| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 5.0, |
| "eval_steps": 500, |
| "global_step": 3125, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.008, |
| "grad_norm": 8.019209505645138, |
| "learning_rate": 5.111821086261981e-07, |
| "loss": 0.9704, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.9074519872665405, |
| "step": 5, |
| "valid_targets_mean": 4634.1, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 0.016, |
| "grad_norm": 8.056913572069162, |
| "learning_rate": 1.1501597444089457e-06, |
| "loss": 0.9831, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.9801443815231323, |
| "step": 10, |
| "valid_targets_mean": 3432.8, |
| "valid_targets_min": 888 |
| }, |
| { |
| "epoch": 0.024, |
| "grad_norm": 5.890567315411599, |
| "learning_rate": 1.7891373801916933e-06, |
| "loss": 0.9216, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.9462229609489441, |
| "step": 15, |
| "valid_targets_mean": 3846.5, |
| "valid_targets_min": 622 |
| }, |
| { |
| "epoch": 0.032, |
| "grad_norm": 4.815413753270494, |
| "learning_rate": 2.428115015974441e-06, |
| "loss": 0.9003, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.9328570365905762, |
| "step": 20, |
| "valid_targets_mean": 2826.7, |
| "valid_targets_min": 592 |
| }, |
| { |
| "epoch": 0.04, |
| "grad_norm": 3.2132576974472262, |
| "learning_rate": 3.0670926517571885e-06, |
| "loss": 0.9021, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.9593732357025146, |
| "step": 25, |
| "valid_targets_mean": 5006.8, |
| "valid_targets_min": 602 |
| }, |
| { |
| "epoch": 0.048, |
| "grad_norm": 2.1277286968257494, |
| "learning_rate": 3.7060702875399364e-06, |
| "loss": 0.8146, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.8108715415000916, |
| "step": 30, |
| "valid_targets_mean": 4988.4, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 0.056, |
| "grad_norm": 1.8316025070705764, |
| "learning_rate": 4.345047923322684e-06, |
| "loss": 0.8371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.8360515832901001, |
| "step": 35, |
| "valid_targets_mean": 2703.9, |
| "valid_targets_min": 958 |
| }, |
| { |
| "epoch": 0.064, |
| "grad_norm": 0.974656430471622, |
| "learning_rate": 4.984025559105431e-06, |
| "loss": 0.8143, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.738986611366272, |
| "step": 40, |
| "valid_targets_mean": 5619.8, |
| "valid_targets_min": 496 |
| }, |
| { |
| "epoch": 0.072, |
| "grad_norm": 1.2003528549510873, |
| "learning_rate": 5.623003194888179e-06, |
| "loss": 0.8001, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.834333598613739, |
| "step": 45, |
| "valid_targets_mean": 2614.9, |
| "valid_targets_min": 534 |
| }, |
| { |
| "epoch": 0.08, |
| "grad_norm": 0.9375989258322925, |
| "learning_rate": 6.261980830670928e-06, |
| "loss": 0.7226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7906558513641357, |
| "step": 50, |
| "valid_targets_mean": 3902.1, |
| "valid_targets_min": 678 |
| }, |
| { |
| "epoch": 0.088, |
| "grad_norm": 0.947503985003781, |
| "learning_rate": 6.900958466453675e-06, |
| "loss": 0.738, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7608993053436279, |
| "step": 55, |
| "valid_targets_mean": 2622.6, |
| "valid_targets_min": 640 |
| }, |
| { |
| "epoch": 0.096, |
| "grad_norm": 1.062618699465369, |
| "learning_rate": 7.5399361022364225e-06, |
| "loss": 0.7835, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.8355079293251038, |
| "step": 60, |
| "valid_targets_mean": 2038.3, |
| "valid_targets_min": 616 |
| }, |
| { |
| "epoch": 0.104, |
| "grad_norm": 0.6519054791996576, |
| "learning_rate": 8.17891373801917e-06, |
| "loss": 0.7377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6525523662567139, |
| "step": 65, |
| "valid_targets_mean": 4931.8, |
| "valid_targets_min": 879 |
| }, |
| { |
| "epoch": 0.112, |
| "grad_norm": 0.8181807803276753, |
| "learning_rate": 8.817891373801917e-06, |
| "loss": 0.7468, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7478166818618774, |
| "step": 70, |
| "valid_targets_mean": 2875.1, |
| "valid_targets_min": 684 |
| }, |
| { |
| "epoch": 0.12, |
| "grad_norm": 0.8077200310163509, |
| "learning_rate": 9.456869009584665e-06, |
| "loss": 0.7, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7476080656051636, |
| "step": 75, |
| "valid_targets_mean": 2929.1, |
| "valid_targets_min": 918 |
| }, |
| { |
| "epoch": 0.128, |
| "grad_norm": 0.6645101628172224, |
| "learning_rate": 1.0095846645367413e-05, |
| "loss": 0.6873, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7070491313934326, |
| "step": 80, |
| "valid_targets_mean": 3938.2, |
| "valid_targets_min": 1116 |
| }, |
| { |
| "epoch": 0.136, |
| "grad_norm": 0.5668806150237382, |
| "learning_rate": 1.073482428115016e-05, |
| "loss": 0.6795, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6539013385772705, |
| "step": 85, |
| "valid_targets_mean": 4791.0, |
| "valid_targets_min": 575 |
| }, |
| { |
| "epoch": 0.144, |
| "grad_norm": 0.631829978989927, |
| "learning_rate": 1.1373801916932907e-05, |
| "loss": 0.6727, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6462942957878113, |
| "step": 90, |
| "valid_targets_mean": 4125.3, |
| "valid_targets_min": 746 |
| }, |
| { |
| "epoch": 0.152, |
| "grad_norm": 0.7522145077585198, |
| "learning_rate": 1.2012779552715656e-05, |
| "loss": 0.6913, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6524339914321899, |
| "step": 95, |
| "valid_targets_mean": 2768.8, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 0.16, |
| "grad_norm": 0.6436788258939267, |
| "learning_rate": 1.2651757188498404e-05, |
| "loss": 0.6746, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6656085848808289, |
| "step": 100, |
| "valid_targets_mean": 4070.2, |
| "valid_targets_min": 693 |
| }, |
| { |
| "epoch": 0.168, |
| "grad_norm": 0.6829360551307666, |
| "learning_rate": 1.329073482428115e-05, |
| "loss": 0.6581, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.696590006351471, |
| "step": 105, |
| "valid_targets_mean": 3549.9, |
| "valid_targets_min": 985 |
| }, |
| { |
| "epoch": 0.176, |
| "grad_norm": 0.6518448907533483, |
| "learning_rate": 1.39297124600639e-05, |
| "loss": 0.6693, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6982806921005249, |
| "step": 110, |
| "valid_targets_mean": 3914.5, |
| "valid_targets_min": 923 |
| }, |
| { |
| "epoch": 0.184, |
| "grad_norm": 0.7108542582881471, |
| "learning_rate": 1.4568690095846648e-05, |
| "loss": 0.6437, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6319302916526794, |
| "step": 115, |
| "valid_targets_mean": 3878.2, |
| "valid_targets_min": 667 |
| }, |
| { |
| "epoch": 0.192, |
| "grad_norm": 0.6111508324238974, |
| "learning_rate": 1.5207667731629394e-05, |
| "loss": 0.6237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6298456788063049, |
| "step": 120, |
| "valid_targets_mean": 4183.7, |
| "valid_targets_min": 831 |
| }, |
| { |
| "epoch": 0.2, |
| "grad_norm": 0.636766820540127, |
| "learning_rate": 1.584664536741214e-05, |
| "loss": 0.6248, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6381227970123291, |
| "step": 125, |
| "valid_targets_mean": 4150.9, |
| "valid_targets_min": 279 |
| }, |
| { |
| "epoch": 0.208, |
| "grad_norm": 0.7297687830940379, |
| "learning_rate": 1.648562300319489e-05, |
| "loss": 0.6313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6899981498718262, |
| "step": 130, |
| "valid_targets_mean": 3481.6, |
| "valid_targets_min": 879 |
| }, |
| { |
| "epoch": 0.216, |
| "grad_norm": 0.6994014106747448, |
| "learning_rate": 1.712460063897764e-05, |
| "loss": 0.6099, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.641595184803009, |
| "step": 135, |
| "valid_targets_mean": 4528.6, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 0.224, |
| "grad_norm": 0.682372825217958, |
| "learning_rate": 1.7763578274760385e-05, |
| "loss": 0.6255, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6313050985336304, |
| "step": 140, |
| "valid_targets_mean": 4437.3, |
| "valid_targets_min": 572 |
| }, |
| { |
| "epoch": 0.232, |
| "grad_norm": 0.6870311865242572, |
| "learning_rate": 1.840255591054313e-05, |
| "loss": 0.6237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5983741283416748, |
| "step": 145, |
| "valid_targets_mean": 3244.9, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 0.24, |
| "grad_norm": 0.7883183108248433, |
| "learning_rate": 1.904153354632588e-05, |
| "loss": 0.6184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6510107517242432, |
| "step": 150, |
| "valid_targets_mean": 3022.1, |
| "valid_targets_min": 687 |
| }, |
| { |
| "epoch": 0.248, |
| "grad_norm": 0.6559928431572688, |
| "learning_rate": 1.9680511182108627e-05, |
| "loss": 0.6168, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6277835369110107, |
| "step": 155, |
| "valid_targets_mean": 4691.1, |
| "valid_targets_min": 774 |
| }, |
| { |
| "epoch": 0.256, |
| "grad_norm": 0.7040494425503686, |
| "learning_rate": 2.0319488817891376e-05, |
| "loss": 0.6178, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5613512992858887, |
| "step": 160, |
| "valid_targets_mean": 3659.9, |
| "valid_targets_min": 691 |
| }, |
| { |
| "epoch": 0.264, |
| "grad_norm": 0.7293885471955911, |
| "learning_rate": 2.0958466453674126e-05, |
| "loss": 0.578, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.546758234500885, |
| "step": 165, |
| "valid_targets_mean": 3741.1, |
| "valid_targets_min": 651 |
| }, |
| { |
| "epoch": 0.272, |
| "grad_norm": 0.772438514428517, |
| "learning_rate": 2.1597444089456872e-05, |
| "loss": 0.5705, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5695148706436157, |
| "step": 170, |
| "valid_targets_mean": 3372.0, |
| "valid_targets_min": 744 |
| }, |
| { |
| "epoch": 0.28, |
| "grad_norm": 0.8184517511660472, |
| "learning_rate": 2.2236421725239618e-05, |
| "loss": 0.6274, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6200259923934937, |
| "step": 175, |
| "valid_targets_mean": 2952.8, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 0.288, |
| "grad_norm": 0.6401835135693486, |
| "learning_rate": 2.2875399361022364e-05, |
| "loss": 0.5871, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6663122773170471, |
| "step": 180, |
| "valid_targets_mean": 4482.8, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 0.296, |
| "grad_norm": 0.562579817398789, |
| "learning_rate": 2.3514376996805114e-05, |
| "loss": 0.594, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5693539977073669, |
| "step": 185, |
| "valid_targets_mean": 5274.5, |
| "valid_targets_min": 669 |
| }, |
| { |
| "epoch": 0.304, |
| "grad_norm": 3.023582033179472, |
| "learning_rate": 2.415335463258786e-05, |
| "loss": 0.5726, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6133388876914978, |
| "step": 190, |
| "valid_targets_mean": 3461.5, |
| "valid_targets_min": 534 |
| }, |
| { |
| "epoch": 0.312, |
| "grad_norm": 0.63776000505745, |
| "learning_rate": 2.4792332268370606e-05, |
| "loss": 0.6095, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5332133173942566, |
| "step": 195, |
| "valid_targets_mean": 4770.1, |
| "valid_targets_min": 613 |
| }, |
| { |
| "epoch": 0.32, |
| "grad_norm": 0.6537457497646995, |
| "learning_rate": 2.543130990415336e-05, |
| "loss": 0.5961, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6032932996749878, |
| "step": 200, |
| "valid_targets_mean": 4235.5, |
| "valid_targets_min": 418 |
| }, |
| { |
| "epoch": 0.328, |
| "grad_norm": 0.679784159844079, |
| "learning_rate": 2.6070287539936105e-05, |
| "loss": 0.5781, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5432563424110413, |
| "step": 205, |
| "valid_targets_mean": 3482.2, |
| "valid_targets_min": 474 |
| }, |
| { |
| "epoch": 0.336, |
| "grad_norm": 0.6536420335215318, |
| "learning_rate": 2.670926517571885e-05, |
| "loss": 0.5527, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5386021137237549, |
| "step": 210, |
| "valid_targets_mean": 4072.3, |
| "valid_targets_min": 288 |
| }, |
| { |
| "epoch": 0.344, |
| "grad_norm": 0.7579600175695036, |
| "learning_rate": 2.73482428115016e-05, |
| "loss": 0.5917, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5911605358123779, |
| "step": 215, |
| "valid_targets_mean": 2998.2, |
| "valid_targets_min": 506 |
| }, |
| { |
| "epoch": 0.352, |
| "grad_norm": 0.8790312000878335, |
| "learning_rate": 2.7987220447284347e-05, |
| "loss": 0.6193, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5623832941055298, |
| "step": 220, |
| "valid_targets_mean": 2856.6, |
| "valid_targets_min": 660 |
| }, |
| { |
| "epoch": 0.36, |
| "grad_norm": 0.8411714839123007, |
| "learning_rate": 2.8626198083067093e-05, |
| "loss": 0.5871, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6712679862976074, |
| "step": 225, |
| "valid_targets_mean": 3072.1, |
| "valid_targets_min": 810 |
| }, |
| { |
| "epoch": 0.368, |
| "grad_norm": 0.6480362190445711, |
| "learning_rate": 2.9265175718849843e-05, |
| "loss": 0.5827, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5256530046463013, |
| "step": 230, |
| "valid_targets_mean": 3665.3, |
| "valid_targets_min": 536 |
| }, |
| { |
| "epoch": 0.376, |
| "grad_norm": 0.6347344055975906, |
| "learning_rate": 2.9904153354632592e-05, |
| "loss": 0.5871, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6053222417831421, |
| "step": 235, |
| "valid_targets_mean": 4444.1, |
| "valid_targets_min": 850 |
| }, |
| { |
| "epoch": 0.384, |
| "grad_norm": 0.7522140681943507, |
| "learning_rate": 3.054313099041534e-05, |
| "loss": 0.5885, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6026010513305664, |
| "step": 240, |
| "valid_targets_mean": 3201.2, |
| "valid_targets_min": 447 |
| }, |
| { |
| "epoch": 0.392, |
| "grad_norm": 0.59444256273076, |
| "learning_rate": 3.1182108626198084e-05, |
| "loss": 0.5656, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5105293989181519, |
| "step": 245, |
| "valid_targets_mean": 4720.0, |
| "valid_targets_min": 511 |
| }, |
| { |
| "epoch": 0.4, |
| "grad_norm": 0.7082748798819095, |
| "learning_rate": 3.1821086261980834e-05, |
| "loss": 0.5715, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5332015752792358, |
| "step": 250, |
| "valid_targets_mean": 3776.2, |
| "valid_targets_min": 552 |
| }, |
| { |
| "epoch": 0.408, |
| "grad_norm": 0.5935423684362258, |
| "learning_rate": 3.246006389776358e-05, |
| "loss": 0.5579, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5111772418022156, |
| "step": 255, |
| "valid_targets_mean": 4674.4, |
| "valid_targets_min": 667 |
| }, |
| { |
| "epoch": 0.416, |
| "grad_norm": 0.8490750724860248, |
| "learning_rate": 3.3099041533546326e-05, |
| "loss": 0.5582, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5234127044677734, |
| "step": 260, |
| "valid_targets_mean": 4834.0, |
| "valid_targets_min": 767 |
| }, |
| { |
| "epoch": 0.424, |
| "grad_norm": 0.6787686121432547, |
| "learning_rate": 3.3738019169329076e-05, |
| "loss": 0.5422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5584520101547241, |
| "step": 265, |
| "valid_targets_mean": 3572.9, |
| "valid_targets_min": 739 |
| }, |
| { |
| "epoch": 0.432, |
| "grad_norm": 0.7210865201308775, |
| "learning_rate": 3.4376996805111825e-05, |
| "loss": 0.5439, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5728445649147034, |
| "step": 270, |
| "valid_targets_mean": 3147.6, |
| "valid_targets_min": 709 |
| }, |
| { |
| "epoch": 0.44, |
| "grad_norm": 0.9390958887116584, |
| "learning_rate": 3.5015974440894575e-05, |
| "loss": 0.572, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6330382823944092, |
| "step": 275, |
| "valid_targets_mean": 2163.2, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 0.448, |
| "grad_norm": 0.7172744444294396, |
| "learning_rate": 3.565495207667732e-05, |
| "loss": 0.5708, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.547631561756134, |
| "step": 280, |
| "valid_targets_mean": 3426.3, |
| "valid_targets_min": 492 |
| }, |
| { |
| "epoch": 0.456, |
| "grad_norm": 0.6587837222708766, |
| "learning_rate": 3.629392971246007e-05, |
| "loss": 0.5714, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5626546144485474, |
| "step": 285, |
| "valid_targets_mean": 4249.6, |
| "valid_targets_min": 615 |
| }, |
| { |
| "epoch": 0.464, |
| "grad_norm": 0.5946489699255438, |
| "learning_rate": 3.6932907348242816e-05, |
| "loss": 0.5554, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6051210165023804, |
| "step": 290, |
| "valid_targets_mean": 4729.2, |
| "valid_targets_min": 655 |
| }, |
| { |
| "epoch": 0.472, |
| "grad_norm": 0.7892760841457642, |
| "learning_rate": 3.757188498402556e-05, |
| "loss": 0.5811, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6125881671905518, |
| "step": 295, |
| "valid_targets_mean": 3489.5, |
| "valid_targets_min": 607 |
| }, |
| { |
| "epoch": 0.48, |
| "grad_norm": 0.6623319928800957, |
| "learning_rate": 3.821086261980831e-05, |
| "loss": 0.5852, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5646267533302307, |
| "step": 300, |
| "valid_targets_mean": 3532.5, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 0.488, |
| "grad_norm": 0.6226375467862089, |
| "learning_rate": 3.884984025559106e-05, |
| "loss": 0.5714, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5953035354614258, |
| "step": 305, |
| "valid_targets_mean": 4777.0, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 0.496, |
| "grad_norm": 0.7584022572749837, |
| "learning_rate": 3.94888178913738e-05, |
| "loss": 0.5374, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5502133369445801, |
| "step": 310, |
| "valid_targets_mean": 3806.0, |
| "valid_targets_min": 679 |
| }, |
| { |
| "epoch": 0.504, |
| "grad_norm": 0.7226522137453772, |
| "learning_rate": 3.9999987518434296e-05, |
| "loss": 0.5273, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5224813222885132, |
| "step": 315, |
| "valid_targets_mean": 3150.1, |
| "valid_targets_min": 655 |
| }, |
| { |
| "epoch": 0.512, |
| "grad_norm": 0.6141445710145204, |
| "learning_rate": 3.999955066527015e-05, |
| "loss": 0.5718, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5544633269309998, |
| "step": 320, |
| "valid_targets_mean": 4412.2, |
| "valid_targets_min": 535 |
| }, |
| { |
| "epoch": 0.52, |
| "grad_norm": 0.8461142526243931, |
| "learning_rate": 3.999848974939926e-05, |
| "loss": 0.5414, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5990142822265625, |
| "step": 325, |
| "valid_targets_mean": 2348.9, |
| "valid_targets_min": 499 |
| }, |
| { |
| "epoch": 0.528, |
| "grad_norm": 0.8250606491282058, |
| "learning_rate": 3.999680480392626e-05, |
| "loss": 0.5356, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5679922103881836, |
| "step": 330, |
| "valid_targets_mean": 3415.2, |
| "valid_targets_min": 629 |
| }, |
| { |
| "epoch": 0.536, |
| "grad_norm": 0.637073771711946, |
| "learning_rate": 3.999449588142792e-05, |
| "loss": 0.534, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4861408770084381, |
| "step": 335, |
| "valid_targets_mean": 3706.7, |
| "valid_targets_min": 631 |
| }, |
| { |
| "epoch": 0.544, |
| "grad_norm": 0.8398228483120712, |
| "learning_rate": 3.9991563053951476e-05, |
| "loss": 0.5715, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.646068811416626, |
| "step": 340, |
| "valid_targets_mean": 2273.1, |
| "valid_targets_min": 433 |
| }, |
| { |
| "epoch": 0.552, |
| "grad_norm": 0.6121671335659467, |
| "learning_rate": 3.99880064130124e-05, |
| "loss": 0.5564, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5839049816131592, |
| "step": 345, |
| "valid_targets_mean": 4629.9, |
| "valid_targets_min": 878 |
| }, |
| { |
| "epoch": 0.56, |
| "grad_norm": 0.7358980937968385, |
| "learning_rate": 3.9983826069591535e-05, |
| "loss": 0.5492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5735797882080078, |
| "step": 350, |
| "valid_targets_mean": 3127.5, |
| "valid_targets_min": 725 |
| }, |
| { |
| "epoch": 0.568, |
| "grad_norm": 0.6083717163603471, |
| "learning_rate": 3.997902215413163e-05, |
| "loss": 0.5249, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5112648010253906, |
| "step": 355, |
| "valid_targets_mean": 4718.0, |
| "valid_targets_min": 507 |
| }, |
| { |
| "epoch": 0.576, |
| "grad_norm": 0.6072026819837963, |
| "learning_rate": 3.997359481653327e-05, |
| "loss": 0.5393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5504653453826904, |
| "step": 360, |
| "valid_targets_mean": 4234.2, |
| "valid_targets_min": 578 |
| }, |
| { |
| "epoch": 0.584, |
| "grad_norm": 0.5970402191237862, |
| "learning_rate": 3.996754422615023e-05, |
| "loss": 0.547, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5461069345474243, |
| "step": 365, |
| "valid_targets_mean": 4407.4, |
| "valid_targets_min": 606 |
| }, |
| { |
| "epoch": 0.592, |
| "grad_norm": 0.7672984290731282, |
| "learning_rate": 3.996087057178411e-05, |
| "loss": 0.5539, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6059372425079346, |
| "step": 370, |
| "valid_targets_mean": 3555.2, |
| "valid_targets_min": 599 |
| }, |
| { |
| "epoch": 0.6, |
| "grad_norm": 0.7939676092215848, |
| "learning_rate": 3.995357406167856e-05, |
| "loss": 0.5563, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5978292226791382, |
| "step": 375, |
| "valid_targets_mean": 2776.0, |
| "valid_targets_min": 734 |
| }, |
| { |
| "epoch": 0.608, |
| "grad_norm": 0.6525335154383021, |
| "learning_rate": 3.994565492351267e-05, |
| "loss": 0.5352, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5540485382080078, |
| "step": 380, |
| "valid_targets_mean": 4123.6, |
| "valid_targets_min": 583 |
| }, |
| { |
| "epoch": 0.616, |
| "grad_norm": 0.6989698695789527, |
| "learning_rate": 3.993711340439394e-05, |
| "loss": 0.5343, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6023855209350586, |
| "step": 385, |
| "valid_targets_mean": 3758.4, |
| "valid_targets_min": 463 |
| }, |
| { |
| "epoch": 0.624, |
| "grad_norm": 0.7124652823560258, |
| "learning_rate": 3.9927949770850535e-05, |
| "loss": 0.5326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5666008591651917, |
| "step": 390, |
| "valid_targets_mean": 3212.3, |
| "valid_targets_min": 754 |
| }, |
| { |
| "epoch": 0.632, |
| "grad_norm": 0.6842076356533583, |
| "learning_rate": 3.991816430882297e-05, |
| "loss": 0.5375, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5813860297203064, |
