| { |
| "best_global_step": null, |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 7.0, |
| "eval_steps": 500, |
| "global_step": 3850, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.00909090909090909, |
| "grad_norm": 6.339214137826888, |
| "learning_rate": 4.155844155844156e-07, |
| "loss": 0.6263, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5477997660636902, |
| "step": 5, |
| "valid_targets_mean": 2903.2, |
| "valid_targets_min": 957 |
| }, |
| { |
| "epoch": 0.01818181818181818, |
| "grad_norm": 6.691577065976262, |
| "learning_rate": 9.350649350649352e-07, |
| "loss": 0.658, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5785444378852844, |
| "step": 10, |
| "valid_targets_mean": 2720.8, |
| "valid_targets_min": 1308 |
| }, |
| { |
| "epoch": 0.02727272727272727, |
| "grad_norm": 6.888217378747393, |
| "learning_rate": 1.4545454545454546e-06, |
| "loss": 0.649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6698688268661499, |
| "step": 15, |
| "valid_targets_mean": 2414.9, |
| "valid_targets_min": 1087 |
| }, |
| { |
| "epoch": 0.03636363636363636, |
| "grad_norm": 6.872383497711192, |
| "learning_rate": 1.9740259740259743e-06, |
| "loss": 0.6331, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.7004647850990295, |
| "step": 20, |
| "valid_targets_mean": 2721.0, |
| "valid_targets_min": 743 |
| }, |
| { |
| "epoch": 0.045454545454545456, |
| "grad_norm": 4.103200743218897, |
| "learning_rate": 2.4935064935064936e-06, |
| "loss": 0.5927, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6205931901931763, |
| "step": 25, |
| "valid_targets_mean": 2484.2, |
| "valid_targets_min": 1271 |
| }, |
| { |
| "epoch": 0.05454545454545454, |
| "grad_norm": 2.397410279781038, |
| "learning_rate": 3.0129870129870133e-06, |
| "loss": 0.5956, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5478155612945557, |
| "step": 30, |
| "valid_targets_mean": 3119.8, |
| "valid_targets_min": 1385 |
| }, |
| { |
| "epoch": 0.06363636363636363, |
| "grad_norm": 1.9320989591881759, |
| "learning_rate": 3.532467532467533e-06, |
| "loss": 0.5349, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.6627036333084106, |
| "step": 35, |
| "valid_targets_mean": 3209.0, |
| "valid_targets_min": 1036 |
| }, |
| { |
| "epoch": 0.07272727272727272, |
| "grad_norm": 1.2577860380350014, |
| "learning_rate": 4.051948051948053e-06, |
| "loss": 0.5456, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4294567108154297, |
| "step": 40, |
| "valid_targets_mean": 2889.9, |
| "valid_targets_min": 723 |
| }, |
| { |
| "epoch": 0.08181818181818182, |
| "grad_norm": 1.3213819490142387, |
| "learning_rate": 4.571428571428572e-06, |
| "loss": 0.512, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.5345403552055359, |
| "step": 45, |
| "valid_targets_mean": 3041.4, |
| "valid_targets_min": 1188 |
| }, |
| { |
| "epoch": 0.09090909090909091, |
| "grad_norm": 0.9381336982813961, |
| "learning_rate": 5.090909090909091e-06, |
| "loss": 0.4759, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46422749757766724, |
| "step": 50, |
| "valid_targets_mean": 2768.2, |
| "valid_targets_min": 918 |
| }, |
| { |
| "epoch": 0.1, |
| "grad_norm": 1.1483543606640796, |
| "learning_rate": 5.6103896103896105e-06, |
| "loss": 0.4981, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4253108501434326, |
| "step": 55, |
| "valid_targets_mean": 3121.2, |
| "valid_targets_min": 1242 |
| }, |
| { |
| "epoch": 0.10909090909090909, |
| "grad_norm": 0.8457378397953323, |
| "learning_rate": 6.129870129870131e-06, |
| "loss": 0.466, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46757400035858154, |
| "step": 60, |
| "valid_targets_mean": 2539.6, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 0.11818181818181818, |
| "grad_norm": 0.7504641099947844, |
| "learning_rate": 6.64935064935065e-06, |
| "loss": 0.4836, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.48012202978134155, |
| "step": 65, |
| "valid_targets_mean": 3173.1, |
| "valid_targets_min": 1911 |
| }, |
| { |
| "epoch": 0.12727272727272726, |
| "grad_norm": 0.7849813069658419, |
| "learning_rate": 7.16883116883117e-06, |
| "loss": 0.4367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44680148363113403, |
| "step": 70, |
| "valid_targets_mean": 1945.4, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 0.13636363636363635, |
| "grad_norm": 0.8079545894514485, |
| "learning_rate": 7.68831168831169e-06, |
| "loss": 0.4871, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47567713260650635, |
| "step": 75, |
| "valid_targets_mean": 2781.6, |
| "valid_targets_min": 1103 |
| }, |
| { |
| "epoch": 0.14545454545454545, |
| "grad_norm": 0.6825461454759197, |
| "learning_rate": 8.20779220779221e-06, |
| "loss": 0.4073, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39913469552993774, |
| "step": 80, |
| "valid_targets_mean": 2483.6, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 0.15454545454545454, |
| "grad_norm": 0.6259764551440307, |
| "learning_rate": 8.727272727272728e-06, |
| "loss": 0.4403, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3804532289505005, |
| "step": 85, |
| "valid_targets_mean": 3047.2, |
| "valid_targets_min": 1145 |
| }, |
| { |
| "epoch": 0.16363636363636364, |
| "grad_norm": 0.643861228084392, |
| "learning_rate": 9.246753246753248e-06, |
| "loss": 0.47, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.47174426913261414, |
| "step": 90, |
| "valid_targets_mean": 3787.8, |
| "valid_targets_min": 1296 |
| }, |
| { |
| "epoch": 0.17272727272727273, |
| "grad_norm": 0.5906721673556469, |
| "learning_rate": 9.766233766233766e-06, |
| "loss": 0.3948, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3424094021320343, |
| "step": 95, |
| "valid_targets_mean": 2835.1, |
| "valid_targets_min": 1020 |
| }, |
| { |
| "epoch": 0.18181818181818182, |
| "grad_norm": 0.6753239700311199, |
| "learning_rate": 1.0285714285714285e-05, |
| "loss": 0.3798, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39557331800460815, |
| "step": 100, |
| "valid_targets_mean": 2038.4, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 0.19090909090909092, |
| "grad_norm": 0.6149179303076256, |
| "learning_rate": 1.0805194805194805e-05, |
| "loss": 0.3984, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37537679076194763, |
| "step": 105, |
| "valid_targets_mean": 2919.0, |
| "valid_targets_min": 1386 |
| }, |
| { |
| "epoch": 0.2, |
| "grad_norm": 0.6013440558412507, |
| "learning_rate": 1.1324675324675325e-05, |
| "loss": 0.3929, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35833466053009033, |
| "step": 110, |
| "valid_targets_mean": 2751.6, |
| "valid_targets_min": 796 |
| }, |
| { |
| "epoch": 0.20909090909090908, |
| "grad_norm": 0.6789497483457126, |
| "learning_rate": 1.1844155844155845e-05, |
| "loss": 0.392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38917845487594604, |
| "step": 115, |
| "valid_targets_mean": 2357.2, |
| "valid_targets_min": 953 |
| }, |
| { |
| "epoch": 0.21818181818181817, |
| "grad_norm": 0.7856489772298153, |
| "learning_rate": 1.2363636363636364e-05, |
| "loss": 0.4254, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4267646074295044, |
| "step": 120, |
| "valid_targets_mean": 2231.1, |
| "valid_targets_min": 934 |
| }, |
| { |
| "epoch": 0.22727272727272727, |
| "grad_norm": 0.5962614603101419, |
| "learning_rate": 1.2883116883116884e-05, |
| "loss": 0.3905, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3281465172767639, |
| "step": 125, |
| "valid_targets_mean": 2614.0, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 0.23636363636363636, |
| "grad_norm": 0.6331353154984325, |
| "learning_rate": 1.3402597402597404e-05, |
| "loss": 0.4166, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3900526762008667, |
| "step": 130, |
| "valid_targets_mean": 2835.6, |
| "valid_targets_min": 1360 |
| }, |
| { |
| "epoch": 0.24545454545454545, |
| "grad_norm": 0.6153943492748803, |
| "learning_rate": 1.3922077922077924e-05, |
| "loss": 0.3767, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30019235610961914, |
| "step": 135, |
| "valid_targets_mean": 2453.8, |
| "valid_targets_min": 1125 |
| }, |
| { |
| "epoch": 0.2545454545454545, |
| "grad_norm": 0.6197478318878169, |
