm": 0.38770252593075355, + "learning_rate": 3.6506112864374394e-05, + "loss": 0.1169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07005764544010162, + "step": 3635, + "valid_targets_mean": 2555.0, + "valid_targets_min": 821 + }, + { + "epoch": 1.9067574646411733, + "grad_norm": 0.8249178240940236, + "learning_rate": 3.6491347297574874e-05, + "loss": 0.1326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07876034826040268, + "step": 3640, + "valid_targets_mean": 867.1, + "valid_targets_min": 612 + }, + { + "epoch": 1.9093766369827134, + "grad_norm": 0.5000614216693436, + "learning_rate": 3.647655359535351e-05, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08283238857984543, + "step": 3645, + "valid_targets_mean": 2202.1, + "valid_targets_min": 830 + }, + { + "epoch": 1.9119958093242535, + "grad_norm": 0.4525706922927935, + "learning_rate": 3.646173178294942e-05, + "loss": 0.1278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05131802707910538, + "step": 3650, + "valid_targets_mean": 3027.0, + "valid_targets_min": 1146 + }, + { + "epoch": 1.9146149816657936, + "grad_norm": 0.3907620341886926, + "learning_rate": 3.644688188564968e-05, + "loss": 0.1132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06965774297714233, + "step": 3655, + "valid_targets_mean": 3296.6, + "valid_targets_min": 2760 + }, + { + "epoch": 1.9172341540073337, + "grad_norm": 0.5826796249967633, + "learning_rate": 3.643200392878928e-05, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07157107442617416, + "step": 3660, + "valid_targets_mean": 1469.5, + "valid_targets_min": 628 + }, + { + "epoch": 1.9198533263488737, + "grad_norm": 0.5642581338660182, + "learning_rate": 3.641709793775108e-05, + "loss": 0.1315, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1086469516158104, + "step": 3665, + "valid_targets_mean": 1752.1, + "valid_targets_min": 691 + }, + { + "epoch": 1.9224724986904138, + "grad_norm": 0.28170895368425586, + "learning_rate": 3.640216393796578e-05, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03042084351181984, + "step": 3670, + "valid_targets_mean": 3863.9, + "valid_targets_min": 3613 + }, + { + "epoch": 1.9250916710319539, + "grad_norm": 0.3248620416461367, + "learning_rate": 3.6387201954911835e-05, + "loss": 0.1023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05128296837210655, + "step": 3675, + "valid_targets_mean": 4304.2, + "valid_targets_min": 1773 + }, + { + "epoch": 1.927710843373494, + "grad_norm": 0.3721422506091931, + "learning_rate": 3.637221201411549e-05, + "loss": 0.1101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05849291384220123, + "step": 3680, + "valid_targets_mean": 1971.9, + "valid_targets_min": 762 + }, + { + "epoch": 1.930330015715034, + "grad_norm": 0.3909415371860603, + "learning_rate": 3.6357194141150636e-05, + "loss": 0.0991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07564368844032288, + "step": 3685, + "valid_targets_mean": 4554.1, + "valid_targets_min": 946 + }, + { + "epoch": 1.9329491880565741, + "grad_norm": 0.35501488521999336, + "learning_rate": 3.634214836163885e-05, + "loss": 0.0924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06378553062677383, + "step": 3690, + "valid_targets_mean": 5015.5, + "valid_targets_min": 4104 + }, + { + "epoch": 1.9355683603981142, + "grad_norm": 0.36430960317131245, + "learning_rate": 3.632707470124931e-05, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052411675453186035, + "step": 3695, + "valid_targets_mean": 3064.4, + "valid_targets_min": 739 + }, + { + "epoch": 1.9381875327396543, + "grad_norm": 0.43852732444058823, + "learning_rate": 3.631197318569877e-05, + "loss": 0.1294, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05728181079030037, + "step": 3700, + "valid_targets_mean": 3863.6, + "valid_targets_min": 3550 + }, + { + "epoch": 1.9408067050811943, + "grad_norm": 0.34574181915012264, + "learning_rate": 3.629684384075148e-05, + "loss": 0.1088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06230345368385315, + "step": 3705, + "valid_targets_mean": 3474.9, + "valid_targets_min": 704 + }, + { + "epoch": 1.9434258774227344, + "grad_norm": 0.323229309378288, + "learning_rate": 3.6281686692219215e-05, + "loss": 0.1099, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047364018857479095, + "step": 3710, + "valid_targets_mean": 2568.8, + "valid_targets_min": 774 + }, + { + "epoch": 1.9460450497642745, + "grad_norm": 0.2865805180471261, + "learning_rate": 3.626650176596114e-05, + "loss": 0.1351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045200690627098083, + "step": 3715, + "valid_targets_mean": 3486.2, + "valid_targets_min": 888 + }, + { + "epoch": 1.9486642221058146, + "grad_norm": 0.3645101643349168, + "learning_rate": 3.625128908788383e-05, + "loss": 0.1058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05560200661420822, + "step": 3720, + "valid_targets_mean": 3371.6, + "valid_targets_min": 1048 + }, + { + "epoch": 1.9512833944473547, + "grad_norm": 0.35727012775933253, + "learning_rate": 3.6236048683941207e-05, + "loss": 0.1192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05395163595676422, + "step": 3725, + "valid_targets_mean": 2941.5, + "valid_targets_min": 722 + }, + { + "epoch": 1.9539025667888947, + "grad_norm": 0.398407406039969, + "learning_rate": 3.62207805801345e-05, + "loss": 0.1142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06683570146560669, + "step": 3730, + "valid_targets_mean": 2359.2, + "valid_targets_min": 865 + }, + { + "epoch": 1.9565217391304348, + "grad_norm": 0.4526976053702537, + "learning_rate": 3.620548480251218e-05, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04311978816986084, + "step": 3735, + "valid_targets_mean": 1747.0, + "valid_targets_min": 570 + }, + { + "epoch": 1.9591409114719749, + "grad_norm": 0.39229483287627864, + "learning_rate": 3.6190161377169946e-05, + "loss": 0.1332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06888778507709503, + "step": 3740, + "valid_targets_mean": 3704.2, + "valid_targets_min": 2811 + }, + { + "epoch": 1.961760083813515, + "grad_norm": 0.4147123294533103, + "learning_rate": 3.6174810330250664e-05, + "loss": 0.1095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04733257740736008, + "step": 3745, + "valid_targets_mean": 1749.1, + "valid_targets_min": 635 + }, + { + "epoch": 1.964379256155055, + "grad_norm": 0.36115289080590474, + "learning_rate": 3.615943168794432e-05, + "loss": 0.1078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054300181567668915, + "step": 3750, + "valid_targets_mean": 2985.1, + "valid_targets_min": 1591 + }, + { + "epoch": 1.9669984284965951, + "grad_norm": 0.3483828778339568, + "learning_rate": 3.6144025476487976e-05, + "loss": 0.107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04829784855246544, + "step": 3755, + "valid_targets_mean": 2912.6, + "valid_targets_min": 800 + }, + { + "epoch": 1.9696176008381352, + "grad_norm": 0.324736990433432, + "learning_rate": 3.612859172216574e-05, + "loss": 0.0993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055288832634687424, + "step": 3760, + "valid_targets_mean": 3084.9, + "valid_targets_min": 860 + }, + { + "epoch": 1.9722367731796753, + "grad_norm": 0.45400580407489743, + "learning_rate": 3.61131304513087e-05, + "loss": 0.1373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06422509253025055, + "step": 3765, + "valid_targets_mean": 2482.6, + "valid_targets_min": 878 + }, + { + "epoch": 1.9748559455212154, + "grad_norm": 0.364513666423315, + "learning_rate": 3.60976416902949e-05, + "loss": 0.0945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048155076801776886, + "step": 3770, + "valid_targets_mean": 3403.2, + "valid_targets_min": 1024 + }, + { + "epoch": 1.9774751178627552, + "grad_norm": 0.4202418643860163, + "learning_rate": 3.608212546554927e-05, + "loss": 0.107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05272798240184784, + "step": 3775, + "valid_targets_mean": 2173.1, + "valid_targets_min": 786 + }, + { + "epoch": 1.9800942902042955, + "grad_norm": 0.4253734066193847, + "learning_rate": 3.6066581803543605e-05, + "loss": 0.11, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.066901296377182, + "step": 3780, + "valid_targets_mean": 3233.9, + "valid_targets_min": 2573 + }, + { + "epoch": 1.9827134625458354, + "grad_norm": 0.29672768926812404, + "learning_rate": 3.605101073079651e-05, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05633041262626648, + "step": 3785, + "valid_targets_mean": 3821.9, + "valid_targets_min": 2274 + }, + { + "epoch": 1.9853326348873757, + "grad_norm": 0.3847755589599964, + "learning_rate": 3.6035412273873355e-05, + "loss": 0.0951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06555216759443283, + "step": 3790, + "valid_targets_mean": 3791.2, + "valid_targets_min": 892 + }, + { + "epoch": 1.9879518072289155, + "grad_norm": 0.3035623229498264, + "learning_rate": 3.601978645938623e-05, + "loss": 0.1055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04228395223617554, + "step": 3795, + "valid_targets_mean": 4820.4, + "valid_targets_min": 4491 + }, + { + "epoch": 1.9905709795704558, + "grad_norm": 0.5257397519577061, + "learning_rate": 3.6004133313993874e-05, + "loss": 0.1488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11688600480556488, + "step": 3800, + "valid_targets_mean": 2159.1, + "valid_targets_min": 709 + }, + { + "epoch": 1.9931901519119957, + "grad_norm": 0.3019179724403995, + "learning_rate": 3.5988452864401695e-05, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03450121358036995, + "step": 3805, + "valid_targets_mean": 3430.8, + "valid_targets_min": 1097 + }, + { + "epoch": 1.995809324253536, + "grad_norm": 0.3278317606992317, + "learning_rate": 3.597274513736167e-05, + "loss": 0.1092, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04503103345632553, + "step": 3810, + "valid_targets_mean": 3032.0, + "valid_targets_min": 840 + }, + { + "epoch": 1.9984284965950758, + "grad_norm": 0.3813647615593932, + "learning_rate": 3.595701015967229e-05, + "loss": 0.1022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0429418683052063, + "step": 3815, + "valid_targets_mean": 3812.2, + "valid_targets_min": 3106 + }, + { + "epoch": 2.001047668936616, + "grad_norm": 0.4284499140312006, + "learning_rate": 3.5941247958178574e-05, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10122483223676682, + "step": 3820, + "valid_targets_mean": 5904.0, + "valid_targets_min": 4967 + }, + { + "epoch": 2.003666841278156, + "grad_norm": 0.40019471016790836, + "learning_rate": 3.5925458559771954e-05, + "loss": 0.2, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11647690087556839, + "step": 3825, + "valid_targets_mean": 8641.4, + "valid_targets_min": 5164 + }, + { + "epoch": 2.0062860136196963, + "grad_norm": 0.34016108535872486, + "learning_rate": 3.590964199139029e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09456409513950348, + "step": 3830, + "valid_targets_mean": 6838.2, + "valid_targets_min": 5200 + }, + { + "epoch": 2.008905185961236, + "grad_norm": 0.3903785178392229, + "learning_rate": 3.589379828001777e-05, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0786237046122551, + "step": 3835, + "valid_targets_mean": 5411.6, + "valid_targets_min": 3853 + }, + { + "epoch": 2.0115243583027764, + "grad_norm": 0.3915887485394028, + "learning_rate": 3.587792745268492e-05, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10382388532161713, + "step": 3840, + "valid_targets_mean": 5598.0, + "valid_targets_min": 4669 + }, + { + "epoch": 2.0141435306443163, + "grad_norm": 0.4104848371003733, + "learning_rate": 3.586202953646848e-05, + "loss": 0.2059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10249824821949005, + "step": 3845, + "valid_targets_mean": 5953.0, + "valid_targets_min": 4216 + }, + { + "epoch": 2.0167627029858566, + "grad_norm": 0.39021228681004444, + "learning_rate": 3.5846104558491465e-05, + "loss": 0.1989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10285435616970062, + "step": 3850, + "valid_targets_mean": 5914.0, + "valid_targets_min": 4589 + }, + { + "epoch": 2.0193818753273964, + "grad_norm": 0.365326511528838, + "learning_rate": 3.5830152545923024e-05, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09683631360530853, + "step": 3855, + "valid_targets_mean": 6515.4, + "valid_targets_min": 4588 + }, + { + "epoch": 2.0220010476689367, + "grad_norm": 1.4095097808980237, + "learning_rate": 3.581417352597843e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09095996618270874, + "step": 3860, + "valid_targets_mean": 7015.2, + "valid_targets_min": 4530 + }, + { + "epoch": 2.0246202200104766, + "grad_norm": 0.36169510092850965, + "learning_rate": 3.579816752591903e-05, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1063755601644516, + "step": 3865, + "valid_targets_mean": 6202.6, + "valid_targets_min": 4837 + }, + { + "epoch": 2.027239392352017, + "grad_norm": 0.4849200189112995, + "learning_rate": 3.578213457305222e-05, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06249552220106125, + "step": 3870, + "valid_targets_mean": 1720.9, + "valid_targets_min": 229 + }, + { + "epoch": 2.0298585646935567, + "grad_norm": 0.3707007660827228, + "learning_rate": 3.5766074694731374e-05, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09450245648622513, + "step": 3875, + "valid_targets_mean": 5963.8, + "valid_targets_min": 4216 + }, + { + "epoch": 2.032477737035097, + "grad_norm": 0.37984112059796205, + "learning_rate": 3.574998791835578e-05, + "loss": 0.1977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0914495438337326, + "step": 3880, + "valid_targets_mean": 5973.9, + "valid_targets_min": 4835 + }, + { + "epoch": 2.035096909376637, + "grad_norm": 0.3451717856261538, + "learning_rate": 3.573387427137064e-05, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08319023251533508, + "step": 3885, + "valid_targets_mean": 6669.8, + "valid_targets_min": 4995 + }, + { + "epoch": 2.037716081718177, + "grad_norm": 0.4409189921348946, + "learning_rate": 3.5717733781266994e-05, + "loss": 0.1771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12701013684272766, + "step": 3890, + "valid_targets_mean": 5705.5, + "valid_targets_min": 4108 + }, + { + "epoch": 2.040335254059717, + "grad_norm": 0.38595345220799343, + "learning_rate": 3.5701566475581675e-05, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11929184943437576, + "step": 3895, + "valid_targets_mean": 6594.0, + "valid_targets_min": 4527 + }, + { + "epoch": 2.0429544264012574, + "grad_norm": 0.411996005083078, + "learning_rate": 3.568537238189727e-05, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1032436341047287, + "step": 3900, + "valid_targets_mean": 3100.9, + "valid_targets_min": 2104 + }, + { + "epoch": 2.045573598742797, + "grad_norm": 0.3603437289904686, + "learning_rate": 3.566915152784206e-05, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08257415890693665, + "step": 3905, + "valid_targets_mean": 5817.8, + "valid_targets_min": 4757 + }, + { + "epoch": 2.0481927710843375, + "grad_norm": 0.4009799461293387, + "learning_rate": 3.565290394108998e-05, + "loss": 0.1749, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11234372854232788, + "step": 3910, + "valid_targets_mean": 6333.0, + "valid_targets_min": 5056 + }, + { + "epoch": 2.0508119434258774, + "grad_norm": 0.3420090992748751, + "learning_rate": 3.56366296493606e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0842496007680893, + "step": 3915, + "valid_targets_mean": 5956.0, + "valid_targets_min": 2834 + }, + { + "epoch": 2.0534311157674177, + "grad_norm": 0.3779894625008644, + "learning_rate": 3.562032868041901e-05, + "loss": 0.2073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12641143798828125, + "step": 3920, + "valid_targets_mean": 6984.6, + "valid_targets_min": 5744 + }, + { + "epoch": 2.0560502881089575, + "grad_norm": 0.3418320657225218, + "learning_rate": 3.560400106207584e-05, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08700765669345856, + "step": 3925, + "valid_targets_mean": 7290.2, + "valid_targets_min": 5640 + }, + { + "epoch": 2.058669460450498, + "grad_norm": 0.38835626160752634, + "learning_rate": 3.558764682218717e-05, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08968793600797653, + "step": 3930, + "valid_targets_mean": 7036.0, + "valid_targets_min": 5210 + }, + { + "epoch": 2.0612886327920377, + "grad_norm": 0.389476080087014, + "learning_rate": 3.557126598865452e-05, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12502798438072205, + "step": 3935, + "valid_targets_mean": 7138.5, + "valid_targets_min": 5292 + }, + { + "epoch": 2.063907805133578, + "grad_norm": 0.39016530275430894, + "learning_rate": 3.5554858589424755e-05, + "loss": 0.2066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10163958370685577, + "step": 3940, + "valid_targets_mean": 6134.2, + "valid_targets_min": 3531 + }, + { + "epoch": 2.066526977475118, + "grad_norm": 0.39984838191690064, + "learning_rate": 3.553842465249009e-05, + "loss": 0.2232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10293346643447876, + "step": 3945, + "valid_targets_mean": 5968.2, + "valid_targets_min": 4640 + }, + { + "epoch": 2.069146149816658, + "grad_norm": 0.38704520187509334, + "learning_rate": 3.552196420588799e-05, + "loss": 0.2064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09838110208511353, + "step": 3950, + "valid_targets_mean": 6249.6, + "valid_targets_min": 4328 + }, + { + "epoch": 2.071765322158198, + "grad_norm": 0.4008425419448513, + "learning_rate": 3.550547727770115e-05, + "loss": 0.1955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10637270659208298, + "step": 3955, + "valid_targets_mean": 6002.4, + "valid_targets_min": 3123 + }, + { + "epoch": 2.0743844944997383, + "grad_norm": 0.3911921061067867, + "learning_rate": 3.548896389605748e-05, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09466516971588135, + "step": 3960, + "valid_targets_mean": 6372.2, + "valid_targets_min": 4256 + }, + { + "epoch": 2.077003666841278, + "grad_norm": 0.4247336082006023, + "learning_rate": 3.547242408912997e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10660803318023682, + "step": 3965, + "valid_targets_mean": 5927.8, + "valid_targets_min": 4753 + }, + { + "epoch": 2.0796228391828184, + "grad_norm": 0.4161444115232309, + "learning_rate": 3.545585788513672e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09794335067272186, + "step": 3970, + "valid_targets_mean": 6106.1, + "valid_targets_min": 5370 + }, + { + "epoch": 2.0822420115243583, + "grad_norm": 0.417431112162858, + "learning_rate": 3.5439265312340874e-05, + "loss": 0.1975, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10755012929439545, + "step": 3975, + "valid_targets_mean": 6240.6, + "valid_targets_min": 4608 + }, + { + "epoch": 2.0848611838658986, + "grad_norm": 0.42915133121570626, + "learning_rate": 3.542264639905054e-05, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09163561463356018, + "step": 3980, + "valid_targets_mean": 5733.2, + "valid_targets_min": 3801 + }, + { + "epoch": 2.0874803562074384, + "grad_norm": 0.39365966964337057, + "learning_rate": 3.5406001173618785e-05, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10349412262439728, + "step": 3985, + "valid_targets_mean": 6430.1, + "valid_targets_min": 4754 + }, + { + "epoch": 2.0900995285489787, + "grad_norm": 0.3678649229546877, + "learning_rate": 3.5389329664443546e-05, + "loss": 0.1691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07494372129440308, + "step": 3990, + "valid_targets_mean": 7375.2, + "valid_targets_min": 6417 + }, + { + "epoch": 2.0927187008905186, + "grad_norm": 0.4707631783030055, + "learning_rate": 3.5372631899967626e-05, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10619588196277618, + "step": 3995, + "valid_targets_mean": 6281.0, + "valid_targets_min": 3468 + }, + { + "epoch": 2.095337873232059, + "grad_norm": 0.37484837164589674, + "learning_rate": 3.535590790867861e-05, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09556078165769577, + "step": 4000, + "valid_targets_mean": 6532.4, + "valid_targets_min": 5221 + }, + { + "epoch": 2.0979570455735987, + "grad_norm": 0.39600830799691644, + "learning_rate": 3.5339157719108825e-05, + "loss": 0.1992, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08089323341846466, + "step": 4005, + "valid_targets_mean": 5552.6, + "valid_targets_min": 4438 + }, + { + "epoch": 2.1005762179151386, + "grad_norm": 0.4075541642332971, + "learning_rate": 3.532238135983529e-05, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10310782492160797, + "step": 4010, + "valid_targets_mean": 5878.1, + "valid_targets_min": 3915 + }, + { + "epoch": 2.103195390256679, + "grad_norm": 0.3685244263079544, + "learning_rate": 3.530557885947968e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07567736506462097, + "step": 4015, + "valid_targets_mean": 7082.6, + "valid_targets_min": 4135 + }, + { + "epoch": 2.1058145625982188, + "grad_norm": 0.4303828445875738, + "learning_rate": 3.528875024670827e-05, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09841020405292511, + "step": 4020, + "valid_targets_mean": 5936.4, + "valid_targets_min": 3947 + }, + { + "epoch": 2.108433734939759, + "grad_norm": 0.41147636679565364, + "learning_rate": 3.5271895550231874e-05, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12070822715759277, + "step": 4025, + "valid_targets_mean": 6836.2, + "valid_targets_min": 4865 + }, + { + "epoch": 2.111052907281299, + "grad_norm": 0.404827778216781, + "learning_rate": 3.525501479880583e-05, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09458012878894806, + "step": 4030, + "valid_targets_mean": 6539.4, + "valid_targets_min": 4643 + }, + { + "epoch": 2.113672079622839, + "grad_norm": 0.4119557200130413, + "learning_rate": 3.523810802122989e-05, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09437651932239532, + "step": 4035, + "valid_targets_mean": 6613.6, + "valid_targets_min": 4959 + }, + { + "epoch": 2.116291251964379, + "grad_norm": 0.3857551741776284, + "learning_rate": 3.522117524634824e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10727666318416595, + "step": 4040, + "valid_targets_mean": 6044.1, + "valid_targets_min": 4652 + }, + { + "epoch": 2.1189104243059194, + "grad_norm": 0.3629802694808262, + "learning_rate": 3.520421650304942e-05, + "loss": 0.1698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08399035036563873, + "step": 4045, + "valid_targets_mean": 5893.9, + "valid_targets_min": 4938 + }, + { + "epoch": 2.121529596647459, + "grad_norm": 0.3562351553986034, + "learning_rate": 3.518723182026625e-05, + "loss": 0.1643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07706134766340256, + "step": 4050, + "valid_targets_mean": 5849.6, + "valid_targets_min": 4546 + }, + { + "epoch": 2.1241487689889995, + "grad_norm": 0.36545122903316773, + "learning_rate": 3.517022122697581e-05, + "loss": 0.1692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08789099752902985, + "step": 4055, + "valid_targets_mean": 5751.5, + "valid_targets_min": 3200 + }, + { + "epoch": 2.1267679413305394, + "grad_norm": 0.38134729107418, + "learning_rate": 3.5153184752199407e-05, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09545209258794785, + "step": 4060, + "valid_targets_mean": 6149.6, + "valid_targets_min": 5010 + }, + { + "epoch": 2.1293871136720797, + "grad_norm": 0.3796478109667821, + "learning_rate": 3.513612242500248e-05, + "loss": 0.1946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10478127747774124, + "step": 4065, + "valid_targets_mean": 6551.2, + "valid_targets_min": 4904 + }, + { + "epoch": 2.1320062860136195, + "grad_norm": 0.43154816567740006, + "learning_rate": 3.511903427449459e-05, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1344555914402008, + "step": 4070, + "valid_targets_mean": 5728.9, + "valid_targets_min": 4812 + }, + { + "epoch": 2.13462545835516, + "grad_norm": 0.40743025524321763, + "learning_rate": 3.510192032982935e-05, + "loss": 0.1951, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0798969641327858, + "step": 4075, + "valid_targets_mean": 5047.9, + "valid_targets_min": 3629 + }, + { + "epoch": 2.1372446306966997, + "grad_norm": 0.3967243019722531, + "learning_rate": 3.508478062020437e-05, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09613733738660812, + "step": 4080, + "valid_targets_mean": 6260.0, + "valid_targets_min": 4324 + }, + { + "epoch": 2.13986380303824, + "grad_norm": 0.360089690411804, + "learning_rate": 3.506761517486123e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08278235793113708, + "step": 4085, + "valid_targets_mean": 7853.1, + "valid_targets_min": 4920 + }, + { + "epoch": 2.14248297537978, + "grad_norm": 0.40877177483736243, + "learning_rate": 3.505042402308541e-05, + "loss": 0.199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09078847616910934, + "step": 4090, + "valid_targets_mean": 6231.0, + "valid_targets_min": 4882 + }, + { + "epoch": 2.14510214772132, + "grad_norm": 0.3643484064304972, + "learning_rate": 3.5033207194206245e-05, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09040717035531998, + "step": 4095, + "valid_targets_mean": 7472.9, + "valid_targets_min": 3943 + }, + { + "epoch": 2.14772132006286, + "grad_norm": 0.38805898228600455, + "learning_rate": 3.501596471759689e-05, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09874925017356873, + "step": 4100, + "valid_targets_mean": 6840.2, + "valid_targets_min": 5034 + }, + { + "epoch": 2.1503404924044003, + "grad_norm": 0.40459874559163334, + "learning_rate": 3.499869662267423e-05, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1005554050207138, + "step": 4105, + "valid_targets_mean": 6458.5, + "valid_targets_min": 4461 + }, + { + "epoch": 2.15295966474594, + "grad_norm": 0.4021288703461126, + "learning_rate": 3.498140293889889e-05, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08713238686323166, + "step": 4110, + "valid_targets_mean": 6645.4, + "valid_targets_min": 5188 + }, + { + "epoch": 2.1555788370874804, + "grad_norm": 0.39549728125746436, + "learning_rate": 3.4964083695775136e-05, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10138913244009018, + "step": 4115, + "valid_targets_mean": 6002.9, + "valid_targets_min": 5009 + }, + { + "epoch": 2.1581980094290203, + "grad_norm": 0.3957837737870197, + "learning_rate": 3.4946738922850825e-05, + "loss": 0.2356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10216537117958069, + "step": 4120, + "valid_targets_mean": 6025.9, + "valid_targets_min": 4551 + }, + { + "epoch": 2.1608171817705606, + "grad_norm": 0.38313190921765966, + "learning_rate": 3.49293686497174e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09715119749307632, + "step": 4125, + "valid_targets_mean": 7210.0, + "valid_targets_min": 4492 + }, + { + "epoch": 2.1634363541121004, + "grad_norm": 0.3777165911050129, + "learning_rate": 3.491197290600978e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09169087558984756, + "step": 4130, + "valid_targets_mean": 7051.4, + "valid_targets_min": 4718 + }, + { + "epoch": 2.1660555264536407, + "grad_norm": 0.3796098792830878, + "learning_rate": 3.489455172140637e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10031376779079437, + "step": 4135, + "valid_targets_mean": 6686.0, + "valid_targets_min": 4827 + }, + { + "epoch": 2.1686746987951806, + "grad_norm": 0.385524732732391, + "learning_rate": 3.487710512562894e-05, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07671770453453064, + "step": 4140, + "valid_targets_mean": 5311.9, + "valid_targets_min": 3828 + }, + { + "epoch": 2.171293871136721, + "grad_norm": 0.3750188275016031, + "learning_rate": 3.4859633148442644e-05, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08107757568359375, + "step": 4145, + "valid_targets_mean": 7078.9, + "valid_targets_min": 4594 + }, + { + "epoch": 2.1739130434782608, + "grad_norm": 0.39016940873503453, + "learning_rate": 3.4842135819655926e-05, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07679791748523712, + "step": 4150, + "valid_targets_mean": 6393.8, + "valid_targets_min": 4509 + }, + { + "epoch": 2.176532215819801, + "grad_norm": 0.37788350391475306, + "learning_rate": 3.482461316912049e-05, + "loss": 0.1721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10497917234897614, + "step": 4155, + "valid_targets_mean": 7848.9, + "valid_targets_min": 5649 + }, + { + "epoch": 2.179151388161341, + "grad_norm": 0.36917606059640495, + "learning_rate": 3.480706522673124e-05, + "loss": 0.1734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08077841252088547, + "step": 4160, + "valid_targets_mean": 6623.4, + "valid_targets_min": 4343 + }, + { + "epoch": 2.181770560502881, + "grad_norm": 0.3667327632355009, + "learning_rate": 3.47894920224262e-05, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0841049775481224, + "step": 4165, + "valid_targets_mean": 6255.9, + "valid_targets_min": 4660 + }, + { + "epoch": 2.184389732844421, + "grad_norm": 0.34449765711462293, + "learning_rate": 3.477189358618655e-05, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09171411395072937, + "step": 4170, + "valid_targets_mean": 8243.0, + "valid_targets_min": 6281 + }, + { + "epoch": 2.1870089051859614, + "grad_norm": 0.36014210926871604, + "learning_rate": 3.4754269948036465e-05, + "loss": 0.1732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09017133712768555, + "step": 4175, + "valid_targets_mean": 7083.5, + "valid_targets_min": 4657 + }, + { + "epoch": 2.189628077527501, + "grad_norm": 0.4061604533549802, + "learning_rate": 3.473662113804314e-05, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1070374995470047, + "step": 4180, + "valid_targets_mean": 6399.8, + "valid_targets_min": 5098 + }, + { + "epoch": 2.1922472498690415, + "grad_norm": 0.44422493113454775, + "learning_rate": 3.4718947186316714e-05, + "loss": 0.1902, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1003928929567337, + "step": 4185, + "valid_targets_mean": 5257.4, + "valid_targets_min": 1879 + }, + { + "epoch": 2.1948664222105814, + "grad_norm": 0.4231432868586818, + "learning_rate": 3.4701248123010214e-05, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10550573468208313, + "step": 4190, + "valid_targets_mean": 5260.5, + "valid_targets_min": 3619 + }, + { + "epoch": 2.1974855945521217, + "grad_norm": 0.7042590921457861, + "learning_rate": 3.4683523978319514e-05, + "loss": 0.2157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08072187006473541, + "step": 4195, + "valid_targets_mean": 1370.4, + "valid_targets_min": 715 + }, + { + "epoch": 2.2001047668936615, + "grad_norm": 0.7164844766967332, + "learning_rate": 3.4665774782483275e-05, + "loss": 0.1991, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10653487592935562, + "step": 4200, + "valid_targets_mean": 1688.0, + "valid_targets_min": 1302 + }, + { + "epoch": 2.202723939235202, + "grad_norm": 0.8018418571503082, + "learning_rate": 3.464800056578289e-05, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09308301657438278, + "step": 4205, + "valid_targets_mean": 1349.2, + "valid_targets_min": 714 + }, + { + "epoch": 2.2053431115767417, + "grad_norm": 0.7962970473341674, + "learning_rate": 3.4630201358542476e-05, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08686232566833496, + "step": 4210, + "valid_targets_mean": 1373.9, + "valid_targets_min": 754 + }, + { + "epoch": 2.207962283918282, + "grad_norm": 0.7923508786740827, + "learning_rate": 3.461237719112872e-05, + "loss": 0.1969, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11983009427785873, + "step": 4215, + "valid_targets_mean": 1645.9, + "valid_targets_min": 869 + }, + { + "epoch": 2.210581456259822, + "grad_norm": 0.8850069176240013, + "learning_rate": 3.459452809395097e-05, + "loss": 0.1983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0863831490278244, + "step": 4220, + "valid_targets_mean": 1497.4, + "valid_targets_min": 731 + }, + { + "epoch": 2.213200628601362, + "grad_norm": 0.7261723990510998, + "learning_rate": 3.457665409746103e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08254241943359375, + "step": 4225, + "valid_targets_mean": 1338.2, + "valid_targets_min": 761 + }, + { + "epoch": 2.215819800942902, + "grad_norm": 0.7612789632096344, + "learning_rate": 3.455875523215325e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10166557133197784, + "step": 4230, + "valid_targets_mean": 1536.1, + "valid_targets_min": 1150 + }, + { + "epoch": 2.2184389732844423, + "grad_norm": 0.7310375700166778, + "learning_rate": 3.454083152856435e-05, + "loss": 0.1872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10035544633865356, + "step": 4235, + "valid_targets_mean": 1645.2, + "valid_targets_min": 1309 + }, + { + "epoch": 2.221058145625982, + "grad_norm": 0.7606843433820253, + "learning_rate": 3.452288301727348e-05, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056063294410705566, + "step": 4240, + "valid_targets_mean": 1178.9, + "valid_targets_min": 835 + }, + { + "epoch": 2.2236773179675224, + "grad_norm": 0.7668872824325965, + "learning_rate": 3.450490972890208e-05, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10496079921722412, + "step": 4245, + "valid_targets_mean": 1589.1, + "valid_targets_min": 1046 + }, + { + "epoch": 2.2262964903090623, + "grad_norm": 0.706948851912517, + "learning_rate": 3.448691169411386e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0922703966498375, + "step": 4250, + "valid_targets_mean": 1562.2, + "valid_targets_min": 797 + }, + { + "epoch": 2.2289156626506026, + "grad_norm": 0.7098704824335834, + "learning_rate": 3.446888894361477e-05, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08851149678230286, + "step": 4255, + "valid_targets_mean": 1487.0, + "valid_targets_min": 570 + }, + { + "epoch": 2.2315348349921424, + "grad_norm": 0.7441872683118494, + "learning_rate": 3.445084150815292e-05, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07137249410152435, + "step": 4260, + "valid_targets_mean": 1291.1, + "valid_targets_min": 548 + }, + { + "epoch": 2.2341540073336827, + "grad_norm": 0.7976270831681815, + "learning_rate": 3.443276941851851e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09706650674343109, + "step": 4265, + "valid_targets_mean": 1432.6, + "valid_targets_min": 682 + }, + { + "epoch": 2.2367731796752226, + "grad_norm": 1.122013582019588, + "learning_rate": 3.4414672705543836e-05, + "loss": 0.1966, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10958822071552277, + "step": 4270, + "valid_targets_mean": 1758.6, + "valid_targets_min": 896 + }, + { + "epoch": 2.239392352016763, + "grad_norm": 0.7290821910749447, + "learning_rate": 3.43965514001032e-05, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11781454086303711, + "step": 4275, + "valid_targets_mean": 1769.0, + "valid_targets_min": 597 + }, + { + "epoch": 2.2420115243583028, + "grad_norm": 0.7047571065780476, + "learning_rate": 3.437840553311282e-05, + "loss": 0.1962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07106795907020569, + "step": 4280, + "valid_targets_mean": 1352.9, + "valid_targets_min": 706 + }, + { + "epoch": 2.244630696699843, + "grad_norm": 0.7098750856445948, + "learning_rate": 3.436023513553087e-05, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07538577914237976, + "step": 4285, + "valid_targets_mean": 1328.0, + "valid_targets_min": 754 + }, + { + "epoch": 2.247249869041383, + "grad_norm": 0.6768651521900432, + "learning_rate": 3.434204023835735e-05, + "loss": 0.1882, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09894634038209915, + "step": 4290, + "valid_targets_mean": 1559.1, + "valid_targets_min": 819 + }, + { + "epoch": 2.249869041382923, + "grad_norm": 0.7043317440115251, + "learning_rate": 3.432382087263405e-05, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07986357808113098, + "step": 4295, + "valid_targets_mean": 1364.0, + "valid_targets_min": 684 + }, + { + "epoch": 2.252488213724463, + "grad_norm": 0.7979918153874863, + "learning_rate": 3.430557706944452e-05, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10084125399589539, + "step": 4300, + "valid_targets_mean": 1220.1, + "valid_targets_min": 711 + }, + { + "epoch": 2.255107386066003, + "grad_norm": 0.7410258036395652, + "learning_rate": 3.4287308859914e-05, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09822061657905579, + "step": 4305, + "valid_targets_mean": 1694.8, + "valid_targets_min": 1184 + }, + { + "epoch": 2.257726558407543, + "grad_norm": 0.7195779531702623, + "learning_rate": 3.426901627520937e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08750243484973907, + "step": 4310, + "valid_targets_mean": 1366.5, + "valid_targets_min": 774 + }, + { + "epoch": 2.2603457307490835, + "grad_norm": 0.6513345082189461, + "learning_rate": 3.425069934653907e-05, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09453466534614563, + "step": 4315, + "valid_targets_mean": 1786.2, + "valid_targets_min": 734 + }, + { + "epoch": 2.2629649030906234, + "grad_norm": 0.7390912800899102, + "learning_rate": 3.423235810515313e-05, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08612161874771118, + "step": 4320, + "valid_targets_mean": 1416.6, + "valid_targets_min": 1011 + }, + { + "epoch": 2.2655840754321632, + "grad_norm": 1.0368501460776134, + "learning_rate": 3.4213992582343e-05, + "loss": 0.1869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1044670045375824, + "step": 4325, + "valid_targets_mean": 1608.4, + "valid_targets_min": 843 + }, + { + "epoch": 2.2682032477737035, + "grad_norm": 0.7094198657277399, + "learning_rate": 3.419560280944158e-05, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09039437770843506, + "step": 4330, + "valid_targets_mean": 1298.5, + "valid_targets_min": 782 + }, + { + "epoch": 2.270822420115244, + "grad_norm": 0.7218013271471241, + "learning_rate": 3.417718881782315e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09985606372356415, + "step": 4335, + "valid_targets_mean": 1679.4, + "valid_targets_min": 646 + }, + { + "epoch": 2.2734415924567837, + "grad_norm": 0.7422215546439532, + "learning_rate": 3.41587506389033e-05, + "loss": 0.1905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10869371891021729, + "step": 4340, + "valid_targets_mean": 1493.9, + "valid_targets_min": 713 + }, + { + "epoch": 2.2760607647983235, + "grad_norm": 0.7704296977620054, + "learning_rate": 3.4140288304138885e-05, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.113609679043293, + "step": 4345, + "valid_targets_mean": 1921.1, + "valid_targets_min": 796 + }, + { + "epoch": 2.278679937139864, + "grad_norm": 0.6745332746403837, + "learning_rate": 3.4121801845027965e-05, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08519363403320312, + "step": 4350, + "valid_targets_mean": 1489.0, + "valid_targets_min": 903 + }, + { + "epoch": 2.2812991094814037, + "grad_norm": 0.7527579900176155, + "learning_rate": 3.410329129310977e-05, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09835608303546906, + "step": 4355, + "valid_targets_mean": 1340.1, + "valid_targets_min": 718 + }, + { + "epoch": 2.283918281822944, + "grad_norm": 0.6634487654273571, + "learning_rate": 3.4084756679964645e-05, + "loss": 0.1774, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08597914129495621, + "step": 4360, + "valid_targets_mean": 1230.5, + "valid_targets_min": 640 + }, + { + "epoch": 2.286537454164484, + "grad_norm": 0.7187577304897397, + "learning_rate": 3.4066198037213956e-05, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08094550669193268, + "step": 4365, + "valid_targets_mean": 1472.5, + "valid_targets_min": 662 + }, + { + "epoch": 2.289156626506024, + "grad_norm": 0.7038320551795163, + "learning_rate": 3.4047615396520073e-05, + "loss": 0.1833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1072240024805069, + "step": 4370, + "valid_targets_mean": 1496.6, + "valid_targets_min": 724 + }, + { + "epoch": 2.291775798847564, + "grad_norm": 0.7250600584269985, + "learning_rate": 3.402900878958633e-05, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10097962617874146, + "step": 4375, + "valid_targets_mean": 1489.9, + "valid_targets_min": 966 + }, + { + "epoch": 2.2943949711891043, + "grad_norm": 0.7011687906830416, + "learning_rate": 3.4010378248156925e-05, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07624346017837524, + "step": 4380, + "valid_targets_mean": 1243.5, + "valid_targets_min": 773 + }, + { + "epoch": 2.297014143530644, + "grad_norm": 0.7461275707907234, + "learning_rate": 3.39917238040169e-05, + "loss": 0.1842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05731279030442238, + "step": 4385, + "valid_targets_mean": 1047.5, + "valid_targets_min": 692 + }, + { + "epoch": 2.2996333158721844, + "grad_norm": 0.768348495929264, + "learning_rate": 3.397304548899207e-05, + "loss": 0.1875, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07975031435489655, + "step": 4390, + "valid_targets_mean": 1156.2, + "valid_targets_min": 776 + }, + { + "epoch": 2.3022524882137243, + "grad_norm": 0.6193993248648041, + "learning_rate": 3.395434333494899e-05, + "loss": 0.1801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12261155992746353, + "step": 4395, + "valid_targets_mean": 2362.4, + "valid_targets_min": 1195 + }, + { + "epoch": 2.3048716605552646, + "grad_norm": 0.667675832331397, + "learning_rate": 3.3935617373794875e-05, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0989714115858078, + "step": 4400, + "valid_targets_mean": 1732.0, + "valid_targets_min": 573 + }, + { + "epoch": 2.3074908328968045, + "grad_norm": 0.6511861740775647, + "learning_rate": 3.391686763747755e-05, + "loss": 0.186, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.101724773645401, + "step": 4405, + "valid_targets_mean": 1854.2, + "valid_targets_min": 754 + }, + { + "epoch": 2.3101100052383448, + "grad_norm": 0.7029757759828048, + "learning_rate": 3.3898094157985424e-05, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07613162696361542, + "step": 4410, + "valid_targets_mean": 1315.4, + "valid_targets_min": 841 + }, + { + "epoch": 2.3127291775798846, + "grad_norm": 0.7369305306221655, + "learning_rate": 3.3879296967347386e-05, + "loss": 0.1792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09967932105064392, + "step": 4415, + "valid_targets_mean": 1610.9, + "valid_targets_min": 956 + }, + { + "epoch": 2.315348349921425, + "grad_norm": 0.8052382865416222, + "learning_rate": 3.38604760976328e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09552525728940964, + "step": 4420, + "valid_targets_mean": 1279.4, + "valid_targets_min": 857 + }, + { + "epoch": 2.3179675222629648, + "grad_norm": 0.6434857609193912, + "learning_rate": 3.3841631580951414e-05, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09495727717876434, + "step": 4425, + "valid_targets_mean": 1312.8, + "valid_targets_min": 780 + }, + { + "epoch": 2.320586694604505, + "grad_norm": 0.660313314972628, + "learning_rate": 3.382276344945334e-05, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07882377505302429, + "step": 4430, + "valid_targets_mean": 1308.9, + "valid_targets_min": 677 + }, + { + "epoch": 2.323205866946045, + "grad_norm": 0.7239783525738875, + "learning_rate": 3.380387173532895e-05, + "loss": 0.1782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06723085045814514, + "step": 4435, + "valid_targets_mean": 1102.0, + "valid_targets_min": 727 + }, + { + "epoch": 2.325825039287585, + "grad_norm": 0.7057564939603069, + "learning_rate": 3.378495647080886e-05, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08943489193916321, + "step": 4440, + "valid_targets_mean": 1489.4, + "valid_targets_min": 726 + }, + { + "epoch": 2.328444211629125, + "grad_norm": 0.7953675082373738, + "learning_rate": 3.376601768816388e-05, + "loss": 0.1877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1159796416759491, + "step": 4445, + "valid_targets_mean": 1940.0, + "valid_targets_min": 669 + }, + { + "epoch": 2.3310633839706654, + "grad_norm": 0.6181280101129222, + "learning_rate": 3.374705541970494e-05, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10113310813903809, + "step": 4450, + "valid_targets_mean": 1807.5, + "valid_targets_min": 987 + }, + { + "epoch": 2.3336825563122052, + "grad_norm": 0.6823960843751892, + "learning_rate": 3.372806969778302e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08754397928714752, + "step": 4455, + "valid_targets_mean": 1522.1, + "valid_targets_min": 736 + }, + { + "epoch": 2.3363017286537455, + "grad_norm": 0.7786972149653073, + "learning_rate": 3.3709060554789114e-05, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12344034016132355, + "step": 4460, + "valid_targets_mean": 1859.2, + "valid_targets_min": 902 + }, + { + "epoch": 2.3389209009952854, + "grad_norm": 0.6819263485156057, + "learning_rate": 3.36900280231542e-05, + "loss": 0.18, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11073195934295654, + "step": 4465, + "valid_targets_mean": 1713.6, + "valid_targets_min": 753 + }, + { + "epoch": 2.3415400733368257, + "grad_norm": 0.698982737436564, + "learning_rate": 3.3670972135349136e-05, + "loss": 0.1773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09778452664613724, + "step": 4470, + "valid_targets_mean": 1662.1, + "valid_targets_min": 814 + }, + { + "epoch": 2.3441592456783655, + "grad_norm": 0.7532594699372448, + "learning_rate": 3.365189292388463e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09430748969316483, + "step": 4475, + "valid_targets_mean": 1488.2, + "valid_targets_min": 790 + }, + { + "epoch": 2.346778418019906, + "grad_norm": 0.7979166116865222, + "learning_rate": 3.36327904213112e-05, + "loss": 0.1698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10330542922019958, + "step": 4480, + "valid_targets_mean": 1607.1, + "valid_targets_min": 989 + }, + { + "epoch": 2.3493975903614457, + "grad_norm": 0.7719585566598948, + "learning_rate": 3.361366466021907e-05, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10411646217107773, + "step": 4485, + "valid_targets_mean": 1665.4, + "valid_targets_min": 826 + }, + { + "epoch": 2.352016762702986, + "grad_norm": 0.6934654348817533, + "learning_rate": 3.3594515673238164e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0847630649805069, + "step": 4490, + "valid_targets_mean": 1526.1, + "valid_targets_min": 907 + }, + { + "epoch": 2.354635935044526, + "grad_norm": 0.7478445096034769, + "learning_rate": 3.357534349303804e-05, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09797757863998413, + "step": 4495, + "valid_targets_mean": 1500.4, + "valid_targets_min": 1093 + }, + { + "epoch": 2.357255107386066, + "grad_norm": 0.7738779499608447, + "learning_rate": 3.3556148152327796e-05, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09403999149799347, + "step": 4500, + "valid_targets_mean": 1698.5, + "valid_targets_min": 1119 + }, + { + "epoch": 2.359874279727606, + "grad_norm": 0.8349671016469592, + "learning_rate": 3.3536929683856075e-05, + "loss": 0.1839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10461972653865814, + "step": 4505, + "valid_targets_mean": 1440.8, + "valid_targets_min": 700 + }, + { + "epoch": 2.3624934520691463, + "grad_norm": 0.6959363116916693, + "learning_rate": 3.351768812041095e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08033595234155655, + "step": 4510, + "valid_targets_mean": 1361.4, + "valid_targets_min": 810 + }, + { + "epoch": 2.365112624410686, + "grad_norm": 0.7724885095337418, + "learning_rate": 3.349842349481992e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07367488741874695, + "step": 4515, + "valid_targets_mean": 1202.1, + "valid_targets_min": 589 + }, + { + "epoch": 2.3677317967522264, + "grad_norm": 0.785656135977402, + "learning_rate": 3.347913583994982e-05, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1174265518784523, + "step": 4520, + "valid_targets_mean": 1551.0, + "valid_targets_min": 932 + }, + { + "epoch": 2.3703509690937663, + "grad_norm": 0.771454492922549, + "learning_rate": 3.345982518870676e-05, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10904380679130554, + "step": 4525, + "valid_targets_mean": 1306.9, + "valid_targets_min": 641 + }, + { + "epoch": 2.3729701414353066, + "grad_norm": 0.7132091224556273, + "learning_rate": 3.344049157403611e-05, + "loss": 0.1784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09000833332538605, + "step": 4530, + "valid_targets_mean": 1349.0, + "valid_targets_min": 701 + }, + { + "epoch": 2.3755893137768465, + "grad_norm": 0.7815019163491806, + "learning_rate": 3.342113502892239e-05, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09496913850307465, + "step": 4535, + "valid_targets_mean": 1900.6, + "valid_targets_min": 1202 + }, + { + "epoch": 2.3782084861183868, + "grad_norm": 0.6827548222687466, + "learning_rate": 3.340175558638927e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06467549502849579, + "step": 4540, + "valid_targets_mean": 1272.1, + "valid_targets_min": 657 + }, + { + "epoch": 2.3808276584599266, + "grad_norm": 0.6981246279697261, + "learning_rate": 3.338235327949946e-05, + "loss": 0.1859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07643115520477295, + "step": 4545, + "valid_targets_mean": 1320.8, + "valid_targets_min": 672 + }, + { + "epoch": 2.383446830801467, + "grad_norm": 0.6801257599230895, + "learning_rate": 3.3362928141354697e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09384497255086899, + "step": 4550, + "valid_targets_mean": 1583.4, + "valid_targets_min": 973 + }, + { + "epoch": 2.3860660031430068, + "grad_norm": 0.7026046569821849, + "learning_rate": 3.334348020509565e-05, + "loss": 0.1767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07073940336704254, + "step": 4555, + "valid_targets_mean": 1251.8, + "valid_targets_min": 732 + }, + { + "epoch": 2.388685175484547, + "grad_norm": 0.8580010038803302, + "learning_rate": 3.332400950390189e-05, + "loss": 0.185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07378994673490524, + "step": 4560, + "valid_targets_mean": 1250.2, + "valid_targets_min": 634 + }, + { + "epoch": 2.391304347826087, + "grad_norm": 0.6628888145564554, + "learning_rate": 3.330451607099185e-05, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0788576751947403, + "step": 4565, + "valid_targets_mean": 1478.1, + "valid_targets_min": 905 + }, + { + "epoch": 2.393923520167627, + "grad_norm": 0.6700112665820487, + "learning_rate": 3.328499993962271e-05, + "loss": 0.1739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0925033837556839, + "step": 4570, + "valid_targets_mean": 1563.1, + "valid_targets_min": 1080 + }, + { + "epoch": 2.396542692509167, + "grad_norm": 0.7392065080237232, + "learning_rate": 3.326546114309039e-05, + "loss": 0.1733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07319460809230804, + "step": 4575, + "valid_targets_mean": 1139.5, + "valid_targets_min": 687 + }, + { + "epoch": 2.3991618648507074, + "grad_norm": 0.6909972896034731, + "learning_rate": 3.324589971472949e-05, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09644930064678192, + "step": 4580, + "valid_targets_mean": 1775.0, + "valid_targets_min": 567 + }, + { + "epoch": 2.4017810371922472, + "grad_norm": 0.6146173418968691, + "learning_rate": 3.32263156879132e-05, + "loss": 0.1771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06641840934753418, + "step": 4585, + "valid_targets_mean": 1457.4, + "valid_targets_min": 810 + }, + { + "epoch": 2.4044002095337875, + "grad_norm": 0.6220474714158003, + "learning_rate": 3.320670909605328e-05, + "loss": 0.1745, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08907969295978546, + "step": 4590, + "valid_targets_mean": 1842.9, + "valid_targets_min": 1122 + }, + { + "epoch": 2.4070193818753274, + "grad_norm": 0.6654061377222593, + "learning_rate": 3.318707997259998e-05, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09738963097333908, + "step": 4595, + "valid_targets_mean": 1870.1, + "valid_targets_min": 934 + }, + { + "epoch": 2.4096385542168672, + "grad_norm": 0.6909269880681186, + "learning_rate": 3.3167428351042e-05, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10850033164024353, + "step": 4600, + "valid_targets_mean": 1894.8, + "valid_targets_min": 1318 + }, + { + "epoch": 2.4122577265584075, + "grad_norm": 0.689433160254431, + "learning_rate": 3.31477542649064e-05, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1018475890159607, + "step": 4605, + "valid_targets_mean": 1567.1, + "valid_targets_min": 1043 + }, + { + "epoch": 2.414876898899948, + "grad_norm": 0.6159510567189528, + "learning_rate": 3.31280577477586e-05, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11649971455335617, + "step": 4610, + "valid_targets_mean": 1980.1, + "valid_targets_min": 920 + }, + { + "epoch": 2.4174960712414877, + "grad_norm": 0.6730036682633775, + "learning_rate": 3.310833883320227e-05, + "loss": 0.1772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0985543429851532, + "step": 4615, + "valid_targets_mean": 1605.9, + "valid_targets_min": 847 + }, + { + "epoch": 2.4201152435830275, + "grad_norm": 0.6222403284088253, + "learning_rate": 3.3088597554879275e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09073705971240997, + "step": 4620, + "valid_targets_mean": 1654.4, + "valid_targets_min": 738 + }, + { + "epoch": 2.422734415924568, + "grad_norm": 0.7157821234797928, + "learning_rate": 3.3068833946469677e-05, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0750696137547493, + "step": 4625, + "valid_targets_mean": 1251.8, + "valid_targets_min": 645 + }, + { + "epoch": 2.425353588266108, + "grad_norm": 0.6027484316310288, + "learning_rate": 3.30490480416916e-05, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0879921168088913, + "step": 4630, + "valid_targets_mean": 2133.9, + "valid_targets_min": 1393 + }, + { + "epoch": 2.427972760607648, + "grad_norm": 0.5746699121713026, + "learning_rate": 3.302923987430121e-05, + "loss": 0.295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14860950410366058, + "step": 4635, + "valid_targets_mean": 3486.5, + "valid_targets_min": 770 + }, + { + "epoch": 2.430591932949188, + "grad_norm": 0.5147431052093084, + "learning_rate": 3.3009409478092664e-05, + "loss": 0.2788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12876978516578674, + "step": 4640, + "valid_targets_mean": 4703.8, + "valid_targets_min": 1799 + }, + { + "epoch": 2.433211105290728, + "grad_norm": 0.46143920934581817, + "learning_rate": 3.298955688689805e-05, + "loss": 0.269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13457268476486206, + "step": 4645, + "valid_targets_mean": 5730.6, + "valid_targets_min": 3015 + }, + { + "epoch": 2.4358302776322684, + "grad_norm": 0.45771665149533075, + "learning_rate": 3.296968213458731e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1267644464969635, + "step": 4650, + "valid_targets_mean": 5475.6, + "valid_targets_min": 1841 + }, + { + "epoch": 2.4384494499738083, + "grad_norm": 0.5676379990224636, + "learning_rate": 3.294978525506819e-05, + "loss": 0.2804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17016258835792542, + "step": 4655, + "valid_targets_mean": 4238.2, + "valid_targets_min": 1374 + }, + { + "epoch": 2.441068622315348, + "grad_norm": 0.5043507738641803, + "learning_rate": 3.29298662822862e-05, + "loss": 0.2754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14034417271614075, + "step": 4660, + "valid_targets_mean": 4297.9, + "valid_targets_min": 1951 + }, + { + "epoch": 2.4436877946568885, + "grad_norm": 0.4668729695806562, + "learning_rate": 3.290992525022455e-05, + "loss": 0.2638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10268481075763702, + "step": 4665, + "valid_targets_mean": 3986.4, + "valid_targets_min": 967 + }, + { + "epoch": 2.4463069669984283, + "grad_norm": 0.48272220883355993, + "learning_rate": 3.288996219290405e-05, + "loss": 0.2742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1456737220287323, + "step": 4670, + "valid_targets_mean": 5266.1, + "valid_targets_min": 1831 + }, + { + "epoch": 2.4489261393399686, + "grad_norm": 0.52937412926189, + "learning_rate": 3.2869977144383114e-05, + "loss": 0.2798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13847965002059937, + "step": 4675, + "valid_targets_mean": 3626.9, + "valid_targets_min": 2077 + }, + { + "epoch": 2.4515453116815085, + "grad_norm": 0.49959843017796596, + "learning_rate": 3.2849970138757675e-05, + "loss": 0.2718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12568604946136475, + "step": 4680, + "valid_targets_mean": 4431.5, + "valid_targets_min": 2834 + }, + { + "epoch": 2.4541644840230488, + "grad_norm": 0.5202151120133254, + "learning_rate": 3.2829941210161106e-05, + "loss": 0.2773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1487361490726471, + "step": 4685, + "valid_targets_mean": 4406.6, + "valid_targets_min": 1873 + }, + { + "epoch": 2.4567836563645886, + "grad_norm": 0.45934037131259303, + "learning_rate": 3.2809890392764216e-05, + "loss": 0.2662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12119745463132858, + "step": 4690, + "valid_targets_mean": 3873.5, + "valid_targets_min": 1013 + }, + { + "epoch": 2.459402828706129, + "grad_norm": 0.5223002576439512, + "learning_rate": 3.2789817720775126e-05, + "loss": 0.2732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14195717871189117, + "step": 4695, + "valid_targets_mean": 4464.8, + "valid_targets_min": 1939 + }, + { + "epoch": 2.4620220010476688, + "grad_norm": 0.5548829495535289, + "learning_rate": 3.2769723228439265e-05, + "loss": 0.269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16730615496635437, + "step": 4700, + "valid_targets_mean": 4397.6, + "valid_targets_min": 2391 + }, + { + "epoch": 2.464641173389209, + "grad_norm": 0.5899416470130255, + "learning_rate": 3.274960695003927e-05, + "loss": 0.28, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1873127520084381, + "step": 4705, + "valid_targets_mean": 4258.1, + "valid_targets_min": 1589 + }, + { + "epoch": 2.467260345730749, + "grad_norm": 0.5218022421153831, + "learning_rate": 3.272946891989495e-05, + "loss": 0.2647, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12697389721870422, + "step": 4710, + "valid_targets_mean": 4111.2, + "valid_targets_min": 1840 + }, + { + "epoch": 2.4698795180722892, + "grad_norm": 0.5402231980986321, + "learning_rate": 3.2709309172363236e-05, + "loss": 0.2731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13955043256282806, + "step": 4715, + "valid_targets_mean": 3903.1, + "valid_targets_min": 1269 + }, + { + "epoch": 2.472498690413829, + "grad_norm": 0.5928751992955947, + "learning_rate": 3.268912774183811e-05, + "loss": 0.2858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13107025623321533, + "step": 4720, + "valid_targets_mean": 3429.4, + "valid_targets_min": 1695 + }, + { + "epoch": 2.4751178627553694, + "grad_norm": 0.5582455207541235, + "learning_rate": 3.2668924662750535e-05, + "loss": 0.2836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.17198556661605835, + "step": 4725, + "valid_targets_mean": 4396.5, + "valid_targets_min": 1542 + }, + { + "epoch": 2.4777370350969092, + "grad_norm": 0.5635659000070361, + "learning_rate": 3.2648699969568415e-05, + "loss": 0.2673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12544940412044525, + "step": 4730, + "valid_targets_mean": 3763.1, + "valid_targets_min": 1438 + }, + { + "epoch": 2.4803562074384495, + "grad_norm": 0.601092825919765, + "learning_rate": 3.262845369679652e-05, + "loss": 0.2777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14525574445724487, + "step": 4735, + "valid_targets_mean": 4322.0, + "valid_targets_min": 2080 + }, + { + "epoch": 2.4829753797799894, + "grad_norm": 0.5392591846360169, + "learning_rate": 3.260818587897646e-05, + "loss": 0.2652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11406882107257843, + "step": 4740, + "valid_targets_mean": 3656.5, + "valid_targets_min": 1026 + }, + { + "epoch": 2.4855945521215297, + "grad_norm": 0.6370707935427085, + "learning_rate": 3.258789655068657e-05, + "loss": 0.2534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13879691064357758, + "step": 4745, + "valid_targets_mean": 3275.9, + "valid_targets_min": 1772 + }, + { + "epoch": 2.4882137244630695, + "grad_norm": 0.5865762731253249, + "learning_rate": 3.256758574654191e-05, + "loss": 0.2717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14946459233760834, + "step": 4750, + "valid_targets_mean": 3340.6, + "valid_targets_min": 1588 + }, + { + "epoch": 2.49083289680461, + "grad_norm": 0.5982332036655983, + "learning_rate": 3.254725350119416e-05, + "loss": 0.2718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1254604309797287, + "step": 4755, + "valid_targets_mean": 3175.5, + "valid_targets_min": 1276 + }, + { + "epoch": 2.4934520691461497, + "grad_norm": 0.49700938242861475, + "learning_rate": 3.2526899849331604e-05, + "loss": 0.2659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11317140609025955, + "step": 4760, + "valid_targets_mean": 4294.9, + "valid_targets_min": 1631 + }, + { + "epoch": 2.49607124148769, + "grad_norm": 0.506393655694655, + "learning_rate": 3.250652482567902e-05, + "loss": 0.2702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15846803784370422, + "step": 4765, + "valid_targets_mean": 5289.8, + "valid_targets_min": 3090 + }, + { + "epoch": 2.49869041382923, + "grad_norm": 0.5642257213461602, + "learning_rate": 3.248612846499767e-05, + "loss": 0.2582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13269807398319244, + "step": 4770, + "valid_targets_mean": 3234.0, + "valid_targets_min": 1446 + }, + { + "epoch": 2.50130958617077, + "grad_norm": 0.6190633867102372, + "learning_rate": 3.24657108020852e-05, + "loss": 0.2743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13396970927715302, + "step": 4775, + "valid_targets_mean": 3117.9, + "valid_targets_min": 1685 + }, + { + "epoch": 2.50392875851231, + "grad_norm": 0.5519889067789996, + "learning_rate": 3.244527187177561e-05, + "loss": 0.2596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12618090212345123, + "step": 4780, + "valid_targets_mean": 3541.6, + "valid_targets_min": 1596 + }, + { + "epoch": 2.5065479308538503, + "grad_norm": 0.5437322461668556, + "learning_rate": 3.2424811708939186e-05, + "loss": 0.2634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14589673280715942, + "step": 4785, + "valid_targets_mean": 5101.5, + "valid_targets_min": 2251 + }, + { + "epoch": 2.50916710319539, + "grad_norm": 0.6260981520646298, + "learning_rate": 3.240433034848243e-05, + "loss": 0.2615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12305879592895508, + "step": 4790, + "valid_targets_mean": 2717.9, + "valid_targets_min": 1458 + }, + { + "epoch": 2.5117862755369305, + "grad_norm": 0.5919409570928328, + "learning_rate": 3.238382782534802e-05, + "loss": 0.267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11653664708137512, + "step": 4795, + "valid_targets_mean": 3202.1, + "valid_targets_min": 1162 + }, + { + "epoch": 2.5144054478784703, + "grad_norm": 0.6142925456771567, + "learning_rate": 3.236330417451472e-05, + "loss": 0.2623, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1277245283126831, + "step": 4800, + "valid_targets_mean": 3012.8, + "valid_targets_min": 1480 + }, + { + "epoch": 2.5170246202200106, + "grad_norm": 0.5759770620381431, + "learning_rate": 3.234275943099736e-05, + "loss": 0.2735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1483064591884613, + "step": 4805, + "valid_targets_mean": 3682.4, + "valid_targets_min": 1229 + }, + { + "epoch": 2.5196437925615505, + "grad_norm": 0.5800436125578038, + "learning_rate": 3.232219362984674e-05, + "loss": 0.2676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14053216576576233, + "step": 4810, + "valid_targets_mean": 3763.1, + "valid_targets_min": 1276 + }, + { + "epoch": 2.5222629649030908, + "grad_norm": 0.5418468138092595, + "learning_rate": 3.23016068061496e-05, + "loss": 0.2624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11864572018384933, + "step": 4815, + "valid_targets_mean": 3172.8, + "valid_targets_min": 1208 + }, + { + "epoch": 2.5248821372446306, + "grad_norm": 0.6141798316789778, + "learning_rate": 3.228099899502853e-05, + "loss": 0.2597, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12610368430614471, + "step": 4820, + "valid_targets_mean": 3354.0, + "valid_targets_min": 924 + }, + { + "epoch": 2.527501309586171, + "grad_norm": 0.5666461799628468, + "learning_rate": 3.226037023164194e-05, + "loss": 0.2665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13936147093772888, + "step": 4825, + "valid_targets_mean": 4218.9, + "valid_targets_min": 1531 + }, + { + "epoch": 2.5301204819277108, + "grad_norm": 0.601054532159182, + "learning_rate": 3.223972055118398e-05, + "loss": 0.268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13349241018295288, + "step": 4830, + "valid_targets_mean": 3040.9, + "valid_targets_min": 1643 + }, + { + "epoch": 2.532739654269251, + "grad_norm": 0.5606930311169707, + "learning_rate": 3.221904998888448e-05, + "loss": 0.2573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15308856964111328, + "step": 4835, + "valid_targets_mean": 4129.6, + "valid_targets_min": 1244 + }, + { + "epoch": 2.535358826610791, + "grad_norm": 0.5499378228725914, + "learning_rate": 3.219835858000892e-05, + "loss": 0.2716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11361494660377502, + "step": 4840, + "valid_targets_mean": 3491.1, + "valid_targets_min": 730 + }, + { + "epoch": 2.5379779989523312, + "grad_norm": 0.5948091881039818, + "learning_rate": 3.217764635985831e-05, + "loss": 0.2645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11753956973552704, + "step": 4845, + "valid_targets_mean": 3085.4, + "valid_targets_min": 977 + }, + { + "epoch": 2.540597171293871, + "grad_norm": 0.5213409786562941, + "learning_rate": 3.215691336376919e-05, + "loss": 0.2559, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12941592931747437, + "step": 4850, + "valid_targets_mean": 4896.9, + "valid_targets_min": 2091 + }, + { + "epoch": 2.543216343635411, + "grad_norm": 0.5828548856997207, + "learning_rate": 3.213615962711354e-05, + "loss": 0.2551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10764909535646439, + "step": 4855, + "valid_targets_mean": 3125.9, + "valid_targets_min": 1654 + }, + { + "epoch": 2.5458355159769512, + "grad_norm": 0.5226309295403885, + "learning_rate": 3.211538518529872e-05, + "loss": 0.271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1245841532945633, + "step": 4860, + "valid_targets_mean": 4228.1, + "valid_targets_min": 1432 + }, + { + "epoch": 2.5484546883184915, + "grad_norm": 0.5288498707276463, + "learning_rate": 3.209459007376744e-05, + "loss": 0.2584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1541016697883606, + "step": 4865, + "valid_targets_mean": 4512.6, + "valid_targets_min": 2422 + }, + { + "epoch": 2.5510738606600314, + "grad_norm": 0.5881802772406762, + "learning_rate": 3.207377432799762e-05, + "loss": 0.2695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12315770238637924, + "step": 4870, + "valid_targets_mean": 