| "step": 395, |
| "valid_targets_mean": 4066.8, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 0.64, |
| "grad_norm": 0.529144826449527, |
| "learning_rate": 3.9907757323655206e-05, |
| "loss": 0.5396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.555700957775116, |
| "step": 400, |
| "valid_targets_mean": 7514.6, |
| "valid_targets_min": 874 |
| }, |
| { |
| "epoch": 0.648, |
| "grad_norm": 0.687588424871281, |
| "learning_rate": 3.98967291400851e-05, |
| "loss": 0.5169, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5300942659378052, |
| "step": 405, |
| "valid_targets_mean": 3809.2, |
| "valid_targets_min": 695 |
| }, |
| { |
| "epoch": 0.656, |
| "grad_norm": 0.6988433600033485, |
| "learning_rate": 3.98850801022343e-05, |
| "loss": 0.5232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44089508056640625, |
| "step": 410, |
| "valid_targets_mean": 4746.0, |
| "valid_targets_min": 570 |
| }, |
| { |
| "epoch": 0.664, |
| "grad_norm": 0.6587643148830101, |
| "learning_rate": 3.987281057359746e-05, |
| "loss": 0.5335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5711066126823425, |
| "step": 415, |
| "valid_targets_mean": 3932.2, |
| "valid_targets_min": 507 |
| }, |
| { |
| "epoch": 0.672, |
| "grad_norm": 0.6308013522460244, |
| "learning_rate": 3.985992093703096e-05, |
| "loss": 0.5377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5582001805305481, |
| "step": 420, |
| "valid_targets_mean": 3887.5, |
| "valid_targets_min": 589 |
| }, |
| { |
| "epoch": 0.68, |
| "grad_norm": 0.5418196566696337, |
| "learning_rate": 3.98464115947409e-05, |
| "loss": 0.5323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49433648586273193, |
| "step": 425, |
| "valid_targets_mean": 5441.3, |
| "valid_targets_min": 720 |
| }, |
| { |
| "epoch": 0.688, |
| "grad_norm": 0.7175445202648967, |
| "learning_rate": 3.9832282968270595e-05, |
| "loss": 0.5287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.552021861076355, |
| "step": 430, |
| "valid_targets_mean": 2823.0, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 0.696, |
| "grad_norm": 0.6367170247884075, |
| "learning_rate": 3.9817535498487385e-05, |
| "loss": 0.5304, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5349035263061523, |
| "step": 435, |
| "valid_targets_mean": 4406.4, |
| "valid_targets_min": 761 |
| }, |
| { |
| "epoch": 0.704, |
| "grad_norm": 0.6231922620233467, |
| "learning_rate": 3.980216964556892e-05, |
| "loss": 0.5176, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5187514424324036, |
| "step": 440, |
| "valid_targets_mean": 4343.2, |
| "valid_targets_min": 724 |
| }, |
| { |
| "epoch": 0.712, |
| "grad_norm": 0.7123358352627772, |
| "learning_rate": 3.978618588898873e-05, |
| "loss": 0.5295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5067673921585083, |
| "step": 445, |
| "valid_targets_mean": 3212.3, |
| "valid_targets_min": 872 |
| }, |
| { |
| "epoch": 0.72, |
| "grad_norm": 0.7454664370057654, |
| "learning_rate": 3.976958472750137e-05, |
| "loss": 0.537, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5686776638031006, |
| "step": 450, |
| "valid_targets_mean": 3169.4, |
| "valid_targets_min": 646 |
| }, |
| { |
| "epoch": 0.728, |
| "grad_norm": 0.5174561677012455, |
| "learning_rate": 3.9752366679126754e-05, |
| "loss": 0.528, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5228891968727112, |
| "step": 455, |
| "valid_targets_mean": 5854.7, |
| "valid_targets_min": 495 |
| }, |
| { |
| "epoch": 0.736, |
| "grad_norm": 0.7061480122954578, |
| "learning_rate": 3.973453228113405e-05, |
| "loss": 0.5422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5699916481971741, |
| "step": 460, |
| "valid_targets_mean": 3457.9, |
| "valid_targets_min": 631 |
| }, |
| { |
| "epoch": 0.744, |
| "grad_norm": 0.6372505538914572, |
| "learning_rate": 3.971608209002489e-05, |
| "loss": 0.5036, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.510094940662384, |
| "step": 465, |
| "valid_targets_mean": 3599.1, |
| "valid_targets_min": 827 |
| }, |
| { |
| "epoch": 0.752, |
| "grad_norm": 0.6192730292845192, |
| "learning_rate": 3.969701668151603e-05, |
| "loss": 0.4938, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5506471395492554, |
| "step": 470, |
| "valid_targets_mean": 4186.9, |
| "valid_targets_min": 605 |
| }, |
| { |
| "epoch": 0.76, |
| "grad_norm": 0.6458176448811681, |
| "learning_rate": 3.9677336650521336e-05, |
| "loss": 0.51, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4645317792892456, |
| "step": 475, |
| "valid_targets_mean": 5438.2, |
| "valid_targets_min": 691 |
| }, |
| { |
| "epoch": 0.768, |
| "grad_norm": 0.6877541407077931, |
| "learning_rate": 3.9657042611133294e-05, |
| "loss": 0.5309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5832065343856812, |
| "step": 480, |
| "valid_targets_mean": 4794.7, |
| "valid_targets_min": 1011 |
| }, |
| { |
| "epoch": 0.776, |
| "grad_norm": 0.7261934590292777, |
| "learning_rate": 3.963613519660379e-05, |
| "loss": 0.5469, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5814934968948364, |
| "step": 485, |
| "valid_targets_mean": 3265.4, |
| "valid_targets_min": 814 |
| }, |
| { |
| "epoch": 0.784, |
| "grad_norm": 0.5727199057578912, |
| "learning_rate": 3.961461505932435e-05, |
| "loss": 0.5226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5271404385566711, |
| "step": 490, |
| "valid_targets_mean": 5650.6, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 0.792, |
| "grad_norm": 0.6931137062168018, |
| "learning_rate": 3.959248287080583e-05, |
| "loss": 0.4874, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5182836055755615, |
| "step": 495, |
| "valid_targets_mean": 2894.6, |
| "valid_targets_min": 634 |
| }, |
| { |
| "epoch": 0.8, |
| "grad_norm": 0.5579037807480154, |
| "learning_rate": 3.9569739321657416e-05, |
| "loss": 0.5329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4890630543231964, |
| "step": 500, |
| "valid_targets_mean": 4849.6, |
| "valid_targets_min": 882 |
| }, |
| { |
| "epoch": 0.808, |
| "grad_norm": 0.5278555941049078, |
| "learning_rate": 3.9546385121565095e-05, |
| "loss": 0.5214, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4740561842918396, |
| "step": 505, |
| "valid_targets_mean": 5560.9, |
| "valid_targets_min": 386 |
| }, |
| { |
| "epoch": 0.816, |
| "grad_norm": 0.7629284256030876, |
| "learning_rate": 3.952242099926951e-05, |
| "loss": 0.5019, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5348608493804932, |
| "step": 510, |
| "valid_targets_mean": 3005.8, |
| "valid_targets_min": 485 |
| }, |
| { |
| "epoch": 0.824, |
| "grad_norm": 0.5984390769675876, |
| "learning_rate": 3.9497847702543196e-05, |
| "loss": 0.5011, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49162226915359497, |
| "step": 515, |
| "valid_targets_mean": 4003.1, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 0.832, |
| "grad_norm": 0.5425603828548777, |
| "learning_rate": 3.94726659981673e-05, |
| "loss": 0.5236, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4525638222694397, |
| "step": 520, |
| "valid_targets_mean": 4395.4, |
| "valid_targets_min": 617 |
| }, |
| { |
| "epoch": 0.84, |
| "grad_norm": 0.527115667278833, |
| "learning_rate": 3.94468766719076e-05, |
| "loss": 0.5203, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5383976101875305, |
| "step": 525, |
| "valid_targets_mean": 6022.6, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 0.848, |
| "grad_norm": 0.48267274772473334, |
| "learning_rate": 3.942048052849001e-05, |
| "loss": 0.4667, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40104740858078003, |
| "step": 530, |
| "valid_targets_mean": 5891.2, |
| "valid_targets_min": 675 |
| }, |
| { |
| "epoch": 0.856, |
| "grad_norm": 0.7791947549458919, |
| "learning_rate": 3.939347839157548e-05, |
| "loss": 0.5056, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5501073598861694, |
| "step": 535, |
| "valid_targets_mean": 2821.9, |
| "valid_targets_min": 780 |
| }, |
| { |
| "epoch": 0.864, |
| "grad_norm": 0.7336120776006877, |
| "learning_rate": 3.9365871103734264e-05, |
| "loss": 0.5003, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5346959829330444, |
| "step": 540, |
| "valid_targets_mean": 3434.2, |
| "valid_targets_min": 733 |
| }, |
| { |
| "epoch": 0.872, |
| "grad_norm": 0.5526825963889843, |
| "learning_rate": 3.933765952641965e-05, |
| "loss": 0.4971, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4672126770019531, |
| "step": 545, |
| "valid_targets_mean": 4462.4, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 0.88, |
| "grad_norm": 0.6189947196809612, |
| "learning_rate": 3.930884453994109e-05, |
| "loss": 0.4983, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4648451507091522, |
| "step": 550, |
| "valid_targets_mean": 3606.6, |
| "valid_targets_min": 624 |
| }, |
| { |
| "epoch": 0.888, |
| "grad_norm": 0.6586170474235856, |
| "learning_rate": 3.9279427043436706e-05, |
| "loss": 0.5388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5403224229812622, |
| "step": 555, |
| "valid_targets_mean": 3831.8, |
| "valid_targets_min": 415 |
| }, |
| { |
| "epoch": 0.896, |
| "grad_norm": 0.5313121728145141, |
| "learning_rate": 3.924940795484525e-05, |
| "loss": 0.5063, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.521693766117096, |
| "step": 560, |
| "valid_targets_mean": 5235.8, |
| "valid_targets_min": 828 |
| }, |
| { |
| "epoch": 0.904, |
| "grad_norm": 0.6424391584952766, |
| "learning_rate": 3.9218788210877436e-05, |
| "loss": 0.5359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5287529826164246, |
| "step": 565, |
| "valid_targets_mean": 3747.9, |
| "valid_targets_min": 597 |
| }, |
| { |
| "epoch": 0.912, |
| "grad_norm": 0.7979713052864025, |
| "learning_rate": 3.918756876698676e-05, |
| "loss": 0.517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5333600640296936, |
| "step": 570, |
| "valid_targets_mean": 4772.2, |
| "valid_targets_min": 426 |
| }, |
| { |
| "epoch": 0.92, |
| "grad_norm": 0.5943189435660875, |
| "learning_rate": 3.9155750597339634e-05, |
| "loss": 0.4952, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5387169718742371, |
| "step": 575, |
| "valid_targets_mean": 4364.7, |
| "valid_targets_min": 743 |
| }, |
| { |
| "epoch": 0.928, |
| "grad_norm": 0.6520342943163902, |
| "learning_rate": 3.912333469478502e-05, |
| "loss": 0.5197, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5179246068000793, |
| "step": 580, |
| "valid_targets_mean": 3263.0, |
| "valid_targets_min": 735 |
| }, |
| { |
| "epoch": 0.936, |
| "grad_norm": 0.6103078060063878, |
| "learning_rate": 3.909032207082344e-05, |
| "loss": 0.4996, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49960577487945557, |
| "step": 585, |
| "valid_targets_mean": 4096.7, |
| "valid_targets_min": 505 |
| }, |
| { |
| "epoch": 0.944, |
| "grad_norm": 0.585543703034473, |
| "learning_rate": 3.90567137555754e-05, |
| "loss": 0.5287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5424899458885193, |
| "step": 590, |
| "valid_targets_mean": 4775.8, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 0.952, |
| "grad_norm": 0.6304193255405047, |
| "learning_rate": 3.9022510797749286e-05, |
| "loss": 0.5087, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5315080285072327, |
| "step": 595, |
| "valid_targets_mean": 4235.9, |
| "valid_targets_min": 907 |
| }, |
| { |
| "epoch": 0.96, |
| "grad_norm": 0.585015028731908, |
| "learning_rate": 3.898771426460859e-05, |
| "loss": 0.4945, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4809357523918152, |
| "step": 600, |
| "valid_targets_mean": 3830.8, |
| "valid_targets_min": 628 |
| }, |
| { |
| "epoch": 0.968, |
| "grad_norm": 0.68245860217036, |
| "learning_rate": 3.8952325241938635e-05, |
| "loss": 0.5126, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5318968296051025, |
| "step": 605, |
| "valid_targets_mean": 4136.1, |
| "valid_targets_min": 635 |
| }, |
| { |
| "epoch": 0.976, |
| "grad_norm": 0.6148876956191819, |
| "learning_rate": 3.8916344834012695e-05, |
| "loss": 0.5311, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5257702469825745, |
| "step": 610, |
| "valid_targets_mean": 4727.1, |
| "valid_targets_min": 644 |
| }, |
| { |
| "epoch": 0.984, |
| "grad_norm": 0.6143278822542023, |
| "learning_rate": 3.887977416355754e-05, |
| "loss": 0.5224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4943949580192566, |
| "step": 615, |
| "valid_targets_mean": 3667.9, |
| "valid_targets_min": 775 |
| }, |
| { |
| "epoch": 0.992, |
| "grad_norm": 0.639756859129058, |
| "learning_rate": 3.884261437171838e-05, |
| "loss": 0.5033, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5292478799819946, |
| "step": 620, |
| "valid_targets_mean": 3723.4, |
| "valid_targets_min": 588 |
| }, |
| { |
| "epoch": 1.0, |
| "grad_norm": 0.6034626688615775, |
| "learning_rate": 3.8804866618023284e-05, |
| "loss": 0.4778, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5043004751205444, |
| "step": 625, |
| "valid_targets_mean": 4286.8, |
| "valid_targets_min": 593 |
| }, |
| { |
| "epoch": 1.008, |
| "grad_norm": 0.5749723492360072, |
| "learning_rate": 3.876653208034698e-05, |
| "loss": 0.5212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4886097311973572, |
| "step": 630, |
| "valid_targets_mean": 4484.8, |
| "valid_targets_min": 710 |
| }, |
| { |
| "epoch": 1.016, |
| "grad_norm": 0.5975830354439139, |
| "learning_rate": 3.8727611954874114e-05, |
| "loss": 0.4941, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4872697591781616, |
| "step": 635, |
| "valid_targets_mean": 4030.0, |
| "valid_targets_min": 701 |
| }, |
| { |
| "epoch": 1.024, |
| "grad_norm": 0.6626052135788788, |
| "learning_rate": 3.8688107456061904e-05, |
| "loss": 0.4666, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48330509662628174, |
| "step": 640, |
| "valid_targets_mean": 3877.7, |
| "valid_targets_min": 751 |
| }, |
| { |
| "epoch": 1.032, |
| "grad_norm": 0.5825477612458702, |
| "learning_rate": 3.864801981660227e-05, |
| "loss": 0.4876, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.50968998670578, |
| "step": 645, |
| "valid_targets_mean": 4582.6, |
| "valid_targets_min": 678 |
| }, |
| { |
| "epoch": 1.04, |
| "grad_norm": 0.5888038935122548, |
| "learning_rate": 3.860735028738337e-05, |
| "loss": 0.4909, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4983733296394348, |
| "step": 650, |
| "valid_targets_mean": 5661.3, |
| "valid_targets_min": 946 |
| }, |
| { |
| "epoch": 1.048, |
| "grad_norm": 0.7028901777525807, |
| "learning_rate": 3.856610013745051e-05, |
| "loss": 0.4675, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43987807631492615, |
| "step": 655, |
| "valid_targets_mean": 2972.2, |
| "valid_targets_min": 540 |
| }, |
| { |
| "epoch": 1.056, |
| "grad_norm": 0.6837565819150657, |
| "learning_rate": 3.852427065396665e-05, |
| "loss": 0.5386, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.54156893491745, |
| "step": 660, |
| "valid_targets_mean": 3283.2, |
| "valid_targets_min": 984 |
| }, |
| { |
| "epoch": 1.064, |
| "grad_norm": 0.7103924713439609, |
| "learning_rate": 3.848186314217213e-05, |
| "loss": 0.4915, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5271738767623901, |
| "step": 665, |
| "valid_targets_mean": 3783.6, |
| "valid_targets_min": 591 |
| }, |
| { |
| "epoch": 1.072, |
| "grad_norm": 0.5777473820322118, |
| "learning_rate": 3.843887892534402e-05, |
| "loss": 0.4949, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4844241142272949, |
| "step": 670, |
| "valid_targets_mean": 5510.7, |
| "valid_targets_min": 801 |
| }, |
| { |
| "epoch": 1.08, |
| "grad_norm": 0.5592068039838433, |
| "learning_rate": 3.8395319344754776e-05, |
| "loss": 0.4824, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4355768859386444, |
| "step": 675, |
| "valid_targets_mean": 4872.8, |
| "valid_targets_min": 775 |
| }, |
| { |
| "epoch": 1.088, |
| "grad_norm": 0.5196780282819776, |
| "learning_rate": 3.8351185759630435e-05, |
| "loss": 0.5028, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4816337525844574, |
| "step": 680, |
| "valid_targets_mean": 6182.8, |
| "valid_targets_min": 689 |
| }, |
| { |
| "epoch": 1.096, |
| "grad_norm": 0.4974070083115012, |
| "learning_rate": 3.830647954710816e-05, |
| "loss": 0.4799, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4324873685836792, |
| "step": 685, |
| "valid_targets_mean": 5340.4, |
| "valid_targets_min": 622 |
| }, |
| { |
| "epoch": 1.104, |
| "grad_norm": 0.5227557174907695, |
| "learning_rate": 3.826120210219331e-05, |
| "loss": 0.4665, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4405563771724701, |
| "step": 690, |
| "valid_targets_mean": 5245.5, |
| "valid_targets_min": 822 |
| }, |
| { |
| "epoch": 1.112, |
| "grad_norm": 0.5415434424451512, |
| "learning_rate": 3.8215354837715836e-05, |
| "loss": 0.4753, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.443329781293869, |
| "step": 695, |
| "valid_targets_mean": 6236.1, |
| "valid_targets_min": 756 |
| }, |
| { |
| "epoch": 1.12, |
| "grad_norm": 0.6938512610507073, |
| "learning_rate": 3.816893918428631e-05, |
| "loss": 0.5224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5610541701316833, |
| "step": 700, |
| "valid_targets_mean": 3587.6, |
| "valid_targets_min": 975 |
| }, |
| { |
| "epoch": 1.1280000000000001, |
| "grad_norm": 0.7257198661742723, |
| "learning_rate": 3.8121956590251153e-05, |
| "loss": 0.4849, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.511725902557373, |
| "step": 705, |
| "valid_targets_mean": 2714.5, |
| "valid_targets_min": 704 |
| }, |
| { |
| "epoch": 1.1360000000000001, |
| "grad_norm": 0.7628570201199647, |
| "learning_rate": 3.8074408521647576e-05, |
| "loss": 0.4942, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5041239857673645, |
| "step": 710, |
| "valid_targets_mean": 4192.9, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 1.144, |
| "grad_norm": 0.7232575314594617, |
| "learning_rate": 3.802629646215771e-05, |
| "loss": 0.5357, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5536401271820068, |
| "step": 715, |
| "valid_targets_mean": 3258.6, |
| "valid_targets_min": 492 |
| }, |
| { |
| "epoch": 1.152, |
| "grad_norm": 0.5322461206523738, |
| "learning_rate": 3.79776219130624e-05, |
| "loss": 0.5255, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4725511074066162, |
| "step": 720, |
| "valid_targets_mean": 5086.8, |
| "valid_targets_min": 754 |
| }, |
| { |
| "epoch": 1.16, |
| "grad_norm": 0.5597590750275225, |
| "learning_rate": 3.792838639319431e-05, |
| "loss": 0.4851, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4070642590522766, |
| "step": 725, |
| "valid_targets_mean": 3765.9, |
| "valid_targets_min": 598 |
| }, |
| { |
| "epoch": 1.168, |
| "grad_norm": 0.5995633156774554, |
| "learning_rate": 3.787859143889054e-05, |
| "loss": 0.4643, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5027796030044556, |
| "step": 730, |
| "valid_targets_mean": 4211.0, |
| "valid_targets_min": 689 |
| }, |
| { |
| "epoch": 1.176, |
| "grad_norm": 0.7391042973959496, |
| "learning_rate": 3.782823860394469e-05, |
| "loss": 0.535, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.478831022977829, |
| "step": 735, |
| "valid_targets_mean": 3402.7, |
| "valid_targets_min": 428 |
| }, |
| { |
| "epoch": 1.184, |
| "grad_norm": 0.655094019532376, |
| "learning_rate": 3.777732945955841e-05, |
| "loss": 0.4662, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49596506357192993, |
| "step": 740, |
| "valid_targets_mean": 4551.3, |
| "valid_targets_min": 786 |
| }, |
| { |
| "epoch": 1.192, |
| "grad_norm": 0.6319621366301474, |
| "learning_rate": 3.772586559429229e-05, |
| "loss": 0.4796, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5001382231712341, |
| "step": 745, |
| "valid_targets_mean": 3626.9, |
| "valid_targets_min": 925 |
| }, |
| { |
| "epoch": 1.2, |
| "grad_norm": 0.5834982530634643, |
| "learning_rate": 3.767384861401636e-05, |
| "loss": 0.4846, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4496903419494629, |
| "step": 750, |
| "valid_targets_mean": 3643.1, |
| "valid_targets_min": 861 |
| }, |
| { |
| "epoch": 1.208, |
| "grad_norm": 0.7608316000603859, |
| "learning_rate": 3.762128014185998e-05, |
| "loss": 0.5195, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5569829344749451, |
| "step": 755, |
| "valid_targets_mean": 2473.0, |
| "valid_targets_min": 473 |
| }, |
| { |
| "epoch": 1.216, |
| "grad_norm": 0.5591412298342779, |
| "learning_rate": 3.7568161818161135e-05, |
| "loss": 0.4618, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49754035472869873, |
| "step": 760, |
| "valid_targets_mean": 6402.6, |
| "valid_targets_min": 1106 |
| }, |
| { |
| "epoch": 1.224, |
| "grad_norm": 0.8123260766469027, |
| "learning_rate": 3.751449530041532e-05, |
| "loss": 0.4975, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5815908908843994, |
| "step": 765, |
| "valid_targets_mean": 2641.3, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 1.232, |
| "grad_norm": 0.6101133475673763, |
| "learning_rate": 3.7460282263223764e-05, |
| "loss": 0.4819, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4854985475540161, |
| "step": 770, |
| "valid_targets_mean": 4249.1, |
| "valid_targets_min": 590 |
| }, |
| { |
| "epoch": 1.24, |
| "grad_norm": 0.5375018128815826, |
| "learning_rate": 3.740552439824122e-05, |
| "loss": 0.486, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4772304594516754, |
| "step": 775, |
| "valid_targets_mean": 4520.9, |
| "valid_targets_min": 801 |
| }, |
| { |
| "epoch": 1.248, |
| "grad_norm": 0.51139105848858, |
| "learning_rate": 3.735022341412314e-05, |
| "loss": 0.477, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47216999530792236, |
| "step": 780, |
| "valid_targets_mean": 6126.1, |
| "valid_targets_min": 738 |
| }, |
| { |
| "epoch": 1.256, |
| "grad_norm": 0.5195276484071466, |
| "learning_rate": 3.7294381036472386e-05, |
| "loss": 0.4797, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46269387006759644, |
| "step": 785, |
| "valid_targets_mean": 4958.2, |
| "valid_targets_min": 667 |
| }, |
| { |
| "epoch": 1.264, |
| "grad_norm": 0.6267531451482399, |
| "learning_rate": 3.723799900778538e-05, |
| "loss": 0.4883, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4804888665676117, |