| "learning_rate": 1.4441558441558442e-05, |
| "loss": 0.4251, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3900189995765686, |
| "step": 140, |
| "valid_targets_mean": 2988.1, |
| "valid_targets_min": 1448 |
| }, |
| { |
| "epoch": 0.2636363636363636, |
| "grad_norm": 0.6848995712566689, |
| "learning_rate": 1.4961038961038962e-05, |
| "loss": 0.4072, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4560791552066803, |
| "step": 145, |
| "valid_targets_mean": 2818.6, |
| "valid_targets_min": 1274 |
| }, |
| { |
| "epoch": 0.2727272727272727, |
| "grad_norm": 0.6611848899773384, |
| "learning_rate": 1.548051948051948e-05, |
| "loss": 0.3951, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43154481053352356, |
| "step": 150, |
| "valid_targets_mean": 3036.3, |
| "valid_targets_min": 1637 |
| }, |
| { |
| "epoch": 0.2818181818181818, |
| "grad_norm": 0.6630744668701863, |
| "learning_rate": 1.6000000000000003e-05, |
| "loss": 0.39, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.44840043783187866, |
| "step": 155, |
| "valid_targets_mean": 3600.2, |
| "valid_targets_min": 1235 |
| }, |
| { |
| "epoch": 0.2909090909090909, |
| "grad_norm": 0.6582951442716148, |
| "learning_rate": 1.651948051948052e-05, |
| "loss": 0.4335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3294903635978699, |
| "step": 160, |
| "valid_targets_mean": 2506.0, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 0.3, |
| "grad_norm": 0.6563223575769623, |
| "learning_rate": 1.703896103896104e-05, |
| "loss": 0.3676, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41344767808914185, |
| "step": 165, |
| "valid_targets_mean": 3668.6, |
| "valid_targets_min": 1414 |
| }, |
| { |
| "epoch": 0.3090909090909091, |
| "grad_norm": 0.6459971009182044, |
| "learning_rate": 1.7558441558441558e-05, |
| "loss": 0.349, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3544159531593323, |
| "step": 170, |
| "valid_targets_mean": 2733.3, |
| "valid_targets_min": 884 |
| }, |
| { |
| "epoch": 0.3181818181818182, |
| "grad_norm": 0.5674634529192868, |
| "learning_rate": 1.807792207792208e-05, |
| "loss": 0.3579, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2982725501060486, |
| "step": 175, |
| "valid_targets_mean": 2959.8, |
| "valid_targets_min": 964 |
| }, |
| { |
| "epoch": 0.32727272727272727, |
| "grad_norm": 0.6303381761627781, |
| "learning_rate": 1.8597402597402598e-05, |
| "loss": 0.3505, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3669891953468323, |
| "step": 180, |
| "valid_targets_mean": 2843.2, |
| "valid_targets_min": 854 |
| }, |
| { |
| "epoch": 0.33636363636363636, |
| "grad_norm": 0.7564327052685856, |
| "learning_rate": 1.9116883116883117e-05, |
| "loss": 0.3954, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4271281957626343, |
| "step": 185, |
| "valid_targets_mean": 2196.0, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 0.34545454545454546, |
| "grad_norm": 0.5774191959868498, |
| "learning_rate": 1.963636363636364e-05, |
| "loss": 0.3696, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2974555492401123, |
| "step": 190, |
| "valid_targets_mean": 2704.4, |
| "valid_targets_min": 1089 |
| }, |
| { |
| "epoch": 0.35454545454545455, |
| "grad_norm": 0.5845657238298496, |
| "learning_rate": 2.0155844155844157e-05, |
| "loss": 0.3301, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2901238799095154, |
| "step": 195, |
| "valid_targets_mean": 2632.3, |
| "valid_targets_min": 868 |
| }, |
| { |
| "epoch": 0.36363636363636365, |
| "grad_norm": 0.6635921725960787, |
| "learning_rate": 2.0675324675324675e-05, |
| "loss": 0.3493, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37927788496017456, |
| "step": 200, |
| "valid_targets_mean": 2613.2, |
| "valid_targets_min": 818 |
| }, |
| { |
| "epoch": 0.37272727272727274, |
| "grad_norm": 1.0884508967232478, |
| "learning_rate": 2.1194805194805194e-05, |
| "loss": 0.3386, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3851262032985687, |
| "step": 205, |
| "valid_targets_mean": 2818.7, |
| "valid_targets_min": 1403 |
| }, |
| { |
| "epoch": 0.38181818181818183, |
| "grad_norm": 0.668205633210714, |
| "learning_rate": 2.1714285714285715e-05, |
| "loss": 0.3898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38827866315841675, |
| "step": 210, |
| "valid_targets_mean": 3074.4, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 0.39090909090909093, |
| "grad_norm": 0.6948506532864128, |
| "learning_rate": 2.2233766233766234e-05, |
| "loss": 0.3548, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3508937358856201, |
| "step": 215, |
| "valid_targets_mean": 2631.0, |
| "valid_targets_min": 1329 |
| }, |
| { |
| "epoch": 0.4, |
| "grad_norm": 0.6483556182899709, |
| "learning_rate": 2.2753246753246752e-05, |
| "loss": 0.3389, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36243948340415955, |
| "step": 220, |
| "valid_targets_mean": 2725.5, |
| "valid_targets_min": 944 |
| }, |
| { |
| "epoch": 0.4090909090909091, |
| "grad_norm": 0.6208051012663126, |
| "learning_rate": 2.3272727272727274e-05, |
| "loss": 0.3561, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3889102637767792, |
| "step": 225, |
| "valid_targets_mean": 3122.2, |
| "valid_targets_min": 1287 |
| }, |
| { |
| "epoch": 0.41818181818181815, |
| "grad_norm": 0.6336603399119323, |
| "learning_rate": 2.3792207792207793e-05, |
| "loss": 0.3457, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31206339597702026, |
| "step": 230, |
| "valid_targets_mean": 2648.8, |
| "valid_targets_min": 833 |
| }, |
| { |
| "epoch": 0.42727272727272725, |
| "grad_norm": 0.9001451266995485, |
| "learning_rate": 2.4311688311688314e-05, |
| "loss": 0.3625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4581303298473358, |
| "step": 235, |
| "valid_targets_mean": 2362.3, |
| "valid_targets_min": 1068 |
| }, |
| { |
| "epoch": 0.43636363636363634, |
| "grad_norm": 0.6370411644388215, |
| "learning_rate": 2.4831168831168833e-05, |
| "loss": 0.3745, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.46398934721946716, |
| "step": 240, |
| "valid_targets_mean": 3442.1, |
| "valid_targets_min": 1364 |
| }, |
| { |
| "epoch": 0.44545454545454544, |
| "grad_norm": 0.6449951274638801, |
| "learning_rate": 2.535064935064935e-05, |
| "loss": 0.337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3919605612754822, |
| "step": 245, |
| "valid_targets_mean": 2908.2, |
| "valid_targets_min": 1500 |
| }, |
| { |
| "epoch": 0.45454545454545453, |
| "grad_norm": 0.6521200383438808, |
| "learning_rate": 2.5870129870129873e-05, |
| "loss": 0.3661, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4429694414138794, |
| "step": 250, |
| "valid_targets_mean": 3033.2, |
| "valid_targets_min": 992 |
| }, |
| { |
| "epoch": 0.4636363636363636, |
| "grad_norm": 0.6773464252143542, |
| "learning_rate": 2.638961038961039e-05, |
| "loss": 0.3656, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.367068886756897, |
| "step": 255, |
| "valid_targets_mean": 3075.7, |
| "valid_targets_min": 1082 |
| }, |
| { |
| "epoch": 0.4727272727272727, |
| "grad_norm": 1.4569160111773742, |
| "learning_rate": 2.690909090909091e-05, |
| "loss": 0.3427, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.40190428495407104, |
| "step": 260, |
| "valid_targets_mean": 3337.8, |
| "valid_targets_min": 1620 |
| }, |
| { |
| "epoch": 0.4818181818181818, |
| "grad_norm": 0.6924018302373937, |
| "learning_rate": 2.742857142857143e-05, |
| "loss": 0.3483, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31538522243499756, |
| "step": 265, |
| "valid_targets_mean": 2467.6, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 0.4909090909090909, |
| "grad_norm": 0.6869692427452243, |
| "learning_rate": 2.794805194805195e-05, |
| "loss": 0.3196, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3204636573791504, |
| "step": 270, |
| "valid_targets_mean": 2205.6, |
| "valid_targets_min": 1208 |
| }, |
| { |
| "epoch": 0.5, |
| "grad_norm": 0.6370154405500698, |
| "learning_rate": 2.8467532467532472e-05, |
| "loss": 0.3277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3479101359844208, |
| "step": 275, |
| "valid_targets_mean": 3041.8, |
| "valid_targets_min": 1407 |
| }, |
| { |
| "epoch": 0.509090909090909, |