3584.2, + "valid_targets_min": 1124 + }, + { + "epoch": 2.5536930330015712, + "grad_norm": 0.5504476564505342, + "learning_rate": 3.205293798350243e-05, + "loss": 0.2632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13040104508399963, + "step": 4875, + "valid_targets_mean": 3950.6, + "valid_targets_min": 1219 + }, + { + "epoch": 2.5563122053431115, + "grad_norm": 0.5675742742201335, + "learning_rate": 3.203208107583017e-05, + "loss": 0.275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1442309021949768, + "step": 4880, + "valid_targets_mean": 3843.0, + "valid_targets_min": 1328 + }, + { + "epoch": 2.558931377684652, + "grad_norm": 0.5899265411754202, + "learning_rate": 3.201120364056422e-05, + "loss": 0.2538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11218120157718658, + "step": 4885, + "valid_targets_mean": 2625.0, + "valid_targets_min": 916 + }, + { + "epoch": 2.5615505500261917, + "grad_norm": 0.5405042322855896, + "learning_rate": 3.199030571332299e-05, + "loss": 0.2716, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08701777458190918, + "step": 4890, + "valid_targets_mean": 3081.5, + "valid_targets_min": 1302 + }, + { + "epoch": 2.5641697223677316, + "grad_norm": 0.5942100313352924, + "learning_rate": 3.1969387329759823e-05, + "loss": 0.2617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18310458958148956, + "step": 4895, + "valid_targets_mean": 4390.5, + "valid_targets_min": 1500 + }, + { + "epoch": 2.566788894709272, + "grad_norm": 0.5098418097949111, + "learning_rate": 3.194844852556301e-05, + "loss": 0.2476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14060355722904205, + "step": 4900, + "valid_targets_mean": 4516.0, + "valid_targets_min": 1909 + }, + { + "epoch": 2.569408067050812, + "grad_norm": 0.5579946175688458, + "learning_rate": 3.1927489336455634e-05, + "loss": 0.2452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11655856668949127, + "step": 4905, + "valid_targets_mean": 3179.0, + "valid_targets_min": 809 + }, + { + "epoch": 2.572027239392352, + "grad_norm": 0.556264035056816, + "learning_rate": 3.190650979819558e-05, + "loss": 0.2701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12420783936977386, + "step": 4910, + "valid_targets_mean": 3833.9, + "valid_targets_min": 2662 + }, + { + "epoch": 2.574646411733892, + "grad_norm": 0.5212102410554529, + "learning_rate": 3.188550994657544e-05, + "loss": 0.2601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12855055928230286, + "step": 4915, + "valid_targets_mean": 4343.4, + "valid_targets_min": 1460 + }, + { + "epoch": 2.577265584075432, + "grad_norm": 0.4928515019150177, + "learning_rate": 3.186448981742249e-05, + "loss": 0.2614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1473008394241333, + "step": 4920, + "valid_targets_mean": 5864.5, + "valid_targets_min": 3945 + }, + { + "epoch": 2.5798847564169725, + "grad_norm": 0.5600108805198146, + "learning_rate": 3.184344944659855e-05, + "loss": 0.2615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15682847797870636, + "step": 4925, + "valid_targets_mean": 4052.5, + "valid_targets_min": 2504 + }, + { + "epoch": 2.5825039287585123, + "grad_norm": 0.5379148204137963, + "learning_rate": 3.182238887000002e-05, + "loss": 0.266, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15107464790344238, + "step": 4930, + "valid_targets_mean": 5593.9, + "valid_targets_min": 3596 + }, + { + "epoch": 2.585123101100052, + "grad_norm": 0.5868043253375208, + "learning_rate": 3.180130812355775e-05, + "loss": 0.268, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12723787128925323, + "step": 4935, + "valid_targets_mean": 3224.1, + "valid_targets_min": 1352 + }, + { + "epoch": 2.5877422734415925, + "grad_norm": 0.48086641376021894, + "learning_rate": 3.1780207243237e-05, + "loss": 0.2552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12989693880081177, + "step": 4940, + "valid_targets_mean": 5189.8, + "valid_targets_min": 1722 + }, + { + "epoch": 2.5903614457831328, + "grad_norm": 0.5274726333964913, + "learning_rate": 3.17590862650374e-05, + "loss": 0.2567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13214188814163208, + "step": 4945, + "valid_targets_mean": 4270.8, + "valid_targets_min": 2090 + }, + { + "epoch": 2.5929806181246726, + "grad_norm": 0.603974573481881, + "learning_rate": 3.173794522499284e-05, + "loss": 0.2513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13388465344905853, + "step": 4950, + "valid_targets_mean": 3596.2, + "valid_targets_min": 1731 + }, + { + "epoch": 2.5955997904662125, + "grad_norm": 0.6111527326141825, + "learning_rate": 3.171678415917146e-05, + "loss": 0.2538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13563989102840424, + "step": 4955, + "valid_targets_mean": 3311.6, + "valid_targets_min": 1460 + }, + { + "epoch": 2.5982189628077528, + "grad_norm": 0.5707197907356751, + "learning_rate": 3.169560310367556e-05, + "loss": 0.2443, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1540435254573822, + "step": 4960, + "valid_targets_mean": 4707.2, + "valid_targets_min": 2357 + }, + { + "epoch": 2.600838135149293, + "grad_norm": 0.7817120525865038, + "learning_rate": 3.1674402094641534e-05, + "loss": 0.2446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12648361921310425, + "step": 4965, + "valid_targets_mean": 3636.2, + "valid_targets_min": 1511 + }, + { + "epoch": 2.603457307490833, + "grad_norm": 0.6027690329614337, + "learning_rate": 3.1653181168239824e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14051169157028198, + "step": 4970, + "valid_targets_mean": 4583.0, + "valid_targets_min": 1764 + }, + { + "epoch": 2.606076479832373, + "grad_norm": 0.6410957048819071, + "learning_rate": 3.163194036067485e-05, + "loss": 0.2709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14867913722991943, + "step": 4975, + "valid_targets_mean": 3437.5, + "valid_targets_min": 1175 + }, + { + "epoch": 2.608695652173913, + "grad_norm": 0.5965422693887537, + "learning_rate": 3.161067970818496e-05, + "loss": 0.2604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09836294502019882, + "step": 4980, + "valid_targets_mean": 3982.1, + "valid_targets_min": 2546 + }, + { + "epoch": 2.6113148245154534, + "grad_norm": 0.6012462988490247, + "learning_rate": 3.158939924704236e-05, + "loss": 0.2624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14792278409004211, + "step": 4985, + "valid_targets_mean": 4124.0, + "valid_targets_min": 1244 + }, + { + "epoch": 2.6139339968569932, + "grad_norm": 0.5181497815578354, + "learning_rate": 3.156809901355303e-05, + "loss": 0.2541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13290362060070038, + "step": 4990, + "valid_targets_mean": 5047.1, + "valid_targets_min": 1856 + }, + { + "epoch": 2.616553169198533, + "grad_norm": 0.5470170145287421, + "learning_rate": 3.154677904405671e-05, + "loss": 0.253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1422726958990097, + "step": 4995, + "valid_targets_mean": 3927.4, + "valid_targets_min": 1948 + }, + { + "epoch": 2.6191723415400734, + "grad_norm": 0.5294702194312926, + "learning_rate": 3.152543937492678e-05, + "loss": 0.2478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12066562473773956, + "step": 5000, + "valid_targets_mean": 3516.5, + "valid_targets_min": 1248 + }, + { + "epoch": 2.6217915138816132, + "grad_norm": 0.47725403747545225, + "learning_rate": 3.150408004257026e-05, + "loss": 0.2506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12193520367145538, + "step": 5005, + "valid_targets_mean": 4611.0, + "valid_targets_min": 1145 + }, + { + "epoch": 2.6244106862231535, + "grad_norm": 0.5671363002346567, + "learning_rate": 3.1482701083427696e-05, + "loss": 0.255, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12533944845199585, + "step": 5010, + "valid_targets_mean": 3262.6, + "valid_targets_min": 2058 + }, + { + "epoch": 2.6270298585646934, + "grad_norm": 0.5284193073769989, + "learning_rate": 3.1461302533973136e-05, + "loss": 0.2532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14680202305316925, + "step": 5015, + "valid_targets_mean": 4546.5, + "valid_targets_min": 1708 + }, + { + "epoch": 2.6296490309062337, + "grad_norm": 0.5448716807282528, + "learning_rate": 3.143988443071403e-05, + "loss": 0.2484, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11072678864002228, + "step": 5020, + "valid_targets_mean": 3555.8, + "valid_targets_min": 1301 + }, + { + "epoch": 2.6322682032477736, + "grad_norm": 0.9661070312186927, + "learning_rate": 3.141844681019121e-05, + "loss": 0.2706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09883260726928711, + "step": 5025, + "valid_targets_mean": 3068.1, + "valid_targets_min": 1971 + }, + { + "epoch": 2.634887375589314, + "grad_norm": 0.5272848414165704, + "learning_rate": 3.139698970897879e-05, + "loss": 0.2664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16034264862537384, + "step": 5030, + "valid_targets_mean": 5172.4, + "valid_targets_min": 1054 + }, + { + "epoch": 2.6375065479308537, + "grad_norm": 0.5991507186668584, + "learning_rate": 3.1375513163684116e-05, + "loss": 0.2607, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14417916536331177, + "step": 5035, + "valid_targets_mean": 3143.4, + "valid_targets_min": 1612 + }, + { + "epoch": 2.640125720272394, + "grad_norm": 0.5456148861170084, + "learning_rate": 3.135401721094773e-05, + "loss": 0.2698, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11884725093841553, + "step": 5040, + "valid_targets_mean": 3499.6, + "valid_targets_min": 1274 + }, + { + "epoch": 2.642744892613934, + "grad_norm": 0.5901230380413761, + "learning_rate": 3.1332501887443264e-05, + "loss": 0.2596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13044601678848267, + "step": 5045, + "valid_targets_mean": 3628.4, + "valid_targets_min": 1625 + }, + { + "epoch": 2.645364064955474, + "grad_norm": 0.5335049214778567, + "learning_rate": 3.13109672298774e-05, + "loss": 0.2681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1493724286556244, + "step": 5050, + "valid_targets_mean": 4642.5, + "valid_targets_min": 1381 + }, + { + "epoch": 2.647983237297014, + "grad_norm": 0.6157842844248287, + "learning_rate": 3.128941327498982e-05, + "loss": 0.2707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15138354897499084, + "step": 5055, + "valid_targets_mean": 3540.1, + "valid_targets_min": 1591 + }, + { + "epoch": 2.6506024096385543, + "grad_norm": 0.6470346869905242, + "learning_rate": 3.12678400595531e-05, + "loss": 0.254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11125011742115021, + "step": 5060, + "valid_targets_mean": 3714.9, + "valid_targets_min": 1227 + }, + { + "epoch": 2.653221581980094, + "grad_norm": 0.574747316982881, + "learning_rate": 3.1246247620372705e-05, + "loss": 0.2598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.16843965649604797, + "step": 5065, + "valid_targets_mean": 4004.1, + "valid_targets_min": 1409 + }, + { + "epoch": 2.6558407543216345, + "grad_norm": 0.6192484248763278, + "learning_rate": 3.122463599428688e-05, + "loss": 0.2617, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14312565326690674, + "step": 5070, + "valid_targets_mean": 2940.8, + "valid_targets_min": 885 + }, + { + "epoch": 2.6584599266631743, + "grad_norm": 0.6113358057498799, + "learning_rate": 3.120300521816662e-05, + "loss": 0.2666, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13297387957572937, + "step": 5075, + "valid_targets_mean": 3441.9, + "valid_targets_min": 814 + }, + { + "epoch": 2.6610790990047146, + "grad_norm": 0.525732354390588, + "learning_rate": 3.118135532891557e-05, + "loss": 0.26, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1263016164302826, + "step": 5080, + "valid_targets_mean": 4079.0, + "valid_targets_min": 2331 + }, + { + "epoch": 2.6636982713462545, + "grad_norm": 0.4933878793248295, + "learning_rate": 3.1159686363470004e-05, + "loss": 0.2406, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1353895366191864, + "step": 5085, + "valid_targets_mean": 5371.5, + "valid_targets_min": 2618 + }, + { + "epoch": 2.6663174436877948, + "grad_norm": 0.5931054244629573, + "learning_rate": 3.1137998358798725e-05, + "loss": 0.2552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12467679381370544, + "step": 5090, + "valid_targets_mean": 3675.1, + "valid_targets_min": 1434 + }, + { + "epoch": 2.6689366160293346, + "grad_norm": 0.5652239242300231, + "learning_rate": 3.1116291351903024e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13629873096942902, + "step": 5095, + "valid_targets_mean": 3813.4, + "valid_targets_min": 1074 + }, + { + "epoch": 2.671555788370875, + "grad_norm": 0.5950160330240889, + "learning_rate": 3.109456537981662e-05, + "loss": 0.2594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12180252373218536, + "step": 5100, + "valid_targets_mean": 3690.8, + "valid_targets_min": 980 + }, + { + "epoch": 2.674174960712415, + "grad_norm": 0.5511113430816735, + "learning_rate": 3.107282047960557e-05, + "loss": 0.2533, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1455768644809723, + "step": 5105, + "valid_targets_mean": 4903.5, + "valid_targets_min": 2366 + }, + { + "epoch": 2.676794133053955, + "grad_norm": 0.5472510916701973, + "learning_rate": 3.1051056688368245e-05, + "loss": 0.2614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13943399488925934, + "step": 5110, + "valid_targets_mean": 4654.6, + "valid_targets_min": 1460 + }, + { + "epoch": 2.679413305395495, + "grad_norm": 0.6540388241465086, + "learning_rate": 3.102927404323524e-05, + "loss": 0.2508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13243556022644043, + "step": 5115, + "valid_targets_mean": 3612.0, + "valid_targets_min": 1351 + }, + { + "epoch": 2.6820324777370352, + "grad_norm": 0.588325110523456, + "learning_rate": 3.1007472581369284e-05, + "loss": 0.254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1372610628604889, + "step": 5120, + "valid_targets_mean": 3632.5, + "valid_targets_min": 2014 + }, + { + "epoch": 2.684651650078575, + "grad_norm": 0.5712435504981204, + "learning_rate": 3.098565233996525e-05, + "loss": 0.2599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11531564593315125, + "step": 5125, + "valid_targets_mean": 3135.9, + "valid_targets_min": 1084 + }, + { + "epoch": 2.6872708224201154, + "grad_norm": 0.5784760452375123, + "learning_rate": 3.096381335625005e-05, + "loss": 0.251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13585948944091797, + "step": 5130, + "valid_targets_mean": 3479.2, + "valid_targets_min": 1618 + }, + { + "epoch": 2.6898899947616552, + "grad_norm": 0.6183374457109351, + "learning_rate": 3.094195566748254e-05, + "loss": 0.2399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13121755421161652, + "step": 5135, + "valid_targets_mean": 3399.5, + "valid_targets_min": 1747 + }, + { + "epoch": 2.6925091671031955, + "grad_norm": 0.5335837326048039, + "learning_rate": 3.092007931095351e-05, + "loss": 0.249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1428021937608719, + "step": 5140, + "valid_targets_mean": 5037.9, + "valid_targets_min": 2092 + }, + { + "epoch": 2.6951283394447354, + "grad_norm": 0.5830793336264763, + "learning_rate": 3.0898184323985594e-05, + "loss": 0.2482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1379297971725464, + "step": 5145, + "valid_targets_mean": 3843.4, + "valid_targets_min": 2786 + }, + { + "epoch": 2.6977475117862757, + "grad_norm": 0.49231767272466725, + "learning_rate": 3.0876270743933214e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10020609945058823, + "step": 5150, + "valid_targets_mean": 4153.8, + "valid_targets_min": 1246 + }, + { + "epoch": 2.7003666841278156, + "grad_norm": 0.5692420998825817, + "learning_rate": 3.08543386081825e-05, + "loss": 0.2423, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10907959938049316, + "step": 5155, + "valid_targets_mean": 3389.8, + "valid_targets_min": 1484 + }, + { + "epoch": 2.702985856469356, + "grad_norm": 0.5131763488003471, + "learning_rate": 3.083238795415126e-05, + "loss": 0.2468, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13922926783561707, + "step": 5160, + "valid_targets_mean": 4818.0, + "valid_targets_min": 3185 + }, + { + "epoch": 2.7056050288108957, + "grad_norm": 0.5399148208715924, + "learning_rate": 3.0810418819288886e-05, + "loss": 0.2472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11872880160808563, + "step": 5165, + "valid_targets_mean": 3818.9, + "valid_targets_min": 1530 + }, + { + "epoch": 2.7082242011524356, + "grad_norm": 0.6266206857666278, + "learning_rate": 3.0788431241076286e-05, + "loss": 0.2709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14760111272335052, + "step": 5170, + "valid_targets_mean": 3010.2, + "valid_targets_min": 887 + }, + { + "epoch": 2.710843373493976, + "grad_norm": 0.5538974890939105, + "learning_rate": 3.076642525702586e-05, + "loss": 0.2434, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12455149739980698, + "step": 5175, + "valid_targets_mean": 4294.1, + "valid_targets_min": 2525 + }, + { + "epoch": 2.713462545835516, + "grad_norm": 0.5661799469653299, + "learning_rate": 3.074440090468138e-05, + "loss": 0.2463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12784098088741302, + "step": 5180, + "valid_targets_mean": 4256.4, + "valid_targets_min": 1759 + }, + { + "epoch": 2.716081718177056, + "grad_norm": 0.5249362591243726, + "learning_rate": 3.072235822161799e-05, + "loss": 0.2452, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12351245433092117, + "step": 5185, + "valid_targets_mean": 4883.9, + "valid_targets_min": 2155 + }, + { + "epoch": 2.718700890518596, + "grad_norm": 0.5769694312613989, + "learning_rate": 3.070029724544208e-05, + "loss": 0.2551, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1250094622373581, + "step": 5190, + "valid_targets_mean": 3758.9, + "valid_targets_min": 2048 + }, + { + "epoch": 2.721320062860136, + "grad_norm": 0.6012125163632736, + "learning_rate": 3.067821801379125e-05, + "loss": 0.2511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10830901563167572, + "step": 5195, + "valid_targets_mean": 2550.0, + "valid_targets_min": 1233 + }, + { + "epoch": 2.7239392352016765, + "grad_norm": 0.5393212791635524, + "learning_rate": 3.0656120564334265e-05, + "loss": 0.2572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09750883281230927, + "step": 5200, + "valid_targets_mean": 2796.5, + "valid_targets_min": 732 + }, + { + "epoch": 2.7265584075432163, + "grad_norm": 0.5302916892300166, + "learning_rate": 3.063400493477097e-05, + "loss": 0.2493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12134997546672821, + "step": 5205, + "valid_targets_mean": 4085.5, + "valid_targets_min": 1800 + }, + { + "epoch": 2.729177579884756, + "grad_norm": 0.5786293338432769, + "learning_rate": 3.061187116283219e-05, + "loss": 0.253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12166250497102737, + "step": 5210, + "valid_targets_mean": 3429.1, + "valid_targets_min": 2327 + }, + { + "epoch": 2.7317967522262965, + "grad_norm": 0.5937421573606068, + "learning_rate": 3.0589719286279736e-05, + "loss": 0.2508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12522894144058228, + "step": 5215, + "valid_targets_mean": 3262.4, + "valid_targets_min": 1597 + }, + { + "epoch": 2.7344159245678368, + "grad_norm": 0.4247641196000233, + "learning_rate": 3.056754934290631e-05, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07670685648918152, + "step": 5220, + "valid_targets_mean": 3631.8, + "valid_targets_min": 2342 + }, + { + "epoch": 2.7370350969093766, + "grad_norm": 0.3762373477143184, + "learning_rate": 3.054536137053542e-05, + "loss": 0.1057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039833299815654755, + "step": 5225, + "valid_targets_mean": 3542.6, + "valid_targets_min": 2731 + }, + { + "epoch": 2.7396542692509165, + "grad_norm": 0.3549950623690654, + "learning_rate": 3.0523155407021343e-05, + "loss": 0.0987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03244061768054962, + "step": 5230, + "valid_targets_mean": 3569.4, + "valid_targets_min": 2927 + }, + { + "epoch": 2.742273441592457, + "grad_norm": 0.4012813053165466, + "learning_rate": 3.0500931490249032e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.069857656955719, + "step": 5235, + "valid_targets_mean": 3426.9, + "valid_targets_min": 2213 + }, + { + "epoch": 2.744892613933997, + "grad_norm": 0.4543359497680009, + "learning_rate": 3.04786896581341e-05, + "loss": 0.0949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05636785551905632, + "step": 5240, + "valid_targets_mean": 2539.8, + "valid_targets_min": 560 + }, + { + "epoch": 2.747511786275537, + "grad_norm": 0.40094251257185143, + "learning_rate": 3.0456429948622702e-05, + "loss": 0.1014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052668213844299316, + "step": 5245, + "valid_targets_mean": 2907.5, + "valid_targets_min": 1002 + }, + { + "epoch": 2.750130958617077, + "grad_norm": 0.6615971266257469, + "learning_rate": 3.04341523996915e-05, + "loss": 0.1629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07997287809848785, + "step": 5250, + "valid_targets_mean": 1458.6, + "valid_targets_min": 650 + }, + { + "epoch": 2.752750130958617, + "grad_norm": 0.3677402195090524, + "learning_rate": 3.041185704934759e-05, + "loss": 0.0989, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04826483130455017, + "step": 5255, + "valid_targets_mean": 3396.1, + "valid_targets_min": 1037 + }, + { + "epoch": 2.7553693033001574, + "grad_norm": 0.5157941638441402, + "learning_rate": 3.0389543935628438e-05, + "loss": 0.1244, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11214305460453033, + "step": 5260, + "valid_targets_mean": 2438.9, + "valid_targets_min": 883 + }, + { + "epoch": 2.7579884756416972, + "grad_norm": 0.239571558697111, + "learning_rate": 3.0367213096601822e-05, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031629160046577454, + "step": 5265, + "valid_targets_mean": 4958.4, + "valid_targets_min": 3285 + }, + { + "epoch": 2.760607647983237, + "grad_norm": 0.33528267197653405, + "learning_rate": 3.0344864570365752e-05, + "loss": 0.0925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039592109620571136, + "step": 5270, + "valid_targets_mean": 4092.4, + "valid_targets_min": 764 + }, + { + "epoch": 2.7632268203247774, + "grad_norm": 0.3669028453664444, + "learning_rate": 3.0322498395048422e-05, + "loss": 0.0986, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05135112628340721, + "step": 5275, + "valid_targets_mean": 3804.9, + "valid_targets_min": 1290 + }, + { + "epoch": 2.7658459926663177, + "grad_norm": 0.4655747166188781, + "learning_rate": 3.030011460880812e-05, + "loss": 0.0981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061896875500679016, + "step": 5280, + "valid_targets_mean": 2547.8, + "valid_targets_min": 729 + }, + { + "epoch": 2.7684651650078576, + "grad_norm": 0.3504688470923911, + "learning_rate": 3.027771324983321e-05, + "loss": 0.1366, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04357130080461502, + "step": 5285, + "valid_targets_mean": 4258.8, + "valid_targets_min": 1961 + }, + { + "epoch": 2.7710843373493974, + "grad_norm": 0.3077217186366561, + "learning_rate": 3.0255294356342e-05, + "loss": 0.0923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04204597324132919, + "step": 5290, + "valid_targets_mean": 3557.5, + "valid_targets_min": 1822 + }, + { + "epoch": 2.7737035096909377, + "grad_norm": 0.3900619780088872, + "learning_rate": 3.0232857966582746e-05, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05191579833626747, + "step": 5295, + "valid_targets_mean": 2159.6, + "valid_targets_min": 841 + }, + { + "epoch": 2.776322682032478, + "grad_norm": 0.3671184202142866, + "learning_rate": 3.0210404118833527e-05, + "loss": 0.0852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04537142813205719, + "step": 5300, + "valid_targets_mean": 2735.1, + "valid_targets_min": 732 + }, + { + "epoch": 2.778941854374018, + "grad_norm": 0.6422929422227227, + "learning_rate": 3.0187932851402225e-05, + "loss": 0.1075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06347440183162689, + "step": 5305, + "valid_targets_mean": 1669.0, + "valid_targets_min": 770 + }, + { + "epoch": 2.7815610267155577, + "grad_norm": 0.3846533986815464, + "learning_rate": 3.0165444202626433e-05, + "loss": 0.11, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04249446839094162, + "step": 5310, + "valid_targets_mean": 2415.0, + "valid_targets_min": 847 + }, + { + "epoch": 2.784180199057098, + "grad_norm": 0.41200280777426784, + "learning_rate": 3.01429382108734e-05, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05815918743610382, + "step": 5315, + "valid_targets_mean": 3279.4, + "valid_targets_min": 2224 + }, + { + "epoch": 2.786799371398638, + "grad_norm": 0.40412963754584913, + "learning_rate": 3.012041491453996e-05, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05089302361011505, + "step": 5320, + "valid_targets_mean": 2944.5, + "valid_targets_min": 1341 + }, + { + "epoch": 2.789418543740178, + "grad_norm": 0.3712719373241418, + "learning_rate": 3.0097874352052477e-05, + "loss": 0.1469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05846469849348068, + "step": 5325, + "valid_targets_mean": 3912.9, + "valid_targets_min": 3005 + }, + { + "epoch": 2.792037716081718, + "grad_norm": 0.37410591516756875, + "learning_rate": 3.0075316561866766e-05, + "loss": 0.0923, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04242626205086708, + "step": 5330, + "valid_targets_mean": 2799.0, + "valid_targets_min": 669 + }, + { + "epoch": 2.7946568884232583, + "grad_norm": 0.6385642702390517, + "learning_rate": 3.0052741582468034e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.077068030834198, + "step": 5335, + "valid_targets_mean": 1381.5, + "valid_targets_min": 702 + }, + { + "epoch": 2.797276060764798, + "grad_norm": 0.7020392259425521, + "learning_rate": 3.003014945237082e-05, + "loss": 0.1309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051507942378520966, + "step": 5340, + "valid_targets_mean": 869.9, + "valid_targets_min": 15 + }, + { + "epoch": 2.7998952331063385, + "grad_norm": 0.3459274266211798, + "learning_rate": 3.0007540210118916e-05, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05308345705270767, + "step": 5345, + "valid_targets_mean": 3761.5, + "valid_targets_min": 2827 + }, + { + "epoch": 2.8025144054478783, + "grad_norm": 0.43614324400601717, + "learning_rate": 2.998491389428531e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10737846791744232, + "step": 5350, + "valid_targets_mean": 3383.4, + "valid_targets_min": 2079 + }, + { + "epoch": 2.8051335777894186, + "grad_norm": 0.3752088181866319, + "learning_rate": 2.9962270543472125e-05, + "loss": 0.2561, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050318267196416855, + "step": 5355, + "valid_targets_mean": 3586.4, + "valid_targets_min": 1932 + }, + { + "epoch": 2.8077527501309585, + "grad_norm": 0.5085442655985506, + "learning_rate": 2.993961019631054e-05, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05650688707828522, + "step": 5360, + "valid_targets_mean": 1462.1, + "valid_targets_min": 820 + }, + { + "epoch": 2.810371922472499, + "grad_norm": 0.4556426702339277, + "learning_rate": 2.9916932891460732e-05, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10121418535709381, + "step": 5365, + "valid_targets_mean": 2709.1, + "valid_targets_min": 965 + }, + { + "epoch": 2.8129910948140386, + "grad_norm": 0.3330007424348888, + "learning_rate": 2.9894238667611816e-05, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05032259225845337, + "step": 5370, + "valid_targets_mean": 4498.0, + "valid_targets_min": 2262 + }, + { + "epoch": 2.815610267155579, + "grad_norm": 0.3833932090400698, + "learning_rate": 2.987152756348176e-05, + "loss": 0.0961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08112305402755737, + "step": 5375, + "valid_targets_mean": 4270.9, + "valid_targets_min": 2733 + }, + { + "epoch": 2.818229439497119, + "grad_norm": 0.37835029205336035, + "learning_rate": 2.9848799617817336e-05, + "loss": 0.1034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07339747250080109, + "step": 5380, + "valid_targets_mean": 4916.5, + "valid_targets_min": 3028 + }, + { + "epoch": 2.820848611838659, + "grad_norm": 0.313477156313271, + "learning_rate": 2.9826054869394054e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03555566817522049, + "step": 5385, + "valid_targets_mean": 4158.5, + "valid_targets_min": 3317 + }, + { + "epoch": 2.823467784180199, + "grad_norm": 0.3303703496531289, + "learning_rate": 2.9803293357016087e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041689179837703705, + "step": 5390, + "valid_targets_mean": 4017.0, + "valid_targets_min": 748 + }, + { + "epoch": 2.8260869565217392, + "grad_norm": 0.40742236420626987, + "learning_rate": 2.9780515119516208e-05, + "loss": 0.0957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05982310697436333, + "step": 5395, + "valid_targets_mean": 3627.6, + "valid_targets_min": 2904 + }, + { + "epoch": 2.828706128863279, + "grad_norm": 0.4401531933262904, + "learning_rate": 2.975772019575572e-05, + "loss": 0.0999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052537109702825546, + "step": 5400, + "valid_targets_mean": 2574.8, + "valid_targets_min": 1939 + }, + { + "epoch": 2.8313253012048194, + "grad_norm": 0.4116628006297776, + "learning_rate": 2.9734908624624407e-05, + "loss": 0.0835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04146304726600647, + "step": 5405, + "valid_targets_mean": 2599.0, + "valid_targets_min": 772 + }, + { + "epoch": 2.8339444735463593, + "grad_norm": 0.5245056345829602, + "learning_rate": 2.971208044504044e-05, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09298737347126007, + "step": 5410, + "valid_targets_mean": 2805.9, + "valid_targets_min": 1289 + }, + { + "epoch": 2.8365636458878996, + "grad_norm": 0.3664485337414988, + "learning_rate": 2.968923569595034e-05, + "loss": 0.0778, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042247503995895386, + "step": 5415, + "valid_targets_mean": 3667.9, + "valid_targets_min": 2730 + }, + { + "epoch": 2.8391828182294394, + "grad_norm": 0.3509961370106701, + "learning_rate": 2.9666374416328874e-05, + "loss": 0.0906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03986793011426926, + "step": 5420, + "valid_targets_mean": 2270.9, + "valid_targets_min": 616 + }, + { + "epoch": 2.8418019905709797, + "grad_norm": 0.4117277347359568, + "learning_rate": 2.9643496645179034e-05, + "loss": 0.0859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048586420714855194, + "step": 5425, + "valid_targets_mean": 4381.4, + "valid_targets_min": 3979 + }, + { + "epoch": 2.8444211629125196, + "grad_norm": 0.3907556582677194, + "learning_rate": 2.9620602421531944e-05, + "loss": 0.1026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06444384902715683, + "step": 5430, + "valid_targets_mean": 2635.8, + "valid_targets_min": 795 + }, + { + "epoch": 2.84704033525406, + "grad_norm": 0.3069314363237155, + "learning_rate": 2.959769178444678e-05, + "loss": 0.13, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043156784027814865, + "step": 5435, + "valid_targets_mean": 3182.4, + "valid_targets_min": 729 + }, + { + "epoch": 2.8496595075955997, + "grad_norm": 0.4965324430254591, + "learning_rate": 2.9574764773010743e-05, + "loss": 0.1245, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05202588066458702, + "step": 5440, + "valid_targets_mean": 1878.5, + "valid_targets_min": 745 + }, + { + "epoch": 2.85227867993714, + "grad_norm": 0.6067897837495378, + "learning_rate": 2.9551821426338958e-05, + "loss": 0.1133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09202402830123901, + "step": 5445, + "valid_targets_mean": 2249.8, + "valid_targets_min": 664 + }, + { + "epoch": 2.85489785227868, + "grad_norm": 0.43608150673484214, + "learning_rate": 2.9528861783574416e-05, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04395114257931709, + "step": 5450, + "valid_targets_mean": 2284.8, + "valid_targets_min": 890 + }, + { + "epoch": 2.85751702462022, + "grad_norm": 0.3226488476771746, + "learning_rate": 2.950588588388792e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04564734548330307, + "step": 5455, + "valid_targets_mean": 3838.2, + "valid_targets_min": 2440 + }, + { + "epoch": 2.86013619696176, + "grad_norm": 0.38167756032253647, + "learning_rate": 2.9482893766478e-05, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041757866740226746, + "step": 5460, + "valid_targets_mean": 3074.5, + "valid_targets_min": 840 + }, + { + "epoch": 2.8627553693033003, + "grad_norm": 0.45897008326321065, + "learning_rate": 2.9459885470570856e-05, + "loss": 0.1253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0498652383685112, + "step": 5465, + "valid_targets_mean": 3609.8, + "valid_targets_min": 1014 + }, + { + "epoch": 2.86537454164484, + "grad_norm": 0.41579464310523456, + "learning_rate": 2.9436861035420298e-05, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045070018619298935, + "step": 5470, + "valid_targets_mean": 1142.9, + "valid_targets_min": 485 + }, + { + "epoch": 2.8679937139863805, + "grad_norm": 0.43798990120505504, + "learning_rate": 2.9413820500307665e-05, + "loss": 0.0968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04424760863184929, + "step": 5475, + "valid_targets_mean": 2665.6, + "valid_targets_min": 605 + }, + { + "epoch": 2.8706128863279203, + "grad_norm": 0.29802022411304274, + "learning_rate": 2.939076390454176e-05, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035394035279750824, + "step": 5480, + "valid_targets_mean": 4443.1, + "valid_targets_min": 848 + }, + { + "epoch": 2.87323205866946, + "grad_norm": 0.364157884038563, + "learning_rate": 2.9367691287458786e-05, + "loss": 0.0934, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05612578243017197, + "step": 5485, + "valid_targets_mean": 4255.2, + "valid_targets_min": 884 + }, + { + "epoch": 2.8758512310110005, + "grad_norm": 0.34580992630502244, + "learning_rate": 2.9344602688422286e-05, + "loss": 0.0925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04737825319170952, + "step": 5490, + "valid_targets_mean": 3826.2, + "valid_targets_min": 1992 + }, + { + "epoch": 2.878470403352541, + "grad_norm": 0.4031281573671315, + "learning_rate": 2.932149814682308e-05, + "loss": 0.1066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06583282351493835, + "step": 5495, + "valid_targets_mean": 3189.2, + "valid_targets_min": 808 + }, + { + "epoch": 2.8810895756940806, + "grad_norm": 0.40597988889316455, + "learning_rate": 2.929837770207915e-05, + "loss": 0.106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05000992491841316, + "step": 5500, + "valid_targets_mean": 3188.5, + "valid_targets_min": 1814 + }, + { + "epoch": 2.8837087480356205, + "grad_norm": 0.4054835876368696, + "learning_rate": 2.927524139363565e-05, + "loss": 0.1114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056005027145147324, + "step": 5505, + "valid_targets_mean": 2962.6, + "valid_targets_min": 931 + }, + { + "epoch": 2.886327920377161, + "grad_norm": 0.31187928940718423, + "learning_rate": 2.925208926096478e-05, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043278176337480545, + "step": 5510, + "valid_targets_mean": 4098.5, + "valid_targets_min": 3848 + }, + { + "epoch": 2.888947092718701, + "grad_norm": 0.4968707756183297, + "learning_rate": 2.9228921343565733e-05, + "loss": 0.1021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05354234203696251, + "step": 5515, + "valid_targets_mean": 1851.2, + "valid_targets_min": 664 + }, + { + "epoch": 2.891566265060241, + "grad_norm": 0.3412669939198899, + "learning_rate": 2.9205737680964645e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05105409026145935, + "step": 5520, + "valid_targets_mean": 3527.9, + "valid_targets_min": 2309 + }, + { + "epoch": 2.894185437401781, + "grad_norm": 0.4081878903455051, + "learning_rate": 2.9182538312714506e-05, + "loss": 0.0831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04212471470236778, + "step": 5525, + "valid_targets_mean": 2707.0, + "valid_targets_min": 1003 + }, + { + "epoch": 2.896804609743321, + "grad_norm": 0.3128239949382503, + "learning_rate": 2.91593232783951e-05, + "loss": 0.1017, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03760182112455368, + "step": 5530, + "valid_targets_mean": 3107.2, + "valid_targets_min": 531 + }, + { + "epoch": 2.8994237820848614, + "grad_norm": 0.341513379082836, + "learning_rate": 2.913609261761295e-05, + "loss": 0.1016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04579591006040573, + "step": 5535, + "valid_targets_mean": 2845.9, + "valid_targets_min": 1003 + }, + { + "epoch": 2.9020429544264013, + "grad_norm": 0.4570396323728713, + "learning_rate": 2.9112846370001223e-05, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03929852321743965, + "step": 5540, + "valid_targets_mean": 1568.5, + "valid_targets_min": 597 + }, + { + "epoch": 2.904662126767941, + "grad_norm": 0.39337596059812086, + "learning_rate": 2.9089584575219688e-05, + "loss": 0.1, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03869905322790146, + "step": 5545, + "valid_targets_mean": 1980.2, + "valid_targets_min": 665 + }, + { + "epoch": 2.9072812991094814, + "grad_norm": 0.486176813679766, + "learning_rate": 2.906630727295463e-05, + "loss": 0.1191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046391211450099945, + "step": 5550, + "valid_targets_mean": 1768.5, + "valid_targets_min": 853 + }, + { + "epoch": 2.9099004714510217, + "grad_norm": 0.4739488386783305, + "learning_rate": 2.9043014502918807e-05, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07332383096218109, + "step": 5555, + "valid_targets_mean": 2583.2, + "valid_targets_min": 997 + }, + { + "epoch": 2.9125196437925616, + "grad_norm": 0.4979513025384694, + "learning_rate": 2.9019706304851348e-05, + "loss": 0.1134, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06526502966880798, + "step": 5560, + "valid_targets_mean": 2746.6, + "valid_targets_min": 740 + }, + { + "epoch": 2.9151388161341014, + "grad_norm": 0.30230530384320564, + "learning_rate": 2.8996382718517724e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037894781678915024, + "step": 5565, + "valid_targets_mean": 3555.2, + "valid_targets_min": 2887 + }, + { + "epoch": 2.9177579884756417, + "grad_norm": 0.3249073347826196, + "learning_rate": 2.8973043783709628e-05, + "loss": 0.095, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03607039898633957, + "step": 5570, + "valid_targets_mean": 3754.2, + "valid_targets_min": 2543 + }, + { + "epoch": 2.920377160817182, + "grad_norm": 0.4669138569456446, + "learning_rate": 2.894968954024497e-05, + "loss": 0.1254, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05429444462060928, + "step": 5575, + "valid_targets_mean": 1742.1, + "valid_targets_min": 446 + }, + { + "epoch": 2.922996333158722, + "grad_norm": 0.3782232445208203, + "learning_rate": 2.8926320027967763e-05, + "loss": 0.0876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03995092958211899, + "step": 5580, + "valid_targets_mean": 2742.4, + "valid_targets_min": 681 + }, + { + "epoch": 2.9256155055002617, + "grad_norm": 0.40296794796980645, + "learning_rate": 2.8902935286748066e-05, + "loss": 0.0967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07102715969085693, + "step": 5585, + "valid_targets_mean": 3844.6, + "valid_targets_min": 799 + }, + { + "epoch": 2.928234677841802, + "grad_norm": 0.3460413667162216, + "learning_rate": 2.8879535356481933e-05, + "loss": 0.0948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04568317532539368, + "step": 5590, + "valid_targets_mean": 3546.4, + "valid_targets_min": 940 + }, + { + "epoch": 2.9308538501833423, + "grad_norm": 0.2923979536789042, + "learning_rate": 2.8856120277091326e-05, + "loss": 0.0854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034960515797138214, + "step": 5595, + "valid_targets_mean": 3592.6, + "valid_targets_min": 861 + }, + { + "epoch": 2.933473022524882, + "grad_norm": 0.48491536058441137, + "learning_rate": 2.8832690088524045e-05, + "loss": 0.0922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06779657304286957, + "step": 5600, + "valid_targets_mean": 4078.1, + "valid_targets_min": 3024 + }, + { + "epoch": 2.936092194866422, + "grad_norm": 0.30899940131186365, + "learning_rate": 2.8809244830753667e-05, + "loss": 0.0976, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05456120893359184, + "step": 5605, + "valid_targets_mean": 3932.1, + "valid_targets_min": 3175 + }, + { + "epoch": 2.9387113672079623, + "grad_norm": 0.3401818959741277, + "learning_rate": 2.8785784543779497e-05, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048924513161182404, + "step": 5610, + "valid_targets_mean": 3798.4, + "valid_targets_min": 2652 + }, + { + "epoch": 2.9413305395495026, + "grad_norm": 0.44811555988296475, + "learning_rate": 2.8762309267626465e-05, + "loss": 0.1187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10950879007577896, + "step": 5615, + "valid_targets_mean": 2981.9, + "valid_targets_min": 764 + }, + { + "epoch": 2.9439497118910425, + "grad_norm": 0.42233886049722147, + "learning_rate": 2.8738819042345073e-05, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05082385241985321, + "step": 5620, + "valid_targets_mean": 1483.8, + "valid_targets_min": 815 + }, + { + "epoch": 2.9465688842325823, + "grad_norm": 0.37088343667755846, + "learning_rate": 2.8715313908011332e-05, + "loss": 0.1188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056612737476825714, + "step": 5625, + "valid_targets_mean": 3776.4, + "valid_targets_min": 2447 + }, + { + "epoch": 2.9491880565741226, + "grad_norm": 0.37472101978595784, + "learning_rate": 2.8691793904726685e-05, + "loss": 0.094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05304092913866043, + "step": 5630, + "valid_targets_mean": 3455.1, + "valid_targets_min": 854 + }, + { + "epoch": 2.9518072289156625, + "grad_norm": 0.322259649312754, + "learning_rate": 2.8668259072617946e-05, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033347636461257935, + "step": 5635, + "valid_targets_mean": 3706.0, + "valid_targets_min": 743 + }, + { + "epoch": 2.954426401257203, + "grad_norm": 0.4904364088147268, + "learning_rate": 2.864470945183724e-05, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09178433567285538, + "step": 5640, + "valid_targets_mean": 2587.1, + "valid_targets_min": 706 + }, + { + "epoch": 2.9570455735987426, + "grad_norm": 0.4047425338458928, + "learning_rate": 2.8621145082561886e-05, + "loss": 0.1404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09862107038497925, + "step": 5645, + "valid_targets_mean": 3028.0, + "valid_targets_min": 2457 + }, + { + "epoch": 2.959664745940283, + "grad_norm": 0.4132173267602903, + "learning_rate": 2.8597566004994407e-05, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04023023694753647, + "step": 5650, + "valid_targets_mean": 2147.2, + "valid_targets_min": 663 + }, + { + "epoch": 2.962283918281823, + "grad_norm": 0.45860496650064586, + "learning_rate": 2.8573972259362405e-05, + "loss": 0.0998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0492866113781929, + "step": 5655, + "valid_targets_mean": 1507.4, + "valid_targets_min": 716 + }, + { + "epoch": 2.964903090623363, + "grad_norm": 0.3703573130233748, + "learning_rate": 2.855036388591848e-05, + "loss": 0.0946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05824311822652817, + "step": 5660, + "valid_targets_mean": 3088.1, + "valid_targets_min": 760 + }, + { + "epoch": 2.967522262964903, + "grad_norm": 0.38526219067747297, + "learning_rate": 2.8526740924940234e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052568789571523666, + "step": 5665, + "valid_targets_mean": 3074.9, + "valid_targets_min": 920 + }, + { + "epoch": 2.9701414353064433, + "grad_norm": 0.3950348786637089, + "learning_rate": 2.8503103416730123e-05, + "loss": 0.0841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03225678578019142, + "step": 5670, + "valid_targets_mean": 997.5, + "valid_targets_min": 635 + }, + { + "epoch": 2.972760607647983, + "grad_norm": 0.3370166936949218, + "learning_rate": 2.8479451401615434e-05, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04371637478470802, + "step": 5675, + "valid_targets_mean": 2942.0, + "valid_targets_min": 920 + }, + { + "epoch": 2.9753797799895234, + "grad_norm": 0.38682761904474483, + "learning_rate": 2.8455784919948206e-05, + "loss": 0.0861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04161805659532547, + "step": 5680, + "valid_targets_mean": 1271.1, + "valid_targets_min": 704 + }, + { + "epoch": 2.9779989523310633, + "grad_norm": 0.37162212135591083, + "learning_rate": 2.8432104012105144e-05, + "loss": 0.0945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0344364196062088, + "step": 5685, + "valid_targets_mean": 2149.4, + "valid_targets_min": 529 + }, + { + "epoch": 2.9806181246726036, + "grad_norm": 0.41291850004413577, + "learning_rate": 2.8408408718487582e-05, + "loss": 0.104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060467127710580826, + "step": 5690, + "valid_targets_mean": 3471.9, + "valid_targets_min": 2017 + }, + { + "epoch": 2.9832372970141434, + "grad_norm": 0.29147832380156924, + "learning_rate": 2.8384699079521398e-05, + "loss": 0.0932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04975741356611252, + "step": 5695, + "valid_targets_mean": 4240.2, + "valid_targets_min": 516 + }, + { + "epoch": 2.9858564693556837, + "grad_norm": 0.3054199600735001, + "learning_rate": 2.836097513565693e-05, + "loss": 0.0859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041876085102558136, + "step": 5700, + "valid_targets_mean": 2583.9, + "valid_targets_min": 775 + }, + { + "epoch": 2.9884756416972236, + "grad_norm": 0.2869486183540889, + "learning_rate": 2.833723692736893e-05, + "loss": 0.0905, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04576297849416733, + "step": 5705, + "valid_targets_mean": 3741.6, + "valid_targets_min": 605 + }, + { + "epoch": 2.991094814038764, + "grad_norm": 0.46336310702083605, + "learning_rate": 2.8313484495156486e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1666039526462555, + "step": 5710, + "valid_targets_mean": 2971.0, + "valid_targets_min": 1907 + }, + { + "epoch": 2.9937139863803037, + "grad_norm": 0.34111806022981245, + "learning_rate": 2.828971787954295e-05, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04475843533873558, + "step": 5715, + "valid_targets_mean": 3138.5, + "valid_targets_min": 1235 + }, + { + "epoch": 2.996333158721844, + "grad_norm": 0.36114737073999276, + "learning_rate": 2.8265937121075873e-05, + "loss": 0.1003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045034777373075485, + "step": 5720, + "valid_targets_mean": 2935.4, + "valid_targets_min": 759 + }, + { + "epoch": 2.998952331063384, + "grad_norm": 0.42379165298960936, + "learning_rate": 2.8242142260326937e-05, + "loss": 0.093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06786138564348221, + "step": 5725, + "valid_targets_mean": 2050.0, + "valid_targets_min": 927 + }, + { + "epoch": 3.001571503404924, + "grad_norm": 1.1006848831604712, + "learning_rate": 2.8218333337891882e-05, + "loss": 0.1834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08112776279449463, + "step": 5730, + "valid_targets_mean": 6116.6, + "valid_targets_min": 4858 + }, + { + "epoch": 3.004190675746464, + "grad_norm": 0.37478351146646555, + "learning_rate": 2.8194510394390443e-05, + "loss": 0.1832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08934018015861511, + "step": 5735, + "valid_targets_mean": 6616.0, + "valid_targets_min": 4759 + }, + { + "epoch": 3.0068098480880043, + "grad_norm": 0.3704816978297245, + "learning_rate": 2.817067347046626e-05, + "loss": 0.1714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08171521127223969, + "step": 5740, + "valid_targets_mean": 5258.1, + "valid_targets_min": 4373 + }, + { + "epoch": 3.009429020429544, + "grad_norm": 0.3566983496955074, + "learning_rate": 2.814682260678684e-05, + "loss": 0.1747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09679475426673889, + "step": 5745, + "valid_targets_mean": 5849.6, + "valid_targets_min": 4813 + }, + { + "epoch": 3.0120481927710845, + "grad_norm": 0.3926702848216614, + "learning_rate": 2.8122957844043474e-05, + "loss": 0.1823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09491441398859024, + "step": 5750, + "valid_targets_mean": 5604.5, + "valid_targets_min": 3983 + }, + { + "epoch": 3.0146673651126243, + "grad_norm": 0.36650809914856325, + "learning_rate": 2.809907922295115e-05, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08622777462005615, + "step": 5755, + "valid_targets_mean": 6438.2, + "valid_targets_min": 5154 + }, + { + "epoch": 3.0172865374541646, + "grad_norm": 0.374103704808098, + "learning_rate": 2.8075186784248517e-05, + "loss": 0.1856, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07890258729457855, + "step": 5760, + "valid_targets_mean": 5509.6, + "valid_targets_min": 5040 + }, + { + "epoch": 3.0199057097957045, + "grad_norm": 0.4049974596952702, + "learning_rate": 2.805128056869779e-05, + "loss": 0.1726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11895795166492462, + "step": 5765, + "valid_targets_mean": 6410.5, + "valid_targets_min": 5066 + }, + { + "epoch": 3.022524882137245, + "grad_norm": 0.5834360534408537, + "learning_rate": 2.8027360617084684e-05, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0787908136844635, + "step": 5770, + "valid_targets_mean": 6616.2, + "valid_targets_min": 4500 + }, + { + "epoch": 3.0251440544787846, + "grad_norm": 0.37846258351854156, + "learning_rate": 2.800342697021835e-05, + "loss": 0.1769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.101019948720932, + "step": 5775, + "valid_targets_mean": 7507.6, + "valid_targets_min": 5455 + }, + { + "epoch": 3.027763226820325, + "grad_norm": 0.37528250186205403, + "learning_rate": 2.797947966893131e-05, + "loss": 0.1806, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09880384802818298, + "step": 5780, + "valid_targets_mean": 7383.8, + "valid_targets_min": 4715 + }, + { + "epoch": 3.030382399161865, + "grad_norm": 0.350766545229544, + "learning_rate": 2.7955518754079382e-05, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08563157916069031, + "step": 5785, + "valid_targets_mean": 6624.9, + "valid_targets_min": 5315 + }, + { + "epoch": 3.033001571503405, + "grad_norm": 0.37633052761397, + "learning_rate": 2.7931544266541596e-05, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09435087442398071, + "step": 5790, + "valid_targets_mean": 7394.5, + "valid_targets_min": 4353 + }, + { + "epoch": 3.035620743844945, + "grad_norm": 0.3468736036426763, + "learning_rate": 2.7907556247220166e-05, + "loss": 0.1616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08253604173660278, + "step": 5795, + "valid_targets_mean": 6296.0, + "valid_targets_min": 4890 + }, + { + "epoch": 3.0382399161864853, + "grad_norm": 0.3846274274020628, + "learning_rate": 2.788355473704036e-05, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07230623811483383, + "step": 5800, + "valid_targets_mean": 5848.5, + "valid_targets_min": 4560 + }, + { + "epoch": 3.040859088528025, + "grad_norm": 0.41890484429602914, + "learning_rate": 2.7859539776950478e-05, + "loss": 0.179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09072008728981018, + "step": 5805, + "valid_targets_mean": 5550.9, + "valid_targets_min": 3640 + }, + { + "epoch": 3.0434782608695654, + "grad_norm": 0.5071253976580835, + "learning_rate": 2.7835511407921765e-05, + "loss": 0.1754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0984182208776474, + "step": 5810, + "valid_targets_mean": 5894.4, + "valid_targets_min": 3967 + }, + { + "epoch": 3.0460974332111053, + "grad_norm": 0.37387718199578546, + "learning_rate": 2.7811469670948348e-05, + "loss": 0.1546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08167785406112671, + "step": 5815, + "valid_targets_mean": 6153.9, + "valid_targets_min": 4521 + }, + { + "epoch": 3.0487166055526456, + "grad_norm": 0.3852439875067998, + "learning_rate": 2.778741460704715e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08166977763175964, + "step": 5820, + "valid_targets_mean": 6440.5, + "valid_targets_min": 5067 + }, + { + "epoch": 3.0513357778941854, + "grad_norm": 0.38629346612397264, + "learning_rate": 2.7763346257257843e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07773430645465851, + "step": 5825, + "valid_targets_mean": 6049.2, + "valid_targets_min": 4385 + }, + { + "epoch": 3.0539549502357257, + "grad_norm": 0.3421804696905982, + "learning_rate": 2.7739264662642754e-05, + "loss": 0.187, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07998912036418915, + "step": 5830, + "valid_targets_mean": 7771.6, + "valid_targets_min": 4566 + }, + { + "epoch": 3.0565741225772656, + "grad_norm": 0.4084496477496339, + "learning_rate": 2.771516986428681e-05, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08300460129976273, + "step": 5835, + "valid_targets_mean": 6633.6, + "valid_targets_min": 5067 + }, + { + "epoch": 3.059193294918806, + "grad_norm": 0.4375648221906735, + "learning_rate": 2.7691061903297466e-05, + "loss": 0.1687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0828254446387291, + "step": 5840, + "valid_targets_mean": 6862.6, + "valid_targets_min": 5178 + }, + { + "epoch": 3.0618124672603457, + "grad_norm": 0.36189186832677733, + "learning_rate": 2.7666940820804633e-05, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0890597552061081, + "step": 5845, + "valid_targets_mean": 6075.1, + "valid_targets_min": 3854 + }, + { + "epoch": 3.064431639601886, + "grad_norm": 0.389210733944373, + "learning_rate": 2.7642806657960613e-05, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10441012680530548, + "step": 5850, + "valid_targets_mean": 6840.1, + "valid_targets_min": 4743 + }, + { + "epoch": 3.067050811943426, + "grad_norm": 0.3999714076113869, + "learning_rate": 2.761865945594002e-05, + "loss": 0.2065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10989060997962952, + "step": 5855, + "valid_targets_mean": 6480.9, + "valid_targets_min": 4761 + }, + { + "epoch": 3.0696699842849657, + "grad_norm": 0.3721363018808316, + "learning_rate": 2.759449925593971e-05, + "loss": 0.1794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08316189795732498, + "step": 5860, + "valid_targets_mean": 6224.5, + "valid_targets_min": 4789 + }, + { + "epoch": 3.072289156626506, + "grad_norm": 0.39009824377853225, + "learning_rate": 2.7570326099178716e-05, + "loss": 0.1798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08990475535392761, + "step": 5865, + "valid_targets_mean": 6686.4, + "valid_targets_min": 5403 + }, + { + "epoch": 3.074908328968046, + "grad_norm": 0.3795726008737177, + "learning_rate": 2.754614002689818e-05, + "loss": 0.1732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08434467762708664, + "step": 5870, + "valid_targets_mean": 6285.0, + "valid_targets_min": 4970 + }, + { + "epoch": 3.077527501309586, + "grad_norm": 0.37637541868683716, + "learning_rate": 2.7521941080361275e-05, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10039840638637543, + "step": 5875, + "valid_targets_mean": 6926.9, + "valid_targets_min": 5299 + }, + { + "epoch": 3.080146673651126, + "grad_norm": 0.38022332019401267, + "learning_rate": 2.749772930085315e-05, + "loss": 0.1734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08434297889471054, + "step": 5880, + "valid_targets_mean": 6121.8, + "valid_targets_min": 4951 + }, + { + "epoch": 3.0827658459926663, + "grad_norm": 0.38806435552580565, + "learning_rate": 2.7473504729680823e-05, + "loss": 0.181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08984823524951935, + "step": 5885, + "valid_targets_mean": 8113.1, + "valid_targets_min": 4794 + }, + { + "epoch": 3.085385018334206, + "grad_norm": 0.41652425357471284, + "learning_rate": 2.744926740817316e-05, + "loss": 0.1824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0975089967250824, + "step": 5890, + "valid_targets_mean": 5281.6, + "valid_targets_min": 593 + }, + { + "epoch": 3.0880041906757465, + "grad_norm": 0.3331239672780886, + "learning_rate": 2.742501737768077e-05, + "loss": 0.1724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07473309338092804, + "step": 5895, + "valid_targets_mean": 7339.1, + "valid_targets_min": 4595 + }, + { + "epoch": 3.0906233630172864, + "grad_norm": 0.3794171172986709, + "learning_rate": 2.7400754679575948e-05, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07807215303182602, + "step": 5900, + "valid_targets_mean": 6674.0, + "valid_targets_min": 5411 + }, + { + "epoch": 3.0932425353588267, + "grad_norm": 0.3994468768709936, + "learning_rate": 2.7376479355252588e-05, + "loss": 0.1692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08967799693346024, + "step": 5905, + "valid_targets_mean": 6596.2, + "valid_targets_min": 4451 + }, + { + "epoch": 3.0958617077003665, + "grad_norm": 0.43304702006297846, + "learning_rate": 2.7352191446126145e-05, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09482640773057938, + "step": 5910, + "valid_targets_mean": 7159.2, + "valid_targets_min": 5568 + }, + { + "epoch": 3.098480880041907, + "grad_norm": 0.40796849630599863, + "learning_rate": 2.732789099363353e-05, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1004834994673729, + "step": 5915, + "valid_targets_mean": 6923.1, + "valid_targets_min": 4839 + }, + { + "epoch": 3.1011000523834467, + "grad_norm": 0.4091315333143437, + "learning_rate": 2.7303578039233055e-05, + "loss": 0.1703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08091986179351807, + "step": 5920, + "valid_targets_mean": 4921.4, + "valid_targets_min": 3517 + }, + { + "epoch": 3.103719224724987, + "grad_norm": 0.3928269940192115, + "learning_rate": 2.7279252624404374e-05, + "loss": 0.1643, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0792890191078186, + "step": 5925, + "valid_targets_mean": 5929.9, + "valid_targets_min": 3551 + }, + { + "epoch": 3.106338397066527, + "grad_norm": 0.39187360564620877, + "learning_rate": 2.7254914790648387e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07602248340845108, + "step": 5930, + "valid_targets_mean": 5854.9, + "valid_targets_min": 4614 + }, + { + "epoch": 3.108957569408067, + "grad_norm": 0.3750097753627125, + "learning_rate": 2.723056457948718e-05, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09644874930381775, + "step": 5935, + "valid_targets_mean": 6779.1, + "valid_targets_min": 5050 + }, + { + "epoch": 3.111576741749607, + "grad_norm": 0.3757860797823471, + "learning_rate": 2.7206202032463963e-05, + "loss": 0.1679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08521732687950134, + "step": 5940, + "valid_targets_mean": 5993.4, + "valid_targets_min": 4734 + }, + { + "epoch": 3.1141959140911473, + "grad_norm": 0.34855395999237143, + "learning_rate": 2.7181827191142988e-05, + "loss": 0.1665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08455619215965271, + "step": 5945, + "valid_targets_mean": 6826.9, + "valid_targets_min": 4966 + }, + { + "epoch": 3.116815086432687, + "grad_norm": 0.3629306245827195, + "learning_rate": 2.7157440097109496e-05, + "loss": 0.1865, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07667134702205658, + "step": 5950, + "valid_targets_mean": 6366.5, + "valid_targets_min": 4510 + }, + { + "epoch": 3.1194342587742274, + "grad_norm": 0.3659755257584252, + "learning_rate": 2.7133040791969597e-05, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06763496994972229, + "step": 5955, + "valid_targets_mean": 5248.9, + "valid_targets_min": 4430 + }, + { + "epoch": 3.1220534311157673, + "grad_norm": 0.3664115118233695, + "learning_rate": 2.7108629317350278e-05, + "loss": 0.1496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0714559257030487, + "step": 5960, + "valid_targets_mean": 5854.6, + "valid_targets_min": 4701 + }, + { + "epoch": 3.1246726034573076, + "grad_norm": 0.3688029594300618, + "learning_rate": 2.7084205714899247e-05, + "loss": 0.1571, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0836070328950882, + "step": 5965, + "valid_targets_mean": 6520.2, + "valid_targets_min": 4501 + }, + { + "epoch": 3.1272917757988474, + "grad_norm": 0.35531086640459225, + "learning_rate": 2.705977002628494e-05, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08988002687692642, + "step": 5970, + "valid_targets_mean": 6796.4, + "valid_targets_min": 3640 + }, + { + "epoch": 3.1299109481403877, + "grad_norm": 0.36621127475427373, + "learning_rate": 2.703532229319638e-05, + "loss": 0.1803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08791564404964447, + "step": 5975, + "valid_targets_mean": 6024.8, + "valid_targets_min": 5024 + }, + { + "epoch": 3.1325301204819276, + "grad_norm": 0.3963512060400468, + "learning_rate": 2.7010862557343168e-05, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09890920668840408, + "step": 5980, + "valid_targets_mean": 6549.4, + "valid_targets_min": 4715 + }, + { + "epoch": 3.135149292823468, + "grad_norm": 0.40434474990319935, + "learning_rate": 2.698639086045536e-05, + "loss": 0.1751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07355411350727081, + "step": 5985, + "valid_targets_mean": 5681.1, + "valid_targets_min": 4730 + }, + { + "epoch": 3.1377684651650077, + "grad_norm": 0.39609893448203354, + "learning_rate": 2.6961907244283424e-05, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08784209191799164, + "step": 5990, + "valid_targets_mean": 5787.6, + "valid_targets_min": 4657 + }, + { + "epoch": 3.140387637506548, + "grad_norm": 0.37661510463536013, + "learning_rate": 2.6937411750598172e-05, + "loss": 0.1712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08997480571269989, + "step": 5995, + "valid_targets_mean": 6862.9, + "valid_targets_min": 5599 + }, + { + "epoch": 3.143006809848088, + "grad_norm": 0.3737978264602746, + "learning_rate": 2.6912904421190672e-05, + "loss": 0.1864, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1133907288312912, + "step": 6000, + "valid_targets_mean": 6471.2, + "valid_targets_min": 4948 + }, + { + "epoch": 3.145625982189628, + "grad_norm": 0.37312818209600035, + "learning_rate": 2.6888385297872194e-05, + "loss": 0.1832, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09566065669059753, + "step": 6005, + "valid_targets_mean": 7644.9, + "valid_targets_min": 4876 + }, + { + "epoch": 3.148245154531168, + "grad_norm": 0.35200702373864323, + "learning_rate": 2.6863854422474108e-05, + "loss": 0.1751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07421612739562988, + "step": 6010, + "valid_targets_mean": 5750.1, + "valid_targets_min": 3599 + }, + { + "epoch": 3.1508643268727083, + "grad_norm": 0.3825591974529176, + "learning_rate": 2.6839311836847856e-05, + "loss": 0.1825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10305662453174591, + "step": 6015, + "valid_targets_mean": 7335.2, + "valid_targets_min": 4800 + }, + { + "epoch": 3.153483499214248, + "grad_norm": 0.31396319928135724, + "learning_rate": 2.6814757582864847e-05, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06879167258739471, + "step": 6020, + "valid_targets_mean": 7088.9, + "valid_targets_min": 4228 + }, + { + "epoch": 3.1561026715557885, + "grad_norm": 0.3782189665231664, + "learning_rate": 2.6790191702416405e-05, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0891866683959961, + "step": 6025, + "valid_targets_mean": 6421.2, + "valid_targets_min": 5314 + }, + { + "epoch": 3.1587218438973284, + "grad_norm": 0.35636421350791364, + "learning_rate": 2.676561423741368e-05, + "loss": 0.2165, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08212357759475708, + "step": 6030, + "valid_targets_mean": 6345.4, + "valid_targets_min": 4634 + }, + { + "epoch": 3.1613410162388687, + "grad_norm": 0.3677988359573353, + "learning_rate": 2.6741025229787596e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07506930083036423, + "step": 6035, + "valid_targets_mean": 5758.2, + "valid_targets_min": 4521 + }, + { + "epoch": 3.1639601885804085, + "grad_norm": 0.407110438717293, + "learning_rate": 2.6716424721488752e-05, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08141658455133438, + "step": 6040, + "valid_targets_mean": 5346.1, + "valid_targets_min": 4180 + }, + { + "epoch": 3.166579360921949, + "grad_norm": 0.35246613213627176, + "learning_rate": 2.6691812754487395e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09361865371465683, + "step": 6045, + "valid_targets_mean": 7561.1, + "valid_targets_min": 4791 + }, + { + "epoch": 3.1691985332634887, + "grad_norm": 0.4463228623924752, + "learning_rate": 2.6667189370773302e-05, + "loss": 0.174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07970430701971054, + "step": 6050, + "valid_targets_mean": 6135.1, + "valid_targets_min": 4670 + }, + { + "epoch": 3.171817705605029, + "grad_norm": 0.44318857511406334, + "learning_rate": 2.664255461235573e-05, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0893382728099823, + "step": 6055, + "valid_targets_mean": 6526.8, + "valid_targets_min": 4806 + }, + { + "epoch": 3.174436877946569, + "grad_norm": 0.3953076988625492, + "learning_rate": 2.6617908521263344e-05, + "loss": 0.1663, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0709521621465683, + "step": 6060, + "valid_targets_mean": 5486.9, + "valid_targets_min": 4453 + }, + { + "epoch": 3.177056050288109, + "grad_norm": 0.38157949773531097, + "learning_rate": 2.659325113954415e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09218616783618927, + "step": 6065, + "valid_targets_mean": 7182.4, + "valid_targets_min": 5687 + }, + { + "epoch": 3.179675222629649, + "grad_norm": 0.40825030583785493, + "learning_rate": 2.6568582509265403e-05, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07747182995080948, + "step": 6070, + "valid_targets_mean": 6945.2, + "valid_targets_min": 4964 + }, + { + "epoch": 3.1822943949711893, + "grad_norm": 0.34753108904440033, + "learning_rate": 2.6543902672513562e-05, + "loss": 0.1676, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0850880891084671, + "step": 6075, + "valid_targets_mean": 7281.2, + "valid_targets_min": 5625 + }, + { + "epoch": 3.184913567312729, + "grad_norm": 0.3482754984997392, + "learning_rate": 2.6519211671394196e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07551047950983047, + "step": 6080, + "valid_targets_mean": 6706.9, + "valid_targets_min": 5006 + }, + { + "epoch": 3.1875327396542694, + "grad_norm": 0.392218516349976, + "learning_rate": 2.6494509548031926e-05, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0884195938706398, + "step": 6085, + "valid_targets_mean": 6947.8, + "valid_targets_min": 5252 + }, + { + "epoch": 3.1901519119958093, + "grad_norm": 0.35417064310780705, + "learning_rate": 2.6469796344570348e-05, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08547120541334152, + "step": 6090, + "valid_targets_mean": 6373.1, + "valid_targets_min": 5071 + }, + { + "epoch": 3.1927710843373496, + "grad_norm": 