| "step": 790, |
| "valid_targets_mean": 3517.1, |
| "valid_targets_min": 578 |
| }, |
| { |
| "epoch": 1.272, |
| "grad_norm": 0.5195057436049215, |
| "learning_rate": 3.7181079087397705e-05, |
| "loss": 0.5087, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5404967665672302, |
| "step": 795, |
| "valid_targets_mean": 6603.4, |
| "valid_targets_min": 989 |
| }, |
| { |
| "epoch": 1.28, |
| "grad_norm": 0.5398694980743434, |
| "learning_rate": 3.712362305142926e-05, |
| "loss": 0.4393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4667036831378937, |
| "step": 800, |
| "valid_targets_mean": 4789.3, |
| "valid_targets_min": 512 |
| }, |
| { |
| "epoch": 1.288, |
| "grad_norm": 0.700472880561129, |
| "learning_rate": 3.706563269272878e-05, |
| "loss": 0.4852, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45391908288002014, |
| "step": 805, |
| "valid_targets_mean": 2936.9, |
| "valid_targets_min": 646 |
| }, |
| { |
| "epoch": 1.296, |
| "grad_norm": 0.6212205565818121, |
| "learning_rate": 3.700710982081794e-05, |
| "loss": 0.4758, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4718417823314667, |
| "step": 810, |
| "valid_targets_mean": 3432.1, |
| "valid_targets_min": 782 |
| }, |
| { |
| "epoch": 1.304, |
| "grad_norm": 0.6408053164272295, |
| "learning_rate": 3.694805626183486e-05, |
| "loss": 0.4872, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5090664625167847, |
| "step": 815, |
| "valid_targets_mean": 4112.0, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 1.312, |
| "grad_norm": 0.7001179282352733, |
| "learning_rate": 3.688847385847711e-05, |
| "loss": 0.4719, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.427803099155426, |
| "step": 820, |
| "valid_targets_mean": 3749.1, |
| "valid_targets_min": 1009 |
| }, |
| { |
| "epoch": 1.32, |
| "grad_norm": 0.6511396074795565, |
| "learning_rate": 3.682836446994428e-05, |
| "loss": 0.5126, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4967302083969116, |
| "step": 825, |
| "valid_targets_mean": 3997.8, |
| "valid_targets_min": 718 |
| }, |
| { |
| "epoch": 1.328, |
| "grad_norm": 0.593371544756325, |
| "learning_rate": 3.676772997187989e-05, |
| "loss": 0.4859, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.475099116563797, |
| "step": 830, |
| "valid_targets_mean": 4375.9, |
| "valid_targets_min": 719 |
| }, |
| { |
| "epoch": 1.336, |
| "grad_norm": 0.6550709845667457, |
| "learning_rate": 3.670657225631289e-05, |
| "loss": 0.4717, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5210568904876709, |
| "step": 835, |
| "valid_targets_mean": 3314.4, |
| "valid_targets_min": 662 |
| }, |
| { |
| "epoch": 1.3439999999999999, |
| "grad_norm": 0.6168115512142734, |
| "learning_rate": 3.6644893231598635e-05, |
| "loss": 0.4695, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5161671042442322, |
| "step": 840, |
| "valid_targets_mean": 3743.0, |
| "valid_targets_min": 552 |
| }, |
| { |
| "epoch": 1.3519999999999999, |
| "grad_norm": 1.2636079758657017, |
| "learning_rate": 3.658269482235932e-05, |
| "loss": 0.4564, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.433046817779541, |
| "step": 845, |
| "valid_targets_mean": 2702.0, |
| "valid_targets_min": 640 |
| }, |
| { |
| "epoch": 1.3599999999999999, |
| "grad_norm": 0.644979165166496, |
| "learning_rate": 3.651997896942394e-05, |
| "loss": 0.525, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5414267778396606, |
| "step": 850, |
| "valid_targets_mean": 4083.8, |
| "valid_targets_min": 815 |
| }, |
| { |
| "epoch": 1.3679999999999999, |
| "grad_norm": 0.5287402349737144, |
| "learning_rate": 3.645674762976769e-05, |
| "loss": 0.4819, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43094199895858765, |
| "step": 855, |
| "valid_targets_mean": 4626.4, |
| "valid_targets_min": 583 |
| }, |
| { |
| "epoch": 1.376, |
| "grad_norm": 0.609279043498901, |
| "learning_rate": 3.639300277645096e-05, |
| "loss": 0.475, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49126923084259033, |
| "step": 860, |
| "valid_targets_mean": 4487.2, |
| "valid_targets_min": 900 |
| }, |
| { |
| "epoch": 1.384, |
| "grad_norm": 0.645279015628696, |
| "learning_rate": 3.6328746398557715e-05, |
| "loss": 0.4776, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.500410795211792, |
| "step": 865, |
| "valid_targets_mean": 3312.8, |
| "valid_targets_min": 447 |
| }, |
| { |
| "epoch": 1.392, |
| "grad_norm": 0.6099569963138785, |
| "learning_rate": 3.6263980501133466e-05, |
| "loss": 0.4815, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4480360150337219, |
| "step": 870, |
| "valid_targets_mean": 4288.2, |
| "valid_targets_min": 658 |
| }, |
| { |
| "epoch": 1.4, |
| "grad_norm": 0.5594438345209423, |
| "learning_rate": 3.619870710512268e-05, |
| "loss": 0.4831, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46603310108184814, |
| "step": 875, |
| "valid_targets_mean": 4127.5, |
| "valid_targets_min": 810 |
| }, |
| { |
| "epoch": 1.408, |
| "grad_norm": 0.7040863724335125, |
| "learning_rate": 3.6132928247305713e-05, |
| "loss": 0.4717, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4972262978553772, |
| "step": 880, |
| "valid_targets_mean": 4002.6, |
| "valid_targets_min": 811 |
| }, |
| { |
| "epoch": 1.416, |
| "grad_norm": 0.4723530474869513, |
| "learning_rate": 3.60666459802353e-05, |
| "loss": 0.4683, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45891261100769043, |
| "step": 885, |
| "valid_targets_mean": 6404.6, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 1.424, |
| "grad_norm": 0.6504389615505566, |
| "learning_rate": 3.599986237217245e-05, |
| "loss": 0.4589, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49261537194252014, |
| "step": 890, |
| "valid_targets_mean": 3836.1, |
| "valid_targets_min": 597 |
| }, |
| { |
| "epoch": 1.432, |
| "grad_norm": 0.5829046317294024, |
| "learning_rate": 3.593257950702194e-05, |
| "loss": 0.4371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41449180245399475, |
| "step": 895, |
| "valid_targets_mean": 4351.4, |
| "valid_targets_min": 373 |
| }, |
| { |
| "epoch": 1.44, |
| "grad_norm": 0.6928805501763816, |
| "learning_rate": 3.586479948426728e-05, |
| "loss": 0.483, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44609037041664124, |
| "step": 900, |
| "valid_targets_mean": 4319.5, |
| "valid_targets_min": 763 |
| }, |
| { |
| "epoch": 1.448, |
| "grad_norm": 0.6643739326321101, |
| "learning_rate": 3.579652441890523e-05, |
| "loss": 0.4797, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49757933616638184, |
| "step": 905, |
| "valid_targets_mean": 3803.0, |
| "valid_targets_min": 657 |
| }, |
| { |
| "epoch": 1.456, |
| "grad_norm": 0.6143346831296209, |
| "learning_rate": 3.572775644137974e-05, |
| "loss": 0.4814, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4909932315349579, |
| "step": 910, |
| "valid_targets_mean": 3776.1, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 1.464, |
| "grad_norm": 0.529180684369256, |
| "learning_rate": 3.5658497697515534e-05, |
| "loss": 0.4566, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5000677704811096, |
| "step": 915, |
| "valid_targets_mean": 6069.6, |
| "valid_targets_min": 812 |
| }, |
| { |
| "epoch": 1.472, |
| "grad_norm": 0.6497553129314787, |
| "learning_rate": 3.558875034845113e-05, |
| "loss": 0.4521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4506397545337677, |
| "step": 920, |
| "valid_targets_mean": 3804.3, |
| "valid_targets_min": 560 |
| }, |
| { |
| "epoch": 1.48, |
| "grad_norm": 0.4920648541618837, |
| "learning_rate": 3.551851657057139e-05, |
| "loss": 0.4807, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4525104761123657, |
| "step": 925, |
| "valid_targets_mean": 7107.8, |
| "valid_targets_min": 635 |
| }, |
| { |
| "epoch": 1.488, |
| "grad_norm": 0.7772212863915389, |
| "learning_rate": 3.544779855543963e-05, |
| "loss": 0.4908, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4923700988292694, |
| "step": 930, |
| "valid_targets_mean": 2380.3, |
| "valid_targets_min": 624 |
| }, |
| { |
| "epoch": 1.496, |
| "grad_norm": 0.5981404469738086, |
| "learning_rate": 3.5376598509729226e-05, |
| "loss": 0.4821, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4714764952659607, |
| "step": 935, |
| "valid_targets_mean": 4965.3, |
| "valid_targets_min": 563 |
| }, |
| { |
| "epoch": 1.504, |
| "grad_norm": 0.7007792353553981, |
| "learning_rate": 3.5304918655154754e-05, |
| "loss": 0.4921, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4638800024986267, |
| "step": 940, |
| "valid_targets_mean": 3031.9, |
| "valid_targets_min": 817 |
| }, |
| { |
| "epoch": 1.512, |
| "grad_norm": 0.6097210410936542, |
| "learning_rate": 3.523276122840266e-05, |
| "loss": 0.4893, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.483889102935791, |
| "step": 945, |
| "valid_targets_mean": 3922.5, |
| "valid_targets_min": 767 |
| }, |
| { |
| "epoch": 1.52, |
| "grad_norm": 0.7526194303601134, |
| "learning_rate": 3.516012848106149e-05, |
| "loss": 0.4721, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5017504692077637, |
| "step": 950, |
| "valid_targets_mean": 2572.9, |
| "valid_targets_min": 755 |
| }, |
| { |
| "epoch": 1.528, |
| "grad_norm": 0.46052003404390657, |
| "learning_rate": 3.5087022679551614e-05, |
| "loss": 0.4661, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41806378960609436, |
| "step": 955, |
| "valid_targets_mean": 5992.2, |
| "valid_targets_min": 751 |
| }, |
| { |
| "epoch": 1.536, |
| "grad_norm": 0.6989167426008138, |
| "learning_rate": 3.5013446105054486e-05, |
| "loss": 0.4931, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5233873128890991, |
| "step": 960, |
| "valid_targets_mean": 2915.4, |
| "valid_targets_min": 935 |
| }, |
| { |
| "epoch": 1.544, |
| "grad_norm": 0.7730262333694343, |
| "learning_rate": 3.493940105344152e-05, |
| "loss": 0.4765, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45950740575790405, |
| "step": 965, |
| "valid_targets_mean": 3003.8, |
| "valid_targets_min": 601 |
| }, |
| { |
| "epoch": 1.552, |
| "grad_norm": 0.6796211235694684, |
| "learning_rate": 3.4864889835202366e-05, |
| "loss": 0.4963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4867255389690399, |
| "step": 970, |
| "valid_targets_mean": 5517.5, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 1.56, |
| "grad_norm": 0.882468209296317, |
| "learning_rate": 3.4789914775372905e-05, |
| "loss": 0.4956, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5269094705581665, |
| "step": 975, |
| "valid_targets_mean": 2169.9, |
| "valid_targets_min": 816 |
| }, |
| { |
| "epoch": 1.568, |
| "grad_norm": 0.6427886583581234, |
| "learning_rate": 3.471447821346264e-05, |
| "loss": 0.4946, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5064438581466675, |
| "step": 980, |
| "valid_targets_mean": 3668.9, |
| "valid_targets_min": 802 |
| }, |
| { |
| "epoch": 1.576, |
| "grad_norm": 0.4751807016740937, |
| "learning_rate": 3.463858250338168e-05, |
| "loss": 0.478, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4542385935783386, |
| "step": 985, |
| "valid_targets_mean": 7338.5, |
| "valid_targets_min": 371 |
| }, |
| { |
| "epoch": 1.584, |
| "grad_norm": 0.5520255318021472, |
| "learning_rate": 3.4562230013367374e-05, |
| "loss": 0.467, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47316858172416687, |
| "step": 990, |
| "valid_targets_mean": 4990.0, |
| "valid_targets_min": 786 |
| }, |
| { |
| "epoch": 1.592, |
| "grad_norm": 0.5538940893817148, |
| "learning_rate": 3.448542312591032e-05, |
| "loss": 0.4678, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43099331855773926, |
| "step": 995, |
| "valid_targets_mean": 4630.3, |
| "valid_targets_min": 766 |
| }, |
| { |
| "epoch": 1.6, |
| "grad_norm": 0.49936258088092805, |
| "learning_rate": 3.440816423768007e-05, |
| "loss": 0.4788, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4556550085544586, |
| "step": 1000, |
| "valid_targets_mean": 5382.2, |
| "valid_targets_min": 557 |
| }, |
| { |
| "epoch": 1.608, |
| "grad_norm": 0.6836362049629878, |
| "learning_rate": 3.433045575945031e-05, |
| "loss": 0.4665, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5310341119766235, |
| "step": 1005, |
| "valid_targets_mean": 2997.4, |
| "valid_targets_min": 705 |
| }, |
| { |
| "epoch": 1.616, |
| "grad_norm": 0.5410029620184749, |
| "learning_rate": 3.42523001160237e-05, |
| "loss": 0.4847, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47429001331329346, |
| "step": 1010, |
| "valid_targets_mean": 4692.7, |
| "valid_targets_min": 418 |
| }, |
| { |
| "epoch": 1.624, |
| "grad_norm": 0.536446225642749, |
| "learning_rate": 3.417369974615615e-05, |
| "loss": 0.4739, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44192248582839966, |
| "step": 1015, |
| "valid_targets_mean": 4533.6, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 1.6320000000000001, |
| "grad_norm": 0.654586453788874, |
| "learning_rate": 3.409465710248074e-05, |
| "loss": 0.4511, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4620225727558136, |
| "step": 1020, |
| "valid_targets_mean": 3155.8, |
| "valid_targets_min": 497 |
| }, |
| { |
| "epoch": 1.6400000000000001, |
| "grad_norm": 0.6885658704413448, |
| "learning_rate": 3.401517465143119e-05, |
| "loss": 0.4594, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47382691502571106, |
| "step": 1025, |
| "valid_targets_mean": 2903.8, |
| "valid_targets_min": 511 |
| }, |
| { |
| "epoch": 1.6480000000000001, |
| "grad_norm": 0.5838313777450732, |
| "learning_rate": 3.393525487316489e-05, |
| "loss": 0.4802, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4386270344257355, |
| "step": 1030, |
| "valid_targets_mean": 3802.1, |
| "valid_targets_min": 917 |
| }, |
| { |
| "epoch": 1.6560000000000001, |
| "grad_norm": 0.5475012361833616, |
| "learning_rate": 3.385490026148554e-05, |
| "loss": 0.4882, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47070324420928955, |
| "step": 1035, |
| "valid_targets_mean": 4426.2, |
| "valid_targets_min": 556 |
| }, |
| { |
| "epoch": 1.6640000000000001, |
| "grad_norm": 0.5825446614405908, |
| "learning_rate": 3.377411332376529e-05, |
| "loss": 0.4871, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49477773904800415, |
| "step": 1040, |
| "valid_targets_mean": 3675.8, |
| "valid_targets_min": 693 |
| }, |
| { |
| "epoch": 1.6720000000000002, |
| "grad_norm": 0.5522875927498029, |
| "learning_rate": 3.369289658086651e-05, |
| "loss": 0.4774, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4486830234527588, |
| "step": 1045, |
| "valid_targets_mean": 4307.8, |
| "valid_targets_min": 513 |
| }, |
| { |
| "epoch": 1.6800000000000002, |
| "grad_norm": 0.5439529345170723, |
| "learning_rate": 3.3611252567063184e-05, |
| "loss": 0.4639, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4867696166038513, |
| "step": 1050, |
| "valid_targets_mean": 4714.7, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 1.688, |
| "grad_norm": 0.753150960893063, |
| "learning_rate": 3.352918382996174e-05, |
| "loss": 0.4951, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5364462733268738, |
| "step": 1055, |
| "valid_targets_mean": 3034.1, |
| "valid_targets_min": 622 |
| }, |
| { |
| "epoch": 1.696, |
| "grad_norm": 0.5840969444722356, |
| "learning_rate": 3.344669293042163e-05, |
| "loss": 0.4687, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5160173177719116, |
| "step": 1060, |
| "valid_targets_mean": 4426.8, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 1.704, |
| "grad_norm": 0.6291215602430366, |
| "learning_rate": 3.336378244247539e-05, |
| "loss": 0.4774, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48828327655792236, |
| "step": 1065, |
| "valid_targets_mean": 3159.0, |
| "valid_targets_min": 579 |
| }, |
| { |
| "epoch": 1.712, |
| "grad_norm": 0.7157439156221113, |
| "learning_rate": 3.3280454953248326e-05, |
| "loss": 0.4719, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46565982699394226, |
| "step": 1070, |
| "valid_targets_mean": 2837.4, |
| "valid_targets_min": 479 |
| }, |
| { |
| "epoch": 1.72, |
| "grad_norm": 0.8834848909249748, |
| "learning_rate": 3.3196713062877765e-05, |
| "loss": 0.4555, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47797664999961853, |
| "step": 1075, |
| "valid_targets_mean": 2625.6, |
| "valid_targets_min": 588 |
| }, |
| { |
| "epoch": 1.728, |
| "grad_norm": 0.47069533464521623, |
| "learning_rate": 3.311255938443196e-05, |
| "loss": 0.467, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4377485513687134, |
| "step": 1080, |
| "valid_targets_mean": 5656.1, |
| "valid_targets_min": 735 |
| }, |
| { |
| "epoch": 1.736, |
| "grad_norm": 0.5994529019656026, |
| "learning_rate": 3.3027996543828524e-05, |
| "loss": 0.4708, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5046743154525757, |
| "step": 1085, |
| "valid_targets_mean": 3614.6, |
| "valid_targets_min": 619 |
| }, |
| { |
| "epoch": 1.744, |
| "grad_norm": 0.6262011025182842, |
| "learning_rate": 3.2943027179752494e-05, |
| "loss": 0.4717, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4625643789768219, |
| "step": 1090, |
| "valid_targets_mean": 4390.8, |
| "valid_targets_min": 701 |
| }, |
| { |
| "epoch": 1.752, |
| "grad_norm": 0.505027541492022, |
| "learning_rate": 3.285765394357401e-05, |
| "loss": 0.4815, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46330204606056213, |
| "step": 1095, |
| "valid_targets_mean": 5306.1, |
| "valid_targets_min": 661 |
| }, |
| { |
| "epoch": 1.76, |
| "grad_norm": 0.514752245979312, |
| "learning_rate": 3.277187949926556e-05, |
| "loss": 0.4531, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41437840461730957, |
| "step": 1100, |
| "valid_targets_mean": 6349.6, |
| "valid_targets_min": 759 |
| }, |
| { |
| "epoch": 1.768, |
| "grad_norm": 0.5757785700580377, |
| "learning_rate": 3.268570652331888e-05, |
| "loss": 0.4744, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43744391202926636, |
| "step": 1105, |
| "valid_targets_mean": 3915.2, |
| "valid_targets_min": 686 |
| }, |
| { |
| "epoch": 1.776, |
| "grad_norm": 0.5909523847823849, |
| "learning_rate": 3.2599137704661405e-05, |
| "loss": 0.4395, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46480992436408997, |
| "step": 1110, |
| "valid_targets_mean": 3621.0, |
| "valid_targets_min": 468 |
| }, |
| { |
| "epoch": 1.784, |
| "grad_norm": 0.6522784519010079, |
| "learning_rate": 3.251217574457239e-05, |
| "loss": 0.4716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4484942555427551, |
| "step": 1115, |
| "valid_targets_mean": 3802.8, |
| "valid_targets_min": 686 |
| }, |
| { |
| "epoch": 1.792, |
| "grad_norm": 0.6076639765753343, |
| "learning_rate": 3.242482335659861e-05, |
| "loss": 0.4756, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41383862495422363, |
| "step": 1120, |
| "valid_targets_mean": 3632.5, |
| "valid_targets_min": 737 |
| }, |
| { |
| "epoch": 1.8, |
| "grad_norm": 1.4930741994185592, |
| "learning_rate": 3.2337083266469687e-05, |
| "loss": 0.4672, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49689725041389465, |
| "step": 1125, |
| "valid_targets_mean": 4094.2, |
| "valid_targets_min": 918 |
| }, |
| { |
| "epoch": 1.808, |
| "grad_norm": 0.5983765842540781, |
| "learning_rate": 3.224895821201304e-05, |
| "loss": 0.4772, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43095701932907104, |
| "step": 1130, |
| "valid_targets_mean": 3158.9, |
| "valid_targets_min": 505 |
| }, |
| { |
| "epoch": 1.8159999999999998, |
| "grad_norm": 0.7377527167711196, |
| "learning_rate": 3.2160450943068446e-05, |
| "loss": 0.494, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4552053213119507, |
| "step": 1135, |
| "valid_targets_mean": 3169.7, |
| "valid_targets_min": 615 |
| }, |
| { |
| "epoch": 1.8239999999999998, |
| "grad_norm": 0.48209247220176665, |
| "learning_rate": 3.207156422140225e-05, |
| "loss": 0.4665, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40306127071380615, |
| "step": 1140, |
| "valid_targets_mean": 5754.3, |
| "valid_targets_min": 633 |
| }, |
| { |
| "epoch": 1.8319999999999999, |
| "grad_norm": 0.6434012245459063, |
| "learning_rate": 3.198230082062115e-05, |
| "loss": 0.4792, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4668983817100525, |
| "step": 1145, |
| "valid_targets_mean": 3197.2, |
| "valid_targets_min": 538 |
| }, |
| { |
| "epoch": 1.8399999999999999, |
| "grad_norm": 0.6995497102738215, |
| "learning_rate": 3.189266352608574e-05, |
| "loss": 0.4839, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5054548978805542, |
| "step": 1150, |
| "valid_targets_mean": 2973.6, |
| "valid_targets_min": 614 |
| }, |
| { |
| "epoch": 1.8479999999999999, |
| "grad_norm": 0.569099812286609, |
| "learning_rate": 3.180265513482345e-05, |
| "loss": 0.4843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5099635720252991, |
| "step": 1155, |
| "valid_targets_mean": 4445.8, |
| "valid_targets_min": 1057 |
| }, |
| { |
| "epoch": 1.8559999999999999, |
| "grad_norm": 0.7118854281438147, |
| "learning_rate": 3.171227845544143e-05, |
| "loss": 0.5015, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5411723852157593, |
| "step": 1160, |
| "valid_targets_mean": 3062.9, |
| "valid_targets_min": 1004 |
| }, |
| { |
| "epoch": 1.8639999999999999, |
| "grad_norm": 0.4667067376045219, |
| "learning_rate": 3.162153630803877e-05, |
| "loss": 0.4776, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42969876527786255, |
| "step": 1165, |
| "valid_targets_mean": 6657.2, |
| "valid_targets_min": 701 |
| }, |
| { |
| "epoch": 1.8719999999999999, |
| "grad_norm": 0.5574951952526179, |
| "learning_rate": 3.153043152411861e-05, |
| "loss": 0.4466, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3883597254753113, |
| "step": 1170, |
| "valid_targets_mean": 3929.1, |
| "valid_targets_min": 714 |
| }, |
| { |
| "epoch": 1.88, |
| "grad_norm": 0.49857166985641654, |
| "learning_rate": 3.14389669464997e-05, |
| "loss": 0.4621, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4549972712993622, |
| "step": 1175, |
| "valid_targets_mean": 5751.6, |
| "valid_targets_min": 602 |
| }, |
| { |
| "epoch": 1.888, |
| "grad_norm": 0.7211688221308987, |
| "learning_rate": 3.134714542922777e-05, |
| "loss": 0.4731, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4644148647785187, |