| "grad_norm": 0.6425903782342028, |
| "learning_rate": 2.898701298701299e-05, |
| "loss": 0.3708, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37169209122657776, |
| "step": 280, |
| "valid_targets_mean": 2951.7, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 0.5181818181818182, |
| "grad_norm": 0.6611801390841658, |
| "learning_rate": 2.950649350649351e-05, |
| "loss": 0.3314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3530304431915283, |
| "step": 285, |
| "valid_targets_mean": 2644.7, |
| "valid_targets_min": 683 |
| }, |
| { |
| "epoch": 0.5272727272727272, |
| "grad_norm": 0.6357035859012629, |
| "learning_rate": 3.002597402597403e-05, |
| "loss": 0.3537, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3512324094772339, |
| "step": 290, |
| "valid_targets_mean": 2653.9, |
| "valid_targets_min": 1441 |
| }, |
| { |
| "epoch": 0.5363636363636364, |
| "grad_norm": 0.6939851083287091, |
| "learning_rate": 3.054545454545455e-05, |
| "loss": 0.3769, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42176488041877747, |
| "step": 295, |
| "valid_targets_mean": 2781.9, |
| "valid_targets_min": 922 |
| }, |
| { |
| "epoch": 0.5454545454545454, |
| "grad_norm": 0.9499594024592395, |
| "learning_rate": 3.106493506493507e-05, |
| "loss": 0.355, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4803914427757263, |
| "step": 300, |
| "valid_targets_mean": 2959.4, |
| "valid_targets_min": 1322 |
| }, |
| { |
| "epoch": 0.5545454545454546, |
| "grad_norm": 0.6488791936318495, |
| "learning_rate": 3.158441558441559e-05, |
| "loss": 0.3629, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35879287123680115, |
| "step": 305, |
| "valid_targets_mean": 2696.9, |
| "valid_targets_min": 1041 |
| }, |
| { |
| "epoch": 0.5636363636363636, |
| "grad_norm": 0.6231805329329607, |
| "learning_rate": 3.210389610389611e-05, |
| "loss": 0.3662, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3448548913002014, |
| "step": 310, |
| "valid_targets_mean": 2770.7, |
| "valid_targets_min": 1559 |
| }, |
| { |
| "epoch": 0.5727272727272728, |
| "grad_norm": 0.5744702058158829, |
| "learning_rate": 3.2623376623376626e-05, |
| "loss": 0.3682, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34380805492401123, |
| "step": 315, |
| "valid_targets_mean": 2960.0, |
| "valid_targets_min": 1214 |
| }, |
| { |
| "epoch": 0.5818181818181818, |
| "grad_norm": 0.6620931219456028, |
| "learning_rate": 3.314285714285715e-05, |
| "loss": 0.3326, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34156909584999084, |
| "step": 320, |
| "valid_targets_mean": 2785.3, |
| "valid_targets_min": 521 |
| }, |
| { |
| "epoch": 0.5909090909090909, |
| "grad_norm": 0.6403063875961951, |
| "learning_rate": 3.366233766233766e-05, |
| "loss": 0.3039, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3152548670768738, |
| "step": 325, |
| "valid_targets_mean": 3012.6, |
| "valid_targets_min": 1198 |
| }, |
| { |
| "epoch": 0.6, |
| "grad_norm": 0.6672907050177364, |
| "learning_rate": 3.4181818181818185e-05, |
| "loss": 0.3384, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3686109781265259, |
| "step": 330, |
| "valid_targets_mean": 2802.0, |
| "valid_targets_min": 1371 |
| }, |
| { |
| "epoch": 0.6090909090909091, |
| "grad_norm": 0.7499566408657522, |
| "learning_rate": 3.47012987012987e-05, |
| "loss": 0.3784, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.412411630153656, |
| "step": 335, |
| "valid_targets_mean": 2592.4, |
| "valid_targets_min": 1169 |
| }, |
| { |
| "epoch": 0.6181818181818182, |
| "grad_norm": 0.6208601795738106, |
| "learning_rate": 3.522077922077922e-05, |
| "loss": 0.3234, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2600933313369751, |
| "step": 340, |
| "valid_targets_mean": 2492.6, |
| "valid_targets_min": 761 |
| }, |
| { |
| "epoch": 0.6272727272727273, |
| "grad_norm": 0.5997046648171032, |
| "learning_rate": 3.5740259740259743e-05, |
| "loss": 0.3524, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3274972438812256, |
| "step": 345, |
| "valid_targets_mean": 2641.0, |
| "valid_targets_min": 752 |
| }, |
| { |
| "epoch": 0.6363636363636364, |
| "grad_norm": 0.5773178226077512, |
| "learning_rate": 3.625974025974026e-05, |
| "loss": 0.3517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35591813921928406, |
| "step": 350, |
| "valid_targets_mean": 3244.9, |
| "valid_targets_min": 1371 |
| }, |
| { |
| "epoch": 0.6454545454545455, |
| "grad_norm": 0.6612206892362726, |
| "learning_rate": 3.677922077922078e-05, |
| "loss": 0.3378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3410378098487854, |
| "step": 355, |
| "valid_targets_mean": 2283.9, |
| "valid_targets_min": 1155 |
| }, |
| { |
| "epoch": 0.6545454545454545, |
| "grad_norm": 0.634520882439268, |
| "learning_rate": 3.72987012987013e-05, |
| "loss": 0.315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3190503716468811, |
| "step": 360, |
| "valid_targets_mean": 2368.0, |
| "valid_targets_min": 1201 |
| }, |
| { |
| "epoch": 0.6636363636363637, |
| "grad_norm": 0.7186007261011443, |
| "learning_rate": 3.7818181818181824e-05, |
| "loss": 0.3663, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.43607550859451294, |
| "step": 365, |
| "valid_targets_mean": 2984.2, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 0.6727272727272727, |
| "grad_norm": 0.6626084696102383, |
| "learning_rate": 3.833766233766234e-05, |
| "loss": 0.3417, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28545016050338745, |
| "step": 370, |
| "valid_targets_mean": 2105.5, |
| "valid_targets_min": 1027 |
| }, |
| { |
| "epoch": 0.6818181818181818, |
| "grad_norm": 0.7422089667942458, |
| "learning_rate": 3.885714285714286e-05, |
| "loss": 0.3501, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.337421715259552, |
| "step": 375, |
| "valid_targets_mean": 2343.6, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 0.6909090909090909, |
| "grad_norm": 0.6377484020727952, |
| "learning_rate": 3.937662337662338e-05, |
| "loss": 0.4028, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.39789557456970215, |
| "step": 380, |
| "valid_targets_mean": 2773.6, |
| "valid_targets_min": 1463 |
| }, |
| { |
| "epoch": 0.7, |
| "grad_norm": 1.2550270765338776, |
| "learning_rate": 3.98961038961039e-05, |
| "loss": 0.3573, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3744387626647949, |
| "step": 385, |
| "valid_targets_mean": 2553.1, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 0.7090909090909091, |
| "grad_norm": 0.6439863150267273, |
| "learning_rate": 3.999986847364818e-05, |
| "loss": 0.3513, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32510316371917725, |
| "step": 390, |
| "valid_targets_mean": 2936.0, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 0.7181818181818181, |
| "grad_norm": 0.5549572539070133, |
| "learning_rate": 3.999933415080877e-05, |
| "loss": 0.3555, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34088924527168274, |
| "step": 395, |
| "valid_targets_mean": 2975.9, |
| "valid_targets_min": 1628 |
| }, |
| { |
| "epoch": 0.7272727272727273, |
| "grad_norm": 0.5728271061572586, |
| "learning_rate": 3.999838882205719e-05, |
| "loss": 0.3517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3326980471611023, |
| "step": 400, |
| "valid_targets_mean": 3024.2, |
| "valid_targets_min": 1244 |
| }, |
| { |
| "epoch": 0.7363636363636363, |
| "grad_norm": 0.5933107195266674, |
| "learning_rate": 3.999703250682087e-05, |
| "loss": 0.3487, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3289388120174408, |
| "step": 405, |
| "valid_targets_mean": 3081.2, |
| "valid_targets_min": 946 |
| }, |
| { |
| "epoch": 0.7454545454545455, |
| "grad_norm": 0.5704829683945404, |
| "learning_rate": 3.9995265232973414e-05, |
| "loss": 0.3492, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33433687686920166, |
| "step": 410, |
| "valid_targets_mean": 2857.6, |
| "valid_targets_min": 977 |
| }, |
| { |
| "epoch": 0.7545454545454545, |
| "grad_norm": 0.6306739160212078, |
| "learning_rate": 3.9993087036834034e-05, |
| "loss": 0.3117, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3217725455760956, |
| "step": 415, |
| "valid_targets_mean": 2499.9, |
| "valid_targets_min": 1433 |
| }, |
| { |
| "epoch": 0.7636363636363637, |