0.3841436072670584, + "learning_rate": 2.6445072103171952e-05, + "loss": 0.1737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09343600273132324, + "step": 6095, + "valid_targets_mean": 6664.0, + "valid_targets_min": 3440 + }, + { + "epoch": 3.1953902566788894, + "grad_norm": 0.45559483827697905, + "learning_rate": 2.6420336866018087e-05, + "loss": 0.178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06918194890022278, + "step": 6100, + "valid_targets_mean": 4241.2, + "valid_targets_min": 791 + }, + { + "epoch": 3.1980094290204297, + "grad_norm": 0.725399793351473, + "learning_rate": 2.6395590675308826e-05, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08077594637870789, + "step": 6105, + "valid_targets_mean": 1367.1, + "valid_targets_min": 817 + }, + { + "epoch": 3.2006286013619696, + "grad_norm": 0.6940465636376866, + "learning_rate": 2.6370833573262946e-05, + "loss": 0.1734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09896910190582275, + "step": 6110, + "valid_targets_mean": 1758.2, + "valid_targets_min": 661 + }, + { + "epoch": 3.20324777370351, + "grad_norm": 0.7711853117394459, + "learning_rate": 2.634606560211785e-05, + "loss": 0.1709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08756490796804428, + "step": 6115, + "valid_targets_mean": 1521.6, + "valid_targets_min": 826 + }, + { + "epoch": 3.2058669460450497, + "grad_norm": 0.77831360128479, + "learning_rate": 2.6321286804129465e-05, + "loss": 0.1651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08230245113372803, + "step": 6120, + "valid_targets_mean": 1366.8, + "valid_targets_min": 742 + }, + { + "epoch": 3.20848611838659, + "grad_norm": 0.7320861316054422, + "learning_rate": 2.62964972215722e-05, + "loss": 0.1765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05615902692079544, + "step": 6125, + "valid_targets_mean": 1165.5, + "valid_targets_min": 718 + }, + { + "epoch": 3.21110529072813, + "grad_norm": 0.6613785362056553, + "learning_rate": 2.6271696896738874e-05, + "loss": 0.1734, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0980529934167862, + "step": 6130, + "valid_targets_mean": 1766.2, + "valid_targets_min": 847 + }, + { + "epoch": 3.21372446306967, + "grad_norm": 0.7866352650348061, + "learning_rate": 2.62468858719406e-05, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08308720588684082, + "step": 6135, + "valid_targets_mean": 1583.8, + "valid_targets_min": 814 + }, + { + "epoch": 3.21634363541121, + "grad_norm": 0.729383476475166, + "learning_rate": 2.6222064189506782e-05, + "loss": 0.1792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07057338953018188, + "step": 6140, + "valid_targets_mean": 1134.5, + "valid_targets_min": 610 + }, + { + "epoch": 3.2189628077527503, + "grad_norm": 0.6927157851342404, + "learning_rate": 2.619723189178498e-05, + "loss": 0.1639, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.093279629945755, + "step": 6145, + "valid_targets_mean": 1884.8, + "valid_targets_min": 819 + }, + { + "epoch": 3.22158198009429, + "grad_norm": 0.6390723228944544, + "learning_rate": 2.6172389021140886e-05, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08048640936613083, + "step": 6150, + "valid_targets_mean": 1438.5, + "valid_targets_min": 615 + }, + { + "epoch": 3.22420115243583, + "grad_norm": 0.7797999459844217, + "learning_rate": 2.6147535619958216e-05, + "loss": 0.1697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12054004520177841, + "step": 6155, + "valid_targets_mean": 1667.1, + "valid_targets_min": 748 + }, + { + "epoch": 3.2268203247773704, + "grad_norm": 0.6875139396294999, + "learning_rate": 2.6122671730638657e-05, + "loss": 0.1692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07025996595621109, + "step": 6160, + "valid_targets_mean": 1348.2, + "valid_targets_min": 704 + }, + { + "epoch": 3.2294394971189107, + "grad_norm": 0.6696933320606057, + "learning_rate": 2.6097797395601782e-05, + "loss": 0.1577, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06189568340778351, + "step": 6165, + "valid_targets_mean": 1212.4, + "valid_targets_min": 722 + }, + { + "epoch": 3.2320586694604505, + "grad_norm": 0.702567995222485, + "learning_rate": 2.6072912657285002e-05, + "loss": 0.1627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08648046851158142, + "step": 6170, + "valid_targets_mean": 1444.4, + "valid_targets_min": 642 + }, + { + "epoch": 3.2346778418019904, + "grad_norm": 0.8104376969768211, + "learning_rate": 2.6048017558143462e-05, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07088565081357956, + "step": 6175, + "valid_targets_mean": 1289.6, + "valid_targets_min": 518 + }, + { + "epoch": 3.2372970141435307, + "grad_norm": 0.7255023410193053, + "learning_rate": 2.602311214064998e-05, + "loss": 0.1753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11163045465946198, + "step": 6180, + "valid_targets_mean": 1796.4, + "valid_targets_min": 728 + }, + { + "epoch": 3.2399161864850705, + "grad_norm": 0.6781031559393812, + "learning_rate": 2.5998196447294996e-05, + "loss": 0.1708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10201551020145416, + "step": 6185, + "valid_targets_mean": 1814.6, + "valid_targets_min": 820 + }, + { + "epoch": 3.242535358826611, + "grad_norm": 0.7286450583637957, + "learning_rate": 2.5973270520586465e-05, + "loss": 0.1724, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08340763300657272, + "step": 6190, + "valid_targets_mean": 1437.2, + "valid_targets_min": 818 + }, + { + "epoch": 3.2451545311681507, + "grad_norm": 0.7492141574566015, + "learning_rate": 2.5948334403049806e-05, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08512294292449951, + "step": 6195, + "valid_targets_mean": 1477.9, + "valid_targets_min": 625 + }, + { + "epoch": 3.247773703509691, + "grad_norm": 0.636264410480963, + "learning_rate": 2.5923388137227825e-05, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09342190623283386, + "step": 6200, + "valid_targets_mean": 2156.2, + "valid_targets_min": 790 + }, + { + "epoch": 3.250392875851231, + "grad_norm": 0.7198652717308373, + "learning_rate": 2.5898431765680644e-05, + "loss": 0.1711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09737849235534668, + "step": 6205, + "valid_targets_mean": 2027.8, + "valid_targets_min": 1043 + }, + { + "epoch": 3.253012048192771, + "grad_norm": 0.6630084494974814, + "learning_rate": 2.5873465330985625e-05, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09584726393222809, + "step": 6210, + "valid_targets_mean": 1829.5, + "valid_targets_min": 719 + }, + { + "epoch": 3.255631220534311, + "grad_norm": 0.7102054735958612, + "learning_rate": 2.5848488875737293e-05, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0718170702457428, + "step": 6215, + "valid_targets_mean": 1326.5, + "valid_targets_min": 697 + }, + { + "epoch": 3.2582503928758513, + "grad_norm": 0.7429130821750529, + "learning_rate": 2.582350244254728e-05, + "loss": 0.1603, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08813682943582535, + "step": 6220, + "valid_targets_mean": 1462.4, + "valid_targets_min": 824 + }, + { + "epoch": 3.260869565217391, + "grad_norm": 0.6941738318759202, + "learning_rate": 2.5798506074044225e-05, + "loss": 0.162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08306953310966492, + "step": 6225, + "valid_targets_mean": 1548.0, + "valid_targets_min": 629 + }, + { + "epoch": 3.2634887375589314, + "grad_norm": 0.7229424119535742, + "learning_rate": 2.577349981287373e-05, + "loss": 0.1624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08312453329563141, + "step": 6230, + "valid_targets_mean": 1610.4, + "valid_targets_min": 1033 + }, + { + "epoch": 3.2661079099004713, + "grad_norm": 0.6695282519425605, + "learning_rate": 2.5748483701698268e-05, + "loss": 0.1645, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08673971146345139, + "step": 6235, + "valid_targets_mean": 1837.6, + "valid_targets_min": 1075 + }, + { + "epoch": 3.2687270822420116, + "grad_norm": 0.699408572371278, + "learning_rate": 2.5723457783197123e-05, + "loss": 0.1653, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07823971658945084, + "step": 6240, + "valid_targets_mean": 1572.5, + "valid_targets_min": 816 + }, + { + "epoch": 3.2713462545835514, + "grad_norm": 0.7279345678997962, + "learning_rate": 2.569842210006631e-05, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0817808210849762, + "step": 6245, + "valid_targets_mean": 1510.4, + "valid_targets_min": 957 + }, + { + "epoch": 3.2739654269250917, + "grad_norm": 0.7034516442678546, + "learning_rate": 2.56733766950185e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05076921731233597, + "step": 6250, + "valid_targets_mean": 1002.4, + "valid_targets_min": 643 + }, + { + "epoch": 3.2765845992666316, + "grad_norm": 0.9957676803575626, + "learning_rate": 2.5648321610782943e-05, + "loss": 0.1692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08878156542778015, + "step": 6255, + "valid_targets_mean": 1574.1, + "valid_targets_min": 1111 + }, + { + "epoch": 3.279203771608172, + "grad_norm": 0.6465046786343321, + "learning_rate": 2.5623256890105422e-05, + "loss": 0.1536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07464969903230667, + "step": 6260, + "valid_targets_mean": 1644.5, + "valid_targets_min": 705 + }, + { + "epoch": 3.2818229439497117, + "grad_norm": 0.816195525548091, + "learning_rate": 2.559818257574814e-05, + "loss": 0.1602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08383892476558685, + "step": 6265, + "valid_targets_mean": 1555.6, + "valid_targets_min": 716 + }, + { + "epoch": 3.284442116291252, + "grad_norm": 0.721627622041745, + "learning_rate": 2.5573098710489683e-05, + "loss": 0.1502, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0769277960062027, + "step": 6270, + "valid_targets_mean": 1269.2, + "valid_targets_min": 627 + }, + { + "epoch": 3.287061288632792, + "grad_norm": 0.6997653881354113, + "learning_rate": 2.5548005337124924e-05, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07156737893819809, + "step": 6275, + "valid_targets_mean": 1424.5, + "valid_targets_min": 817 + }, + { + "epoch": 3.289680460974332, + "grad_norm": 0.690853398746686, + "learning_rate": 2.552290249846496e-05, + "loss": 0.1581, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07242089509963989, + "step": 6280, + "valid_targets_mean": 1382.5, + "valid_targets_min": 974 + }, + { + "epoch": 3.292299633315872, + "grad_norm": 0.729012052110836, + "learning_rate": 2.5497790237337024e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07587244361639023, + "step": 6285, + "valid_targets_mean": 1400.1, + "valid_targets_min": 700 + }, + { + "epoch": 3.2949188056574124, + "grad_norm": 0.7801429955446634, + "learning_rate": 2.5472668596584448e-05, + "loss": 0.1614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09984627366065979, + "step": 6290, + "valid_targets_mean": 1836.5, + "valid_targets_min": 880 + }, + { + "epoch": 3.297537977998952, + "grad_norm": 0.743176432178358, + "learning_rate": 2.5447537619066543e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08273142576217651, + "step": 6295, + "valid_targets_mean": 1480.4, + "valid_targets_min": 903 + }, + { + "epoch": 3.3001571503404925, + "grad_norm": 0.7132081297055632, + "learning_rate": 2.5422397347658575e-05, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07163409888744354, + "step": 6300, + "valid_targets_mean": 1437.9, + "valid_targets_min": 682 + }, + { + "epoch": 3.3027763226820324, + "grad_norm": 0.7677933454268899, + "learning_rate": 2.539724782525164e-05, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08004729449748993, + "step": 6305, + "valid_targets_mean": 1251.8, + "valid_targets_min": 703 + }, + { + "epoch": 3.3053954950235727, + "grad_norm": 0.7943407165697173, + "learning_rate": 2.537208909475263e-05, + "loss": 0.1679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0743921548128128, + "step": 6310, + "valid_targets_mean": 1405.6, + "valid_targets_min": 934 + }, + { + "epoch": 3.3080146673651125, + "grad_norm": 0.7389762940947008, + "learning_rate": 2.534692119908414e-05, + "loss": 0.1726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12644554674625397, + "step": 6315, + "valid_targets_mean": 1885.8, + "valid_targets_min": 1122 + }, + { + "epoch": 3.310633839706653, + "grad_norm": 0.6723448699275575, + "learning_rate": 2.532174418118442e-05, + "loss": 0.157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08895815163850784, + "step": 6320, + "valid_targets_mean": 1871.0, + "valid_targets_min": 882 + }, + { + "epoch": 3.3132530120481927, + "grad_norm": 0.7159734610386331, + "learning_rate": 2.5296558084007253e-05, + "loss": 0.1563, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06577227264642715, + "step": 6325, + "valid_targets_mean": 1421.4, + "valid_targets_min": 551 + }, + { + "epoch": 3.315872184389733, + "grad_norm": 0.7475482122922238, + "learning_rate": 2.5271362950521935e-05, + "loss": 0.1629, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08857665210962296, + "step": 6330, + "valid_targets_mean": 1634.0, + "valid_targets_min": 791 + }, + { + "epoch": 3.318491356731273, + "grad_norm": 0.6802764224126414, + "learning_rate": 2.5246158823713176e-05, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0757802426815033, + "step": 6335, + "valid_targets_mean": 1377.0, + "valid_targets_min": 904 + }, + { + "epoch": 3.321110529072813, + "grad_norm": 0.6860274792428837, + "learning_rate": 2.5220945746581014e-05, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08174487948417664, + "step": 6340, + "valid_targets_mean": 1642.8, + "valid_targets_min": 859 + }, + { + "epoch": 3.323729701414353, + "grad_norm": 0.7237389938468328, + "learning_rate": 2.5195723762140777e-05, + "loss": 0.149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0694696381688118, + "step": 6345, + "valid_targets_mean": 1274.0, + "valid_targets_min": 756 + }, + { + "epoch": 3.3263488737558933, + "grad_norm": 0.787128259336043, + "learning_rate": 2.5170492913422973e-05, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08977150917053223, + "step": 6350, + "valid_targets_mean": 1453.8, + "valid_targets_min": 836 + }, + { + "epoch": 3.328968046097433, + "grad_norm": 0.8018754334520489, + "learning_rate": 2.514525324347325e-05, + "loss": 0.1621, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0953463762998581, + "step": 6355, + "valid_targets_mean": 1690.9, + "valid_targets_min": 1462 + }, + { + "epoch": 3.3315872184389734, + "grad_norm": 0.6697156130227047, + "learning_rate": 2.5120004795352287e-05, + "loss": 0.1549, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0697600394487381, + "step": 6360, + "valid_targets_mean": 1457.9, + "valid_targets_min": 967 + }, + { + "epoch": 3.3342063907805133, + "grad_norm": 0.660076956571282, + "learning_rate": 2.509474761213576e-05, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07847227901220322, + "step": 6365, + "valid_targets_mean": 1588.8, + "valid_targets_min": 1227 + }, + { + "epoch": 3.3368255631220536, + "grad_norm": 0.7022754027795831, + "learning_rate": 2.5069481736914218e-05, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06762699782848358, + "step": 6370, + "valid_targets_mean": 1243.0, + "valid_targets_min": 740 + }, + { + "epoch": 3.3394447354635934, + "grad_norm": 0.692283883843287, + "learning_rate": 2.5044207212793073e-05, + "loss": 0.1593, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09675326943397522, + "step": 6375, + "valid_targets_mean": 2075.1, + "valid_targets_min": 881 + }, + { + "epoch": 3.3420639078051337, + "grad_norm": 0.7596393501203929, + "learning_rate": 2.5018924082892465e-05, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08390866219997406, + "step": 6380, + "valid_targets_mean": 1332.2, + "valid_targets_min": 636 + }, + { + "epoch": 3.3446830801466736, + "grad_norm": 0.9804017283381228, + "learning_rate": 2.4993632390347234e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06037019193172455, + "step": 6385, + "valid_targets_mean": 1061.9, + "valid_targets_min": 814 + }, + { + "epoch": 3.347302252488214, + "grad_norm": 0.7397752344314142, + "learning_rate": 2.4968332178306826e-05, + "loss": 0.1478, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07326038926839828, + "step": 6390, + "valid_targets_mean": 1354.0, + "valid_targets_min": 820 + }, + { + "epoch": 3.3499214248297537, + "grad_norm": 0.7562290377336969, + "learning_rate": 2.4943023489935215e-05, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07187385857105255, + "step": 6395, + "valid_targets_mean": 1292.9, + "valid_targets_min": 1008 + }, + { + "epoch": 3.352540597171294, + "grad_norm": 0.6971071706540642, + "learning_rate": 2.4917706368410838e-05, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06977003067731857, + "step": 6400, + "valid_targets_mean": 1513.5, + "valid_targets_min": 728 + }, + { + "epoch": 3.355159769512834, + "grad_norm": 0.8194784102608055, + "learning_rate": 2.4892380856926517e-05, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07805469632148743, + "step": 6405, + "valid_targets_mean": 1389.9, + "valid_targets_min": 692 + }, + { + "epoch": 3.357778941854374, + "grad_norm": 0.763931804191717, + "learning_rate": 2.48670469986894e-05, + "loss": 0.1578, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08736824989318848, + "step": 6410, + "valid_targets_mean": 1351.2, + "valid_targets_min": 840 + }, + { + "epoch": 3.360398114195914, + "grad_norm": 0.7115013265253044, + "learning_rate": 2.4841704836920862e-05, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07284218817949295, + "step": 6415, + "valid_targets_mean": 1458.2, + "valid_targets_min": 950 + }, + { + "epoch": 3.3630172865374544, + "grad_norm": 0.6857982783305582, + "learning_rate": 2.481635441485645e-05, + "loss": 0.1548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07983918488025665, + "step": 6420, + "valid_targets_mean": 1749.0, + "valid_targets_min": 802 + }, + { + "epoch": 3.365636458878994, + "grad_norm": 0.7598799035380043, + "learning_rate": 2.479099577574581e-05, + "loss": 0.1583, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07860361039638519, + "step": 6425, + "valid_targets_mean": 1506.4, + "valid_targets_min": 750 + }, + { + "epoch": 3.368255631220534, + "grad_norm": 0.8084854794038734, + "learning_rate": 2.4765628962852586e-05, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0805310606956482, + "step": 6430, + "valid_targets_mean": 1637.9, + "valid_targets_min": 1013 + }, + { + "epoch": 3.3708748035620744, + "grad_norm": 0.7221848851134685, + "learning_rate": 2.474025401945439e-05, + "loss": 0.1595, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06069980189204216, + "step": 6435, + "valid_targets_mean": 1228.9, + "valid_targets_min": 748 + }, + { + "epoch": 3.3734939759036147, + "grad_norm": 0.7345477184476626, + "learning_rate": 2.4714870988842696e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07106143236160278, + "step": 6440, + "valid_targets_mean": 1397.4, + "valid_targets_min": 729 + }, + { + "epoch": 3.3761131482451545, + "grad_norm": 0.700447897736445, + "learning_rate": 2.4689479914322777e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07165275514125824, + "step": 6445, + "valid_targets_mean": 1639.9, + "valid_targets_min": 1264 + }, + { + "epoch": 3.3787323205866944, + "grad_norm": 0.6741223108167599, + "learning_rate": 2.466408083921363e-05, + "loss": 0.1596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06849883496761322, + "step": 6450, + "valid_targets_mean": 1377.9, + "valid_targets_min": 916 + }, + { + "epoch": 3.3813514929282347, + "grad_norm": 0.7201663680310889, + "learning_rate": 2.4638673806847883e-05, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07745347917079926, + "step": 6455, + "valid_targets_mean": 1434.2, + "valid_targets_min": 748 + }, + { + "epoch": 3.383970665269775, + "grad_norm": 0.7236053807817885, + "learning_rate": 2.4613258860571782e-05, + "loss": 0.1569, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08043740689754486, + "step": 6460, + "valid_targets_mean": 1448.2, + "valid_targets_min": 916 + }, + { + "epoch": 3.386589837611315, + "grad_norm": 0.7230529608547769, + "learning_rate": 2.4587836043745027e-05, + "loss": 0.1521, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07702300697565079, + "step": 6465, + "valid_targets_mean": 1561.6, + "valid_targets_min": 848 + }, + { + "epoch": 3.3892090099528547, + "grad_norm": 0.8070482307933063, + "learning_rate": 2.4562405399740778e-05, + "loss": 0.1634, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07397100329399109, + "step": 6470, + "valid_targets_mean": 1103.4, + "valid_targets_min": 687 + }, + { + "epoch": 3.391828182294395, + "grad_norm": 0.7273087734194071, + "learning_rate": 2.4536966971945536e-05, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08558210730552673, + "step": 6475, + "valid_targets_mean": 1570.6, + "valid_targets_min": 1022 + }, + { + "epoch": 3.3944473546359353, + "grad_norm": 8.94249684292272, + "learning_rate": 2.4511520803759083e-05, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06441488862037659, + "step": 6480, + "valid_targets_mean": 1260.6, + "valid_targets_min": 1013 + }, + { + "epoch": 3.397066526977475, + "grad_norm": 0.7033247150238093, + "learning_rate": 2.448606693859442e-05, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07360631227493286, + "step": 6485, + "valid_targets_mean": 1469.6, + "valid_targets_min": 824 + }, + { + "epoch": 3.399685699319015, + "grad_norm": 0.7863362439057723, + "learning_rate": 2.4460605419877652e-05, + "loss": 0.1669, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08512599766254425, + "step": 6490, + "valid_targets_mean": 1283.8, + "valid_targets_min": 548 + }, + { + "epoch": 3.4023048716605553, + "grad_norm": 0.7051970588155935, + "learning_rate": 2.443513629104796e-05, + "loss": 0.1501, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07364164292812347, + "step": 6495, + "valid_targets_mean": 1326.2, + "valid_targets_min": 694 + }, + { + "epoch": 3.4049240440020956, + "grad_norm": 0.6511836504853266, + "learning_rate": 2.4409659595557507e-05, + "loss": 0.1575, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06439782679080963, + "step": 6500, + "valid_targets_mean": 1619.8, + "valid_targets_min": 1030 + }, + { + "epoch": 3.4075432163436354, + "grad_norm": 0.6271923103255855, + "learning_rate": 2.4384175376871366e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06951231509447098, + "step": 6505, + "valid_targets_mean": 1632.9, + "valid_targets_min": 745 + }, + { + "epoch": 3.4101623886851753, + "grad_norm": 0.7840905730868706, + "learning_rate": 2.4358683678467442e-05, + "loss": 0.1505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08921942114830017, + "step": 6510, + "valid_targets_mean": 1553.2, + "valid_targets_min": 990 + }, + { + "epoch": 3.4127815610267156, + "grad_norm": 0.6928157066182933, + "learning_rate": 2.43331845438364e-05, + "loss": 0.1572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050745025277137756, + "step": 6515, + "valid_targets_mean": 1129.8, + "valid_targets_min": 639 + }, + { + "epoch": 3.4154007333682554, + "grad_norm": 0.6981470631251446, + "learning_rate": 2.4307678016481595e-05, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07331757247447968, + "step": 6520, + "valid_targets_mean": 1595.0, + "valid_targets_min": 721 + }, + { + "epoch": 3.4180199057097957, + "grad_norm": 0.72419476507239, + "learning_rate": 2.428216413991899e-05, + "loss": 0.1542, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0771220326423645, + "step": 6525, + "valid_targets_mean": 1401.1, + "valid_targets_min": 874 + }, + { + "epoch": 3.4206390780513356, + "grad_norm": 0.7183118403348777, + "learning_rate": 2.4256642957677096e-05, + "loss": 0.156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07753396034240723, + "step": 6530, + "valid_targets_mean": 1617.1, + "valid_targets_min": 686 + }, + { + "epoch": 3.423258250392876, + "grad_norm": 0.7093166080435358, + "learning_rate": 2.4231114513296868e-05, + "loss": 0.1605, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08326885104179382, + "step": 6535, + "valid_targets_mean": 1559.4, + "valid_targets_min": 713 + }, + { + "epoch": 3.4258774227344158, + "grad_norm": 0.8546442685115246, + "learning_rate": 2.420557885033168e-05, + "loss": 0.175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18318283557891846, + "step": 6540, + "valid_targets_mean": 3018.8, + "valid_targets_min": 920 + }, + { + "epoch": 3.428496595075956, + "grad_norm": 0.5412621726003334, + "learning_rate": 2.4180036012347197e-05, + "loss": 0.2674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08558709919452667, + "step": 6545, + "valid_targets_mean": 3768.8, + "valid_targets_min": 746 + }, + { + "epoch": 3.431115767417496, + "grad_norm": 0.47493460451138947, + "learning_rate": 2.415448604292133e-05, + "loss": 0.2496, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12001121789216995, + "step": 6550, + "valid_targets_mean": 5086.6, + "valid_targets_min": 1846 + }, + { + "epoch": 3.433734939759036, + "grad_norm": 0.4489361583361423, + "learning_rate": 2.4128928985644165e-05, + "loss": 0.2392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09062111377716064, + "step": 6555, + "valid_targets_mean": 3826.8, + "valid_targets_min": 1321 + }, + { + "epoch": 3.436354112100576, + "grad_norm": 0.44374587608332783, + "learning_rate": 2.4103364884117878e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12933379411697388, + "step": 6560, + "valid_targets_mean": 6475.0, + "valid_targets_min": 3049 + }, + { + "epoch": 3.4389732844421164, + "grad_norm": 0.5337228874956089, + "learning_rate": 2.4077793781956652e-05, + "loss": 0.2534, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.127142533659935, + "step": 6565, + "valid_targets_mean": 4328.6, + "valid_targets_min": 2075 + }, + { + "epoch": 3.441592456783656, + "grad_norm": 0.4643121804132278, + "learning_rate": 2.4052215722786635e-05, + "loss": 0.2466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12487849593162537, + "step": 6570, + "valid_targets_mean": 5135.4, + "valid_targets_min": 1991 + }, + { + "epoch": 3.4442116291251965, + "grad_norm": 0.5357770445376597, + "learning_rate": 2.4026630750245823e-05, + "loss": 0.2369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1093788743019104, + "step": 6575, + "valid_targets_mean": 3699.0, + "valid_targets_min": 1665 + }, + { + "epoch": 3.4468308014667364, + "grad_norm": 0.48048205150664014, + "learning_rate": 2.4001038907984015e-05, + "loss": 0.2475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1195569708943367, + "step": 6580, + "valid_targets_mean": 4666.8, + "valid_targets_min": 578 + }, + { + "epoch": 3.4494499738082767, + "grad_norm": 0.4636560937257869, + "learning_rate": 2.3975440239662724e-05, + "loss": 0.2519, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11389769613742828, + "step": 6585, + "valid_targets_mean": 4916.1, + "valid_targets_min": 1241 + }, + { + "epoch": 3.4520691461498165, + "grad_norm": 0.5081239454775448, + "learning_rate": 2.3949834788955136e-05, + "loss": 0.2456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1260935515165329, + "step": 6590, + "valid_targets_mean": 4525.4, + "valid_targets_min": 2726 + }, + { + "epoch": 3.454688318491357, + "grad_norm": 0.4237746910615088, + "learning_rate": 2.3924222599545967e-05, + "loss": 0.242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11075103282928467, + "step": 6595, + "valid_targets_mean": 6055.0, + "valid_targets_min": 2639 + }, + { + "epoch": 3.4573074908328967, + "grad_norm": 0.4972113562704315, + "learning_rate": 2.3898603715131455e-05, + "loss": 0.2437, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1271575391292572, + "step": 6600, + "valid_targets_mean": 5447.0, + "valid_targets_min": 2244 + }, + { + "epoch": 3.459926663174437, + "grad_norm": 0.46080297058500136, + "learning_rate": 2.387297817941926e-05, + "loss": 0.2461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1249583438038826, + "step": 6605, + "valid_targets_mean": 5634.1, + "valid_targets_min": 3076 + }, + { + "epoch": 3.462545835515977, + "grad_norm": 0.5524691591959915, + "learning_rate": 2.384734603612838e-05, + "loss": 0.242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.14067986607551575, + "step": 6610, + "valid_targets_mean": 4215.6, + "valid_targets_min": 1721 + }, + { + "epoch": 3.465165007857517, + "grad_norm": 0.5753055954226567, + "learning_rate": 2.382170732898909e-05, + "loss": 0.2481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11744669824838638, + "step": 6615, + "valid_targets_mean": 3667.0, + "valid_targets_min": 1705 + }, + { + "epoch": 3.467784180199057, + "grad_norm": 0.5620312218064215, + "learning_rate": 2.3796062101742864e-05, + "loss": 0.232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10161575675010681, + "step": 6620, + "valid_targets_mean": 3136.5, + "valid_targets_min": 1538 + }, + { + "epoch": 3.4704033525405973, + "grad_norm": 0.5839952431744498, + "learning_rate": 2.3770410398142307e-05, + "loss": 0.2422, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09144783020019531, + "step": 6625, + "valid_targets_mean": 3045.5, + "valid_targets_min": 1451 + }, + { + "epoch": 3.473022524882137, + "grad_norm": 0.5722177721005322, + "learning_rate": 2.374475226195106e-05, + "loss": 0.2538, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13419339060783386, + "step": 6630, + "valid_targets_mean": 4635.6, + "valid_targets_min": 2560 + }, + { + "epoch": 3.4756416972236774, + "grad_norm": 0.5731540854960034, + "learning_rate": 2.371908773694375e-05, + "loss": 0.2511, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11543504893779755, + "step": 6635, + "valid_targets_mean": 3925.4, + "valid_targets_min": 1612 + }, + { + "epoch": 3.4782608695652173, + "grad_norm": 0.5605624941760808, + "learning_rate": 2.3693416866905885e-05, + "loss": 0.2462, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13869768381118774, + "step": 6640, + "valid_targets_mean": 4462.9, + "valid_targets_min": 1921 + }, + { + "epoch": 3.4808800419067576, + "grad_norm": 0.5904918095113585, + "learning_rate": 2.3667739695633825e-05, + "loss": 0.2328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11713632941246033, + "step": 6645, + "valid_targets_mean": 4697.6, + "valid_targets_min": 2004 + }, + { + "epoch": 3.4834992142482974, + "grad_norm": 0.5654588666799971, + "learning_rate": 2.364205626693466e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09012836217880249, + "step": 6650, + "valid_targets_mean": 3302.0, + "valid_targets_min": 1896 + }, + { + "epoch": 3.4861183865898377, + "grad_norm": 0.6103347647365203, + "learning_rate": 2.3616366624626165e-05, + "loss": 0.2222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12937378883361816, + "step": 6655, + "valid_targets_mean": 3412.4, + "valid_targets_min": 1846 + }, + { + "epoch": 3.4887375589313776, + "grad_norm": 0.5749052069572483, + "learning_rate": 2.3590670812536713e-05, + "loss": 0.2469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.15378713607788086, + "step": 6660, + "valid_targets_mean": 4289.0, + "valid_targets_min": 3360 + }, + { + "epoch": 3.491356731272918, + "grad_norm": 0.5565439811657013, + "learning_rate": 2.3564968874505202e-05, + "loss": 0.2367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11687811464071274, + "step": 6665, + "valid_targets_mean": 3717.1, + "valid_targets_min": 1475 + }, + { + "epoch": 3.4939759036144578, + "grad_norm": 0.5532024019876822, + "learning_rate": 2.3539260854380982e-05, + "loss": 0.2404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1277935802936554, + "step": 6670, + "valid_targets_mean": 4400.9, + "valid_targets_min": 2229 + }, + { + "epoch": 3.496595075955998, + "grad_norm": 0.5791167537263643, + "learning_rate": 2.3513546796023783e-05, + "loss": 0.2384, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10298140347003937, + "step": 6675, + "valid_targets_mean": 3238.2, + "valid_targets_min": 1233 + }, + { + "epoch": 3.499214248297538, + "grad_norm": 0.5887424926472161, + "learning_rate": 2.348782674330363e-05, + "loss": 0.2251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10596165060997009, + "step": 6680, + "valid_targets_mean": 3109.8, + "valid_targets_min": 869 + }, + { + "epoch": 3.501833420639078, + "grad_norm": 0.5185673239521797, + "learning_rate": 2.346210074010078e-05, + "loss": 0.2385, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11567020416259766, + "step": 6685, + "valid_targets_mean": 4352.0, + "valid_targets_min": 1781 + }, + { + "epoch": 3.504452592980618, + "grad_norm": 0.6396724661953933, + "learning_rate": 2.343636883030564e-05, + "loss": 0.2376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13516461849212646, + "step": 6690, + "valid_targets_mean": 3687.0, + "valid_targets_min": 1108 + }, + { + "epoch": 3.5070717653221584, + "grad_norm": 0.6020937868220277, + "learning_rate": 2.3410631057818703e-05, + "loss": 0.2295, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11052301526069641, + "step": 6695, + "valid_targets_mean": 2814.9, + "valid_targets_min": 1913 + }, + { + "epoch": 3.509690937663698, + "grad_norm": 0.5854231638430263, + "learning_rate": 2.338488746655044e-05, + "loss": 0.2312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13174648582935333, + "step": 6700, + "valid_targets_mean": 4176.2, + "valid_targets_min": 1828 + }, + { + "epoch": 3.512310110005238, + "grad_norm": 0.5531224417414006, + "learning_rate": 2.3359138100421274e-05, + "loss": 0.2363, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10763726383447647, + "step": 6705, + "valid_targets_mean": 3962.1, + "valid_targets_min": 1589 + }, + { + "epoch": 3.5149292823467784, + "grad_norm": 0.5987313877267432, + "learning_rate": 2.333338300336147e-05, + "loss": 0.2276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11427710205316544, + "step": 6710, + "valid_targets_mean": 3307.2, + "valid_targets_min": 2289 + }, + { + "epoch": 3.5175484546883187, + "grad_norm": 0.5772416854249409, + "learning_rate": 2.3307622219311078e-05, + "loss": 0.234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08628536760807037, + "step": 6715, + "valid_targets_mean": 2650.5, + "valid_targets_min": 1248 + }, + { + "epoch": 3.5201676270298585, + "grad_norm": 0.5746127588561513, + "learning_rate": 2.3281855792219834e-05, + "loss": 0.2433, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10547653585672379, + "step": 6720, + "valid_targets_mean": 3181.0, + "valid_targets_min": 1727 + }, + { + "epoch": 3.5227867993713984, + "grad_norm": 0.5477655009619727, + "learning_rate": 2.3256083766047113e-05, + "loss": 0.2354, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12699449062347412, + "step": 6725, + "valid_targets_mean": 4625.5, + "valid_targets_min": 1965 + }, + { + "epoch": 3.5254059717129387, + "grad_norm": 0.5690976026566256, + "learning_rate": 2.3230306184761853e-05, + "loss": 0.2287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13122272491455078, + "step": 6730, + "valid_targets_mean": 3945.2, + "valid_targets_min": 2018 + }, + { + "epoch": 3.528025144054479, + "grad_norm": 0.5539172521944679, + "learning_rate": 2.3204523092342442e-05, + "loss": 0.2323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1102069765329361, + "step": 6735, + "valid_targets_mean": 4180.6, + "valid_targets_min": 1730 + }, + { + "epoch": 3.530644316396019, + "grad_norm": 0.5775422269004854, + "learning_rate": 2.3178734532776685e-05, + "loss": 0.2431, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11924617737531662, + "step": 6740, + "valid_targets_mean": 3331.2, + "valid_targets_min": 2097 + }, + { + "epoch": 3.5332634887375587, + "grad_norm": 0.6388277324268377, + "learning_rate": 2.3152940550061728e-05, + "loss": 0.2286, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13986080884933472, + "step": 6745, + "valid_targets_mean": 3802.9, + "valid_targets_min": 1648 + }, + { + "epoch": 3.535882661079099, + "grad_norm": 0.5637266652475762, + "learning_rate": 2.3127141188203948e-05, + "loss": 0.237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10528875887393951, + "step": 6750, + "valid_targets_mean": 3595.1, + "valid_targets_min": 1421 + }, + { + "epoch": 3.5385018334206393, + "grad_norm": 0.5751877425102137, + "learning_rate": 2.3101336491218916e-05, + "loss": 0.2334, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1265600472688675, + "step": 6755, + "valid_targets_mean": 3734.4, + "valid_targets_min": 2284 + }, + { + "epoch": 3.541121005762179, + "grad_norm": 0.5232442873757781, + "learning_rate": 2.3075526503131283e-05, + "loss": 0.2221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10531005263328552, + "step": 6760, + "valid_targets_mean": 4147.4, + "valid_targets_min": 2095 + }, + { + "epoch": 3.543740178103719, + "grad_norm": 0.5855439512538047, + "learning_rate": 2.304971126797475e-05, + "loss": 0.2324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09644465148448944, + "step": 6765, + "valid_targets_mean": 3100.2, + "valid_targets_min": 1673 + }, + { + "epoch": 3.5463593504452593, + "grad_norm": 0.5221365137695247, + "learning_rate": 2.302389082979196e-05, + "loss": 0.237, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12844055891036987, + "step": 6770, + "valid_targets_mean": 4951.2, + "valid_targets_min": 2067 + }, + { + "epoch": 3.5489785227867996, + "grad_norm": 0.5730074505577188, + "learning_rate": 2.2998065232634428e-05, + "loss": 0.2277, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0867147445678711, + "step": 6775, + "valid_targets_mean": 2375.2, + "valid_targets_min": 742 + }, + { + "epoch": 3.5515976951283394, + "grad_norm": 0.5609140272850285, + "learning_rate": 2.2972234520562485e-05, + "loss": 0.2416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13488733768463135, + "step": 6780, + "valid_targets_mean": 4786.8, + "valid_targets_min": 1766 + }, + { + "epoch": 3.5542168674698793, + "grad_norm": 0.5889544076046962, + "learning_rate": 2.2946398737645173e-05, + "loss": 0.23, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11075693368911743, + "step": 6785, + "valid_targets_mean": 3331.2, + "valid_targets_min": 1504 + }, + { + "epoch": 3.5568360398114196, + "grad_norm": 0.5429091381579527, + "learning_rate": 2.292055792796019e-05, + "loss": 0.2398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10074573755264282, + "step": 6790, + "valid_targets_mean": 3742.8, + "valid_targets_min": 1170 + }, + { + "epoch": 3.55945521215296, + "grad_norm": 0.5712724051794679, + "learning_rate": 2.2894712135593816e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11813041567802429, + "step": 6795, + "valid_targets_mean": 3477.5, + "valid_targets_min": 2669 + }, + { + "epoch": 3.5620743844944998, + "grad_norm": 0.5801978324207829, + "learning_rate": 2.2868861404640826e-05, + "loss": 0.2394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11976154893636703, + "step": 6800, + "valid_targets_mean": 3669.4, + "valid_targets_min": 2923 + }, + { + "epoch": 3.5646935568360396, + "grad_norm": 0.5341846370115644, + "learning_rate": 2.2843005779204424e-05, + "loss": 0.2283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09514598548412323, + "step": 6805, + "valid_targets_mean": 3643.6, + "valid_targets_min": 2756 + }, + { + "epoch": 3.56731272917758, + "grad_norm": 0.5405718842163569, + "learning_rate": 2.2817145303396155e-05, + "loss": 0.2219, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11385028809309006, + "step": 6810, + "valid_targets_mean": 4241.8, + "valid_targets_min": 2526 + }, + { + "epoch": 3.56993190151912, + "grad_norm": 0.5247567122855383, + "learning_rate": 2.279128002133586e-05, + "loss": 0.2168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1277405023574829, + "step": 6815, + "valid_targets_mean": 5317.1, + "valid_targets_min": 464 + }, + { + "epoch": 3.57255107386066, + "grad_norm": 0.5515397018372545, + "learning_rate": 2.276540997715155e-05, + "loss": 0.2427, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13250070810317993, + "step": 6820, + "valid_targets_mean": 4387.4, + "valid_targets_min": 1003 + }, + { + "epoch": 3.5751702462022, + "grad_norm": 0.5595826445666005, + "learning_rate": 2.273953521497938e-05, + "loss": 0.2309, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12329387664794922, + "step": 6825, + "valid_targets_mean": 3934.6, + "valid_targets_min": 1632 + }, + { + "epoch": 3.57778941854374, + "grad_norm": 0.6001471929850768, + "learning_rate": 2.2713655778963563e-05, + "loss": 0.2343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12080959975719452, + "step": 6830, + "valid_targets_mean": 3808.1, + "valid_targets_min": 1969 + }, + { + "epoch": 3.5804085908852805, + "grad_norm": 0.7086703407027434, + "learning_rate": 2.268777171325626e-05, + "loss": 0.2269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09347039461135864, + "step": 6835, + "valid_targets_mean": 3318.8, + "valid_targets_min": 1491 + }, + { + "epoch": 3.5830277632268204, + "grad_norm": 0.6047094919850572, + "learning_rate": 2.2661883062017558e-05, + "loss": 0.2447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1348264068365097, + "step": 6840, + "valid_targets_mean": 4169.9, + "valid_targets_min": 1637 + }, + { + "epoch": 3.5856469355683602, + "grad_norm": 0.5570782908965607, + "learning_rate": 2.263598986941534e-05, + "loss": 0.2323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10942605137825012, + "step": 6845, + "valid_targets_mean": 3640.1, + "valid_targets_min": 1818 + }, + { + "epoch": 3.5882661079099005, + "grad_norm": 0.6066054375904381, + "learning_rate": 2.261009217962526e-05, + "loss": 0.2284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12729772925376892, + "step": 6850, + "valid_targets_mean": 3803.0, + "valid_targets_min": 1523 + }, + { + "epoch": 3.5908852802514404, + "grad_norm": 0.5487739593002455, + "learning_rate": 2.2584190036830635e-05, + "loss": 0.2259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13107003271579742, + "step": 6855, + "valid_targets_mean": 4696.1, + "valid_targets_min": 1814 + }, + { + "epoch": 3.5935044525929807, + "grad_norm": 0.6187882265924753, + "learning_rate": 2.2558283485222367e-05, + "loss": 0.2206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12287470698356628, + "step": 6860, + "valid_targets_mean": 3022.9, + "valid_targets_min": 1073 + }, + { + "epoch": 3.5961236249345205, + "grad_norm": 0.5653764710318611, + "learning_rate": 2.2532372568998903e-05, + "loss": 0.221, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10427041351795197, + "step": 6865, + "valid_targets_mean": 3355.5, + "valid_targets_min": 1654 + }, + { + "epoch": 3.598742797276061, + "grad_norm": 0.5948764521192724, + "learning_rate": 2.2506457332366124e-05, + "loss": 0.2148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.115760937333107, + "step": 6870, + "valid_targets_mean": 3669.1, + "valid_targets_min": 1079 + }, + { + "epoch": 3.6013619696176007, + "grad_norm": 0.5582943951014767, + "learning_rate": 2.2480537819537293e-05, + "loss": 0.2132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0872621089220047, + "step": 6875, + "valid_targets_mean": 3073.5, + "valid_targets_min": 1143 + }, + { + "epoch": 3.603981141959141, + "grad_norm": 0.563262523476544, + "learning_rate": 2.2454614074732936e-05, + "loss": 0.2201, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09740658104419708, + "step": 6880, + "valid_targets_mean": 3300.2, + "valid_targets_min": 1853 + }, + { + "epoch": 3.606600314300681, + "grad_norm": 0.6641448530915551, + "learning_rate": 2.242868614218084e-05, + "loss": 0.251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11844535171985626, + "step": 6885, + "valid_targets_mean": 3122.8, + "valid_targets_min": 1410 + }, + { + "epoch": 3.609219486642221, + "grad_norm": 0.6447384569210771, + "learning_rate": 2.2402754066115905e-05, + "loss": 0.2224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09699396789073944, + "step": 6890, + "valid_targets_mean": 2564.8, + "valid_targets_min": 1566 + }, + { + "epoch": 3.611838658983761, + "grad_norm": 0.592209176007707, + "learning_rate": 2.2376817890780128e-05, + "loss": 0.2346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11458073556423187, + "step": 6895, + "valid_targets_mean": 3865.4, + "valid_targets_min": 2335 + }, + { + "epoch": 3.6144578313253013, + "grad_norm": 0.5875876100016537, + "learning_rate": 2.2350877660422475e-05, + "loss": 0.2241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11619818955659866, + "step": 6900, + "valid_targets_mean": 3269.5, + "valid_targets_min": 1255 + }, + { + "epoch": 3.617077003666841, + "grad_norm": 0.5762951345901648, + "learning_rate": 2.232493341929885e-05, + "loss": 0.2225, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12324737012386322, + "step": 6905, + "valid_targets_mean": 3915.5, + "valid_targets_min": 2355 + }, + { + "epoch": 3.6196961760083814, + "grad_norm": 0.5827419064301562, + "learning_rate": 2.2298985211671975e-05, + "loss": 0.2196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09965714812278748, + "step": 6910, + "valid_targets_mean": 3704.9, + "valid_targets_min": 1281 + }, + { + "epoch": 3.6223153483499213, + "grad_norm": 0.589138398013212, + "learning_rate": 2.2273033081811367e-05, + "loss": 0.2227, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12496821582317352, + "step": 6915, + "valid_targets_mean": 3788.6, + "valid_targets_min": 1836 + }, + { + "epoch": 3.6249345206914616, + "grad_norm": 0.5503619080652081, + "learning_rate": 2.2247077073993217e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1558687388896942, + "step": 6920, + "valid_targets_mean": 5778.1, + "valid_targets_min": 4218 + }, + { + "epoch": 3.6275536930330015, + "grad_norm": 0.5459978807624811, + "learning_rate": 2.222111723250034e-05, + "loss": 0.222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11263719201087952, + "step": 6925, + "valid_targets_mean": 3982.6, + "valid_targets_min": 2717 + }, + { + "epoch": 3.6301728653745418, + "grad_norm": 0.6073482618129737, + "learning_rate": 2.219515360162208e-05, + "loss": 0.223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11600340157747269, + "step": 6930, + "valid_targets_mean": 2804.9, + "valid_targets_min": 1171 + }, + { + "epoch": 3.6327920377160816, + "grad_norm": 0.5824948052376526, + "learning_rate": 2.2169186225654266e-05, + "loss": 0.2426, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12364323437213898, + "step": 6935, + "valid_targets_mean": 4265.1, + "valid_targets_min": 2177 + }, + { + "epoch": 3.635411210057622, + "grad_norm": 0.5069483275684302, + "learning_rate": 2.2143215148899094e-05, + "loss": 0.2316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11704994738101959, + "step": 6940, + "valid_targets_mean": 4090.0, + "valid_targets_min": 2368 + }, + { + "epoch": 3.6380303823991618, + "grad_norm": 0.5509749114728657, + "learning_rate": 2.2117240415665093e-05, + "loss": 0.2391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10130609571933746, + "step": 6945, + "valid_targets_mean": 4149.6, + "valid_targets_min": 2294 + }, + { + "epoch": 3.640649554740702, + "grad_norm": 0.6699929852879776, + "learning_rate": 2.2091262070267017e-05, + "loss": 0.2368, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13128294050693512, + "step": 6950, + "valid_targets_mean": 3977.1, + "valid_targets_min": 1135 + }, + { + "epoch": 3.643268727082242, + "grad_norm": 0.5844214510751357, + "learning_rate": 2.2065280157025786e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12205550819635391, + "step": 6955, + "valid_targets_mean": 4319.0, + "valid_targets_min": 1449 + }, + { + "epoch": 3.645887899423782, + "grad_norm": 0.6183671048355622, + "learning_rate": 2.2039294720268413e-05, + "loss": 0.2409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12880170345306396, + "step": 6960, + "valid_targets_mean": 3227.5, + "valid_targets_min": 1400 + }, + { + "epoch": 3.648507071765322, + "grad_norm": 0.548420129345648, + "learning_rate": 2.2013305804327923e-05, + "loss": 0.2405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09395840764045715, + "step": 6965, + "valid_targets_mean": 3268.4, + "valid_targets_min": 1433 + }, + { + "epoch": 3.6511262441068624, + "grad_norm": 0.592091964883014, + "learning_rate": 2.1987313453543254e-05, + "loss": 0.2229, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0836934745311737, + "step": 6970, + "valid_targets_mean": 2946.5, + "valid_targets_min": 614 + }, + { + "epoch": 3.6537454164484022, + "grad_norm": 0.5989412408823244, + "learning_rate": 2.1961317712259237e-05, + "loss": 0.2278, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0994023010134697, + "step": 6975, + "valid_targets_mean": 3125.0, + "valid_targets_min": 2041 + }, + { + "epoch": 3.6563645887899425, + "grad_norm": 0.6348009504581797, + "learning_rate": 2.193531862482646e-05, + "loss": 0.2396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12506204843521118, + "step": 6980, + "valid_targets_mean": 3852.9, + "valid_targets_min": 1333 + }, + { + "epoch": 3.6589837611314824, + "grad_norm": 0.5524264083494778, + "learning_rate": 2.1909316235601246e-05, + "loss": 0.2297, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09926673024892807, + "step": 6985, + "valid_targets_mean": 3973.2, + "valid_targets_min": 1325 + }, + { + "epoch": 3.6616029334730227, + "grad_norm": 0.5010378997520444, + "learning_rate": 2.1883310588945523e-05, + "loss": 0.2251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08619755506515503, + "step": 6990, + "valid_targets_mean": 3510.1, + "valid_targets_min": 2439 + }, + { + "epoch": 3.6642221058145625, + "grad_norm": 0.5867378174433812, + "learning_rate": 2.18573017292268e-05, + "loss": 0.2172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10960593819618225, + "step": 6995, + "valid_targets_mean": 3699.4, + "valid_targets_min": 2651 + }, + { + "epoch": 3.666841278156103, + "grad_norm": 0.5873937984830838, + "learning_rate": 2.1831289700818058e-05, + "loss": 0.2308, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13545319437980652, + "step": 7000, + "valid_targets_mean": 4253.4, + "valid_targets_min": 1076 + }, + { + "epoch": 3.6694604504976427, + "grad_norm": 0.5384263202572114, + "learning_rate": 2.1805274548097675e-05, + "loss": 0.2117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12658920884132385, + "step": 7005, + "valid_targets_mean": 5076.9, + "valid_targets_min": 2127 + }, + { + "epoch": 3.672079622839183, + "grad_norm": 0.5693117959387839, + "learning_rate": 2.177925631544937e-05, + "loss": 0.2316, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1129695400595665, + "step": 7010, + "valid_targets_mean": 4415.2, + "valid_targets_min": 3904 + }, + { + "epoch": 3.674698795180723, + "grad_norm": 0.6012527915626795, + "learning_rate": 2.1753235047262124e-05, + "loss": 0.2272, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08253274857997894, + "step": 7015, + "valid_targets_mean": 2814.0, + "valid_targets_min": 793 + }, + { + "epoch": 3.6773179675222627, + "grad_norm": 0.5471177256049523, + "learning_rate": 2.172721078793008e-05, + "loss": 0.2267, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11194857209920883, + "step": 7020, + "valid_targets_mean": 4488.0, + "valid_targets_min": 1777 + }, + { + "epoch": 3.679937139863803, + "grad_norm": 0.5585907801126335, + "learning_rate": 2.1701183581852497e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09934152662754059, + "step": 7025, + "valid_targets_mean": 3447.1, + "valid_targets_min": 1965 + }, + { + "epoch": 3.6825563122053433, + "grad_norm": 0.5962801917924513, + "learning_rate": 2.1675153473433653e-05, + "loss": 0.2319, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12726783752441406, + "step": 7030, + "valid_targets_mean": 4516.6, + "valid_targets_min": 1206 + }, + { + "epoch": 3.685175484546883, + "grad_norm": 0.5294001970990402, + "learning_rate": 2.1649120507082776e-05, + "loss": 0.2304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12501150369644165, + "step": 7035, + "valid_targets_mean": 5388.2, + "valid_targets_min": 3599 + }, + { + "epoch": 3.687794656888423, + "grad_norm": 0.5538210281001408, + "learning_rate": 2.162308472721398e-05, + "loss": 0.2206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10568016767501831, + "step": 7040, + "valid_targets_mean": 3637.1, + "valid_targets_min": 1250 + }, + { + "epoch": 3.6904138292299633, + "grad_norm": 0.5417249555470527, + "learning_rate": 2.1597046178246174e-05, + "loss": 0.2176, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11221703886985779, + "step": 7045, + "valid_targets_mean": 4201.2, + "valid_targets_min": 1876 + }, + { + "epoch": 3.6930330015715036, + "grad_norm": 0.5382568442486038, + "learning_rate": 2.1571004904602993e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11344371736049652, + "step": 7050, + "valid_targets_mean": 4462.0, + "valid_targets_min": 2487 + }, + { + "epoch": 3.6956521739130435, + "grad_norm": 0.5710459438577438, + "learning_rate": 2.154496095071272e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0923365131020546, + "step": 7055, + "valid_targets_mean": 3380.9, + "valid_targets_min": 1360 + }, + { + "epoch": 3.6982713462545833, + "grad_norm": 0.5608129507480488, + "learning_rate": 2.1518914361008197e-05, + "loss": 0.214, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11632062494754791, + "step": 7060, + "valid_targets_mean": 4094.6, + "valid_targets_min": 1496 + }, + { + "epoch": 3.7008905185961236, + "grad_norm": 0.6159659135935731, + "learning_rate": 2.1492865179926786e-05, + "loss": 0.2145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1226055696606636, + "step": 7065, + "valid_targets_mean": 3533.5, + "valid_targets_min": 2059 + }, + { + "epoch": 3.703509690937664, + "grad_norm": 0.5922107534926981, + "learning_rate": 2.1466813451910253e-05, + "loss": 0.2153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11977484822273254, + "step": 7070, + "valid_targets_mean": 4585.9, + "valid_targets_min": 2890 + }, + { + "epoch": 3.7061288632792038, + "grad_norm": 0.7172908830730714, + "learning_rate": 2.144075922140472e-05, + "loss": 0.2234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08903069794178009, + "step": 7075, + "valid_targets_mean": 3349.1, + "valid_targets_min": 1598 + }, + { + "epoch": 3.7087480356207436, + "grad_norm": 0.5512463394944735, + "learning_rate": 2.1414702532860576e-05, + "loss": 0.2355, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09501080960035324, + "step": 7080, + "valid_targets_mean": 3761.4, + "valid_targets_min": 2267 + }, + { + "epoch": 3.711367207962284, + "grad_norm": 0.6285891791369463, + "learning_rate": 2.1388643430732396e-05, + "loss": 0.2152, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10063962638378143, + "step": 7085, + "valid_targets_mean": 2590.0, + "valid_targets_min": 1279 + }, + { + "epoch": 3.713986380303824, + "grad_norm": 0.6176121071949148, + "learning_rate": 2.1362581959478877e-05, + "loss": 0.2181, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1172870546579361, + "step": 7090, + "valid_targets_mean": 3247.0, + "valid_targets_min": 1730 + }, + { + "epoch": 3.716605552645364, + "grad_norm": 0.6665565885663419, + "learning_rate": 2.133651816356276e-05, + "loss": 0.2231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11746871471405029, + "step": 7095, + "valid_targets_mean": 3144.8, + "valid_targets_min": 1664 + }, + { + "epoch": 3.719224724986904, + "grad_norm": 0.644759156491007, + "learning_rate": 2.1310452087450755e-05, + "loss": 0.216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08383020758628845, + "step": 7100, + "valid_targets_mean": 3659.5, + "valid_targets_min": 1945 + }, + { + "epoch": 3.7218438973284442, + "grad_norm": 0.6126868383162574, + "learning_rate": 2.1284383775613456e-05, + "loss": 0.2328, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12941987812519073, + "step": 7105, + "valid_targets_mean": 4016.2, + "valid_targets_min": 2418 + }, + { + "epoch": 3.7244630696699845, + "grad_norm": 0.5620786119548035, + "learning_rate": 2.1258313272525277e-05, + "loss": 0.2249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11668717116117477, + "step": 7110, + "valid_targets_mean": 4044.8, + "valid_targets_min": 2100 + }, + { + "epoch": 3.7270822420115244, + "grad_norm": 0.5986464512923034, + "learning_rate": 2.123224062266436e-05, + "loss": 0.2202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11451369524002075, + "step": 7115, + "valid_targets_mean": 3743.8, + "valid_targets_min": 1365 + }, + { + "epoch": 3.7297014143530642, + "grad_norm": 0.549946872451894, + "learning_rate": 2.1206165870512522e-05, + "loss": 0.2241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10152804851531982, + "step": 7120, + "valid_targets_mean": 4017.1, + "valid_targets_min": 1572 + }, + { + "epoch": 3.7323205866946045, + "grad_norm": 0.6302653929330909, + "learning_rate": 2.1180089060555162e-05, + "loss": 0.2298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10293979197740555, + "step": 7125, + "valid_targets_mean": 3171.4, + "valid_targets_min": 876 + }, + { + "epoch": 3.734939759036145, + "grad_norm": 0.4160110504811606, + "learning_rate": 2.115401023728118e-05, + "loss": 0.1216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0484163798391819, + "step": 7130, + "valid_targets_mean": 2722.6, + "valid_targets_min": 458 + }, + { + "epoch": 3.7375589313776847, + "grad_norm": 0.4056044651742565, + "learning_rate": 2.112792944518293e-05, + "loss": 0.0917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046069204807281494, + "step": 7135, + "valid_targets_mean": 3113.1, + "valid_targets_min": 2461 + }, + { + "epoch": 3.7401781037192245, + "grad_norm": 0.35401091971804155, + "learning_rate": 2.1101846728756107e-05, + "loss": 0.0839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04864474758505821, + "step": 7140, + "valid_targets_mean": 3648.1, + "valid_targets_min": 2068 + }, + { + "epoch": 3.742797276060765, + "grad_norm": 0.44052829127698223, + "learning_rate": 2.1075762132499702e-05, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049874428659677505, + "step": 7145, + "valid_targets_mean": 3423.6, + "valid_targets_min": 2425 + }, + { + "epoch": 3.745416448402305, + "grad_norm": 0.45832517949006607, + "learning_rate": 2.104967570091591e-05, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04273105412721634, + "step": 7150, + "valid_targets_mean": 2724.1, + "valid_targets_min": 1198 + }, + { + "epoch": 3.748035620743845, + "grad_norm": 0.6908025070624498, + "learning_rate": 2.1023587478510045e-05, + "loss": 0.1011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0726727768778801, + "step": 7155, + "valid_targets_mean": 1232.5, + "valid_targets_min": 807 + }, + { + "epoch": 3.750654793085385, + "grad_norm": 0.4487537007971137, + "learning_rate": 2.0997497509790498e-05, + "loss": 0.1383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05464408919215202, + "step": 7160, + "valid_targets_mean": 3487.4, + "valid_targets_min": 2351 + }, + { + "epoch": 3.753273965426925, + "grad_norm": 0.46932742552426077, + "learning_rate": 2.097140583926862e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07225260883569717, + "step": 7165, + "valid_targets_mean": 3126.4, + "valid_targets_min": 781 + }, + { + "epoch": 3.755893137768465, + "grad_norm": 0.3353048040606108, + "learning_rate": 2.094531251145868e-05, + "loss": 0.1041, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04661848023533821, + "step": 7170, + "valid_targets_mean": 3474.8, + "valid_targets_min": 799 + }, + { + "epoch": 3.7585123101100053, + "grad_norm": 0.3056280910177452, + "learning_rate": 2.0919217570877758e-05, + "loss": 0.1087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04121142625808716, + "step": 7175, + "valid_targets_mean": 5026.1, + "valid_targets_min": 2802 + }, + { + "epoch": 3.761131482451545, + "grad_norm": 0.485070425264739, + "learning_rate": 2.0893121062045708e-05, + "loss": 0.0826, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0324074923992157, + "step": 7180, + "valid_targets_mean": 2611.0, + "valid_targets_min": 574 + }, + { + "epoch": 3.7637506547930855, + "grad_norm": 0.335213389340215, + "learning_rate": 2.0867023029485032e-05, + "loss": 0.0907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04309866577386856, + "step": 7185, + "valid_targets_mean": 4036.5, + "valid_targets_min": 2642 + }, + { + "epoch": 3.7663698271346253, + "grad_norm": 0.5066021264924445, + "learning_rate": 2.0840923517720854e-05, + "loss": 0.0935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06152278184890747, + "step": 7190, + "valid_targets_mean": 1960.5, + "valid_targets_min": 593 + }, + { + "epoch": 3.7689889994761656, + "grad_norm": 0.4316307995276316, + "learning_rate": 2.081482257128082e-05, + "loss": 0.1184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03745308145880699, + "step": 7195, + "valid_targets_mean": 3354.2, + "valid_targets_min": 1353 + }, + { + "epoch": 3.7716081718177055, + "grad_norm": 0.46209727079168716, + "learning_rate": 2.078872023469501e-05, + "loss": 0.0858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04949069023132324, + "step": 7200, + "valid_targets_mean": 2300.9, + "valid_targets_min": 818 + }, + { + "epoch": 3.7742273441592458, + "grad_norm": 0.30903460192052334, + "learning_rate": 2.076261655249588e-05, + "loss": 0.0855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029910128563642502, + "step": 7205, + "valid_targets_mean": 2295.8, + "valid_targets_min": 934 + }, + { + "epoch": 3.7768465165007856, + "grad_norm": 0.3458789313427291, + "learning_rate": 2.0736511569218186e-05, + "loss": 0.0786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04335102438926697, + "step": 7210, + "valid_targets_mean": 3524.4, + "valid_targets_min": 2798 + }, + { + "epoch": 3.779465688842326, + "grad_norm": 0.42898136088142014, + "learning_rate": 2.071040532939891e-05, + "loss": 0.0987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039288848638534546, + "step": 7215, + "valid_targets_mean": 2810.1, + "valid_targets_min": 471 + }, + { + "epoch": 3.7820848611838658, + "grad_norm": 0.41360948784950297, + "learning_rate": 2.0684297877577165e-05, + "loss": 0.1001, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0632447600364685, + "step": 7220, + "valid_targets_mean": 3042.8, + "valid_targets_min": 903 + }, + { + "epoch": 3.784704033525406, + "grad_norm": 0.4230208985969971, + "learning_rate": 2.0658189258294144e-05, + "loss": 0.0999, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04112110659480095, + "step": 7225, + "valid_targets_mean": 3081.6, + "valid_targets_min": 474 + }, + { + "epoch": 3.787323205866946, + "grad_norm": 0.39799952163467855, + "learning_rate": 2.0632079516093015e-05, + "loss": 0.1011, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04688946530222893, + "step": 7230, + "valid_targets_mean": 3468.0, + "valid_targets_min": 640 + }, + { + "epoch": 3.7899423782084862, + "grad_norm": 0.329745560958774, + "learning_rate": 2.0605968695518873e-05, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03248301148414612, + "step": 7235, + "valid_targets_mean": 3438.4, + "valid_targets_min": 902 + }, + { + "epoch": 3.792561550550026, + "grad_norm": 0.39561398130590325, + "learning_rate": 2.057985684111866e-05, + "loss": 0.0839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04233215004205704, + "step": 7240, + "valid_targets_mean": 3212.8, + "valid_targets_min": 1837 + }, + { + "epoch": 3.7951807228915664, + "grad_norm": 0.6648216549438575, + "learning_rate": 2.0553743997441066e-05, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08644934743642807, + "step": 7245, + "valid_targets_mean": 1358.8, + "valid_targets_min": 756 + }, + { + "epoch": 3.7977998952331062, + "grad_norm": 0.5279683786067549, + "learning_rate": 2.0527630209036474e-05, + "loss": 0.1125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08262963593006134, + "step": 7250, + "valid_targets_mean": 2896.2, + "valid_targets_min": 1072 + }, + { + "epoch": 3.8004190675746465, + "grad_norm": 0.3684273942546229, + "learning_rate": 2.0501515520456895e-05, + "loss": 0.0904, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029133372008800507, + "step": 7255, + "valid_targets_mean": 2908.1, + "valid_targets_min": 895 + }, + { + "epoch": 3.8030382399161864, + "grad_norm": 0.5443844973474393, + "learning_rate": 2.047539997625584e-05, + "loss": 0.2054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18386998772621155, + "step": 7260, + "valid_targets_mean": 2562.2, + "valid_targets_min": 878 + }, + { + "epoch": 3.8056574122577267, + "grad_norm": 0.43643402431058953, + "learning_rate": 2.0449283620988322e-05, + "loss": 0.1899, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04157412052154541, + "step": 7265, + "valid_targets_mean": 2087.6, + "valid_targets_min": 756 + }, + { + "epoch": 3.8082765845992665, + "grad_norm": 0.3383277384752161, + "learning_rate": 2.04231664992107e-05, + "loss": 0.1019, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03680277615785599, + "step": 7270, + "valid_targets_mean": 3421.8, + "valid_targets_min": 2963 + }, + { + "epoch": 3.810895756940807, + "grad_norm": 0.7113310097745058, + "learning_rate": 2.039704865548067e-05, + "loss": 0.1704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.22442680597305298, + "step": 7275, + "valid_targets_mean": 2375.2, + "valid_targets_min": 1153 + }, + { + "epoch": 3.8135149292823467, + "grad_norm": 0.3151565963037599, + "learning_rate": 2.0370930134357142e-05, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03634630888700485, + "step": 7280, + "valid_targets_mean": 4886.2, + "valid_targets_min": 1700 + }, + { + "epoch": 3.816134101623887, + "grad_norm": 0.3330309683018269, + "learning_rate": 2.0344810980400185e-05, + "loss": 0.0898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029430480673909187, + "step": 7285, + "valid_targets_mean": 3965.1, + "valid_targets_min": 789 + }, + { + "epoch": 3.818753273965427, + "grad_norm": 0.378476474527609, + "learning_rate": 2.031869123817096e-05, + "loss": 0.097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0501452311873436, + "step": 7290, + "valid_targets_mean": 3771.5, + "valid_targets_min": 995 + }, + { + "epoch": 3.821372446306967, + "grad_norm": 0.35063006936867425, + "learning_rate": 2.029257095223161e-05, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030551211908459663, + "step": 7295, + "valid_targets_mean": 2220.8, + "valid_targets_min": 488 + }, + { + "epoch": 3.823991618648507, + "grad_norm": 0.43616254819968925, + "learning_rate": 2.026645016714522e-05, + "loss": 0.0889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0527697391808033, + "step": 7300, + "valid_targets_mean": 3635.4, + "valid_targets_min": 2437 + }, + { + "epoch": 3.8266107909900473, + "grad_norm": 0.45631116816128553, + "learning_rate": 2.024032892747573e-05, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03690987825393677, + "step": 7305, + "valid_targets_mean": 1752.1, + "valid_targets_min": 712 + }, + { + "epoch": 3.829229963331587, + "grad_norm": 0.3838933685566221, + "learning_rate": 2.021420727778784e-05, + "loss": 