| "step": 1180, |
| "valid_targets_mean": 4079.8, |
| "valid_targets_min": 539 |
| }, |
| { |
| "epoch": 1.896, |
| "grad_norm": 0.5649019306883653, |
| "learning_rate": 3.1254969837486425e-05, |
| "loss": 0.4801, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47578728199005127, |
| "step": 1185, |
| "valid_targets_mean": 4451.9, |
| "valid_targets_min": 496 |
| }, |
| { |
| "epoch": 1.904, |
| "grad_norm": 0.8113490782507405, |
| "learning_rate": 3.116244304750774e-05, |
| "loss": 0.4841, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49812304973602295, |
| "step": 1190, |
| "valid_targets_mean": 4256.6, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 1.912, |
| "grad_norm": 0.6475596871318647, |
| "learning_rate": 3.106956794648254e-05, |
| "loss": 0.4662, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4609223008155823, |
| "step": 1195, |
| "valid_targets_mean": 3238.9, |
| "valid_targets_min": 746 |
| }, |
| { |
| "epoch": 1.92, |
| "grad_norm": 0.5392636163279259, |
| "learning_rate": 3.097634743247026e-05, |
| "loss": 0.4366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4826831519603729, |
| "step": 1200, |
| "valid_targets_mean": 4662.3, |
| "valid_targets_min": 553 |
| }, |
| { |
| "epoch": 1.928, |
| "grad_norm": 0.61215815551841, |
| "learning_rate": 3.08827844143086e-05, |
| "loss": 0.4761, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5427641868591309, |
| "step": 1205, |
| "valid_targets_mean": 4324.7, |
| "valid_targets_min": 824 |
| }, |
| { |
| "epoch": 1.936, |
| "grad_norm": 0.5987033728383583, |
| "learning_rate": 3.078888181152264e-05, |
| "loss": 0.4722, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42406606674194336, |
| "step": 1210, |
| "valid_targets_mean": 3636.4, |
| "valid_targets_min": 684 |
| }, |
| { |
| "epoch": 1.944, |
| "grad_norm": 0.6515070925773007, |
| "learning_rate": 3.0694642554233855e-05, |
| "loss": 0.4707, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47039297223091125, |
| "step": 1215, |
| "valid_targets_mean": 2975.4, |
| "valid_targets_min": 592 |
| }, |
| { |
| "epoch": 1.952, |
| "grad_norm": 0.5388317669629852, |
| "learning_rate": 3.0600069583068594e-05, |
| "loss": 0.4779, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4415266513824463, |
| "step": 1220, |
| "valid_targets_mean": 4852.2, |
| "valid_targets_min": 347 |
| }, |
| { |
| "epoch": 1.96, |
| "grad_norm": 0.5138321633547656, |
| "learning_rate": 3.0505165849066394e-05, |
| "loss": 0.4681, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4434703290462494, |
| "step": 1225, |
| "valid_targets_mean": 5056.9, |
| "valid_targets_min": 907 |
| }, |
| { |
| "epoch": 1.968, |
| "grad_norm": 0.7612326702075074, |
| "learning_rate": 3.040993431358782e-05, |
| "loss": 0.4736, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4785926938056946, |
| "step": 1230, |
| "valid_targets_mean": 2302.5, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 1.976, |
| "grad_norm": 0.7430168909983449, |
| "learning_rate": 3.031437794822215e-05, |
| "loss": 0.4701, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46587690711021423, |
| "step": 1235, |
| "valid_targets_mean": 3480.7, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 1.984, |
| "grad_norm": 0.8871840326954018, |
| "learning_rate": 3.021849973469455e-05, |
| "loss": 0.5202, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5849260687828064, |
| "step": 1240, |
| "valid_targets_mean": 2373.0, |
| "valid_targets_min": 824 |
| }, |
| { |
| "epoch": 1.992, |
| "grad_norm": 0.8005852985566412, |
| "learning_rate": 3.012230266477313e-05, |
| "loss": 0.4786, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5191680788993835, |
| "step": 1245, |
| "valid_targets_mean": 2152.9, |
| "valid_targets_min": 834 |
| }, |
| { |
| "epoch": 2.0, |
| "grad_norm": 1.423544343942953, |
| "learning_rate": 3.0025789740175502e-05, |
| "loss": 0.4934, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5457210540771484, |
| "step": 1250, |
| "valid_targets_mean": 2222.7, |
| "valid_targets_min": 561 |
| }, |
| { |
| "epoch": 2.008, |
| "grad_norm": 0.5955977900794318, |
| "learning_rate": 2.9928963972475186e-05, |
| "loss": 0.4681, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4416769742965698, |
| "step": 1255, |
| "valid_targets_mean": 4610.1, |
| "valid_targets_min": 615 |
| }, |
| { |
| "epoch": 2.016, |
| "grad_norm": 0.6090260458793183, |
| "learning_rate": 2.9831828383007585e-05, |
| "loss": 0.4479, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4574715495109558, |
| "step": 1260, |
| "valid_targets_mean": 4182.2, |
| "valid_targets_min": 645 |
| }, |
| { |
| "epoch": 2.024, |
| "grad_norm": 0.7360060643061078, |
| "learning_rate": 2.9734386002775754e-05, |
| "loss": 0.4695, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5532656908035278, |
| "step": 1265, |
| "valid_targets_mean": 3207.1, |
| "valid_targets_min": 512 |
| }, |
| { |
| "epoch": 2.032, |
| "grad_norm": 0.6687884359084095, |
| "learning_rate": 2.963663987235577e-05, |
| "loss": 0.4552, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.426003634929657, |
| "step": 1270, |
| "valid_targets_mean": 2995.4, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 2.04, |
| "grad_norm": 0.7116484185211367, |
| "learning_rate": 2.95385930418019e-05, |
| "loss": 0.4567, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5189186334609985, |
| "step": 1275, |
| "valid_targets_mean": 3110.1, |
| "valid_targets_min": 742 |
| }, |
| { |
| "epoch": 2.048, |
| "grad_norm": 0.7186019588818685, |
| "learning_rate": 2.9440248570551406e-05, |
| "loss": 0.4459, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41865086555480957, |
| "step": 1280, |
| "valid_targets_mean": 3671.8, |
| "valid_targets_min": 906 |
| }, |
| { |
| "epoch": 2.056, |
| "grad_norm": 1.4747315343165368, |
| "learning_rate": 2.934160952732907e-05, |
| "loss": 0.449, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4162527322769165, |
| "step": 1285, |
| "valid_targets_mean": 3745.3, |
| "valid_targets_min": 306 |
| }, |
| { |
| "epoch": 2.064, |
| "grad_norm": 0.6471086351380495, |
| "learning_rate": 2.9242678990051462e-05, |
| "loss": 0.4433, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44822385907173157, |
| "step": 1290, |
| "valid_targets_mean": 4509.1, |
| "valid_targets_min": 576 |
| }, |
| { |
| "epoch": 2.072, |
| "grad_norm": 0.6346182269430595, |
| "learning_rate": 2.9143460045730886e-05, |
| "loss": 0.4487, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48546066880226135, |
| "step": 1295, |
| "valid_targets_mean": 3940.9, |
| "valid_targets_min": 876 |
| }, |
| { |
| "epoch": 2.08, |
| "grad_norm": 2.031709545540969, |
| "learning_rate": 2.9043955790379035e-05, |
| "loss": 0.4714, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43055105209350586, |
| "step": 1300, |
| "valid_targets_mean": 3915.9, |
| "valid_targets_min": 635 |
| }, |
| { |
| "epoch": 2.088, |
| "grad_norm": 0.6903051949163922, |
| "learning_rate": 2.8944169328910427e-05, |
| "loss": 0.446, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48170751333236694, |
| "step": 1305, |
| "valid_targets_mean": 3417.4, |
| "valid_targets_min": 681 |
| }, |
| { |
| "epoch": 2.096, |
| "grad_norm": 0.5332653284325558, |
| "learning_rate": 2.884410377504547e-05, |
| "loss": 0.4159, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42663711309432983, |
| "step": 1310, |
| "valid_targets_mean": 5080.2, |
| "valid_targets_min": 640 |
| }, |
| { |
| "epoch": 2.104, |
| "grad_norm": 0.602437990002518, |
| "learning_rate": 2.8743762251213333e-05, |
| "loss": 0.4508, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47259843349456787, |
| "step": 1315, |
| "valid_targets_mean": 4112.4, |
| "valid_targets_min": 646 |
| }, |
| { |
| "epoch": 2.112, |
| "grad_norm": 0.676214101245154, |
| "learning_rate": 2.8643147888454507e-05, |
| "loss": 0.4432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4584727883338928, |
| "step": 1320, |
| "valid_targets_mean": 3043.8, |
| "valid_targets_min": 415 |
| }, |
| { |
| "epoch": 2.12, |
| "grad_norm": 0.7428353700968319, |
| "learning_rate": 2.854226382632312e-05, |
| "loss": 0.4713, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5539952516555786, |
| "step": 1325, |
| "valid_targets_mean": 3083.6, |
| "valid_targets_min": 624 |
| }, |
| { |
| "epoch": 2.128, |
| "grad_norm": 0.5464368058251405, |
| "learning_rate": 2.844111321278893e-05, |
| "loss": 0.4385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3453948497772217, |
| "step": 1330, |
| "valid_targets_mean": 3928.3, |
| "valid_targets_min": 553 |
| }, |
| { |
| "epoch": 2.136, |
| "grad_norm": 0.5772508201873505, |
| "learning_rate": 2.833969920413913e-05, |
| "loss": 0.4465, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4165865182876587, |
| "step": 1335, |
| "valid_targets_mean": 3898.2, |
| "valid_targets_min": 480 |
| }, |
| { |
| "epoch": 2.144, |
| "grad_norm": 0.6774356448413013, |
| "learning_rate": 2.8238024964879857e-05, |
| "loss": 0.4465, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4822053015232086, |
| "step": 1340, |
| "valid_targets_mean": 3180.1, |
| "valid_targets_min": 686 |
| }, |
| { |
| "epoch": 2.152, |
| "grad_norm": 0.654073300893228, |
| "learning_rate": 2.8136093667637438e-05, |
| "loss": 0.4496, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4197562336921692, |
| "step": 1345, |
| "valid_targets_mean": 2752.7, |
| "valid_targets_min": 518 |
| }, |
| { |
| "epoch": 2.16, |
| "grad_norm": 0.7260228311473961, |
| "learning_rate": 2.8033908493059394e-05, |
| "loss": 0.4643, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4236224591732025, |
| "step": 1350, |
| "valid_targets_mean": 3862.9, |
| "valid_targets_min": 977 |
| }, |
| { |
| "epoch": 2.168, |
| "grad_norm": 0.5308213424796718, |
| "learning_rate": 2.793147262971519e-05, |
| "loss": 0.4484, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4658902883529663, |
| "step": 1355, |
| "valid_targets_mean": 5550.6, |
| "valid_targets_min": 763 |
| }, |
| { |
| "epoch": 2.176, |
| "grad_norm": 0.5816937352471572, |
| "learning_rate": 2.7828789273996748e-05, |
| "loss": 0.4431, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.414844274520874, |
| "step": 1360, |
| "valid_targets_mean": 4192.6, |
| "valid_targets_min": 643 |
| }, |
| { |
| "epoch": 2.184, |
| "grad_norm": 0.6690509201863886, |
| "learning_rate": 2.7725861630018703e-05, |
| "loss": 0.4441, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41888707876205444, |
| "step": 1365, |
| "valid_targets_mean": 3258.7, |
| "valid_targets_min": 857 |
| }, |
| { |
| "epoch": 2.192, |
| "grad_norm": 0.5315576725263342, |
| "learning_rate": 2.7622692909518423e-05, |
| "loss": 0.4088, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4389989376068115, |
| "step": 1370, |
| "valid_targets_mean": 5194.4, |
| "valid_targets_min": 651 |
| }, |
| { |
| "epoch": 2.2, |
| "grad_norm": 0.43861560505715885, |
| "learning_rate": 2.7519286331755766e-05, |
| "loss": 0.4343, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36606454849243164, |
| "step": 1375, |
| "valid_targets_mean": 7803.1, |
| "valid_targets_min": 915 |
| }, |
| { |
| "epoch": 2.208, |
| "grad_norm": 0.6860438916170236, |
| "learning_rate": 2.7415645123412672e-05, |
| "loss": 0.4345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4727722704410553, |
| "step": 1380, |
| "valid_targets_mean": 3681.9, |
| "valid_targets_min": 354 |
| }, |
| { |
| "epoch": 2.216, |
| "grad_norm": 0.6233752088365587, |
| "learning_rate": 2.731177251849246e-05, |
| "loss": 0.4469, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45124104619026184, |
| "step": 1385, |
| "valid_targets_mean": 3928.4, |
| "valid_targets_min": 409 |
| }, |
| { |
| "epoch": 2.224, |
| "grad_norm": 0.7054017714889622, |
| "learning_rate": 2.7207671758218884e-05, |
| "loss": 0.4511, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5190399885177612, |
| "step": 1390, |
| "valid_targets_mean": 3278.7, |
| "valid_targets_min": 668 |
| }, |
| { |
| "epoch": 2.232, |
| "grad_norm": 0.5481452159873532, |
| "learning_rate": 2.710334609093504e-05, |
| "loss": 0.4338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40331241488456726, |
| "step": 1395, |
| "valid_targets_mean": 4237.9, |
| "valid_targets_min": 929 |
| }, |
| { |
| "epoch": 2.24, |
| "grad_norm": 0.5209483042162194, |
| "learning_rate": 2.699879877200198e-05, |
| "loss": 0.4247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.367615282535553, |
| "step": 1400, |
| "valid_targets_mean": 4907.4, |
| "valid_targets_min": 718 |
| }, |
| { |
| "epoch": 2.248, |
| "grad_norm": 0.6180610305765888, |
| "learning_rate": 2.6894033063697143e-05, |
| "loss": 0.4649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4836103916168213, |
| "step": 1405, |
| "valid_targets_mean": 4520.9, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 2.2560000000000002, |
| "grad_norm": 0.624263121954133, |
| "learning_rate": 2.6789052235112554e-05, |
| "loss": 0.4516, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4151955544948578, |
| "step": 1410, |
| "valid_targets_mean": 3549.2, |
| "valid_targets_min": 739 |
| }, |
| { |
| "epoch": 2.2640000000000002, |
| "grad_norm": 0.5255415368728256, |
| "learning_rate": 2.66838595620528e-05, |
| "loss": 0.4243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4413053095340729, |
| "step": 1415, |
| "valid_targets_mean": 5782.1, |
| "valid_targets_min": 655 |
| }, |
| { |
| "epoch": 2.2720000000000002, |
| "grad_norm": 0.6397618135971025, |
| "learning_rate": 2.6578458326932842e-05, |
| "loss": 0.4536, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44738149642944336, |
| "step": 1420, |
| "valid_targets_mean": 3827.2, |
| "valid_targets_min": 505 |
| }, |
| { |
| "epoch": 2.2800000000000002, |
| "grad_norm": 0.7085382486155277, |
| "learning_rate": 2.6472851818675583e-05, |
| "loss": 0.4633, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4663681387901306, |
| "step": 1425, |
| "valid_targets_mean": 2644.9, |
| "valid_targets_min": 646 |
| }, |
| { |
| "epoch": 2.288, |
| "grad_norm": 0.9979890243748089, |
| "learning_rate": 2.6367043332609223e-05, |
| "loss": 0.4459, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4871729016304016, |
| "step": 1430, |
| "valid_targets_mean": 1448.7, |
| "valid_targets_min": 744 |
| }, |
| { |
| "epoch": 2.296, |
| "grad_norm": 0.5174927037809555, |
| "learning_rate": 2.6261036170364448e-05, |
| "loss": 0.4251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43367236852645874, |
| "step": 1435, |
| "valid_targets_mean": 5433.6, |
| "valid_targets_min": 783 |
| }, |
| { |
| "epoch": 2.304, |
| "grad_norm": 0.5663046987401196, |
| "learning_rate": 2.6154833639771415e-05, |
| "loss": 0.4422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46524518728256226, |
| "step": 1440, |
| "valid_targets_mean": 5141.4, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 2.312, |
| "grad_norm": 0.5337110127394135, |
| "learning_rate": 2.6048439054756492e-05, |
| "loss": 0.4463, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4362761676311493, |
| "step": 1445, |
| "valid_targets_mean": 5409.9, |
| "valid_targets_min": 1159 |
| }, |
| { |
| "epoch": 2.32, |
| "grad_norm": 0.5463782845574509, |
| "learning_rate": 2.594185573523892e-05, |
| "loss": 0.4565, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45417675375938416, |
| "step": 1450, |
| "valid_targets_mean": 4838.8, |
| "valid_targets_min": 619 |
| }, |
| { |
| "epoch": 2.328, |
| "grad_norm": 0.6942365338980392, |
| "learning_rate": 2.583508700702716e-05, |
| "loss": 0.4481, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4617985486984253, |
| "step": 1455, |
| "valid_targets_mean": 2802.6, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 2.336, |
| "grad_norm": 0.6070000633127984, |
| "learning_rate": 2.572813620171513e-05, |
| "loss": 0.4513, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42797911167144775, |
| "step": 1460, |
| "valid_targets_mean": 3719.3, |
| "valid_targets_min": 635 |
| }, |
| { |
| "epoch": 2.344, |
| "grad_norm": 0.4753849685324334, |
| "learning_rate": 2.5621006656578267e-05, |
| "loss": 0.4213, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3997453451156616, |
| "step": 1465, |
| "valid_targets_mean": 5544.6, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 2.352, |
| "grad_norm": 0.5989976974737479, |
| "learning_rate": 2.5513701714469373e-05, |
| "loss": 0.4413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4619044363498688, |
| "step": 1470, |
| "valid_targets_mean": 4001.4, |
| "valid_targets_min": 672 |
| }, |
| { |
| "epoch": 2.36, |
| "grad_norm": 0.5487842792272467, |
| "learning_rate": 2.540622472371429e-05, |
| "loss": 0.4343, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4185061454772949, |
| "step": 1475, |
| "valid_targets_mean": 4278.5, |
| "valid_targets_min": 426 |
| }, |
| { |
| "epoch": 2.368, |
| "grad_norm": 0.4402218121898237, |
| "learning_rate": 2.5298579038007478e-05, |
| "loss": 0.4342, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3919063210487366, |
| "step": 1480, |
| "valid_targets_mean": 6910.7, |
| "valid_targets_min": 815 |
| }, |
| { |
| "epoch": 2.376, |
| "grad_norm": 0.7373247521439833, |
| "learning_rate": 2.519076801630727e-05, |
| "loss": 0.4723, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5013588070869446, |
| "step": 1485, |
| "valid_targets_mean": 2875.9, |
| "valid_targets_min": 658 |
| }, |
| { |
| "epoch": 2.384, |
| "grad_norm": 0.556610693165223, |
| "learning_rate": 2.508279502273117e-05, |
| "loss": 0.4419, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43320780992507935, |
| "step": 1490, |
| "valid_targets_mean": 5407.4, |
| "valid_targets_min": 836 |
| }, |
| { |
| "epoch": 2.392, |
| "grad_norm": 0.5098627681422288, |
| "learning_rate": 2.4974663426450798e-05, |
| "loss": 0.4392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4344981908798218, |
| "step": 1495, |
| "valid_targets_mean": 5123.3, |
| "valid_targets_min": 601 |
| }, |
| { |
| "epoch": 2.4, |
| "grad_norm": 0.5039771982978674, |
| "learning_rate": 2.4866376601586798e-05, |
| "loss": 0.4464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4104907810688019, |
| "step": 1500, |
| "valid_targets_mean": 5579.6, |
| "valid_targets_min": 754 |
| }, |
| { |
| "epoch": 2.408, |
| "grad_norm": 0.5096044181569787, |
| "learning_rate": 2.475793792710352e-05, |
| "loss": 0.4174, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4063170552253723, |
| "step": 1505, |
| "valid_targets_mean": 5081.4, |
| "valid_targets_min": 488 |
| }, |
| { |
| "epoch": 2.416, |
| "grad_norm": 0.7479128529316125, |
| "learning_rate": 2.4649350786703637e-05, |
| "loss": 0.448, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47363007068634033, |
| "step": 1510, |
| "valid_targets_mean": 2561.8, |
| "valid_targets_min": 708 |
| }, |
| { |
| "epoch": 2.424, |
| "grad_norm": 0.5885671758624585, |
| "learning_rate": 2.45406185687225e-05, |
| "loss": 0.4514, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44971561431884766, |
| "step": 1515, |
| "valid_targets_mean": 4042.2, |
| "valid_targets_min": 678 |
| }, |
| { |
| "epoch": 2.432, |
| "grad_norm": 0.5915561518938581, |
| "learning_rate": 2.443174466602246e-05, |
| "loss": 0.4559, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4462353587150574, |
| "step": 1520, |
| "valid_targets_mean": 3938.5, |
| "valid_targets_min": 765 |
| }, |
| { |
| "epoch": 2.44, |
| "grad_norm": 0.663840439463274, |
| "learning_rate": 2.4322732475886953e-05, |
| "loss": 0.4457, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.426447331905365, |
| "step": 1525, |
| "valid_targets_mean": 2940.4, |
| "valid_targets_min": 494 |
| }, |
| { |
| "epoch": 2.448, |
| "grad_norm": 0.5258251709290092, |
| "learning_rate": 2.4213585399914528e-05, |
| "loss": 0.4099, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4359118640422821, |
| "step": 1530, |
| "valid_targets_mean": 4996.3, |
| "valid_targets_min": 832 |
| }, |
| { |
| "epoch": 2.456, |
| "grad_norm": 0.5279872310300597, |
| "learning_rate": 2.4104306843912687e-05, |
| "loss": 0.4368, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4399365186691284, |
| "step": 1535, |
| "valid_targets_mean": 5331.9, |
| "valid_targets_min": 521 |
| }, |
| { |
| "epoch": 2.464, |
| "grad_norm": 0.5035482920980331, |
| "learning_rate": 2.3994900217791615e-05, |
| "loss": 0.4564, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4546802341938019, |
| "step": 1540, |
| "valid_targets_mean": 6091.1, |
| "valid_targets_min": 553 |
| }, |
| { |
| "epoch": 2.472, |
| "grad_norm": 0.5503280071758032, |
| "learning_rate": 2.3885368935457762e-05, |
| "loss": 0.4442, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38341644406318665, |
| "step": 1545, |
| "valid_targets_mean": 4704.2, |
| "valid_targets_min": 534 |
| }, |
| { |
| "epoch": 2.48, |
| "grad_norm": 0.5122727644758672, |
| "learning_rate": 2.3775716414707355e-05, |
| "loss": 0.4236, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39771783351898193, |
| "step": 1550, |
| "valid_targets_mean": 5978.4, |
| "valid_targets_min": 737 |
| }, |
| { |
| "epoch": 2.488, |
| "grad_norm": 0.577359891725372, |
| "learning_rate": 2.36659460771197e-05, |
| "loss": 0.4408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.426598459482193, |
| "step": 1555, |
| "valid_targets_mean": 4298.5, |
| "valid_targets_min": 758 |
| }, |
| { |
| "epoch": 2.496, |
| "grad_norm": 0.534664454111424, |
| "learning_rate": 2.3556061347950455e-05, |
| "loss": 0.422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41438737511634827, |
| "step": 1560, |
| "valid_targets_mean": 4637.9, |
| "valid_targets_min": 1024 |
| }, |
| { |
| "epoch": 2.504, |
| "grad_norm": 0.742643154280012, |
| "learning_rate": 2.3446065656024734e-05, |
| "loss": 0.4412, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45946621894836426, |
| "step": 1565, |
| "valid_targets_mean": 2557.0, |
| "valid_targets_min": 568 |
| }, |
| { |
| "epoch": 2.512, |
| "grad_norm": 0.5946087395842765, |
| "learning_rate": 2.33359624336301e-05, |
| "loss": 0.4429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44135582447052, |
| "step": 1570, |
| "valid_targets_mean": 3980.1, |
| "valid_targets_min": 614 |
| }, |
| { |
| "epoch": 2.52, |