| "grad_norm": 1.5946938923112275, |
| "learning_rate": 3.9990497963166797e-05, |
| "loss": 0.3382, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2770576477050781, |
| "step": 420, |
| "valid_targets_mean": 2605.4, |
| "valid_targets_min": 962 |
| }, |
| { |
| "epoch": 0.7727272727272727, |
| "grad_norm": 0.600595195060008, |
| "learning_rate": 3.99874980651797e-05, |
| "loss": 0.3466, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37298667430877686, |
| "step": 425, |
| "valid_targets_mean": 2681.2, |
| "valid_targets_min": 856 |
| }, |
| { |
| "epoch": 0.7818181818181819, |
| "grad_norm": 1.27833536268593, |
| "learning_rate": 3.998408740452359e-05, |
| "loss": 0.3358, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2970236539840698, |
| "step": 430, |
| "valid_targets_mean": 2670.9, |
| "valid_targets_min": 1294 |
| }, |
| { |
| "epoch": 0.7909090909090909, |
| "grad_norm": 0.556446397533044, |
| "learning_rate": 3.998026605129088e-05, |
| "loss": 0.3138, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32350656390190125, |
| "step": 435, |
| "valid_targets_mean": 2978.1, |
| "valid_targets_min": 805 |
| }, |
| { |
| "epoch": 0.8, |
| "grad_norm": 0.5867403646744739, |
| "learning_rate": 3.997603408401413e-05, |
| "loss": 0.3452, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3522062301635742, |
| "step": 440, |
| "valid_targets_mean": 3308.9, |
| "valid_targets_min": 703 |
| }, |
| { |
| "epoch": 0.8090909090909091, |
| "grad_norm": 0.5864500142871352, |
| "learning_rate": 3.997139158966441e-05, |
| "loss": 0.334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30179262161254883, |
| "step": 445, |
| "valid_targets_mean": 2586.1, |
| "valid_targets_min": 1202 |
| }, |
| { |
| "epoch": 0.8181818181818182, |
| "grad_norm": 0.7051801097423364, |
| "learning_rate": 3.996633866364953e-05, |
| "loss": 0.3519, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.41929012537002563, |
| "step": 450, |
| "valid_targets_mean": 2292.3, |
| "valid_targets_min": 849 |
| }, |
| { |
| "epoch": 0.8272727272727273, |
| "grad_norm": 0.6665108592747994, |
| "learning_rate": 3.996087540981206e-05, |
| "loss": 0.3488, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.42378395795822144, |
| "step": 455, |
| "valid_targets_mean": 2839.5, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 0.8363636363636363, |
| "grad_norm": 0.5610962836076394, |
| "learning_rate": 3.9955001940427236e-05, |
| "loss": 0.3157, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3496928811073303, |
| "step": 460, |
| "valid_targets_mean": 3111.1, |
| "valid_targets_min": 1193 |
| }, |
| { |
| "epoch": 0.8454545454545455, |
| "grad_norm": 0.6683085302834227, |
| "learning_rate": 3.99487183762006e-05, |
| "loss": 0.3339, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3613266050815582, |
| "step": 465, |
| "valid_targets_mean": 2604.9, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 0.8545454545454545, |
| "grad_norm": 0.6083634878868931, |
| "learning_rate": 3.994202484626555e-05, |
| "loss": 0.3391, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32465070486068726, |
| "step": 470, |
| "valid_targets_mean": 2452.5, |
| "valid_targets_min": 1520 |
| }, |
| { |
| "epoch": 0.8636363636363636, |
| "grad_norm": 0.612777955562894, |
| "learning_rate": 3.993492148818069e-05, |
| "loss": 0.3187, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2962433695793152, |
| "step": 475, |
| "valid_targets_mean": 2897.3, |
| "valid_targets_min": 1415 |
| }, |
| { |
| "epoch": 0.8727272727272727, |
| "grad_norm": 0.5845598358859424, |
| "learning_rate": 3.992740844792699e-05, |
| "loss": 0.316, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2899693250656128, |
| "step": 480, |
| "valid_targets_mean": 2515.9, |
| "valid_targets_min": 883 |
| }, |
| { |
| "epoch": 0.8818181818181818, |
| "grad_norm": 0.5412938052268608, |
| "learning_rate": 3.991948587990479e-05, |
| "loss": 0.3055, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2653069496154785, |
| "step": 485, |
| "valid_targets_mean": 3022.9, |
| "valid_targets_min": 1033 |
| }, |
| { |
| "epoch": 0.8909090909090909, |
| "grad_norm": 0.6492200255248272, |
| "learning_rate": 3.991115394693061e-05, |
| "loss": 0.3303, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2680763602256775, |
| "step": 490, |
| "valid_targets_mean": 2127.9, |
| "valid_targets_min": 1310 |
| }, |
| { |
| "epoch": 0.9, |
| "grad_norm": 0.5141368409068433, |
| "learning_rate": 3.990241282023385e-05, |
| "loss": 0.3307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.286845326423645, |
| "step": 495, |
| "valid_targets_mean": 3001.3, |
| "valid_targets_min": 1225 |
| }, |
| { |
| "epoch": 0.9090909090909091, |
| "grad_norm": 0.574684130591376, |
| "learning_rate": 3.989326267945323e-05, |
| "loss": 0.3325, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29765868186950684, |
| "step": 500, |
| "valid_targets_mean": 2777.2, |
| "valid_targets_min": 1084 |
| }, |
| { |
| "epoch": 0.9181818181818182, |
| "grad_norm": 0.5692282689065302, |
| "learning_rate": 3.98837037126331e-05, |
| "loss": 0.344, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3814631700515747, |
| "step": 505, |
| "valid_targets_mean": 3176.1, |
| "valid_targets_min": 957 |
| }, |
| { |
| "epoch": 0.9272727272727272, |
| "grad_norm": 0.5977294617991848, |
| "learning_rate": 3.98737361162196e-05, |
| "loss": 0.3378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3976469933986664, |
| "step": 510, |
| "valid_targets_mean": 3157.1, |
| "valid_targets_min": 1135 |
| }, |
| { |
| "epoch": 0.9363636363636364, |
| "grad_norm": 0.6220654599149978, |
| "learning_rate": 3.986336009505659e-05, |
| "loss": 0.337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3083745837211609, |
| "step": 515, |
| "valid_targets_mean": 2279.6, |
| "valid_targets_min": 891 |
| }, |
| { |
| "epoch": 0.9454545454545454, |
| "grad_norm": 0.546118021121679, |
| "learning_rate": 3.985257586238149e-05, |
| "loss": 0.3207, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28898388147354126, |
| "step": 520, |
| "valid_targets_mean": 2713.4, |
| "valid_targets_min": 1345 |
| }, |
| { |
| "epoch": 0.9545454545454546, |
| "grad_norm": 0.5482830720326405, |
| "learning_rate": 3.984138363982084e-05, |
| "loss": 0.3378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27064448595046997, |
| "step": 525, |
| "valid_targets_mean": 2486.1, |
| "valid_targets_min": 705 |
| }, |
| { |
| "epoch": 0.9636363636363636, |
| "grad_norm": 0.6226203497672143, |
| "learning_rate": 3.982978365738578e-05, |
| "loss": 0.3361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3750750422477722, |
| "step": 530, |
| "valid_targets_mean": 2812.8, |
| "valid_targets_min": 1062 |
| }, |
| { |
| "epoch": 0.9727272727272728, |
| "grad_norm": 0.5612366252493312, |
| "learning_rate": 3.981777615346731e-05, |
| "loss": 0.3232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3804115653038025, |
| "step": 535, |
| "valid_targets_mean": 2720.2, |
| "valid_targets_min": 927 |
| }, |
| { |
| "epoch": 0.9818181818181818, |
| "grad_norm": 0.6294458190395656, |
| "learning_rate": 3.980536137483141e-05, |
| "loss": 0.3434, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.4118061363697052, |
| "step": 540, |
| "valid_targets_mean": 2462.1, |
| "valid_targets_min": 1454 |
| }, |
| { |
| "epoch": 0.990909090909091, |
| "grad_norm": 0.6220192316895428, |
| "learning_rate": 3.9792539576613934e-05, |
| "loss": 0.323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32386863231658936, |
| "step": 545, |
| "valid_targets_mean": 2925.1, |
| "valid_targets_min": 531 |
| }, |
| { |
| "epoch": 1.0, |
| "grad_norm": 0.5842866957763833, |
| "learning_rate": 3.9779311022315405e-05, |
| "loss": 0.2908, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2880892753601074, |
| "step": 550, |
| "valid_targets_mean": 2579.1, |
| "valid_targets_min": 1080 |
| }, |
| { |
| "epoch": 1.009090909090909, |
| "grad_norm": 0.6447275138647391, |
| "learning_rate": 3.976567598379558e-05, |
| "loss": 0.3405, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3465537428855896, |
| "step": 555, |
| "valid_targets_mean": 2912.1, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 1.018181818181818, |