0.0858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042619407176971436, + "step": 7310, + "valid_targets_mean": 3713.4, + "valid_targets_min": 2422 + }, + { + "epoch": 3.831849135673127, + "grad_norm": 0.4624646777898602, + "learning_rate": 2.018808526264697e-05, + "loss": 0.0764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040911175310611725, + "step": 7315, + "valid_targets_mean": 3998.5, + "valid_targets_min": 3643 + }, + { + "epoch": 3.8344683080146673, + "grad_norm": 0.3312099465745374, + "learning_rate": 2.0161962926619144e-05, + "loss": 0.1317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031162984669208527, + "step": 7320, + "valid_targets_mean": 3319.0, + "valid_targets_min": 772 + }, + { + "epoch": 3.8370874803562076, + "grad_norm": 0.35766875913643204, + "learning_rate": 2.013584031427094e-05, + "loss": 0.0701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029681073501706123, + "step": 7325, + "valid_targets_mean": 2934.8, + "valid_targets_min": 722 + }, + { + "epoch": 3.8397066526977475, + "grad_norm": 0.325182321457188, + "learning_rate": 2.010971747016941e-05, + "loss": 0.0776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03128822147846222, + "step": 7330, + "valid_targets_mean": 3656.4, + "valid_targets_min": 2788 + }, + { + "epoch": 3.8423258250392873, + "grad_norm": 0.3796258363732989, + "learning_rate": 2.0083594438882006e-05, + "loss": 0.0804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041404254734516144, + "step": 7335, + "valid_targets_mean": 3202.5, + "valid_targets_min": 948 + }, + { + "epoch": 3.8449449973808276, + "grad_norm": 0.5370441827609428, + "learning_rate": 2.0057471264976493e-05, + "loss": 0.098, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047868311405181885, + "step": 7340, + "valid_targets_mean": 1002.0, + "valid_targets_min": 490 + }, + { + "epoch": 3.847564169722368, + "grad_norm": 0.36151419425053055, + "learning_rate": 2.0031347993020876e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04746914282441139, + "step": 7345, + "valid_targets_mean": 3751.9, + "valid_targets_min": 2800 + }, + { + "epoch": 3.8501833420639078, + "grad_norm": 0.48650446970590927, + "learning_rate": 2.0005224667583337e-05, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07138540595769882, + "step": 7350, + "valid_targets_mean": 2641.8, + "valid_targets_min": 896 + }, + { + "epoch": 3.8528025144054476, + "grad_norm": 0.43996135113983476, + "learning_rate": 1.9979101333232136e-05, + "loss": 0.0945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052057426422834396, + "step": 7355, + "valid_targets_mean": 3199.0, + "valid_targets_min": 1333 + }, + { + "epoch": 3.855421686746988, + "grad_norm": 0.48053871730861536, + "learning_rate": 1.9952978034535556e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049403298646211624, + "step": 7360, + "valid_targets_mean": 2548.1, + "valid_targets_min": 693 + }, + { + "epoch": 3.8580408590885282, + "grad_norm": 0.37707605809553124, + "learning_rate": 1.9926854816061833e-05, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03836170583963394, + "step": 7365, + "valid_targets_mean": 2482.9, + "valid_targets_min": 803 + }, + { + "epoch": 3.860660031430068, + "grad_norm": 0.37591192757871905, + "learning_rate": 1.990073172237904e-05, + "loss": 0.0927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04763051122426987, + "step": 7370, + "valid_targets_mean": 3239.2, + "valid_targets_min": 2525 + }, + { + "epoch": 3.863279203771608, + "grad_norm": 0.5092058090562325, + "learning_rate": 1.9874608798055045e-05, + "loss": 0.1144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04198355972766876, + "step": 7375, + "valid_targets_mean": 3806.1, + "valid_targets_min": 2393 + }, + { + "epoch": 3.8658983761131482, + "grad_norm": 0.4208071842163223, + "learning_rate": 1.9848486087657447e-05, + "loss": 0.0977, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04279317706823349, + "step": 7380, + "valid_targets_mean": 2990.9, + "valid_targets_min": 804 + }, + { + "epoch": 3.8685175484546885, + "grad_norm": 0.359992259446852, + "learning_rate": 1.982236363575345e-05, + "loss": 0.0893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04400169476866722, + "step": 7385, + "valid_targets_mean": 4885.8, + "valid_targets_min": 3695 + }, + { + "epoch": 3.8711367207962284, + "grad_norm": 0.40927072887229127, + "learning_rate": 1.9796241486909836e-05, + "loss": 0.081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04953446239233017, + "step": 7390, + "valid_targets_mean": 3128.8, + "valid_targets_min": 596 + }, + { + "epoch": 3.8737558931377682, + "grad_norm": 0.2709069876826699, + "learning_rate": 1.9770119685692874e-05, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029946565628051758, + "step": 7395, + "valid_targets_mean": 4567.8, + "valid_targets_min": 3804 + }, + { + "epoch": 3.8763750654793085, + "grad_norm": 0.34901230839138486, + "learning_rate": 1.974399827666822e-05, + "loss": 0.0852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033761754631996155, + "step": 7400, + "valid_targets_mean": 3853.1, + "valid_targets_min": 2862 + }, + { + "epoch": 3.878994237820849, + "grad_norm": 0.5379326016494378, + "learning_rate": 1.971787730440087e-05, + "loss": 0.1054, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060941699892282486, + "step": 7405, + "valid_targets_mean": 1072.6, + "valid_targets_min": 497 + }, + { + "epoch": 3.8816134101623887, + "grad_norm": 0.4057047828128052, + "learning_rate": 1.9691756813455084e-05, + "loss": 0.0893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051376499235630035, + "step": 7410, + "valid_targets_mean": 4191.0, + "valid_targets_min": 3918 + }, + { + "epoch": 3.8842325825039286, + "grad_norm": 0.5401601213083609, + "learning_rate": 1.9665636848394297e-05, + "loss": 0.1007, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04499197006225586, + "step": 7415, + "valid_targets_mean": 2121.9, + "valid_targets_min": 785 + }, + { + "epoch": 3.886851754845469, + "grad_norm": 0.32132723882855646, + "learning_rate": 1.9639517453781032e-05, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03625746816396713, + "step": 7420, + "valid_targets_mean": 3451.6, + "valid_targets_min": 905 + }, + { + "epoch": 3.889470927187009, + "grad_norm": 0.4135392476031905, + "learning_rate": 1.961339867417687e-05, + "loss": 0.0965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0686865895986557, + "step": 7425, + "valid_targets_mean": 2929.2, + "valid_targets_min": 657 + }, + { + "epoch": 3.892090099528549, + "grad_norm": 0.37515512830809106, + "learning_rate": 1.9587280554142307e-05, + "loss": 0.0878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0452505387365818, + "step": 7430, + "valid_targets_mean": 3798.9, + "valid_targets_min": 3184 + }, + { + "epoch": 3.894709271870089, + "grad_norm": 0.5023228883386553, + "learning_rate": 1.9561163138236723e-05, + "loss": 0.0792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04881703853607178, + "step": 7435, + "valid_targets_mean": 1463.9, + "valid_targets_min": 627 + }, + { + "epoch": 3.897328444211629, + "grad_norm": 0.47230694191226646, + "learning_rate": 1.9535046471018326e-05, + "loss": 0.0883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04295075684785843, + "step": 7440, + "valid_targets_mean": 1701.5, + "valid_targets_min": 827 + }, + { + "epoch": 3.8999476165531695, + "grad_norm": 0.45915026381292917, + "learning_rate": 1.9508930597044e-05, + "loss": 0.0916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04063272103667259, + "step": 7445, + "valid_targets_mean": 1789.9, + "valid_targets_min": 940 + }, + { + "epoch": 3.9025667888947093, + "grad_norm": 0.3954345884768852, + "learning_rate": 1.9482815560869327e-05, + "loss": 0.1062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03934101015329361, + "step": 7450, + "valid_targets_mean": 2943.8, + "valid_targets_min": 958 + }, + { + "epoch": 3.905185961236249, + "grad_norm": 0.4364668314569832, + "learning_rate": 1.9456701407048414e-05, + "loss": 0.0879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03706955537199974, + "step": 7455, + "valid_targets_mean": 1859.2, + "valid_targets_min": 719 + }, + { + "epoch": 3.9078051335777895, + "grad_norm": 0.47194378323597347, + "learning_rate": 1.943058818013388e-05, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034562960267066956, + "step": 7460, + "valid_targets_mean": 1703.9, + "valid_targets_min": 759 + }, + { + "epoch": 3.9104243059193298, + "grad_norm": 0.700268989503475, + "learning_rate": 1.9404475924676777e-05, + "loss": 0.1096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11131960153579712, + "step": 7465, + "valid_targets_mean": 1556.6, + "valid_targets_min": 521 + }, + { + "epoch": 3.9130434782608696, + "grad_norm": 0.43658523058972476, + "learning_rate": 1.9378364685226468e-05, + "loss": 0.0885, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05440928041934967, + "step": 7470, + "valid_targets_mean": 3267.6, + "valid_targets_min": 1032 + }, + { + "epoch": 3.9156626506024095, + "grad_norm": 0.28844422616853477, + "learning_rate": 1.9352254506330626e-05, + "loss": 0.0754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030360206961631775, + "step": 7475, + "valid_targets_mean": 3609.0, + "valid_targets_min": 2741 + }, + { + "epoch": 3.9182818229439498, + "grad_norm": 0.3715990132553998, + "learning_rate": 1.932614543253506e-05, + "loss": 0.0868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0373227596282959, + "step": 7480, + "valid_targets_mean": 2735.8, + "valid_targets_min": 1692 + }, + { + "epoch": 3.9209009952854896, + "grad_norm": 0.3981016191642541, + "learning_rate": 1.9300037508383753e-05, + "loss": 0.1156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049582041800022125, + "step": 7485, + "valid_targets_mean": 3693.4, + "valid_targets_min": 2215 + }, + { + "epoch": 3.92352016762703, + "grad_norm": 0.34583298706238536, + "learning_rate": 1.9273930778418685e-05, + "loss": 0.0727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03250321000814438, + "step": 7490, + "valid_targets_mean": 3109.9, + "valid_targets_min": 723 + }, + { + "epoch": 3.92613933996857, + "grad_norm": 0.31314955780607784, + "learning_rate": 1.9247825287179796e-05, + "loss": 0.0906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03753659129142761, + "step": 7495, + "valid_targets_mean": 4984.2, + "valid_targets_min": 3282 + }, + { + "epoch": 3.92875851231011, + "grad_norm": 0.2801049301493354, + "learning_rate": 1.922172107920495e-05, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035190075635910034, + "step": 7500, + "valid_targets_mean": 3863.4, + "valid_targets_min": 995 + }, + { + "epoch": 3.93137768465165, + "grad_norm": 0.2595165676984985, + "learning_rate": 1.9195618199029783e-05, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034311793744564056, + "step": 7505, + "valid_targets_mean": 4201.0, + "valid_targets_min": 949 + }, + { + "epoch": 3.9339968569931902, + "grad_norm": 0.38550787364432304, + "learning_rate": 1.916951669118769e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058845240622758865, + "step": 7510, + "valid_targets_mean": 4321.6, + "valid_targets_min": 2724 + }, + { + "epoch": 3.93661602933473, + "grad_norm": 0.3234258924666239, + "learning_rate": 1.9143416600209716e-05, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0298369862139225, + "step": 7515, + "valid_targets_mean": 3384.5, + "valid_targets_min": 424 + }, + { + "epoch": 3.9392352016762704, + "grad_norm": 0.38794089776795426, + "learning_rate": 1.9117317970624484e-05, + "loss": 0.1107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03481116145849228, + "step": 7520, + "valid_targets_mean": 2987.2, + "valid_targets_min": 531 + }, + { + "epoch": 3.9418543740178102, + "grad_norm": 0.35515793157957803, + "learning_rate": 1.9091220846958133e-05, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034674271941185, + "step": 7525, + "valid_targets_mean": 3469.5, + "valid_targets_min": 1195 + }, + { + "epoch": 3.9444735463593505, + "grad_norm": 0.3608196817905148, + "learning_rate": 1.9065125273734228e-05, + "loss": 0.0747, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06690368056297302, + "step": 7530, + "valid_targets_mean": 2682.8, + "valid_targets_min": 722 + }, + { + "epoch": 3.9470927187008904, + "grad_norm": 0.35202453577339904, + "learning_rate": 1.903903129547369e-05, + "loss": 0.1039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03734735772013664, + "step": 7535, + "valid_targets_mean": 3496.9, + "valid_targets_min": 1251 + }, + { + "epoch": 3.9497118910424307, + "grad_norm": 0.38729346471328996, + "learning_rate": 1.9012938956694714e-05, + "loss": 0.0833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0460013672709465, + "step": 7540, + "valid_targets_mean": 2912.5, + "valid_targets_min": 970 + }, + { + "epoch": 3.9523310633839706, + "grad_norm": 0.5250253138141149, + "learning_rate": 1.8986848301912707e-05, + "loss": 0.0952, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04620261862874031, + "step": 7545, + "valid_targets_mean": 1423.6, + "valid_targets_min": 634 + }, + { + "epoch": 3.954950235725511, + "grad_norm": 0.5484336799506357, + "learning_rate": 1.8960759375640204e-05, + "loss": 0.1086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07207681238651276, + "step": 7550, + "valid_targets_mean": 1423.9, + "valid_targets_min": 685 + }, + { + "epoch": 3.9575694080670507, + "grad_norm": 0.3855867199379245, + "learning_rate": 1.8934672222386762e-05, + "loss": 0.1179, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03402852267026901, + "step": 7555, + "valid_targets_mean": 2932.5, + "valid_targets_min": 593 + }, + { + "epoch": 3.960188580408591, + "grad_norm": 0.3505139205213625, + "learning_rate": 1.8908586886658967e-05, + "loss": 0.0909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03425735980272293, + "step": 7560, + "valid_targets_mean": 3945.8, + "valid_targets_min": 3129 + }, + { + "epoch": 3.962807752750131, + "grad_norm": 0.410923239791014, + "learning_rate": 1.8882503412960247e-05, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051767900586128235, + "step": 7565, + "valid_targets_mean": 3621.1, + "valid_targets_min": 859 + }, + { + "epoch": 3.965426925091671, + "grad_norm": 0.4460120714969015, + "learning_rate": 1.88564218457909e-05, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04497465491294861, + "step": 7570, + "valid_targets_mean": 2553.0, + "valid_targets_min": 616 + }, + { + "epoch": 3.968046097433211, + "grad_norm": 0.3871970576508809, + "learning_rate": 1.8830342229647945e-05, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03743528574705124, + "step": 7575, + "valid_targets_mean": 2690.5, + "valid_targets_min": 541 + }, + { + "epoch": 3.9706652697747513, + "grad_norm": 0.6847408857050856, + "learning_rate": 1.880426460902507e-05, + "loss": 0.0807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053018633276224136, + "step": 7580, + "valid_targets_mean": 906.4, + "valid_targets_min": 598 + }, + { + "epoch": 3.973284442116291, + "grad_norm": 0.4077014400207548, + "learning_rate": 1.8778189028412574e-05, + "loss": 0.0968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042825065553188324, + "step": 7585, + "valid_targets_mean": 3147.2, + "valid_targets_min": 1101 + }, + { + "epoch": 3.9759036144578315, + "grad_norm": 0.4951055792239926, + "learning_rate": 1.8752115532297263e-05, + "loss": 0.0792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0355534702539444, + "step": 7590, + "valid_targets_mean": 2331.8, + "valid_targets_min": 680 + }, + { + "epoch": 3.9785227867993713, + "grad_norm": 0.36487242951948884, + "learning_rate": 1.8726044165162406e-05, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04377260059118271, + "step": 7595, + "valid_targets_mean": 2577.0, + "valid_targets_min": 774 + }, + { + "epoch": 3.9811419591409116, + "grad_norm": 0.3924887363373228, + "learning_rate": 1.869997497148761e-05, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050407372415065765, + "step": 7600, + "valid_targets_mean": 4140.1, + "valid_targets_min": 1474 + }, + { + "epoch": 3.9837611314824515, + "grad_norm": 0.24620775061700506, + "learning_rate": 1.8673907995748802e-05, + "loss": 0.0746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02671591192483902, + "step": 7605, + "valid_targets_mean": 5498.9, + "valid_targets_min": 3677 + }, + { + "epoch": 3.9863803038239918, + "grad_norm": 0.3645694595521768, + "learning_rate": 1.8647843282418114e-05, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04993020370602608, + "step": 7610, + "valid_targets_mean": 3352.5, + "valid_targets_min": 675 + }, + { + "epoch": 3.9889994761655316, + "grad_norm": 0.33029737536102133, + "learning_rate": 1.86217808759638e-05, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03337732329964638, + "step": 7615, + "valid_targets_mean": 3902.5, + "valid_targets_min": 795 + }, + { + "epoch": 3.991618648507072, + "grad_norm": 0.49572464239368597, + "learning_rate": 1.8595720820850214e-05, + "loss": 0.1829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0809839516878128, + "step": 7620, + "valid_targets_mean": 2495.0, + "valid_targets_min": 605 + }, + { + "epoch": 3.994237820848612, + "grad_norm": 0.32130503344509675, + "learning_rate": 1.8569663161537665e-05, + "loss": 0.0779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03973918780684471, + "step": 7625, + "valid_targets_mean": 3298.6, + "valid_targets_min": 985 + }, + { + "epoch": 3.9968569931901516, + "grad_norm": 0.38638293193878603, + "learning_rate": 1.854360794248241e-05, + "loss": 0.0903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044368430972099304, + "step": 7630, + "valid_targets_mean": 3505.2, + "valid_targets_min": 878 + }, + { + "epoch": 3.999476165531692, + "grad_norm": 0.5361200585789292, + "learning_rate": 1.8517555208136508e-05, + "loss": 0.1111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11933934688568115, + "step": 7635, + "valid_targets_mean": 2289.5, + "valid_targets_min": 1214 + }, + { + "epoch": 4.002095337873232, + "grad_norm": 0.5046993117304381, + "learning_rate": 1.8491505002947784e-05, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09715577214956284, + "step": 7640, + "valid_targets_mean": 6794.6, + "valid_targets_min": 4960 + }, + { + "epoch": 4.0047145102147725, + "grad_norm": 0.35678226300518445, + "learning_rate": 1.846545737135978e-05, + "loss": 0.1658, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08897572755813599, + "step": 7645, + "valid_targets_mean": 7524.9, + "valid_targets_min": 4541 + }, + { + "epoch": 4.007333682556312, + "grad_norm": 0.3861077669799444, + "learning_rate": 1.8439412357811607e-05, + "loss": 0.159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08123362809419632, + "step": 7650, + "valid_targets_mean": 5443.9, + "valid_targets_min": 4245 + }, + { + "epoch": 4.009952854897852, + "grad_norm": 0.3775283336118568, + "learning_rate": 1.8413370006737934e-05, + "loss": 0.1596, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08923494815826416, + "step": 7655, + "valid_targets_mean": 7527.2, + "valid_targets_min": 5149 + }, + { + "epoch": 4.0125720272393925, + "grad_norm": 0.35977041198566295, + "learning_rate": 1.838733036256888e-05, + "loss": 0.1667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0835188627243042, + "step": 7660, + "valid_targets_mean": 7006.4, + "valid_targets_min": 4892 + }, + { + "epoch": 4.015191199580933, + "grad_norm": 0.4100107439671739, + "learning_rate": 1.836129346972995e-05, + "loss": 0.1671, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0873408168554306, + "step": 7665, + "valid_targets_mean": 6395.8, + "valid_targets_min": 4741 + }, + { + "epoch": 4.017810371922472, + "grad_norm": 0.36796112964425054, + "learning_rate": 1.8335259372641945e-05, + "loss": 0.1696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07623052597045898, + "step": 7670, + "valid_targets_mean": 6463.6, + "valid_targets_min": 4236 + }, + { + "epoch": 4.0204295442640126, + "grad_norm": 0.39553165460460893, + "learning_rate": 1.8309228115720907e-05, + "loss": 0.1638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09139275550842285, + "step": 7675, + "valid_targets_mean": 6933.2, + "valid_targets_min": 5516 + }, + { + "epoch": 4.023048716605553, + "grad_norm": 0.35164961313014875, + "learning_rate": 1.8283199743378028e-05, + "loss": 0.1797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08002959191799164, + "step": 7680, + "valid_targets_mean": 7346.1, + "valid_targets_min": 4970 + }, + { + "epoch": 4.025667888947093, + "grad_norm": 0.3553261312378511, + "learning_rate": 1.8257174300019565e-05, + "loss": 0.1586, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06670039892196655, + "step": 7685, + "valid_targets_mean": 6472.9, + "valid_targets_min": 4552 + }, + { + "epoch": 4.028287061288633, + "grad_norm": 0.4693893334940823, + "learning_rate": 1.8231151830046818e-05, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07813999056816101, + "step": 7690, + "valid_targets_mean": 7325.1, + "valid_targets_min": 5398 + }, + { + "epoch": 4.030906233630173, + "grad_norm": 0.37213628380758396, + "learning_rate": 1.8205132377855974e-05, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08336234837770462, + "step": 7695, + "valid_targets_mean": 7650.2, + "valid_targets_min": 5696 + }, + { + "epoch": 4.033525405971713, + "grad_norm": 0.363102284615978, + "learning_rate": 1.8179115987838078e-05, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07469281554222107, + "step": 7700, + "valid_targets_mean": 5666.0, + "valid_targets_min": 4515 + }, + { + "epoch": 4.036144578313253, + "grad_norm": 0.35073208721536664, + "learning_rate": 1.8153102704378975e-05, + "loss": 0.1476, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06807784736156464, + "step": 7705, + "valid_targets_mean": 5900.4, + "valid_targets_min": 4084 + }, + { + "epoch": 4.038763750654793, + "grad_norm": 0.3898715278076427, + "learning_rate": 1.8127092571859173e-05, + "loss": 0.1489, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08689968287944794, + "step": 7710, + "valid_targets_mean": 7969.1, + "valid_targets_min": 4672 + }, + { + "epoch": 4.041382922996333, + "grad_norm": 0.36385328461769995, + "learning_rate": 1.8101085634653846e-05, + "loss": 0.1649, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08636879920959473, + "step": 7715, + "valid_targets_mean": 7626.1, + "valid_targets_min": 6281 + }, + { + "epoch": 4.0440020953378735, + "grad_norm": 0.3333481201545469, + "learning_rate": 1.8075081937132676e-05, + "loss": 0.1585, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059382036328315735, + "step": 7720, + "valid_targets_mean": 6799.5, + "valid_targets_min": 4384 + }, + { + "epoch": 4.046621267679413, + "grad_norm": 0.35198912685964345, + "learning_rate": 1.8049081523659855e-05, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07585656642913818, + "step": 7725, + "valid_targets_mean": 6568.0, + "valid_targets_min": 4614 + }, + { + "epoch": 4.049240440020953, + "grad_norm": 0.36003043597604506, + "learning_rate": 1.8023084438593945e-05, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0751047134399414, + "step": 7730, + "valid_targets_mean": 6748.0, + "valid_targets_min": 5386 + }, + { + "epoch": 4.0518596123624935, + "grad_norm": 0.40188756309665574, + "learning_rate": 1.7997090726287828e-05, + "loss": 0.1548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08476327359676361, + "step": 7735, + "valid_targets_mean": 6902.2, + "valid_targets_min": 5685 + }, + { + "epoch": 4.054478784704034, + "grad_norm": 0.3492151882018201, + "learning_rate": 1.797110043108866e-05, + "loss": 0.1735, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09153053164482117, + "step": 7740, + "valid_targets_mean": 7076.2, + "valid_targets_min": 4898 + }, + { + "epoch": 4.057097957045573, + "grad_norm": 0.41184439116480687, + "learning_rate": 1.794511359733774e-05, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07922191917896271, + "step": 7745, + "valid_targets_mean": 4991.8, + "valid_targets_min": 3834 + }, + { + "epoch": 4.0597171293871135, + "grad_norm": 0.42697565451189406, + "learning_rate": 1.791913026937048e-05, + "loss": 0.1576, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09586077183485031, + "step": 7750, + "valid_targets_mean": 6576.6, + "valid_targets_min": 4432 + }, + { + "epoch": 4.062336301728654, + "grad_norm": 0.44547692156935037, + "learning_rate": 1.789315049151629e-05, + "loss": 0.1686, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07448678463697433, + "step": 7755, + "valid_targets_mean": 6308.6, + "valid_targets_min": 4371 + }, + { + "epoch": 4.064955474070194, + "grad_norm": 0.3603777540343446, + "learning_rate": 1.7867174308098537e-05, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08962307125329971, + "step": 7760, + "valid_targets_mean": 6699.1, + "valid_targets_min": 4770 + }, + { + "epoch": 4.0675746464117335, + "grad_norm": 0.36704542678303614, + "learning_rate": 1.7841201763434465e-05, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08683955669403076, + "step": 7765, + "valid_targets_mean": 6969.2, + "valid_targets_min": 5190 + }, + { + "epoch": 4.070193818753274, + "grad_norm": 0.37014460687358286, + "learning_rate": 1.781523290183508e-05, + "loss": 0.1601, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07665061950683594, + "step": 7770, + "valid_targets_mean": 6307.1, + "valid_targets_min": 5647 + }, + { + "epoch": 4.072812991094814, + "grad_norm": 0.36301629553729164, + "learning_rate": 1.7789267767605145e-05, + "loss": 0.1648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08621418476104736, + "step": 7775, + "valid_targets_mean": 6924.8, + "valid_targets_min": 4398 + }, + { + "epoch": 4.075432163436354, + "grad_norm": 0.40666680352945367, + "learning_rate": 1.7763306405043027e-05, + "loss": 0.1619, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08380205184221268, + "step": 7780, + "valid_targets_mean": 5753.4, + "valid_targets_min": 3814 + }, + { + "epoch": 4.078051335777894, + "grad_norm": 0.4092511337486787, + "learning_rate": 1.773734885844068e-05, + "loss": 0.1714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08741191774606705, + "step": 7785, + "valid_targets_mean": 6429.1, + "valid_targets_min": 5029 + }, + { + "epoch": 4.080670508119434, + "grad_norm": 0.39789126185891305, + "learning_rate": 1.7711395172083546e-05, + "loss": 0.1599, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08730865269899368, + "step": 7790, + "valid_targets_mean": 6615.4, + "valid_targets_min": 5042 + }, + { + "epoch": 4.083289680460974, + "grad_norm": 0.39111331795201215, + "learning_rate": 1.768544539025046e-05, + "loss": 0.1707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08760306984186172, + "step": 7795, + "valid_targets_mean": 7438.6, + "valid_targets_min": 4625 + }, + { + "epoch": 4.085908852802515, + "grad_norm": 0.39295508548901675, + "learning_rate": 1.7659499557213624e-05, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07737244665622711, + "step": 7800, + "valid_targets_mean": 5983.5, + "valid_targets_min": 4820 + }, + { + "epoch": 4.088528025144054, + "grad_norm": 0.3590028884013462, + "learning_rate": 1.763355771723849e-05, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06421305984258652, + "step": 7805, + "valid_targets_mean": 5466.4, + "valid_targets_min": 4337 + }, + { + "epoch": 4.091147197485594, + "grad_norm": 0.33147411082404826, + "learning_rate": 1.76076199145837e-05, + "loss": 0.1413, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07399965077638626, + "step": 7810, + "valid_targets_mean": 7386.4, + "valid_targets_min": 3794 + }, + { + "epoch": 4.093766369827135, + "grad_norm": 0.39650430718364593, + "learning_rate": 1.7581686193500997e-05, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07049068808555603, + "step": 7815, + "valid_targets_mean": 6136.9, + "valid_targets_min": 5697 + }, + { + "epoch": 4.096385542168675, + "grad_norm": 0.41449179428693694, + "learning_rate": 1.7555756598235178e-05, + "loss": 0.1746, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08700234442949295, + "step": 7820, + "valid_targets_mean": 5969.4, + "valid_targets_min": 4632 + }, + { + "epoch": 4.099004714510214, + "grad_norm": 0.36826644630894345, + "learning_rate": 1.7529831173023994e-05, + "loss": 0.1664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08240143954753876, + "step": 7825, + "valid_targets_mean": 6647.2, + "valid_targets_min": 4713 + }, + { + "epoch": 4.101623886851755, + "grad_norm": 0.3658315689336967, + "learning_rate": 1.7503909962098075e-05, + "loss": 0.1567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0915922299027443, + "step": 7830, + "valid_targets_mean": 6687.8, + "valid_targets_min": 5103 + }, + { + "epoch": 4.104243059193295, + "grad_norm": 0.3552165438856329, + "learning_rate": 1.7477993009680874e-05, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07887688279151917, + "step": 7835, + "valid_targets_mean": 6908.6, + "valid_targets_min": 3450 + }, + { + "epoch": 4.106862231534835, + "grad_norm": 0.5763615782651337, + "learning_rate": 1.745208035998857e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07748524844646454, + "step": 7840, + "valid_targets_mean": 6493.6, + "valid_targets_min": 4549 + }, + { + "epoch": 4.109481403876375, + "grad_norm": 0.3957220479027356, + "learning_rate": 1.7426172057229987e-05, + "loss": 0.1673, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07997159659862518, + "step": 7845, + "valid_targets_mean": 6096.6, + "valid_targets_min": 4838 + }, + { + "epoch": 4.112100576217915, + "grad_norm": 0.3824992868016784, + "learning_rate": 1.7400268145606576e-05, + "loss": 0.154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08249804377555847, + "step": 7850, + "valid_targets_mean": 6304.5, + "valid_targets_min": 4588 + }, + { + "epoch": 4.114719748559455, + "grad_norm": 0.4444588523690061, + "learning_rate": 1.7374368669312245e-05, + "loss": 0.1592, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08841138333082199, + "step": 7855, + "valid_targets_mean": 5490.1, + "valid_targets_min": 3764 + }, + { + "epoch": 4.117338920900996, + "grad_norm": 0.46062071352679934, + "learning_rate": 1.7348473672533377e-05, + "loss": 0.1682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0907268226146698, + "step": 7860, + "valid_targets_mean": 6737.8, + "valid_targets_min": 4923 + }, + { + "epoch": 4.119958093242535, + "grad_norm": 0.34978360201953385, + "learning_rate": 1.732258319944868e-05, + "loss": 0.1386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07136829942464828, + "step": 7865, + "valid_targets_mean": 6511.5, + "valid_targets_min": 4112 + }, + { + "epoch": 4.122577265584075, + "grad_norm": 0.3491043816183656, + "learning_rate": 1.729669729422917e-05, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07058646529912949, + "step": 7870, + "valid_targets_mean": 5693.1, + "valid_targets_min": 3761 + }, + { + "epoch": 4.125196437925616, + "grad_norm": 0.3681642809514946, + "learning_rate": 1.727081600103806e-05, + "loss": 0.1482, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06867656856775284, + "step": 7875, + "valid_targets_mean": 6110.8, + "valid_targets_min": 4928 + }, + { + "epoch": 4.127815610267156, + "grad_norm": 0.41791522677397996, + "learning_rate": 1.7244939364030676e-05, + "loss": 0.1539, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09475985169410706, + "step": 7880, + "valid_targets_mean": 6740.8, + "valid_targets_min": 4874 + }, + { + "epoch": 4.130434782608695, + "grad_norm": 0.4773839620240532, + "learning_rate": 1.7219067427354436e-05, + "loss": 0.1726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10236172378063202, + "step": 7885, + "valid_targets_mean": 7385.4, + "valid_targets_min": 4989 + }, + { + "epoch": 4.133053954950236, + "grad_norm": 0.4227949051784635, + "learning_rate": 1.7193200235148706e-05, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08922139555215836, + "step": 7890, + "valid_targets_mean": 6639.4, + "valid_targets_min": 4534 + }, + { + "epoch": 4.135673127291776, + "grad_norm": 0.37090123303531725, + "learning_rate": 1.716733783154479e-05, + "loss": 0.1564, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06595587730407715, + "step": 7895, + "valid_targets_mean": 5464.0, + "valid_targets_min": 4773 + }, + { + "epoch": 4.138292299633316, + "grad_norm": 0.3809779760694723, + "learning_rate": 1.7141480260665794e-05, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08241276443004608, + "step": 7900, + "valid_targets_mean": 5868.6, + "valid_targets_min": 4597 + }, + { + "epoch": 4.140911471974856, + "grad_norm": 0.36180796258677017, + "learning_rate": 1.711562756662658e-05, + "loss": 0.1567, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07521642744541168, + "step": 7905, + "valid_targets_mean": 6699.6, + "valid_targets_min": 1046 + }, + { + "epoch": 4.143530644316396, + "grad_norm": 0.3579165961212639, + "learning_rate": 1.7089779793533716e-05, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08528405427932739, + "step": 7910, + "valid_targets_mean": 6446.4, + "valid_targets_min": 4240 + }, + { + "epoch": 4.146149816657936, + "grad_norm": 0.3562636456324392, + "learning_rate": 1.706393698548534e-05, + "loss": 0.1644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0732632726430893, + "step": 7915, + "valid_targets_mean": 6580.9, + "valid_targets_min": 4489 + }, + { + "epoch": 4.1487689889994765, + "grad_norm": 0.38934628179457265, + "learning_rate": 1.703809918657116e-05, + "loss": 0.1646, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0822799801826477, + "step": 7920, + "valid_targets_mean": 5510.4, + "valid_targets_min": 4865 + }, + { + "epoch": 4.151388161341016, + "grad_norm": 0.38039942134419014, + "learning_rate": 1.7012266440872294e-05, + "loss": 0.1644, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06634122878313065, + "step": 7925, + "valid_targets_mean": 5621.1, + "valid_targets_min": 4452 + }, + { + "epoch": 4.154007333682556, + "grad_norm": 0.3881829840056501, + "learning_rate": 1.6986438792461272e-05, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07326548546552658, + "step": 7930, + "valid_targets_mean": 5921.0, + "valid_targets_min": 3918 + }, + { + "epoch": 4.156626506024097, + "grad_norm": 0.35486454495371633, + "learning_rate": 1.696061628540191e-05, + "loss": 0.1589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06416025012731552, + "step": 7935, + "valid_targets_mean": 5588.9, + "valid_targets_min": 3727 + }, + { + "epoch": 4.159245678365637, + "grad_norm": 0.3625989559195439, + "learning_rate": 1.6934798963749256e-05, + "loss": 0.2026, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07746327668428421, + "step": 7940, + "valid_targets_mean": 6208.5, + "valid_targets_min": 4482 + }, + { + "epoch": 4.161864850707176, + "grad_norm": 0.3545423001409932, + "learning_rate": 1.6908986871549525e-05, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059910908341407776, + "step": 7945, + "valid_targets_mean": 5420.5, + "valid_targets_min": 3362 + }, + { + "epoch": 4.164484023048717, + "grad_norm": 0.39227559870493695, + "learning_rate": 1.6883180052839975e-05, + "loss": 0.1625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06310118734836578, + "step": 7950, + "valid_targets_mean": 5243.5, + "valid_targets_min": 4395 + }, + { + "epoch": 4.167103195390257, + "grad_norm": 0.41319003014361494, + "learning_rate": 1.6857378551648918e-05, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09356322884559631, + "step": 7955, + "valid_targets_mean": 5888.6, + "valid_targets_min": 4135 + }, + { + "epoch": 4.169722367731797, + "grad_norm": 0.35460062015713045, + "learning_rate": 1.6831582411995556e-05, + "loss": 0.1555, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06868542730808258, + "step": 7960, + "valid_targets_mean": 6252.8, + "valid_targets_min": 4262 + }, + { + "epoch": 4.172341540073337, + "grad_norm": 0.3624071077925172, + "learning_rate": 1.680579167788994e-05, + "loss": 0.1439, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0699419304728508, + "step": 7965, + "valid_targets_mean": 6526.4, + "valid_targets_min": 4785 + }, + { + "epoch": 4.174960712414877, + "grad_norm": 0.38272783208930233, + "learning_rate": 1.6780006393332934e-05, + "loss": 0.1524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062110498547554016, + "step": 7970, + "valid_targets_mean": 5490.5, + "valid_targets_min": 4229 + }, + { + "epoch": 4.177579884756417, + "grad_norm": 0.4541688243924822, + "learning_rate": 1.675422660231607e-05, + "loss": 0.1568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08226379007101059, + "step": 7975, + "valid_targets_mean": 7176.1, + "valid_targets_min": 5110 + }, + { + "epoch": 4.1801990570979575, + "grad_norm": 0.375908420893291, + "learning_rate": 1.672845234882153e-05, + "loss": 0.1491, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09539134800434113, + "step": 7980, + "valid_targets_mean": 7298.8, + "valid_targets_min": 4822 + }, + { + "epoch": 4.182818229439497, + "grad_norm": 0.3512697655421249, + "learning_rate": 1.6702683676822028e-05, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0776122510433197, + "step": 7985, + "valid_targets_mean": 7062.1, + "valid_targets_min": 4987 + }, + { + "epoch": 4.185437401781037, + "grad_norm": 0.3548047999068206, + "learning_rate": 1.6676920630280785e-05, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07747262716293335, + "step": 7990, + "valid_targets_mean": 6830.5, + "valid_targets_min": 4726 + }, + { + "epoch": 4.1880565741225775, + "grad_norm": 0.3819430207008536, + "learning_rate": 1.66511632531514e-05, + "loss": 0.1417, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07341688871383667, + "step": 7995, + "valid_targets_mean": 5633.0, + "valid_targets_min": 4536 + }, + { + "epoch": 4.190675746464118, + "grad_norm": 0.40392595392851477, + "learning_rate": 1.6625411589377795e-05, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07395326346158981, + "step": 8000, + "valid_targets_mean": 6318.0, + "valid_targets_min": 5070 + }, + { + "epoch": 4.193294918805657, + "grad_norm": 0.4441317380842661, + "learning_rate": 1.659966568289417e-05, + "loss": 0.1626, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09643901884555817, + "step": 8005, + "valid_targets_mean": 6617.1, + "valid_targets_min": 3774 + }, + { + "epoch": 4.1959140911471975, + "grad_norm": 0.8914091416833568, + "learning_rate": 1.6573925577624885e-05, + "loss": 0.1662, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08490767329931259, + "step": 8010, + "valid_targets_mean": 1381.4, + "valid_targets_min": 864 + }, + { + "epoch": 4.198533263488738, + "grad_norm": 0.6879974649926835, + "learning_rate": 1.6548191317484405e-05, + "loss": 0.1695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08852212876081467, + "step": 8015, + "valid_targets_mean": 1637.2, + "valid_targets_min": 956 + }, + { + "epoch": 4.201152435830277, + "grad_norm": 0.791605871787273, + "learning_rate": 1.6522462946377233e-05, + "loss": 0.1513, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08068401366472244, + "step": 8020, + "valid_targets_mean": 1660.4, + "valid_targets_min": 875 + }, + { + "epoch": 4.2037716081718175, + "grad_norm": 0.9285054914740514, + "learning_rate": 1.6496740508197797e-05, + "loss": 0.1557, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08627966046333313, + "step": 8025, + "valid_targets_mean": 1455.4, + "valid_targets_min": 750 + }, + { + "epoch": 4.206390780513358, + "grad_norm": 0.7954167187943217, + "learning_rate": 1.647102404683044e-05, + "loss": 0.1471, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09543976187705994, + "step": 8030, + "valid_targets_mean": 2055.0, + "valid_targets_min": 898 + }, + { + "epoch": 4.209009952854898, + "grad_norm": 0.7350704570236828, + "learning_rate": 1.6445313606149278e-05, + "loss": 0.1506, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07267652451992035, + "step": 8035, + "valid_targets_mean": 1432.1, + "valid_targets_min": 1070 + }, + { + "epoch": 4.2116291251964375, + "grad_norm": 0.7022969627528745, + "learning_rate": 1.6419609230018183e-05, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07557923346757889, + "step": 8040, + "valid_targets_mean": 1676.1, + "valid_targets_min": 970 + }, + { + "epoch": 4.214248297537978, + "grad_norm": 0.7331649096655076, + "learning_rate": 1.639391096229065e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06656152009963989, + "step": 8045, + "valid_targets_mean": 1273.8, + "valid_targets_min": 648 + }, + { + "epoch": 4.216867469879518, + "grad_norm": 0.756637184055864, + "learning_rate": 1.6368218846809777e-05, + "loss": 0.1584, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08959043025970459, + "step": 8050, + "valid_targets_mean": 1795.0, + "valid_targets_min": 1204 + }, + { + "epoch": 4.219486642221058, + "grad_norm": 0.7007644790954075, + "learning_rate": 1.6342532927408153e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055157940834760666, + "step": 8055, + "valid_targets_mean": 1266.9, + "valid_targets_min": 770 + }, + { + "epoch": 4.222105814562598, + "grad_norm": 0.8192471344982334, + "learning_rate": 1.63168532479078e-05, + "loss": 0.1499, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05294179171323776, + "step": 8060, + "valid_targets_mean": 1062.0, + "valid_targets_min": 572 + }, + { + "epoch": 4.224724986904138, + "grad_norm": 0.7339945402391059, + "learning_rate": 1.6291179852120095e-05, + "loss": 0.1522, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0885145291686058, + "step": 8065, + "valid_targets_mean": 1749.0, + "valid_targets_min": 712 + }, + { + "epoch": 4.227344159245678, + "grad_norm": 0.7493953881579103, + "learning_rate": 1.626551278384569e-05, + "loss": 0.1475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0800328403711319, + "step": 8070, + "valid_targets_mean": 1578.0, + "valid_targets_min": 923 + }, + { + "epoch": 4.229963331587219, + "grad_norm": 0.7660237210220724, + "learning_rate": 1.623985208687445e-05, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06861114501953125, + "step": 8075, + "valid_targets_mean": 1515.8, + "valid_targets_min": 842 + }, + { + "epoch": 4.232582503928758, + "grad_norm": 0.7526630191864675, + "learning_rate": 1.6214197804985364e-05, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06550689041614532, + "step": 8080, + "valid_targets_mean": 1479.9, + "valid_targets_min": 822 + }, + { + "epoch": 4.235201676270298, + "grad_norm": 0.726276514595483, + "learning_rate": 1.6188549981946475e-05, + "loss": 0.1407, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09295983612537384, + "step": 8085, + "valid_targets_mean": 1839.1, + "valid_targets_min": 1042 + }, + { + "epoch": 4.237820848611839, + "grad_norm": 0.7257518985241499, + "learning_rate": 1.616290866151481e-05, + "loss": 0.1469, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0574457123875618, + "step": 8090, + "valid_targets_mean": 1152.8, + "valid_targets_min": 870 + }, + { + "epoch": 4.240440020953379, + "grad_norm": 0.7600775709380813, + "learning_rate": 1.6137273887436302e-05, + "loss": 0.1503, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07445642352104187, + "step": 8095, + "valid_targets_mean": 1510.9, + "valid_targets_min": 740 + }, + { + "epoch": 4.243059193294918, + "grad_norm": 0.6576176366920666, + "learning_rate": 1.611164570344572e-05, + "loss": 0.1481, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061112623661756516, + "step": 8100, + "valid_targets_mean": 1350.0, + "valid_targets_min": 681 + }, + { + "epoch": 4.245678365636459, + "grad_norm": 0.7243345461069208, + "learning_rate": 1.6086024153266574e-05, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08239995688199997, + "step": 8105, + "valid_targets_mean": 1824.9, + "valid_targets_min": 976 + }, + { + "epoch": 4.248297537977999, + "grad_norm": 0.6761318842858626, + "learning_rate": 1.6060409280611083e-05, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06249365210533142, + "step": 8110, + "valid_targets_mean": 1590.8, + "valid_targets_min": 832 + }, + { + "epoch": 4.250916710319539, + "grad_norm": 0.6916134047219931, + "learning_rate": 1.6034801129180047e-05, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06040395796298981, + "step": 8115, + "valid_targets_mean": 1461.9, + "valid_targets_min": 891 + }, + { + "epoch": 4.253535882661079, + "grad_norm": 0.8089997687505759, + "learning_rate": 1.6009199742662804e-05, + "loss": 0.1455, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07450135052204132, + "step": 8120, + "valid_targets_mean": 1290.0, + "valid_targets_min": 734 + }, + { + "epoch": 4.256155055002619, + "grad_norm": 0.733226386131972, + "learning_rate": 1.5983605164737172e-05, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0565100722014904, + "step": 8125, + "valid_targets_mean": 1226.6, + "valid_targets_min": 704 + }, + { + "epoch": 4.258774227344159, + "grad_norm": 0.8463668051931974, + "learning_rate": 1.595801743906932e-05, + "loss": 0.143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07282941043376923, + "step": 8130, + "valid_targets_mean": 1603.0, + "valid_targets_min": 872 + }, + { + "epoch": 4.2613933996857, + "grad_norm": 0.7740810392564101, + "learning_rate": 1.5932436609313762e-05, + "loss": 0.1371, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06875354051589966, + "step": 8135, + "valid_targets_mean": 1451.5, + "valid_targets_min": 805 + }, + { + "epoch": 4.264012572027239, + "grad_norm": 0.7141212075884272, + "learning_rate": 1.5906862719113223e-05, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07992750406265259, + "step": 8140, + "valid_targets_mean": 2075.8, + "valid_targets_min": 999 + }, + { + "epoch": 4.266631744368779, + "grad_norm": 0.7044187142564264, + "learning_rate": 1.5881295812098584e-05, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0789269506931305, + "step": 8145, + "valid_targets_mean": 2077.5, + "valid_targets_min": 750 + }, + { + "epoch": 4.26925091671032, + "grad_norm": 0.6953707280610457, + "learning_rate": 1.5855735931888837e-05, + "loss": 0.1397, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05323594808578491, + "step": 8150, + "valid_targets_mean": 1137.2, + "valid_targets_min": 613 + }, + { + "epoch": 4.27187008905186, + "grad_norm": 0.7227839694518329, + "learning_rate": 1.583018312209096e-05, + "loss": 0.1429, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08068192005157471, + "step": 8155, + "valid_targets_mean": 1883.4, + "valid_targets_min": 1184 + }, + { + "epoch": 4.274489261393399, + "grad_norm": 0.7474081095760743, + "learning_rate": 1.5804637426299894e-05, + "loss": 0.138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07572872936725616, + "step": 8160, + "valid_targets_mean": 1535.0, + "valid_targets_min": 758 + }, + { + "epoch": 4.27710843373494, + "grad_norm": 0.7463737788737524, + "learning_rate": 1.5779098888098413e-05, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0545760802924633, + "step": 8165, + "valid_targets_mean": 1206.6, + "valid_targets_min": 800 + }, + { + "epoch": 4.27972760607648, + "grad_norm": 0.7188725614276276, + "learning_rate": 1.5753567551057112e-05, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07694423943758011, + "step": 8170, + "valid_targets_mean": 1754.8, + "valid_targets_min": 933 + }, + { + "epoch": 4.28234677841802, + "grad_norm": 0.8020355022579448, + "learning_rate": 1.5728043458734278e-05, + "loss": 0.1372, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06083884835243225, + "step": 8175, + "valid_targets_mean": 1073.6, + "valid_targets_min": 746 + }, + { + "epoch": 4.28496595075956, + "grad_norm": 0.7541823559114252, + "learning_rate": 1.570252665467583e-05, + "loss": 0.1311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07423181086778641, + "step": 8180, + "valid_targets_mean": 1458.6, + "valid_targets_min": 1106 + }, + { + "epoch": 4.2875851231011, + "grad_norm": 0.8554473415702296, + "learning_rate": 1.567701718241529e-05, + "loss": 0.1403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07971362769603729, + "step": 8185, + "valid_targets_mean": 1646.9, + "valid_targets_min": 635 + }, + { + "epoch": 4.29020429544264, + "grad_norm": 0.7563970805069881, + "learning_rate": 1.565151508547363e-05, + "loss": 0.1369, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05597660690546036, + "step": 8190, + "valid_targets_mean": 1174.4, + "valid_targets_min": 739 + }, + { + "epoch": 4.292823467784181, + "grad_norm": 0.7452191252960292, + "learning_rate": 1.5626020407359266e-05, + "loss": 0.1398, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08104661107063293, + "step": 8195, + "valid_targets_mean": 1692.8, + "valid_targets_min": 831 + }, + { + "epoch": 4.29544264012572, + "grad_norm": 0.7435208818142135, + "learning_rate": 1.560053319156795e-05, + "loss": 0.1386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07135645300149918, + "step": 8200, + "valid_targets_mean": 1495.0, + "valid_targets_min": 722 + }, + { + "epoch": 4.29806181246726, + "grad_norm": 0.7229346807685144, + "learning_rate": 1.5575053481582694e-05, + "loss": 0.1394, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0740927904844284, + "step": 8205, + "valid_targets_mean": 1382.2, + "valid_targets_min": 711 + }, + { + "epoch": 4.300680984808801, + "grad_norm": 0.7213323736904493, + "learning_rate": 1.5549581320873715e-05, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044076889753341675, + "step": 8210, + "valid_targets_mean": 1137.9, + "valid_targets_min": 733 + }, + { + "epoch": 4.303300157150341, + "grad_norm": 0.7819325099247775, + "learning_rate": 1.5524116752898347e-05, + "loss": 0.1404, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07010834664106369, + "step": 8215, + "valid_targets_mean": 1539.8, + "valid_targets_min": 953 + }, + { + "epoch": 4.30591932949188, + "grad_norm": 0.7215041140333152, + "learning_rate": 1.5498659821100975e-05, + "loss": 0.1458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06586821377277374, + "step": 8220, + "valid_targets_mean": 1616.5, + "valid_targets_min": 1011 + }, + { + "epoch": 4.308538501833421, + "grad_norm": 0.8095159631567369, + "learning_rate": 1.5473210568912935e-05, + "loss": 0.1505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08267545700073242, + "step": 8225, + "valid_targets_mean": 1566.6, + "valid_targets_min": 996 + }, + { + "epoch": 4.311157674174961, + "grad_norm": 0.7608794339845889, + "learning_rate": 1.54477690397525e-05, + "loss": 0.1336, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059193000197410583, + "step": 8230, + "valid_targets_mean": 1453.4, + "valid_targets_min": 699 + }, + { + "epoch": 4.313776846516501, + "grad_norm": 0.7318760445934105, + "learning_rate": 1.5422335277024733e-05, + "loss": 0.1312, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07046680897474289, + "step": 8235, + "valid_targets_mean": 1445.5, + "valid_targets_min": 957 + }, + { + "epoch": 4.316396018858041, + "grad_norm": 0.7961872321007037, + "learning_rate": 1.5396909324121446e-05, + "loss": 0.1453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07809412479400635, + "step": 8240, + "valid_targets_mean": 1587.1, + "valid_targets_min": 787 + }, + { + "epoch": 4.319015191199581, + "grad_norm": 0.7538996368078313, + "learning_rate": 1.5371491224421165e-05, + "loss": 0.1344, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043929800391197205, + "step": 8245, + "valid_targets_mean": 1065.2, + "valid_targets_min": 549 + }, + { + "epoch": 4.321634363541121, + "grad_norm": 0.740589928884431, + "learning_rate": 1.5346081021288967e-05, + "loss": 0.1298, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055525489151477814, + "step": 8250, + "valid_targets_mean": 1241.8, + "valid_targets_min": 725 + }, + { + "epoch": 4.3242535358826615, + "grad_norm": 0.756913325671026, + "learning_rate": 1.5320678758076504e-05, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07678687572479248, + "step": 8255, + "valid_targets_mean": 1640.1, + "valid_targets_min": 852 + }, + { + "epoch": 4.326872708224201, + "grad_norm": 0.7359798464998011, + "learning_rate": 1.5295284478121847e-05, + "loss": 0.1405, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06366491317749023, + "step": 8260, + "valid_targets_mean": 1439.6, + "valid_targets_min": 1010 + }, + { + "epoch": 4.329491880565741, + "grad_norm": 1.135048786385237, + "learning_rate": 1.5269898224749454e-05, + "loss": 0.1392, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06554259359836578, + "step": 8265, + "valid_targets_mean": 1374.6, + "valid_targets_min": 545 + }, + { + "epoch": 4.3321110529072815, + "grad_norm": 0.7369299584504874, + "learning_rate": 1.5244520041270114e-05, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07120509445667267, + "step": 8270, + "valid_targets_mean": 1519.4, + "valid_targets_min": 1136 + }, + { + "epoch": 4.334730225248821, + "grad_norm": 0.6480445838801647, + "learning_rate": 1.5219149970980809e-05, + "loss": 0.1346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08611080050468445, + "step": 8275, + "valid_targets_mean": 1715.6, + "valid_targets_min": 849 + }, + { + "epoch": 4.337349397590361, + "grad_norm": 0.7497983537727215, + "learning_rate": 1.5193788057164716e-05, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057314857840538025, + "step": 8280, + "valid_targets_mean": 1272.1, + "valid_targets_min": 789 + }, + { + "epoch": 4.3399685699319015, + "grad_norm": 0.7585584784172799, + "learning_rate": 1.5168434343091061e-05, + "loss": 0.137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061569757759571075, + "step": 8285, + "valid_targets_mean": 1362.1, + "valid_targets_min": 673 + }, + { + "epoch": 4.342587742273442, + "grad_norm": 0.8604649477178995, + "learning_rate": 1.5143088872015121e-05, + "loss": 0.1364, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06527075171470642, + "step": 8290, + "valid_targets_mean": 1213.4, + "valid_targets_min": 842 + }, + { + "epoch": 4.345206914614982, + "grad_norm": 0.6933227142138495, + "learning_rate": 1.511775168717808e-05, + "loss": 0.1317, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05573000758886337, + "step": 8295, + "valid_targets_mean": 1335.8, + "valid_targets_min": 672 + }, + { + "epoch": 4.3478260869565215, + "grad_norm": 0.6811881961358933, + "learning_rate": 1.5092422831806981e-05, + "loss": 0.1253, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05557628720998764, + "step": 8300, + "valid_targets_mean": 1314.4, + "valid_targets_min": 727 + }, + { + "epoch": 4.350445259298062, + "grad_norm": 0.7337898539081581, + "learning_rate": 1.5067102349114686e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06094259396195412, + "step": 8305, + "valid_targets_mean": 1436.2, + "valid_targets_min": 647 + }, + { + "epoch": 4.353064431639602, + "grad_norm": 0.7983289132034042, + "learning_rate": 1.5041790282299742e-05, + "loss": 0.1388, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.059410758316516876, + "step": 8310, + "valid_targets_mean": 1253.5, + "valid_targets_min": 613 + }, + { + "epoch": 4.3556836039811415, + "grad_norm": 0.7938882651701654, + "learning_rate": 1.5016486674546367e-05, + "loss": 0.1352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06341471523046494, + "step": 8315, + "valid_targets_mean": 1290.0, + "valid_targets_min": 746 + }, + { + "epoch": 4.358302776322682, + "grad_norm": 0.7415994509082007, + "learning_rate": 1.499119156902432e-05, + "loss": 0.1352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06149760261178017, + "step": 8320, + "valid_targets_mean": 1586.2, + "valid_targets_min": 892 + }, + { + "epoch": 4.360921948664222, + "grad_norm": 0.7526385843246455, + "learning_rate": 1.4965905008888864e-05, + "loss": 0.1414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07845599949359894, + "step": 8325, + "valid_targets_mean": 1519.9, + "valid_targets_min": 914 + }, + { + "epoch": 4.363541121005762, + "grad_norm": 0.7490275954127474, + "learning_rate": 1.4940627037280693e-05, + "loss": 0.1358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061721738427877426, + "step": 8330, + "valid_targets_mean": 1562.8, + "valid_targets_min": 636 + }, + { + "epoch": 4.366160293347303, + "grad_norm": 0.7910846535308033, + "learning_rate": 1.4915357697325834e-05, + "loss": 0.1343, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07377348840236664, + "step": 8335, + "valid_targets_mean": 1533.5, + "valid_targets_min": 1045 + }, + { + "epoch": 4.368779465688842, + "grad_norm": 0.8531621779926405, + "learning_rate": 1.48900970321356e-05, + "loss": 0.1396, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07574078440666199, + "step": 8340, + "valid_targets_mean": 1525.6, + "valid_targets_min": 1008 + }, + { + "epoch": 4.371398638030382, + "grad_norm": 0.8103605748533639, + "learning_rate": 1.4864845084806492e-05, + "loss": 0.1332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06176677346229553, + "step": 8345, + "valid_targets_mean": 1285.1, + "valid_targets_min": 729 + }, + { + "epoch": 4.374017810371923, + "grad_norm": 0.8058638481024218, + "learning_rate": 1.4839601898420153e-05, + "loss": 0.1411, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08079956471920013, + "step": 8350, + "valid_targets_mean": 1551.1, + "valid_targets_min": 1001 + }, + { + "epoch": 4.376636982713462, + "grad_norm": 0.796392217944281, + "learning_rate": 1.4814367516043263e-05, + "loss": 0.143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06639496982097626, + "step": 8355, + "valid_targets_mean": 1565.9, + "valid_targets_min": 1175 + }, + { + "epoch": 4.379256155055002, + "grad_norm": 0.8379626959646866, + "learning_rate": 1.4789141980727488e-05, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06710278987884521, + "step": 8360, + "valid_targets_mean": 1294.1, + "valid_targets_min": 779 + }, + { + "epoch": 4.381875327396543, + "grad_norm": 0.7903116687298194, + "learning_rate": 1.4763925335509411e-05, + "loss": 0.1408, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05957753211259842, + "step": 8365, + "valid_targets_mean": 1328.1, + "valid_targets_min": 801 + }, + { + "epoch": 4.384494499738083, + "grad_norm": 0.7292967799964774, + "learning_rate": 1.473871762341042e-05, + "loss": 0.129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06875885277986526, + "step": 8370, + "valid_targets_mean": 1635.6, + "valid_targets_min": 1020 + }, + { + "epoch": 4.387113672079622, + "grad_norm": 0.7156243819767485, + "learning_rate": 1.4713518887436702e-05, + "loss": 0.1321, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07740089297294617, + "step": 8375, + "valid_targets_mean": 1837.4, + "valid_targets_min": 1042 + }, + { + "epoch": 4.389732844421163, + "grad_norm": 1.3945087214797716, + "learning_rate": 1.46883291705791e-05, + "loss": 0.1383, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04777291789650917, + "step": 8380, + "valid_targets_mean": 1129.4, + "valid_targets_min": 716 + }, + { + "epoch": 4.392352016762703, + "grad_norm": 0.7379006624284808, + "learning_rate": 1.4663148515813066e-05, + "loss": 0.133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0646188035607338, + "step": 8385, + "valid_targets_mean": 1372.5, + "valid_targets_min": 707 + }, + { + "epoch": 4.394971189104243, + "grad_norm": 0.6852837713154475, + "learning_rate": 1.4637976966098628e-05, + "loss": 0.132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05881188064813614, + "step": 8390, + "valid_targets_mean": 1358.1, + "valid_targets_min": 1218 + }, + { + "epoch": 4.397590361445783, + "grad_norm": 0.74280798442123, + "learning_rate": 1.4612814564380231e-05, + "loss": 0.1349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0652073323726654, + "step": 8395, + "valid_targets_mean": 1529.0, + "valid_targets_min": 793 + }, + { + "epoch": 4.400209533787323, + "grad_norm": 0.7415650956747493, + "learning_rate": 1.4587661353586761e-05, + "loss": 0.145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06057053059339523, + "step": 8400, + "valid_targets_mean": 1374.6, + "valid_targets_min": 812 + }, + { + "epoch": 4.402828706128863, + "grad_norm": 0.7351610943137601, + "learning_rate": 1.456251737663138e-05, + "loss": 0.129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05575305223464966, + "step": 8405, + "valid_targets_mean": 1465.5, + "valid_targets_min": 979 + }, + { + "epoch": 4.405447878470404, + "grad_norm": 0.7253945566295781, + "learning_rate": 1.4537382676411536e-05, + "loss": 0.1382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07179462909698486, + "step": 8410, + "valid_targets_mean": 1695.2, + "valid_targets_min": 1344 + }, + { + "epoch": 4.408067050811943, + "grad_norm": 0.6688591587697432, + "learning_rate": 1.451225729580882e-05, + "loss": 0.1161, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06504763662815094, + "step": 8415, + "valid_targets_mean": 1684.1, + "valid_targets_min": 933 + }, + { + "epoch": 4.410686223153483, + "grad_norm": 0.871125450560016, + "learning_rate": 1.448714127768893e-05, + "loss": 0.1378, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08041547238826752, + "step": 8420, + "valid_targets_mean": 1425.9, + "valid_targets_min": 868 + }, + { + "epoch": 4.413305395495024, + "grad_norm": 0.7513210640802018, + "learning_rate": 1.4462034664901608e-05, + "loss": 0.135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08151189237833023, + "step": 8425, + "valid_targets_mean": 1714.0, + "valid_targets_min": 1064 + }, + { + "epoch": 4.415924567836564, + "grad_norm": 0.6658653252379988, + "learning_rate": 1.4436937500280522e-05, + "loss": 0.1331, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0863686203956604, + "step": 8430, + "valid_targets_mean": 2258.8, + "valid_targets_min": 1070 + }, + { + "epoch": 4.418543740178103, + "grad_norm": 0.7582159250380298, + "learning_rate": 1.4411849826643252e-05, + "loss": 0.1346, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08660875260829926, + "step": 8435, + "valid_targets_mean": 1788.9, + "valid_targets_min": 658 + }, + { + "epoch": 4.421162912519644, + "grad_norm": 0.8403620280349441, + "learning_rate": 1.4386771686791161e-05, + "loss": 0.135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0593537837266922, + "step": 8440, + "valid_targets_mean": 1278.2, + "valid_targets_min": 1018 + }, + { + "epoch": 4.423782084861184, + "grad_norm": 0.803379205842192, + "learning_rate": 1.4361703123509347e-05, + "loss": 0.1379, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08067694306373596, + "step": 8445, + "valid_targets_mean": 1629.4, + "valid_targets_min": 922 + }, + { + "epoch": 4.426401257202724, + "grad_norm": 1.1355549553613313, + "learning_rate": 1.4336644179566595e-05, + "loss": 0.1788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1881808340549469, + "step": 8450, + "valid_targets_mean": 5807.8, + "valid_targets_min": 2317 + }, + { + "epoch": 4.429020429544264, + "grad_norm": 0.6739038545042699, + "learning_rate": 1.4311594897715249e-05, + "loss": 0.2401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13510861992835999, + "step": 8455, + "valid_targets_mean": 5801.6, + "valid_targets_min": 2453 + }, + { + "epoch": 4.431639601885804, + "grad_norm": 0.5635264900723713, + "learning_rate": 1.4286555320691182e-05, + "loss": 0.2302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12830252945423126, + "step": 8460, + "valid_targets_mean": 5075.4, + "valid_targets_min": 835 + }, + { + "epoch": 4.434258774227344, + "grad_norm": 0.5961420877228959, + "learning_rate": 1.4261525491213712e-05, + "loss": 0.2259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12928646802902222, + "step": 8465, + "valid_targets_mean": 4082.1, + "valid_targets_min": 1201 + }, + { + "epoch": 4.436877946568885, + "grad_norm": 0.516998816586594, + "learning_rate": 1.4236505451985533e-05, + "loss": 0.2159, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10965876281261444, + "step": 8470, + "valid_targets_mean": 4560.0, + "valid_targets_min": 1880 + }, + { + "epoch": 4.439497118910424, + "grad_norm": 0.5072086441686843, + "learning_rate": 1.4211495245692608e-05, + "loss": 0.2387, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1697598397731781, + "step": 8475, + "valid_targets_mean": 6637.9, + "valid_targets_min": 2171 + }, + { + "epoch": 4.442116291251964, + "grad_norm": 0.5112366484598851, + "learning_rate": 1.4186494915004159e-05, + "loss": 0.2155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12562744319438934, + "step": 8480, + "valid_targets_mean": 6562.5, + "valid_targets_min": 1789 + }, + { + "epoch": 4.444735463593505, + "grad_norm": 0.5114706708375617, + "learning_rate": 1.4161504502572541e-05, + "loss": 0.2235, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10230296105146408, + "step": 8485, + "valid_targets_mean": 5572.5, + "valid_targets_min": 1855 + }, + { + "epoch": 4.447354635935045, + "grad_norm": 0.5475733007938988, + "learning_rate": 1.4136524051033181e-05, + "loss": 0.2236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1181647926568985, + "step": 8490, + "valid_targets_mean": 3913.8, + "valid_targets_min": 1024 + }, + { + "epoch": 4.449973808276584, + "grad_norm": 0.5072273973376359, + "learning_rate": 1.4111553603004538e-05, + "loss": 0.2301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11360402405261993, + "step": 8495, + "valid_targets_mean": 4908.0, + "valid_targets_min": 2107 + }, + { + "epoch": 4.452592980618125, + "grad_norm": 0.5474851914438034, + "learning_rate": 1.4086593201087979e-05, + "loss": 0.2257, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12396378070116043, + "step": 8500, + "valid_targets_mean": 4423.2, + "valid_targets_min": 802 + }, + { + "epoch": 4.455212152959665, + "grad_norm": 0.4563725455055399, + "learning_rate": 1.4061642887867732e-05, + "loss": 0.2149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08844783902168274, + "step": 8505, + "valid_targets_mean": 4354.5, + "valid_targets_min": 1597 + }, + { + "epoch": 4.457831325301205, + "grad_norm": 0.5962265185412776, + "learning_rate": 1.4036702705910842e-05, + "loss": 0.2287, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11159521341323853, + "step": 8510, + "valid_targets_mean": 3518.2, + "valid_targets_min": 1650 + }, + { + "epoch": 4.460450497642745, + "grad_norm": 0.6117588079878431, + "learning_rate": 1.401177269776703e-05, + "loss": 0.2195, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11195825785398483, + "step": 8515, + "valid_targets_mean": 4340.0, + "valid_targets_min": 1385 + }, + { + "epoch": 4.463069669984285, + "grad_norm": 0.5656053248881259, + "learning_rate": 1.3986852905968694e-05, + "loss": 0.2163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12745608389377594, + "step": 8520, + "valid_targets_mean": 4061.2, + "valid_targets_min": 3087 + }, + { + "epoch": 4.465688842325825, + "grad_norm": 0.48728447394557517, + "learning_rate": 1.396194337303078e-05, + "loss": 0.2172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10228261351585388, + "step": 8525, + "valid_targets_mean": 5073.4, + "valid_targets_min": 2785 + }, + { + "epoch": 4.4683080146673655, + "grad_norm": 