| "grad_norm": 0.7692556639202142, |
| "learning_rate": 2.3225755116409497e-05, |
| "loss": 0.4855, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.49565717577934265, |
| "step": 1575, |
| "valid_targets_mean": 2437.8, |
| "valid_targets_min": 702 |
| }, |
| { |
| "epoch": 2.528, |
| "grad_norm": 0.6518050042070153, |
| "learning_rate": 2.311544714325403e-05, |
| "loss": 0.4278, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4329824149608612, |
| "step": 1580, |
| "valid_targets_mean": 3308.9, |
| "valid_targets_min": 837 |
| }, |
| { |
| "epoch": 2.536, |
| "grad_norm": 0.7001492648412008, |
| "learning_rate": 2.300504195619563e-05, |
| "loss": 0.4865, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47200626134872437, |
| "step": 1585, |
| "valid_targets_mean": 3019.2, |
| "valid_targets_min": 723 |
| }, |
| { |
| "epoch": 2.544, |
| "grad_norm": 0.5775943044286223, |
| "learning_rate": 2.2894543000299697e-05, |
| "loss": 0.4546, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4556680917739868, |
| "step": 1590, |
| "valid_targets_mean": 4622.2, |
| "valid_targets_min": 746 |
| }, |
| { |
| "epoch": 2.552, |
| "grad_norm": 0.6300239166139451, |
| "learning_rate": 2.2783953723557572e-05, |
| "loss": 0.4442, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47904396057128906, |
| "step": 1595, |
| "valid_targets_mean": 3739.5, |
| "valid_targets_min": 691 |
| }, |
| { |
| "epoch": 2.56, |
| "grad_norm": 0.5521871360992379, |
| "learning_rate": 2.2673277576778946e-05, |
| "loss": 0.4353, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4208674728870392, |
| "step": 1600, |
| "valid_targets_mean": 4564.6, |
| "valid_targets_min": 555 |
| }, |
| { |
| "epoch": 2.568, |
| "grad_norm": 0.6820540018619283, |
| "learning_rate": 2.2562518013484208e-05, |
| "loss": 0.4273, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40282294154167175, |
| "step": 1605, |
| "valid_targets_mean": 2706.5, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 2.576, |
| "grad_norm": 0.6077922915119577, |
| "learning_rate": 2.245167848979664e-05, |
| "loss": 0.4142, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42859092354774475, |
| "step": 1610, |
| "valid_targets_mean": 3625.1, |
| "valid_targets_min": 509 |
| }, |
| { |
| "epoch": 2.584, |
| "grad_norm": 0.5854962148493413, |
| "learning_rate": 2.23407624643346e-05, |
| "loss": 0.4648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42316490411758423, |
| "step": 1615, |
| "valid_targets_mean": 4010.8, |
| "valid_targets_min": 713 |
| }, |
| { |
| "epoch": 2.592, |
| "grad_norm": 0.544244197674275, |
| "learning_rate": 2.2229773398103606e-05, |
| "loss": 0.4167, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40714818239212036, |
| "step": 1620, |
| "valid_targets_mean": 4601.6, |
| "valid_targets_min": 847 |
| }, |
| { |
| "epoch": 2.6, |
| "grad_norm": 0.7315101815116987, |
| "learning_rate": 2.2118714754388323e-05, |
| "loss": 0.4689, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4557625651359558, |
| "step": 1625, |
| "valid_targets_mean": 2673.6, |
| "valid_targets_min": 627 |
| }, |
| { |
| "epoch": 2.608, |
| "grad_norm": 0.6392348215128343, |
| "learning_rate": 2.200758999864449e-05, |
| "loss": 0.4095, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41079527139663696, |
| "step": 1630, |
| "valid_targets_mean": 3180.8, |
| "valid_targets_min": 492 |
| }, |
| { |
| "epoch": 2.616, |
| "grad_norm": 0.553837921244433, |
| "learning_rate": 2.1896402598390818e-05, |
| "loss": 0.419, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3827151358127594, |
| "step": 1635, |
| "valid_targets_mean": 4638.4, |
| "valid_targets_min": 616 |
| }, |
| { |
| "epoch": 2.624, |
| "grad_norm": 0.5193517813868557, |
| "learning_rate": 2.178515602310074e-05, |
| "loss": 0.4533, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40580642223358154, |
| "step": 1640, |
| "valid_targets_mean": 4868.9, |
| "valid_targets_min": 799 |
| }, |
| { |
| "epoch": 2.632, |
| "grad_norm": 0.6317219989613208, |
| "learning_rate": 2.1673853744094193e-05, |
| "loss": 0.4467, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4618436098098755, |
| "step": 1645, |
| "valid_targets_mean": 3528.4, |
| "valid_targets_min": 583 |
| }, |
| { |
| "epoch": 2.64, |
| "grad_norm": 0.8058706372384056, |
| "learning_rate": 2.1562499234429283e-05, |
| "loss": 0.4757, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4898325800895691, |
| "step": 1650, |
| "valid_targets_mean": 2231.3, |
| "valid_targets_min": 507 |
| }, |
| { |
| "epoch": 2.648, |
| "grad_norm": 0.7352262306367001, |
| "learning_rate": 2.1451095968793908e-05, |
| "loss": 0.4389, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46522995829582214, |
| "step": 1655, |
| "valid_targets_mean": 3700.2, |
| "valid_targets_min": 542 |
| }, |
| { |
| "epoch": 2.656, |
| "grad_norm": 0.6500232390169828, |
| "learning_rate": 2.1339647423397337e-05, |
| "loss": 0.4385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4323245882987976, |
| "step": 1660, |
| "valid_targets_mean": 3278.4, |
| "valid_targets_min": 774 |
| }, |
| { |
| "epoch": 2.664, |
| "grad_norm": 0.6031196836473796, |
| "learning_rate": 2.122815707586176e-05, |
| "loss": 0.4358, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4227977991104126, |
| "step": 1665, |
| "valid_targets_mean": 3835.8, |
| "valid_targets_min": 517 |
| }, |
| { |
| "epoch": 2.672, |
| "grad_norm": 0.5631354627031511, |
| "learning_rate": 2.111662840511373e-05, |
| "loss": 0.4372, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40425512194633484, |
| "step": 1670, |
| "valid_targets_mean": 4473.6, |
| "valid_targets_min": 591 |
| }, |
| { |
| "epoch": 2.68, |
| "grad_norm": 0.6281174285703458, |
| "learning_rate": 2.1005064891275638e-05, |
| "loss": 0.4534, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4625401496887207, |
| "step": 1675, |
| "valid_targets_mean": 3325.9, |
| "valid_targets_min": 618 |
| }, |
| { |
| "epoch": 2.6879999999999997, |
| "grad_norm": 0.5324007779451334, |
| "learning_rate": 2.0893470015557126e-05, |
| "loss": 0.4137, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3770933151245117, |
| "step": 1680, |
| "valid_targets_mean": 4407.2, |
| "valid_targets_min": 854 |
| }, |
| { |
| "epoch": 2.6959999999999997, |
| "grad_norm": 0.5704221627115216, |
| "learning_rate": 2.078184726014643e-05, |
| "loss": 0.4524, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4448625445365906, |
| "step": 1685, |
| "valid_targets_mean": 4212.2, |
| "valid_targets_min": 669 |
| }, |
| { |
| "epoch": 2.7039999999999997, |
| "grad_norm": 0.5813144458895672, |
| "learning_rate": 2.0670200108101754e-05, |
| "loss": 0.4548, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4791436195373535, |
| "step": 1690, |
| "valid_targets_mean": 4126.9, |
| "valid_targets_min": 473 |
| }, |
| { |
| "epoch": 2.7119999999999997, |
| "grad_norm": 0.8782653466918007, |
| "learning_rate": 2.0558532043242557e-05, |
| "loss": 0.464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.453347384929657, |
| "step": 1695, |
| "valid_targets_mean": 1782.1, |
| "valid_targets_min": 564 |
| }, |
| { |
| "epoch": 2.7199999999999998, |
| "grad_norm": 0.5583160859664104, |
| "learning_rate": 2.0446846550040863e-05, |
| "loss": 0.428, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4322124719619751, |
| "step": 1700, |
| "valid_targets_mean": 4821.1, |
| "valid_targets_min": 669 |
| }, |
| { |
| "epoch": 2.7279999999999998, |
| "grad_norm": 0.5277977577152152, |
| "learning_rate": 2.033514711351253e-05, |
| "loss": 0.4315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42123162746429443, |
| "step": 1705, |
| "valid_targets_mean": 5344.1, |
| "valid_targets_min": 709 |
| }, |
| { |
| "epoch": 2.7359999999999998, |
| "grad_norm": 0.498988724319745, |
| "learning_rate": 2.022343721910851e-05, |
| "loss": 0.4515, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44440481066703796, |
| "step": 1710, |
| "valid_targets_mean": 6055.7, |
| "valid_targets_min": 701 |
| }, |
| { |
| "epoch": 2.7439999999999998, |
| "grad_norm": 0.5688318880815759, |
| "learning_rate": 2.0111720352606054e-05, |
| "loss": 0.4293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4387132227420807, |
| "step": 1715, |
| "valid_targets_mean": 4293.4, |
| "valid_targets_min": 1000 |
| }, |
| { |
| "epoch": 2.752, |
| "grad_norm": 0.5541189383448685, |
| "learning_rate": 2e-05, |
| "loss": 0.4345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42854511737823486, |
| "step": 1720, |
| "valid_targets_mean": 4060.4, |
| "valid_targets_min": 1006 |
| }, |
| { |
| "epoch": 2.76, |
| "grad_norm": 0.5636891444562256, |
| "learning_rate": 1.988827964739395e-05, |
| "loss": 0.4569, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4016704559326172, |
| "step": 1725, |
| "valid_targets_mean": 4026.6, |
| "valid_targets_min": 556 |
| }, |
| { |
| "epoch": 2.768, |
| "grad_norm": 0.5707069098467574, |
| "learning_rate": 1.9776562780891494e-05, |
| "loss": 0.4277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3943563997745514, |
| "step": 1730, |
| "valid_targets_mean": 3987.3, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 2.776, |
| "grad_norm": 0.46274734962368136, |
| "learning_rate": 1.966485288648747e-05, |
| "loss": 0.4408, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38267430663108826, |
| "step": 1735, |
| "valid_targets_mean": 6487.2, |
| "valid_targets_min": 785 |
| }, |
| { |
| "epoch": 2.784, |
| "grad_norm": 0.5235450558438844, |
| "learning_rate": 1.9553153449959144e-05, |
| "loss": 0.4465, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40462425351142883, |
| "step": 1740, |
| "valid_targets_mean": 4937.6, |
| "valid_targets_min": 741 |
| }, |
| { |
| "epoch": 2.792, |
| "grad_norm": 0.4835906686418437, |
| "learning_rate": 1.9441467956757453e-05, |
| "loss": 0.4716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4714353680610657, |
| "step": 1745, |
| "valid_targets_mean": 6360.5, |
| "valid_targets_min": 654 |
| }, |
| { |
| "epoch": 2.8, |
| "grad_norm": 0.5100225254528137, |
| "learning_rate": 1.9329799891898256e-05, |
| "loss": 0.4314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4558408260345459, |
| "step": 1750, |
| "valid_targets_mean": 5252.3, |
| "valid_targets_min": 707 |
| }, |
| { |
| "epoch": 2.808, |
| "grad_norm": 0.7039108297346979, |
| "learning_rate": 1.9218152739853576e-05, |
| "loss": 0.4289, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3978843688964844, |
| "step": 1755, |
| "valid_targets_mean": 4321.9, |
| "valid_targets_min": 725 |
| }, |
| { |
| "epoch": 2.816, |
| "grad_norm": 0.5642168770075947, |
| "learning_rate": 1.9106529984442884e-05, |
| "loss": 0.4398, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45182371139526367, |
| "step": 1760, |
| "valid_targets_mean": 4244.4, |
| "valid_targets_min": 552 |
| }, |
| { |
| "epoch": 2.824, |
| "grad_norm": 0.6525684551205236, |
| "learning_rate": 1.8994935108724366e-05, |
| "loss": 0.4521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4625537395477295, |
| "step": 1765, |
| "valid_targets_mean": 3384.4, |
| "valid_targets_min": 838 |
| }, |
| { |
| "epoch": 2.832, |
| "grad_norm": 0.5604654454033197, |
| "learning_rate": 1.8883371594886276e-05, |
| "loss": 0.4497, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4158444404602051, |
| "step": 1770, |
| "valid_targets_mean": 5318.2, |
| "valid_targets_min": 598 |
| }, |
| { |
| "epoch": 2.84, |
| "grad_norm": 0.7805033021518426, |
| "learning_rate": 1.877184292413824e-05, |
| "loss": 0.4403, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46476149559020996, |
| "step": 1775, |
| "valid_targets_mean": 2218.4, |
| "valid_targets_min": 762 |
| }, |
| { |
| "epoch": 2.848, |
| "grad_norm": 0.632248312453592, |
| "learning_rate": 1.8660352576602663e-05, |
| "loss": 0.4715, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4902727007865906, |
| "step": 1780, |
| "valid_targets_mean": 4020.6, |
| "valid_targets_min": 848 |
| }, |
| { |
| "epoch": 2.856, |
| "grad_norm": 0.8500750674610906, |
| "learning_rate": 1.8548904031206102e-05, |
| "loss": 0.4558, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5282151103019714, |
| "step": 1785, |
| "valid_targets_mean": 2049.2, |
| "valid_targets_min": 489 |
| }, |
| { |
| "epoch": 2.864, |
| "grad_norm": 0.5478175008230401, |
| "learning_rate": 1.843750076557072e-05, |
| "loss": 0.4351, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4625953137874603, |
| "step": 1790, |
| "valid_targets_mean": 4885.2, |
| "valid_targets_min": 506 |
| }, |
| { |
| "epoch": 2.872, |
| "grad_norm": 0.6244527229546376, |
| "learning_rate": 1.832614625590581e-05, |
| "loss": 0.4246, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43290358781814575, |
| "step": 1795, |
| "valid_targets_mean": 3414.3, |
| "valid_targets_min": 574 |
| }, |
| { |
| "epoch": 2.88, |
| "grad_norm": 0.8234587446370712, |
| "learning_rate": 1.8214843976899264e-05, |
| "loss": 0.454, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4970802664756775, |
| "step": 1800, |
| "valid_targets_mean": 2389.6, |
| "valid_targets_min": 511 |
| }, |
| { |
| "epoch": 2.888, |
| "grad_norm": 0.7474318673230876, |
| "learning_rate": 1.810359740160919e-05, |
| "loss": 0.4377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44770461320877075, |
| "step": 1805, |
| "valid_targets_mean": 2616.4, |
| "valid_targets_min": 629 |
| }, |
| { |
| "epoch": 2.896, |
| "grad_norm": 0.507392374752142, |
| "learning_rate": 1.7992410001355515e-05, |
| "loss": 0.4388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4395235776901245, |
| "step": 1810, |
| "valid_targets_mean": 5488.2, |
| "valid_targets_min": 817 |
| }, |
| { |
| "epoch": 2.904, |
| "grad_norm": 0.7243529623293884, |
| "learning_rate": 1.788128524561168e-05, |
| "loss": 0.4574, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5047330260276794, |
| "step": 1815, |
| "valid_targets_mean": 3967.6, |
| "valid_targets_min": 812 |
| }, |
| { |
| "epoch": 2.912, |
| "grad_norm": 0.633168445724496, |
| "learning_rate": 1.7770226601896397e-05, |
| "loss": 0.4324, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47217708826065063, |
| "step": 1820, |
| "valid_targets_mean": 3495.6, |
| "valid_targets_min": 767 |
| }, |
| { |
| "epoch": 2.92, |
| "grad_norm": 0.5327318709010629, |
| "learning_rate": 1.7659237535665404e-05, |
| "loss": 0.4572, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4306362569332123, |
| "step": 1825, |
| "valid_targets_mean": 4756.9, |
| "valid_targets_min": 754 |
| }, |
| { |
| "epoch": 2.928, |
| "grad_norm": 0.7771946593440502, |
| "learning_rate": 1.754832151020337e-05, |
| "loss": 0.4335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48200926184654236, |
| "step": 1830, |
| "valid_targets_mean": 2481.1, |
| "valid_targets_min": 794 |
| }, |
| { |
| "epoch": 2.936, |
| "grad_norm": 0.7503956631971685, |
| "learning_rate": 1.74374819865158e-05, |
| "loss": 0.435, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43350616097450256, |
| "step": 1835, |
| "valid_targets_mean": 2832.5, |
| "valid_targets_min": 751 |
| }, |
| { |
| "epoch": 2.944, |
| "grad_norm": 0.5813547416857596, |
| "learning_rate": 1.7326722423221057e-05, |
| "loss": 0.4394, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39921069145202637, |
| "step": 1840, |
| "valid_targets_mean": 3993.1, |
| "valid_targets_min": 597 |
| }, |
| { |
| "epoch": 2.952, |
| "grad_norm": 0.5326305159342818, |
| "learning_rate": 1.7216046276442438e-05, |
| "loss": 0.4184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40970325469970703, |
| "step": 1845, |
| "valid_targets_mean": 5141.3, |
| "valid_targets_min": 868 |
| }, |
| { |
| "epoch": 2.96, |
| "grad_norm": 0.5726519569305819, |
| "learning_rate": 1.7105456999700306e-05, |
| "loss": 0.4379, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4494893550872803, |
| "step": 1850, |
| "valid_targets_mean": 4538.5, |
| "valid_targets_min": 1067 |
| }, |
| { |
| "epoch": 2.968, |
| "grad_norm": 0.5532240290014173, |
| "learning_rate": 1.6994958043804374e-05, |
| "loss": 0.4503, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41899892687797546, |
| "step": 1855, |
| "valid_targets_mean": 4828.1, |
| "valid_targets_min": 602 |
| }, |
| { |
| "epoch": 2.976, |
| "grad_norm": 0.721900867482457, |
| "learning_rate": 1.6884552856745972e-05, |
| "loss": 0.4625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46213817596435547, |
| "step": 1860, |
| "valid_targets_mean": 2615.0, |
| "valid_targets_min": 288 |
| }, |
| { |
| "epoch": 2.984, |
| "grad_norm": 0.7315015103460375, |
| "learning_rate": 1.6774244883590503e-05, |
| "loss": 0.429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4381354749202728, |
| "step": 1865, |
| "valid_targets_mean": 2520.1, |
| "valid_targets_min": 626 |
| }, |
| { |
| "epoch": 2.992, |
| "grad_norm": 0.5819438726684737, |
| "learning_rate": 1.6664037566369905e-05, |
| "loss": 0.4362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4385232925415039, |
| "step": 1870, |
| "valid_targets_mean": 4100.1, |
| "valid_targets_min": 512 |
| }, |
| { |
| "epoch": 3.0, |
| "grad_norm": 0.6054633776816779, |
| "learning_rate": 1.6553934343975273e-05, |
| "loss": 0.4334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.390241801738739, |
| "step": 1875, |
| "valid_targets_mean": 3394.1, |
| "valid_targets_min": 613 |
| }, |
| { |
| "epoch": 3.008, |
| "grad_norm": 0.5137093432064929, |
| "learning_rate": 1.644393865204955e-05, |
| "loss": 0.4403, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4124566316604614, |
| "step": 1880, |
| "valid_targets_mean": 4975.0, |
| "valid_targets_min": 279 |
| }, |
| { |
| "epoch": 3.016, |
| "grad_norm": 0.6040955563332363, |
| "learning_rate": 1.6334053922880304e-05, |
| "loss": 0.4166, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4080016613006592, |
| "step": 1885, |
| "valid_targets_mean": 4139.9, |
| "valid_targets_min": 1014 |
| }, |
| { |
| "epoch": 3.024, |
| "grad_norm": 0.5643520837040242, |
| "learning_rate": 1.622428358529265e-05, |
| "loss": 0.4374, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41588467359542847, |
| "step": 1890, |
| "valid_targets_mean": 4808.7, |
| "valid_targets_min": 837 |
| }, |
| { |
| "epoch": 3.032, |
| "grad_norm": 0.5036209628878098, |
| "learning_rate": 1.611463106454224e-05, |
| "loss": 0.4334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36360183358192444, |
| "step": 1895, |
| "valid_targets_mean": 5222.1, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 3.04, |
| "grad_norm": 0.643260239413669, |
| "learning_rate": 1.6005099782208392e-05, |
| "loss": 0.4385, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4102736711502075, |
| "step": 1900, |
| "valid_targets_mean": 3426.6, |
| "valid_targets_min": 624 |
| }, |
| { |
| "epoch": 3.048, |
| "grad_norm": 0.5073496084996462, |
| "learning_rate": 1.5895693156087317e-05, |
| "loss": 0.4128, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41783252358436584, |
| "step": 1905, |
| "valid_targets_mean": 6147.1, |
| "valid_targets_min": 912 |
| }, |
| { |
| "epoch": 3.056, |
| "grad_norm": 0.6943111946599407, |
| "learning_rate": 1.578641460008548e-05, |
| "loss": 0.4402, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5079389214515686, |
| "step": 1910, |
| "valid_targets_mean": 3470.3, |
| "valid_targets_min": 508 |
| }, |
| { |
| "epoch": 3.064, |
| "grad_norm": 0.7091661125569685, |
| "learning_rate": 1.5677267524113054e-05, |
| "loss": 0.4107, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4140545129776001, |
| "step": 1915, |
| "valid_targets_mean": 3084.1, |
| "valid_targets_min": 626 |
| }, |
| { |
| "epoch": 3.072, |
| "grad_norm": 0.6859858593138877, |
| "learning_rate": 1.5568255333977547e-05, |
| "loss": 0.4185, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4497836232185364, |
| "step": 1920, |
| "valid_targets_mean": 3122.9, |
| "valid_targets_min": 664 |
| }, |
| { |
| "epoch": 3.08, |
| "grad_norm": 0.7339049628922232, |
| "learning_rate": 1.5459381431277506e-05, |
| "loss": 0.4082, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3879634737968445, |
| "step": 1925, |
| "valid_targets_mean": 2320.3, |
| "valid_targets_min": 447 |
| }, |
| { |
| "epoch": 3.088, |
| "grad_norm": 1.7541410222491485, |
| "learning_rate": 1.5350649213296373e-05, |
| "loss": 0.4249, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46820366382598877, |
| "step": 1930, |
| "valid_targets_mean": 3518.1, |
| "valid_targets_min": 646 |
| }, |
| { |
| "epoch": 3.096, |
| "grad_norm": 0.6039401909562907, |
| "learning_rate": 1.5242062072896483e-05, |
| "loss": 0.4204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4237516522407532, |
| "step": 1935, |
| "valid_targets_mean": 4190.0, |
| "valid_targets_min": 884 |
| }, |
| { |
| "epoch": 3.104, |
| "grad_norm": 0.6053990380451738, |
| "learning_rate": 1.5133623398413209e-05, |
| "loss": 0.3991, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38883084058761597, |
| "step": 1940, |
| "valid_targets_mean": 3890.1, |
| "valid_targets_min": 621 |
| }, |
| { |
| "epoch": 3.112, |
| "grad_norm": 0.5153567551089437, |
| "learning_rate": 1.50253365735492e-05, |
| "loss": 0.4124, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3877636790275574, |
| "step": 1945, |
| "valid_targets_mean": 4963.4, |
| "valid_targets_min": 519 |
| }, |
| { |
| "epoch": 3.12, |
| "grad_norm": 0.7398879295621316, |
| "learning_rate": 1.4917204977268833e-05, |
| "loss": 0.432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4899936318397522, |
| "step": 1950, |
| "valid_targets_mean": 3293.1, |
| "valid_targets_min": 745 |
| }, |
| { |
| "epoch": 3.128, |
| "grad_norm": 0.5271555062833402, |
| "learning_rate": 1.4809231983692733e-05, |
| "loss": 0.4097, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37848031520843506, |
| "step": 1955, |
| "valid_targets_mean": 4639.6, |
| "valid_targets_min": 826 |
| }, |
| { |
| "epoch": 3.136, |
| "grad_norm": 0.5550015442785227, |
| "learning_rate": 1.4701420961992533e-05, |
| "loss": 0.424, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4084477126598358, |
| "step": 1960, |
| "valid_targets_mean": 4837.2, |
| "valid_targets_min": 556 |
| }, |
| { |
| "epoch": 3.144, |
| "grad_norm": 0.7096847519468681, |
| "learning_rate": 1.459377527628571e-05, |