| "grad_norm": 0.5563184580746852, |
| "learning_rate": 3.975163474126785e-05, |
| "loss": 0.3241, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34276115894317627, |
| "step": 560, |
| "valid_targets_mean": 3331.6, |
| "valid_targets_min": 1656 |
| }, |
| { |
| "epoch": 1.0272727272727273, |
| "grad_norm": 0.6091974583109037, |
| "learning_rate": 3.9737187583293505e-05, |
| "loss": 0.3216, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28951388597488403, |
| "step": 565, |
| "valid_targets_mean": 2659.5, |
| "valid_targets_min": 1432 |
| }, |
| { |
| "epoch": 1.0363636363636364, |
| "grad_norm": 0.5131660550916598, |
| "learning_rate": 3.9722334806775806e-05, |
| "loss": 0.3127, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32671698927879333, |
| "step": 570, |
| "valid_targets_mean": 3052.7, |
| "valid_targets_min": 1228 |
| }, |
| { |
| "epoch": 1.0454545454545454, |
| "grad_norm": 0.5675543992731116, |
| "learning_rate": 3.9707076716953866e-05, |
| "loss": 0.2979, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3123570680618286, |
| "step": 575, |
| "valid_targets_mean": 2878.4, |
| "valid_targets_min": 1736 |
| }, |
| { |
| "epoch": 1.0545454545454545, |
| "grad_norm": 0.6335186826998919, |
| "learning_rate": 3.969141362739636e-05, |
| "loss": 0.2983, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32032886147499084, |
| "step": 580, |
| "valid_targets_mean": 2756.3, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 1.0636363636363637, |
| "grad_norm": 0.5740150543951968, |
| "learning_rate": 3.967534585999515e-05, |
| "loss": 0.2861, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25155892968177795, |
| "step": 585, |
| "valid_targets_mean": 2694.0, |
| "valid_targets_min": 1068 |
| }, |
| { |
| "epoch": 1.0727272727272728, |
| "grad_norm": 0.588406146061277, |
| "learning_rate": 3.965887374495859e-05, |
| "loss": 0.3043, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2900567352771759, |
| "step": 590, |
| "valid_targets_mean": 2990.4, |
| "valid_targets_min": 947 |
| }, |
| { |
| "epoch": 1.0818181818181818, |
| "grad_norm": 0.5601881961473839, |
| "learning_rate": 3.964199762080478e-05, |
| "loss": 0.3065, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3262474238872528, |
| "step": 595, |
| "valid_targets_mean": 3018.2, |
| "valid_targets_min": 1111 |
| }, |
| { |
| "epoch": 1.0909090909090908, |
| "grad_norm": 0.5520777767766365, |
| "learning_rate": 3.9624717834354606e-05, |
| "loss": 0.2908, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2716611921787262, |
| "step": 600, |
| "valid_targets_mean": 2441.5, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 1.1, |
| "grad_norm": 0.5612527579381237, |
| "learning_rate": 3.9607034740724615e-05, |
| "loss": 0.2902, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31123465299606323, |
| "step": 605, |
| "valid_targets_mean": 2829.6, |
| "valid_targets_min": 436 |
| }, |
| { |
| "epoch": 1.1090909090909091, |
| "grad_norm": 0.5797580457154696, |
| "learning_rate": 3.958894870331971e-05, |
| "loss": 0.2809, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26564833521842957, |
| "step": 610, |
| "valid_targets_mean": 2660.3, |
| "valid_targets_min": 1203 |
| }, |
| { |
| "epoch": 1.1181818181818182, |
| "grad_norm": 2.3781597689886103, |
| "learning_rate": 3.9570460093825664e-05, |
| "loss": 0.347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3429916501045227, |
| "step": 615, |
| "valid_targets_mean": 3107.9, |
| "valid_targets_min": 1230 |
| }, |
| { |
| "epoch": 1.1272727272727272, |
| "grad_norm": 0.552110516149831, |
| "learning_rate": 3.9551569292201536e-05, |
| "loss": 0.2715, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30601733922958374, |
| "step": 620, |
| "valid_targets_mean": 2663.9, |
| "valid_targets_min": 1271 |
| }, |
| { |
| "epoch": 1.1363636363636362, |
| "grad_norm": 0.6417975384157059, |
| "learning_rate": 3.9532276686671804e-05, |
| "loss": 0.3171, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33833664655685425, |
| "step": 625, |
| "valid_targets_mean": 3195.1, |
| "valid_targets_min": 1222 |
| }, |
| { |
| "epoch": 1.1454545454545455, |
| "grad_norm": 0.6137360698285799, |
| "learning_rate": 3.951258267371841e-05, |
| "loss": 0.2971, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3272460401058197, |
| "step": 630, |
| "valid_targets_mean": 2482.5, |
| "valid_targets_min": 1562 |
| }, |
| { |
| "epoch": 1.1545454545454545, |
| "grad_norm": 0.5820527391442863, |
| "learning_rate": 3.9492487658072615e-05, |
| "loss": 0.2888, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29666125774383545, |
| "step": 635, |
| "valid_targets_mean": 2572.6, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 1.1636363636363636, |
| "grad_norm": 0.5678713238695126, |
| "learning_rate": 3.947199205270668e-05, |
| "loss": 0.3227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34240779280662537, |
| "step": 640, |
| "valid_targets_mean": 3045.4, |
| "valid_targets_min": 1172 |
| }, |
| { |
| "epoch": 1.1727272727272728, |
| "grad_norm": 0.5266377557583839, |
| "learning_rate": 3.9451096278825386e-05, |
| "loss": 0.3286, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36640381813049316, |
| "step": 645, |
| "valid_targets_mean": 3615.8, |
| "valid_targets_min": 1346 |
| }, |
| { |
| "epoch": 1.1818181818181819, |
| "grad_norm": 0.5322441323754087, |
| "learning_rate": 3.942980076585735e-05, |
| "loss": 0.2964, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25494956970214844, |
| "step": 650, |
| "valid_targets_mean": 2572.3, |
| "valid_targets_min": 900 |
| }, |
| { |
| "epoch": 1.190909090909091, |
| "grad_norm": 0.5414703426964603, |
| "learning_rate": 3.940810595144624e-05, |
| "loss": 0.2882, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28678566217422485, |
| "step": 655, |
| "valid_targets_mean": 2838.4, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 1.2, |
| "grad_norm": 0.9276630436508214, |
| "learning_rate": 3.938601228144173e-05, |
| "loss": 0.284, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2708185613155365, |
| "step": 660, |
| "valid_targets_mean": 3114.6, |
| "valid_targets_min": 1752 |
| }, |
| { |
| "epoch": 1.209090909090909, |
| "grad_norm": 0.5900825119721383, |
| "learning_rate": 3.9363520209890405e-05, |
| "loss": 0.3043, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3425714671611786, |
| "step": 665, |
| "valid_targets_mean": 3101.9, |
| "valid_targets_min": 1035 |
| }, |
| { |
| "epoch": 1.2181818181818183, |
| "grad_norm": 0.5479378833478837, |
| "learning_rate": 3.9340630199026365e-05, |
| "loss": 0.2791, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31353360414505005, |
| "step": 670, |
| "valid_targets_mean": 3066.2, |
| "valid_targets_min": 1100 |
| }, |
| { |
| "epoch": 1.2272727272727273, |
| "grad_norm": 0.6726858996451737, |
| "learning_rate": 3.931734271926176e-05, |
| "loss": 0.301, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36476853489875793, |
| "step": 675, |
| "valid_targets_mean": 2387.9, |
| "valid_targets_min": 846 |
| }, |
| { |
| "epoch": 1.2363636363636363, |
| "grad_norm": 0.5964455414003474, |
| "learning_rate": 3.929365824917712e-05, |
| "loss": 0.3236, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32583141326904297, |
| "step": 680, |
| "valid_targets_mean": 2670.8, |
| "valid_targets_min": 1297 |
| }, |
| { |
| "epoch": 1.2454545454545454, |
| "grad_norm": 0.5781040411209731, |
| "learning_rate": 3.9269577275511504e-05, |
| "loss": 0.3417, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30773332715034485, |
| "step": 685, |
| "valid_targets_mean": 2672.5, |
| "valid_targets_min": 957 |
| }, |
| { |
| "epoch": 1.2545454545454544, |
| "grad_norm": 0.5639647192049679, |
| "learning_rate": 3.924510029315253e-05, |
| "loss": 0.2942, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2964206635951996, |
| "step": 690, |
| "valid_targets_mean": 2741.2, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 1.2636363636363637, |
| "grad_norm": 0.5627064673073386, |
| "learning_rate": 3.922022780512614e-05, |
| "loss": 0.3456, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.355957567691803, |
| "step": 695, |
| "valid_targets_mean": 3049.9, |
| "valid_targets_min": 1020 |