0.5918271091658465, + "learning_rate": 1.393704414145073e-05, + "loss": 0.2138, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1172756776213646, + "step": 8530, + "valid_targets_mean": 4354.0, + "valid_targets_min": 1210 + }, + { + "epoch": 4.470927187008905, + "grad_norm": 0.5996756264617664, + "learning_rate": 1.3912155253708437e-05, + "loss": 0.2199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1186710000038147, + "step": 8535, + "valid_targets_mean": 3266.6, + "valid_targets_min": 1064 + }, + { + "epoch": 4.473546359350445, + "grad_norm": 0.6501591111893317, + "learning_rate": 1.3887276752266113e-05, + "loss": 0.2304, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12010496854782104, + "step": 8540, + "valid_targets_mean": 3360.6, + "valid_targets_min": 1678 + }, + { + "epoch": 4.4761655316919855, + "grad_norm": 0.5942735702190691, + "learning_rate": 1.3862408679568282e-05, + "loss": 0.2236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10922303795814514, + "step": 8545, + "valid_targets_mean": 3776.0, + "valid_targets_min": 671 + }, + { + "epoch": 4.478784704033526, + "grad_norm": 0.6041960206661602, + "learning_rate": 1.3837551078041644e-05, + "loss": 0.2207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11250011622905731, + "step": 8550, + "valid_targets_mean": 3642.9, + "valid_targets_min": 2000 + }, + { + "epoch": 4.481403876375065, + "grad_norm": 0.5919120996944863, + "learning_rate": 1.381270399009507e-05, + "loss": 0.2075, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12203384935855865, + "step": 8555, + "valid_targets_mean": 4261.9, + "valid_targets_min": 1532 + }, + { + "epoch": 4.4840230487166055, + "grad_norm": 0.552342437187867, + "learning_rate": 1.3787867458119464e-05, + "loss": 0.2113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07645145058631897, + "step": 8560, + "valid_targets_mean": 3298.9, + "valid_targets_min": 1051 + }, + { + "epoch": 4.486642221058146, + "grad_norm": 0.5753541481329446, + "learning_rate": 1.3763041524487721e-05, + "loss": 0.2034, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09878991544246674, + "step": 8565, + "valid_targets_mean": 3058.1, + "valid_targets_min": 1345 + }, + { + "epoch": 4.489261393399686, + "grad_norm": 0.5568476513605404, + "learning_rate": 1.3738226231554687e-05, + "loss": 0.2191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10006771981716156, + "step": 8570, + "valid_targets_mean": 4157.6, + "valid_targets_min": 2231 + }, + { + "epoch": 4.4918805657412255, + "grad_norm": 0.610187810290809, + "learning_rate": 1.3713421621657015e-05, + "loss": 0.2203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08782511204481125, + "step": 8575, + "valid_targets_mean": 3561.4, + "valid_targets_min": 2127 + }, + { + "epoch": 4.494499738082766, + "grad_norm": 0.6199379022678227, + "learning_rate": 1.3688627737113163e-05, + "loss": 0.2119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11463864147663116, + "step": 8580, + "valid_targets_mean": 3363.4, + "valid_targets_min": 1265 + }, + { + "epoch": 4.497118910424306, + "grad_norm": 0.5566260927340954, + "learning_rate": 1.3663844620223272e-05, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08432582020759583, + "step": 8585, + "valid_targets_mean": 3700.0, + "valid_targets_min": 908 + }, + { + "epoch": 4.499738082765846, + "grad_norm": 0.6676270058527579, + "learning_rate": 1.3639072313269106e-05, + "loss": 0.2119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1117234155535698, + "step": 8590, + "valid_targets_mean": 2910.9, + "valid_targets_min": 1264 + }, + { + "epoch": 4.502357255107386, + "grad_norm": 0.6150505130401227, + "learning_rate": 1.3614310858514016e-05, + "loss": 0.2094, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11026929318904877, + "step": 8595, + "valid_targets_mean": 2855.2, + "valid_targets_min": 926 + }, + { + "epoch": 4.504976427448926, + "grad_norm": 0.6234223131710177, + "learning_rate": 1.3589560298202811e-05, + "loss": 0.2121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08696039766073227, + "step": 8600, + "valid_targets_mean": 2905.6, + "valid_targets_min": 1016 + }, + { + "epoch": 4.507595599790466, + "grad_norm": 0.5756158294374238, + "learning_rate": 1.3564820674561726e-05, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09233200550079346, + "step": 8605, + "valid_targets_mean": 3499.9, + "valid_targets_min": 2019 + }, + { + "epoch": 4.510214772132006, + "grad_norm": 0.6000420463407551, + "learning_rate": 1.354009202979833e-05, + "loss": 0.2122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13629981875419617, + "step": 8610, + "valid_targets_mean": 4911.0, + "valid_targets_min": 3583 + }, + { + "epoch": 4.512833944473546, + "grad_norm": 0.6140025542567284, + "learning_rate": 1.3515374406101476e-05, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08976728469133377, + "step": 8615, + "valid_targets_mean": 3381.5, + "valid_targets_min": 1110 + }, + { + "epoch": 4.515453116815086, + "grad_norm": 0.5983989519981994, + "learning_rate": 1.3490667845641188e-05, + "loss": 0.2049, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12762027978897095, + "step": 8620, + "valid_targets_mean": 4497.9, + "valid_targets_min": 1460 + }, + { + "epoch": 4.518072289156627, + "grad_norm": 0.5996405003785557, + "learning_rate": 1.3465972390568643e-05, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11904385685920715, + "step": 8625, + "valid_targets_mean": 3962.4, + "valid_targets_min": 2719 + }, + { + "epoch": 4.520691461498167, + "grad_norm": 0.6081929788172725, + "learning_rate": 1.3441288083016058e-05, + "loss": 0.2168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0968737006187439, + "step": 8630, + "valid_targets_mean": 3175.9, + "valid_targets_min": 1673 + }, + { + "epoch": 4.523310633839706, + "grad_norm": 0.562972136820254, + "learning_rate": 1.3416614965096622e-05, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09831330180168152, + "step": 8635, + "valid_targets_mean": 3482.6, + "valid_targets_min": 1860 + }, + { + "epoch": 4.525929806181247, + "grad_norm": 0.5303813548554197, + "learning_rate": 1.3391953078904458e-05, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1032838225364685, + "step": 8640, + "valid_targets_mean": 4597.0, + "valid_targets_min": 2934 + }, + { + "epoch": 4.528548978522787, + "grad_norm": 0.5784020993018973, + "learning_rate": 1.3367302466514514e-05, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10434633493423462, + "step": 8645, + "valid_targets_mean": 3667.6, + "valid_targets_min": 1945 + }, + { + "epoch": 4.5311681508643264, + "grad_norm": 0.5605266917043563, + "learning_rate": 1.3342663169982486e-05, + "loss": 0.2145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10432937741279602, + "step": 8650, + "valid_targets_mean": 4618.9, + "valid_targets_min": 1908 + }, + { + "epoch": 4.533787323205867, + "grad_norm": 0.5783949109839808, + "learning_rate": 1.3318035231344806e-05, + "loss": 0.2093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11130911111831665, + "step": 8655, + "valid_targets_mean": 3561.0, + "valid_targets_min": 2250 + }, + { + "epoch": 4.536406495547407, + "grad_norm": 0.5739450321596407, + "learning_rate": 1.3293418692618483e-05, + "loss": 0.2085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08710279315710068, + "step": 8660, + "valid_targets_mean": 4291.5, + "valid_targets_min": 2492 + }, + { + "epoch": 4.539025667888947, + "grad_norm": 0.5583244584290854, + "learning_rate": 1.3268813595801113e-05, + "loss": 0.2106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10895667970180511, + "step": 8665, + "valid_targets_mean": 4467.1, + "valid_targets_min": 3059 + }, + { + "epoch": 4.541644840230488, + "grad_norm": 0.6100907000364568, + "learning_rate": 1.3244219982870748e-05, + "loss": 0.203, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10925273597240448, + "step": 8670, + "valid_targets_mean": 4255.2, + "valid_targets_min": 2283 + }, + { + "epoch": 4.544264012572027, + "grad_norm": 0.6245531949844386, + "learning_rate": 1.3219637895785864e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10976527631282806, + "step": 8675, + "valid_targets_mean": 3564.6, + "valid_targets_min": 1182 + }, + { + "epoch": 4.546883184913567, + "grad_norm": 0.5193275661920408, + "learning_rate": 1.3195067376485264e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09766323864459991, + "step": 8680, + "valid_targets_mean": 4895.6, + "valid_targets_min": 1869 + }, + { + "epoch": 4.549502357255108, + "grad_norm": 0.5542019246807562, + "learning_rate": 1.3170508466888009e-05, + "loss": 0.209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12730374932289124, + "step": 8685, + "valid_targets_mean": 4721.1, + "valid_targets_min": 2498 + }, + { + "epoch": 4.552121529596647, + "grad_norm": 0.63634751725488, + "learning_rate": 1.3145961208893372e-05, + "loss": 0.2169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10462123155593872, + "step": 8690, + "valid_targets_mean": 2935.6, + "valid_targets_min": 988 + }, + { + "epoch": 4.554740701938187, + "grad_norm": 0.6427429519421998, + "learning_rate": 1.312142564438072e-05, + "loss": 0.2076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11896716058254242, + "step": 8695, + "valid_targets_mean": 4191.4, + "valid_targets_min": 1534 + }, + { + "epoch": 4.557359874279728, + "grad_norm": 0.5784347769758553, + "learning_rate": 1.309690181520951e-05, + "loss": 0.2096, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10362888872623444, + "step": 8700, + "valid_targets_mean": 3546.4, + "valid_targets_min": 2252 + }, + { + "epoch": 4.559979046621268, + "grad_norm": 0.7312580868267099, + "learning_rate": 1.3072389763219144e-05, + "loss": 0.2113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11491537094116211, + "step": 8705, + "valid_targets_mean": 3548.0, + "valid_targets_min": 1436 + }, + { + "epoch": 4.562598218962807, + "grad_norm": 0.585793910877953, + "learning_rate": 1.3047889530228928e-05, + "loss": 0.2087, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10064186155796051, + "step": 8710, + "valid_targets_mean": 4002.1, + "valid_targets_min": 2333 + }, + { + "epoch": 4.565217391304348, + "grad_norm": 0.5832890454646561, + "learning_rate": 1.3023401158038038e-05, + "loss": 0.2069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08964493870735168, + "step": 8715, + "valid_targets_mean": 3507.9, + "valid_targets_min": 1259 + }, + { + "epoch": 4.567836563645888, + "grad_norm": 0.7514588487465512, + "learning_rate": 1.2998924688425377e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08776670694351196, + "step": 8720, + "valid_targets_mean": 2891.5, + "valid_targets_min": 1319 + }, + { + "epoch": 4.570455735987428, + "grad_norm": 0.6825442976719487, + "learning_rate": 1.2974460163149575e-05, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09235363453626633, + "step": 8725, + "valid_targets_mean": 3002.8, + "valid_targets_min": 1194 + }, + { + "epoch": 4.573074908328968, + "grad_norm": 0.6380077271378126, + "learning_rate": 1.2950007623948857e-05, + "loss": 0.2216, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09646183997392654, + "step": 8730, + "valid_targets_mean": 2704.0, + "valid_targets_min": 1453 + }, + { + "epoch": 4.575694080670508, + "grad_norm": 0.5204611273917362, + "learning_rate": 1.292556711254101e-05, + "loss": 0.202, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09181611984968185, + "step": 8735, + "valid_targets_mean": 5175.9, + "valid_targets_min": 3011 + }, + { + "epoch": 4.578313253012048, + "grad_norm": 0.5788594812890054, + "learning_rate": 1.2901138670623308e-05, + "loss": 0.2142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0812472477555275, + "step": 8740, + "valid_targets_mean": 2915.9, + "valid_targets_min": 1256 + }, + { + "epoch": 4.580932425353589, + "grad_norm": 0.6229753211884391, + "learning_rate": 1.2876722339872414e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13599498569965363, + "step": 8745, + "valid_targets_mean": 4661.6, + "valid_targets_min": 1336 + }, + { + "epoch": 4.583551597695128, + "grad_norm": 0.5970142439794883, + "learning_rate": 1.2852318161944348e-05, + "loss": 0.2174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10285273939371109, + "step": 8750, + "valid_targets_mean": 3390.4, + "valid_targets_min": 1275 + }, + { + "epoch": 4.586170770036668, + "grad_norm": 0.6032636206424804, + "learning_rate": 1.2827926178474388e-05, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1137130931019783, + "step": 8755, + "valid_targets_mean": 3660.0, + "valid_targets_min": 2202 + }, + { + "epoch": 4.588789942378209, + "grad_norm": 0.5904769474050618, + "learning_rate": 1.2803546431077008e-05, + "loss": 0.2051, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11178098618984222, + "step": 8760, + "valid_targets_mean": 4094.2, + "valid_targets_min": 3236 + }, + { + "epoch": 4.591409114719749, + "grad_norm": 0.6121303729014081, + "learning_rate": 1.2779178961345804e-05, + "loss": 0.2008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10016481578350067, + "step": 8765, + "valid_targets_mean": 3057.6, + "valid_targets_min": 1573 + }, + { + "epoch": 4.594028287061288, + "grad_norm": 0.5965664716395885, + "learning_rate": 1.2754823810853427e-05, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0960645079612732, + "step": 8770, + "valid_targets_mean": 3780.9, + "valid_targets_min": 2343 + }, + { + "epoch": 4.596647459402829, + "grad_norm": 0.5846086370918094, + "learning_rate": 1.2730481021151518e-05, + "loss": 0.1926, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08448463678359985, + "step": 8775, + "valid_targets_mean": 2815.8, + "valid_targets_min": 1026 + }, + { + "epoch": 4.599266631744369, + "grad_norm": 0.6480939691943781, + "learning_rate": 1.2706150633770612e-05, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09494088590145111, + "step": 8780, + "valid_targets_mean": 3460.9, + "valid_targets_min": 1042 + }, + { + "epoch": 4.601885804085909, + "grad_norm": 0.588488405650157, + "learning_rate": 1.268183269022011e-05, + "loss": 0.1844, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08301176875829697, + "step": 8785, + "valid_targets_mean": 3172.8, + "valid_targets_min": 1274 + }, + { + "epoch": 4.604504976427449, + "grad_norm": 0.64821912050279, + "learning_rate": 1.2657527231988151e-05, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10646785795688629, + "step": 8790, + "valid_targets_mean": 3939.5, + "valid_targets_min": 1575 + }, + { + "epoch": 4.607124148768989, + "grad_norm": 0.5872239639323615, + "learning_rate": 1.2633234300541612e-05, + "loss": 0.2258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07948024570941925, + "step": 8795, + "valid_targets_mean": 3739.0, + "valid_targets_min": 1312 + }, + { + "epoch": 4.609743321110529, + "grad_norm": 0.614964206330233, + "learning_rate": 1.2608953937325965e-05, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09030885994434357, + "step": 8800, + "valid_targets_mean": 3187.8, + "valid_targets_min": 1683 + }, + { + "epoch": 4.6123624934520695, + "grad_norm": 0.6273925560449134, + "learning_rate": 1.2584686183765245e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09735693037509918, + "step": 8805, + "valid_targets_mean": 3945.8, + "valid_targets_min": 1743 + }, + { + "epoch": 4.614981665793609, + "grad_norm": 0.5786449168789363, + "learning_rate": 1.2560431081261997e-05, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10264867544174194, + "step": 8810, + "valid_targets_mean": 3953.8, + "valid_targets_min": 2213 + }, + { + "epoch": 4.617600838135149, + "grad_norm": 0.5987735333734895, + "learning_rate": 1.253618867119715e-05, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1019289642572403, + "step": 8815, + "valid_targets_mean": 3690.4, + "valid_targets_min": 1586 + }, + { + "epoch": 4.6202200104766895, + "grad_norm": 0.6428371289531681, + "learning_rate": 1.251195899493001e-05, + "loss": 0.1993, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10397177934646606, + "step": 8820, + "valid_targets_mean": 3381.4, + "valid_targets_min": 2033 + }, + { + "epoch": 4.62283918281823, + "grad_norm": 0.5712917660290828, + "learning_rate": 1.248774209379814e-05, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08461996912956238, + "step": 8825, + "valid_targets_mean": 3651.2, + "valid_targets_min": 1616 + }, + { + "epoch": 4.625458355159769, + "grad_norm": 0.617775477940132, + "learning_rate": 1.2463538009117296e-05, + "loss": 0.2093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09714353829622269, + "step": 8830, + "valid_targets_mean": 3462.6, + "valid_targets_min": 2000 + }, + { + "epoch": 4.6280775275013095, + "grad_norm": 0.6085594306495448, + "learning_rate": 1.2439346782181407e-05, + "loss": 0.1972, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10350202769041061, + "step": 8835, + "valid_targets_mean": 3591.0, + "valid_targets_min": 1750 + }, + { + "epoch": 4.63069669984285, + "grad_norm": 0.6877011330716615, + "learning_rate": 1.2415168454262418e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1348126232624054, + "step": 8840, + "valid_targets_mean": 4291.8, + "valid_targets_min": 2429 + }, + { + "epoch": 4.63331587218439, + "grad_norm": 0.6583227986299749, + "learning_rate": 1.2391003066610314e-05, + "loss": 0.2178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10792477428913116, + "step": 8845, + "valid_targets_mean": 3645.9, + "valid_targets_min": 2342 + }, + { + "epoch": 4.6359350445259295, + "grad_norm": 0.6519960158301006, + "learning_rate": 1.2366850660452965e-05, + "loss": 0.2078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09717158228158951, + "step": 8850, + "valid_targets_mean": 3556.8, + "valid_targets_min": 1571 + }, + { + "epoch": 4.63855421686747, + "grad_norm": 0.6638671396378051, + "learning_rate": 1.2342711276996113e-05, + "loss": 0.2146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09164178371429443, + "step": 8855, + "valid_targets_mean": 2768.0, + "valid_targets_min": 1487 + }, + { + "epoch": 4.64117338920901, + "grad_norm": 0.5943330020742307, + "learning_rate": 1.2318584957423286e-05, + "loss": 0.206, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08871106803417206, + "step": 8860, + "valid_targets_mean": 3589.1, + "valid_targets_min": 1970 + }, + { + "epoch": 4.6437925615505495, + "grad_norm": 0.5775191376124311, + "learning_rate": 1.22944717428957e-05, + "loss": 0.2156, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09700791537761688, + "step": 8865, + "valid_targets_mean": 4087.2, + "valid_targets_min": 2794 + }, + { + "epoch": 4.64641173389209, + "grad_norm": 0.6084110899127034, + "learning_rate": 1.2270371674552242e-05, + "loss": 0.2102, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07788115739822388, + "step": 8870, + "valid_targets_mean": 2785.5, + "valid_targets_min": 1066 + }, + { + "epoch": 4.64903090623363, + "grad_norm": 0.5894367654447638, + "learning_rate": 1.2246284793509355e-05, + "loss": 0.2133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08926661312580109, + "step": 8875, + "valid_targets_mean": 3409.9, + "valid_targets_min": 2281 + }, + { + "epoch": 4.65165007857517, + "grad_norm": 0.6010400069578848, + "learning_rate": 1.222221114086099e-05, + "loss": 0.2101, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1068127304315567, + "step": 8880, + "valid_targets_mean": 4227.4, + "valid_targets_min": 1969 + }, + { + "epoch": 4.654269250916711, + "grad_norm": 0.6110943340704044, + "learning_rate": 1.2198150757678518e-05, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08686664700508118, + "step": 8885, + "valid_targets_mean": 2942.6, + "valid_targets_min": 1005 + }, + { + "epoch": 4.65688842325825, + "grad_norm": 0.6613189745416628, + "learning_rate": 1.2174103685010682e-05, + "loss": 0.2084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0999584048986435, + "step": 8890, + "valid_targets_mean": 3895.9, + "valid_targets_min": 1980 + }, + { + "epoch": 4.65950759559979, + "grad_norm": 0.6144052420711653, + "learning_rate": 1.2150069963883517e-05, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11973720788955688, + "step": 8895, + "valid_targets_mean": 4768.2, + "valid_targets_min": 1257 + }, + { + "epoch": 4.662126767941331, + "grad_norm": 0.6462378145777582, + "learning_rate": 1.2126049635300268e-05, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10318487137556076, + "step": 8900, + "valid_targets_mean": 3240.4, + "valid_targets_min": 1357 + }, + { + "epoch": 4.66474594028287, + "grad_norm": 0.5549609168642301, + "learning_rate": 1.2102042740241351e-05, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08621539175510406, + "step": 8905, + "valid_targets_mean": 4587.8, + "valid_targets_min": 1410 + }, + { + "epoch": 4.6673651126244105, + "grad_norm": 0.5926365166681564, + "learning_rate": 1.207804931966424e-05, + "loss": 0.2067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1017741784453392, + "step": 8910, + "valid_targets_mean": 3696.5, + "valid_targets_min": 1390 + }, + { + "epoch": 4.669984284965951, + "grad_norm": 0.6184214952217437, + "learning_rate": 1.2054069414503442e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08176013827323914, + "step": 8915, + "valid_targets_mean": 2997.4, + "valid_targets_min": 790 + }, + { + "epoch": 4.672603457307491, + "grad_norm": 0.5640817212445918, + "learning_rate": 1.2030103065670393e-05, + "loss": 0.2081, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10405165702104568, + "step": 8920, + "valid_targets_mean": 4428.5, + "valid_targets_min": 1901 + }, + { + "epoch": 4.675222629649031, + "grad_norm": 0.6645622137299424, + "learning_rate": 1.2006150314053393e-05, + "loss": 0.205, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09197638928890228, + "step": 8925, + "valid_targets_mean": 2790.1, + "valid_targets_min": 1460 + }, + { + "epoch": 4.677841801990571, + "grad_norm": 0.5869601237873328, + "learning_rate": 1.1982211200517573e-05, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10966682434082031, + "step": 8930, + "valid_targets_mean": 4607.4, + "valid_targets_min": 952 + }, + { + "epoch": 4.680460974332111, + "grad_norm": 0.7032947942517622, + "learning_rate": 1.195828576590476e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09690295159816742, + "step": 8935, + "valid_targets_mean": 2605.1, + "valid_targets_min": 1551 + }, + { + "epoch": 4.683080146673651, + "grad_norm": 0.5679371399683334, + "learning_rate": 1.1934374051033482e-05, + "loss": 0.2003, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0817055031657219, + "step": 8940, + "valid_targets_mean": 3755.8, + "valid_targets_min": 2267 + }, + { + "epoch": 4.685699319015191, + "grad_norm": 0.6091870981296118, + "learning_rate": 1.1910476096698828e-05, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09081467986106873, + "step": 8945, + "valid_targets_mean": 5198.8, + "valid_targets_min": 1233 + }, + { + "epoch": 4.688318491356731, + "grad_norm": 0.5623519306160905, + "learning_rate": 1.1886591943672418e-05, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08707252889871597, + "step": 8950, + "valid_targets_mean": 4899.5, + "valid_targets_min": 1741 + }, + { + "epoch": 4.690937663698271, + "grad_norm": 0.562387776471201, + "learning_rate": 1.1862721632702347e-05, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10288101434707642, + "step": 8955, + "valid_targets_mean": 4473.5, + "valid_targets_min": 856 + }, + { + "epoch": 4.693556836039812, + "grad_norm": 0.6209562551383604, + "learning_rate": 1.1838865204513061e-05, + "loss": 0.1981, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11920629441738129, + "step": 8960, + "valid_targets_mean": 4047.8, + "valid_targets_min": 1062 + }, + { + "epoch": 4.696176008381352, + "grad_norm": 0.5700712570773924, + "learning_rate": 1.1815022699805355e-05, + "loss": 0.19, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08797033131122589, + "step": 8965, + "valid_targets_mean": 4106.0, + "valid_targets_min": 974 + }, + { + "epoch": 4.698795180722891, + "grad_norm": 0.6692593805250973, + "learning_rate": 1.1791194159256235e-05, + "loss": 0.193, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09842033684253693, + "step": 8970, + "valid_targets_mean": 3092.2, + "valid_targets_min": 1447 + }, + { + "epoch": 4.701414353064432, + "grad_norm": 0.6041003165377862, + "learning_rate": 1.1767379623518919e-05, + "loss": 0.1901, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10699644684791565, + "step": 8975, + "valid_targets_mean": 4404.8, + "valid_targets_min": 1744 + }, + { + "epoch": 4.704033525405972, + "grad_norm": 0.6319936791409297, + "learning_rate": 1.1743579133222707e-05, + "loss": 0.1921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11326825618743896, + "step": 8980, + "valid_targets_mean": 4678.9, + "valid_targets_min": 1515 + }, + { + "epoch": 4.706652697747511, + "grad_norm": 0.6856971416603618, + "learning_rate": 1.1719792728972932e-05, + "loss": 0.2116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1220206767320633, + "step": 8985, + "valid_targets_mean": 3396.1, + "valid_targets_min": 1349 + }, + { + "epoch": 4.709271870089052, + "grad_norm": 0.6303279094222795, + "learning_rate": 1.1696020451350924e-05, + "loss": 0.1998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09829162806272507, + "step": 8990, + "valid_targets_mean": 3056.6, + "valid_targets_min": 1156 + }, + { + "epoch": 4.711891042430592, + "grad_norm": 0.6384663625152177, + "learning_rate": 1.1672262340913884e-05, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10249923169612885, + "step": 8995, + "valid_targets_mean": 3856.4, + "valid_targets_min": 2076 + }, + { + "epoch": 4.714510214772132, + "grad_norm": 0.5857701917977421, + "learning_rate": 1.1648518438194853e-05, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09439747035503387, + "step": 9000, + "valid_targets_mean": 3352.8, + "valid_targets_min": 1021 + }, + { + "epoch": 4.717129387113673, + "grad_norm": 0.658705211845484, + "learning_rate": 1.1624788783702647e-05, + "loss": 0.2042, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10137808322906494, + "step": 9005, + "valid_targets_mean": 3367.0, + "valid_targets_min": 654 + }, + { + "epoch": 4.719748559455212, + "grad_norm": 0.6236316801539902, + "learning_rate": 1.1601073417921729e-05, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10994338244199753, + "step": 9010, + "valid_targets_mean": 4698.5, + "valid_targets_min": 2336 + }, + { + "epoch": 4.722367731796752, + "grad_norm": 0.7131262743615736, + "learning_rate": 1.1577372381312251e-05, + "loss": 0.2089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09219497442245483, + "step": 9015, + "valid_targets_mean": 2237.0, + "valid_targets_min": 815 + }, + { + "epoch": 4.724986904138293, + "grad_norm": 0.5957692371412628, + "learning_rate": 1.155368571430985e-05, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08926228433847427, + "step": 9020, + "valid_targets_mean": 3623.5, + "valid_targets_min": 2389 + }, + { + "epoch": 4.727606076479832, + "grad_norm": 0.5969100290636394, + "learning_rate": 1.1530013457325708e-05, + "loss": 0.2029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10672080516815186, + "step": 9025, + "valid_targets_mean": 3766.9, + "valid_targets_min": 1793 + }, + { + "epoch": 4.730225248821372, + "grad_norm": 0.6588725046419033, + "learning_rate": 1.1506355650746364e-05, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09532336890697479, + "step": 9030, + "valid_targets_mean": 3133.1, + "valid_targets_min": 1217 + }, + { + "epoch": 4.732844421162913, + "grad_norm": 0.6309093896039423, + "learning_rate": 1.1482712334933769e-05, + "loss": 0.1874, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060718268156051636, + "step": 9035, + "valid_targets_mean": 1525.2, + "valid_targets_min": 534 + }, + { + "epoch": 4.735463593504453, + "grad_norm": 0.3885328245516775, + "learning_rate": 1.1459083550225088e-05, + "loss": 0.1005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03458120673894882, + "step": 9040, + "valid_targets_mean": 2735.4, + "valid_targets_min": 834 + }, + { + "epoch": 4.738082765845992, + "grad_norm": 0.3636562284485851, + "learning_rate": 1.1435469336932731e-05, + "loss": 0.0828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03856399655342102, + "step": 9045, + "valid_targets_mean": 3253.9, + "valid_targets_min": 2790 + }, + { + "epoch": 4.740701938187533, + "grad_norm": 0.3862809125108849, + "learning_rate": 1.141186973534424e-05, + "loss": 0.0771, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038694508373737335, + "step": 9050, + "valid_targets_mean": 3383.1, + "valid_targets_min": 2133 + }, + { + "epoch": 4.743321110529073, + "grad_norm": 0.41070606817145666, + "learning_rate": 1.1388284785722235e-05, + "loss": 0.0967, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04316247999668121, + "step": 9055, + "valid_targets_mean": 2954.8, + "valid_targets_min": 2338 + }, + { + "epoch": 4.745940282870613, + "grad_norm": 0.43047098119773924, + "learning_rate": 1.1364714528304325e-05, + "loss": 0.0766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03911907598376274, + "step": 9060, + "valid_targets_mean": 2680.4, + "valid_targets_min": 662 + }, + { + "epoch": 4.748559455212153, + "grad_norm": 1.0370700253451957, + "learning_rate": 1.1341159003303075e-05, + "loss": 0.1052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07769293338060379, + "step": 9065, + "valid_targets_mean": 1101.0, + "valid_targets_min": 559 + }, + { + "epoch": 4.751178627553693, + "grad_norm": 0.4185714612428154, + "learning_rate": 1.131761825090588e-05, + "loss": 0.1142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03520479053258896, + "step": 9070, + "valid_targets_mean": 3490.6, + "valid_targets_min": 1091 + }, + { + "epoch": 4.753797799895233, + "grad_norm": 0.3979703719222556, + "learning_rate": 1.129409231127496e-05, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040776483714580536, + "step": 9075, + "valid_targets_mean": 3061.8, + "valid_targets_min": 947 + }, + { + "epoch": 4.7564169722367735, + "grad_norm": 0.5246335636978204, + "learning_rate": 1.1270581224547258e-05, + "loss": 0.112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09622713923454285, + "step": 9080, + "valid_targets_mean": 2125.4, + "valid_targets_min": 859 + }, + { + "epoch": 4.759036144578313, + "grad_norm": 0.2880616851967164, + "learning_rate": 1.1247085030834378e-05, + "loss": 0.0798, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04296407103538513, + "step": 9085, + "valid_targets_mean": 5837.8, + "valid_targets_min": 636 + }, + { + "epoch": 4.761655316919853, + "grad_norm": 0.3870107808118631, + "learning_rate": 1.1223603770222509e-05, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044115059077739716, + "step": 9090, + "valid_targets_mean": 2185.0, + "valid_targets_min": 826 + }, + { + "epoch": 4.7642744892613935, + "grad_norm": 0.31312418094636496, + "learning_rate": 1.1200137482772369e-05, + "loss": 0.078, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03482242301106453, + "step": 9095, + "valid_targets_mean": 3446.5, + "valid_targets_min": 1096 + }, + { + "epoch": 4.766893661602934, + "grad_norm": 0.4510915835331537, + "learning_rate": 1.1176686208519131e-05, + "loss": 0.0956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04061276465654373, + "step": 9100, + "valid_targets_mean": 2343.1, + "valid_targets_min": 688 + }, + { + "epoch": 4.769512833944473, + "grad_norm": 0.3663711754274614, + "learning_rate": 1.1153249987472332e-05, + "loss": 0.0971, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028953764587640762, + "step": 9105, + "valid_targets_mean": 2938.5, + "valid_targets_min": 553 + }, + { + "epoch": 4.7721320062860135, + "grad_norm": 0.4756827239507784, + "learning_rate": 1.1129828859615879e-05, + "loss": 0.0791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0428026020526886, + "step": 9110, + "valid_targets_mean": 2988.9, + "valid_targets_min": 566 + }, + { + "epoch": 4.774751178627554, + "grad_norm": 0.32947379184775777, + "learning_rate": 1.1106422864907863e-05, + "loss": 0.073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030004404485225677, + "step": 9115, + "valid_targets_mean": 3230.2, + "valid_targets_min": 1111 + }, + { + "epoch": 4.777370350969094, + "grad_norm": 0.4460483467027478, + "learning_rate": 1.1083032043280624e-05, + "loss": 0.0782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054130714386701584, + "step": 9120, + "valid_targets_mean": 1232.1, + "valid_targets_min": 622 + }, + { + "epoch": 4.7799895233106335, + "grad_norm": 0.5754521811626009, + "learning_rate": 1.1059656434640557e-05, + "loss": 0.0898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04246404021978378, + "step": 9125, + "valid_targets_mean": 2132.5, + "valid_targets_min": 696 + }, + { + "epoch": 4.782608695652174, + "grad_norm": 0.47194891500723624, + "learning_rate": 1.1036296078868136e-05, + "loss": 0.088, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04515007510781288, + "step": 9130, + "valid_targets_mean": 3225.2, + "valid_targets_min": 1922 + }, + { + "epoch": 4.785227867993714, + "grad_norm": 0.46523936381688163, + "learning_rate": 1.1012951015817804e-05, + "loss": 0.0893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033740557730197906, + "step": 9135, + "valid_targets_mean": 2008.5, + "valid_targets_min": 802 + }, + { + "epoch": 4.787847040335254, + "grad_norm": 0.5086398379720268, + "learning_rate": 1.0989621285317913e-05, + "loss": 0.0963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06698470562696457, + "step": 9140, + "valid_targets_mean": 1754.1, + "valid_targets_min": 957 + }, + { + "epoch": 4.790466212676794, + "grad_norm": 0.3848292805462421, + "learning_rate": 1.0966306927170658e-05, + "loss": 0.1126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04305237531661987, + "step": 9145, + "valid_targets_mean": 2498.5, + "valid_targets_min": 641 + }, + { + "epoch": 4.793085385018334, + "grad_norm": 0.42888435339467895, + "learning_rate": 1.094300798115201e-05, + "loss": 0.0751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04948770999908447, + "step": 9150, + "valid_targets_mean": 2364.6, + "valid_targets_min": 924 + }, + { + "epoch": 4.795704557359874, + "grad_norm": 0.5169038984285168, + "learning_rate": 1.0919724487011636e-05, + "loss": 0.1123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.080439493060112, + "step": 9155, + "valid_targets_mean": 2385.5, + "valid_targets_min": 700 + }, + { + "epoch": 4.798323729701415, + "grad_norm": 0.4854786192263136, + "learning_rate": 1.0896456484472862e-05, + "loss": 0.0946, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0754244327545166, + "step": 9160, + "valid_targets_mean": 3363.2, + "valid_targets_min": 2731 + }, + { + "epoch": 4.800942902042954, + "grad_norm": 0.5247983437432087, + "learning_rate": 1.0873204013232544e-05, + "loss": 0.0841, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07040989398956299, + "step": 9165, + "valid_targets_mean": 2615.0, + "valid_targets_min": 893 + }, + { + "epoch": 4.8035620743844945, + "grad_norm": 0.6332762654797041, + "learning_rate": 1.0849967112961097e-05, + "loss": 0.2457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.26801955699920654, + "step": 9170, + "valid_targets_mean": 2615.9, + "valid_targets_min": 922 + }, + { + "epoch": 4.806181246726035, + "grad_norm": 0.4505779202601575, + "learning_rate": 1.0826745823302312e-05, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030948026105761528, + "step": 9175, + "valid_targets_mean": 1357.4, + "valid_targets_min": 507 + }, + { + "epoch": 4.808800419067575, + "grad_norm": 0.39769796099027094, + "learning_rate": 1.0803540183873399e-05, + "loss": 0.0965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047426559031009674, + "step": 9180, + "valid_targets_mean": 3236.9, + "valid_targets_min": 615 + }, + { + "epoch": 4.8114195914091145, + "grad_norm": 0.44864526784910297, + "learning_rate": 1.078035023426483e-05, + "loss": 0.1755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06312110275030136, + "step": 9185, + "valid_targets_mean": 3023.1, + "valid_targets_min": 826 + }, + { + "epoch": 4.814038763750655, + "grad_norm": 0.42368332393686664, + "learning_rate": 1.0757176014040325e-05, + "loss": 0.0825, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.041739609092473984, + "step": 9190, + "valid_targets_mean": 2924.4, + "valid_targets_min": 851 + }, + { + "epoch": 4.816657936092195, + "grad_norm": 0.4238813341789226, + "learning_rate": 1.073401756273677e-05, + "loss": 0.0824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05370740592479706, + "step": 9195, + "valid_targets_mean": 4528.4, + "valid_targets_min": 2665 + }, + { + "epoch": 4.8192771084337345, + "grad_norm": 0.3523251882136089, + "learning_rate": 1.0710874919864143e-05, + "loss": 0.0848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029187655076384544, + "step": 9200, + "valid_targets_mean": 2820.6, + "valid_targets_min": 595 + }, + { + "epoch": 4.821896280775275, + "grad_norm": 0.4207070866316958, + "learning_rate": 1.0687748124905451e-05, + "loss": 0.0776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04913245514035225, + "step": 9205, + "valid_targets_mean": 2583.1, + "valid_targets_min": 825 + }, + { + "epoch": 4.824515453116815, + "grad_norm": 0.43565822810950644, + "learning_rate": 1.066463721731667e-05, + "loss": 0.0789, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03805457428097725, + "step": 9210, + "valid_targets_mean": 3771.5, + "valid_targets_min": 672 + }, + { + "epoch": 4.827134625458355, + "grad_norm": 0.3886853563096883, + "learning_rate": 1.0641542236526672e-05, + "loss": 0.0725, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02948293834924698, + "step": 9215, + "valid_targets_mean": 3165.0, + "valid_targets_min": 1956 + }, + { + "epoch": 4.829753797799896, + "grad_norm": 0.38044400941168516, + "learning_rate": 1.0618463221937136e-05, + "loss": 0.0797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03678198903799057, + "step": 9220, + "valid_targets_mean": 3719.8, + "valid_targets_min": 2540 + }, + { + "epoch": 4.832372970141435, + "grad_norm": 0.42656011291326623, + "learning_rate": 1.0595400212922526e-05, + "loss": 0.0706, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049015581607818604, + "step": 9225, + "valid_targets_mean": 3447.5, + "valid_targets_min": 1240 + }, + { + "epoch": 4.834992142482975, + "grad_norm": 0.2731718010599396, + "learning_rate": 1.0572353248829988e-05, + "loss": 0.111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022322287783026695, + "step": 9230, + "valid_targets_mean": 2975.2, + "valid_targets_min": 623 + }, + { + "epoch": 4.837611314824516, + "grad_norm": 0.4003110398084211, + "learning_rate": 1.0549322368979298e-05, + "loss": 0.0665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03430616110563278, + "step": 9235, + "valid_targets_mean": 2960.1, + "valid_targets_min": 583 + }, + { + "epoch": 4.840230487166055, + "grad_norm": 0.3738422547624986, + "learning_rate": 1.0526307612662792e-05, + "loss": 0.0703, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033656422048807144, + "step": 9240, + "valid_targets_mean": 3723.5, + "valid_targets_min": 3230 + }, + { + "epoch": 4.842849659507595, + "grad_norm": 0.413651615950664, + "learning_rate": 1.05033090191453e-05, + "loss": 0.0693, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02839820832014084, + "step": 9245, + "valid_targets_mean": 2520.9, + "valid_targets_min": 647 + }, + { + "epoch": 4.845468831849136, + "grad_norm": 0.7566790245345728, + "learning_rate": 1.048032662766405e-05, + "loss": 0.1146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08817195892333984, + "step": 9250, + "valid_targets_mean": 1094.5, + "valid_targets_min": 195 + }, + { + "epoch": 4.848088004190676, + "grad_norm": 0.37846533120167325, + "learning_rate": 1.0457360477428682e-05, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05591708794236183, + "step": 9255, + "valid_targets_mean": 3635.9, + "valid_targets_min": 3050 + }, + { + "epoch": 4.850707176532216, + "grad_norm": 0.5551310070150359, + "learning_rate": 1.0434410607621065e-05, + "loss": 0.1093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06684407591819763, + "step": 9260, + "valid_targets_mean": 2382.9, + "valid_targets_min": 1087 + }, + { + "epoch": 4.853326348873756, + "grad_norm": 0.3879526447794273, + "learning_rate": 1.0411477057395358e-05, + "loss": 0.0776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03387272730469704, + "step": 9265, + "valid_targets_mean": 3025.5, + "valid_targets_min": 847 + }, + { + "epoch": 4.855945521215296, + "grad_norm": 0.5467806127265231, + "learning_rate": 1.0388559865877816e-05, + "loss": 0.1036, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05753830075263977, + "step": 9270, + "valid_targets_mean": 2263.1, + "valid_targets_min": 916 + }, + { + "epoch": 4.858564693556836, + "grad_norm": 0.46981403028756086, + "learning_rate": 1.0365659072166823e-05, + "loss": 0.1074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0404030978679657, + "step": 9275, + "valid_targets_mean": 2567.0, + "valid_targets_min": 1062 + }, + { + "epoch": 4.861183865898376, + "grad_norm": 0.6367641161933895, + "learning_rate": 1.0342774715332776e-05, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06532709300518036, + "step": 9280, + "valid_targets_mean": 1443.4, + "valid_targets_min": 771 + }, + { + "epoch": 4.863803038239916, + "grad_norm": 0.4112965980758365, + "learning_rate": 1.0319906834418027e-05, + "loss": 0.079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0373617485165596, + "step": 9285, + "valid_targets_mean": 2753.2, + "valid_targets_min": 707 + }, + { + "epoch": 4.866422210581456, + "grad_norm": 0.3772559625858325, + "learning_rate": 1.0297055468436823e-05, + "loss": 0.0862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038474299013614655, + "step": 9290, + "valid_targets_mean": 3665.8, + "valid_targets_min": 2850 + }, + { + "epoch": 4.869041382922997, + "grad_norm": 0.41255509052506123, + "learning_rate": 1.0274220656375235e-05, + "loss": 0.0834, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031541645526885986, + "step": 9295, + "valid_targets_mean": 2238.2, + "valid_targets_min": 749 + }, + { + "epoch": 4.871660555264537, + "grad_norm": 0.3511844272654673, + "learning_rate": 1.0251402437191089e-05, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038034938275814056, + "step": 9300, + "valid_targets_mean": 4488.2, + "valid_targets_min": 1052 + }, + { + "epoch": 4.874279727606076, + "grad_norm": 0.3139669336754645, + "learning_rate": 1.0228600849813915e-05, + "loss": 0.0668, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03635767474770546, + "step": 9305, + "valid_targets_mean": 4258.2, + "valid_targets_min": 2540 + }, + { + "epoch": 4.876898899947617, + "grad_norm": 0.3349043594554718, + "learning_rate": 1.0205815933144833e-05, + "loss": 0.0788, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029315005987882614, + "step": 9310, + "valid_targets_mean": 3782.5, + "valid_targets_min": 1859 + }, + { + "epoch": 4.879518072289157, + "grad_norm": 0.38057616203586175, + "learning_rate": 1.0183047726056574e-05, + "loss": 0.0931, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02993633598089218, + "step": 9315, + "valid_targets_mean": 2879.5, + "valid_targets_min": 893 + }, + { + "epoch": 4.882137244630696, + "grad_norm": 0.5203800580710363, + "learning_rate": 1.0160296267393308e-05, + "loss": 0.0888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04160454124212265, + "step": 9320, + "valid_targets_mean": 2072.6, + "valid_targets_min": 753 + }, + { + "epoch": 4.884756416972237, + "grad_norm": 0.3040895272766178, + "learning_rate": 1.013756159597068e-05, + "loss": 0.0824, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02434583194553852, + "step": 9325, + "valid_targets_mean": 4205.6, + "valid_targets_min": 3934 + }, + { + "epoch": 4.887375589313777, + "grad_norm": 0.4491421587888301, + "learning_rate": 1.0114843750575654e-05, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04224167764186859, + "step": 9330, + "valid_targets_mean": 3174.9, + "valid_targets_min": 1576 + }, + { + "epoch": 4.889994761655317, + "grad_norm": 0.4658221903856158, + "learning_rate": 1.0092142769966506e-05, + "loss": 0.0859, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04499458521604538, + "step": 9335, + "valid_targets_mean": 3241.8, + "valid_targets_min": 1006 + }, + { + "epoch": 4.892613933996857, + "grad_norm": 0.3577552340286057, + "learning_rate": 1.0069458692872742e-05, + "loss": 0.071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024907685816287994, + "step": 9340, + "valid_targets_mean": 3445.5, + "valid_targets_min": 980 + }, + { + "epoch": 4.895233106338397, + "grad_norm": 0.5530245185860783, + "learning_rate": 1.0046791557995023e-05, + "loss": 0.0805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05116071552038193, + "step": 9345, + "valid_targets_mean": 2083.9, + "valid_targets_min": 745 + }, + { + "epoch": 4.897852278679937, + "grad_norm": 0.5319416341003791, + "learning_rate": 1.002414140400511e-05, + "loss": 0.0785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049515582621097565, + "step": 9350, + "valid_targets_mean": 2768.8, + "valid_targets_min": 1020 + }, + { + "epoch": 4.9004714510214775, + "grad_norm": 0.46736897934250193, + "learning_rate": 1.000150826954579e-05, + "loss": 0.0797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05170634388923645, + "step": 9355, + "valid_targets_mean": 2314.8, + "valid_targets_min": 967 + }, + { + "epoch": 4.903090623363017, + "grad_norm": 0.4764884165451955, + "learning_rate": 9.978892193230821e-06, + "loss": 0.0962, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0348917581140995, + "step": 9360, + "valid_targets_mean": 2462.0, + "valid_targets_min": 1015 + }, + { + "epoch": 4.905709795704557, + "grad_norm": 0.5815973718141937, + "learning_rate": 9.956293213644839e-06, + "loss": 0.0784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0444444939494133, + "step": 9365, + "valid_targets_mean": 1214.9, + "valid_targets_min": 705 + }, + { + "epoch": 4.9083289680460975, + "grad_norm": 0.3843792514609499, + "learning_rate": 9.933711369343329e-06, + "loss": 0.0878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027883168309926987, + "step": 9370, + "valid_targets_mean": 2364.8, + "valid_targets_min": 622 + }, + { + "epoch": 4.910948140387638, + "grad_norm": 0.4277008067956395, + "learning_rate": 9.91114669885254e-06, + "loss": 0.1028, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03156109154224396, + "step": 9375, + "valid_targets_mean": 3702.2, + "valid_targets_min": 1322 + }, + { + "epoch": 4.913567312729177, + "grad_norm": 0.39996412337164605, + "learning_rate": 9.888599240669419e-06, + "loss": 0.0797, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039226680994033813, + "step": 9380, + "valid_targets_mean": 2863.1, + "valid_targets_min": 1011 + }, + { + "epoch": 4.9161864850707175, + "grad_norm": 0.4507681899326495, + "learning_rate": 9.866069033261548e-06, + "loss": 0.0701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04473258554935455, + "step": 9385, + "valid_targets_mean": 3106.1, + "valid_targets_min": 874 + }, + { + "epoch": 4.918805657412258, + "grad_norm": 0.46792715886752356, + "learning_rate": 9.843556115067083e-06, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05716276168823242, + "step": 9390, + "valid_targets_mean": 3161.4, + "valid_targets_min": 1420 + }, + { + "epoch": 4.921424829753798, + "grad_norm": 0.4056962666000174, + "learning_rate": 9.821060524494658e-06, + "loss": 0.0945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032773829996585846, + "step": 9395, + "valid_targets_mean": 2796.0, + "valid_targets_min": 1109 + }, + { + "epoch": 4.9240440020953375, + "grad_norm": 0.39625295730691934, + "learning_rate": 9.798582299923393e-06, + "loss": 0.0627, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03169768676161766, + "step": 9400, + "valid_targets_mean": 1822.9, + "valid_targets_min": 684 + }, + { + "epoch": 4.926663174436878, + "grad_norm": 0.4851165513493398, + "learning_rate": 9.776121479702722e-06, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03963569551706314, + "step": 9405, + "valid_targets_mean": 2389.8, + "valid_targets_min": 671 + }, + { + "epoch": 4.929282346778418, + "grad_norm": 0.37566566044442923, + "learning_rate": 9.753678102152449e-06, + "loss": 0.0732, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03486199304461479, + "step": 9410, + "valid_targets_mean": 4513.8, + "valid_targets_min": 1057 + }, + { + "epoch": 4.931901519119958, + "grad_norm": 0.3614742104167863, + "learning_rate": 9.731252205562551e-06, + "loss": 0.0665, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031139519065618515, + "step": 9415, + "valid_targets_mean": 3533.1, + "valid_targets_min": 1243 + }, + { + "epoch": 4.934520691461498, + "grad_norm": 0.35676337242311723, + "learning_rate": 9.708843828193251e-06, + "loss": 0.0871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038030996918678284, + "step": 9420, + "valid_targets_mean": 3911.2, + "valid_targets_min": 3322 + }, + { + "epoch": 4.937139863803038, + "grad_norm": 0.502189310973067, + "learning_rate": 9.686453008274825e-06, + "loss": 0.0744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0454743355512619, + "step": 9425, + "valid_targets_mean": 1820.0, + "valid_targets_min": 680 + }, + { + "epoch": 4.9397590361445785, + "grad_norm": 0.41894087876609704, + "learning_rate": 9.664079784007627e-06, + "loss": 0.1014, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04343470558524132, + "step": 9430, + "valid_targets_mean": 3371.9, + "valid_targets_min": 640 + }, + { + "epoch": 4.942378208486119, + "grad_norm": 0.2792855993214479, + "learning_rate": 9.64172419356198e-06, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.021821660920977592, + "step": 9435, + "valid_targets_mean": 3327.0, + "valid_targets_min": 929 + }, + { + "epoch": 4.944997380827658, + "grad_norm": 0.5318348169324256, + "learning_rate": 9.619386275078129e-06, + "loss": 0.0937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057028867304325104, + "step": 9440, + "valid_targets_mean": 3318.0, + "valid_targets_min": 2380 + }, + { + "epoch": 4.9476165531691985, + "grad_norm": 0.3813103838617628, + "learning_rate": 9.597066066666164e-06, + "loss": 0.0718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037043243646621704, + "step": 9445, + "valid_targets_mean": 3709.4, + "valid_targets_min": 2619 + }, + { + "epoch": 4.950235725510739, + "grad_norm": 0.6477991161376521, + "learning_rate": 9.574763606405968e-06, + "loss": 0.0816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0495147705078125, + "step": 9450, + "valid_targets_mean": 1574.6, + "valid_targets_min": 693 + }, + { + "epoch": 4.952854897852279, + "grad_norm": 0.408370764045725, + "learning_rate": 9.55247893234712e-06, + "loss": 0.0776, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042133379727602005, + "step": 9455, + "valid_targets_mean": 3885.1, + "valid_targets_min": 3129 + }, + { + "epoch": 4.9554740701938185, + "grad_norm": 0.7488973122866266, + "learning_rate": 9.5302120825089e-06, + "loss": 0.1033, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049062225967645645, + "step": 9460, + "valid_targets_mean": 1611.0, + "valid_targets_min": 543 + }, + { + "epoch": 4.958093242535359, + "grad_norm": 0.4413157045511703, + "learning_rate": 9.507963094880124e-06, + "loss": 0.1043, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04812631011009216, + "step": 9465, + "valid_targets_mean": 2319.6, + "valid_targets_min": 503 + }, + { + "epoch": 4.960712414876899, + "grad_norm": 0.5096418919473332, + "learning_rate": 9.485732007419188e-06, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05371435731649399, + "step": 9470, + "valid_targets_mean": 1987.9, + "valid_targets_min": 625 + }, + { + "epoch": 4.963331587218439, + "grad_norm": 0.46327809990399793, + "learning_rate": 9.463518858053904e-06, + "loss": 0.0786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034822236746549606, + "step": 9475, + "valid_targets_mean": 3822.8, + "valid_targets_min": 2822 + }, + { + "epoch": 4.965950759559979, + "grad_norm": 0.4245950929755056, + "learning_rate": 9.441323684681502e-06, + "loss": 0.0777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0466957613825798, + "step": 9480, + "valid_targets_mean": 3724.5, + "valid_targets_min": 2987 + }, + { + "epoch": 4.968569931901519, + "grad_norm": 0.36685123790091356, + "learning_rate": 9.419146525168545e-06, + "loss": 0.0685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028822703287005424, + "step": 9485, + "valid_targets_mean": 3022.0, + "valid_targets_min": 856 + }, + { + "epoch": 4.971189104243059, + "grad_norm": 0.9315212054751016, + "learning_rate": 9.396987417350856e-06, + "loss": 0.0803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0436936616897583, + "step": 9490, + "valid_targets_mean": 764.2, + "valid_targets_min": 563 + }, + { + "epoch": 4.973808276584599, + "grad_norm": 0.393555698091191, + "learning_rate": 9.374846399033469e-06, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03284937143325806, + "step": 9495, + "valid_targets_mean": 3117.1, + "valid_targets_min": 751 + }, + { + "epoch": 4.976427448926139, + "grad_norm": 0.5184553638302952, + "learning_rate": 9.352723507990528e-06, + "loss": 0.0719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04663628339767456, + "step": 9500, + "valid_targets_mean": 3566.1, + "valid_targets_min": 980 + }, + { + "epoch": 4.979046621267679, + "grad_norm": 0.4562573056132558, + "learning_rate": 9.330618781965305e-06, + "loss": 0.0726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038732901215553284, + "step": 9505, + "valid_targets_mean": 3589.8, + "valid_targets_min": 2538 + }, + { + "epoch": 4.98166579360922, + "grad_norm": 0.3036424051720614, + "learning_rate": 9.308532258670025e-06, + "loss": 0.0817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028068505227565765, + "step": 9510, + "valid_targets_mean": 4182.8, + "valid_targets_min": 485 + }, + { + "epoch": 4.98428496595076, + "grad_norm": 0.28761388918511854, + "learning_rate": 9.28646397578589e-06, + "loss": 0.069, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03375013917684555, + "step": 9515, + "valid_targets_mean": 4832.8, + "valid_targets_min": 2551 + }, + { + "epoch": 4.986904138292299, + "grad_norm": 0.32327604601205945, + "learning_rate": 9.264413970962969e-06, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037574365735054016, + "step": 9520, + "valid_targets_mean": 4733.6, + "valid_targets_min": 3365 + }, + { + "epoch": 4.98952331063384, + "grad_norm": 0.4416179363446702, + "learning_rate": 9.24238228182016e-06, + "loss": 0.0794, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05937991291284561, + "step": 9525, + "valid_targets_mean": 2954.4, + "valid_targets_min": 872 + }, + { + "epoch": 4.99214248297538, + "grad_norm": 0.36503077226780584, + "learning_rate": 9.220368945945103e-06, + "loss": 0.1636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03802693635225296, + "step": 9530, + "valid_targets_mean": 2590.9, + "valid_targets_min": 1015 + }, + { + "epoch": 4.994761655316919, + "grad_norm": 0.41033605107517607, + "learning_rate": 9.198374000894123e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05921158567070961, + "step": 9535, + "valid_targets_mean": 2342.4, + "valid_targets_min": 967 + }, + { + "epoch": 4.99738082765846, + "grad_norm": 0.3642781476199765, + "learning_rate": 9.176397484192184e-06, + "loss": 0.0748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03135570138692856, + "step": 9540, + "valid_targets_mean": 3335.1, + "valid_targets_min": 886 + }, + { + "epoch": 5.0, + "grad_norm": 0.40865859037767577, + "learning_rate": 9.1544394333328e-06, + "loss": 0.1082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05871417373418808, + "step": 9545, + "valid_targets_mean": 5394.5, + "valid_targets_min": 3513 + }, + { + "epoch": 5.00261917234154, + "grad_norm": 0.5395630571687405, + "learning_rate": 9.132499885777964e-06, + "loss": 0.1655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08791050314903259, + "step": 9550, + "valid_targets_mean": 7214.4, + "valid_targets_min": 4847 + }, + { + "epoch": 5.00523834468308, + "grad_norm": 0.44157383770841946, + "learning_rate": 9.110578878958145e-06, + "loss": 0.1573, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07886155694723129, + "step": 9555, + "valid_targets_mean": 6417.1, + "valid_targets_min": 4480 + }, + { + "epoch": 5.00785751702462, + "grad_norm": 0.43682047423851106, + "learning_rate": 9.088676450272124e-06, + "loss": 0.1505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08467578142881393, + "step": 9560, + "valid_targets_mean": 7605.4, + "valid_targets_min": 5236 + }, + { + "epoch": 5.01047668936616, + "grad_norm": 0.38113245866468276, + "learning_rate": 9.06679263708705e-06, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06564006209373474, + "step": 9565, + "valid_targets_mean": 5827.9, + "valid_targets_min": 4727 + }, + { + "epoch": 5.013095861707701, + "grad_norm": 0.39617142040080333, + "learning_rate": 9.044927476738252e-06, + "loss": 0.163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08957388997077942, + "step": 9570, + "valid_targets_mean": 6448.2, + "valid_targets_min": 4211 + }, + { + "epoch": 5.01571503404924, + "grad_norm": 0.4320602288317964, + "learning_rate": 9.02308100652927e-06, + "loss": 0.1568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09174610674381256, + "step": 9575, + "valid_targets_mean": 6108.1, + "valid_targets_min": 4279 + }, + { + "epoch": 5.01833420639078, + "grad_norm": 0.37751973936978805, + "learning_rate": 9.001253263731753e-06, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06624109297990799, + "step": 9580, + "valid_targets_mean": 5962.4, + "valid_targets_min": 4589 + }, + { + "epoch": 5.020953378732321, + "grad_norm": 0.4702614323708166, + "learning_rate": 8.979444285585392e-06, + "loss": 0.1659, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0781085193157196, + "step": 9585, + "valid_targets_mean": 3503.2, + "valid_targets_min": 2015 + }, + { + "epoch": 5.023572551073861, + "grad_norm": 0.3949773013133077, + "learning_rate": 8.957654109297874e-06, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06408419460058212, + "step": 9590, + "valid_targets_mean": 6102.5, + "valid_targets_min": 4778 + }, + { + "epoch": 5.0261917234154, + "grad_norm": 0.35899135192101694, + "learning_rate": 8.935882772044798e-06, + "loss": 0.1504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07194272428750992, + "step": 9595, + "valid_targets_mean": 5868.2, + "valid_targets_min": 4612 + }, + { + "epoch": 5.028810895756941, + "grad_norm": 0.4008046263140558, + "learning_rate": 8.91413031096963e-06, + "loss": 0.1615, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08334788680076599, + "step": 9600, + "valid_targets_mean": 6105.8, + "valid_targets_min": 4491 + }, + { + "epoch": 5.031430068098481, + "grad_norm": 0.39060149618863954, + "learning_rate": 8.892396763183637e-06, + "loss": 0.1488, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07954488694667816, + "step": 9605, + "valid_targets_mean": 6962.0, + "valid_targets_min": 4754 + }, + { + "epoch": 5.034049240440021, + "grad_norm": 0.36645296335708355, + "learning_rate": 8.870682165765798e-06, + "loss": 0.1466, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07006168365478516, + "step": 9610, + "valid_targets_mean": 6035.1, + "valid_targets_min": 4787 + }, + { + "epoch": 5.036668412781561, + "grad_norm": 0.3568079492539736, + "learning_rate": 8.848986555762777e-06, + "loss": 0.1399, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.068642757833004, + "step": 9615, + "valid_targets_mean": 6028.5, + "valid_targets_min": 5213 + }, + { + "epoch": 5.039287585123101, + "grad_norm": 0.3795579466371004, + "learning_rate": 8.827309970188847e-06, + "loss": 0.1401, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06725533306598663, + "step": 9620, + "valid_targets_mean": 6814.1, + "valid_targets_min": 4664 + }, + { + "epoch": 5.041906757464641, + "grad_norm": 0.38303237810983, + "learning_rate": 8.805652446025815e-06, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06851939111948013, + "step": 9625, + "valid_targets_mean": 6281.5, + "valid_targets_min": 4996 + }, + { + "epoch": 5.0445259298061815, + "grad_norm": 0.38777770952300034, + "learning_rate": 8.784014020222968e-06, + "loss": 0.1474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06911195814609528, + "step": 9630, + "valid_targets_mean": 6628.9, + "valid_targets_min": 4582 + }, + { + "epoch": 5.047145102147721, + "grad_norm": 0.3563359927656082, + "learning_rate": 8.762394729697022e-06, + "loss": 0.1349, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06103801727294922, + "step": 9635, + "valid_targets_mean": 6536.6, + "valid_targets_min": 5205 + }, + { + "epoch": 5.049764274489261, + "grad_norm": 0.35120735120786944, + "learning_rate": 8.740794611332038e-06, + "loss": 0.1373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07135598361492157, + "step": 9640, + "valid_targets_mean": 7031.0, + "valid_targets_min": 5098 + }, + { + "epoch": 5.0523834468308015, + "grad_norm": 0.3992926513728043, + "learning_rate": 8.719213701979348e-06, + "loss": 0.1504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08695392310619354, + "step": 9645, + "valid_targets_mean": 6514.2, + "valid_targets_min": 5016 + }, + { + "epoch": 5.055002619172342, + "grad_norm": 0.4250054952188991, + "learning_rate": 8.697652038457563e-06, + "loss": 0.169, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10635842382907867, + "step": 9650, + "valid_targets_mean": 6330.0, + "valid_targets_min": 965 + }, + { + "epoch": 5.057621791513881, + "grad_norm": 0.3664272345129127, + "learning_rate": 8.676109657552396e-06, + "loss": 0.1391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07179813832044601, + "step": 9655, + "valid_targets_mean": 6293.1, + "valid_targets_min": 5218 + }, + { + "epoch": 5.0602409638554215, + "grad_norm": 0.4089685715186158, + "learning_rate": 8.654586596016726e-06, + "loss": 0.148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06715304404497147, + "step": 9660, + "valid_targets_mean": 6957.4, + "valid_targets_min": 4213 + }, + { + "epoch": 5.062860136196962, + "grad_norm": 0.4722414415791888, + "learning_rate": 8.633082890570417e-06, + "loss": 0.1633, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0850965604186058, + "step": 9665, + "valid_targets_mean": 5888.5, + "valid_targets_min": 4923 + }, + { + "epoch": 5.065479308538502, + "grad_norm": 0.40676198745231246, + "learning_rate": 8.611598577900343e-06, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10953730344772339, + "step": 9670, + "valid_targets_mean": 7822.9, + "valid_targets_min": 4993 + }, + { + "epoch": 5.068098480880042, + "grad_norm": 0.3930340169870409, + "learning_rate": 8.590133694660278e-06, + "loss": 0.1744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08052397519350052, + "step": 9675, + "valid_targets_mean": 6098.6, + "valid_targets_min": 4872 + }, + { + "epoch": 5.070717653221582, + "grad_norm": 0.3731311088096667, + "learning_rate": 8.568688277470858e-06, + "loss": 0.1515, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07839261740446091, + "step": 9680, + "valid_targets_mean": 6892.2, + "valid_targets_min": 4706 + }, + { + "epoch": 5.073336825563122, + "grad_norm": 0.39831999006834823, + "learning_rate": 8.547262362919504e-06, + "loss": 0.153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07230527698993683, + "step": 9685, + "valid_targets_mean": 6352.0, + "valid_targets_min": 4889 + }, + { + "epoch": 5.0759559979046625, + "grad_norm": 0.4155451358038843, + "learning_rate": 8.525855987560368e-06, + "loss": 0.153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0722023993730545, + "step": 9690, + "valid_targets_mean": 5948.9, + "valid_targets_min": 3888 + }, + { + "epoch": 5.078575170246202, + "grad_norm": 0.41610785583455384, + "learning_rate": 8.504469187914242e-06, + "loss": 0.1591, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06958805024623871, + "step": 9695, + "valid_targets_mean": 5846.2, + "valid_targets_min": 4051 + }, + { + "epoch": 5.081194342587742, + "grad_norm": 0.4347135251917439, + "learning_rate": 8.483102000468569e-06, + "loss": 0.1524, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07873499393463135, + "step": 9700, + "valid_targets_mean": 7319.2, + "valid_targets_min": 5230 + }, + { + "epoch": 5.0838135149292825, + "grad_norm": 0.3819688938611039, + "learning_rate": 8.461754461677275e-06, + "loss": 0.1614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0817309096455574, + "step": 9705, + "valid_targets_mean": 8235.1, + "valid_targets_min": 4997 + }, + { + "epoch": 5.086432687270823, + "grad_norm": 0.4097446493828587, + "learning_rate": 8.440426607960821e-06, + "loss": 0.1556, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07728447020053864, + "step": 9710, + "valid_targets_mean": 6402.5, + "valid_targets_min": 5364 + }, + { + "epoch": 5.089051859612362, + "grad_norm": 0.36786655439525595, + "learning_rate": 8.419118475706032e-06, + "loss": 0.1456, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08483774214982986, + "step": 9715, + "valid_targets_mean": 7117.4, + "valid_targets_min": 3391 + }, + { + "epoch": 5.0916710319539025, + "grad_norm": 0.3558875008935785, + "learning_rate": 8.397830101266133e-06, + "loss": 0.1326, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07120934128761292, + "step": 9720, + "valid_targets_mean": 7542.8, + "valid_targets_min": 4829 + }, + { + "epoch": 5.094290204295443, + "grad_norm": 0.410557932583765, + "learning_rate": 8.376561520960603e-06, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07391059398651123, + "step": 9725, + "valid_targets_mean": 6284.4, + "valid_targets_min": 5180 + }, + { + "epoch": 5.096909376636983, + "grad_norm": 0.4278585703070951, + "learning_rate": 8.355312771075165e-06, + "loss": 0.1604, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07944594323635101, + "step": 9730, + "valid_targets_mean": 5806.2, + "valid_targets_min": 5101 + }, + { + "epoch": 5.0995285489785225, + "grad_norm": 0.41039825691911347, + "learning_rate": 8.33408388786172e-06, + "loss": 0.1553, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07008364796638489, + "step": 9735, + "valid_targets_mean": 6697.9, + "valid_targets_min": 4527 + }, + { + "epoch": 5.102147721320063, + "grad_norm": 0.3848642861216921, + "learning_rate": 8.312874907538259e-06, + "loss": 0.1483, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08209452033042908, + "step": 9740, + "valid_targets_mean": 6410.4, + "valid_targets_min": 3082 + }, + { + "epoch": 5.104766893661603, + "grad_norm": 0.3851385062138823, + "learning_rate": 8.29168586628883e-06, + "loss": 0.1409, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07119680941104889, + "step": 9745, + "valid_targets_mean": 6847.1, + "valid_targets_min": 4685 + }, + { + "epoch": 