| "loss": 0.4275, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4384467899799347, |
| "step": 1965, |
| "valid_targets_mean": 2920.6, |
| "valid_targets_min": 723 |
| }, |
| { |
| "epoch": 3.152, |
| "grad_norm": 0.7301446675432169, |
| "learning_rate": 1.4486298285530634e-05, |
| "loss": 0.4267, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41845598816871643, |
| "step": 1970, |
| "valid_targets_mean": 2488.9, |
| "valid_targets_min": 774 |
| }, |
| { |
| "epoch": 3.16, |
| "grad_norm": 0.675361855978539, |
| "learning_rate": 1.4378993343421736e-05, |
| "loss": 0.4293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41537049412727356, |
| "step": 1975, |
| "valid_targets_mean": 3094.2, |
| "valid_targets_min": 416 |
| }, |
| { |
| "epoch": 3.168, |
| "grad_norm": 0.5632344378219207, |
| "learning_rate": 1.4271863798284877e-05, |
| "loss": 0.4141, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4148460030555725, |
| "step": 1980, |
| "valid_targets_mean": 4743.3, |
| "valid_targets_min": 986 |
| }, |
| { |
| "epoch": 3.176, |
| "grad_norm": 0.6386673015303019, |
| "learning_rate": 1.4164912992972846e-05, |
| "loss": 0.4376, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4006524682044983, |
| "step": 1985, |
| "valid_targets_mean": 3333.8, |
| "valid_targets_min": 523 |
| }, |
| { |
| "epoch": 3.184, |
| "grad_norm": 0.7341708673500656, |
| "learning_rate": 1.4058144264761087e-05, |
| "loss": 0.4131, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4031445384025574, |
| "step": 1990, |
| "valid_targets_mean": 5348.2, |
| "valid_targets_min": 653 |
| }, |
| { |
| "epoch": 3.192, |
| "grad_norm": 0.6108805400299323, |
| "learning_rate": 1.3951560945243517e-05, |
| "loss": 0.4285, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39428362250328064, |
| "step": 1995, |
| "valid_targets_mean": 3951.9, |
| "valid_targets_min": 742 |
| }, |
| { |
| "epoch": 3.2, |
| "grad_norm": 0.5744519338390691, |
| "learning_rate": 1.3845166360228597e-05, |
| "loss": 0.4248, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40177762508392334, |
| "step": 2000, |
| "valid_targets_mean": 4564.3, |
| "valid_targets_min": 684 |
| }, |
| { |
| "epoch": 3.208, |
| "grad_norm": 0.6863671838454662, |
| "learning_rate": 1.3738963829635559e-05, |
| "loss": 0.4504, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46749067306518555, |
| "step": 2005, |
| "valid_targets_mean": 3531.8, |
| "valid_targets_min": 492 |
| }, |
| { |
| "epoch": 3.216, |
| "grad_norm": 0.6671591173606514, |
| "learning_rate": 1.3632956667390784e-05, |
| "loss": 0.407, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41678929328918457, |
| "step": 2010, |
| "valid_targets_mean": 3826.6, |
| "valid_targets_min": 561 |
| }, |
| { |
| "epoch": 3.224, |
| "grad_norm": 0.588235248316702, |
| "learning_rate": 1.3527148181324425e-05, |
| "loss": 0.4276, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4322141706943512, |
| "step": 2015, |
| "valid_targets_mean": 4300.5, |
| "valid_targets_min": 661 |
| }, |
| { |
| "epoch": 3.232, |
| "grad_norm": 0.5254294071213138, |
| "learning_rate": 1.3421541673067168e-05, |
| "loss": 0.4247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38688722252845764, |
| "step": 2020, |
| "valid_targets_mean": 4773.9, |
| "valid_targets_min": 1061 |
| }, |
| { |
| "epoch": 3.24, |
| "grad_norm": 0.7778220547959938, |
| "learning_rate": 1.3316140437947207e-05, |
| "loss": 0.4248, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4182344675064087, |
| "step": 2025, |
| "valid_targets_mean": 4412.4, |
| "valid_targets_min": 572 |
| }, |
| { |
| "epoch": 3.248, |
| "grad_norm": 0.5354024193090587, |
| "learning_rate": 1.321094776488745e-05, |
| "loss": 0.4155, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38844114542007446, |
| "step": 2030, |
| "valid_targets_mean": 4513.8, |
| "valid_targets_min": 901 |
| }, |
| { |
| "epoch": 3.2560000000000002, |
| "grad_norm": 0.6307442771592611, |
| "learning_rate": 1.3105966936302856e-05, |
| "loss": 0.4213, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3858344554901123, |
| "step": 2035, |
| "valid_targets_mean": 3388.1, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 3.2640000000000002, |
| "grad_norm": 0.7489956045920521, |
| "learning_rate": 1.3001201227998023e-05, |
| "loss": 0.4295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4188210964202881, |
| "step": 2040, |
| "valid_targets_mean": 2661.7, |
| "valid_targets_min": 659 |
| }, |
| { |
| "epoch": 3.2720000000000002, |
| "grad_norm": 0.650505657663387, |
| "learning_rate": 1.2896653909064964e-05, |
| "loss": 0.4111, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4127410650253296, |
| "step": 2045, |
| "valid_targets_mean": 4998.3, |
| "valid_targets_min": 813 |
| }, |
| { |
| "epoch": 3.2800000000000002, |
| "grad_norm": 0.5978107263413696, |
| "learning_rate": 1.2792328241781124e-05, |
| "loss": 0.4016, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4069294035434723, |
| "step": 2050, |
| "valid_targets_mean": 4051.4, |
| "valid_targets_min": 744 |
| }, |
| { |
| "epoch": 3.288, |
| "grad_norm": 0.629812123697598, |
| "learning_rate": 1.2688227481507546e-05, |
| "loss": 0.4182, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4324589669704437, |
| "step": 2055, |
| "valid_targets_mean": 3744.4, |
| "valid_targets_min": 681 |
| }, |
| { |
| "epoch": 3.296, |
| "grad_norm": 0.5549507988646019, |
| "learning_rate": 1.258435487658733e-05, |
| "loss": 0.4164, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40244030952453613, |
| "step": 2060, |
| "valid_targets_mean": 4940.0, |
| "valid_targets_min": 649 |
| }, |
| { |
| "epoch": 3.304, |
| "grad_norm": 0.5869431749420226, |
| "learning_rate": 1.2480713668244243e-05, |
| "loss": 0.4177, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41933634877204895, |
| "step": 2065, |
| "valid_targets_mean": 4774.3, |
| "valid_targets_min": 601 |
| }, |
| { |
| "epoch": 3.312, |
| "grad_norm": 0.5560654604661646, |
| "learning_rate": 1.2377307090481586e-05, |
| "loss": 0.4348, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40384188294410706, |
| "step": 2070, |
| "valid_targets_mean": 4333.8, |
| "valid_targets_min": 629 |
| }, |
| { |
| "epoch": 3.32, |
| "grad_norm": 0.5258484186346938, |
| "learning_rate": 1.2274138369981298e-05, |
| "loss": 0.4312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41006770730018616, |
| "step": 2075, |
| "valid_targets_mean": 5825.9, |
| "valid_targets_min": 816 |
| }, |
| { |
| "epoch": 3.328, |
| "grad_norm": 0.721798719675094, |
| "learning_rate": 1.2171210726003256e-05, |
| "loss": 0.4107, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4058680534362793, |
| "step": 2080, |
| "valid_targets_mean": 3225.6, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 3.336, |
| "grad_norm": 0.6048643181166588, |
| "learning_rate": 1.2068527370284815e-05, |
| "loss": 0.4163, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.45129895210266113, |
| "step": 2085, |
| "valid_targets_mean": 3978.1, |
| "valid_targets_min": 534 |
| }, |
| { |
| "epoch": 3.344, |
| "grad_norm": 0.6339488417554544, |
| "learning_rate": 1.1966091506940616e-05, |
| "loss": 0.4184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43796777725219727, |
| "step": 2090, |
| "valid_targets_mean": 3807.9, |
| "valid_targets_min": 753 |
| }, |
| { |
| "epoch": 3.352, |
| "grad_norm": 0.6016289665121101, |
| "learning_rate": 1.1863906332362569e-05, |
| "loss": 0.4001, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41488558053970337, |
| "step": 2095, |
| "valid_targets_mean": 3921.8, |
| "valid_targets_min": 687 |
| }, |
| { |
| "epoch": 3.36, |
| "grad_norm": 0.6737156145382773, |
| "learning_rate": 1.176197503512015e-05, |
| "loss": 0.4234, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46443474292755127, |
| "step": 2100, |
| "valid_targets_mean": 3559.7, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 3.368, |
| "grad_norm": 0.6983626223742545, |
| "learning_rate": 1.1660300795860877e-05, |
| "loss": 0.427, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.432370662689209, |
| "step": 2105, |
| "valid_targets_mean": 3061.8, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 3.376, |
| "grad_norm": 0.4979082402374389, |
| "learning_rate": 1.1558886787211071e-05, |
| "loss": 0.4313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3811950087547302, |
| "step": 2110, |
| "valid_targets_mean": 5262.1, |
| "valid_targets_min": 892 |
| }, |
| { |
| "epoch": 3.384, |
| "grad_norm": 0.5658616278719996, |
| "learning_rate": 1.1457736173676883e-05, |
| "loss": 0.384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39314180612564087, |
| "step": 2115, |
| "valid_targets_mean": 5609.1, |
| "valid_targets_min": 701 |
| }, |
| { |
| "epoch": 3.392, |
| "grad_norm": 0.5742144020806174, |
| "learning_rate": 1.1356852111545493e-05, |
| "loss": 0.4091, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3987872302532196, |
| "step": 2120, |
| "valid_targets_mean": 4422.2, |
| "valid_targets_min": 775 |
| }, |
| { |
| "epoch": 3.4, |
| "grad_norm": 0.5596475137152013, |
| "learning_rate": 1.1256237748786675e-05, |
| "loss": 0.3993, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3944924771785736, |
| "step": 2125, |
| "valid_targets_mean": 4539.4, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 3.408, |
| "grad_norm": 0.6913259682743764, |
| "learning_rate": 1.1155896224954543e-05, |
| "loss": 0.4201, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46018218994140625, |
| "step": 2130, |
| "valid_targets_mean": 3704.2, |
| "valid_targets_min": 719 |
| }, |
| { |
| "epoch": 3.416, |
| "grad_norm": 0.5521206700287924, |
| "learning_rate": 1.1055830671089578e-05, |
| "loss": 0.4269, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41626229882240295, |
| "step": 2135, |
| "valid_targets_mean": 5602.0, |
| "valid_targets_min": 737 |
| }, |
| { |
| "epoch": 3.424, |
| "grad_norm": 0.626791492401228, |
| "learning_rate": 1.0956044209620966e-05, |
| "loss": 0.4124, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.435993492603302, |
| "step": 2140, |
| "valid_targets_mean": 3991.2, |
| "valid_targets_min": 496 |
| }, |
| { |
| "epoch": 3.432, |
| "grad_norm": 0.476094171103533, |
| "learning_rate": 1.0856539954269121e-05, |
| "loss": 0.4253, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.377435564994812, |
| "step": 2145, |
| "valid_targets_mean": 5789.9, |
| "valid_targets_min": 884 |
| }, |
| { |
| "epoch": 3.44, |
| "grad_norm": 0.6128716789365134, |
| "learning_rate": 1.0757321009948543e-05, |
| "loss": 0.419, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40004780888557434, |
| "step": 2150, |
| "valid_targets_mean": 3649.6, |
| "valid_targets_min": 495 |
| }, |
| { |
| "epoch": 3.448, |
| "grad_norm": 0.5575841204894988, |
| "learning_rate": 1.0658390472670938e-05, |
| "loss": 0.3902, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4247322976589203, |
| "step": 2155, |
| "valid_targets_mean": 6032.9, |
| "valid_targets_min": 985 |
| }, |
| { |
| "epoch": 3.456, |
| "grad_norm": 0.6753891110732698, |
| "learning_rate": 1.0559751429448597e-05, |
| "loss": 0.42, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4023727774620056, |
| "step": 2160, |
| "valid_targets_mean": 4760.7, |
| "valid_targets_min": 647 |
| }, |
| { |
| "epoch": 3.464, |
| "grad_norm": 0.6369181074836571, |
| "learning_rate": 1.0461406958198101e-05, |
| "loss": 0.4353, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4552972912788391, |
| "step": 2165, |
| "valid_targets_mean": 4412.8, |
| "valid_targets_min": 600 |
| }, |
| { |
| "epoch": 3.472, |
| "grad_norm": 0.4819966107421562, |
| "learning_rate": 1.0363360127644235e-05, |
| "loss": 0.4066, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.394325315952301, |
| "step": 2170, |
| "valid_targets_mean": 7356.6, |
| "valid_targets_min": 622 |
| }, |
| { |
| "epoch": 3.48, |
| "grad_norm": 0.6128025340406501, |
| "learning_rate": 1.0265613997224255e-05, |
| "loss": 0.4125, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3975856900215149, |
| "step": 2175, |
| "valid_targets_mean": 3477.9, |
| "valid_targets_min": 605 |
| }, |
| { |
| "epoch": 3.488, |
| "grad_norm": 0.6384157096000654, |
| "learning_rate": 1.0168171616992422e-05, |
| "loss": 0.4358, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4252591133117676, |
| "step": 2180, |
| "valid_targets_mean": 4226.9, |
| "valid_targets_min": 498 |
| }, |
| { |
| "epoch": 3.496, |
| "grad_norm": 0.5767516682589051, |
| "learning_rate": 1.007103602752483e-05, |
| "loss": 0.4353, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4446565508842468, |
| "step": 2185, |
| "valid_targets_mean": 4517.8, |
| "valid_targets_min": 576 |
| }, |
| { |
| "epoch": 3.504, |
| "grad_norm": 0.5724798059794518, |
| "learning_rate": 9.974210259824505e-06, |
| "loss": 0.4143, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4005405604839325, |
| "step": 2190, |
| "valid_targets_mean": 4404.2, |
| "valid_targets_min": 678 |
| }, |
| { |
| "epoch": 3.512, |
| "grad_norm": 0.5733569057551399, |
| "learning_rate": 9.877697335226872e-06, |
| "loss": 0.4355, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4807167053222656, |
| "step": 2195, |
| "valid_targets_mean": 4826.6, |
| "valid_targets_min": 592 |
| }, |
| { |
| "epoch": 3.52, |
| "grad_norm": 1.1802646891503334, |
| "learning_rate": 9.781500265305448e-06, |
| "loss": 0.422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43163758516311646, |
| "step": 2200, |
| "valid_targets_mean": 3795.0, |
| "valid_targets_min": 705 |
| }, |
| { |
| "epoch": 3.528, |
| "grad_norm": 0.8845021260253212, |
| "learning_rate": 9.685622051777856e-06, |
| "loss": 0.4302, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46266594529151917, |
| "step": 2205, |
| "valid_targets_mean": 2079.0, |
| "valid_targets_min": 708 |
| }, |
| { |
| "epoch": 3.536, |
| "grad_norm": 0.5646534622070288, |
| "learning_rate": 9.590065686412182e-06, |
| "loss": 0.3905, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39034974575042725, |
| "step": 2210, |
| "valid_targets_mean": 4726.2, |
| "valid_targets_min": 376 |
| }, |
| { |
| "epoch": 3.544, |
| "grad_norm": 0.5508012766023374, |
| "learning_rate": 9.494834150933616e-06, |
| "loss": 0.412, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42684414982795715, |
| "step": 2215, |
| "valid_targets_mean": 5314.9, |
| "valid_targets_min": 552 |
| }, |
| { |
| "epoch": 3.552, |
| "grad_norm": 0.6966622976087259, |
| "learning_rate": 9.399930416931404e-06, |
| "loss": 0.4038, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38812291622161865, |
| "step": 2220, |
| "valid_targets_mean": 4849.1, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 3.56, |
| "grad_norm": 0.5793443000507197, |
| "learning_rate": 9.30535744576615e-06, |
| "loss": 0.4427, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4145427346229553, |
| "step": 2225, |
| "valid_targets_mean": 4751.8, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 3.568, |
| "grad_norm": 0.5440153611777978, |
| "learning_rate": 9.211118188477362e-06, |
| "loss": 0.4129, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42668813467025757, |
| "step": 2230, |
| "valid_targets_mean": 4886.1, |
| "valid_targets_min": 825 |
| }, |
| { |
| "epoch": 3.576, |
| "grad_norm": 0.6942559664808036, |
| "learning_rate": 9.117215585691408e-06, |
| "loss": 0.4318, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.408440500497818, |
| "step": 2235, |
| "valid_targets_mean": 2879.3, |
| "valid_targets_min": 575 |
| }, |
| { |
| "epoch": 3.584, |
| "grad_norm": 0.5283718370515997, |
| "learning_rate": 9.023652567529744e-06, |
| "loss": 0.4322, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39787888526916504, |
| "step": 2240, |
| "valid_targets_mean": 5380.6, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 3.592, |
| "grad_norm": 0.7189589849266416, |
| "learning_rate": 8.930432053517465e-06, |
| "loss": 0.4315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4863574802875519, |
| "step": 2245, |
| "valid_targets_mean": 3135.4, |
| "valid_targets_min": 598 |
| }, |
| { |
| "epoch": 3.6, |
| "grad_norm": 0.5688319197328309, |
| "learning_rate": 8.837556952492264e-06, |
| "loss": 0.427, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.443569153547287, |
| "step": 2250, |
| "valid_targets_mean": 4849.6, |
| "valid_targets_min": 491 |
| }, |
| { |
| "epoch": 3.608, |
| "grad_norm": 0.6372972828422979, |
| "learning_rate": 8.745030162513582e-06, |
| "loss": 0.4112, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43911778926849365, |
| "step": 2255, |
| "valid_targets_mean": 3977.4, |
| "valid_targets_min": 674 |
| }, |
| { |
| "epoch": 3.616, |
| "grad_norm": 0.646929389349214, |
| "learning_rate": 8.652854570772236e-06, |
| "loss": 0.4331, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.452898770570755, |
| "step": 2260, |
| "valid_targets_mean": 4541.3, |
| "valid_targets_min": 622 |
| }, |
| { |
| "epoch": 3.624, |
| "grad_norm": 0.536120974383535, |
| "learning_rate": 8.561033053500312e-06, |
| "loss": 0.4012, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39861825108528137, |
| "step": 2265, |
| "valid_targets_mean": 5880.2, |
| "valid_targets_min": 680 |
| }, |
| { |
| "epoch": 3.632, |
| "grad_norm": 0.5883785010261273, |
| "learning_rate": 8.46956847588141e-06, |
| "loss": 0.4105, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4199955463409424, |
| "step": 2270, |
| "valid_targets_mean": 4599.0, |
| "valid_targets_min": 569 |
| }, |
| { |
| "epoch": 3.64, |
| "grad_norm": 0.43011548249040454, |
| "learning_rate": 8.378463691961237e-06, |
| "loss": 0.4052, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3818497955799103, |
| "step": 2275, |
| "valid_targets_mean": 7853.8, |
| "valid_targets_min": 1082 |
| }, |
| { |
| "epoch": 3.648, |
| "grad_norm": 0.7813234320320073, |
| "learning_rate": 8.287721544558574e-06, |
| "loss": 0.423, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4120236039161682, |
| "step": 2280, |
| "valid_targets_mean": 2544.3, |
| "valid_targets_min": 777 |
| }, |
| { |
| "epoch": 3.656, |
| "grad_norm": 0.5267738721602246, |
| "learning_rate": 8.197344865176548e-06, |
| "loss": 0.4039, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3992099165916443, |
| "step": 2285, |
| "valid_targets_mean": 4824.1, |
| "valid_targets_min": 421 |
| }, |
| { |
| "epoch": 3.664, |
| "grad_norm": 0.5857399421262088, |
| "learning_rate": 8.10733647391427e-06, |
| "loss": 0.4192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3996756076812744, |
| "step": 2290, |
| "valid_targets_mean": 4680.4, |
| "valid_targets_min": 573 |
| }, |
| { |
| "epoch": 3.672, |
| "grad_norm": 0.6229880410561257, |
| "learning_rate": 8.017699179378849e-06, |
| "loss": 0.4102, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4350461959838867, |
| "step": 2295, |
| "valid_targets_mean": 4063.1, |
| "valid_targets_min": 535 |
| }, |
| { |
| "epoch": 3.68, |
| "grad_norm": 0.5935713053030977, |
| "learning_rate": 7.928435778597763e-06, |
| "loss": 0.4112, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4190603494644165, |
| "step": 2300, |
| "valid_targets_mean": 4264.1, |
| "valid_targets_min": 754 |
| }, |
| { |
| "epoch": 3.6879999999999997, |
| "grad_norm": 0.8421241937446357, |
| "learning_rate": 7.839549056931557e-06, |
| "loss": 0.4345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4648500084877014, |
| "step": 2305, |
| "valid_targets_mean": 2191.2, |
| "valid_targets_min": 812 |
| }, |
| { |
| "epoch": 3.6959999999999997, |
| "grad_norm": 0.627712707242927, |
| "learning_rate": 7.751041787986965e-06, |
| "loss": 0.4379, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43756556510925293, |
| "step": 2310, |
| "valid_targets_mean": 3952.6, |
| "valid_targets_min": 644 |
| }, |
| { |
| "epoch": 3.7039999999999997, |
| "grad_norm": 0.55692187150058, |
| "learning_rate": 7.662916733530317e-06, |
| "loss": 0.4214, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.410952627658844, |
| "step": 2315, |
| "valid_targets_mean": 5188.4, |
| "valid_targets_min": 1013 |
| }, |
| { |
| "epoch": 3.7119999999999997, |
| "grad_norm": 0.5841435249769223, |
| "learning_rate": 7.575176643401394e-06, |
| "loss": 0.3974, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39622873067855835, |
| "step": 2320, |
| "valid_targets_mean": 4232.2, |
| "valid_targets_min": 691 |
| }, |
| { |
| "epoch": 3.7199999999999998, |
| "grad_norm": 0.6334245660960619, |
| "learning_rate": 7.487824255427616e-06, |
| "loss": 0.4123, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43315714597702026, |
| "step": 2325, |
| "valid_targets_mean": 3952.1, |
| "valid_targets_min": 679 |
| }, |
| { |
| "epoch": 3.7279999999999998, |
| "grad_norm": 0.7125490272965322, |
| "learning_rate": 7.400862295338595e-06, |
| "loss": 0.4025, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3960513770580292, |
| "step": 2330, |
| "valid_targets_mean": 2979.1, |
| "valid_targets_min": 568 |
| }, |
| { |
| "epoch": 3.7359999999999998, |
| "grad_norm": 0.7047871741832065, |
| "learning_rate": 7.314293476681122e-06, |
| "loss": 0.4278, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4047684669494629, |
| "step": 2335, |
| "valid_targets_mean": 2885.2, |
| "valid_targets_min": 769 |
| }, |
| { |
| "epoch": 3.7439999999999998, |
| "grad_norm": 0.5235697909405775, |
| "learning_rate": 7.228120500734443e-06, |
| "loss": 0.4066, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39039039611816406, |
| "step": 2340, |
| "valid_targets_mean": 5031.9, |
| "valid_targets_min": 693 |
| }, |
| { |
| "epoch": 3.752, |
| "grad_norm": 0.6309610439555812, |
| "learning_rate": 7.1423460564259995e-06, |
| "loss": 0.425, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41119787096977234, |
| "step": 2345, |
| "valid_targets_mean": 3696.8, |
| "valid_targets_min": 615 |
| }, |
| { |
| "epoch": 3.76, |
| "grad_norm": 0.6046248254536285, |
| "learning_rate": 7.056972820247516e-06, |
| "loss": 0.4312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44950827956199646, |
| "step": 2350, |
| "valid_targets_mean": 4173.8, |
| "valid_targets_min": 468 |
| }, |
| { |
| "epoch": 3.768, |
| "grad_norm": 0.5813802818171795, |
| "learning_rate": 6.97200345617149e-06, |