| }, |
| { |
| "epoch": 1.2727272727272727, |
| "grad_norm": 0.49715452818653605, |
| "learning_rate": 3.919496032258637e-05, |
| "loss": 0.2891, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23877036571502686, |
| "step": 700, |
| "valid_targets_mean": 3023.6, |
| "valid_targets_min": 1371 |
| }, |
| { |
| "epoch": 1.2818181818181817, |
| "grad_norm": 0.49529413287800456, |
| "learning_rate": 3.9169298364804716e-05, |
| "loss": 0.3041, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22517327964305878, |
| "step": 705, |
| "valid_targets_mean": 2738.3, |
| "valid_targets_min": 538 |
| }, |
| { |
| "epoch": 1.290909090909091, |
| "grad_norm": 0.6026202338409374, |
| "learning_rate": 3.914324245915956e-05, |
| "loss": 0.311, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.36687037348747253, |
| "step": 710, |
| "valid_targets_mean": 3185.6, |
| "valid_targets_min": 1202 |
| }, |
| { |
| "epoch": 1.3, |
| "grad_norm": 0.580349480036476, |
| "learning_rate": 3.91167931411253e-05, |
| "loss": 0.3153, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25920552015304565, |
| "step": 715, |
| "valid_targets_mean": 3105.7, |
| "valid_targets_min": 1270 |
| }, |
| { |
| "epoch": 1.309090909090909, |
| "grad_norm": 0.5139387937516867, |
| "learning_rate": 3.908995095426134e-05, |
| "loss": 0.3288, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3409667909145355, |
| "step": 720, |
| "valid_targets_mean": 3377.1, |
| "valid_targets_min": 865 |
| }, |
| { |
| "epoch": 1.3181818181818181, |
| "grad_norm": 0.6127282175817513, |
| "learning_rate": 3.90627164502009e-05, |
| "loss": 0.273, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26966482400894165, |
| "step": 725, |
| "valid_targets_mean": 2139.5, |
| "valid_targets_min": 926 |
| }, |
| { |
| "epoch": 1.3272727272727272, |
| "grad_norm": 0.5506882869535827, |
| "learning_rate": 3.903509018863974e-05, |
| "loss": 0.2881, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2546706199645996, |
| "step": 730, |
| "valid_targets_mean": 2567.1, |
| "valid_targets_min": 1240 |
| }, |
| { |
| "epoch": 1.3363636363636364, |
| "grad_norm": 0.555751244337636, |
| "learning_rate": 3.90070727373246e-05, |
| "loss": 0.2925, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3293226957321167, |
| "step": 735, |
| "valid_targets_mean": 2879.5, |
| "valid_targets_min": 989 |
| }, |
| { |
| "epoch": 1.3454545454545455, |
| "grad_norm": 1.4072423287991722, |
| "learning_rate": 3.897866467204155e-05, |
| "loss": 0.2898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2524782419204712, |
| "step": 740, |
| "valid_targets_mean": 2753.0, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 1.3545454545454545, |
| "grad_norm": 0.5528867464762154, |
| "learning_rate": 3.894986657660418e-05, |
| "loss": 0.301, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32227402925491333, |
| "step": 745, |
| "valid_targets_mean": 2946.1, |
| "valid_targets_min": 1568 |
| }, |
| { |
| "epoch": 1.3636363636363638, |
| "grad_norm": 0.577032285353784, |
| "learning_rate": 3.892067904284154e-05, |
| "loss": 0.3165, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2747570872306824, |
| "step": 750, |
| "valid_targets_mean": 2776.9, |
| "valid_targets_min": 942 |
| }, |
| { |
| "epoch": 1.3727272727272728, |
| "grad_norm": 0.6292241710341524, |
| "learning_rate": 3.889110267058608e-05, |
| "loss": 0.2927, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33666500449180603, |
| "step": 755, |
| "valid_targets_mean": 2910.2, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 1.3818181818181818, |
| "grad_norm": 0.563288737269173, |
| "learning_rate": 3.886113806766121e-05, |
| "loss": 0.3378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2766912877559662, |
| "step": 760, |
| "valid_targets_mean": 2442.6, |
| "valid_targets_min": 603 |
| }, |
| { |
| "epoch": 1.3909090909090909, |
| "grad_norm": 0.6656004786051898, |
| "learning_rate": 3.883078584986888e-05, |
| "loss": 0.2884, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29978060722351074, |
| "step": 765, |
| "valid_targets_mean": 2543.6, |
| "valid_targets_min": 1096 |
| }, |
| { |
| "epoch": 1.4, |
| "grad_norm": 0.5236885976409229, |
| "learning_rate": 3.8800046640976916e-05, |
| "loss": 0.3098, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33730077743530273, |
| "step": 770, |
| "valid_targets_mean": 3164.1, |
| "valid_targets_min": 1218 |
| }, |
| { |
| "epoch": 1.4090909090909092, |
| "grad_norm": 0.5593924531226494, |
| "learning_rate": 3.876892107270616e-05, |
| "loss": 0.3122, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2947944700717926, |
| "step": 775, |
| "valid_targets_mean": 2482.2, |
| "valid_targets_min": 1080 |
| }, |
| { |
| "epoch": 1.4181818181818182, |
| "grad_norm": 0.5906084708968861, |
| "learning_rate": 3.873740978471755e-05, |
| "loss": 0.2911, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23599432408809662, |
| "step": 780, |
| "valid_targets_mean": 2478.7, |
| "valid_targets_min": 1493 |
| }, |
| { |
| "epoch": 1.4272727272727272, |
| "grad_norm": 0.5850744993629736, |
| "learning_rate": 3.8705513424598934e-05, |
| "loss": 0.3283, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30988335609436035, |
| "step": 785, |
| "valid_targets_mean": 2411.6, |
| "valid_targets_min": 1177 |
| }, |
| { |
| "epoch": 1.4363636363636363, |
| "grad_norm": 0.5894689684496299, |
| "learning_rate": 3.8673232647851756e-05, |
| "loss": 0.2962, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31927740573883057, |
| "step": 790, |
| "valid_targets_mean": 2514.2, |
| "valid_targets_min": 804 |
| }, |
| { |
| "epoch": 1.4454545454545453, |
| "grad_norm": 0.5495995497086523, |
| "learning_rate": 3.8640568117877594e-05, |
| "loss": 0.3044, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33518069982528687, |
| "step": 795, |
| "valid_targets_mean": 3436.6, |
| "valid_targets_min": 1066 |
| }, |
| { |
| "epoch": 1.4545454545454546, |
| "grad_norm": 0.6941187533631331, |
| "learning_rate": 3.8607520505964574e-05, |
| "loss": 0.2927, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2904342710971832, |
| "step": 800, |
| "valid_targets_mean": 2603.8, |
| "valid_targets_min": 1062 |
| }, |
| { |
| "epoch": 1.4636363636363636, |
| "grad_norm": 0.5460794833502489, |
| "learning_rate": 3.857409049127348e-05, |
| "loss": 0.2973, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.279479444026947, |
| "step": 805, |
| "valid_targets_mean": 2939.2, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 1.4727272727272727, |
| "grad_norm": 0.5927685903990932, |
| "learning_rate": 3.8540278760823866e-05, |
| "loss": 0.3024, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3844236135482788, |
| "step": 810, |
| "valid_targets_mean": 2929.4, |
| "valid_targets_min": 1126 |
| }, |
| { |
| "epoch": 1.481818181818182, |
| "grad_norm": 0.6004710936297869, |
| "learning_rate": 3.8506086009479934e-05, |
| "loss": 0.297, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35591375827789307, |
| "step": 815, |
| "valid_targets_mean": 2592.2, |
| "valid_targets_min": 890 |
| }, |
| { |
| "epoch": 1.490909090909091, |
| "grad_norm": 0.5137662741104242, |
| "learning_rate": 3.8471512939936224e-05, |
| "loss": 0.28, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28532475233078003, |
| "step": 820, |
| "valid_targets_mean": 3796.9, |
| "valid_targets_min": 1147 |
| }, |
| { |
| "epoch": 1.5, |
| "grad_norm": 0.580257790045676, |
| "learning_rate": 3.843656026270319e-05, |
| "loss": 0.3151, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32185226678848267, |
| "step": 825, |
| "valid_targets_mean": 2719.7, |
| "valid_targets_min": 602 |
| }, |
| { |
| "epoch": 1.509090909090909, |
| "grad_norm": 0.5166030598464046, |
| "learning_rate": 3.840122869609258e-05, |
| "loss": 0.3075, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25254085659980774, |
| "step": 830, |
| "valid_targets_mean": 3248.6, |
| "valid_targets_min": 1070 |
| }, |
| { |
| "epoch": 1.518181818181818, |
| "grad_norm": 0.5921245945686834, |
| "learning_rate": 3.8365518966202724e-05, |
| "loss": 0.3173, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3269743323326111, |
| "step": 835, |
| "valid_targets_mean": 2656.5, |
| "valid_targets_min": 1177 |