5.107386066003143, + "grad_norm": 0.40331122919175133, + "learning_rate": 8.270516800263461e-06, + "loss": 0.1536, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07771193981170654, + "step": 9750, + "valid_targets_mean": 6194.0, + "valid_targets_min": 4644 + }, + { + "epoch": 5.110005238344683, + "grad_norm": 0.3654455263195733, + "learning_rate": 8.249367745578086e-06, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0662379339337349, + "step": 9755, + "valid_targets_mean": 6283.8, + "valid_targets_min": 4433 + }, + { + "epoch": 5.112624410686223, + "grad_norm": 0.39275477682874005, + "learning_rate": 8.228238738314514e-06, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07109342515468597, + "step": 9760, + "valid_targets_mean": 5917.2, + "valid_targets_min": 4857 + }, + { + "epoch": 5.115243583027763, + "grad_norm": 0.4418496425335142, + "learning_rate": 8.207129814520349e-06, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09638451039791107, + "step": 9765, + "valid_targets_mean": 6986.4, + "valid_targets_min": 5101 + }, + { + "epoch": 5.117862755369304, + "grad_norm": 0.39706689128284944, + "learning_rate": 8.186041010208932e-06, + "loss": 0.151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.062459833920001984, + "step": 9770, + "valid_targets_mean": 5299.5, + "valid_targets_min": 4442 + }, + { + "epoch": 5.120481927710843, + "grad_norm": 0.3697232372597734, + "learning_rate": 8.164972361359274e-06, + "loss": 0.1292, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05542721226811409, + "step": 9775, + "valid_targets_mean": 6133.9, + "valid_targets_min": 5130 + }, + { + "epoch": 5.123101100052383, + "grad_norm": 0.36758741991224736, + "learning_rate": 8.143923903916e-06, + "loss": 0.1313, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07765291631221771, + "step": 9780, + "valid_targets_mean": 6477.4, + "valid_targets_min": 4162 + }, + { + "epoch": 5.125720272393924, + "grad_norm": 0.3987240830806586, + "learning_rate": 8.1228956737893e-06, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06947469711303711, + "step": 9785, + "valid_targets_mean": 6350.4, + "valid_targets_min": 3940 + }, + { + "epoch": 5.128339444735464, + "grad_norm": 0.3921966822341434, + "learning_rate": 8.10188770685482e-06, + "loss": 0.146, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08241759240627289, + "step": 9790, + "valid_targets_mean": 6625.1, + "valid_targets_min": 4642 + }, + { + "epoch": 5.130958617077003, + "grad_norm": 0.4424832185814462, + "learning_rate": 8.080900038953686e-06, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08623035252094269, + "step": 9795, + "valid_targets_mean": 4907.0, + "valid_targets_min": 4431 + }, + { + "epoch": 5.133577789418544, + "grad_norm": 0.4190066238419264, + "learning_rate": 8.05993270589234e-06, + "loss": 0.1751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0806969627737999, + "step": 9800, + "valid_targets_mean": 6465.0, + "valid_targets_min": 5333 + }, + { + "epoch": 5.136196961760084, + "grad_norm": 0.4126281342912384, + "learning_rate": 8.038985743442582e-06, + "loss": 0.144, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07737930864095688, + "step": 9805, + "valid_targets_mean": 5650.5, + "valid_targets_min": 4769 + }, + { + "epoch": 5.138816134101624, + "grad_norm": 0.4248778682938229, + "learning_rate": 8.018059187341414e-06, + "loss": 0.1509, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07351960241794586, + "step": 9810, + "valid_targets_mean": 5540.0, + "valid_targets_min": 4281 + }, + { + "epoch": 5.141435306443164, + "grad_norm": 0.5618555325728233, + "learning_rate": 7.997153073291049e-06, + "loss": 0.1546, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.107383131980896, + "step": 9815, + "valid_targets_mean": 6499.0, + "valid_targets_min": 4844 + }, + { + "epoch": 5.144054478784704, + "grad_norm": 0.40162232309599816, + "learning_rate": 7.976267436958817e-06, + "loss": 0.166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06817017495632172, + "step": 9820, + "valid_targets_mean": 5736.5, + "valid_targets_min": 5227 + }, + { + "epoch": 5.146673651126244, + "grad_norm": 0.38165815546516785, + "learning_rate": 7.955402313977115e-06, + "loss": 0.1495, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0587364099919796, + "step": 9825, + "valid_targets_mean": 5857.2, + "valid_targets_min": 2461 + }, + { + "epoch": 5.149292823467785, + "grad_norm": 0.39394755170994955, + "learning_rate": 7.934557739943333e-06, + "loss": 0.1582, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06818060576915741, + "step": 9830, + "valid_targets_mean": 5793.4, + "valid_targets_min": 2815 + }, + { + "epoch": 5.151911995809324, + "grad_norm": 0.36602462853860307, + "learning_rate": 7.91373375041982e-06, + "loss": 0.1529, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06513991951942444, + "step": 9835, + "valid_targets_mean": 6588.8, + "valid_targets_min": 4875 + }, + { + "epoch": 5.154531168150864, + "grad_norm": 0.4070722313128332, + "learning_rate": 7.89293038093379e-06, + "loss": 0.1389, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06802505254745483, + "step": 9840, + "valid_targets_mean": 5984.8, + "valid_targets_min": 4901 + }, + { + "epoch": 5.157150340492405, + "grad_norm": 0.34919848395047715, + "learning_rate": 7.87214766697729e-06, + "loss": 0.1485, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06530314683914185, + "step": 9845, + "valid_targets_mean": 6067.4, + "valid_targets_min": 3786 + }, + { + "epoch": 5.159769512833945, + "grad_norm": 0.43677013449556995, + "learning_rate": 7.851385644007103e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07367599010467529, + "step": 9850, + "valid_targets_mean": 5852.5, + "valid_targets_min": 5057 + }, + { + "epoch": 5.162388685175484, + "grad_norm": 0.3897132448525935, + "learning_rate": 7.830644347444761e-06, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07792486995458603, + "step": 9855, + "valid_targets_mean": 6362.2, + "valid_targets_min": 3952 + }, + { + "epoch": 5.165007857517025, + "grad_norm": 0.4665080680126876, + "learning_rate": 7.80992381267637e-06, + "loss": 0.154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07359066605567932, + "step": 9860, + "valid_targets_mean": 5360.2, + "valid_targets_min": 4267 + }, + { + "epoch": 5.167627029858565, + "grad_norm": 0.4269467163505103, + "learning_rate": 7.78922407505268e-06, + "loss": 0.1598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0926523506641388, + "step": 9865, + "valid_targets_mean": 6842.1, + "valid_targets_min": 4785 + }, + { + "epoch": 5.170246202200104, + "grad_norm": 0.38311175886442733, + "learning_rate": 7.768545169888909e-06, + "loss": 0.1395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06768403202295303, + "step": 9870, + "valid_targets_mean": 6188.5, + "valid_targets_min": 5438 + }, + { + "epoch": 5.172865374541645, + "grad_norm": 0.42390866433189944, + "learning_rate": 7.747887132464757e-06, + "loss": 0.1414, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08370433747768402, + "step": 9875, + "valid_targets_mean": 5848.6, + "valid_targets_min": 4734 + }, + { + "epoch": 5.175484546883185, + "grad_norm": 0.36195790511947135, + "learning_rate": 7.72724999802432e-06, + "loss": 0.1332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058599796146154404, + "step": 9880, + "valid_targets_mean": 6278.2, + "valid_targets_min": 5636 + }, + { + "epoch": 5.178103719224725, + "grad_norm": 0.4003048259119387, + "learning_rate": 7.706633801776031e-06, + "loss": 0.1518, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06315216422080994, + "step": 9885, + "valid_targets_mean": 6168.9, + "valid_targets_min": 4206 + }, + { + "epoch": 5.180722891566265, + "grad_norm": 0.386392742493416, + "learning_rate": 7.686038578892599e-06, + "loss": 0.1436, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06524616479873657, + "step": 9890, + "valid_targets_mean": 6226.5, + "valid_targets_min": 4535 + }, + { + "epoch": 5.183342063907805, + "grad_norm": 0.34214283526730643, + "learning_rate": 7.665464364510954e-06, + "loss": 0.1447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06979788839817047, + "step": 9895, + "valid_targets_mean": 7821.6, + "valid_targets_min": 5471 + }, + { + "epoch": 5.185961236249345, + "grad_norm": 0.3563196857983805, + "learning_rate": 7.644911193732185e-06, + "loss": 0.1516, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06242002919316292, + "step": 9900, + "valid_targets_mean": 6843.0, + "valid_targets_min": 5243 + }, + { + "epoch": 5.1885804085908855, + "grad_norm": 0.3989606043113984, + "learning_rate": 7.624379101621464e-06, + "loss": 0.1351, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0840473622083664, + "step": 9905, + "valid_targets_mean": 6658.6, + "valid_targets_min": 4906 + }, + { + "epoch": 5.191199580932425, + "grad_norm": 0.423777512801697, + "learning_rate": 7.603868123208023e-06, + "loss": 0.1541, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07457157969474792, + "step": 9910, + "valid_targets_mean": 6799.5, + "valid_targets_min": 5168 + }, + { + "epoch": 5.193818753273965, + "grad_norm": 0.4483238511136525, + "learning_rate": 7.583378293485059e-06, + "loss": 0.1547, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07715138792991638, + "step": 9915, + "valid_targets_mean": 5376.2, + "valid_targets_min": 3130 + }, + { + "epoch": 5.1964379256155055, + "grad_norm": 0.8386333029669407, + "learning_rate": 7.5629096474096954e-06, + "loss": 0.1614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12272357940673828, + "step": 9920, + "valid_targets_mean": 2226.6, + "valid_targets_min": 1111 + }, + { + "epoch": 5.199057097957046, + "grad_norm": 0.7326890382774619, + "learning_rate": 7.542462219902909e-06, + "loss": 0.147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08280377835035324, + "step": 9925, + "valid_targets_mean": 1648.0, + "valid_targets_min": 929 + }, + { + "epoch": 5.201676270298585, + "grad_norm": 0.7754705657641278, + "learning_rate": 7.5220360458494855e-06, + "loss": 0.1391, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0577634796500206, + "step": 9930, + "valid_targets_mean": 1430.0, + "valid_targets_min": 914 + }, + { + "epoch": 5.204295442640126, + "grad_norm": 0.7224611584403995, + "learning_rate": 7.5016311600979245e-06, + "loss": 0.1362, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05936136841773987, + "step": 9935, + "valid_targets_mean": 1400.1, + "valid_targets_min": 684 + }, + { + "epoch": 5.206914614981666, + "grad_norm": 0.7624226867341292, + "learning_rate": 7.481247597460448e-06, + "loss": 0.1381, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06537143141031265, + "step": 9940, + "valid_targets_mean": 1884.6, + "valid_targets_min": 876 + }, + { + "epoch": 5.209533787323206, + "grad_norm": 0.7902014349749125, + "learning_rate": 7.460885392712856e-06, + "loss": 0.1323, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058743298053741455, + "step": 9945, + "valid_targets_mean": 1341.0, + "valid_targets_min": 772 + }, + { + "epoch": 5.212152959664746, + "grad_norm": 0.7753376233015081, + "learning_rate": 7.440544580594553e-06, + "loss": 0.1403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058752238750457764, + "step": 9950, + "valid_targets_mean": 1462.6, + "valid_targets_min": 655 + }, + { + "epoch": 5.214772132006286, + "grad_norm": 0.7910100637676735, + "learning_rate": 7.420225195808397e-06, + "loss": 0.1246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06303153932094574, + "step": 9955, + "valid_targets_mean": 1407.8, + "valid_targets_min": 737 + }, + { + "epoch": 5.217391304347826, + "grad_norm": 0.8178167536719031, + "learning_rate": 7.399927273020748e-06, + "loss": 0.1415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057194191962480545, + "step": 9960, + "valid_targets_mean": 1260.0, + "valid_targets_min": 759 + }, + { + "epoch": 5.2200104766893665, + "grad_norm": 0.7999444274090964, + "learning_rate": 7.379650846861295e-06, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051078036427497864, + "step": 9965, + "valid_targets_mean": 1251.4, + "valid_targets_min": 750 + }, + { + "epoch": 5.222629649030906, + "grad_norm": 0.7710773167654943, + "learning_rate": 7.359395951923081e-06, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08422595262527466, + "step": 9970, + "valid_targets_mean": 2022.1, + "valid_targets_min": 927 + }, + { + "epoch": 5.225248821372446, + "grad_norm": 0.7774073867155006, + "learning_rate": 7.3391626227624145e-06, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06413210183382034, + "step": 9975, + "valid_targets_mean": 1364.9, + "valid_targets_min": 679 + }, + { + "epoch": 5.2278679937139865, + "grad_norm": 0.7869337381972866, + "learning_rate": 7.318950893898804e-06, + "loss": 0.1311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06664053350687027, + "step": 9980, + "valid_targets_mean": 1375.4, + "valid_targets_min": 771 + }, + { + "epoch": 5.230487166055527, + "grad_norm": 0.7766123709077877, + "learning_rate": 7.29876079981491e-06, + "loss": 0.1209, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07158958166837692, + "step": 9985, + "valid_targets_mean": 1776.8, + "valid_targets_min": 878 + }, + { + "epoch": 5.233106338397066, + "grad_norm": 0.7564363447217451, + "learning_rate": 7.2785923749564905e-06, + "loss": 0.1231, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0588599368929863, + "step": 9990, + "valid_targets_mean": 1513.5, + "valid_targets_min": 1103 + }, + { + "epoch": 5.2357255107386065, + "grad_norm": 0.8073202911849479, + "learning_rate": 7.2584456537323025e-06, + "loss": 0.1259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0702177882194519, + "step": 9995, + "valid_targets_mean": 1680.0, + "valid_targets_min": 821 + }, + { + "epoch": 5.238344683080147, + "grad_norm": 0.8267091483484147, + "learning_rate": 7.238320670514125e-06, + "loss": 0.1332, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08260713517665863, + "step": 10000, + "valid_targets_mean": 1960.6, + "valid_targets_min": 974 + }, + { + "epoch": 5.240963855421687, + "grad_norm": 0.7770790158825492, + "learning_rate": 7.218217459636605e-06, + "loss": 0.1307, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06306066364049911, + "step": 10005, + "valid_targets_mean": 1495.6, + "valid_targets_min": 869 + }, + { + "epoch": 5.2435830277632265, + "grad_norm": 0.7486113167393021, + "learning_rate": 7.198136055397269e-06, + "loss": 0.13, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056537315249443054, + "step": 10010, + "valid_targets_mean": 1296.5, + "valid_targets_min": 1127 + }, + { + "epoch": 5.246202200104767, + "grad_norm": 0.7524854159195199, + "learning_rate": 7.178076492056434e-06, + "loss": 0.1302, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06375785917043686, + "step": 10015, + "valid_targets_mean": 1711.9, + "valid_targets_min": 709 + }, + { + "epoch": 5.248821372446307, + "grad_norm": 0.7942396731884592, + "learning_rate": 7.1580388038371525e-06, + "loss": 0.128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058079786598682404, + "step": 10020, + "valid_targets_mean": 1241.1, + "valid_targets_min": 720 + }, + { + "epoch": 5.251440544787847, + "grad_norm": 0.8001584685827973, + "learning_rate": 7.1380230249251605e-06, + "loss": 0.1301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06254139542579651, + "step": 10025, + "valid_targets_mean": 1343.6, + "valid_targets_min": 662 + }, + { + "epoch": 5.254059717129387, + "grad_norm": 0.7562944869247711, + "learning_rate": 7.118029189468812e-06, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05077434331178665, + "step": 10030, + "valid_targets_mean": 1273.8, + "valid_targets_min": 859 + }, + { + "epoch": 5.256678889470927, + "grad_norm": 0.7516924837444345, + "learning_rate": 7.098057331579031e-06, + "loss": 0.116, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06456160545349121, + "step": 10035, + "valid_targets_mean": 1475.9, + "valid_targets_min": 950 + }, + { + "epoch": 5.259298061812467, + "grad_norm": 0.8490994380072254, + "learning_rate": 7.078107485329222e-06, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06808802485466003, + "step": 10040, + "valid_targets_mean": 1643.9, + "valid_targets_min": 686 + }, + { + "epoch": 5.261917234154008, + "grad_norm": 0.819747570963924, + "learning_rate": 7.05817968475528e-06, + "loss": 0.1251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06246265023946762, + "step": 10045, + "valid_targets_mean": 1486.2, + "valid_targets_min": 702 + }, + { + "epoch": 5.264536406495547, + "grad_norm": 0.8057669155730638, + "learning_rate": 7.038273963855442e-06, + "loss": 0.1248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05086313933134079, + "step": 10050, + "valid_targets_mean": 1281.2, + "valid_targets_min": 779 + }, + { + "epoch": 5.267155578837087, + "grad_norm": 0.8255548876472795, + "learning_rate": 7.018390356590303e-06, + "loss": 0.1358, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06761796772480011, + "step": 10055, + "valid_targets_mean": 1427.4, + "valid_targets_min": 822 + }, + { + "epoch": 5.269774751178628, + "grad_norm": 0.8083852436547692, + "learning_rate": 6.998528896882726e-06, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04791147634387016, + "step": 10060, + "valid_targets_mean": 1262.0, + "valid_targets_min": 868 + }, + { + "epoch": 5.272393923520168, + "grad_norm": 0.752390714181271, + "learning_rate": 6.978689618617784e-06, + "loss": 0.1276, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053519390523433685, + "step": 10065, + "valid_targets_mean": 1303.9, + "valid_targets_min": 765 + }, + { + "epoch": 5.275013095861707, + "grad_norm": 0.7912010665702333, + "learning_rate": 6.958872555642711e-06, + "loss": 0.1242, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06376153975725174, + "step": 10070, + "valid_targets_mean": 1463.9, + "valid_targets_min": 793 + }, + { + "epoch": 5.277632268203248, + "grad_norm": 0.795159270610698, + "learning_rate": 6.939077741766842e-06, + "loss": 0.1291, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07671696692705154, + "step": 10075, + "valid_targets_mean": 1570.0, + "valid_targets_min": 857 + }, + { + "epoch": 5.280251440544788, + "grad_norm": 0.7586798849751026, + "learning_rate": 6.9193052107615335e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057947032153606415, + "step": 10080, + "valid_targets_mean": 1509.6, + "valid_targets_min": 1219 + }, + { + "epoch": 5.282870612886328, + "grad_norm": 0.8326527348513357, + "learning_rate": 6.8995549963601675e-06, + "loss": 0.125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06129676103591919, + "step": 10085, + "valid_targets_mean": 1448.9, + "valid_targets_min": 660 + }, + { + "epoch": 5.285489785227868, + "grad_norm": 0.7693957186981843, + "learning_rate": 6.8798271322580015e-06, + "loss": 0.1137, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055658403784036636, + "step": 10090, + "valid_targets_mean": 1586.9, + "valid_targets_min": 853 + }, + { + "epoch": 5.288108957569408, + "grad_norm": 0.8774745761168313, + "learning_rate": 6.860121652112211e-06, + "loss": 0.1236, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06589289009571075, + "step": 10095, + "valid_targets_mean": 1853.2, + "valid_targets_min": 890 + }, + { + "epoch": 5.290728129910948, + "grad_norm": 0.8449626080857656, + "learning_rate": 6.840438589541733e-06, + "loss": 0.1251, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048303451389074326, + "step": 10100, + "valid_targets_mean": 1249.4, + "valid_targets_min": 765 + }, + { + "epoch": 5.293347302252489, + "grad_norm": 0.8377593232202983, + "learning_rate": 6.820777978127315e-06, + "loss": 0.121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07175786048173904, + "step": 10105, + "valid_targets_mean": 1580.0, + "valid_targets_min": 727 + }, + { + "epoch": 5.295966474594028, + "grad_norm": 0.7706779821951633, + "learning_rate": 6.801139851411347e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05244048684835434, + "step": 10110, + "valid_targets_mean": 1220.2, + "valid_targets_min": 722 + }, + { + "epoch": 5.298585646935568, + "grad_norm": 0.8125134596174877, + "learning_rate": 6.78152424289789e-06, + "loss": 0.1261, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05607043579220772, + "step": 10115, + "valid_targets_mean": 1463.0, + "valid_targets_min": 760 + }, + { + "epoch": 5.301204819277109, + "grad_norm": 0.8148234486964837, + "learning_rate": 6.761931186052579e-06, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05072994902729988, + "step": 10120, + "valid_targets_mean": 1404.2, + "valid_targets_min": 1046 + }, + { + "epoch": 5.303823991618648, + "grad_norm": 0.8283579915124643, + "learning_rate": 6.742360714302576e-06, + "loss": 0.1279, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05837526544928551, + "step": 10125, + "valid_targets_mean": 1383.8, + "valid_targets_min": 692 + }, + { + "epoch": 5.306443163960188, + "grad_norm": 0.7806275615739623, + "learning_rate": 6.7228128610365116e-06, + "loss": 0.1271, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05380524322390556, + "step": 10130, + "valid_targets_mean": 1496.4, + "valid_targets_min": 907 + }, + { + "epoch": 5.309062336301729, + "grad_norm": 0.7764979686934466, + "learning_rate": 6.703287659604429e-06, + "loss": 0.1324, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06216323748230934, + "step": 10135, + "valid_targets_mean": 1602.6, + "valid_targets_min": 1214 + }, + { + "epoch": 5.311681508643269, + "grad_norm": 0.7698105630229095, + "learning_rate": 6.683785143317707e-06, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05006994307041168, + "step": 10140, + "valid_targets_mean": 1383.2, + "valid_targets_min": 822 + }, + { + "epoch": 5.314300680984809, + "grad_norm": 0.7662834869458358, + "learning_rate": 6.6643053454490605e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06229148060083389, + "step": 10145, + "valid_targets_mean": 1601.6, + "valid_targets_min": 982 + }, + { + "epoch": 5.316919853326349, + "grad_norm": 0.8142743770439129, + "learning_rate": 6.644848299232407e-06, + "loss": 0.1258, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047794342041015625, + "step": 10150, + "valid_targets_mean": 1240.2, + "valid_targets_min": 713 + }, + { + "epoch": 5.319539025667889, + "grad_norm": 0.7707573306047922, + "learning_rate": 6.625414037862865e-06, + "loss": 0.1168, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056698039174079895, + "step": 10155, + "valid_targets_mean": 1525.5, + "valid_targets_min": 721 + }, + { + "epoch": 5.322158198009429, + "grad_norm": 0.7882250552267522, + "learning_rate": 6.606002594496683e-06, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05490151420235634, + "step": 10160, + "valid_targets_mean": 1275.1, + "valid_targets_min": 741 + }, + { + "epoch": 5.324777370350969, + "grad_norm": 0.8653629678325007, + "learning_rate": 6.586614002251173e-06, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06029641255736351, + "step": 10165, + "valid_targets_mean": 1554.5, + "valid_targets_min": 734 + }, + { + "epoch": 5.327396542692509, + "grad_norm": 0.7609800120302717, + "learning_rate": 6.567248294204669e-06, + "loss": 0.1224, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06260424107313156, + "step": 10170, + "valid_targets_mean": 1809.5, + "valid_targets_min": 744 + }, + { + "epoch": 5.330015715034049, + "grad_norm": 0.8135605283196446, + "learning_rate": 6.547905503396454e-06, + "loss": 0.1252, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06977517902851105, + "step": 10175, + "valid_targets_mean": 1836.4, + "valid_targets_min": 1295 + }, + { + "epoch": 5.3326348873755895, + "grad_norm": 0.8604110298998684, + "learning_rate": 6.528585662826725e-06, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05917800962924957, + "step": 10180, + "valid_targets_mean": 1427.8, + "valid_targets_min": 785 + }, + { + "epoch": 5.33525405971713, + "grad_norm": 0.7245215235802898, + "learning_rate": 6.5092888054564995e-06, + "loss": 0.122, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06001307815313339, + "step": 10185, + "valid_targets_mean": 1787.8, + "valid_targets_min": 740 + }, + { + "epoch": 5.337873232058669, + "grad_norm": 0.783561161605677, + "learning_rate": 6.490014964207622e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05108339339494705, + "step": 10190, + "valid_targets_mean": 1259.0, + "valid_targets_min": 711 + }, + { + "epoch": 5.34049240440021, + "grad_norm": 0.7475376127655651, + "learning_rate": 6.470764171962633e-06, + "loss": 0.123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0684993788599968, + "step": 10195, + "valid_targets_mean": 1798.9, + "valid_targets_min": 1122 + }, + { + "epoch": 5.34311157674175, + "grad_norm": 0.7498089951728653, + "learning_rate": 6.451536461564774e-06, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06576815992593765, + "step": 10200, + "valid_targets_mean": 1666.1, + "valid_targets_min": 875 + }, + { + "epoch": 5.345730749083289, + "grad_norm": 0.7758137453771179, + "learning_rate": 6.432331865817891e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051991116255521774, + "step": 10205, + "valid_targets_mean": 1226.6, + "valid_targets_min": 567 + }, + { + "epoch": 5.34834992142483, + "grad_norm": 0.7682706434272608, + "learning_rate": 6.413150417486409e-06, + "loss": 0.1123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054126303642988205, + "step": 10210, + "valid_targets_mean": 1456.9, + "valid_targets_min": 912 + }, + { + "epoch": 5.35096909376637, + "grad_norm": 0.8171250257257212, + "learning_rate": 6.3939921492952514e-06, + "loss": 0.1284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0739431381225586, + "step": 10215, + "valid_targets_mean": 1766.4, + "valid_targets_min": 1277 + }, + { + "epoch": 5.35358826610791, + "grad_norm": 0.7567769453206884, + "learning_rate": 6.374857093929802e-06, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05277364328503609, + "step": 10220, + "valid_targets_mean": 1377.2, + "valid_targets_min": 773 + }, + { + "epoch": 5.35620743844945, + "grad_norm": 0.7590441173974665, + "learning_rate": 6.355745284035835e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048600517213344574, + "step": 10225, + "valid_targets_mean": 1212.6, + "valid_targets_min": 679 + }, + { + "epoch": 5.35882661079099, + "grad_norm": 0.916009956649449, + "learning_rate": 6.336656752219479e-06, + "loss": 0.1232, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05700609087944031, + "step": 10230, + "valid_targets_mean": 1523.2, + "valid_targets_min": 818 + }, + { + "epoch": 5.36144578313253, + "grad_norm": 0.7549606242122717, + "learning_rate": 6.3175915310471155e-06, + "loss": 0.121, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05459044128656387, + "step": 10235, + "valid_targets_mean": 1375.5, + "valid_targets_min": 824 + }, + { + "epoch": 5.3640649554740705, + "grad_norm": 0.793419578210093, + "learning_rate": 6.2985496530454115e-06, + "loss": 0.1198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05644887685775757, + "step": 10240, + "valid_targets_mean": 1494.4, + "valid_targets_min": 592 + }, + { + "epoch": 5.36668412781561, + "grad_norm": 0.7888744569307411, + "learning_rate": 6.279531150701146e-06, + "loss": 0.1197, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0674830749630928, + "step": 10245, + "valid_targets_mean": 1652.6, + "valid_targets_min": 789 + }, + { + "epoch": 5.36930330015715, + "grad_norm": 0.8090353007449265, + "learning_rate": 6.260536056461275e-06, + "loss": 0.1207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05096486210823059, + "step": 10250, + "valid_targets_mean": 1337.6, + "valid_targets_min": 622 + }, + { + "epoch": 5.3719224724986905, + "grad_norm": 0.8819760620654713, + "learning_rate": 6.241564402732776e-06, + "loss": 0.1154, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05537041276693344, + "step": 10255, + "valid_targets_mean": 1221.4, + "valid_targets_min": 625 + }, + { + "epoch": 5.374541644840231, + "grad_norm": 0.8296299770424478, + "learning_rate": 6.222616221882658e-06, + "loss": 0.1248, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0559258759021759, + "step": 10260, + "valid_targets_mean": 1695.1, + "valid_targets_min": 888 + }, + { + "epoch": 5.37716081718177, + "grad_norm": 0.8185428531308253, + "learning_rate": 6.203691546237876e-06, + "loss": 0.1274, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07287395745515823, + "step": 10265, + "valid_targets_mean": 1831.0, + "valid_targets_min": 635 + }, + { + "epoch": 5.3797799895233105, + "grad_norm": 0.9315864728454666, + "learning_rate": 6.184790408085281e-06, + "loss": 0.1228, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07765045762062073, + "step": 10270, + "valid_targets_mean": 1643.6, + "valid_targets_min": 522 + }, + { + "epoch": 5.382399161864851, + "grad_norm": 0.7783667456798354, + "learning_rate": 6.165912839671577e-06, + "loss": 0.1182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05467323586344719, + "step": 10275, + "valid_targets_mean": 1264.1, + "valid_targets_min": 697 + }, + { + "epoch": 5.385018334206391, + "grad_norm": 0.7870945999549837, + "learning_rate": 6.147058873203247e-06, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06641581654548645, + "step": 10280, + "valid_targets_mean": 1721.8, + "valid_targets_min": 868 + }, + { + "epoch": 5.3876375065479305, + "grad_norm": 0.7663135500450261, + "learning_rate": 6.12822854084651e-06, + "loss": 0.1142, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05682505667209625, + "step": 10285, + "valid_targets_mean": 1489.8, + "valid_targets_min": 769 + }, + { + "epoch": 5.390256678889471, + "grad_norm": 0.8471754956032208, + "learning_rate": 6.1094218747272725e-06, + "loss": 0.1223, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05401988327503204, + "step": 10290, + "valid_targets_mean": 1321.8, + "valid_targets_min": 745 + }, + { + "epoch": 5.392875851231011, + "grad_norm": 0.8127051305524463, + "learning_rate": 6.0906389069310434e-06, + "loss": 0.1145, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05158001556992531, + "step": 10295, + "valid_targets_mean": 1332.2, + "valid_targets_min": 680 + }, + { + "epoch": 5.395495023572551, + "grad_norm": 0.7767557383969613, + "learning_rate": 6.07187966950292e-06, + "loss": 0.1212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07307745516300201, + "step": 10300, + "valid_targets_mean": 1908.4, + "valid_targets_min": 1347 + }, + { + "epoch": 5.398114195914091, + "grad_norm": 0.896746120806894, + "learning_rate": 6.053144194447507e-06, + "loss": 0.1234, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07278147339820862, + "step": 10305, + "valid_targets_mean": 1539.8, + "valid_targets_min": 851 + }, + { + "epoch": 5.400733368255631, + "grad_norm": 0.7175501741057415, + "learning_rate": 6.0344325137288675e-06, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05111266300082207, + "step": 10310, + "valid_targets_mean": 1462.5, + "valid_targets_min": 903 + }, + { + "epoch": 5.403352540597171, + "grad_norm": 0.7749373103306352, + "learning_rate": 6.015744659270475e-06, + "loss": 0.1158, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0624946653842926, + "step": 10315, + "valid_targets_mean": 1629.5, + "valid_targets_min": 848 + }, + { + "epoch": 5.405971712938712, + "grad_norm": 0.582433969996407, + "learning_rate": 5.997080662955148e-06, + "loss": 0.1157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0461689829826355, + "step": 10320, + "valid_targets_mean": 1626.6, + "valid_targets_min": 702 + }, + { + "epoch": 5.408590885280251, + "grad_norm": 0.7407116699946077, + "learning_rate": 5.978440556625009e-06, + "loss": 0.1058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05523639917373657, + "step": 10325, + "valid_targets_mean": 1578.0, + "valid_targets_min": 957 + }, + { + "epoch": 5.411210057621791, + "grad_norm": 0.7769356479597141, + "learning_rate": 5.9598243720814e-06, + "loss": 0.1222, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0678984671831131, + "step": 10330, + "valid_targets_mean": 1799.2, + "valid_targets_min": 1377 + }, + { + "epoch": 5.413829229963332, + "grad_norm": 0.7419611596301409, + "learning_rate": 5.941232141084892e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06330928206443787, + "step": 10335, + "valid_targets_mean": 1747.6, + "valid_targets_min": 870 + }, + { + "epoch": 5.416448402304872, + "grad_norm": 0.8304026455750404, + "learning_rate": 5.922663895355138e-06, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06456241011619568, + "step": 10340, + "valid_targets_mean": 1658.8, + "valid_targets_min": 822 + }, + { + "epoch": 5.419067574646411, + "grad_norm": 0.8239095135669237, + "learning_rate": 5.904119666570929e-06, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053917400538921356, + "step": 10345, + "valid_targets_mean": 1458.6, + "valid_targets_min": 873 + }, + { + "epoch": 5.421686746987952, + "grad_norm": 0.7564022608532956, + "learning_rate": 5.885599486370024e-06, + "loss": 0.1129, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05188994109630585, + "step": 10350, + "valid_targets_mean": 1291.1, + "valid_targets_min": 591 + }, + { + "epoch": 5.424305919329492, + "grad_norm": 0.7757103349497616, + "learning_rate": 5.867103386349189e-06, + "loss": 0.1192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05520738661289215, + "step": 10355, + "valid_targets_mean": 1378.5, + "valid_targets_min": 1066 + }, + { + "epoch": 5.426925091671032, + "grad_norm": 1.063171461019337, + "learning_rate": 5.8486313980640975e-06, + "loss": 0.1883, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10638885945081711, + "step": 10360, + "valid_targets_mean": 5546.2, + "valid_targets_min": 1971 + }, + { + "epoch": 5.429544264012572, + "grad_norm": 0.90371761404607, + "learning_rate": 5.830183553029285e-06, + "loss": 0.2269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1378684937953949, + "step": 10365, + "valid_targets_mean": 5648.9, + "valid_targets_min": 1631 + }, + { + "epoch": 5.432163436354112, + "grad_norm": 0.6948897390138358, + "learning_rate": 5.811759882718102e-06, + "loss": 0.2233, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12646134197711945, + "step": 10370, + "valid_targets_mean": 5411.4, + "valid_targets_min": 1509 + }, + { + "epoch": 5.434782608695652, + "grad_norm": 0.6083459112699869, + "learning_rate": 5.793360418562659e-06, + "loss": 0.2071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09969645738601685, + "step": 10375, + "valid_targets_mean": 4960.2, + "valid_targets_min": 2736 + }, + { + "epoch": 5.437401781037193, + "grad_norm": 0.576810182279233, + "learning_rate": 5.7749851919537396e-06, + "loss": 0.2111, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09488146752119064, + "step": 10380, + "valid_targets_mean": 3726.8, + "valid_targets_min": 1439 + }, + { + "epoch": 5.440020953378732, + "grad_norm": 0.5480731783743545, + "learning_rate": 5.756634234240825e-06, + "loss": 0.2173, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0950741171836853, + "step": 10385, + "valid_targets_mean": 4269.4, + "valid_targets_min": 1838 + }, + { + "epoch": 5.442640125720272, + "grad_norm": 0.49474575895449224, + "learning_rate": 5.738307576731943e-06, + "loss": 0.2008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12315724790096283, + "step": 10390, + "valid_targets_mean": 5973.6, + "valid_targets_min": 2762 + }, + { + "epoch": 5.445259298061813, + "grad_norm": 0.5236956004381352, + "learning_rate": 5.7200052506937095e-06, + "loss": 0.2133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09620815515518188, + "step": 10395, + "valid_targets_mean": 3987.1, + "valid_targets_min": 1654 + }, + { + "epoch": 5.447878470403353, + "grad_norm": 0.4801525691359519, + "learning_rate": 5.701727287351186e-06, + "loss": 0.207, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09895598888397217, + "step": 10400, + "valid_targets_mean": 5043.6, + "valid_targets_min": 2493 + }, + { + "epoch": 5.450497642744892, + "grad_norm": 0.5556800077239972, + "learning_rate": 5.6834737178879e-06, + "loss": 0.2171, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09787669777870178, + "step": 10405, + "valid_targets_mean": 4468.9, + "valid_targets_min": 874 + }, + { + "epoch": 5.453116815086433, + "grad_norm": 0.4916098022307589, + "learning_rate": 5.665244573445747e-06, + "loss": 0.2082, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09788405895233154, + "step": 10410, + "valid_targets_mean": 5241.0, + "valid_targets_min": 953 + }, + { + "epoch": 5.455735987427973, + "grad_norm": 0.5618035478671081, + "learning_rate": 5.647039885124956e-06, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13313160836696625, + "step": 10415, + "valid_targets_mean": 5277.0, + "valid_targets_min": 1265 + }, + { + "epoch": 5.458355159769513, + "grad_norm": 0.5076422217935207, + "learning_rate": 5.628859683984034e-06, + "loss": 0.2062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10731293261051178, + "step": 10420, + "valid_targets_mean": 6127.8, + "valid_targets_min": 2566 + }, + { + "epoch": 5.460974332111053, + "grad_norm": 0.5396993184924479, + "learning_rate": 5.610704001039706e-06, + "loss": 0.2079, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08992849290370941, + "step": 10425, + "valid_targets_mean": 3446.9, + "valid_targets_min": 1759 + }, + { + "epoch": 5.463593504452593, + "grad_norm": 0.5856340517558397, + "learning_rate": 5.592572867266877e-06, + "loss": 0.2004, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09708355367183685, + "step": 10430, + "valid_targets_mean": 3659.4, + "valid_targets_min": 2179 + }, + { + "epoch": 5.466212676794133, + "grad_norm": 0.5802157708950447, + "learning_rate": 5.57446631359855e-06, + "loss": 0.2006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10070739686489105, + "step": 10435, + "valid_targets_mean": 4161.5, + "valid_targets_min": 1785 + }, + { + "epoch": 5.4688318491356736, + "grad_norm": 0.5953771269649999, + "learning_rate": 5.556384370925812e-06, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11969751119613647, + "step": 10440, + "valid_targets_mean": 4467.9, + "valid_targets_min": 2345 + }, + { + "epoch": 5.471451021477213, + "grad_norm": 0.5819830242540315, + "learning_rate": 5.538327070097753e-06, + "loss": 0.2052, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1149875819683075, + "step": 10445, + "valid_targets_mean": 4057.2, + "valid_targets_min": 2767 + }, + { + "epoch": 5.474070193818753, + "grad_norm": 0.6472730162590515, + "learning_rate": 5.5202944419214236e-06, + "loss": 0.2114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10512992739677429, + "step": 10450, + "valid_targets_mean": 3450.2, + "valid_targets_min": 1205 + }, + { + "epoch": 5.476689366160294, + "grad_norm": 0.5861339164214086, + "learning_rate": 5.502286517161783e-06, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10298838466405869, + "step": 10455, + "valid_targets_mean": 3777.1, + "valid_targets_min": 2004 + }, + { + "epoch": 5.479308538501833, + "grad_norm": 0.603433469041397, + "learning_rate": 5.4843033265416426e-06, + "loss": 0.2086, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11102436482906342, + "step": 10460, + "valid_targets_mean": 4273.0, + "valid_targets_min": 1469 + }, + { + "epoch": 5.481927710843373, + "grad_norm": 0.6678418584076552, + "learning_rate": 5.466344900741616e-06, + "loss": 0.192, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11930456012487411, + "step": 10465, + "valid_targets_mean": 4196.8, + "valid_targets_min": 1889 + }, + { + "epoch": 5.484546883184914, + "grad_norm": 0.6234612238244531, + "learning_rate": 5.448411270400071e-06, + "loss": 0.1919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10475799441337585, + "step": 10470, + "valid_targets_mean": 4319.2, + "valid_targets_min": 2046 + }, + { + "epoch": 5.487166055526454, + "grad_norm": 0.7524683823716459, + "learning_rate": 5.430502466113052e-06, + "loss": 0.1896, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1156826764345169, + "step": 10475, + "valid_targets_mean": 4580.1, + "valid_targets_min": 3373 + }, + { + "epoch": 5.489785227867994, + "grad_norm": 0.6328834178974502, + "learning_rate": 5.4126185184342875e-06, + "loss": 0.2035, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10501977056264877, + "step": 10480, + "valid_targets_mean": 3786.0, + "valid_targets_min": 1747 + }, + { + "epoch": 5.492404400209534, + "grad_norm": 0.6275868892837163, + "learning_rate": 5.39475945787505e-06, + "loss": 0.2023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09050057083368301, + "step": 10485, + "valid_targets_mean": 3339.9, + "valid_targets_min": 1166 + }, + { + "epoch": 5.495023572551074, + "grad_norm": 0.7653887066093367, + "learning_rate": 5.376925314904202e-06, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09916828572750092, + "step": 10490, + "valid_targets_mean": 4114.1, + "valid_targets_min": 1537 + }, + { + "epoch": 5.497642744892614, + "grad_norm": 0.5681857124622235, + "learning_rate": 5.359116119948056e-06, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09252078831195831, + "step": 10495, + "valid_targets_mean": 4550.4, + "valid_targets_min": 3119 + }, + { + "epoch": 5.500261917234154, + "grad_norm": 0.5941366225134208, + "learning_rate": 5.341331903390381e-06, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0952102392911911, + "step": 10500, + "valid_targets_mean": 3810.8, + "valid_targets_min": 2562 + }, + { + "epoch": 5.502881089575694, + "grad_norm": 0.5930542176589342, + "learning_rate": 5.323572695572326e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10075797140598297, + "step": 10505, + "valid_targets_mean": 4045.5, + "valid_targets_min": 2659 + }, + { + "epoch": 5.505500261917234, + "grad_norm": 0.5401652038904653, + "learning_rate": 5.305838526792375e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0801267921924591, + "step": 10510, + "valid_targets_mean": 4064.8, + "valid_targets_min": 1349 + }, + { + "epoch": 5.5081194342587745, + "grad_norm": 0.6346523996231602, + "learning_rate": 5.2881294273062926e-06, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09092675894498825, + "step": 10515, + "valid_targets_mean": 3371.4, + "valid_targets_min": 2474 + }, + { + "epoch": 5.510738606600315, + "grad_norm": 0.6491916765337424, + "learning_rate": 5.270445427327071e-06, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0785207450389862, + "step": 10520, + "valid_targets_mean": 2590.5, + "valid_targets_min": 738 + }, + { + "epoch": 5.513357778941854, + "grad_norm": 0.5908791844899327, + "learning_rate": 5.252786557024885e-06, + "loss": 0.1861, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08739091455936432, + "step": 10525, + "valid_targets_mean": 3570.5, + "valid_targets_min": 2033 + }, + { + "epoch": 5.5159769512833945, + "grad_norm": 0.7371594922058565, + "learning_rate": 5.2351528465270405e-06, + "loss": 0.1936, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10885859280824661, + "step": 10530, + "valid_targets_mean": 2933.1, + "valid_targets_min": 1346 + }, + { + "epoch": 5.518596123624935, + "grad_norm": 0.7345845013449672, + "learning_rate": 5.217544325917893e-06, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10203791409730911, + "step": 10535, + "valid_targets_mean": 2991.1, + "valid_targets_min": 1821 + }, + { + "epoch": 5.521215295966474, + "grad_norm": 0.6632183424145547, + "learning_rate": 5.199961025238871e-06, + "loss": 0.1957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09681926667690277, + "step": 10540, + "valid_targets_mean": 3400.1, + "valid_targets_min": 2535 + }, + { + "epoch": 5.5238344683080145, + "grad_norm": 0.6163983658181954, + "learning_rate": 5.182402974488323e-06, + "loss": 0.1903, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09051080048084259, + "step": 10545, + "valid_targets_mean": 3411.5, + "valid_targets_min": 2352 + }, + { + "epoch": 5.526453640649555, + "grad_norm": 0.6945148683114686, + "learning_rate": 5.164870203621553e-06, + "loss": 0.1961, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09246239811182022, + "step": 10550, + "valid_targets_mean": 3319.8, + "valid_targets_min": 1153 + }, + { + "epoch": 5.529072812991095, + "grad_norm": 0.6207747154653062, + "learning_rate": 5.147362742550725e-06, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10024027526378632, + "step": 10555, + "valid_targets_mean": 3718.0, + "valid_targets_min": 2072 + }, + { + "epoch": 5.5316919853326345, + "grad_norm": 0.628372879403923, + "learning_rate": 5.129880621144827e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08635382354259491, + "step": 10560, + "valid_targets_mean": 3451.0, + "valid_targets_min": 922 + }, + { + "epoch": 5.534311157674175, + "grad_norm": 0.7001741122442561, + "learning_rate": 5.112423869229608e-06, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07553081214427948, + "step": 10565, + "valid_targets_mean": 3381.6, + "valid_targets_min": 1417 + }, + { + "epoch": 5.536930330015715, + "grad_norm": 0.6162281335037177, + "learning_rate": 5.09499251658754e-06, + "loss": 0.2005, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09244462102651596, + "step": 10570, + "valid_targets_mean": 4260.6, + "valid_targets_min": 1844 + }, + { + "epoch": 5.539549502357255, + "grad_norm": 0.6556066988158472, + "learning_rate": 5.077586592957771e-06, + "loss": 0.189, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10219431668519974, + "step": 10575, + "valid_targets_mean": 3525.2, + "valid_targets_min": 1261 + }, + { + "epoch": 5.542168674698795, + "grad_norm": 0.60657759560393, + "learning_rate": 5.060206128036034e-06, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11625860631465912, + "step": 10580, + "valid_targets_mean": 4548.6, + "valid_targets_min": 2077 + }, + { + "epoch": 5.544787847040335, + "grad_norm": 0.6188161456533431, + "learning_rate": 5.042851151474675e-06, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10895156860351562, + "step": 10585, + "valid_targets_mean": 4140.9, + "valid_targets_min": 2724 + }, + { + "epoch": 5.547407019381875, + "grad_norm": 0.5910629821526391, + "learning_rate": 5.025521692882513e-06, + "loss": 0.1884, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12794706225395203, + "step": 10590, + "valid_targets_mean": 5291.4, + "valid_targets_min": 2292 + }, + { + "epoch": 5.550026191723416, + "grad_norm": 0.6892239300392184, + "learning_rate": 5.008217781824849e-06, + "loss": 0.1938, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0943140834569931, + "step": 10595, + "valid_targets_mean": 2965.1, + "valid_targets_min": 1377 + }, + { + "epoch": 5.552645364064955, + "grad_norm": 0.606749113509467, + "learning_rate": 4.990939447823402e-06, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08866129070520401, + "step": 10600, + "valid_targets_mean": 3099.5, + "valid_targets_min": 1350 + }, + { + "epoch": 5.555264536406495, + "grad_norm": 0.6288960799120435, + "learning_rate": 4.973686720356248e-06, + "loss": 0.1949, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09150900691747665, + "step": 10605, + "valid_targets_mean": 3031.9, + "valid_targets_min": 654 + }, + { + "epoch": 5.557883708748036, + "grad_norm": 0.6402776271675373, + "learning_rate": 4.956459628857775e-06, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0792524665594101, + "step": 10610, + "valid_targets_mean": 3463.8, + "valid_targets_min": 1706 + }, + { + "epoch": 5.560502881089576, + "grad_norm": 0.5956880434022487, + "learning_rate": 4.9392582027186466e-06, + "loss": 0.1956, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0998479425907135, + "step": 10615, + "valid_targets_mean": 4105.5, + "valid_targets_min": 1739 + }, + { + "epoch": 5.563122053431115, + "grad_norm": 0.6389086875060449, + "learning_rate": 4.922082471285712e-06, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08533816039562225, + "step": 10620, + "valid_targets_mean": 2944.4, + "valid_targets_min": 1302 + }, + { + "epoch": 5.565741225772656, + "grad_norm": 0.5654606488856281, + "learning_rate": 4.9049324638620245e-06, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07984811812639236, + "step": 10625, + "valid_targets_mean": 3531.5, + "valid_targets_min": 2501 + }, + { + "epoch": 5.568360398114196, + "grad_norm": 0.6034382563674366, + "learning_rate": 4.8878082097067055e-06, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09276079386472702, + "step": 10630, + "valid_targets_mean": 4673.8, + "valid_targets_min": 3232 + }, + { + "epoch": 5.570979570455736, + "grad_norm": 0.5880316878039378, + "learning_rate": 4.870709738034982e-06, + "loss": 0.1801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09958494454622269, + "step": 10635, + "valid_targets_mean": 4465.1, + "valid_targets_min": 1101 + }, + { + "epoch": 5.573598742797276, + "grad_norm": 0.6161775039110391, + "learning_rate": 4.85363707801805e-06, + "loss": 0.2018, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10295313596725464, + "step": 10640, + "valid_targets_mean": 4113.2, + "valid_targets_min": 2079 + }, + { + "epoch": 5.576217915138816, + "grad_norm": 0.6324109116207158, + "learning_rate": 4.836590258783118e-06, + "loss": 0.1918, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09619958698749542, + "step": 10645, + "valid_targets_mean": 3891.4, + "valid_targets_min": 2138 + }, + { + "epoch": 5.578837087480356, + "grad_norm": 0.5851085764717683, + "learning_rate": 4.819569309413266e-06, + "loss": 0.1945, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09492487460374832, + "step": 10650, + "valid_targets_mean": 3470.2, + "valid_targets_min": 1234 + }, + { + "epoch": 5.581456259821897, + "grad_norm": 0.6095036275760621, + "learning_rate": 4.802574258947459e-06, + "loss": 0.1973, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09708787500858307, + "step": 10655, + "valid_targets_mean": 4038.5, + "valid_targets_min": 1822 + }, + { + "epoch": 5.584075432163436, + "grad_norm": 0.6352463231003567, + "learning_rate": 4.7856051363804715e-06, + "loss": 0.2002, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09764095395803452, + "step": 10660, + "valid_targets_mean": 4214.6, + "valid_targets_min": 1651 + }, + { + "epoch": 5.586694604504976, + "grad_norm": 0.6392191953809139, + "learning_rate": 4.768661970662847e-06, + "loss": 0.1925, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11182627081871033, + "step": 10665, + "valid_targets_mean": 3447.9, + "valid_targets_min": 1434 + }, + { + "epoch": 5.589313776846517, + "grad_norm": 0.6009026399723422, + "learning_rate": 4.751744790700841e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10528317838907242, + "step": 10670, + "valid_targets_mean": 4753.4, + "valid_targets_min": 2725 + }, + { + "epoch": 5.591932949188057, + "grad_norm": 0.7007496675306103, + "learning_rate": 4.7348536253563815e-06, + "loss": 0.182, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09295904636383057, + "step": 10675, + "valid_targets_mean": 4005.8, + "valid_targets_min": 1966 + }, + { + "epoch": 5.594552121529596, + "grad_norm": 0.6510037046708538, + "learning_rate": 4.717988503446993e-06, + "loss": 0.1807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08994663506746292, + "step": 10680, + "valid_targets_mean": 3669.0, + "valid_targets_min": 1407 + }, + { + "epoch": 5.597171293871137, + "grad_norm": 0.642042668683365, + "learning_rate": 4.7011494537458035e-06, + "loss": 0.1823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1009426862001419, + "step": 10685, + "valid_targets_mean": 4330.5, + "valid_targets_min": 1444 + }, + { + "epoch": 5.599790466212677, + "grad_norm": 0.612455396100137, + "learning_rate": 4.684336504981426e-06, + "loss": 0.1729, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07925133407115936, + "step": 10690, + "valid_targets_mean": 3077.8, + "valid_targets_min": 1504 + }, + { + "epoch": 5.602409638554217, + "grad_norm": 0.5992845565910465, + "learning_rate": 4.667549685837962e-06, + "loss": 0.1718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0706472396850586, + "step": 10695, + "valid_targets_mean": 3051.6, + "valid_targets_min": 858 + }, + { + "epoch": 5.605028810895757, + "grad_norm": 0.6723047707897838, + "learning_rate": 4.650789024954929e-06, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07880277186632156, + "step": 10700, + "valid_targets_mean": 3231.8, + "valid_targets_min": 1288 + }, + { + "epoch": 5.607647983237297, + "grad_norm": 0.6558981600539739, + "learning_rate": 4.634054550927218e-06, + "loss": 0.2074, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09159151464700699, + "step": 10705, + "valid_targets_mean": 4053.0, + "valid_targets_min": 1537 + }, + { + "epoch": 5.610267155578837, + "grad_norm": 0.6794315543318654, + "learning_rate": 4.6173462923050424e-06, + "loss": 0.1845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09614451229572296, + "step": 10710, + "valid_targets_mean": 3917.1, + "valid_targets_min": 2060 + }, + { + "epoch": 5.612886327920377, + "grad_norm": 0.6146371220196338, + "learning_rate": 4.600664277593892e-06, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10302869975566864, + "step": 10715, + "valid_targets_mean": 4482.6, + "valid_targets_min": 2515 + }, + { + "epoch": 5.615505500261917, + "grad_norm": 0.6278322264122889, + "learning_rate": 4.584008535254482e-06, + "loss": 0.1848, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09306444972753525, + "step": 10720, + "valid_targets_mean": 3847.2, + "valid_targets_min": 1744 + }, + { + "epoch": 5.618124672603457, + "grad_norm": 0.5926537989610805, + "learning_rate": 4.567379093702691e-06, + "loss": 0.1895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09181578457355499, + "step": 10725, + "valid_targets_mean": 5009.9, + "valid_targets_min": 2062 + }, + { + "epoch": 5.620743844944998, + "grad_norm": 0.6115807728453339, + "learning_rate": 4.550775981309561e-06, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08251829445362091, + "step": 10730, + "valid_targets_mean": 3550.4, + "valid_targets_min": 1181 + }, + { + "epoch": 5.623363017286538, + "grad_norm": 0.6452669401446164, + "learning_rate": 4.534199226401177e-06, + "loss": 0.1784, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10271713137626648, + "step": 10735, + "valid_targets_mean": 4010.6, + "valid_targets_min": 1937 + }, + { + "epoch": 5.625982189628077, + "grad_norm": 0.7482822013226555, + "learning_rate": 4.5176488572586765e-06, + "loss": 0.1948, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10714162141084671, + "step": 10740, + "valid_targets_mean": 3624.9, + "valid_targets_min": 1649 + }, + { + "epoch": 5.628601361969618, + "grad_norm": 0.669003508523284, + "learning_rate": 4.501124902118177e-06, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08696283400058746, + "step": 10745, + "valid_targets_mean": 2808.1, + "valid_targets_min": 1197 + }, + { + "epoch": 5.631220534311158, + "grad_norm": 0.642778651103175, + "learning_rate": 4.48462738917073e-06, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09780557453632355, + "step": 10750, + "valid_targets_mean": 3251.4, + "valid_targets_min": 1824 + }, + { + "epoch": 5.633839706652697, + "grad_norm": 0.6624363355855661, + "learning_rate": 4.468156346562282e-06, + "loss": 0.2021, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09658319503068924, + "step": 10755, + "valid_targets_mean": 3145.6, + "valid_targets_min": 538 + }, + { + "epoch": 5.636458878994238, + "grad_norm": 0.6070459984463376, + "learning_rate": 4.451711802393615e-06, + "loss": 0.1924, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0847313404083252, + "step": 10760, + "valid_targets_mean": 3597.1, + "valid_targets_min": 537 + }, + { + "epoch": 5.639078051335778, + "grad_norm": 0.7092424642117867, + "learning_rate": 4.435293784720283e-06, + "loss": 0.2058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09769661724567413, + "step": 10765, + "valid_targets_mean": 3386.5, + "valid_targets_min": 2169 + }, + { + "epoch": 5.641697223677318, + "grad_norm": 0.6355248125845707, + "learning_rate": 4.418902321552629e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09848521649837494, + "step": 10770, + "valid_targets_mean": 3687.6, + "valid_targets_min": 1625 + }, + { + "epoch": 5.6443163960188585, + "grad_norm": 0.6120506803867753, + "learning_rate": 4.4025374408556406e-06, + "loss": 0.1988, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08796226233243942, + "step": 10775, + "valid_targets_mean": 3807.0, + "valid_targets_min": 2642 + }, + { + "epoch": 5.646935568360398, + "grad_norm": 0.6688994234709845, + "learning_rate": 4.386199170549002e-06, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1190839558839798, + "step": 10780, + "valid_targets_mean": 4289.9, + "valid_targets_min": 2935 + }, + { + "epoch": 5.649554740701938, + "grad_norm": 0.5796753802986291, + "learning_rate": 4.36988753850695e-06, + "loss": 0.1907, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0992836281657219, + "step": 10785, + "valid_targets_mean": 5189.9, + "valid_targets_min": 2652 + }, + { + "epoch": 5.6521739130434785, + "grad_norm": 0.6269366324257445, + "learning_rate": 4.353602572558322e-06, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11186525225639343, + "step": 10790, + "valid_targets_mean": 4117.2, + "valid_targets_min": 648 + }, + { + "epoch": 5.654793085385018, + "grad_norm": 0.6295653522809195, + "learning_rate": 4.337344300486426e-06, + "loss": 0.1909, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08556877076625824, + "step": 10795, + "valid_targets_mean": 3492.8, + "valid_targets_min": 637 + }, + { + "epoch": 5.657412257726558, + "grad_norm": 0.6634207010550597, + "learning_rate": 4.3211127500290396e-06, + "loss": 0.1917, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09629739820957184, + "step": 10800, + "valid_targets_mean": 3412.5, + "valid_targets_min": 719 + }, + { + "epoch": 5.6600314300680985, + "grad_norm": 0.5998794325982086, + "learning_rate": 4.3049079488783584e-06, + "loss": 0.1937, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08252845704555511, + "step": 10805, + "valid_targets_mean": 4227.4, + "valid_targets_min": 2182 + }, + { + "epoch": 5.662650602409639, + "grad_norm": 0.6760489578515492, + "learning_rate": 4.288729924680935e-06, + "loss": 0.1804, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08843020349740982, + "step": 10810, + "valid_targets_mean": 3211.1, + "valid_targets_min": 1372 + }, + { + "epoch": 5.665269774751179, + "grad_norm": 0.7914056921810235, + "learning_rate": 4.272578705037638e-06, + "loss": 0.1816, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11147724837064743, + "step": 10815, + "valid_targets_mean": 4490.9, + "valid_targets_min": 2594 + }, + { + "epoch": 5.6678889470927185, + "grad_norm": 0.6177854870163906, + "learning_rate": 4.256454317503611e-06, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08226415514945984, + "step": 10820, + "valid_targets_mean": 3054.6, + "valid_targets_min": 530 + }, + { + "epoch": 5.670508119434259, + "grad_norm": 0.6731217439958592, + "learning_rate": 4.240356789588207e-06, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12775012850761414, + "step": 10825, + "valid_targets_mean": 4110.2, + "valid_targets_min": 611 + }, + { + "epoch": 5.673127291775799, + "grad_norm": 0.6543533872275946, + "learning_rate": 4.22428614875497e-06, + "loss": 0.1867, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09943042695522308, + "step": 10830, + "valid_targets_mean": 3804.2, + "valid_targets_min": 1987 + }, + { + "epoch": 5.6757464641173385, + "grad_norm": 0.6957398378200624, + "learning_rate": 4.208242422421567e-06, + "loss": 0.1876, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09107796847820282, + "step": 10835, + "valid_targets_mean": 3203.2, + "valid_targets_min": 1362 + }, + { + "epoch": 5.678365636458879, + "grad_norm": 0.5821926651243519, + "learning_rate": 4.192225637959743e-06, + "loss": 0.1868, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08990643918514252, + "step": 10840, + "valid_targets_mean": 3245.1, + "valid_targets_min": 852 + }, + { + "epoch": 5.680984808800419, + "grad_norm": 0.6555728953818382, + "learning_rate": 4.176235822695285e-06, + "loss": 0.1881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0719364583492279, + "step": 10845, + "valid_targets_mean": 2537.4, + "valid_targets_min": 1195 + }, + { + "epoch": 5.683603981141959, + "grad_norm": 0.6751150544023213, + "learning_rate": 4.1602730039079645e-06, + "loss": 0.1892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10380406677722931, + "step": 10850, + "valid_targets_mean": 4383.0, + "valid_targets_min": 1397 + }, + { + "epoch": 5.686223153483499, + "grad_norm": 0.7797308498338602, + "learning_rate": 4.144337208831499e-06, + "loss": 0.1922, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08104817569255829, + "step": 10855, + "valid_targets_mean": 3421.0, + "valid_targets_min": 1975 + }, + { + "epoch": 5.688842325825039, + "grad_norm": 0.6492396312984121, + "learning_rate": 4.128428464653484e-06, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09600323438644409, + "step": 10860, + "valid_targets_mean": 4100.6, + "valid_targets_min": 2564 + }, + { + "epoch": 5.691461498166579, + "grad_norm": 0.5397278299680731, + "learning_rate": 4.112546798515402e-06, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08865267783403397, + "step": 10865, + "valid_targets_mean": 4986.9, + "valid_targets_min": 2595 + }, + { + "epoch": 5.69408067050812, + "grad_norm": 0.590934205564606, + "learning_rate": 4.0966922375124894e-06, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1038179099559784, + "step": 10870, + "valid_targets_mean": 3992.9, + "valid_targets_min": 2845 + }, + { + "epoch": 5.696699842849659, + "grad_norm": 0.6706002087332668, + "learning_rate": 4.080864808693787e-06, + "loss": 0.1767, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06809826195240021, + "step": 10875, + "valid_targets_mean": 2390.1, + "valid_targets_min": 993 + }, + { + "epoch": 5.699319015191199, + "grad_norm": 0.579005625033608, + "learning_rate": 4.0650645390620135e-06, + "loss": 0.1736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07173579931259155, + "step": 10880, + "valid_targets_mean": 3729.1, + "valid_targets_min": 1575 + }, + { + "epoch": 5.70193818753274, + "grad_norm": 0.6604994302356326, + "learning_rate": 4.049291455573561e-06, + "loss": 0.1805, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0973813533782959, + "step": 10885, + "valid_targets_mean": 3469.0, + "valid_targets_min": 1360 + }, + { + "epoch": 5.70455735987428, + "grad_norm": 0.6100538868998756, + "learning_rate": 4.033545585138445e-06, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08820445835590363, + "step": 10890, + "valid_targets_mean": 3555.5, + "valid_targets_min": 2278 + }, + { + "epoch": 5.707176532215819, + "grad_norm": 0.6295968369885291, + "learning_rate": 4.01782695462025e-06, + "loss": 0.1964, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10447599738836288, + "step": 10895, + "valid_targets_mean": 3812.2, + "valid_targets_min": 1454 + }, + { + "epoch": 5.70979570455736, + "grad_norm": 0.6447512066853467, + "learning_rate": 4.002135590836089e-06, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09186576306819916, + "step": 10900, + "valid_targets_mean": 3796.6, + "valid_targets_min": 2344 + }, + { + "epoch": 5.7124148768989, + "grad_norm": 0.7408252439339145, + "learning_rate": 3.986471520556552e-06, + "loss": 0.183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09859302639961243, + "step": 10905, + "valid_targets_mean": 3324.1, + "valid_targets_min": 1278 + }, + { + "epoch": 5.71503404924044, + "grad_norm": 0.7306006333524507, + "learning_rate": 3.9708347705056675e-06, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10865725576877594, + "step": 10910, + "valid_targets_mean": 3381.4, + "valid_targets_min": 1858 + }, + { + "epoch": 5.71765322158198, + "grad_norm": 0.7024106650274711, + "learning_rate": 3.95522536736086e-06, + "loss": 0.1843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07249006628990173, + "step": 10915, + "valid_targets_mean": 2566.0, + "valid_targets_min": 420 + }, + { + "epoch": 5.72027239392352, + "grad_norm": 0.6829695160047269, + "learning_rate": 3.9396433377528746e-06, + "loss": 0.1819, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0833345428109169, + "step": 10920, + "valid_targets_mean": 2649.4, + "valid_targets_min": 1039 + }, + { + "epoch": 5.72289156626506, + "grad_norm": 0.6643308503800619, + "learning_rate": 3.924088708265792e-06, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08934235572814941, + "step": 10925, + "valid_targets_mean": 4475.1, + "valid_targets_min": 1737 + }, + { + "epoch": 5.725510738606601, + "grad_norm": 0.6364562383396593, + "learning_rate": 3.908561505436905e-06, + "loss": 0.1827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10796832293272018, + "step": 10930, + "valid_targets_mean": 4150.8, + "valid_targets_min": 809 + }, + { + "epoch": 5.72812991094814, + "grad_norm": 0.6100742752302343, + "learning_rate": 3.89306175575676e-06, + "loss": 0.1823, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08781400322914124, + "step": 10935, + "valid_targets_mean": 3864.1, + "valid_targets_min": 1780 + }, + { + "epoch": 5.73074908328968, + "grad_norm": 0.7412735280295316, + "learning_rate": 3.8775894856690245e-06, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08687148243188858, + "step": 10940, + "valid_targets_mean": 4058.0, + "valid_targets_min": 2185 + }, + { + "epoch": 5.733368255631221, + "grad_norm": 0.45274781108720297, + "learning_rate": 3.862144721570504e-06, + "loss": 0.1548, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04223743826150894, + "step": 10945, + "valid_targets_mean": 2736.0, + "valid_targets_min": 876 + }, + { + "epoch": 5.735987427972761, + "grad_norm": 0.5602968529454337, + "learning_rate": 3.84672748981108e-06, + "loss": 0.0955, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0356433168053627, + "step": 10950, + "valid_targets_mean": 1472.5, + "valid_targets_min": 324 + }, + { + "epoch": 5.7386066003143, + "grad_norm": 0.39052322912921966, + "learning_rate": 3.83133781669365e-06, + "loss": 0.0737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04522166773676872, + "step": 10955, + "valid_targets_mean": 2699.9, + "valid_targets_min": 1064 + }, + { + "epoch": 5.741225772655841, + "grad_norm": 0.3953305828707977, + "learning_rate": 3.815975728474106e-06, + "loss": 0.0721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038538265973329544, + "step": 10960, + "valid_targets_mean": 3137.0, + "valid_targets_min": 710 + }, + { + "epoch": 5.743844944997381, + "grad_norm": 0.32566948001942964, + "learning_rate": 3.800641251361268e-06, + "loss": 0.0866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028854478150606155, + "step": 10965, + "valid_targets_mean": 3472.9, + "valid_targets_min": 1223 + }, + { + "epoch": 5.746464117338921, + "grad_norm": 0.3970693824340154, + "learning_rate": 3.7853344115168633e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04087584465742111, + "step": 