| "loss": 0.4251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4253307580947876, |
| "step": 2355, |
| "valid_targets_mean": 4188.0, |
| "valid_targets_min": 906 |
| }, |
| { |
| "epoch": 3.776, |
| "grad_norm": 0.6480857251816284, |
| "learning_rate": 6.887440615568044e-06, |
| "loss": 0.4063, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3995557427406311, |
| "step": 2360, |
| "valid_targets_mean": 3368.9, |
| "valid_targets_min": 696 |
| }, |
| { |
| "epoch": 3.784, |
| "grad_norm": 0.5978641746105108, |
| "learning_rate": 6.803286937122233e-06, |
| "loss": 0.4212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4335462749004364, |
| "step": 2365, |
| "valid_targets_mean": 4294.8, |
| "valid_targets_min": 717 |
| }, |
| { |
| "epoch": 3.792, |
| "grad_norm": 0.705354339864555, |
| "learning_rate": 6.719545046751674e-06, |
| "loss": 0.4492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4486519694328308, |
| "step": 2370, |
| "valid_targets_mean": 3097.3, |
| "valid_targets_min": 667 |
| }, |
| { |
| "epoch": 3.8, |
| "grad_norm": 0.6447197729667318, |
| "learning_rate": 6.636217557524605e-06, |
| "loss": 0.4089, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4317939281463623, |
| "step": 2375, |
| "valid_targets_mean": 3642.6, |
| "valid_targets_min": 738 |
| }, |
| { |
| "epoch": 3.808, |
| "grad_norm": 0.5784840504063279, |
| "learning_rate": 6.55330706957837e-06, |
| "loss": 0.4183, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.417401522397995, |
| "step": 2380, |
| "valid_targets_mean": 4374.6, |
| "valid_targets_min": 1143 |
| }, |
| { |
| "epoch": 3.816, |
| "grad_norm": 0.5691858028003152, |
| "learning_rate": 6.4708161700382655e-06, |
| "loss": 0.4185, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4285266101360321, |
| "step": 2385, |
| "valid_targets_mean": 4893.5, |
| "valid_targets_min": 527 |
| }, |
| { |
| "epoch": 3.824, |
| "grad_norm": 0.6509918903676563, |
| "learning_rate": 6.388747432936819e-06, |
| "loss": 0.3843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39279234409332275, |
| "step": 2390, |
| "valid_targets_mean": 3421.9, |
| "valid_targets_min": 766 |
| }, |
| { |
| "epoch": 3.832, |
| "grad_norm": 0.5477195713050954, |
| "learning_rate": 6.3071034191334915e-06, |
| "loss": 0.3963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4045974612236023, |
| "step": 2395, |
| "valid_targets_mean": 4609.2, |
| "valid_targets_min": 868 |
| }, |
| { |
| "epoch": 3.84, |
| "grad_norm": 0.6280025691309644, |
| "learning_rate": 6.22588667623472e-06, |
| "loss": 0.4238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4826744794845581, |
| "step": 2400, |
| "valid_targets_mean": 4032.0, |
| "valid_targets_min": 473 |
| }, |
| { |
| "epoch": 3.848, |
| "grad_norm": 0.7680355117957194, |
| "learning_rate": 6.145099738514466e-06, |
| "loss": 0.4578, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4926620125770569, |
| "step": 2405, |
| "valid_targets_mean": 2725.4, |
| "valid_targets_min": 615 |
| }, |
| { |
| "epoch": 3.856, |
| "grad_norm": 0.5711966605379848, |
| "learning_rate": 6.064745126835112e-06, |
| "loss": 0.4111, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4238035976886749, |
| "step": 2410, |
| "valid_targets_mean": 4564.0, |
| "valid_targets_min": 679 |
| }, |
| { |
| "epoch": 3.864, |
| "grad_norm": 0.5326749978154176, |
| "learning_rate": 5.984825348568812e-06, |
| "loss": 0.4065, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4256637394428253, |
| "step": 2415, |
| "valid_targets_mean": 5087.4, |
| "valid_targets_min": 659 |
| }, |
| { |
| "epoch": 3.872, |
| "grad_norm": 0.7369604952522139, |
| "learning_rate": 5.905342897519262e-06, |
| "loss": 0.405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47101157903671265, |
| "step": 2420, |
| "valid_targets_mean": 2868.4, |
| "valid_targets_min": 558 |
| }, |
| { |
| "epoch": 3.88, |
| "grad_norm": 0.7010288266967857, |
| "learning_rate": 5.826300253843851e-06, |
| "loss": 0.421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3836793005466461, |
| "step": 2425, |
| "valid_targets_mean": 2784.3, |
| "valid_targets_min": 575 |
| }, |
| { |
| "epoch": 3.888, |
| "grad_norm": 0.6824318371192227, |
| "learning_rate": 5.7476998839763035e-06, |
| "loss": 0.4208, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4819521903991699, |
| "step": 2430, |
| "valid_targets_mean": 3219.2, |
| "valid_targets_min": 660 |
| }, |
| { |
| "epoch": 3.896, |
| "grad_norm": 0.5989642635294622, |
| "learning_rate": 5.669544240549698e-06, |
| "loss": 0.3991, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40206092596054077, |
| "step": 2435, |
| "valid_targets_mean": 3922.7, |
| "valid_targets_min": 749 |
| }, |
| { |
| "epoch": 3.904, |
| "grad_norm": 0.5819637834251199, |
| "learning_rate": 5.591835762319946e-06, |
| "loss": 0.43, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4356538653373718, |
| "step": 2440, |
| "valid_targets_mean": 4358.2, |
| "valid_targets_min": 989 |
| }, |
| { |
| "epoch": 3.912, |
| "grad_norm": 0.8093003590748824, |
| "learning_rate": 5.514576874089683e-06, |
| "loss": 0.4171, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4149336814880371, |
| "step": 2445, |
| "valid_targets_mean": 3634.3, |
| "valid_targets_min": 761 |
| }, |
| { |
| "epoch": 3.92, |
| "grad_norm": 0.5637333714042546, |
| "learning_rate": 5.437769986632622e-06, |
| "loss": 0.4192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40535950660705566, |
| "step": 2450, |
| "valid_targets_mean": 5450.6, |
| "valid_targets_min": 842 |
| }, |
| { |
| "epoch": 3.928, |
| "grad_norm": 0.5650303402408413, |
| "learning_rate": 5.361417496618315e-06, |
| "loss": 0.4098, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4089083969593048, |
| "step": 2455, |
| "valid_targets_mean": 4491.1, |
| "valid_targets_min": 402 |
| }, |
| { |
| "epoch": 3.936, |
| "grad_norm": 0.5862251884278199, |
| "learning_rate": 5.285521786537368e-06, |
| "loss": 0.4296, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.421395480632782, |
| "step": 2460, |
| "valid_targets_mean": 4045.0, |
| "valid_targets_min": 499 |
| }, |
| { |
| "epoch": 3.944, |
| "grad_norm": 0.5911774172256168, |
| "learning_rate": 5.2100852246270975e-06, |
| "loss": 0.4443, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41850167512893677, |
| "step": 2465, |
| "valid_targets_mean": 4103.7, |
| "valid_targets_min": 1182 |
| }, |
| { |
| "epoch": 3.952, |
| "grad_norm": 0.5488515492585119, |
| "learning_rate": 5.135110164797637e-06, |
| "loss": 0.4105, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4004397988319397, |
| "step": 2470, |
| "valid_targets_mean": 4585.8, |
| "valid_targets_min": 741 |
| }, |
| { |
| "epoch": 3.96, |
| "grad_norm": 0.638564502319262, |
| "learning_rate": 5.060598946558484e-06, |
| "loss": 0.4295, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43971341848373413, |
| "step": 2475, |
| "valid_targets_mean": 3582.0, |
| "valid_targets_min": 847 |
| }, |
| { |
| "epoch": 3.968, |
| "grad_norm": 0.5896713751876114, |
| "learning_rate": 4.986553894945512e-06, |
| "loss": 0.3939, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3950134515762329, |
| "step": 2480, |
| "valid_targets_mean": 3773.5, |
| "valid_targets_min": 801 |
| }, |
| { |
| "epoch": 3.976, |
| "grad_norm": 0.7928195363821494, |
| "learning_rate": 4.912977320448391e-06, |
| "loss": 0.4187, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3833604156970978, |
| "step": 2485, |
| "valid_targets_mean": 3785.5, |
| "valid_targets_min": 473 |
| }, |
| { |
| "epoch": 3.984, |
| "grad_norm": 0.59139939976017, |
| "learning_rate": 4.839871518938513e-06, |
| "loss": 0.4383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4159795343875885, |
| "step": 2490, |
| "valid_targets_mean": 4217.4, |
| "valid_targets_min": 663 |
| }, |
| { |
| "epoch": 3.992, |
| "grad_norm": 0.550470299025946, |
| "learning_rate": 4.767238771597347e-06, |
| "loss": 0.3958, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3765171766281128, |
| "step": 2495, |
| "valid_targets_mean": 5407.9, |
| "valid_targets_min": 1275 |
| }, |
| { |
| "epoch": 4.0, |
| "grad_norm": 0.6285536655979354, |
| "learning_rate": 4.695081344845254e-06, |
| "loss": 0.4467, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4390712380409241, |
| "step": 2500, |
| "valid_targets_mean": 3904.6, |
| "valid_targets_min": 767 |
| }, |
| { |
| "epoch": 4.008, |
| "grad_norm": 0.6426613600523003, |
| "learning_rate": 4.623401490270778e-06, |
| "loss": 0.4065, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4202343821525574, |
| "step": 2505, |
| "valid_targets_mean": 3585.8, |
| "valid_targets_min": 631 |
| }, |
| { |
| "epoch": 4.016, |
| "grad_norm": 0.5585933431067611, |
| "learning_rate": 4.552201444560373e-06, |
| "loss": 0.4059, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4154893457889557, |
| "step": 2510, |
| "valid_targets_mean": 4587.4, |
| "valid_targets_min": 834 |
| }, |
| { |
| "epoch": 4.024, |
| "grad_norm": 0.5849602135579992, |
| "learning_rate": 4.481483429428615e-06, |
| "loss": 0.4119, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3984353244304657, |
| "step": 2515, |
| "valid_targets_mean": 4275.1, |
| "valid_targets_min": 657 |
| }, |
| { |
| "epoch": 4.032, |
| "grad_norm": 0.5834357553067363, |
| "learning_rate": 4.4112496515488765e-06, |
| "loss": 0.4011, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3702242970466614, |
| "step": 2520, |
| "valid_targets_mean": 4599.9, |
| "valid_targets_min": 608 |
| }, |
| { |
| "epoch": 4.04, |
| "grad_norm": 0.7637639117096291, |
| "learning_rate": 4.341502302484472e-06, |
| "loss": 0.425, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4776972532272339, |
| "step": 2525, |
| "valid_targets_mean": 2890.7, |
| "valid_targets_min": 695 |
| }, |
| { |
| "epoch": 4.048, |
| "grad_norm": 0.7468745784687151, |
| "learning_rate": 4.272243558620264e-06, |
| "loss": 0.4098, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4523777961730957, |
| "step": 2530, |
| "valid_targets_mean": 2911.0, |
| "valid_targets_min": 333 |
| }, |
| { |
| "epoch": 4.056, |
| "grad_norm": 0.5802649714021575, |
| "learning_rate": 4.203475581094771e-06, |
| "loss": 0.4373, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.425601601600647, |
| "step": 2535, |
| "valid_targets_mean": 5046.2, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 4.064, |
| "grad_norm": 0.5447288932891161, |
| "learning_rate": 4.135200515732716e-06, |
| "loss": 0.4005, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3529777526855469, |
| "step": 2540, |
| "valid_targets_mean": 4301.4, |
| "valid_targets_min": 653 |
| }, |
| { |
| "epoch": 4.072, |
| "grad_norm": 0.7333561005546733, |
| "learning_rate": 4.067420492978065e-06, |
| "loss": 0.406, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4183364808559418, |
| "step": 2545, |
| "valid_targets_mean": 2979.5, |
| "valid_targets_min": 433 |
| }, |
| { |
| "epoch": 4.08, |
| "grad_norm": 0.8448397467709902, |
| "learning_rate": 4.000137627827554e-06, |
| "loss": 0.411, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4083864092826843, |
| "step": 2550, |
| "valid_targets_mean": 1975.8, |
| "valid_targets_min": 737 |
| }, |
| { |
| "epoch": 4.088, |
| "grad_norm": 0.7039063639446964, |
| "learning_rate": 3.9333540197647035e-06, |
| "loss": 0.3973, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42661112546920776, |
| "step": 2555, |
| "valid_targets_mean": 2891.6, |
| "valid_targets_min": 744 |
| }, |
| { |
| "epoch": 4.096, |
| "grad_norm": 0.6103368310567368, |
| "learning_rate": 3.867071752694282e-06, |
| "loss": 0.4136, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39338967204093933, |
| "step": 2560, |
| "valid_targets_mean": 4202.1, |
| "valid_targets_min": 619 |
| }, |
| { |
| "epoch": 4.104, |
| "grad_norm": 0.5609255390428625, |
| "learning_rate": 3.8012928948773243e-06, |
| "loss": 0.4109, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3590124249458313, |
| "step": 2565, |
| "valid_targets_mean": 4673.2, |
| "valid_targets_min": 534 |
| }, |
| { |
| "epoch": 4.112, |
| "grad_norm": 0.7163062681286491, |
| "learning_rate": 3.7360194988665364e-06, |
| "loss": 0.4282, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4255579113960266, |
| "step": 2570, |
| "valid_targets_mean": 2981.8, |
| "valid_targets_min": 753 |
| }, |
| { |
| "epoch": 4.12, |
| "grad_norm": 0.4784051484306757, |
| "learning_rate": 3.6712536014422885e-06, |
| "loss": 0.405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.350371778011322, |
| "step": 2575, |
| "valid_targets_mean": 6456.9, |
| "valid_targets_min": 503 |
| }, |
| { |
| "epoch": 4.128, |
| "grad_norm": 0.5438453496849911, |
| "learning_rate": 3.606997223549049e-06, |
| "loss": 0.4131, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4073692262172699, |
| "step": 2580, |
| "valid_targets_mean": 5279.0, |
| "valid_targets_min": 801 |
| }, |
| { |
| "epoch": 4.136, |
| "grad_norm": 0.5879969117381249, |
| "learning_rate": 3.543252370232313e-06, |
| "loss": 0.3969, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4108046889305115, |
| "step": 2585, |
| "valid_targets_mean": 4720.8, |
| "valid_targets_min": 959 |
| }, |
| { |
| "epoch": 4.144, |
| "grad_norm": 0.5532047673911502, |
| "learning_rate": 3.4800210305760662e-06, |
| "loss": 0.4201, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36657533049583435, |
| "step": 2590, |
| "valid_targets_mean": 4700.9, |
| "valid_targets_min": 601 |
| }, |
| { |
| "epoch": 4.152, |
| "grad_norm": 0.5901167864484412, |
| "learning_rate": 3.4173051776406817e-06, |
| "loss": 0.4184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4340725541114807, |
| "step": 2595, |
| "valid_targets_mean": 4326.9, |
| "valid_targets_min": 523 |
| }, |
| { |
| "epoch": 4.16, |
| "grad_norm": 0.5475674178491631, |
| "learning_rate": 3.3551067684013706e-06, |
| "loss": 0.4084, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3945809006690979, |
| "step": 2600, |
| "valid_targets_mean": 4706.6, |
| "valid_targets_min": 851 |
| }, |
| { |
| "epoch": 4.168, |
| "grad_norm": 0.5801898131217752, |
| "learning_rate": 3.2934277436871187e-06, |
| "loss": 0.3873, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3601340055465698, |
| "step": 2605, |
| "valid_targets_mean": 4189.4, |
| "valid_targets_min": 534 |
| }, |
| { |
| "epoch": 4.176, |
| "grad_norm": 0.7299558918948404, |
| "learning_rate": 3.232270028120121e-06, |
| "loss": 0.405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4364742338657379, |
| "step": 2610, |
| "valid_targets_mean": 2916.7, |
| "valid_targets_min": 703 |
| }, |
| { |
| "epoch": 4.184, |
| "grad_norm": 0.5996399283446768, |
| "learning_rate": 3.1716355300557256e-06, |
| "loss": 0.3937, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4116308391094208, |
| "step": 2615, |
| "valid_targets_mean": 4109.5, |
| "valid_targets_min": 600 |
| }, |
| { |
| "epoch": 4.192, |
| "grad_norm": 0.5730701668848656, |
| "learning_rate": 3.111526141522896e-06, |
| "loss": 0.4212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41376417875289917, |
| "step": 2620, |
| "valid_targets_mean": 4810.4, |
| "valid_targets_min": 691 |
| }, |
| { |
| "epoch": 4.2, |
| "grad_norm": 0.5519404941783179, |
| "learning_rate": 3.0519437381651507e-06, |
| "loss": 0.4072, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40336206555366516, |
| "step": 2625, |
| "valid_targets_mean": 4866.9, |
| "valid_targets_min": 749 |
| }, |
| { |
| "epoch": 4.208, |
| "grad_norm": 0.5067010081942247, |
| "learning_rate": 2.992890179182062e-06, |
| "loss": 0.3919, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35615670680999756, |
| "step": 2630, |
| "valid_targets_mean": 5492.4, |
| "valid_targets_min": 631 |
| }, |
| { |
| "epoch": 4.216, |
| "grad_norm": 0.6359749406389089, |
| "learning_rate": 2.93436730727122e-06, |
| "loss": 0.4067, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.402889609336853, |
| "step": 2635, |
| "valid_targets_mean": 3929.4, |
| "valid_targets_min": 912 |
| }, |
| { |
| "epoch": 4.224, |
| "grad_norm": 0.6145949359641839, |
| "learning_rate": 2.8763769485707447e-06, |
| "loss": 0.3904, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40841296315193176, |
| "step": 2640, |
| "valid_targets_mean": 4276.6, |
| "valid_targets_min": 799 |
| }, |
| { |
| "epoch": 4.232, |
| "grad_norm": 0.7110071687256778, |
| "learning_rate": 2.818920912602294e-06, |
| "loss": 0.405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4124786853790283, |
| "step": 2645, |
| "valid_targets_mean": 2939.0, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 4.24, |
| "grad_norm": 0.5743817768129006, |
| "learning_rate": 2.762000992214626e-06, |
| "loss": 0.4164, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3974683880805969, |
| "step": 2650, |
| "valid_targets_mean": 4447.1, |
| "valid_targets_min": 511 |
| }, |
| { |
| "epoch": 4.248, |
| "grad_norm": 0.5413226643387624, |
| "learning_rate": 2.7056189635276162e-06, |
| "loss": 0.3864, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3866695463657379, |
| "step": 2655, |
| "valid_targets_mean": 5034.9, |
| "valid_targets_min": 677 |
| }, |
| { |
| "epoch": 4.256, |
| "grad_norm": 0.5726347506605515, |
| "learning_rate": 2.6497765858768643e-06, |
| "loss": 0.4024, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42359739542007446, |
| "step": 2660, |
| "valid_targets_mean": 4637.1, |
| "valid_targets_min": 635 |
| }, |
| { |
| "epoch": 4.264, |
| "grad_norm": 0.4855797229728632, |
| "learning_rate": 2.594475601758786e-06, |
| "loss": 0.3905, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3630528748035431, |
| "step": 2665, |
| "valid_targets_mean": 5871.8, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 4.272, |
| "grad_norm": 0.6421484551855371, |
| "learning_rate": 2.539717736776237e-06, |
| "loss": 0.3901, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42318665981292725, |
| "step": 2670, |
| "valid_targets_mean": 4278.5, |
| "valid_targets_min": 624 |
| }, |
| { |
| "epoch": 4.28, |
| "grad_norm": 0.6235765181112626, |
| "learning_rate": 2.4855046995846844e-06, |
| "loss": 0.3848, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4206055700778961, |
| "step": 2675, |
| "valid_targets_mean": 4279.4, |
| "valid_targets_min": 484 |
| }, |
| { |
| "epoch": 4.288, |
| "grad_norm": 0.5687422916455531, |
| "learning_rate": 2.431838181838868e-06, |
| "loss": 0.4003, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38011544942855835, |
| "step": 2680, |
| "valid_targets_mean": 4600.6, |
| "valid_targets_min": 931 |
| }, |
| { |
| "epoch": 4.296, |
| "grad_norm": 0.5597281309268304, |
| "learning_rate": 2.3787198581400285e-06, |
| "loss": 0.409, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3837645649909973, |
| "step": 2685, |
| "valid_targets_mean": 4807.3, |
| "valid_targets_min": 852 |
| }, |
| { |
| "epoch": 4.304, |
| "grad_norm": 0.5208997470892812, |
| "learning_rate": 2.3261513859836437e-06, |
| "loss": 0.3891, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36763709783554077, |
| "step": 2690, |
| "valid_targets_mean": 5476.7, |
| "valid_targets_min": 713 |
| }, |
| { |
| "epoch": 4.312, |
| "grad_norm": 0.729596593306008, |
| "learning_rate": 2.27413440570772e-06, |
| "loss": 0.4192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47866132855415344, |
| "step": 2695, |
| "valid_targets_mean": 3093.9, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 4.32, |
| "grad_norm": 0.5963682532410031, |
| "learning_rate": 2.222670540441596e-06, |
| "loss": 0.3944, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35257434844970703, |
| "step": 2700, |
| "valid_targets_mean": 5420.8, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 4.328, |
| "grad_norm": 0.5229520235670004, |
| "learning_rate": 2.17176139605531e-06, |
| "loss": 0.4251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3816877603530884, |
| "step": 2705, |
| "valid_targets_mean": 5403.9, |
| "valid_targets_min": 725 |
| }, |
| { |
| "epoch": 4.336, |
| "grad_norm": 0.6555810725268437, |
| "learning_rate": 2.121408561109466e-06, |
| "loss": 0.4173, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38497596979141235, |
| "step": 2710, |
| "valid_targets_mean": 3670.6, |
| "valid_targets_min": 734 |
| }, |
| { |
| "epoch": 4.344, |
| "grad_norm": 0.5360008671439541, |
| "learning_rate": 2.071613606805696e-06, |
| "loss": 0.4179, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4150103032588959, |
| "step": 2715, |
| "valid_targets_mean": 5754.9, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 4.352, |
| "grad_norm": 0.5212551398201117, |
| "learning_rate": 2.0223780869376018e-06, |
| "loss": 0.394, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40325045585632324, |
| "step": 2720, |
| "valid_targets_mean": 5807.8, |
| "valid_targets_min": 691 |
| }, |
| { |
| "epoch": 4.36, |
| "grad_norm": 0.501570979642165, |
| "learning_rate": 1.9737035378422907e-06, |
| "loss": 0.3989, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38142967224121094, |
| "step": 2725, |
| "valid_targets_mean": 5602.2, |
| "valid_targets_min": 696 |
| }, |
| { |
| "epoch": 4.368, |
| "grad_norm": 0.6586918366767852, |
| "learning_rate": 1.925591478352424e-06, |
| "loss": 0.3876, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3896998167037964, |
| "step": 2730, |
| "valid_targets_mean": 3427.3, |
| "valid_targets_min": 671 |
| }, |
| { |
| "epoch": 4.376, |
| "grad_norm": 0.5558539980698498, |
| "learning_rate": 1.8780434097488443e-06, |
| "loss": 0.4082, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4040504992008209, |
| "step": 2735, |
| "valid_targets_mean": 4714.9, |
| "valid_targets_min": 695 |
| }, |
| { |
| "epoch": 4.384, |
| "grad_norm": 0.7867324950433691, |
| "learning_rate": 1.831060815713699e-06, |
| "loss": 0.4048, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4829801917076111, |
| "step": 2740, |
| "valid_targets_mean": 2988.8, |
| "valid_targets_min": 288 |
| }, |
| { |
| "epoch": 4.392, |
| "grad_norm": 0.6214041417079301, |
| "learning_rate": 1.7846451622841643e-06, |
| "loss": 0.4194, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3955761194229126, |
| "step": 2745, |
| "valid_targets_mean": 3676.2, |
| "valid_targets_min": 489 |
| }, |
| { |
| "epoch": 4.4, |
| "grad_norm": 0.5561873136372868, |