| }, |
| { |
| "epoch": 1.5272727272727273, |
| "grad_norm": 0.6134729011732185, |
| "learning_rate": 3.832943180690356e-05, |
| "loss": 0.3075, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3301253020763397, |
| "step": 840, |
| "valid_targets_mean": 2366.7, |
| "valid_targets_min": 1095 |
| }, |
| { |
| "epoch": 1.5363636363636364, |
| "grad_norm": 0.535738543010804, |
| "learning_rate": 3.829296795982156e-05, |
| "loss": 0.2853, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28925201296806335, |
| "step": 845, |
| "valid_targets_mean": 2624.3, |
| "valid_targets_min": 743 |
| }, |
| { |
| "epoch": 1.5454545454545454, |
| "grad_norm": 0.600523233527797, |
| "learning_rate": 3.8256128174324515e-05, |
| "loss": 0.3094, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3454379141330719, |
| "step": 850, |
| "valid_targets_mean": 2845.1, |
| "valid_targets_min": 1278 |
| }, |
| { |
| "epoch": 1.5545454545454547, |
| "grad_norm": 0.6308938968736859, |
| "learning_rate": 3.82189132075061e-05, |
| "loss": 0.3011, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30569013953208923, |
| "step": 855, |
| "valid_targets_mean": 2541.4, |
| "valid_targets_min": 873 |
| }, |
| { |
| "epoch": 1.5636363636363635, |
| "grad_norm": 0.5347500694529886, |
| "learning_rate": 3.818132382417037e-05, |
| "loss": 0.2995, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3351518213748932, |
| "step": 860, |
| "valid_targets_mean": 3025.1, |
| "valid_targets_min": 1379 |
| }, |
| { |
| "epoch": 1.5727272727272728, |
| "grad_norm": 0.5708969532984176, |
| "learning_rate": 3.8143360796815964e-05, |
| "loss": 0.3366, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3305366635322571, |
| "step": 865, |
| "valid_targets_mean": 3047.1, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 1.5818181818181818, |
| "grad_norm": 0.5387436367009513, |
| "learning_rate": 3.81050249056203e-05, |
| "loss": 0.3013, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3135007619857788, |
| "step": 870, |
| "valid_targets_mean": 3312.9, |
| "valid_targets_min": 1369 |
| }, |
| { |
| "epoch": 1.5909090909090908, |
| "grad_norm": 0.5237363567692674, |
| "learning_rate": 3.8066316938423495e-05, |
| "loss": 0.3142, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3161287009716034, |
| "step": 875, |
| "valid_targets_mean": 3033.4, |
| "valid_targets_min": 1441 |
| }, |
| { |
| "epoch": 1.6, |
| "grad_norm": 0.5633442537930327, |
| "learning_rate": 3.8027237690712206e-05, |
| "loss": 0.3001, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29282844066619873, |
| "step": 880, |
| "valid_targets_mean": 2575.6, |
| "valid_targets_min": 1008 |
| }, |
| { |
| "epoch": 1.6090909090909091, |
| "grad_norm": 0.6071962643151043, |
| "learning_rate": 3.798778796560326e-05, |
| "loss": 0.334, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3855435848236084, |
| "step": 885, |
| "valid_targets_mean": 3637.1, |
| "valid_targets_min": 1275 |
| }, |
| { |
| "epoch": 1.6181818181818182, |
| "grad_norm": 0.5144501183205095, |
| "learning_rate": 3.794796857382717e-05, |
| "loss": 0.3243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26045137643814087, |
| "step": 890, |
| "valid_targets_mean": 2535.8, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 1.6272727272727274, |
| "grad_norm": 0.6391108758936941, |
| "learning_rate": 3.790778033371145e-05, |
| "loss": 0.3247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3640952706336975, |
| "step": 895, |
| "valid_targets_mean": 2988.9, |
| "valid_targets_min": 917 |
| }, |
| { |
| "epoch": 1.6363636363636362, |
| "grad_norm": 0.46946518875838034, |
| "learning_rate": 3.786722407116379e-05, |
| "loss": 0.3034, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2868632674217224, |
| "step": 900, |
| "valid_targets_mean": 3317.6, |
| "valid_targets_min": 1161 |
| }, |
| { |
| "epoch": 1.6454545454545455, |
| "grad_norm": 0.5766601735552375, |
| "learning_rate": 3.782630061965515e-05, |
| "loss": 0.2775, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30782195925712585, |
| "step": 905, |
| "valid_targets_mean": 2723.4, |
| "valid_targets_min": 1319 |
| }, |
| { |
| "epoch": 1.6545454545454545, |
| "grad_norm": 0.6297295273887248, |
| "learning_rate": 3.778501082020255e-05, |
| "loss": 0.3104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3077327609062195, |
| "step": 910, |
| "valid_targets_mean": 2310.8, |
| "valid_targets_min": 962 |
| }, |
| { |
| "epoch": 1.6636363636363636, |
| "grad_norm": 0.5724281159599157, |
| "learning_rate": 3.7743355521351814e-05, |
| "loss": 0.2642, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2423037588596344, |
| "step": 915, |
| "valid_targets_mean": 2519.7, |
| "valid_targets_min": 1096 |
| }, |
| { |
| "epoch": 1.6727272727272728, |
| "grad_norm": 0.5784482599606942, |
| "learning_rate": 3.7701335579160147e-05, |
| "loss": 0.2959, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26155000925064087, |
| "step": 920, |
| "valid_targets_mean": 2259.2, |
| "valid_targets_min": 1104 |
| }, |
| { |
| "epoch": 1.6818181818181817, |
| "grad_norm": 0.6257032915898507, |
| "learning_rate": 3.7658951857178544e-05, |
| "loss": 0.2945, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27045929431915283, |
| "step": 925, |
| "valid_targets_mean": 2321.8, |
| "valid_targets_min": 869 |
| }, |
| { |
| "epoch": 1.690909090909091, |
| "grad_norm": 0.47330627816498005, |
| "learning_rate": 3.7616205226434005e-05, |
| "loss": 0.3099, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3237019181251526, |
| "step": 930, |
| "valid_targets_mean": 3809.6, |
| "valid_targets_min": 1843 |
| }, |
| { |
| "epoch": 1.7, |
| "grad_norm": 0.5566942533616496, |
| "learning_rate": 3.7573096565411694e-05, |
| "loss": 0.3062, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3009275496006012, |
| "step": 935, |
| "valid_targets_mean": 2932.1, |
| "valid_targets_min": 513 |
| }, |
| { |
| "epoch": 1.709090909090909, |
| "grad_norm": 0.5553866246810933, |
| "learning_rate": 3.7529626760036814e-05, |
| "loss": 0.2959, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31550776958465576, |
| "step": 940, |
| "valid_targets_mean": 3059.3, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 1.7181818181818183, |
| "grad_norm": 0.5698641256792278, |
| "learning_rate": 3.7485796703656475e-05, |
| "loss": 0.3062, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2999531626701355, |
| "step": 945, |
| "valid_targets_mean": 2500.3, |
| "valid_targets_min": 683 |
| }, |
| { |
| "epoch": 1.7272727272727273, |
| "grad_norm": 0.5049658769969413, |
| "learning_rate": 3.7441607297021254e-05, |
| "loss": 0.2997, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25607892870903015, |
| "step": 950, |
| "valid_targets_mean": 2733.9, |
| "valid_targets_min": 1585 |
| }, |
| { |
| "epoch": 1.7363636363636363, |
| "grad_norm": 0.5106591813226968, |
| "learning_rate": 3.7397059448266786e-05, |
| "loss": 0.2709, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26858973503112793, |
| "step": 955, |
| "valid_targets_mean": 2915.4, |
| "valid_targets_min": 1287 |
| }, |
| { |
| "epoch": 1.7454545454545456, |
| "grad_norm": 0.5583983545217152, |
| "learning_rate": 3.735215407289498e-05, |
| "loss": 0.323, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28842100501060486, |
| "step": 960, |
| "valid_targets_mean": 2971.6, |
| "valid_targets_min": 1471 |
| }, |
| { |
| "epoch": 1.7545454545454544, |
| "grad_norm": 0.7008542358361727, |
| "learning_rate": 3.730689209375533e-05, |
| "loss": 0.3217, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31536877155303955, |
| "step": 965, |
| "valid_targets_mean": 3088.8, |
| "valid_targets_min": 606 |
| }, |
| { |
| "epoch": 1.7636363636363637, |
| "grad_norm": 0.600059036494435, |
| "learning_rate": 3.726127444102583e-05, |
| "loss": 0.3194, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25902581214904785, |
| "step": 970, |
| "valid_targets_mean": 2461.3, |
| "valid_targets_min": 1266 |
| }, |
| { |
| "epoch": 1.7727272727272727, |
| "grad_norm": 0.5146678795415901, |
| "learning_rate": 3.721530205219395e-05, |
| "loss": 0.3125, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2943532168865204, |
| "step": 975, |