10970, + "valid_targets_mean": 3326.0, + "valid_targets_min": 1431 + }, + { + "epoch": 5.749083289680461, + "grad_norm": 0.7347475184526525, + "learning_rate": 3.7700552350554477e-06, + "loss": 0.1066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04949723929166794, + "step": 10975, + "valid_targets_mean": 1161.2, + "valid_targets_min": 559 + }, + { + "epoch": 5.751702462022001, + "grad_norm": 0.37062429917412815, + "learning_rate": 3.7548037480443977e-06, + "loss": 0.0998, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03718848526477814, + "step": 10980, + "valid_targets_mean": 3598.2, + "valid_targets_min": 1623 + }, + { + "epoch": 5.754321634363541, + "grad_norm": 0.41957714332532947, + "learning_rate": 3.7395799765038422e-06, + "loss": 0.0833, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04327414184808731, + "step": 10985, + "valid_targets_mean": 3178.2, + "valid_targets_min": 846 + }, + { + "epoch": 5.756940806705082, + "grad_norm": 0.45919357072590583, + "learning_rate": 3.724383946406631e-06, + "loss": 0.1177, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08314870297908783, + "step": 10990, + "valid_targets_mean": 3622.6, + "valid_targets_min": 1949 + }, + { + "epoch": 5.759559979046621, + "grad_norm": 0.35704699024038183, + "learning_rate": 3.709215683678282e-06, + "loss": 0.061, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03476821631193161, + "step": 10995, + "valid_targets_mean": 3344.9, + "valid_targets_min": 764 + }, + { + "epoch": 5.762179151388161, + "grad_norm": 0.43722995769096873, + "learning_rate": 3.6940752141969415e-06, + "loss": 0.0738, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0408332496881485, + "step": 11000, + "valid_targets_mean": 2391.6, + "valid_targets_min": 776 + }, + { + "epoch": 5.764798323729702, + "grad_norm": 0.387068506676206, + "learning_rate": 3.678962563793325e-06, + "loss": 0.0755, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043632425367832184, + "step": 11005, + "valid_targets_mean": 2818.5, + "valid_targets_min": 754 + }, + { + "epoch": 5.767417496071241, + "grad_norm": 0.47641045695053597, + "learning_rate": 3.6638777582507177e-06, + "loss": 0.0919, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04719253256917, + "step": 11010, + "valid_targets_mean": 1646.6, + "valid_targets_min": 635 + }, + { + "epoch": 5.770036668412781, + "grad_norm": 0.42454993282954606, + "learning_rate": 3.6488208233048594e-06, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038079530000686646, + "step": 11015, + "valid_targets_mean": 3374.8, + "valid_targets_min": 1862 + }, + { + "epoch": 5.772655840754322, + "grad_norm": 0.4110818114492606, + "learning_rate": 3.6337917846439806e-06, + "loss": 0.0726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04183412343263626, + "step": 11020, + "valid_targets_mean": 2837.4, + "valid_targets_min": 574 + }, + { + "epoch": 5.775275013095862, + "grad_norm": 0.44535778897742284, + "learning_rate": 3.6187906679086827e-06, + "loss": 0.0695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03141092509031296, + "step": 11025, + "valid_targets_mean": 2618.4, + "valid_targets_min": 523 + }, + { + "epoch": 5.777894185437402, + "grad_norm": 0.4567450056117344, + "learning_rate": 3.603817498691966e-06, + "loss": 0.0736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040859535336494446, + "step": 11030, + "valid_targets_mean": 3548.9, + "valid_targets_min": 2189 + }, + { + "epoch": 5.780513357778942, + "grad_norm": 0.37637333250965627, + "learning_rate": 3.588872302539117e-06, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027796108275651932, + "step": 11035, + "valid_targets_mean": 3564.6, + "valid_targets_min": 827 + }, + { + "epoch": 5.783132530120482, + "grad_norm": 0.4728268007078894, + "learning_rate": 3.5739551049477174e-06, + "loss": 0.0877, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04373238980770111, + "step": 11040, + "valid_targets_mean": 1936.6, + "valid_targets_min": 597 + }, + { + "epoch": 5.785751702462022, + "grad_norm": 0.5246260899544871, + "learning_rate": 3.5590659313675735e-06, + "loss": 0.0828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037010811269283295, + "step": 11045, + "valid_targets_mean": 1747.5, + "valid_targets_min": 701 + }, + { + "epoch": 5.788370874803562, + "grad_norm": 0.515495857800338, + "learning_rate": 3.5442048072006884e-06, + "loss": 0.1058, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10253942012786865, + "step": 11050, + "valid_targets_mean": 3329.6, + "valid_targets_min": 1652 + }, + { + "epoch": 5.790990047145102, + "grad_norm": 0.41294395582408694, + "learning_rate": 3.5293717578012055e-06, + "loss": 0.0863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03411465883255005, + "step": 11055, + "valid_targets_mean": 3117.1, + "valid_targets_min": 786 + }, + { + "epoch": 5.793609219486642, + "grad_norm": 0.5304924668984271, + "learning_rate": 3.514566808475375e-06, + "loss": 0.0721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04522507265210152, + "step": 11060, + "valid_targets_mean": 2574.4, + "valid_targets_min": 971 + }, + { + "epoch": 5.7962283918281825, + "grad_norm": 0.42088146287950684, + "learning_rate": 3.4997899844814874e-06, + "loss": 0.103, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03895718604326248, + "step": 11065, + "valid_targets_mean": 3320.0, + "valid_targets_min": 2610 + }, + { + "epoch": 5.798847564169723, + "grad_norm": 0.429380124994767, + "learning_rate": 3.4850413110298864e-06, + "loss": 0.0889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02835875377058983, + "step": 11070, + "valid_targets_mean": 2580.6, + "valid_targets_min": 753 + }, + { + "epoch": 5.801466736511262, + "grad_norm": 0.5510185851581492, + "learning_rate": 3.470320813282848e-06, + "loss": 0.0792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03655437007546425, + "step": 11075, + "valid_targets_mean": 1537.1, + "valid_targets_min": 619 + }, + { + "epoch": 5.8040859088528025, + "grad_norm": 0.5742451081771948, + "learning_rate": 3.4556285163546143e-06, + "loss": 0.2611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10610149800777435, + "step": 11080, + "valid_targets_mean": 2271.2, + "valid_targets_min": 987 + }, + { + "epoch": 5.806705081194343, + "grad_norm": 0.6041572892455058, + "learning_rate": 3.4409644453112857e-06, + "loss": 0.0921, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04518938809633255, + "step": 11085, + "valid_targets_mean": 1351.5, + "valid_targets_min": 329 + }, + { + "epoch": 5.809324253535882, + "grad_norm": 0.48957305209779606, + "learning_rate": 3.426328625170823e-06, + "loss": 0.0807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03558412194252014, + "step": 11090, + "valid_targets_mean": 1553.9, + "valid_targets_min": 608 + }, + { + "epoch": 5.8119434258774225, + "grad_norm": 0.47481390074543245, + "learning_rate": 3.4117210809029876e-06, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049004290252923965, + "step": 11095, + "valid_targets_mean": 3645.6, + "valid_targets_min": 873 + }, + { + "epoch": 5.814562598218963, + "grad_norm": 0.2907299273952838, + "learning_rate": 3.3971418374292963e-06, + "loss": 0.0704, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.021421954035758972, + "step": 11100, + "valid_targets_mean": 4720.6, + "valid_targets_min": 3529 + }, + { + "epoch": 5.817181770560503, + "grad_norm": 0.47527925803429083, + "learning_rate": 3.3825909196229858e-06, + "loss": 0.0828, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044109053909778595, + "step": 11105, + "valid_targets_mean": 4437.6, + "valid_targets_min": 2336 + }, + { + "epoch": 5.819800942902043, + "grad_norm": 0.4194258862455567, + "learning_rate": 3.3680683523089662e-06, + "loss": 0.0769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03383781015872955, + "step": 11110, + "valid_targets_mean": 3533.1, + "valid_targets_min": 1226 + }, + { + "epoch": 5.822420115243583, + "grad_norm": 0.39646292321212323, + "learning_rate": 3.353574160263784e-06, + "loss": 0.0722, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03398211672902107, + "step": 11115, + "valid_targets_mean": 3299.2, + "valid_targets_min": 798 + }, + { + "epoch": 5.825039287585123, + "grad_norm": 0.4315862326161756, + "learning_rate": 3.33910836821556e-06, + "loss": 0.0708, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03025813028216362, + "step": 11120, + "valid_targets_mean": 1836.9, + "valid_targets_min": 723 + }, + { + "epoch": 5.827658459926663, + "grad_norm": 0.41660979483531035, + "learning_rate": 3.3246710008439777e-06, + "loss": 0.0695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03674270957708359, + "step": 11125, + "valid_targets_mean": 3316.0, + "valid_targets_min": 898 + }, + { + "epoch": 5.830277632268203, + "grad_norm": 0.30926435439450894, + "learning_rate": 3.3102620827802247e-06, + "loss": 0.068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02628961391746998, + "step": 11130, + "valid_targets_mean": 4027.2, + "valid_targets_min": 3278 + }, + { + "epoch": 5.832896804609743, + "grad_norm": 0.4789669930493723, + "learning_rate": 3.295881638606948e-06, + "loss": 0.0707, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03222496062517166, + "step": 11135, + "valid_targets_mean": 2052.8, + "valid_targets_min": 739 + }, + { + "epoch": 5.835515976951283, + "grad_norm": 0.49519335813458293, + "learning_rate": 3.2815296928582163e-06, + "loss": 0.1023, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03633875772356987, + "step": 11140, + "valid_targets_mean": 1574.9, + "valid_targets_min": 538 + }, + { + "epoch": 5.838135149292824, + "grad_norm": 0.34615854283756625, + "learning_rate": 3.267206270019481e-06, + "loss": 0.0572, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025921868160367012, + "step": 11145, + "valid_targets_mean": 1956.9, + "valid_targets_min": 654 + }, + { + "epoch": 5.840754321634364, + "grad_norm": 0.4470604132230254, + "learning_rate": 3.252911394527527e-06, + "loss": 0.0701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04106149449944496, + "step": 11150, + "valid_targets_mean": 2735.4, + "valid_targets_min": 930 + }, + { + "epoch": 5.843373493975903, + "grad_norm": 0.46559707642320936, + "learning_rate": 3.2386450907704426e-06, + "loss": 0.0638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03752767667174339, + "step": 11155, + "valid_targets_mean": 2965.0, + "valid_targets_min": 684 + }, + { + "epoch": 5.845992666317444, + "grad_norm": 0.420075089731118, + "learning_rate": 3.2244073830875512e-06, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037872493267059326, + "step": 11160, + "valid_targets_mean": 3344.6, + "valid_targets_min": 999 + }, + { + "epoch": 5.848611838658984, + "grad_norm": 0.5504289891817628, + "learning_rate": 3.2101982957694244e-06, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04689952731132507, + "step": 11165, + "valid_targets_mean": 2445.4, + "valid_targets_min": 696 + }, + { + "epoch": 5.8512310110005235, + "grad_norm": 0.3984920841498833, + "learning_rate": 3.1960178530577645e-06, + "loss": 0.0873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03639506921172142, + "step": 11170, + "valid_targets_mean": 3514.9, + "valid_targets_min": 1068 + }, + { + "epoch": 5.853850183342064, + "grad_norm": 0.4796134811644199, + "learning_rate": 3.181866079145446e-06, + "loss": 0.0739, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034779034554958344, + "step": 11175, + "valid_targets_mean": 1602.8, + "valid_targets_min": 536 + }, + { + "epoch": 5.856469355683604, + "grad_norm": 0.5331134032401933, + "learning_rate": 3.167742998176391e-06, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09140349924564362, + "step": 11180, + "valid_targets_mean": 3568.8, + "valid_targets_min": 1507 + }, + { + "epoch": 5.859088528025144, + "grad_norm": 0.463904514329834, + "learning_rate": 3.1536486342455987e-06, + "loss": 0.0831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04113813489675522, + "step": 11185, + "valid_targets_mean": 3292.2, + "valid_targets_min": 2342 + }, + { + "epoch": 5.861707700366684, + "grad_norm": 0.590455371396696, + "learning_rate": 3.1395830113990633e-06, + "loss": 0.1006, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.039802517741918564, + "step": 11190, + "valid_targets_mean": 1567.5, + "valid_targets_min": 687 + }, + { + "epoch": 5.864326872708224, + "grad_norm": 0.5132741548808104, + "learning_rate": 3.125546153633747e-06, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026795022189617157, + "step": 11195, + "valid_targets_mean": 1145.4, + "valid_targets_min": 449 + }, + { + "epoch": 5.866946045049764, + "grad_norm": 0.36334773151804683, + "learning_rate": 3.1115380848975363e-06, + "loss": 0.0773, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029641179367899895, + "step": 11200, + "valid_targets_mean": 3191.0, + "valid_targets_min": 658 + }, + { + "epoch": 5.869565217391305, + "grad_norm": 0.5991195162430977, + "learning_rate": 3.0975588290892e-06, + "loss": 0.0817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04347830265760422, + "step": 11205, + "valid_targets_mean": 1924.8, + "valid_targets_min": 577 + }, + { + "epoch": 5.872184389732844, + "grad_norm": 0.3752108373308556, + "learning_rate": 3.083608410058356e-06, + "loss": 0.0652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032664038240909576, + "step": 11210, + "valid_targets_mean": 3992.5, + "valid_targets_min": 726 + }, + { + "epoch": 5.874803562074384, + "grad_norm": 0.4146117522797057, + "learning_rate": 3.06968685160542e-06, + "loss": 0.0632, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03724570944905281, + "step": 11215, + "valid_targets_mean": 3626.4, + "valid_targets_min": 2252 + }, + { + "epoch": 5.877422734415925, + "grad_norm": 0.46017163166684444, + "learning_rate": 3.055794177481559e-06, + "loss": 0.0709, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029016774147748947, + "step": 11220, + "valid_targets_mean": 1556.2, + "valid_targets_min": 531 + }, + { + "epoch": 5.880041906757465, + "grad_norm": 0.4614670120692765, + "learning_rate": 3.0419304113886894e-06, + "loss": 0.0879, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03797117993235588, + "step": 11225, + "valid_targets_mean": 2410.5, + "valid_targets_min": 744 + }, + { + "epoch": 5.882661079099004, + "grad_norm": 0.4547931160977343, + "learning_rate": 3.028095576979375e-06, + "loss": 0.0845, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04330388084053993, + "step": 11230, + "valid_targets_mean": 3795.1, + "valid_targets_min": 3136 + }, + { + "epoch": 5.885280251440545, + "grad_norm": 0.3284734841589199, + "learning_rate": 3.014289697856845e-06, + "loss": 0.0714, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029763588681817055, + "step": 11235, + "valid_targets_mean": 3324.4, + "valid_targets_min": 2366 + }, + { + "epoch": 5.887899423782085, + "grad_norm": 0.3839860997936193, + "learning_rate": 3.000512797574917e-06, + "loss": 0.0687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027156345546245575, + "step": 11240, + "valid_targets_mean": 1668.1, + "valid_targets_min": 669 + }, + { + "epoch": 5.890518596123625, + "grad_norm": 0.49819940900697596, + "learning_rate": 2.9867648996379706e-06, + "loss": 0.085, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04575673118233681, + "step": 11245, + "valid_targets_mean": 3529.8, + "valid_targets_min": 1785 + }, + { + "epoch": 5.893137768465165, + "grad_norm": 0.34267834657623725, + "learning_rate": 2.973046027500912e-06, + "loss": 0.0598, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03191864490509033, + "step": 11250, + "valid_targets_mean": 4120.1, + "valid_targets_min": 3862 + }, + { + "epoch": 5.895756940806705, + "grad_norm": 0.4051749199558015, + "learning_rate": 2.9593562045691104e-06, + "loss": 0.0741, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03148390352725983, + "step": 11255, + "valid_targets_mean": 3062.2, + "valid_targets_min": 539 + }, + { + "epoch": 5.898376113148245, + "grad_norm": 0.4495493302093107, + "learning_rate": 2.9456954541983982e-06, + "loss": 0.0754, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030369386076927185, + "step": 11260, + "valid_targets_mean": 2781.5, + "valid_targets_min": 1016 + }, + { + "epoch": 5.900995285489786, + "grad_norm": 0.7190006675084865, + "learning_rate": 2.932063799694984e-06, + "loss": 0.0855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04888589680194855, + "step": 11265, + "valid_targets_mean": 1102.9, + "valid_targets_min": 888 + }, + { + "epoch": 5.903614457831325, + "grad_norm": 0.5115158331041757, + "learning_rate": 2.918461264315462e-06, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028645064681768417, + "step": 11270, + "valid_targets_mean": 2602.5, + "valid_targets_min": 697 + }, + { + "epoch": 5.906233630172865, + "grad_norm": 0.6423944381816261, + "learning_rate": 2.904887871266724e-06, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04641034081578255, + "step": 11275, + "valid_targets_mean": 1592.9, + "valid_targets_min": 605 + }, + { + "epoch": 5.908852802514406, + "grad_norm": 0.38574550579583566, + "learning_rate": 2.8913436437059526e-06, + "loss": 0.0758, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03084387257695198, + "step": 11280, + "valid_targets_mean": 2899.1, + "valid_targets_min": 1117 + }, + { + "epoch": 5.911471974855946, + "grad_norm": 0.411781700698398, + "learning_rate": 2.8778286047405755e-06, + "loss": 0.0957, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0337490476667881, + "step": 11285, + "valid_targets_mean": 3376.6, + "valid_targets_min": 1388 + }, + { + "epoch": 5.914091147197485, + "grad_norm": 0.41880001873105893, + "learning_rate": 2.8643427774282174e-06, + "loss": 0.0743, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03259750083088875, + "step": 11290, + "valid_targets_mean": 3243.4, + "valid_targets_min": 524 + }, + { + "epoch": 5.916710319539026, + "grad_norm": 0.45241376587693205, + "learning_rate": 2.8508861847766665e-06, + "loss": 0.0651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03404395282268524, + "step": 11295, + "valid_targets_mean": 2468.0, + "valid_targets_min": 970 + }, + { + "epoch": 5.919329491880566, + "grad_norm": 0.5141758034556363, + "learning_rate": 2.8374588497438415e-06, + "loss": 0.0849, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0388319194316864, + "step": 11300, + "valid_targets_mean": 2959.6, + "valid_targets_min": 877 + }, + { + "epoch": 5.921948664222106, + "grad_norm": 0.4083656887019628, + "learning_rate": 2.8240607952377265e-06, + "loss": 0.0795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029686542227864265, + "step": 11305, + "valid_targets_mean": 3886.5, + "valid_targets_min": 3251 + }, + { + "epoch": 5.924567836563646, + "grad_norm": 0.5959203152852239, + "learning_rate": 2.810692044116381e-06, + "loss": 0.0651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045078739523887634, + "step": 11310, + "valid_targets_mean": 2695.6, + "valid_targets_min": 564 + }, + { + "epoch": 5.927187008905186, + "grad_norm": 0.40801392989892593, + "learning_rate": 2.797352619187834e-06, + "loss": 0.0726, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0393000990152359, + "step": 11315, + "valid_targets_mean": 4404.8, + "valid_targets_min": 2555 + }, + { + "epoch": 5.929806181246726, + "grad_norm": 0.3750414318761594, + "learning_rate": 2.7840425432101237e-06, + "loss": 0.0661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033185385167598724, + "step": 11320, + "valid_targets_mean": 4410.5, + "valid_targets_min": 3261 + }, + { + "epoch": 5.9324253535882665, + "grad_norm": 0.4360988731485157, + "learning_rate": 2.7707618388911715e-06, + "loss": 0.06, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02474169433116913, + "step": 11325, + "valid_targets_mean": 1768.5, + "valid_targets_min": 618 + }, + { + "epoch": 5.935044525929806, + "grad_norm": 0.32693371035018365, + "learning_rate": 2.757510528888836e-06, + "loss": 0.0829, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033438898622989655, + "step": 11330, + "valid_targets_mean": 3985.8, + "valid_targets_min": 3213 + }, + { + "epoch": 5.937663698271346, + "grad_norm": 0.3684767173731949, + "learning_rate": 2.7442886358107834e-06, + "loss": 0.0781, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08166645467281342, + "step": 11335, + "valid_targets_mean": 3149.1, + "valid_targets_min": 1539 + }, + { + "epoch": 5.9402828706128865, + "grad_norm": 0.345073048376949, + "learning_rate": 2.731096182214521e-06, + "loss": 0.084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025820255279541016, + "step": 11340, + "valid_targets_mean": 2094.8, + "valid_targets_min": 509 + }, + { + "epoch": 5.942902042954426, + "grad_norm": 0.33537468898855227, + "learning_rate": 2.717933190607318e-06, + "loss": 0.0803, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.019750487059354782, + "step": 11345, + "valid_targets_mean": 2441.4, + "valid_targets_min": 633 + }, + { + "epoch": 5.945521215295966, + "grad_norm": 0.426706459231261, + "learning_rate": 2.704799683446184e-06, + "loss": 0.0941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03590042144060135, + "step": 11350, + "valid_targets_mean": 2740.9, + "valid_targets_min": 737 + }, + { + "epoch": 5.9481403876375065, + "grad_norm": 0.3921109790857667, + "learning_rate": 2.691695683137827e-06, + "loss": 0.063, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030507143586874008, + "step": 11355, + "valid_targets_mean": 2726.9, + "valid_targets_min": 492 + }, + { + "epoch": 5.950759559979047, + "grad_norm": 0.5705433058207423, + "learning_rate": 2.6786212120386124e-06, + "loss": 0.0801, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0457477867603302, + "step": 11360, + "valid_targets_mean": 1607.9, + "valid_targets_min": 833 + }, + { + "epoch": 5.953378732320587, + "grad_norm": 0.36231962807371804, + "learning_rate": 2.665576292454519e-06, + "loss": 0.0655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.023260414600372314, + "step": 11365, + "valid_targets_mean": 3502.6, + "valid_targets_min": 923 + }, + { + "epoch": 5.9559979046621265, + "grad_norm": 0.663701033012781, + "learning_rate": 2.6525609466411207e-06, + "loss": 0.1029, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05070425197482109, + "step": 11370, + "valid_targets_mean": 1553.9, + "valid_targets_min": 531 + }, + { + "epoch": 5.958617077003667, + "grad_norm": 0.521198285864692, + "learning_rate": 2.6395751968035322e-06, + "loss": 0.0978, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06332775205373764, + "step": 11375, + "valid_targets_mean": 1825.6, + "valid_targets_min": 575 + }, + { + "epoch": 5.961236249345207, + "grad_norm": 0.4663314599273933, + "learning_rate": 2.6266190650963695e-06, + "loss": 0.0718, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038124896585941315, + "step": 11380, + "valid_targets_mean": 3701.9, + "valid_targets_min": 2784 + }, + { + "epoch": 5.9638554216867465, + "grad_norm": 0.38628353071030425, + "learning_rate": 2.6136925736237294e-06, + "loss": 0.07, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031040053814649582, + "step": 11385, + "valid_targets_mean": 3575.8, + "valid_targets_min": 2525 + }, + { + "epoch": 5.966474594028287, + "grad_norm": 0.38785969300380346, + "learning_rate": 2.6007957444391285e-06, + "loss": 0.0691, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.025637367740273476, + "step": 11390, + "valid_targets_mean": 3081.0, + "valid_targets_min": 970 + }, + { + "epoch": 5.969093766369827, + "grad_norm": 0.4880521031126767, + "learning_rate": 2.5879285995454904e-06, + "loss": 0.0638, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03066452592611313, + "step": 11395, + "valid_targets_mean": 1963.5, + "valid_targets_min": 817 + }, + { + "epoch": 5.971712938711367, + "grad_norm": 0.9012906952646811, + "learning_rate": 2.5750911608950756e-06, + "loss": 0.0839, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0594344288110733, + "step": 11400, + "valid_targets_mean": 1270.9, + "valid_targets_min": 637 + }, + { + "epoch": 5.974332111052908, + "grad_norm": 0.3948556091210494, + "learning_rate": 2.562283450389489e-06, + "loss": 0.0624, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031968310475349426, + "step": 11405, + "valid_targets_mean": 2594.1, + "valid_targets_min": 893 + }, + { + "epoch": 5.976951283394447, + "grad_norm": 0.4169703511713825, + "learning_rate": 2.549505489879589e-06, + "loss": 0.066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02886059507727623, + "step": 11410, + "valid_targets_mean": 3222.2, + "valid_targets_min": 1606 + }, + { + "epoch": 5.9795704557359874, + "grad_norm": 0.49222936704396925, + "learning_rate": 2.536757301165511e-06, + "loss": 0.067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029052825644612312, + "step": 11415, + "valid_targets_mean": 3023.8, + "valid_targets_min": 678 + }, + { + "epoch": 5.982189628077528, + "grad_norm": 0.37745082954088605, + "learning_rate": 2.5240389059965642e-06, + "loss": 0.0764, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.031406715512275696, + "step": 11420, + "valid_targets_mean": 4263.8, + "valid_targets_min": 1078 + }, + { + "epoch": 5.984808800419067, + "grad_norm": 0.38951827475607137, + "learning_rate": 2.5113503260712514e-06, + "loss": 0.0655, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03931698203086853, + "step": 11425, + "valid_targets_mean": 4150.0, + "valid_targets_min": 761 + }, + { + "epoch": 5.9874279727606075, + "grad_norm": 0.47461858130541124, + "learning_rate": 2.498691583037196e-06, + "loss": 0.0705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03730637952685356, + "step": 11430, + "valid_targets_mean": 1368.4, + "valid_targets_min": 538 + }, + { + "epoch": 5.990047145102148, + "grad_norm": 0.49041623928216665, + "learning_rate": 2.486062698491125e-06, + "loss": 0.0869, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07755677402019501, + "step": 11435, + "valid_targets_mean": 2630.4, + "valid_targets_min": 780 + }, + { + "epoch": 5.992666317443688, + "grad_norm": 0.3682097475921529, + "learning_rate": 2.4734636939788216e-06, + "loss": 0.141, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0306241475045681, + "step": 11440, + "valid_targets_mean": 3353.9, + "valid_targets_min": 2347 + }, + { + "epoch": 5.995285489785228, + "grad_norm": 0.41164012972971276, + "learning_rate": 2.460894590995091e-06, + "loss": 0.0692, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03697861731052399, + "step": 11445, + "valid_targets_mean": 3880.2, + "valid_targets_min": 2473 + }, + { + "epoch": 5.997904662126768, + "grad_norm": 0.39157628481336426, + "learning_rate": 2.448355410983723e-06, + "loss": 0.0674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03364294394850731, + "step": 11450, + "valid_targets_mean": 3556.4, + "valid_targets_min": 2245 + }, + { + "epoch": 6.000523834468308, + "grad_norm": 0.7635041840677032, + "learning_rate": 2.4358461753374663e-06, + "loss": 0.1194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08429177105426788, + "step": 11455, + "valid_targets_mean": 6935.0, + "valid_targets_min": 5295 + }, + { + "epoch": 6.003143006809848, + "grad_norm": 0.6853715795652247, + "learning_rate": 2.423366905397957e-06, + "loss": 0.1602, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09288565069437027, + "step": 11460, + "valid_targets_mean": 7672.9, + "valid_targets_min": 5066 + }, + { + "epoch": 6.005762179151388, + "grad_norm": 0.6832584811415545, + "learning_rate": 2.4109176224557463e-06, + "loss": 0.1554, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07927744835615158, + "step": 11465, + "valid_targets_mean": 7347.6, + "valid_targets_min": 4923 + }, + { + "epoch": 6.008381351492928, + "grad_norm": 0.5313328852311628, + "learning_rate": 2.3984983477501866e-06, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0711808055639267, + "step": 11470, + "valid_targets_mean": 5446.8, + "valid_targets_min": 3922 + }, + { + "epoch": 6.011000523834468, + "grad_norm": 0.4997046056491872, + "learning_rate": 2.3861091024694673e-06, + "loss": 0.1447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07330195605754852, + "step": 11475, + "valid_targets_mean": 6924.9, + "valid_targets_min": 4797 + }, + { + "epoch": 6.013619696176009, + "grad_norm": 0.4650974418782904, + "learning_rate": 2.373749907750518e-06, + "loss": 0.16, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0791945606470108, + "step": 11480, + "valid_targets_mean": 6443.5, + "valid_targets_min": 3853 + }, + { + "epoch": 6.016238868517548, + "grad_norm": 0.467892781786394, + "learning_rate": 2.361420784679016e-06, + "loss": 0.1531, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07518461346626282, + "step": 11485, + "valid_targets_mean": 5446.5, + "valid_targets_min": 4487 + }, + { + "epoch": 6.018858040859088, + "grad_norm": 0.4275798198215986, + "learning_rate": 2.3491217542893274e-06, + "loss": 0.1428, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07186359167098999, + "step": 11490, + "valid_targets_mean": 6210.2, + "valid_targets_min": 4137 + }, + { + "epoch": 6.021477213200629, + "grad_norm": 0.581977378572657, + "learning_rate": 2.336852837564485e-06, + "loss": 0.1717, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12328876554965973, + "step": 11495, + "valid_targets_mean": 5404.4, + "valid_targets_min": 698 + }, + { + "epoch": 6.024096385542169, + "grad_norm": 0.4047421391516061, + "learning_rate": 2.3246140554361385e-06, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08596757054328918, + "step": 11500, + "valid_targets_mean": 7197.6, + "valid_targets_min": 4499 + }, + { + "epoch": 6.026715557883708, + "grad_norm": 0.4434431195400402, + "learning_rate": 2.312405428784532e-06, + "loss": 0.1545, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10661351680755615, + "step": 11505, + "valid_targets_mean": 5538.4, + "valid_targets_min": 4291 + }, + { + "epoch": 6.029334730225249, + "grad_norm": 0.3992849940410854, + "learning_rate": 2.3002269784384533e-06, + "loss": 0.1415, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05793555825948715, + "step": 11510, + "valid_targets_mean": 5531.4, + "valid_targets_min": 4050 + }, + { + "epoch": 6.031953902566789, + "grad_norm": 0.4123827452324319, + "learning_rate": 2.2880787251752134e-06, + "loss": 0.15, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08464507758617401, + "step": 11515, + "valid_targets_mean": 7526.6, + "valid_targets_min": 5196 + }, + { + "epoch": 6.034573074908329, + "grad_norm": 0.3966454646856446, + "learning_rate": 2.2759606897206043e-06, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06937603652477264, + "step": 11520, + "valid_targets_mean": 6248.9, + "valid_targets_min": 4749 + }, + { + "epoch": 6.037192247249869, + "grad_norm": 0.4284122902210696, + "learning_rate": 2.263872892748866e-06, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.038214460015296936, + "step": 11525, + "valid_targets_mean": 3248.5, + "valid_targets_min": 1908 + }, + { + "epoch": 6.039811419591409, + "grad_norm": 0.39651014020970976, + "learning_rate": 2.251815354882647e-06, + "loss": 0.1367, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06964430958032608, + "step": 11530, + "valid_targets_mean": 6225.6, + "valid_targets_min": 4808 + }, + { + "epoch": 6.042430591932949, + "grad_norm": 0.4186323822718657, + "learning_rate": 2.2397880966929718e-06, + "loss": 0.1458, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07343867421150208, + "step": 11535, + "valid_targets_mean": 5383.0, + "valid_targets_min": 3774 + }, + { + "epoch": 6.04504976427449, + "grad_norm": 0.39504982322902155, + "learning_rate": 2.227791138699211e-06, + "loss": 0.1416, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07207377254962921, + "step": 11540, + "valid_targets_mean": 6291.4, + "valid_targets_min": 5335 + }, + { + "epoch": 6.047668936616029, + "grad_norm": 0.3866475606519103, + "learning_rate": 2.2158245013690216e-06, + "loss": 0.1275, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07138402760028839, + "step": 11545, + "valid_targets_mean": 7749.1, + "valid_targets_min": 6205 + }, + { + "epoch": 6.050288108957569, + "grad_norm": 0.37280615374122056, + "learning_rate": 2.2038882051183606e-06, + "loss": 0.1356, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06929509341716766, + "step": 11550, + "valid_targets_mean": 6882.1, + "valid_targets_min": 4428 + }, + { + "epoch": 6.05290728129911, + "grad_norm": 0.41725884744359426, + "learning_rate": 2.1919822703113945e-06, + "loss": 0.1525, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07632793486118317, + "step": 11555, + "valid_targets_mean": 5751.8, + "valid_targets_min": 4458 + }, + { + "epoch": 6.05552645364065, + "grad_norm": 0.37060182861886776, + "learning_rate": 2.180106717260515e-06, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0640210211277008, + "step": 11560, + "valid_targets_mean": 6003.9, + "valid_targets_min": 4458 + }, + { + "epoch": 6.058145625982189, + "grad_norm": 0.41749330601499113, + "learning_rate": 2.168261566226255e-06, + "loss": 0.1376, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06407776474952698, + "step": 11565, + "valid_targets_mean": 6033.8, + "valid_targets_min": 4066 + }, + { + "epoch": 6.06076479832373, + "grad_norm": 0.4257257090543421, + "learning_rate": 2.1564468374172986e-06, + "loss": 0.1446, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07927943766117096, + "step": 11570, + "valid_targets_mean": 6601.0, + "valid_targets_min": 4708 + }, + { + "epoch": 6.06338397066527, + "grad_norm": 0.4302258917827767, + "learning_rate": 2.144662550990422e-06, + "loss": 0.1594, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08398758620023727, + "step": 11575, + "valid_targets_mean": 6092.1, + "valid_targets_min": 4789 + }, + { + "epoch": 6.06600314300681, + "grad_norm": 0.4191905189849291, + "learning_rate": 2.1329087270504577e-06, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08341837674379349, + "step": 11580, + "valid_targets_mean": 6474.6, + "valid_targets_min": 4224 + }, + { + "epoch": 6.06862231534835, + "grad_norm": 0.42107837698030154, + "learning_rate": 2.1211853856502794e-06, + "loss": 0.1654, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07812417298555374, + "step": 11585, + "valid_targets_mean": 6312.4, + "valid_targets_min": 4617 + }, + { + "epoch": 6.07124148768989, + "grad_norm": 0.4077369014950039, + "learning_rate": 2.1094925467907456e-06, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07644317299127579, + "step": 11590, + "valid_targets_mean": 6333.8, + "valid_targets_min": 4783 + }, + { + "epoch": 6.07386066003143, + "grad_norm": 0.3950929630087697, + "learning_rate": 2.0978302304206766e-06, + "loss": 0.1463, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07442428171634674, + "step": 11595, + "valid_targets_mean": 7380.1, + "valid_targets_min": 6020 + }, + { + "epoch": 6.0764798323729705, + "grad_norm": 0.43151569284994723, + "learning_rate": 2.0861984564368297e-06, + "loss": 0.1527, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07998433709144592, + "step": 11600, + "valid_targets_mean": 6009.0, + "valid_targets_min": 3049 + }, + { + "epoch": 6.07909900471451, + "grad_norm": 0.40491470375477356, + "learning_rate": 2.0745972446838294e-06, + "loss": 0.15, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06832192093133926, + "step": 11605, + "valid_targets_mean": 5973.1, + "valid_targets_min": 4822 + }, + { + "epoch": 6.08171817705605, + "grad_norm": 0.4283898815299016, + "learning_rate": 2.0630266149541934e-06, + "loss": 0.1492, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06974171102046967, + "step": 11610, + "valid_targets_mean": 5754.9, + "valid_targets_min": 2910 + }, + { + "epoch": 6.0843373493975905, + "grad_norm": 0.42404554459164345, + "learning_rate": 2.0514865869882295e-06, + "loss": 0.157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08622089773416519, + "step": 11615, + "valid_targets_mean": 6414.6, + "valid_targets_min": 4791 + }, + { + "epoch": 6.086956521739131, + "grad_norm": 0.4285451130125846, + "learning_rate": 2.039977180474071e-06, + "loss": 0.1493, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07590167224407196, + "step": 11620, + "valid_targets_mean": 5675.9, + "valid_targets_min": 3983 + }, + { + "epoch": 6.08957569408067, + "grad_norm": 0.3607337378177125, + "learning_rate": 2.0284984150475795e-06, + "loss": 0.135, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06058434396982193, + "step": 11625, + "valid_targets_mean": 6975.5, + "valid_targets_min": 4383 + }, + { + "epoch": 6.0921948664222105, + "grad_norm": 0.3857821688003951, + "learning_rate": 2.0170503102923544e-06, + "loss": 0.1301, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07114368677139282, + "step": 11630, + "valid_targets_mean": 6417.1, + "valid_targets_min": 3182 + }, + { + "epoch": 6.094814038763751, + "grad_norm": 0.43896323433695666, + "learning_rate": 2.0056328857396856e-06, + "loss": 0.1558, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.082911916077137, + "step": 11635, + "valid_targets_mean": 6218.6, + "valid_targets_min": 5077 + }, + { + "epoch": 6.097433211105291, + "grad_norm": 0.4650857189291461, + "learning_rate": 1.994246160868516e-06, + "loss": 0.1552, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07567033171653748, + "step": 11640, + "valid_targets_mean": 4774.9, + "valid_targets_min": 3663 + }, + { + "epoch": 6.1000523834468305, + "grad_norm": 0.4929320452045988, + "learning_rate": 1.9828901551054193e-06, + "loss": 0.1447, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07487622648477554, + "step": 11645, + "valid_targets_mean": 6316.8, + "valid_targets_min": 4105 + }, + { + "epoch": 6.102671555788371, + "grad_norm": 0.39618729704248107, + "learning_rate": 1.9715648878245532e-06, + "loss": 0.1457, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06671681255102158, + "step": 11650, + "valid_targets_mean": 6528.9, + "valid_targets_min": 4537 + }, + { + "epoch": 6.105290728129911, + "grad_norm": 0.40322190183421525, + "learning_rate": 1.960270378347642e-06, + "loss": 0.1352, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06754016876220703, + "step": 11655, + "valid_targets_mean": 6416.8, + "valid_targets_min": 2494 + }, + { + "epoch": 6.107909900471451, + "grad_norm": 0.39902395840715554, + "learning_rate": 1.9490066459439184e-06, + "loss": 0.1472, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07599892467260361, + "step": 11660, + "valid_targets_mean": 6070.9, + "valid_targets_min": 4921 + }, + { + "epoch": 6.110529072812991, + "grad_norm": 0.40814362386536235, + "learning_rate": 1.9377737098301263e-06, + "loss": 0.1532, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07961056381464005, + "step": 11665, + "valid_targets_mean": 6701.4, + "valid_targets_min": 5285 + }, + { + "epoch": 6.113148245154531, + "grad_norm": 0.38451422611213415, + "learning_rate": 1.9265715891704583e-06, + "loss": 0.139, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0686163455247879, + "step": 11670, + "valid_targets_mean": 6273.9, + "valid_targets_min": 4464 + }, + { + "epoch": 6.1157674174960714, + "grad_norm": 0.43338852509815573, + "learning_rate": 1.9154003030765377e-06, + "loss": 0.1568, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06851521134376526, + "step": 11675, + "valid_targets_mean": 5554.8, + "valid_targets_min": 4439 + }, + { + "epoch": 6.118386589837612, + "grad_norm": 0.396705010862356, + "learning_rate": 1.9042598706073767e-06, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06214672699570656, + "step": 11680, + "valid_targets_mean": 6043.0, + "valid_targets_min": 4686 + }, + { + "epoch": 6.121005762179151, + "grad_norm": 0.3949250217763038, + "learning_rate": 1.8931503107693605e-06, + "loss": 0.1259, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06108330190181732, + "step": 11685, + "valid_targets_mean": 5912.5, + "valid_targets_min": 3381 + }, + { + "epoch": 6.1236249345206915, + "grad_norm": 0.3903360118144714, + "learning_rate": 1.8820716425161812e-06, + "loss": 0.1284, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0668412297964096, + "step": 11690, + "valid_targets_mean": 6495.5, + "valid_targets_min": 5309 + }, + { + "epoch": 6.126244106862232, + "grad_norm": 0.41718303959093767, + "learning_rate": 1.8710238847488572e-06, + "loss": 0.1341, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07565851509571075, + "step": 11695, + "valid_targets_mean": 6483.9, + "valid_targets_min": 5161 + }, + { + "epoch": 6.128863279203772, + "grad_norm": 0.4160261819354094, + "learning_rate": 1.8600070563156425e-06, + "loss": 0.1474, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07268711179494858, + "step": 11700, + "valid_targets_mean": 5958.9, + "valid_targets_min": 4666 + }, + { + "epoch": 6.1314824515453115, + "grad_norm": 0.5475209517621088, + "learning_rate": 1.8490211760120513e-06, + "loss": 0.1701, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09618830680847168, + "step": 11705, + "valid_targets_mean": 1691.5, + "valid_targets_min": 1440 + }, + { + "epoch": 6.134101623886852, + "grad_norm": 0.43461666805191196, + "learning_rate": 1.8380662625807665e-06, + "loss": 0.1588, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06448307633399963, + "step": 11710, + "valid_targets_mean": 5299.1, + "valid_targets_min": 4547 + }, + { + "epoch": 6.136720796228392, + "grad_norm": 0.3974472972992945, + "learning_rate": 1.8271423347116757e-06, + "loss": 0.1382, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07221479713916779, + "step": 11715, + "valid_targets_mean": 6003.5, + "valid_targets_min": 5031 + }, + { + "epoch": 6.1393399685699315, + "grad_norm": 0.38406936839781486, + "learning_rate": 1.8162494110417683e-06, + "loss": 0.1453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0674646869301796, + "step": 11720, + "valid_targets_mean": 8169.6, + "valid_targets_min": 5487 + }, + { + "epoch": 6.141959140911472, + "grad_norm": 0.41233465492117627, + "learning_rate": 1.805387510155161e-06, + "loss": 0.1508, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0686035305261612, + "step": 11725, + "valid_targets_mean": 6474.9, + "valid_targets_min": 4253 + }, + { + "epoch": 6.144578313253012, + "grad_norm": 0.4150816767805394, + "learning_rate": 1.7945566505830325e-06, + "loss": 0.1611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08474124222993851, + "step": 11730, + "valid_targets_mean": 7021.0, + "valid_targets_min": 4990 + }, + { + "epoch": 6.147197485594552, + "grad_norm": 0.4489944065453429, + "learning_rate": 1.7837568508036108e-06, + "loss": 0.1475, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08974331617355347, + "step": 11735, + "valid_targets_mean": 6356.0, + "valid_targets_min": 4385 + }, + { + "epoch": 6.149816657936092, + "grad_norm": 0.38251335298690886, + "learning_rate": 1.7729881292421236e-06, + "loss": 0.1504, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07843641936779022, + "step": 11740, + "valid_targets_mean": 6815.1, + "valid_targets_min": 5164 + }, + { + "epoch": 6.152435830277632, + "grad_norm": 0.38299409271369916, + "learning_rate": 1.7622505042707927e-06, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054671477526426315, + "step": 11745, + "valid_targets_mean": 5588.5, + "valid_targets_min": 4098 + }, + { + "epoch": 6.155055002619172, + "grad_norm": 0.3849649968314433, + "learning_rate": 1.7515439942087597e-06, + "loss": 0.1365, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0744810700416565, + "step": 11750, + "valid_targets_mean": 6167.6, + "valid_targets_min": 4801 + }, + { + "epoch": 6.157674174960713, + "grad_norm": 0.4659826568819937, + "learning_rate": 1.7408686173221179e-06, + "loss": 0.1696, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.18148547410964966, + "step": 11755, + "valid_targets_mean": 6224.9, + "valid_targets_min": 4772 + }, + { + "epoch": 6.160293347302252, + "grad_norm": 0.39245727324645113, + "learning_rate": 1.7302243918238137e-06, + "loss": 0.1674, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08372463285923004, + "step": 11760, + "valid_targets_mean": 6549.2, + "valid_targets_min": 5025 + }, + { + "epoch": 6.162912519643792, + "grad_norm": 0.4236546968405094, + "learning_rate": 1.7196113358736655e-06, + "loss": 0.1374, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08350541442632675, + "step": 11765, + "valid_targets_mean": 6680.8, + "valid_targets_min": 4818 + }, + { + "epoch": 6.165531691985333, + "grad_norm": 0.4252807871771535, + "learning_rate": 1.7090294675783026e-06, + "loss": 0.1473, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07304557412862778, + "step": 11770, + "valid_targets_mean": 5835.1, + "valid_targets_min": 4111 + }, + { + "epoch": 6.168150864326873, + "grad_norm": 0.41961395095715426, + "learning_rate": 1.6984788049911572e-06, + "loss": 0.151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07110430300235748, + "step": 11775, + "valid_targets_mean": 7029.9, + "valid_targets_min": 2275 + }, + { + "epoch": 6.170770036668412, + "grad_norm": 0.3842603197642495, + "learning_rate": 1.687959366112415e-06, + "loss": 0.1311, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06225533410906792, + "step": 11780, + "valid_targets_mean": 5621.2, + "valid_targets_min": 4826 + }, + { + "epoch": 6.173389209009953, + "grad_norm": 0.4129683822346252, + "learning_rate": 1.6774711688889933e-06, + "loss": 0.1421, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06798767298460007, + "step": 11785, + "valid_targets_mean": 6858.8, + "valid_targets_min": 4710 + }, + { + "epoch": 6.176008381351493, + "grad_norm": 0.37739841344583, + "learning_rate": 1.667014231214512e-06, + "loss": 0.1322, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10182426869869232, + "step": 11790, + "valid_targets_mean": 7012.5, + "valid_targets_min": 5421 + }, + { + "epoch": 6.178627553693033, + "grad_norm": 0.3871077219022217, + "learning_rate": 1.6565885709292496e-06, + "loss": 0.1373, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06399504840373993, + "step": 11795, + "valid_targets_mean": 6967.6, + "valid_targets_min": 4949 + }, + { + "epoch": 6.181246726034573, + "grad_norm": 0.38958123446250925, + "learning_rate": 1.6461942058201397e-06, + "loss": 0.1395, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07286036014556885, + "step": 11800, + "valid_targets_mean": 6348.5, + "valid_targets_min": 4831 + }, + { + "epoch": 6.183865898376113, + "grad_norm": 0.40593610363280064, + "learning_rate": 1.6358311536207084e-06, + "loss": 0.1453, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0912897139787674, + "step": 11805, + "valid_targets_mean": 6597.2, + "valid_targets_min": 4018 + }, + { + "epoch": 6.186485070717653, + "grad_norm": 0.31799328586360026, + "learning_rate": 1.6254994320110707e-06, + "loss": 0.1361, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05004189908504486, + "step": 11810, + "valid_targets_mean": 6420.9, + "valid_targets_min": 5034 + }, + { + "epoch": 6.189104243059194, + "grad_norm": 0.40361880302184194, + "learning_rate": 1.61519905861788e-06, + "loss": 0.1386, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08185481280088425, + "step": 11815, + "valid_targets_mean": 6962.4, + "valid_targets_min": 5442 + }, + { + "epoch": 6.191723415400733, + "grad_norm": 0.4319415525968686, + "learning_rate": 1.6049300510143173e-06, + "loss": 0.149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07399599254131317, + "step": 11820, + "valid_targets_mean": 5610.9, + "valid_targets_min": 4609 + }, + { + "epoch": 6.194342587742273, + "grad_norm": 0.4516256716879042, + "learning_rate": 1.5946924267200437e-06, + "loss": 0.1505, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07311864197254181, + "step": 11825, + "valid_targets_mean": 5038.4, + "valid_targets_min": 2301 + }, + { + "epoch": 6.196961760083814, + "grad_norm": 0.824034701811811, + "learning_rate": 1.584486203201181e-06, + "loss": 0.1512, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07870453596115112, + "step": 11830, + "valid_targets_mean": 1241.9, + "valid_targets_min": 788 + }, + { + "epoch": 6.199580932425354, + "grad_norm": 0.767108493583537, + "learning_rate": 1.5743113978702807e-06, + "loss": 0.1375, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07108853757381439, + "step": 11835, + "valid_targets_mean": 1518.8, + "valid_targets_min": 1244 + }, + { + "epoch": 6.202200104766893, + "grad_norm": 0.7142169034194359, + "learning_rate": 1.5641680280862947e-06, + "loss": 0.1299, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06787177920341492, + "step": 11840, + "valid_targets_mean": 1552.4, + "valid_targets_min": 796 + }, + { + "epoch": 6.204819277108434, + "grad_norm": 0.7329697771543124, + "learning_rate": 1.5540561111545272e-06, + "loss": 0.1283, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06166737899184227, + "step": 11845, + "valid_targets_mean": 1342.1, + "valid_targets_min": 697 + }, + { + "epoch": 6.207438449449974, + "grad_norm": 0.7068878550133145, + "learning_rate": 1.5439756643266468e-06, + "loss": 0.1338, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07681246846914291, + "step": 11850, + "valid_targets_mean": 1741.4, + "valid_targets_min": 638 + }, + { + "epoch": 6.210057621791514, + "grad_norm": 0.6949506439763343, + "learning_rate": 1.5339267048006078e-06, + "loss": 0.1246, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06340840458869934, + "step": 11855, + "valid_targets_mean": 1658.0, + "valid_targets_min": 727 + }, + { + "epoch": 6.212676794133054, + "grad_norm": 0.7242968232972786, + "learning_rate": 1.5239092497206697e-06, + "loss": 0.1314, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06681711226701736, + "step": 11860, + "valid_targets_mean": 1523.4, + "valid_targets_min": 1025 + }, + { + "epoch": 6.215295966474594, + "grad_norm": 0.7995787832735743, + "learning_rate": 1.5139233161773193e-06, + "loss": 0.1155, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06490489095449448, + "step": 11865, + "valid_targets_mean": 1389.6, + "valid_targets_min": 714 + }, + { + "epoch": 6.217915138816134, + "grad_norm": 0.745760853507401, + "learning_rate": 1.5039689212072772e-06, + "loss": 0.1348, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05875927582383156, + "step": 11870, + "valid_targets_mean": 1423.6, + "valid_targets_min": 944 + }, + { + "epoch": 6.2205343111576745, + "grad_norm": 0.7587926185480729, + "learning_rate": 1.4940460817934587e-06, + "loss": 0.1305, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0633007362484932, + "step": 11875, + "valid_targets_mean": 1938.8, + "valid_targets_min": 790 + }, + { + "epoch": 6.223153483499214, + "grad_norm": 0.7846618910521161, + "learning_rate": 1.4841548148649376e-06, + "loss": 0.1215, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06577780842781067, + "step": 11880, + "valid_targets_mean": 1524.5, + "valid_targets_min": 896 + }, + { + "epoch": 6.225772655840754, + "grad_norm": 0.7474114872098311, + "learning_rate": 1.4742951372969262e-06, + "loss": 0.1263, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06239018589258194, + "step": 11885, + "valid_targets_mean": 1766.2, + "valid_targets_min": 1027 + }, + { + "epoch": 6.2283918281822945, + "grad_norm": 0.6915658838565912, + "learning_rate": 1.4644670659107419e-06, + "loss": 0.1199, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054799459874629974, + "step": 11890, + "valid_targets_mean": 1471.9, + "valid_targets_min": 763 + }, + { + "epoch": 6.231011000523835, + "grad_norm": 0.8034093865902141, + "learning_rate": 1.454670617473779e-06, + "loss": 0.1148, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049490250647068024, + "step": 11895, + "valid_targets_mean": 998.5, + "valid_targets_min": 659 + }, + { + "epoch": 6.233630172865374, + "grad_norm": 0.7289906278108554, + "learning_rate": 1.4449058086994839e-06, + "loss": 0.1123, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060606542974710464, + "step": 11900, + "valid_targets_mean": 1568.8, + "valid_targets_min": 744 + }, + { + "epoch": 6.2362493452069145, + "grad_norm": 0.8621652453345776, + "learning_rate": 1.4351726562473166e-06, + "loss": 0.1184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056785475462675095, + "step": 11905, + "valid_targets_mean": 1345.0, + "valid_targets_min": 874 + }, + { + "epoch": 6.238868517548455, + "grad_norm": 0.8732654726511692, + "learning_rate": 1.4254711767227347e-06, + "loss": 0.1288, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06115246191620827, + "step": 11910, + "valid_targets_mean": 1335.8, + "valid_targets_min": 798 + }, + { + "epoch": 6.241487689889995, + "grad_norm": 0.8670396138847691, + "learning_rate": 1.4158013866771624e-06, + "loss": 0.1249, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07378070056438446, + "step": 11915, + "valid_targets_mean": 1635.2, + "valid_targets_min": 802 + }, + { + "epoch": 6.2441068622315345, + "grad_norm": 0.7960833278946176, + "learning_rate": 1.4061633026079501e-06, + "loss": 0.1226, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06942878663539886, + "step": 11920, + "valid_targets_mean": 2001.0, + "valid_targets_min": 940 + }, + { + "epoch": 6.246726034573075, + "grad_norm": 0.744882756420173, + "learning_rate": 1.3965569409583646e-06, + "loss": 0.1185, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049470677971839905, + "step": 11925, + "valid_targets_mean": 1358.5, + "valid_targets_min": 776 + }, + { + "epoch": 6.249345206914615, + "grad_norm": 0.8329318828109886, + "learning_rate": 1.3869823181175468e-06, + "loss": 0.1212, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05968695878982544, + "step": 11930, + "valid_targets_mean": 1450.4, + "valid_targets_min": 640 + }, + { + "epoch": 6.2519643792561554, + "grad_norm": 0.7783909092814958, + "learning_rate": 1.3774394504204925e-06, + "loss": 0.1218, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06345199048519135, + "step": 11935, + "valid_targets_mean": 1593.1, + "valid_targets_min": 693 + }, + { + "epoch": 6.254583551597695, + "grad_norm": 0.7910341657378867, + "learning_rate": 1.3679283541480116e-06, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0432841032743454, + "step": 11940, + "valid_targets_mean": 1005.0, + "valid_targets_min": 634 + }, + { + "epoch": 6.257202723939235, + "grad_norm": 1.1936307297280846, + "learning_rate": 1.3584490455267263e-06, + "loss": 0.1133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05266657471656799, + "step": 11945, + "valid_targets_mean": 1430.8, + "valid_targets_min": 793 + }, + { + "epoch": 6.2598218962807755, + "grad_norm": 0.843722782583509, + "learning_rate": 1.349001540729009e-06, + "loss": 0.1147, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061245668679475784, + "step": 11950, + "valid_targets_mean": 1673.6, + "valid_targets_min": 563 + }, + { + "epoch": 6.262441068622316, + "grad_norm": 0.7987635935599736, + "learning_rate": 1.339585855872989e-06, + "loss": 0.1175, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054979611188173294, + "step": 11955, + "valid_targets_mean": 1586.8, + "valid_targets_min": 837 + }, + { + "epoch": 6.265060240963855, + "grad_norm": 0.8525572473307433, + "learning_rate": 1.3302020070224942e-06, + "loss": 0.1204, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05749509483575821, + "step": 11960, + "valid_targets_mean": 1620.4, + "valid_targets_min": 1107 + }, + { + "epoch": 6.2676794133053955, + "grad_norm": 0.9968884137146439, + "learning_rate": 1.3208500101870446e-06, + "loss": 0.1269, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04865128546953201, + "step": 11965, + "valid_targets_mean": 1290.5, + "valid_targets_min": 680 + }, + { + "epoch": 6.270298585646936, + "grad_norm": 0.8047408328393952, + "learning_rate": 1.3115298813218201e-06, + "loss": 0.1127, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07513408362865448, + "step": 11970, + "valid_targets_mean": 1927.4, + "valid_targets_min": 933 + }, + { + "epoch": 6.272917757988475, + "grad_norm": 0.8066905146976031, + "learning_rate": 1.302241636327628e-06, + "loss": 0.12, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06284064799547195, + "step": 11975, + "valid_targets_mean": 1654.8, + "valid_targets_min": 631 + }, + { + "epoch": 6.2755369303300155, + "grad_norm": 0.7309992104805626, + "learning_rate": 1.2929852910508811e-06, + "loss": 0.1157, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055158913135528564, + "step": 11980, + "valid_targets_mean": 1519.0, + "valid_targets_min": 804 + }, + { + "epoch": 6.278156102671556, + "grad_norm": 0.7547680678570587, + "learning_rate": 1.2837608612835717e-06, + "loss": 0.1198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0578368604183197, + "step": 11985, + "valid_targets_mean": 1539.5, + "valid_targets_min": 1019 + }, + { + "epoch": 6.280775275013096, + "grad_norm": 0.821570826864286, + "learning_rate": 1.274568362763231e-06, + "loss": 0.1128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046871982514858246, + "step": 11990, + "valid_targets_mean": 1112.8, + "valid_targets_min": 856 + }, + { + "epoch": 6.283394447354636, + "grad_norm": 0.7757757718848503, + "learning_rate": 1.2654078111729318e-06, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07132439315319061, + "step": 11995, + "valid_targets_mean": 1914.2, + "valid_targets_min": 1168 + }, + { + "epoch": 6.286013619696176, + "grad_norm": 0.8873179333781386, + "learning_rate": 1.256279222141219e-06, + "loss": 0.1062, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06316615641117096, + "step": 12000, + "valid_targets_mean": 1613.0, + "valid_targets_min": 678 + }, + { + "epoch": 6.288632792037716, + "grad_norm": 0.8769382523563896, + "learning_rate": 1.2471826112421348e-06, + "loss": 0.1172, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07142555713653564, + "step": 12005, + "valid_targets_mean": 1470.0, + "valid_targets_min": 676 + }, + { + "epoch": 6.291251964379256, + "grad_norm": 0.7048799330801979, + "learning_rate": 1.238117993995136e-06, + "loss": 0.1117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044160082936286926, + "step": 12010, + "valid_targets_mean": 1185.5, + "valid_targets_min": 794 + }, + { + "epoch": 6.293871136720796, + "grad_norm": 0.7856551038869302, + "learning_rate": 1.2290853858651252e-06, + "loss": 0.1149, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051200881600379944, + "step": 12015, + "valid_targets_mean": 1259.0, + "valid_targets_min": 742 + }, + { + "epoch": 6.296490309062336, + "grad_norm": 0.9273041323586666, + "learning_rate": 1.220084802262369e-06, + "loss": 0.1132, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0679549127817154, + "step": 12020, + "valid_targets_mean": 1792.5, + "valid_targets_min": 741 + }, + { + "epoch": 6.299109481403876, + "grad_norm": 0.9096330064049609, + "learning_rate": 1.211116258542513e-06, + "loss": 0.119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06267944723367691, + "step": 12025, + "valid_targets_mean": 1599.6, + "valid_targets_min": 1157 + }, + { + "epoch": 6.301728653745417, + "grad_norm": 0.8281325080463295, + "learning_rate": 1.202179770006533e-06, + "loss": 0.1073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08307260274887085, + "step": 12030, + "valid_targets_mean": 1989.1, + "valid_targets_min": 862 + }, + { + "epoch": 6.304347826086957, + "grad_norm": 0.8086757409087395, + "learning_rate": 1.1932753519007246e-06, + "loss": 0.1174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05675650015473366, + "step": 12035, + "valid_targets_mean": 1428.1, + "valid_targets_min": 884 + }, + { + "epoch": 6.306966998428496, + "grad_norm": 0.8869832213789681, + "learning_rate": 1.18440301941666e-06, + "loss": 0.119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058827124536037445, + "step": 12040, + "valid_targets_mean": 1570.4, + "valid_targets_min": 1058 + }, + { + "epoch": 6.309586170770037, + "grad_norm": 0.8104961902323611, + "learning_rate": 1.1755627876911823e-06, + "loss": 0.1241, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05483512580394745, + "step": 12045, + "valid_targets_mean": 1509.6, + "valid_targets_min": 618 + }, + { + "epoch": 6.312205343111577, + "grad_norm": 0.7391950333294132, + "learning_rate": 1.1667546718063493e-06, + "loss": 0.1072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05378376692533493, + "step": 12050, + "valid_targets_mean": 1528.2, + "valid_targets_min": 744 + }, + { + "epoch": 6.314824515453116, + "grad_norm": 0.9181565585111005, + "learning_rate": 1.1579786867894471e-06, + "loss": 0.1153, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06050017848610878, + "step": 12055, + "valid_targets_mean": 1736.6, + "valid_targets_min": 1071 + }, + { + "epoch": 6.317443687794657, + "grad_norm": 0.85435510087702, + "learning_rate": 1.1492348476129323e-06, + "loss": 0.113, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05069496110081673, + "step": 12060, + "valid_targets_mean": 1120.4, + "valid_targets_min": 677 + }, + { + "epoch": 6.320062860136197, + "grad_norm": 0.8063787328557168, + "learning_rate": 1.1405231691944252e-06, + "loss": 0.1057, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04731714725494385, + "step": 12065, + "valid_targets_mean": 1139.6, + "valid_targets_min": 679 + }, + { + "epoch": 6.322682032477737, + "grad_norm": 0.8102016579413351, + "learning_rate": 1.1318436663966703e-06, + "loss": 0.1083, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056190505623817444, + "step": 12070, + "valid_targets_mean": 1509.2, + "valid_targets_min": 504 + }, + { + "epoch": 6.325301204819277, + "grad_norm": 0.7793586753673484, + "learning_rate": 1.123196354027527e-06, + "loss": 0.1126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.052414409816265106, + "step": 12075, + "valid_targets_mean": 1787.6, + "valid_targets_min": 1347 + }, + { + "epoch": 6.327920377160817, + "grad_norm": 0.7771634495228487, + "learning_rate": 1.1145812468399274e-06, + "loss": 0.1124, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050339505076408386, + "step": 12080, + "valid_targets_mean": 1404.1, + "valid_targets_min": 768 + }, + { + "epoch": 6.330539549502357, + "grad_norm": 0.7814958562457943, + "learning_rate": 1.1059983595318613e-06, + "loss": 0.1178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048321641981601715, + "step": 12085, + "valid_targets_mean": 1428.6, + "valid_targets_min": 1184 + }, + { + "epoch": 6.333158721843898, + "grad_norm": 0.7530919530531909, + "learning_rate": 1.0974477067463552e-06, + "loss": 0.1084, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04832286387681961, + "step": 12090, + "valid_targets_mean": 1489.2, + "valid_targets_min": 963 + }, + { + "epoch": 6.335777894185437, + "grad_norm": 0.8419205147839612, + "learning_rate": 1.0889293030714288e-06, + "loss": 0.1125, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.058813080191612244, + "step": 12095, + "valid_targets_mean": 1493.4, + "valid_targets_min": 931 + }, + { + "epoch": 6.338397066526977, + "grad_norm": 0.8169426127667708, + "learning_rate": 1.0804431630401013e-06, + "loss": 0.109, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06472467631101608, + "step": 12100, + "valid_targets_mean": 1552.0, + "valid_targets_min": 861 + }, + { + "epoch": 6.341016238868518, + "grad_norm": 0.7728161323355918, + "learning_rate": 1.071989301130325e-06, + "loss": 0.1115, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04840776324272156, + "step": 12105, + "valid_targets_mean": 1285.2, + "valid_targets_min": 797 + }, + { + "epoch": 6.343635411210058, + "grad_norm": 0.8337519408508389, + "learning_rate": 1.0635677317650006e-06, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.057807717472314835, + "step": 12110, + "valid_targets_mean": 1849.1, + "valid_targets_min": 1220 + }, + { + "epoch": 6.346254583551597, + "grad_norm": 0.7706509839494446, + "learning_rate": 1.0551784693119281e-06, + "loss": 0.1008, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0440559946000576, + "step": 12115, + "valid_targets_mean": 1488.0, + "valid_targets_min": 712 + }, + { + "epoch": 6.348873755893138, + "grad_norm": 0.8752968350948339, + "learning_rate": 1.0468215280837945e-06, + "loss": 0.1106, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06225879117846489, + "step": 12120, + "valid_targets_mean": 1415.0, + "valid_targets_min": 605 + }, + { + "epoch": 6.351492928234678, + "grad_norm": 0.753901515290285, + "learning_rate": 1.038496922338137e-06, + "loss": 0.1178, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04541228711605072, + "step": 12125, + "valid_targets_mean": 1473.4, + "valid_targets_min": 746 + }, + { + "epoch": 6.354112100576218, + "grad_norm": 0.8844491064762435, + "learning_rate": 1.0302046662773301e-06, + "loss": 0.1048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04617415368556976, + "step": 12130, + "valid_targets_mean": 1135.9, + "valid_targets_min": 723 + }, + { + "epoch": 6.356731272917758, + "grad_norm": 0.7871502031235931, + "learning_rate": 1.0219447740485622e-06, + "loss": 0.1118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06085076928138733, + "step": 12135, + "valid_targets_mean": 1624.0, + "valid_targets_min": 855 + }, + { + "epoch": 6.359350445259298, + "grad_norm": 0.767945202704885, + "learning_rate": 1.0137172597438006e-06, + "loss": 0.1151, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054333850741386414, + "step": 12140, + "valid_targets_mean": 1534.1, + "valid_targets_min": 1094 + }, + { + "epoch": 6.361969617600838, + "grad_norm": 0.7499305266087968, + "learning_rate": 1.0055221373997704e-06, + "loss": 0.1117, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.055590756237506866, + "step": 12145, + "valid_targets_mean": 1657.8, + "valid_targets_min": 803 + }, + { + "epoch": 6.3645887899423785, + "grad_norm": 0.9180369691704098, + "learning_rate": 9.973594209979475e-07, + "loss": 0.1143, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05862277001142502, + "step": 12150, + "valid_targets_mean": 1427.2, + "valid_targets_min": 740 + }, + { + "epoch": 6.367207962283918, + "grad_norm": 0.7824321669921074, + "learning_rate": 9.892291244645014e-07, + "loss": 0.1071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04916942119598389, + "step": 12155, + "valid_targets_mean": 1393.8, + "valid_targets_min": 1180 + }, + { + "epoch": 6.369827134625458, + "grad_norm": 0.8223816726277031, + "learning_rate": 9.811312616703072e-07, + "loss": 0.1133, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05651529133319855, + "step": 12160, + "valid_targets_mean": 1632.5, + "valid_targets_min": 1276 + }, + { + "epoch": 6.3724463069669985, + "grad_norm": 0.8568538903584729, + "learning_rate": 9.730658464309007e-07, + "loss": 0.1076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.056821659207344055, + "step": 12165, + "valid_targets_mean": 1496.6, + "valid_targets_min": 812 + }, + { + "epoch": 6.375065479308539, + "grad_norm": 0.8738741829592628, + "learning_rate": 9.650328925064545e-07, + "loss": 0.1163, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06567776203155518, + "step": 