| "learning_rate": 1.7387978978066988e-06, |
| "loss": 0.4089, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3890697956085205, |
| "step": 2750, |
| "valid_targets_mean": 5758.4, |
| "valid_targets_min": 882 |
| }, |
| { |
| "epoch": 4.408, |
| "grad_norm": 0.560721929365539, |
| "learning_rate": 1.6935204528918347e-06, |
| "loss": 0.3973, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40641331672668457, |
| "step": 2755, |
| "valid_targets_mean": 5017.4, |
| "valid_targets_min": 521 |
| }, |
| { |
| "epoch": 4.416, |
| "grad_norm": 0.6889136427973638, |
| "learning_rate": 1.6488142403695651e-06, |
| "loss": 0.4157, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4421613812446594, |
| "step": 2760, |
| "valid_targets_mean": 3295.3, |
| "valid_targets_min": 709 |
| }, |
| { |
| "epoch": 4.424, |
| "grad_norm": 0.7902577945988017, |
| "learning_rate": 1.6046806552452254e-06, |
| "loss": 0.4283, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4924136698246002, |
| "step": 2765, |
| "valid_targets_mean": 3308.9, |
| "valid_targets_min": 751 |
| }, |
| { |
| "epoch": 4.432, |
| "grad_norm": 0.6844284345573052, |
| "learning_rate": 1.5611210746559868e-06, |
| "loss": 0.3938, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40334612131118774, |
| "step": 2770, |
| "valid_targets_mean": 3122.0, |
| "valid_targets_min": 583 |
| }, |
| { |
| "epoch": 4.44, |
| "grad_norm": 0.6162273196275045, |
| "learning_rate": 1.5181368578278744e-06, |
| "loss": 0.4041, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35837671160697937, |
| "step": 2775, |
| "valid_targets_mean": 3665.4, |
| "valid_targets_min": 718 |
| }, |
| { |
| "epoch": 4.448, |
| "grad_norm": 0.5974789854549655, |
| "learning_rate": 1.4757293460333566e-06, |
| "loss": 0.4308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4109255075454712, |
| "step": 2780, |
| "valid_targets_mean": 4179.9, |
| "valid_targets_min": 793 |
| }, |
| { |
| "epoch": 4.456, |
| "grad_norm": 0.5708600383249177, |
| "learning_rate": 1.4338998625494905e-06, |
| "loss": 0.4075, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3754119277000427, |
| "step": 2785, |
| "valid_targets_mean": 4436.0, |
| "valid_targets_min": 773 |
| }, |
| { |
| "epoch": 4.464, |
| "grad_norm": 0.6313362954432985, |
| "learning_rate": 1.3926497126166405e-06, |
| "loss": 0.4204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41806894540786743, |
| "step": 2790, |
| "valid_targets_mean": 4132.2, |
| "valid_targets_min": 624 |
| }, |
| { |
| "epoch": 4.4719999999999995, |
| "grad_norm": 0.5994471438843727, |
| "learning_rate": 1.3519801833977298e-06, |
| "loss": 0.4271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43274080753326416, |
| "step": 2795, |
| "valid_targets_mean": 4554.3, |
| "valid_targets_min": 371 |
| }, |
| { |
| "epoch": 4.48, |
| "grad_norm": 0.5410383078069514, |
| "learning_rate": 1.3118925439381003e-06, |
| "loss": 0.4073, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35545891523361206, |
| "step": 2800, |
| "valid_targets_mean": 5008.1, |
| "valid_targets_min": 563 |
| }, |
| { |
| "epoch": 4.4879999999999995, |
| "grad_norm": 0.7142880244968844, |
| "learning_rate": 1.2723880451258918e-06, |
| "loss": 0.4072, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4319564998149872, |
| "step": 2805, |
| "valid_targets_mean": 3327.8, |
| "valid_targets_min": 717 |
| }, |
| { |
| "epoch": 4.496, |
| "grad_norm": 0.5901784201564468, |
| "learning_rate": 1.2334679196530219e-06, |
| "loss": 0.4332, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42878252267837524, |
| "step": 2810, |
| "valid_targets_mean": 4661.2, |
| "valid_targets_min": 669 |
| }, |
| { |
| "epoch": 4.504, |
| "grad_norm": 0.5840631068232639, |
| "learning_rate": 1.1951333819767163e-06, |
| "loss": 0.4096, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.392315149307251, |
| "step": 2815, |
| "valid_targets_mean": 4338.4, |
| "valid_targets_min": 823 |
| }, |
| { |
| "epoch": 4.5120000000000005, |
| "grad_norm": 0.9670608326476431, |
| "learning_rate": 1.157385628281622e-06, |
| "loss": 0.4121, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.409821093082428, |
| "step": 2820, |
| "valid_targets_mean": 2009.2, |
| "valid_targets_min": 725 |
| }, |
| { |
| "epoch": 4.52, |
| "grad_norm": 0.6605390936042161, |
| "learning_rate": 1.1202258364424633e-06, |
| "loss": 0.4025, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41265183687210083, |
| "step": 2825, |
| "valid_targets_mean": 3653.6, |
| "valid_targets_min": 676 |
| }, |
| { |
| "epoch": 4.5280000000000005, |
| "grad_norm": 0.512314838581097, |
| "learning_rate": 1.0836551659873073e-06, |
| "loss": 0.4052, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3962034583091736, |
| "step": 2830, |
| "valid_targets_mean": 5787.9, |
| "valid_targets_min": 930 |
| }, |
| { |
| "epoch": 4.536, |
| "grad_norm": 0.5526075598864743, |
| "learning_rate": 1.0476747580613723e-06, |
| "loss": 0.4239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37799936532974243, |
| "step": 2835, |
| "valid_targets_mean": 4742.4, |
| "valid_targets_min": 616 |
| }, |
| { |
| "epoch": 4.5440000000000005, |
| "grad_norm": 0.5423110924774888, |
| "learning_rate": 1.012285735391416e-06, |
| "loss": 0.4318, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41332849860191345, |
| "step": 2840, |
| "valid_targets_mean": 5307.6, |
| "valid_targets_min": 598 |
| }, |
| { |
| "epoch": 4.552, |
| "grad_norm": 0.611737571874644, |
| "learning_rate": 9.774892022507166e-07, |
| "loss": 0.4062, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3971940875053406, |
| "step": 2845, |
| "valid_targets_mean": 4247.7, |
| "valid_targets_min": 771 |
| }, |
| { |
| "epoch": 4.5600000000000005, |
| "grad_norm": 0.7135757278290046, |
| "learning_rate": 9.432862444245994e-07, |
| "loss": 0.4359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41892296075820923, |
| "step": 2850, |
| "valid_targets_mean": 3975.4, |
| "valid_targets_min": 789 |
| }, |
| { |
| "epoch": 4.568, |
| "grad_norm": 0.7473593705860603, |
| "learning_rate": 9.096779291765667e-07, |
| "loss": 0.4205, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4587504267692566, |
| "step": 2855, |
| "valid_targets_mean": 2893.6, |
| "valid_targets_min": 813 |
| }, |
| { |
| "epoch": 4.576, |
| "grad_norm": 0.6873974964018028, |
| "learning_rate": 8.766653052149831e-07, |
| "loss": 0.4326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44257092475891113, |
| "step": 2860, |
| "valid_targets_mean": 3811.4, |
| "valid_targets_min": 924 |
| }, |
| { |
| "epoch": 4.584, |
| "grad_norm": 0.7683330781359213, |
| "learning_rate": 8.442494026603709e-07, |
| "loss": 0.4131, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44089746475219727, |
| "step": 2865, |
| "valid_targets_mean": 3014.2, |
| "valid_targets_min": 687 |
| }, |
| { |
| "epoch": 4.592, |
| "grad_norm": 0.5846368499398767, |
| "learning_rate": 8.124312330132423e-07, |
| "loss": 0.3811, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3351588845252991, |
| "step": 2870, |
| "valid_targets_mean": 3851.7, |
| "valid_targets_min": 677 |
| }, |
| { |
| "epoch": 4.6, |
| "grad_norm": 0.5968786949619281, |
| "learning_rate": 7.812117891225667e-07, |
| "loss": 0.405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3558521866798401, |
| "step": 2875, |
| "valid_targets_mean": 3713.2, |
| "valid_targets_min": 783 |
| }, |
| { |
| "epoch": 4.608, |
| "grad_norm": 0.6212588253374268, |
| "learning_rate": 7.505920451547544e-07, |
| "loss": 0.4055, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3719406723976135, |
| "step": 2880, |
| "valid_targets_mean": 3888.3, |
| "valid_targets_min": 516 |
| }, |
| { |
| "epoch": 4.616, |
| "grad_norm": 0.8263222292237222, |
| "learning_rate": 7.205729565632947e-07, |
| "loss": 0.4107, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.415608674287796, |
| "step": 2885, |
| "valid_targets_mean": 2358.3, |
| "valid_targets_min": 738 |
| }, |
| { |
| "epoch": 4.624, |
| "grad_norm": 1.2742274611700906, |
| "learning_rate": 6.911554600589121e-07, |
| "loss": 0.4135, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3934563398361206, |
| "step": 2890, |
| "valid_targets_mean": 4383.9, |
| "valid_targets_min": 621 |
| }, |
| { |
| "epoch": 4.632, |
| "grad_norm": 0.7165845833648357, |
| "learning_rate": 6.62340473580354e-07, |
| "loss": 0.422, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46797922253608704, |
| "step": 2895, |
| "valid_targets_mean": 3301.2, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 4.64, |
| "grad_norm": 0.5652440899945861, |
| "learning_rate": 6.341288962657422e-07, |
| "loss": 0.4163, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4029410779476166, |
| "step": 2900, |
| "valid_targets_mean": 4548.8, |
| "valid_targets_min": 447 |
| }, |
| { |
| "epoch": 4.648, |
| "grad_norm": 0.6700159540879693, |
| "learning_rate": 6.06521608424524e-07, |
| "loss": 0.4329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4570663273334503, |
| "step": 2905, |
| "valid_targets_mean": 4093.9, |
| "valid_targets_min": 597 |
| }, |
| { |
| "epoch": 4.656, |
| "grad_norm": 0.6045546990414774, |
| "learning_rate": 5.795194715099905e-07, |
| "loss": 0.3856, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4214397966861725, |
| "step": 2910, |
| "valid_targets_mean": 5777.8, |
| "valid_targets_min": 940 |
| }, |
| { |
| "epoch": 4.664, |
| "grad_norm": 0.5952012215364537, |
| "learning_rate": 5.531233280924042e-07, |
| "loss": 0.4219, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3828292787075043, |
| "step": 2915, |
| "valid_targets_mean": 4322.4, |
| "valid_targets_min": 615 |
| }, |
| { |
| "epoch": 4.672, |
| "grad_norm": 0.6450876587783286, |
| "learning_rate": 5.273340018327044e-07, |
| "loss": 0.4067, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4427468776702881, |
| "step": 2920, |
| "valid_targets_mean": 3636.7, |
| "valid_targets_min": 686 |
| }, |
| { |
| "epoch": 4.68, |
| "grad_norm": 0.8489893578004784, |
| "learning_rate": 5.02152297456806e-07, |
| "loss": 0.4044, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44238054752349854, |
| "step": 2925, |
| "valid_targets_mean": 2258.9, |
| "valid_targets_min": 535 |
| }, |
| { |
| "epoch": 4.688, |
| "grad_norm": 0.7104670754151398, |
| "learning_rate": 4.775790007304993e-07, |
| "loss": 0.4145, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43017131090164185, |
| "step": 2930, |
| "valid_targets_mean": 3351.2, |
| "valid_targets_min": 573 |
| }, |
| { |
| "epoch": 4.696, |
| "grad_norm": 0.6357385020238071, |
| "learning_rate": 4.5361487843490924e-07, |
| "loss": 0.4049, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4254443347454071, |
| "step": 2935, |
| "valid_targets_mean": 4101.7, |
| "valid_targets_min": 683 |
| }, |
| { |
| "epoch": 4.704, |
| "grad_norm": 0.5732449316192012, |
| "learning_rate": 4.3026067834258667e-07, |
| "loss": 0.4204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4033844470977783, |
| "step": 2940, |
| "valid_targets_mean": 4472.9, |
| "valid_targets_min": 540 |
| }, |
| { |
| "epoch": 4.712, |
| "grad_norm": 0.48553171918656174, |
| "learning_rate": 4.0751712919417484e-07, |
| "loss": 0.3781, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3926544785499573, |
| "step": 2945, |
| "valid_targets_mean": 6882.5, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 4.72, |
| "grad_norm": 0.6821340139530365, |
| "learning_rate": 3.853849406756549e-07, |
| "loss": 0.4099, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42946845293045044, |
| "step": 2950, |
| "valid_targets_mean": 3391.1, |
| "valid_targets_min": 373 |
| }, |
| { |
| "epoch": 4.728, |
| "grad_norm": 0.6654823260184427, |
| "learning_rate": 3.6386480339621886e-07, |
| "loss": 0.3988, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39326146245002747, |
| "step": 2955, |
| "valid_targets_mean": 3489.5, |
| "valid_targets_min": 652 |
| }, |
| { |
| "epoch": 4.736, |
| "grad_norm": 0.6083477077382592, |
| "learning_rate": 3.4295738886670925e-07, |
| "loss": 0.4034, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3781413435935974, |
| "step": 2960, |
| "valid_targets_mean": 4359.6, |
| "valid_targets_min": 617 |
| }, |
| { |
| "epoch": 4.744, |
| "grad_norm": 0.6597022939202071, |
| "learning_rate": 3.226633494786668e-07, |
| "loss": 0.3878, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3716740012168884, |
| "step": 2965, |
| "valid_targets_mean": 3212.9, |
| "valid_targets_min": 729 |
| }, |
| { |
| "epoch": 4.752, |
| "grad_norm": 0.8113317008052021, |
| "learning_rate": 3.0298331848398033e-07, |
| "loss": 0.4039, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4363306760787964, |
| "step": 2970, |
| "valid_targets_mean": 3148.3, |
| "valid_targets_min": 715 |
| }, |
| { |
| "epoch": 4.76, |
| "grad_norm": 0.6243007406816568, |
| "learning_rate": 2.839179099751133e-07, |
| "loss": 0.4149, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4543706178665161, |
| "step": 2975, |
| "valid_targets_mean": 4166.6, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 4.768, |
| "grad_norm": 0.6224022376593623, |
| "learning_rate": 2.654677188659549e-07, |
| "loss": 0.4168, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.451460063457489, |
| "step": 2980, |
| "valid_targets_mean": 4207.4, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 4.776, |
| "grad_norm": 0.5374144174742845, |
| "learning_rate": 2.476333208732462e-07, |
| "loss": 0.4053, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42308709025382996, |
| "step": 2985, |
| "valid_targets_mean": 6995.9, |
| "valid_targets_min": 780 |
| }, |
| { |
| "epoch": 4.784, |
| "grad_norm": 0.6285382216596294, |
| "learning_rate": 2.3041527249863193e-07, |
| "loss": 0.4221, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42085695266723633, |
| "step": 2990, |
| "valid_targets_mean": 3705.4, |
| "valid_targets_min": 915 |
| }, |
| { |
| "epoch": 4.792, |
| "grad_norm": 0.6313424923207209, |
| "learning_rate": 2.1381411101127013e-07, |
| "loss": 0.4242, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4452266991138458, |
| "step": 2995, |
| "valid_targets_mean": 4179.6, |
| "valid_targets_min": 654 |
| }, |
| { |
| "epoch": 4.8, |
| "grad_norm": 0.5405302367882336, |
| "learning_rate": 1.9783035443108999e-07, |
| "loss": 0.3761, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3892124891281128, |
| "step": 3000, |
| "valid_targets_mean": 4694.6, |
| "valid_targets_min": 726 |
| }, |
| { |
| "epoch": 4.808, |
| "grad_norm": 0.6499699005097503, |
| "learning_rate": 1.8246450151261362e-07, |
| "loss": 0.4126, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46307098865509033, |
| "step": 3005, |
| "valid_targets_mean": 4108.7, |
| "valid_targets_min": 734 |
| }, |
| { |
| "epoch": 4.816, |
| "grad_norm": 0.6639498536382377, |
| "learning_rate": 1.6771703172940635e-07, |
| "loss": 0.3848, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44725149869918823, |
| "step": 3010, |
| "valid_targets_mean": 3558.9, |
| "valid_targets_min": 663 |
| }, |
| { |
| "epoch": 4.824, |
| "grad_norm": 0.549526064647749, |
| "learning_rate": 1.5358840525909967e-07, |
| "loss": 0.3986, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4009050130844116, |
| "step": 3015, |
| "valid_targets_mean": 4778.8, |
| "valid_targets_min": 671 |
| }, |
| { |
| "epoch": 4.832, |
| "grad_norm": 0.7167556183864082, |
| "learning_rate": 1.4007906296904072e-07, |
| "loss": 0.3943, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43567198514938354, |
| "step": 3020, |
| "valid_targets_mean": 3120.4, |
| "valid_targets_min": 547 |
| }, |
| { |
| "epoch": 4.84, |
| "grad_norm": 0.7952417316247405, |
| "learning_rate": 1.2718942640254084e-07, |
| "loss": 0.4024, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4332461953163147, |
| "step": 3025, |
| "valid_targets_mean": 2783.6, |
| "valid_targets_min": 646 |
| }, |
| { |
| "epoch": 4.848, |
| "grad_norm": 0.5604252297788266, |
| "learning_rate": 1.1491989776570623e-07, |
| "loss": 0.4079, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3904341757297516, |
| "step": 3030, |
| "valid_targets_mean": 4534.9, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 4.856, |
| "grad_norm": 0.6858298477952322, |
| "learning_rate": 1.0327085991490127e-07, |
| "loss": 0.4213, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4092717170715332, |
| "step": 3035, |
| "valid_targets_mean": 3204.4, |
| "valid_targets_min": 509 |
| }, |
| { |
| "epoch": 4.864, |
| "grad_norm": 0.6401379988844095, |
| "learning_rate": 9.22426763447981e-08, |
| "loss": 0.4118, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41920021176338196, |
| "step": 3040, |
| "valid_targets_mean": 3930.8, |
| "valid_targets_min": 643 |
| }, |
| { |
| "epoch": 4.872, |
| "grad_norm": 0.5334024885638476, |
| "learning_rate": 8.183569117703461e-08, |
| "loss": 0.3955, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3883068263530731, |
| "step": 3045, |
| "valid_targets_mean": 5270.0, |
| "valid_targets_min": 816 |
| }, |
| { |
| "epoch": 4.88, |
| "grad_norm": 0.5549772986705793, |
| "learning_rate": 7.205022914946957e-08, |
| "loss": 0.4232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38646113872528076, |
| "step": 3050, |
| "valid_targets_mean": 4886.3, |
| "valid_targets_min": 707 |
| }, |
| { |
| "epoch": 4.888, |
| "grad_norm": 0.5767956457076785, |
| "learning_rate": 6.288659560606203e-08, |
| "loss": 0.4086, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3830558657646179, |
| "step": 3055, |
| "valid_targets_mean": 4712.8, |
| "valid_targets_min": 727 |
| }, |
| { |
| "epoch": 4.896, |
| "grad_norm": 0.5988889732664718, |
| "learning_rate": 5.4345076487332114e-08, |
| "loss": 0.3863, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4035729169845581, |
| "step": 3060, |
| "valid_targets_mean": 4299.6, |
| "valid_targets_min": 763 |
| }, |
| { |
| "epoch": 4.904, |
| "grad_norm": 0.5205741542745066, |
| "learning_rate": 4.642593832144382e-08, |
| "loss": 0.4224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36331793665885925, |
| "step": 3065, |
| "valid_targets_mean": 4843.8, |
| "valid_targets_min": 939 |
| }, |
| { |
| "epoch": 4.912, |
| "grad_norm": 0.6165591241445428, |
| "learning_rate": 3.912942821589161e-08, |
| "loss": 0.3843, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3959740400314331, |
| "step": 3070, |
| "valid_targets_mean": 4223.8, |
| "valid_targets_min": 912 |
| }, |
| { |
| "epoch": 4.92, |
| "grad_norm": 0.6355765004228577, |
| "learning_rate": 3.2455773849779935e-08, |
| "loss": 0.4104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42105889320373535, |
| "step": 3075, |
| "valid_targets_mean": 3869.8, |
| "valid_targets_min": 743 |
| }, |
| { |
| "epoch": 4.928, |
| "grad_norm": 0.7183753779829565, |
| "learning_rate": 2.6405183466731154e-08, |
| "loss": 0.4293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46973997354507446, |
| "step": 3080, |
| "valid_targets_mean": 3456.1, |
| "valid_targets_min": 903 |
| }, |
| { |
| "epoch": 4.936, |
| "grad_norm": 0.665925414774037, |
| "learning_rate": 2.0977845868375145e-08, |
| "loss": 0.4192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4395608603954315, |
| "step": 3085, |
| "valid_targets_mean": 3503.2, |
| "valid_targets_min": 615 |
| }, |
| { |
| "epoch": 4.944, |
| "grad_norm": 0.6663253811911488, |
| "learning_rate": 1.6173930408467376e-08, |
| "loss": 0.4184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4220019578933716, |
| "step": 3090, |
| "valid_targets_mean": 3666.4, |
| "valid_targets_min": 663 |
| }, |
| { |
| "epoch": 4.952, |
| "grad_norm": 0.628895057963846, |
| "learning_rate": 1.199358698759978e-08, |
| "loss": 0.3879, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3834680914878845, |
| "step": 3095, |
| "valid_targets_mean": 3802.9, |
| "valid_targets_min": 468 |
| }, |
| { |
| "epoch": 4.96, |
| "grad_norm": 0.6447997763250441, |
| "learning_rate": 8.436946048522298e-09, |
| "loss": 0.4109, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4318605065345764, |
| "step": 3100, |
| "valid_targets_mean": 3820.8, |
| "valid_targets_min": 568 |
| }, |
| { |
| "epoch": 4.968, |
| "grad_norm": 0.4941475635121412, |
| "learning_rate": 5.504118572081662e-09, |
| "loss": 0.3983, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3983478844165802, |
| "step": 3105, |
| "valid_targets_mean": 6376.6, |
| "valid_targets_min": 924 |
| }, |
| { |
| "epoch": 4.976, |
| "grad_norm": 0.6289386884950056, |
| "learning_rate": 3.1951960737419686e-09, |
| "loss": 0.4222, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4331316351890564, |
| "step": 3110, |
| "valid_targets_mean": 4068.7, |
| "valid_targets_min": 763 |
| }, |
| { |
| "epoch": 4.984, |
| "grad_norm": 0.9118066646290699, |
| "learning_rate": 1.5102506007447227e-09, |
| "loss": 0.3988, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4566645622253418, |
| "step": 3115, |
| "valid_targets_mean": 2408.6, |
| "valid_targets_min": 734 |
| }, |
| { |
| "epoch": 4.992, |
| "grad_norm": 0.5570389322501734, |
| "learning_rate": 4.493347298528683e-10, |
| "loss": 0.4032, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3776412904262543, |
| "step": 3120, |
| "valid_targets_mean": 4442.2, |
| "valid_targets_min": 657 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 0.6109238323745433, |
| "learning_rate": 1.248156571209691e-11, |
| "loss": 0.3841, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37494388222694397, |
| "step": 3125, |
| "valid_targets_mean": 3849.6, |
| "valid_targets_min": 730 |
| }, |
| { |
| "epoch": 5.0, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37494388222694397, |
| "step": 3125, |
| "total_flos": 883823926247424.0, |
| "train_loss": 0.468077876701355, |
| "train_runtime": 22577.6779, |
| "train_samples_per_second": 2.212, |
| "train_steps_per_second": 0.138, |
| "valid_targets_mean": 3849.6, |
| "valid_targets_min": 730 |
| } |
| ], |
| "logging_steps": 5, |
| "max_steps": 3125, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 5, |
| "save_steps": 500, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": false, |
| "should_training_stop": false |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 883823926247424.0, |
| "train_batch_size": 1, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