12170, + "valid_targets_mean": 1708.0, + "valid_targets_min": 1041 + }, + { + "epoch": 6.377684651650078, + "grad_norm": 0.8573270925314733, + "learning_rate": 9.570324136017683e-07, + "loss": 0.1166, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05247359350323677, + "step": 12175, + "valid_targets_mean": 1354.1, + "valid_targets_min": 777 + }, + { + "epoch": 6.3803038239916186, + "grad_norm": 0.7951492860958485, + "learning_rate": 9.490644233662328e-07, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05267871171236038, + "step": 12180, + "valid_targets_mean": 1665.0, + "valid_targets_min": 1283 + }, + { + "epoch": 6.382922996333159, + "grad_norm": 0.826697298035572, + "learning_rate": 9.411289353938136e-07, + "loss": 0.108, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05282936990261078, + "step": 12185, + "valid_targets_mean": 1153.0, + "valid_targets_min": 620 + }, + { + "epoch": 6.385542168674699, + "grad_norm": 0.7982537704121014, + "learning_rate": 9.332259632230123e-07, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04646070674061775, + "step": 12190, + "valid_targets_mean": 1386.5, + "valid_targets_min": 656 + }, + { + "epoch": 6.388161341016239, + "grad_norm": 0.8535869881021406, + "learning_rate": 9.253555203368792e-07, + "loss": 0.1119, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05572054162621498, + "step": 12195, + "valid_targets_mean": 1611.2, + "valid_targets_min": 984 + }, + { + "epoch": 6.390780513357779, + "grad_norm": 0.7409233156129882, + "learning_rate": 9.175176201629466e-07, + "loss": 0.1048, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04613656550645828, + "step": 12200, + "valid_targets_mean": 1801.8, + "valid_targets_min": 682 + }, + { + "epoch": 6.393399685699319, + "grad_norm": 0.7609047418456126, + "learning_rate": 9.09712276073238e-07, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04728458821773529, + "step": 12205, + "valid_targets_mean": 1508.9, + "valid_targets_min": 850 + }, + { + "epoch": 6.3960188580408595, + "grad_norm": 0.7557854851343144, + "learning_rate": 9.019395013842303e-07, + "loss": 0.1112, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053674980998039246, + "step": 12210, + "valid_targets_mean": 1573.5, + "valid_targets_min": 747 + }, + { + "epoch": 6.398638030382399, + "grad_norm": 0.7787496829918205, + "learning_rate": 8.941993093568314e-07, + "loss": 0.1162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.046790435910224915, + "step": 12215, + "valid_targets_mean": 1439.5, + "valid_targets_min": 933 + }, + { + "epoch": 6.401257202723939, + "grad_norm": 0.786251626067855, + "learning_rate": 8.86491713196369e-07, + "loss": 0.1066, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05557426065206528, + "step": 12220, + "valid_targets_mean": 1379.0, + "valid_targets_min": 799 + }, + { + "epoch": 6.4038763750654795, + "grad_norm": 0.7285433748757622, + "learning_rate": 8.788167260525537e-07, + "loss": 0.1046, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04541241377592087, + "step": 12225, + "valid_targets_mean": 1371.1, + "valid_targets_min": 1074 + }, + { + "epoch": 6.40649554740702, + "grad_norm": 0.7248388458592854, + "learning_rate": 8.7117436101946e-07, + "loss": 0.1065, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.049913328140974045, + "step": 12230, + "valid_targets_mean": 1555.4, + "valid_targets_min": 741 + }, + { + "epoch": 6.409114719748559, + "grad_norm": 0.8184813153052386, + "learning_rate": 8.635646311355206e-07, + "loss": 0.0983, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05012248456478119, + "step": 12235, + "valid_targets_mean": 1237.8, + "valid_targets_min": 720 + }, + { + "epoch": 6.4117338920900995, + "grad_norm": 0.9087359820098273, + "learning_rate": 8.559875493834702e-07, + "loss": 0.1174, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06459945440292358, + "step": 12240, + "valid_targets_mean": 1458.6, + "valid_targets_min": 610 + }, + { + "epoch": 6.41435306443164, + "grad_norm": 0.7684091465430946, + "learning_rate": 8.484431286903616e-07, + "loss": 0.1047, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06463173031806946, + "step": 12245, + "valid_targets_mean": 1887.2, + "valid_targets_min": 713 + }, + { + "epoch": 6.41697223677318, + "grad_norm": 0.8154391126453507, + "learning_rate": 8.409313819275144e-07, + "loss": 0.1089, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0512552410364151, + "step": 12250, + "valid_targets_mean": 1466.2, + "valid_targets_min": 874 + }, + { + "epoch": 6.4195914091147195, + "grad_norm": 0.7765737020584184, + "learning_rate": 8.334523219105084e-07, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054173991084098816, + "step": 12255, + "valid_targets_mean": 1510.6, + "valid_targets_min": 765 + }, + { + "epoch": 6.42221058145626, + "grad_norm": 0.8349893536795133, + "learning_rate": 8.260059613991545e-07, + "loss": 0.1039, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04710109531879425, + "step": 12260, + "valid_targets_mean": 1297.1, + "valid_targets_min": 589 + }, + { + "epoch": 6.4248297537978, + "grad_norm": 0.8274978726547828, + "learning_rate": 8.185923130974816e-07, + "loss": 0.1114, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.060467541217803955, + "step": 12265, + "valid_targets_mean": 1671.8, + "valid_targets_min": 1230 + }, + { + "epoch": 6.42744892613934, + "grad_norm": 0.9612345892318542, + "learning_rate": 8.112113896537033e-07, + "loss": 0.2015, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12336645275354385, + "step": 12270, + "valid_targets_mean": 4792.6, + "valid_targets_min": 702 + }, + { + "epoch": 6.43006809848088, + "grad_norm": 0.8915866166206202, + "learning_rate": 8.038632036602068e-07, + "loss": 0.2162, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11761324107646942, + "step": 12275, + "valid_targets_mean": 5322.8, + "valid_targets_min": 1005 + }, + { + "epoch": 6.43268727082242, + "grad_norm": 0.7836107682794644, + "learning_rate": 7.965477676535216e-07, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07904733717441559, + "step": 12280, + "valid_targets_mean": 3584.6, + "valid_targets_min": 2054 + }, + { + "epoch": 6.43530644316396, + "grad_norm": 0.866053101393214, + "learning_rate": 7.892650941143109e-07, + "loss": 0.2093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09079736471176147, + "step": 12285, + "valid_targets_mean": 3685.6, + "valid_targets_min": 1751 + }, + { + "epoch": 6.437925615505501, + "grad_norm": 0.7290318360116065, + "learning_rate": 7.820151954673316e-07, + "loss": 0.2068, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.12724754214286804, + "step": 12290, + "valid_targets_mean": 5944.2, + "valid_targets_min": 1970 + }, + { + "epoch": 6.44054478784704, + "grad_norm": 0.6995057791594722, + "learning_rate": 7.747980840814384e-07, + "loss": 0.2126, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10071046650409698, + "step": 12295, + "valid_targets_mean": 5916.0, + "valid_targets_min": 920 + }, + { + "epoch": 6.44316396018858, + "grad_norm": 0.7303027679630291, + "learning_rate": 7.676137722695376e-07, + "loss": 0.2031, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10200616717338562, + "step": 12300, + "valid_targets_mean": 6475.4, + "valid_targets_min": 2231 + }, + { + "epoch": 6.445783132530121, + "grad_norm": 0.640239884720967, + "learning_rate": 7.60462272288578e-07, + "loss": 0.2024, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0754626989364624, + "step": 12305, + "valid_targets_mean": 4043.4, + "valid_targets_min": 467 + }, + { + "epoch": 6.44840230487166, + "grad_norm": 0.6380969091227143, + "learning_rate": 7.533435963395375e-07, + "loss": 0.2128, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11595934629440308, + "step": 12310, + "valid_targets_mean": 5549.8, + "valid_targets_min": 3440 + }, + { + "epoch": 6.4510214772132, + "grad_norm": 0.6563780221740477, + "learning_rate": 7.462577565673835e-07, + "loss": 0.2093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10418115556240082, + "step": 12315, + "valid_targets_mean": 4736.9, + "valid_targets_min": 1697 + }, + { + "epoch": 6.453640649554741, + "grad_norm": 0.6569947299630399, + "learning_rate": 7.392047650610679e-07, + "loss": 0.2037, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11170069128274918, + "step": 12320, + "valid_targets_mean": 4838.2, + "valid_targets_min": 1223 + }, + { + "epoch": 6.456259821896281, + "grad_norm": 0.6002573967986714, + "learning_rate": 7.321846338535032e-07, + "loss": 0.2053, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10597524046897888, + "step": 12325, + "valid_targets_mean": 5466.9, + "valid_targets_min": 1552 + }, + { + "epoch": 6.458878994237821, + "grad_norm": 0.6170095055046106, + "learning_rate": 7.251973749215357e-07, + "loss": 0.2055, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11766611039638519, + "step": 12330, + "valid_targets_mean": 5166.0, + "valid_targets_min": 1546 + }, + { + "epoch": 6.461498166579361, + "grad_norm": 0.6095894994412887, + "learning_rate": 7.182430001859253e-07, + "loss": 0.1954, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09352870285511017, + "step": 12335, + "valid_targets_mean": 3848.2, + "valid_targets_min": 1957 + }, + { + "epoch": 6.464117338920901, + "grad_norm": 0.6426035674505436, + "learning_rate": 7.113215215113456e-07, + "loss": 0.1987, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09355652332305908, + "step": 12340, + "valid_targets_mean": 3487.5, + "valid_targets_min": 1538 + }, + { + "epoch": 6.466736511262441, + "grad_norm": 0.6340303519684779, + "learning_rate": 7.044329507063241e-07, + "loss": 0.1916, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09087233245372772, + "step": 12345, + "valid_targets_mean": 3397.2, + "valid_targets_min": 1645 + }, + { + "epoch": 6.469355683603981, + "grad_norm": 0.6885213434588451, + "learning_rate": 6.975772995232621e-07, + "loss": 0.1974, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08457259833812714, + "step": 12350, + "valid_targets_mean": 2859.4, + "valid_targets_min": 935 + }, + { + "epoch": 6.471974855945521, + "grad_norm": 0.6490260382940615, + "learning_rate": 6.907545796583926e-07, + "loss": 0.2022, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10389542579650879, + "step": 12355, + "valid_targets_mean": 3490.6, + "valid_targets_min": 1269 + }, + { + "epoch": 6.474594028287061, + "grad_norm": 0.705080529554098, + "learning_rate": 6.839648027517665e-07, + "loss": 0.2016, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0941087156534195, + "step": 12360, + "valid_targets_mean": 2860.0, + "valid_targets_min": 1173 + }, + { + "epoch": 6.477213200628602, + "grad_norm": 0.6126939509007335, + "learning_rate": 6.772079803872267e-07, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08957060426473618, + "step": 12365, + "valid_targets_mean": 3480.9, + "valid_targets_min": 1392 + }, + { + "epoch": 6.479832372970141, + "grad_norm": 0.6643017124373498, + "learning_rate": 6.704841240924031e-07, + "loss": 0.1997, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1080889105796814, + "step": 12370, + "valid_targets_mean": 3878.6, + "valid_targets_min": 1883 + }, + { + "epoch": 6.482451545311681, + "grad_norm": 0.6453397923957787, + "learning_rate": 6.637932453386664e-07, + "loss": 0.1893, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11066446453332901, + "step": 12375, + "valid_targets_mean": 3721.4, + "valid_targets_min": 2106 + }, + { + "epoch": 6.485070717653222, + "grad_norm": 0.5621856381718705, + "learning_rate": 6.571353555411475e-07, + "loss": 0.1793, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06962023675441742, + "step": 12380, + "valid_targets_mean": 3868.5, + "valid_targets_min": 2222 + }, + { + "epoch": 6.487689889994762, + "grad_norm": 0.6526048622189725, + "learning_rate": 6.505104660586759e-07, + "loss": 0.191, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09235334396362305, + "step": 12385, + "valid_targets_mean": 3010.1, + "valid_targets_min": 958 + }, + { + "epoch": 6.490309062336301, + "grad_norm": 0.6291810081480641, + "learning_rate": 6.439185881937949e-07, + "loss": 0.194, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09143108129501343, + "step": 12390, + "valid_targets_mean": 3789.1, + "valid_targets_min": 771 + }, + { + "epoch": 6.492928234677842, + "grad_norm": 0.5785247955221308, + "learning_rate": 6.373597331927128e-07, + "loss": 0.1984, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06739918887615204, + "step": 12395, + "valid_targets_mean": 3578.9, + "valid_targets_min": 1774 + }, + { + "epoch": 6.495547407019382, + "grad_norm": 0.6910281152576162, + "learning_rate": 6.308339122453189e-07, + "loss": 0.1873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08360813558101654, + "step": 12400, + "valid_targets_mean": 2545.8, + "valid_targets_min": 997 + }, + { + "epoch": 6.498166579360922, + "grad_norm": 0.6203099571113615, + "learning_rate": 6.243411364851204e-07, + "loss": 0.1866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09680942445993423, + "step": 12405, + "valid_targets_mean": 3667.1, + "valid_targets_min": 1741 + }, + { + "epoch": 6.500785751702462, + "grad_norm": 0.6562927817654133, + "learning_rate": 6.178814169892633e-07, + "loss": 0.1906, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09165982902050018, + "step": 12410, + "valid_targets_mean": 3606.1, + "valid_targets_min": 863 + }, + { + "epoch": 6.503404924044002, + "grad_norm": 0.6028858565374556, + "learning_rate": 6.114547647784941e-07, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09807367622852325, + "step": 12415, + "valid_targets_mean": 3876.9, + "valid_targets_min": 1854 + }, + { + "epoch": 6.506024096385542, + "grad_norm": 0.5631945508008784, + "learning_rate": 6.0506119081714e-07, + "loss": 0.1811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0895792618393898, + "step": 12420, + "valid_targets_mean": 4763.2, + "valid_targets_min": 2040 + }, + { + "epoch": 6.5086432687270825, + "grad_norm": 0.6203079207995192, + "learning_rate": 5.987007060130978e-07, + "loss": 0.1889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08201078325510025, + "step": 12425, + "valid_targets_mean": 3521.9, + "valid_targets_min": 1036 + }, + { + "epoch": 6.511262441068622, + "grad_norm": 0.5924661700896477, + "learning_rate": 5.92373321217814e-07, + "loss": 0.1878, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09583745896816254, + "step": 12430, + "valid_targets_mean": 4304.8, + "valid_targets_min": 1279 + }, + { + "epoch": 6.513881613410162, + "grad_norm": 0.6556087747768623, + "learning_rate": 5.860790472262534e-07, + "loss": 0.1852, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09577606618404388, + "step": 12435, + "valid_targets_mean": 3600.0, + "valid_targets_min": 1798 + }, + { + "epoch": 6.5165007857517026, + "grad_norm": 0.6524514137665829, + "learning_rate": 5.798178947769106e-07, + "loss": 0.1871, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08674263954162598, + "step": 12440, + "valid_targets_mean": 3067.2, + "valid_targets_min": 1446 + }, + { + "epoch": 6.519119958093243, + "grad_norm": 0.6001705092369103, + "learning_rate": 5.735898745517499e-07, + "loss": 0.1851, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09648394584655762, + "step": 12445, + "valid_targets_mean": 4183.2, + "valid_targets_min": 1270 + }, + { + "epoch": 6.521739130434782, + "grad_norm": 0.5789108150632485, + "learning_rate": 5.673949971762249e-07, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10458388924598694, + "step": 12450, + "valid_targets_mean": 5485.8, + "valid_targets_min": 1216 + }, + { + "epoch": 6.524358302776323, + "grad_norm": 0.6365862583903048, + "learning_rate": 5.612332732192416e-07, + "loss": 0.1809, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10634689778089523, + "step": 12455, + "valid_targets_mean": 3994.0, + "valid_targets_min": 2028 + }, + { + "epoch": 6.526977475117863, + "grad_norm": 0.6683915895376606, + "learning_rate": 5.551047131931442e-07, + "loss": 0.1963, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1127801313996315, + "step": 12460, + "valid_targets_mean": 3700.6, + "valid_targets_min": 2329 + }, + { + "epoch": 6.529596647459403, + "grad_norm": 0.6658818568358972, + "learning_rate": 5.490093275536934e-07, + "loss": 0.1898, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11775028705596924, + "step": 12465, + "valid_targets_mean": 4075.2, + "valid_targets_min": 1363 + }, + { + "epoch": 6.532215819800943, + "grad_norm": 0.603520375210053, + "learning_rate": 5.429471267000552e-07, + "loss": 0.1799, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08491937816143036, + "step": 12470, + "valid_targets_mean": 3560.8, + "valid_targets_min": 1087 + }, + { + "epoch": 6.534834992142483, + "grad_norm": 0.6651883490296507, + "learning_rate": 5.36918120974781e-07, + "loss": 0.1928, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09837782382965088, + "step": 12475, + "valid_targets_mean": 3860.1, + "valid_targets_min": 2483 + }, + { + "epoch": 6.537454164484023, + "grad_norm": 0.639229291293438, + "learning_rate": 5.309223206637826e-07, + "loss": 0.1943, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10892842710018158, + "step": 12480, + "valid_targets_mean": 4583.2, + "valid_targets_min": 2316 + }, + { + "epoch": 6.5400733368255635, + "grad_norm": 0.6283408829693979, + "learning_rate": 5.249597359963332e-07, + "loss": 0.1835, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.100398950278759, + "step": 12485, + "valid_targets_mean": 4198.9, + "valid_targets_min": 2022 + }, + { + "epoch": 6.542692509167103, + "grad_norm": 0.5885321040286818, + "learning_rate": 5.190303771450222e-07, + "loss": 0.1779, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07413794845342636, + "step": 12490, + "valid_targets_mean": 3228.9, + "valid_targets_min": 1294 + }, + { + "epoch": 6.545311681508643, + "grad_norm": 0.6445410853548745, + "learning_rate": 5.131342542257666e-07, + "loss": 0.1935, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10934539884328842, + "step": 12495, + "valid_targets_mean": 3846.8, + "valid_targets_min": 2716 + }, + { + "epoch": 6.5479308538501835, + "grad_norm": 0.6888642745844418, + "learning_rate": 5.072713772977711e-07, + "loss": 0.1847, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09181120991706848, + "step": 12500, + "valid_targets_mean": 3211.6, + "valid_targets_min": 2060 + }, + { + "epoch": 6.550550026191724, + "grad_norm": 0.691333169436966, + "learning_rate": 5.014417563635276e-07, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09888258576393127, + "step": 12505, + "valid_targets_mean": 3320.4, + "valid_targets_min": 2407 + }, + { + "epoch": 6.553169198533263, + "grad_norm": 0.5979901894796124, + "learning_rate": 4.956454013687873e-07, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09806334227323532, + "step": 12510, + "valid_targets_mean": 4165.1, + "valid_targets_min": 1533 + }, + { + "epoch": 6.5557883708748035, + "grad_norm": 0.6396467931879949, + "learning_rate": 4.898823222025462e-07, + "loss": 0.1968, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11920514702796936, + "step": 12515, + "valid_targets_mean": 5201.2, + "valid_targets_min": 2189 + }, + { + "epoch": 6.558407543216344, + "grad_norm": 0.6300159753489842, + "learning_rate": 4.841525286970327e-07, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1105848178267479, + "step": 12520, + "valid_targets_mean": 4193.1, + "valid_targets_min": 1328 + }, + { + "epoch": 6.561026715557884, + "grad_norm": 0.6454776965573862, + "learning_rate": 4.784560306276875e-07, + "loss": 0.196, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.13785168528556824, + "step": 12525, + "valid_targets_mean": 4500.4, + "valid_targets_min": 1708 + }, + { + "epoch": 6.5636458878994235, + "grad_norm": 0.6087367359513417, + "learning_rate": 4.727928377131363e-07, + "loss": 0.1795, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07544751465320587, + "step": 12530, + "valid_targets_mean": 3494.0, + "valid_targets_min": 1912 + }, + { + "epoch": 6.566265060240964, + "grad_norm": 0.6621317744034372, + "learning_rate": 4.6716295961520384e-07, + "loss": 0.1818, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11185840517282486, + "step": 12535, + "valid_targets_mean": 4145.2, + "valid_targets_min": 2606 + }, + { + "epoch": 6.568884232582504, + "grad_norm": 0.5911671607997898, + "learning_rate": 4.6156640593885805e-07, + "loss": 0.1786, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08351260423660278, + "step": 12540, + "valid_targets_mean": 4189.0, + "valid_targets_min": 1378 + }, + { + "epoch": 6.571503404924044, + "grad_norm": 0.7054889985176381, + "learning_rate": 4.560031862322278e-07, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.113043412566185, + "step": 12545, + "valid_targets_mean": 3983.4, + "valid_targets_min": 1219 + }, + { + "epoch": 6.574122577265584, + "grad_norm": 0.6171820462700326, + "learning_rate": 4.504733099865566e-07, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1003456711769104, + "step": 12550, + "valid_targets_mean": 3997.0, + "valid_targets_min": 2631 + }, + { + "epoch": 6.576741749607124, + "grad_norm": 0.6421828994706504, + "learning_rate": 4.4497678663621756e-07, + "loss": 0.188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1136467456817627, + "step": 12555, + "valid_targets_mean": 4167.4, + "valid_targets_min": 1423 + }, + { + "epoch": 6.579360921948664, + "grad_norm": 0.6408214025088442, + "learning_rate": 4.395136255586696e-07, + "loss": 0.1857, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0737825557589531, + "step": 12560, + "valid_targets_mean": 3381.9, + "valid_targets_min": 2320 + }, + { + "epoch": 6.581980094290204, + "grad_norm": 0.6355136447882026, + "learning_rate": 4.340838360744615e-07, + "loss": 0.1932, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09415724873542786, + "step": 12565, + "valid_targets_mean": 3174.9, + "valid_targets_min": 1188 + }, + { + "epoch": 6.584599266631744, + "grad_norm": 0.5974334102416431, + "learning_rate": 4.286874274471986e-07, + "loss": 0.1929, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07140226662158966, + "step": 12570, + "valid_targets_mean": 3378.6, + "valid_targets_min": 618 + }, + { + "epoch": 6.587218438973284, + "grad_norm": 0.7347012411938242, + "learning_rate": 4.233244088835453e-07, + "loss": 0.1888, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09825826436281204, + "step": 12575, + "valid_targets_mean": 2988.1, + "valid_targets_min": 1364 + }, + { + "epoch": 6.589837611314825, + "grad_norm": 0.6545078147772053, + "learning_rate": 4.1799478953319814e-07, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08172410726547241, + "step": 12580, + "valid_targets_mean": 3199.1, + "valid_targets_min": 1397 + }, + { + "epoch": 6.592456783656365, + "grad_norm": 0.619408448207835, + "learning_rate": 4.1269857848886153e-07, + "loss": 0.1753, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10200368613004684, + "step": 12585, + "valid_targets_mean": 5348.0, + "valid_targets_min": 1778 + }, + { + "epoch": 6.595075955997904, + "grad_norm": 0.6339933832164817, + "learning_rate": 4.074357847862609e-07, + "loss": 0.1763, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09227821975946426, + "step": 12590, + "valid_targets_mean": 3444.9, + "valid_targets_min": 621 + }, + { + "epoch": 6.597695128339445, + "grad_norm": 0.5950223334681427, + "learning_rate": 4.022064174040918e-07, + "loss": 0.1723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07300743460655212, + "step": 12595, + "valid_targets_mean": 3872.9, + "valid_targets_min": 1290 + }, + { + "epoch": 6.600314300680985, + "grad_norm": 1.4005098615945741, + "learning_rate": 3.9701048526403774e-07, + "loss": 0.1721, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08079865574836731, + "step": 12600, + "valid_targets_mean": 2546.6, + "valid_targets_min": 642 + }, + { + "epoch": 6.602933473022524, + "grad_norm": 0.7131855989057194, + "learning_rate": 3.9184799723072764e-07, + "loss": 0.17, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09240078926086426, + "step": 12605, + "valid_targets_mean": 3428.2, + "valid_targets_min": 1978 + }, + { + "epoch": 6.605552645364065, + "grad_norm": 0.7178154653152131, + "learning_rate": 3.867189621117362e-07, + "loss": 0.1894, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.130007803440094, + "step": 12610, + "valid_targets_mean": 4160.0, + "valid_targets_min": 2642 + }, + { + "epoch": 6.608171817705605, + "grad_norm": 0.6302852301092045, + "learning_rate": 3.81623388657566e-07, + "loss": 0.1941, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07194172590970993, + "step": 12615, + "valid_targets_mean": 2937.1, + "valid_targets_min": 1510 + }, + { + "epoch": 6.610790990047145, + "grad_norm": 0.7193541521730871, + "learning_rate": 3.7656128556163407e-07, + "loss": 0.1812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.093415766954422, + "step": 12620, + "valid_targets_mean": 2744.2, + "valid_targets_min": 987 + }, + { + "epoch": 6.613410162388686, + "grad_norm": 0.6271033009757231, + "learning_rate": 3.7153266146024547e-07, + "loss": 0.1846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0849226787686348, + "step": 12625, + "valid_targets_mean": 3638.4, + "valid_targets_min": 1428 + }, + { + "epoch": 6.616029334730225, + "grad_norm": 0.6339918774008351, + "learning_rate": 3.6653752493259976e-07, + "loss": 0.184, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10839584469795227, + "step": 12630, + "valid_targets_mean": 4731.5, + "valid_targets_min": 1700 + }, + { + "epoch": 6.618648507071765, + "grad_norm": 0.5274154764190819, + "learning_rate": 3.615758845007533e-07, + "loss": 0.1756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08318442106246948, + "step": 12635, + "valid_targets_mean": 5699.6, + "valid_targets_min": 1767 + }, + { + "epoch": 6.621267679413306, + "grad_norm": 0.5471570673936963, + "learning_rate": 3.5664774862962605e-07, + "loss": 0.1815, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08453177660703659, + "step": 12640, + "valid_targets_mean": 4101.5, + "valid_targets_min": 2684 + }, + { + "epoch": 6.623886851754845, + "grad_norm": 0.6944313429887211, + "learning_rate": 3.5175312572696795e-07, + "loss": 0.1785, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08602851629257202, + "step": 12645, + "valid_targets_mean": 2761.2, + "valid_targets_min": 1542 + }, + { + "epoch": 6.626506024096385, + "grad_norm": 0.6998607534289536, + "learning_rate": 3.468920241433571e-07, + "loss": 0.1831, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07480040937662125, + "step": 12650, + "valid_targets_mean": 2765.8, + "valid_targets_min": 1476 + }, + { + "epoch": 6.629125196437926, + "grad_norm": 0.628052338276033, + "learning_rate": 3.4206445217217944e-07, + "loss": 0.1792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08613723516464233, + "step": 12655, + "valid_targets_mean": 3884.5, + "valid_targets_min": 1711 + }, + { + "epoch": 6.631744368779466, + "grad_norm": 0.6730873887822488, + "learning_rate": 3.3727041804962447e-07, + "loss": 0.1891, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09674298763275146, + "step": 12660, + "valid_targets_mean": 3235.6, + "valid_targets_min": 1669 + }, + { + "epoch": 6.634363541121006, + "grad_norm": 0.6742550977753634, + "learning_rate": 3.3250992995465416e-07, + "loss": 0.1958, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09208236634731293, + "step": 12665, + "valid_targets_mean": 3388.5, + "valid_targets_min": 1624 + }, + { + "epoch": 6.636982713462546, + "grad_norm": 0.6861287671246876, + "learning_rate": 3.277829960090073e-07, + "loss": 0.1887, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09841510653495789, + "step": 12670, + "valid_targets_mean": 3380.9, + "valid_targets_min": 1676 + }, + { + "epoch": 6.639601885804086, + "grad_norm": 0.6216259541198084, + "learning_rate": 3.230896242771642e-07, + "loss": 0.198, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07941648364067078, + "step": 12675, + "valid_targets_mean": 3148.4, + "valid_targets_min": 1500 + }, + { + "epoch": 6.642221058145626, + "grad_norm": 0.5779295692189107, + "learning_rate": 3.184298227663618e-07, + "loss": 0.1792, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08519819378852844, + "step": 12680, + "valid_targets_mean": 4343.5, + "valid_targets_min": 2556 + }, + { + "epoch": 6.644840230487166, + "grad_norm": 0.6388857436536234, + "learning_rate": 3.13803599426552e-07, + "loss": 0.1995, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07832565903663635, + "step": 12685, + "valid_targets_mean": 3299.5, + "valid_targets_min": 1241 + }, + { + "epoch": 6.647459402828706, + "grad_norm": 0.7076552500989624, + "learning_rate": 3.092109621504058e-07, + "loss": 0.1965, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11633239686489105, + "step": 12690, + "valid_targets_mean": 3547.0, + "valid_targets_min": 992 + }, + { + "epoch": 6.650078575170246, + "grad_norm": 0.7139011472829109, + "learning_rate": 3.0465191877329104e-07, + "loss": 0.1821, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10384678840637207, + "step": 12695, + "valid_targets_mean": 4328.6, + "valid_targets_min": 1133 + }, + { + "epoch": 6.6526977475117866, + "grad_norm": 0.6202654212949067, + "learning_rate": 3.001264770732615e-07, + "loss": 0.1863, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09818834811449051, + "step": 12700, + "valid_targets_mean": 4202.6, + "valid_targets_min": 2369 + }, + { + "epoch": 6.655316919853326, + "grad_norm": 0.6338356248212572, + "learning_rate": 2.9563464477104786e-07, + "loss": 0.1853, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07603020966053009, + "step": 12705, + "valid_targets_mean": 3563.0, + "valid_targets_min": 1919 + }, + { + "epoch": 6.657936092194866, + "grad_norm": 0.669565852450954, + "learning_rate": 2.9117642953003564e-07, + "loss": 0.1927, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10715268552303314, + "step": 12710, + "valid_targets_mean": 3975.5, + "valid_targets_min": 2902 + }, + { + "epoch": 6.660555264536407, + "grad_norm": 0.7322787995442488, + "learning_rate": 2.867518389562629e-07, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.1092028021812439, + "step": 12715, + "valid_targets_mean": 2975.5, + "valid_targets_min": 617 + }, + { + "epoch": 6.663174436877947, + "grad_norm": 0.706181999443342, + "learning_rate": 2.823608805983957e-07, + "loss": 0.1694, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07655573636293411, + "step": 12720, + "valid_targets_mean": 3290.4, + "valid_targets_min": 1349 + }, + { + "epoch": 6.665793609219486, + "grad_norm": 0.6437619171580853, + "learning_rate": 2.7800356194772836e-07, + "loss": 0.1817, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09604845941066742, + "step": 12725, + "valid_targets_mean": 3459.0, + "valid_targets_min": 1769 + }, + { + "epoch": 6.668412781561027, + "grad_norm": 0.6168123575892074, + "learning_rate": 2.7367989043815434e-07, + "loss": 0.1791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08612355589866638, + "step": 12730, + "valid_targets_mean": 4186.1, + "valid_targets_min": 2097 + }, + { + "epoch": 6.671031953902567, + "grad_norm": 0.6837017208333571, + "learning_rate": 2.6938987344616864e-07, + "loss": 0.1854, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09780089557170868, + "step": 12735, + "valid_targets_mean": 3304.0, + "valid_targets_min": 1066 + }, + { + "epoch": 6.673651126244107, + "grad_norm": 0.5778950999854322, + "learning_rate": 2.651335182908499e-07, + "loss": 0.1802, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.090537890791893, + "step": 12740, + "valid_targets_mean": 4457.9, + "valid_targets_min": 2635 + }, + { + "epoch": 6.676270298585647, + "grad_norm": 0.6373690105448861, + "learning_rate": 2.6091083223384496e-07, + "loss": 0.1897, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0995105728507042, + "step": 12745, + "valid_targets_mean": 4296.2, + "valid_targets_min": 981 + }, + { + "epoch": 6.678889470927187, + "grad_norm": 0.6422863095358654, + "learning_rate": 2.5672182247935773e-07, + "loss": 0.1796, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09039589017629623, + "step": 12750, + "valid_targets_mean": 4020.1, + "valid_targets_min": 2131 + }, + { + "epoch": 6.681508643268727, + "grad_norm": 0.7413798369732252, + "learning_rate": 2.525664961741403e-07, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09179878234863281, + "step": 12755, + "valid_targets_mean": 3541.0, + "valid_targets_min": 1353 + }, + { + "epoch": 6.6841278156102675, + "grad_norm": 0.6459676626032188, + "learning_rate": 2.484448604074796e-07, + "loss": 0.1886, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08133266866207123, + "step": 12760, + "valid_targets_mean": 3362.0, + "valid_targets_min": 1509 + }, + { + "epoch": 6.686746987951807, + "grad_norm": 0.6381931915174774, + "learning_rate": 2.4435692221117966e-07, + "loss": 0.1858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09177473187446594, + "step": 12765, + "valid_targets_mean": 3918.9, + "valid_targets_min": 2128 + }, + { + "epoch": 6.689366160293347, + "grad_norm": 0.6167203331179475, + "learning_rate": 2.4030268855955495e-07, + "loss": 0.1728, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07183702290058136, + "step": 12770, + "valid_targets_mean": 2919.0, + "valid_targets_min": 1687 + }, + { + "epoch": 6.6919853326348875, + "grad_norm": 0.6725961593423001, + "learning_rate": 2.362821663694237e-07, + "loss": 0.1787, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10078228265047073, + "step": 12775, + "valid_targets_mean": 3696.9, + "valid_targets_min": 1803 + }, + { + "epoch": 6.694604504976428, + "grad_norm": 0.6227886377393108, + "learning_rate": 2.32295362500079e-07, + "loss": 0.1777, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08572547137737274, + "step": 12780, + "valid_targets_mean": 4102.0, + "valid_targets_min": 2263 + }, + { + "epoch": 6.697223677317967, + "grad_norm": 0.7081838656650461, + "learning_rate": 2.2834228375330003e-07, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10950066149234772, + "step": 12785, + "valid_targets_mean": 4621.4, + "valid_targets_min": 2232 + }, + { + "epoch": 6.6998428496595075, + "grad_norm": 0.6656950881593954, + "learning_rate": 2.2442293687331863e-07, + "loss": 0.1677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10185244679450989, + "step": 12790, + "valid_targets_mean": 4051.6, + "valid_targets_min": 873 + }, + { + "epoch": 6.702462022001048, + "grad_norm": 0.5811853551520246, + "learning_rate": 2.2053732854682598e-07, + "loss": 0.1723, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07740449905395508, + "step": 12795, + "valid_targets_mean": 3340.1, + "valid_targets_min": 1942 + }, + { + "epoch": 6.705081194342588, + "grad_norm": 0.6636169059360069, + "learning_rate": 2.16685465402946e-07, + "loss": 0.1762, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08695822954177856, + "step": 12800, + "valid_targets_mean": 3179.4, + "valid_targets_min": 897 + }, + { + "epoch": 6.7077003666841275, + "grad_norm": 0.6675055391580295, + "learning_rate": 2.128673540132331e-07, + "loss": 0.1911, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09340022504329681, + "step": 12805, + "valid_targets_mean": 3492.2, + "valid_targets_min": 1695 + }, + { + "epoch": 6.710319539025668, + "grad_norm": 0.6073889148051943, + "learning_rate": 2.090830008916589e-07, + "loss": 0.1711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07706175744533539, + "step": 12810, + "valid_targets_mean": 4043.9, + "valid_targets_min": 1712 + }, + { + "epoch": 6.712938711367208, + "grad_norm": 0.6221740037386776, + "learning_rate": 2.0533241249460545e-07, + "loss": 0.1744, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07362157106399536, + "step": 12815, + "valid_targets_mean": 2987.9, + "valid_targets_min": 993 + }, + { + "epoch": 6.715557883708748, + "grad_norm": 0.5868421221063573, + "learning_rate": 2.0161559522084095e-07, + "loss": 0.1742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08614050596952438, + "step": 12820, + "valid_targets_mean": 3854.8, + "valid_targets_min": 1679 + }, + { + "epoch": 6.718177056050288, + "grad_norm": 0.5626451083344456, + "learning_rate": 1.9793255541152633e-07, + "loss": 0.1808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08351093530654907, + "step": 12825, + "valid_targets_mean": 3938.8, + "valid_targets_min": 2345 + }, + { + "epoch": 6.720796228391828, + "grad_norm": 0.6241159417559475, + "learning_rate": 1.9428329935018642e-07, + "loss": 0.1813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09518198668956757, + "step": 12830, + "valid_targets_mean": 4507.9, + "valid_targets_min": 3068 + }, + { + "epoch": 6.723415400733368, + "grad_norm": 0.6779767248606396, + "learning_rate": 1.9066783326272098e-07, + "loss": 0.1855, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0879785567522049, + "step": 12835, + "valid_targets_mean": 4080.9, + "valid_targets_min": 2965 + }, + { + "epoch": 6.726034573074909, + "grad_norm": 0.5981399847390828, + "learning_rate": 1.870861633173693e-07, + "loss": 0.1768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.07661651819944382, + "step": 12840, + "valid_targets_mean": 2993.5, + "valid_targets_min": 1817 + }, + { + "epoch": 6.728653745416448, + "grad_norm": 0.6347100868343822, + "learning_rate": 1.8353829562471893e-07, + "loss": 0.1783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08969065546989441, + "step": 12845, + "valid_targets_mean": 3795.2, + "valid_targets_min": 1727 + }, + { + "epoch": 6.731272917757988, + "grad_norm": 0.5636941377793129, + "learning_rate": 1.8002423623768584e-07, + "loss": 0.1766, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08597507327795029, + "step": 12850, + "valid_targets_mean": 4095.9, + "valid_targets_min": 2491 + }, + { + "epoch": 6.733892090099529, + "grad_norm": 0.5179396427870425, + "learning_rate": 1.7654399115150988e-07, + "loss": 0.1461, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.09012123197317123, + "step": 12855, + "valid_targets_mean": 2046.1, + "valid_targets_min": 333 + }, + { + "epoch": 6.736511262441068, + "grad_norm": 0.37963782489960224, + "learning_rate": 1.7309756630373486e-07, + "loss": 0.0791, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04024237394332886, + "step": 12860, + "valid_targets_mean": 3455.9, + "valid_targets_min": 712 + }, + { + "epoch": 6.739130434782608, + "grad_norm": 0.44535000646563583, + "learning_rate": 1.6968496757421294e-07, + "loss": 0.0742, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03516673296689987, + "step": 12865, + "valid_targets_mean": 2452.6, + "valid_targets_min": 732 + }, + { + "epoch": 6.741749607124149, + "grad_norm": 0.694465933502411, + "learning_rate": 1.663062007850824e-07, + "loss": 0.0807, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03872720152139664, + "step": 12870, + "valid_targets_mean": 2228.9, + "valid_targets_min": 803 + }, + { + "epoch": 6.744368779465689, + "grad_norm": 0.38445161158765984, + "learning_rate": 1.6296127170075894e-07, + "loss": 0.0727, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03502201288938522, + "step": 12875, + "valid_targets_mean": 3038.6, + "valid_targets_min": 818 + }, + { + "epoch": 6.746987951807229, + "grad_norm": 0.4340486642862031, + "learning_rate": 1.5965018602793315e-07, + "loss": 0.0765, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04020746797323227, + "step": 12880, + "valid_targets_mean": 3049.4, + "valid_targets_min": 1524 + }, + { + "epoch": 6.749607124148769, + "grad_norm": 0.8059481744093037, + "learning_rate": 1.5637294941555526e-07, + "loss": 0.1183, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.053411468863487244, + "step": 12885, + "valid_targets_mean": 1256.1, + "valid_targets_min": 539 + }, + { + "epoch": 6.752226296490309, + "grad_norm": 0.414794394458646, + "learning_rate": 1.5312956745482388e-07, + "loss": 0.0872, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03645399212837219, + "step": 12890, + "valid_targets_mean": 3597.0, + "valid_targets_min": 2920 + }, + { + "epoch": 6.754845468831849, + "grad_norm": 0.4786662803371018, + "learning_rate": 1.4992004567918383e-07, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04220326989889145, + "step": 12895, + "valid_targets_mean": 2737.9, + "valid_targets_min": 1048 + }, + { + "epoch": 6.757464641173389, + "grad_norm": 0.31938282030052795, + "learning_rate": 1.4674438956430837e-07, + "loss": 0.1097, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027137383818626404, + "step": 12900, + "valid_targets_mean": 4340.1, + "valid_targets_min": 962 + }, + { + "epoch": 6.760083813514929, + "grad_norm": 0.3738010322411182, + "learning_rate": 1.4360260452809472e-07, + "loss": 0.0642, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04171907529234886, + "step": 12905, + "valid_targets_mean": 4283.1, + "valid_targets_min": 2091 + }, + { + "epoch": 6.762702985856469, + "grad_norm": 0.3708300150615133, + "learning_rate": 1.4049469593065524e-07, + "loss": 0.073, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035861678421497345, + "step": 12910, + "valid_targets_mean": 4320.9, + "valid_targets_min": 3349 + }, + { + "epoch": 6.76532215819801, + "grad_norm": 0.40270466618242945, + "learning_rate": 1.3742066907430185e-07, + "loss": 0.0737, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037723440676927567, + "step": 12915, + "valid_targets_mean": 3895.9, + "valid_targets_min": 518 + }, + { + "epoch": 6.76794133053955, + "grad_norm": 0.5235394881812243, + "learning_rate": 1.3438052920354827e-07, + "loss": 0.1104, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043522994965314865, + "step": 12920, + "valid_targets_mean": 2219.6, + "valid_targets_min": 549 + }, + { + "epoch": 6.770560502881089, + "grad_norm": 0.3836462034946707, + "learning_rate": 1.313742815050878e-07, + "loss": 0.0667, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029054410755634308, + "step": 12925, + "valid_targets_mean": 3700.2, + "valid_targets_min": 2724 + }, + { + "epoch": 6.77317967522263, + "grad_norm": 0.4979062136886603, + "learning_rate": 1.2840193110779776e-07, + "loss": 0.0772, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0390518456697464, + "step": 12930, + "valid_targets_mean": 1672.5, + "valid_targets_min": 775 + }, + { + "epoch": 6.77579884756417, + "grad_norm": 0.33508226379516654, + "learning_rate": 1.2546348308271949e-07, + "loss": 0.0625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.020936738699674606, + "step": 12935, + "valid_targets_mean": 3546.2, + "valid_targets_min": 3183 + }, + { + "epoch": 6.778418019905709, + "grad_norm": 0.44349242015093226, + "learning_rate": 1.2255894244305843e-07, + "loss": 0.0775, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033688582479953766, + "step": 12940, + "valid_targets_mean": 2473.2, + "valid_targets_min": 577 + }, + { + "epoch": 6.78103719224725, + "grad_norm": 0.5417750546586528, + "learning_rate": 1.1968831414416626e-07, + "loss": 0.0843, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05502060055732727, + "step": 12945, + "valid_targets_mean": 1597.0, + "valid_targets_min": 557 + }, + { + "epoch": 6.78365636458879, + "grad_norm": 0.4409383389943552, + "learning_rate": 1.1685160308354315e-07, + "loss": 0.08, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0404217466711998, + "step": 12950, + "valid_targets_mean": 3019.1, + "valid_targets_min": 774 + }, + { + "epoch": 6.78627553693033, + "grad_norm": 0.4222235305201826, + "learning_rate": 1.1404881410082002e-07, + "loss": 0.0827, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03403383865952492, + "step": 12955, + "valid_targets_mean": 1825.8, + "valid_targets_min": 680 + }, + { + "epoch": 6.7888947092718706, + "grad_norm": 0.5011825277601092, + "learning_rate": 1.1127995197775632e-07, + "loss": 0.1188, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06172202527523041, + "step": 12960, + "valid_targets_mean": 3686.9, + "valid_targets_min": 2819 + }, + { + "epoch": 6.79151388161341, + "grad_norm": 0.3240669351138313, + "learning_rate": 1.085450214382311e-07, + "loss": 0.0679, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02920447662472725, + "step": 12965, + "valid_targets_mean": 4137.8, + "valid_targets_min": 3927 + }, + { + "epoch": 6.79413305395495, + "grad_norm": 0.5427469680837108, + "learning_rate": 1.0584402714822972e-07, + "loss": 0.0783, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.051346149295568466, + "step": 12970, + "valid_targets_mean": 1376.1, + "valid_targets_min": 714 + }, + { + "epoch": 6.796752226296491, + "grad_norm": 0.421141501335943, + "learning_rate": 1.031769737158439e-07, + "loss": 0.0994, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03386833518743515, + "step": 12975, + "valid_targets_mean": 3432.0, + "valid_targets_min": 754 + }, + { + "epoch": 6.79937139863803, + "grad_norm": 0.37955973318746317, + "learning_rate": 1.0054386569126051e-07, + "loss": 0.0881, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03767680749297142, + "step": 12980, + "valid_targets_mean": 3855.2, + "valid_targets_min": 3178 + }, + { + "epoch": 6.80199057097957, + "grad_norm": 0.6751338239838669, + "learning_rate": 9.79447075667439e-08, + "loss": 0.107, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.11066795140504837, + "step": 12985, + "valid_targets_mean": 1566.6, + "valid_targets_min": 920 + }, + { + "epoch": 6.804609743321111, + "grad_norm": 0.4873775523005877, + "learning_rate": 9.53795037766514e-08, + "loss": 0.2403, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03654659539461136, + "step": 12990, + "valid_targets_mean": 3478.0, + "valid_targets_min": 2314 + }, + { + "epoch": 6.807228915662651, + "grad_norm": 0.41663560374071595, + "learning_rate": 9.284825869739999e-08, + "loss": 0.0838, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03145845979452133, + "step": 12995, + "valid_targets_mean": 2561.1, + "valid_targets_min": 833 + }, + { + "epoch": 6.809848088004191, + "grad_norm": 0.4430923832609423, + "learning_rate": 9.035097664747972e-08, + "loss": 0.0811, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04930516332387924, + "step": 13000, + "valid_targets_mean": 3509.2, + "valid_targets_min": 1803 + }, + { + "epoch": 6.812467260345731, + "grad_norm": 0.45593320164655093, + "learning_rate": 8.788766188742914e-08, + "loss": 0.1719, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.054887570440769196, + "step": 13005, + "valid_targets_mean": 4291.4, + "valid_targets_min": 3089 + }, + { + "epoch": 6.815086432687271, + "grad_norm": 0.30183471880069057, + "learning_rate": 8.545831861983989e-08, + "loss": 0.0616, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030286788940429688, + "step": 13010, + "valid_targets_mean": 4843.2, + "valid_targets_min": 2664 + }, + { + "epoch": 6.817705605028811, + "grad_norm": 0.36164986284476963, + "learning_rate": 8.306295098934991e-08, + "loss": 0.0846, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.024289684370160103, + "step": 13015, + "valid_targets_mean": 3042.1, + "valid_targets_min": 560 + }, + { + "epoch": 6.820324777370351, + "grad_norm": 0.4480250710139255, + "learning_rate": 8.070156308262356e-08, + "loss": 0.0813, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04077955335378647, + "step": 13020, + "valid_targets_mean": 3863.6, + "valid_targets_min": 1615 + }, + { + "epoch": 6.822943949711891, + "grad_norm": 0.38581206914916344, + "learning_rate": 7.837415892836264e-08, + "loss": 0.0688, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026508484035730362, + "step": 13025, + "valid_targets_mean": 3312.5, + "valid_targets_min": 2172 + }, + { + "epoch": 6.825563122053431, + "grad_norm": 0.38793008128315765, + "learning_rate": 7.608074249728204e-08, + "loss": 0.0675, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030909154564142227, + "step": 13030, + "valid_targets_mean": 3059.9, + "valid_targets_min": 536 + }, + { + "epoch": 6.8281822943949715, + "grad_norm": 0.4043607686105581, + "learning_rate": 7.382131770211409e-08, + "loss": 0.0705, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03388378396630287, + "step": 13035, + "valid_targets_mean": 3881.0, + "valid_targets_min": 3295 + }, + { + "epoch": 6.830801466736511, + "grad_norm": 0.3627923849401524, + "learning_rate": 7.159588839760423e-08, + "loss": 0.0661, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027576766908168793, + "step": 13040, + "valid_targets_mean": 3287.2, + "valid_targets_min": 820 + }, + { + "epoch": 6.833420639078051, + "grad_norm": 0.6399267355361847, + "learning_rate": 6.940445838049093e-08, + "loss": 0.1067, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.10864797979593277, + "step": 13045, + "valid_targets_mean": 2566.6, + "valid_targets_min": 1046 + }, + { + "epoch": 6.8360398114195915, + "grad_norm": 0.31844573735219384, + "learning_rate": 6.724703138951017e-08, + "loss": 0.0636, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030831724405288696, + "step": 13050, + "valid_targets_mean": 3590.1, + "valid_targets_min": 2719 + }, + { + "epoch": 6.838658983761132, + "grad_norm": 0.683724450813192, + "learning_rate": 6.5123611105391e-08, + "loss": 0.0651, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.050329990684986115, + "step": 13055, + "valid_targets_mean": 1638.1, + "valid_targets_min": 699 + }, + { + "epoch": 6.841278156102671, + "grad_norm": 0.34046232286239747, + "learning_rate": 6.303420115083781e-08, + "loss": 0.0625, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027434110641479492, + "step": 13060, + "valid_targets_mean": 3621.6, + "valid_targets_min": 2353 + }, + { + "epoch": 6.8438973284442115, + "grad_norm": 0.3781497710051084, + "learning_rate": 6.097880509053467e-08, + "loss": 0.064, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02926563285291195, + "step": 13065, + "valid_targets_mean": 2328.1, + "valid_targets_min": 549 + }, + { + "epoch": 6.846516500785752, + "grad_norm": 0.41021539723044204, + "learning_rate": 5.8957426431136554e-08, + "loss": 0.1091, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02801881544291973, + "step": 13070, + "valid_targets_mean": 3330.4, + "valid_targets_min": 908 + }, + { + "epoch": 6.849135673127292, + "grad_norm": 0.3794833887627156, + "learning_rate": 5.69700686212582e-08, + "loss": 0.0915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.06556300818920135, + "step": 13075, + "valid_targets_mean": 3458.4, + "valid_targets_min": 2196 + }, + { + "epoch": 6.8517548454688315, + "grad_norm": 0.36788321802629065, + "learning_rate": 5.5016735051474093e-08, + "loss": 0.0761, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.026008859276771545, + "step": 13080, + "valid_targets_mean": 1252.1, + "valid_targets_min": 598 + }, + { + "epoch": 6.854374017810372, + "grad_norm": 0.8975939479089073, + "learning_rate": 5.309742905431403e-08, + "loss": 0.0895, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0811864584684372, + "step": 13085, + "valid_targets_mean": 1353.1, + "valid_targets_min": 578 + }, + { + "epoch": 6.856993190151912, + "grad_norm": 0.642807708833633, + "learning_rate": 5.121215390424761e-08, + "loss": 0.1136, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.061927855014801025, + "step": 13090, + "valid_targets_mean": 1974.2, + "valid_targets_min": 695 + }, + { + "epoch": 6.859612362493452, + "grad_norm": 0.49724818922628733, + "learning_rate": 4.9360912817695286e-08, + "loss": 0.077, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03489340841770172, + "step": 13095, + "valid_targets_mean": 1724.5, + "valid_targets_min": 691 + }, + { + "epoch": 6.862231534834992, + "grad_norm": 0.33598766841754785, + "learning_rate": 4.7543708953003974e-08, + "loss": 0.093, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02932359278202057, + "step": 13100, + "valid_targets_mean": 3842.6, + "valid_targets_min": 2024 + }, + { + "epoch": 6.864850707176532, + "grad_norm": 0.5600497246010242, + "learning_rate": 4.576054541045149e-08, + "loss": 0.0769, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05074271559715271, + "step": 13105, + "valid_targets_mean": 1651.0, + "valid_targets_min": 523 + }, + { + "epoch": 6.867469879518072, + "grad_norm": 0.3189972848933327, + "learning_rate": 4.4011425232248774e-08, + "loss": 0.0713, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028555691242218018, + "step": 13110, + "valid_targets_mean": 3532.5, + "valid_targets_min": 468 + }, + { + "epoch": 6.870089051859613, + "grad_norm": 0.4060338160268741, + "learning_rate": 4.229635140251542e-08, + "loss": 0.0768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.019814886152744293, + "step": 13115, + "valid_targets_mean": 5298.0, + "valid_targets_min": 1195 + }, + { + "epoch": 6.872708224201152, + "grad_norm": 0.37037097662179636, + "learning_rate": 4.061532684729308e-08, + "loss": 0.0678, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.022655967622995377, + "step": 13120, + "valid_targets_mean": 3562.6, + "valid_targets_min": 459 + }, + { + "epoch": 6.875327396542692, + "grad_norm": 0.3640601085037516, + "learning_rate": 3.896835443452762e-08, + "loss": 0.0664, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03081428073346615, + "step": 13125, + "valid_targets_mean": 3548.0, + "valid_targets_min": 981 + }, + { + "epoch": 6.877946568884233, + "grad_norm": 0.45658087071210757, + "learning_rate": 3.735543697407362e-08, + "loss": 0.076, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04716961458325386, + "step": 13130, + "valid_targets_mean": 2130.1, + "valid_targets_min": 697 + }, + { + "epoch": 6.880565741225773, + "grad_norm": 0.3935577188566529, + "learning_rate": 3.577657721768546e-08, + "loss": 0.0836, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04093249887228012, + "step": 13135, + "valid_targets_mean": 3661.6, + "valid_targets_min": 2759 + }, + { + "epoch": 6.883184913567312, + "grad_norm": 0.39969157210822065, + "learning_rate": 3.423177785900844e-08, + "loss": 0.0842, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.042475707828998566, + "step": 13140, + "valid_targets_mean": 3909.5, + "valid_targets_min": 3391 + }, + { + "epoch": 6.885804085908853, + "grad_norm": 0.3810697165835852, + "learning_rate": 3.272104153358324e-08, + "loss": 0.0684, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.028907202184200287, + "step": 13145, + "valid_targets_mean": 3319.9, + "valid_targets_min": 1099 + }, + { + "epoch": 6.888423258250393, + "grad_norm": 0.40129194616477243, + "learning_rate": 3.12443708188348e-08, + "loss": 0.0712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0376894548535347, + "step": 13150, + "valid_targets_mean": 3717.0, + "valid_targets_min": 1874 + }, + { + "epoch": 6.891042430591933, + "grad_norm": 0.35289624638750433, + "learning_rate": 2.9801768234065663e-08, + "loss": 0.0808, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030940137803554535, + "step": 13155, + "valid_targets_mean": 3263.5, + "valid_targets_min": 724 + }, + { + "epoch": 6.893661602933473, + "grad_norm": 0.3177800774401248, + "learning_rate": 2.8393236240464862e-08, + "loss": 0.0589, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027186524122953415, + "step": 13160, + "valid_targets_mean": 3707.5, + "valid_targets_min": 2769 + }, + { + "epoch": 6.896280775275013, + "grad_norm": 0.43924295200663305, + "learning_rate": 2.7018777241083484e-08, + "loss": 0.0782, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.032714247703552246, + "step": 13165, + "valid_targets_mean": 2815.8, + "valid_targets_min": 645 + }, + { + "epoch": 6.898899947616553, + "grad_norm": 0.3860992881253904, + "learning_rate": 2.5678393580852446e-08, + "loss": 0.0733, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02999677136540413, + "step": 13170, + "valid_targets_mean": 3488.8, + "valid_targets_min": 600 + }, + { + "epoch": 6.901519119958094, + "grad_norm": 0.40402674816483036, + "learning_rate": 2.4372087546558064e-08, + "loss": 0.0873, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03606380149722099, + "step": 13175, + "valid_targets_mean": 3373.4, + "valid_targets_min": 1648 + }, + { + "epoch": 6.904138292299633, + "grad_norm": 0.5102169471008514, + "learning_rate": 2.309986136685316e-08, + "loss": 0.0756, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04222027212381363, + "step": 13180, + "valid_targets_mean": 2555.0, + "valid_targets_min": 821 + }, + { + "epoch": 6.906757464641173, + "grad_norm": 0.8465620208298412, + "learning_rate": 2.186171721225039e-08, + "loss": 0.0858, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05596958100795746, + "step": 13185, + "valid_targets_mean": 867.1, + "valid_targets_min": 612 + }, + { + "epoch": 6.909376636982714, + "grad_norm": 0.5492682172792553, + "learning_rate": 2.0657657195106706e-08, + "loss": 0.0751, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05503883212804794, + "step": 13190, + "valid_targets_mean": 2202.1, + "valid_targets_min": 830 + }, + { + "epoch": 6.911995809324253, + "grad_norm": 0.3972474717885821, + "learning_rate": 1.948768336963891e-08, + "loss": 0.0866, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033593904227018356, + "step": 13195, + "valid_targets_mean": 3027.0, + "valid_targets_min": 1146 + }, + { + "epoch": 6.914614981665793, + "grad_norm": 0.4368583506670117, + "learning_rate": 1.8351797731908093e-08, + "loss": 0.0748, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04359440132975578, + "step": 13200, + "valid_targets_mean": 3296.6, + "valid_targets_min": 2760 + }, + { + "epoch": 6.917234154007334, + "grad_norm": 0.597637523428214, + "learning_rate": 1.725000221981077e-08, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.047540731728076935, + "step": 13205, + "valid_targets_mean": 1469.5, + "valid_targets_min": 628 + }, + { + "epoch": 6.919853326348874, + "grad_norm": 0.6648468760956254, + "learning_rate": 1.61822987130944e-08, + "loss": 0.0889, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.071270652115345, + "step": 13210, + "valid_targets_mean": 1752.1, + "valid_targets_min": 691 + }, + { + "epoch": 6.922472498690414, + "grad_norm": 0.3151167981400266, + "learning_rate": 1.5148689033333e-08, + "loss": 0.0677, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.01949170231819153, + "step": 13215, + "valid_targets_mean": 3863.9, + "valid_targets_min": 3613 + }, + { + "epoch": 6.925091671031954, + "grad_norm": 0.3609607709775375, + "learning_rate": 1.4149174943944854e-08, + "loss": 0.0681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.034584157168865204, + "step": 13220, + "valid_targets_mean": 4304.2, + "valid_targets_min": 1773 + }, + { + "epoch": 6.927710843373494, + "grad_norm": 0.3766390380419614, + "learning_rate": 1.3183758150165926e-08, + "loss": 0.0731, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03962358832359314, + "step": 13225, + "valid_targets_mean": 1971.9, + "valid_targets_min": 762 + }, + { + "epoch": 6.930330015715034, + "grad_norm": 0.4152652927103812, + "learning_rate": 1.2252440299072022e-08, + "loss": 0.0652, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.048780351877212524, + "step": 13230, + "valid_targets_mean": 4554.1, + "valid_targets_min": 946 + }, + { + "epoch": 6.932949188056574, + "grad_norm": 0.4052622382021391, + "learning_rate": 1.1355222979556601e-08, + "loss": 0.0611, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.043601371347904205, + "step": 13235, + "valid_targets_mean": 5015.5, + "valid_targets_min": 4104 + }, + { + "epoch": 6.935568360398114, + "grad_norm": 0.3748078819706086, + "learning_rate": 1.049210772233522e-08, + "loss": 0.0812, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033804330974817276, + "step": 13240, + "valid_targets_mean": 3064.4, + "valid_targets_min": 739 + }, + { + "epoch": 6.938187532739654, + "grad_norm": 0.4639722902263239, + "learning_rate": 9.663095999947746e-09, + "loss": 0.0912, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03895917534828186, + "step": 13245, + "valid_targets_mean": 3863.6, + "valid_targets_min": 3550 + }, + { + "epoch": 6.940806705081195, + "grad_norm": 0.39368848340723156, + "learning_rate": 8.868189226742818e-09, + "loss": 0.0711, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.040794141590595245, + "step": 13250, + "valid_targets_mean": 3474.9, + "valid_targets_min": 704 + }, + { + "epoch": 6.943425877422735, + "grad_norm": 0.37828505313020083, + "learning_rate": 8.107388758893386e-09, + "loss": 0.0768, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030433308333158493, + "step": 13255, + "valid_targets_mean": 2568.8, + "valid_targets_min": 774 + }, + { + "epoch": 6.946045049764274, + "grad_norm": 0.3168199773597853, + "learning_rate": 7.380695894376733e-09, + "loss": 0.0915, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027098558843135834, + "step": 13260, + "valid_targets_mean": 3486.2, + "valid_targets_min": 888 + }, + { + "epoch": 6.948664222105815, + "grad_norm": 0.4355654039716683, + "learning_rate": 6.688111872983349e-09, + "loss": 0.0685, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03660779073834419, + "step": 13265, + "valid_targets_mean": 3371.6, + "valid_targets_min": 1048 + }, + { + "epoch": 6.951283394447355, + "grad_norm": 0.39094867570324077, + "learning_rate": 6.0296378763125e-09, + "loss": 0.0757, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0327947624027729, + "step": 13270, + "valid_targets_mean": 2941.5, + "valid_targets_min": 722 + }, + { + "epoch": 6.953902566788894, + "grad_norm": 0.49476977079908663, + "learning_rate": 5.405275027765555e-09, + "loss": 0.0736, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.045730821788311005, + "step": 13275, + "valid_targets_mean": 2359.2, + "valid_targets_min": 865 + }, + { + "epoch": 6.956521739130435, + "grad_norm": 0.49187675984686874, + "learning_rate": 4.815024392552659e-09, + "loss": 0.1059, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03030387870967388, + "step": 13280, + "valid_targets_mean": 1747.0, + "valid_targets_min": 570 + }, + { + "epoch": 6.959140911471975, + "grad_norm": 0.4564650516018697, + "learning_rate": 4.258886977683841e-09, + "loss": 0.0892, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.044109731912612915, + "step": 13285, + "valid_targets_mean": 3704.2, + "valid_targets_min": 2811 + }, + { + "epoch": 6.961760083813515, + "grad_norm": 0.422563426262655, + "learning_rate": 3.7368637319690204e-09, + "loss": 0.0712, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030791914090514183, + "step": 13290, + "valid_targets_mean": 1749.1, + "valid_targets_min": 635 + }, + { + "epoch": 6.9643792561550555, + "grad_norm": 0.4260118166921828, + "learning_rate": 3.2489555460157863e-09, + "loss": 0.0695, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.035841863602399826, + "step": 13295, + "valid_targets_mean": 2985.1, + "valid_targets_min": 1591 + }, + { + "epoch": 6.966998428496595, + "grad_norm": 0.39989493932010906, + "learning_rate": 2.795163252231614e-09, + "loss": 0.0687, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.030270284041762352, + "step": 13300, + "valid_targets_mean": 2912.6, + "valid_targets_min": 800 + }, + { + "epoch": 6.969617600838135, + "grad_norm": 0.3871237712611208, + "learning_rate": 2.3754876248216484e-09, + "loss": 0.0618, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.033040743321180344, + "step": 13305, + "valid_targets_mean": 3084.9, + "valid_targets_min": 860 + }, + { + "epoch": 6.9722367731796755, + "grad_norm": 0.4992136474342194, + "learning_rate": 1.9899293797753792e-09, + "loss": 0.0862, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03879676014184952, + "step": 13310, + "valid_targets_mean": 2482.6, + "valid_targets_min": 878 + }, + { + "epoch": 6.974855945521215, + "grad_norm": 0.3685887101689459, + "learning_rate": 1.6384891748910669e-09, + "loss": 0.0614, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03403729200363159, + "step": 13315, + "valid_targets_mean": 3403.2, + "valid_targets_min": 1024 + }, + { + "epoch": 6.977475117862755, + "grad_norm": 0.43975060817321204, + "learning_rate": 1.3211676097446558e-09, + "loss": 0.0682, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.03496960550546646, + "step": 13320, + "valid_targets_mean": 2173.1, + "valid_targets_min": 786 + }, + { + "epoch": 6.9800942902042955, + "grad_norm": 0.4478083620308506, + "learning_rate": 1.0379652257141991e-09, + "loss": 0.0697, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.0458064079284668, + "step": 13325, + "valid_targets_mean": 3233.9, + "valid_targets_min": 2573 + }, + { + "epoch": 6.982713462545836, + "grad_norm": 0.3557386259180458, + "learning_rate": 7.888825059576555e-10, + "loss": 0.072, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.037160474807024, + "step": 13330, + "valid_targets_mean": 3821.9, + "valid_targets_min": 2274 + }, + { + "epoch": 6.985332634887375, + "grad_norm": 0.4311579463314463, + "learning_rate": 5.739198754350916e-10, + "loss": 0.0635, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.04020063206553459, + "step": 13335, + "valid_targets_mean": 3791.2, + "valid_targets_min": 892 + }, + { + "epoch": 6.9879518072289155, + "grad_norm": 0.3781394475299389, + "learning_rate": 3.9307770088203855e-10, + "loss": 0.0681, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.02724323235452175, + "step": 13340, + "valid_targets_mean": 4820.4, + "valid_targets_min": 4491 + }, + { + "epoch": 6.990570979570456, + "grad_norm": 0.5795160774312458, + "learning_rate": 2.463562908316952e-10, + "loss": 0.1071, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.08761267364025116, + "step": 13345, + "valid_targets_mean": 2159.1, + "valid_targets_min": 709 + }, + { + "epoch": 6.993190151911996, + "grad_norm": 0.32974821703773993, + "learning_rate": 1.3375589559938562e-10, + "loss": 0.118, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.021897807717323303, + "step": 13350, + "valid_targets_mean": 3430.8, + "valid_targets_min": 1097 + }, + { + "epoch": 6.9958093242535355, + "grad_norm": 0.39141871864807426, + "learning_rate": 5.5276707291440637e-11, + "loss": 0.0702, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.029825208708643913, + "step": 13355, + "valid_targets_mean": 3032.0, + "valid_targets_min": 840 + }, + { + "epoch": 6.998428496595076, + "grad_norm": 0.3386180818814195, + "learning_rate": 1.0918859796316127e-11, + "loss": 0.0648, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.027408761903643608, + "step": 13360, + "valid_targets_mean": 3812.2, + "valid_targets_min": 3106 + }, + { + "epoch": 7.0, + "loss_nan_ranks": 0, + "loss_rank_avg": 0.05363672971725464, + "step": 13363, + "total_flos": 4.5166057635687956e+18, + "train_loss": 0.18181621480833796, + "train_runtime": 98002.5659, + "train_samples_per_second": 2.181, + "train_steps_per_second": 0.136, + "valid_targets_mean": 5394.5, + "valid_targets_min": 3513 + } + ], + "logging_steps": 5, + "max_steps": 13363, + "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": 4.5166057635687956e+18, + "train_batch_size": 1, + "trial_name": null, + "trial_params": null +}