| "valid_targets_mean": 2999.1, |
| "valid_targets_min": 1626 |
| }, |
| { |
| "epoch": 1.7818181818181817, |
| "grad_norm": 0.6264974405624288, |
| "learning_rate": 3.716897587203733e-05, |
| "loss": 0.3031, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3075316846370697, |
| "step": 980, |
| "valid_targets_mean": 2197.6, |
| "valid_targets_min": 900 |
| }, |
| { |
| "epoch": 1.790909090909091, |
| "grad_norm": 0.8108650837597063, |
| "learning_rate": 3.712229685260434e-05, |
| "loss": 0.309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32539889216423035, |
| "step": 985, |
| "valid_targets_mean": 2118.2, |
| "valid_targets_min": 935 |
| }, |
| { |
| "epoch": 1.8, |
| "grad_norm": 0.5784815878349024, |
| "learning_rate": 3.707526595319459e-05, |
| "loss": 0.2854, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3252108693122864, |
| "step": 990, |
| "valid_targets_mean": 3014.9, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 1.809090909090909, |
| "grad_norm": 0.5138580399975403, |
| "learning_rate": 3.7027884140339144e-05, |
| "loss": 0.2989, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27519625425338745, |
| "step": 995, |
| "valid_targets_mean": 2874.1, |
| "valid_targets_min": 922 |
| }, |
| { |
| "epoch": 1.8181818181818183, |
| "grad_norm": 0.5033049189948867, |
| "learning_rate": 3.698015238778066e-05, |
| "loss": 0.2913, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.38201528787612915, |
| "step": 1000, |
| "valid_targets_mean": 3500.8, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 1.8272727272727272, |
| "grad_norm": 0.5165562131281177, |
| "learning_rate": 3.693207167645344e-05, |
| "loss": 0.3093, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2646842896938324, |
| "step": 1005, |
| "valid_targets_mean": 2819.9, |
| "valid_targets_min": 529 |
| }, |
| { |
| "epoch": 1.8363636363636364, |
| "grad_norm": 0.5687615871942717, |
| "learning_rate": 3.6883642994463194e-05, |
| "loss": 0.3035, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2546718716621399, |
| "step": 1010, |
| "valid_targets_mean": 2292.6, |
| "valid_targets_min": 1405 |
| }, |
| { |
| "epoch": 1.8454545454545455, |
| "grad_norm": 0.6635357183417174, |
| "learning_rate": 3.6834867337066805e-05, |
| "loss": 0.3032, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3448749780654907, |
| "step": 1015, |
| "valid_targets_mean": 2516.5, |
| "valid_targets_min": 1422 |
| }, |
| { |
| "epoch": 1.8545454545454545, |
| "grad_norm": 0.5378740657548305, |
| "learning_rate": 3.678574570665181e-05, |
| "loss": 0.3104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31566622853279114, |
| "step": 1020, |
| "valid_targets_mean": 2945.9, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 1.8636363636363638, |
| "grad_norm": 0.5035442837344294, |
| "learning_rate": 3.673627911271586e-05, |
| "loss": 0.3294, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2709481716156006, |
| "step": 1025, |
| "valid_targets_mean": 2638.1, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 1.8727272727272726, |
| "grad_norm": 0.5317640331664419, |
| "learning_rate": 3.668646857184591e-05, |
| "loss": 0.3059, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3069203495979309, |
| "step": 1030, |
| "valid_targets_mean": 2766.3, |
| "valid_targets_min": 1116 |
| }, |
| { |
| "epoch": 1.8818181818181818, |
| "grad_norm": 0.5551139811484433, |
| "learning_rate": 3.663631510769739e-05, |
| "loss": 0.314, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2864927053451538, |
| "step": 1035, |
| "valid_targets_mean": 3255.9, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 1.8909090909090909, |
| "grad_norm": 0.5244636101727915, |
| "learning_rate": 3.658581975097311e-05, |
| "loss": 0.3085, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27532529830932617, |
| "step": 1040, |
| "valid_targets_mean": 2916.7, |
| "valid_targets_min": 1303 |
| }, |
| { |
| "epoch": 1.9, |
| "grad_norm": 0.5328499288436037, |
| "learning_rate": 3.653498353940215e-05, |
| "loss": 0.3027, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35427671670913696, |
| "step": 1045, |
| "valid_targets_mean": 3069.6, |
| "valid_targets_min": 1347 |
| }, |
| { |
| "epoch": 1.9090909090909092, |
| "grad_norm": 0.4910307505819529, |
| "learning_rate": 3.648380751771846e-05, |
| "loss": 0.2975, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2594786286354065, |
| "step": 1050, |
| "valid_targets_mean": 2967.8, |
| "valid_targets_min": 1302 |
| }, |
| { |
| "epoch": 1.9181818181818182, |
| "grad_norm": 0.5535789990195293, |
| "learning_rate": 3.6432292737639426e-05, |
| "loss": 0.3227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2898174226284027, |
| "step": 1055, |
| "valid_targets_mean": 2596.1, |
| "valid_targets_min": 1051 |
| }, |
| { |
| "epoch": 1.9272727272727272, |
| "grad_norm": 0.4824675689759683, |
| "learning_rate": 3.638044025784425e-05, |
| "loss": 0.3212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2852433919906616, |
| "step": 1060, |
| "valid_targets_mean": 3208.1, |
| "valid_targets_min": 1167 |
| }, |
| { |
| "epoch": 1.9363636363636365, |
| "grad_norm": 0.578163740136828, |
| "learning_rate": 3.63282511439522e-05, |
| "loss": 0.3112, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3305288851261139, |
| "step": 1065, |
| "valid_targets_mean": 2377.6, |
| "valid_targets_min": 1160 |
| }, |
| { |
| "epoch": 1.9454545454545453, |
| "grad_norm": 0.5864842908917559, |
| "learning_rate": 3.627572646850069e-05, |
| "loss": 0.276, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2787714898586273, |
| "step": 1070, |
| "valid_targets_mean": 2351.8, |
| "valid_targets_min": 944 |
| }, |
| { |
| "epoch": 1.9545454545454546, |
| "grad_norm": 0.6081783924734938, |
| "learning_rate": 3.6222867310923296e-05, |
| "loss": 0.3183, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3721981644630432, |
| "step": 1075, |
| "valid_targets_mean": 3016.4, |
| "valid_targets_min": 1230 |
| }, |
| { |
| "epoch": 1.9636363636363636, |
| "grad_norm": 0.5178622230386561, |
| "learning_rate": 3.6169674757527466e-05, |
| "loss": 0.2887, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2899319529533386, |
| "step": 1080, |
| "valid_targets_mean": 2910.1, |
| "valid_targets_min": 1107 |
| }, |
| { |
| "epoch": 1.9727272727272727, |
| "grad_norm": 0.5345917181059462, |
| "learning_rate": 3.61161499014723e-05, |
| "loss": 0.2973, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25718826055526733, |
| "step": 1085, |
| "valid_targets_mean": 2929.2, |
| "valid_targets_min": 1142 |
| }, |
| { |
| "epoch": 1.981818181818182, |
| "grad_norm": 0.5176401124515433, |
| "learning_rate": 3.606229384274604e-05, |
| "loss": 0.2863, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28826001286506653, |
| "step": 1090, |
| "valid_targets_mean": 2809.6, |
| "valid_targets_min": 1480 |
| }, |
| { |
| "epoch": 1.990909090909091, |
| "grad_norm": 0.5539961041603162, |
| "learning_rate": 3.600810768814345e-05, |
| "loss": 0.3149, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30427122116088867, |
| "step": 1095, |
| "valid_targets_mean": 2813.6, |
| "valid_targets_min": 1552 |
| }, |
| { |
| "epoch": 2.0, |
| "grad_norm": 0.5224215311779324, |
| "learning_rate": 3.595359255124311e-05, |
| "loss": 0.2711, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28612616658210754, |
| "step": 1100, |
| "valid_targets_mean": 2870.6, |
| "valid_targets_min": 719 |
| }, |
| { |
| "epoch": 2.0090909090909093, |
| "grad_norm": 0.489439619755712, |
| "learning_rate": 3.589874955238449e-05, |
| "loss": 0.2496, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2773962616920471, |
| "step": 1105, |
| "valid_targets_mean": 3120.6, |
| "valid_targets_min": 1333 |
| }, |
| { |
| "epoch": 2.018181818181818, |
| "grad_norm": 0.6640014872803903, |
| "learning_rate": 3.5843579818644956e-05, |
| "loss": 0.237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2267443984746933, |
| "step": 1110, |
| "valid_targets_mean": 2743.2, |
| "valid_targets_min": 935 |
| }, |
| { |
| "epoch": 2.0272727272727273, |
| "grad_norm": 0.5835718798395699, |
| "learning_rate": 3.5788084483816587e-05, |
| "loss": 0.2713, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31667560338974, |
| "step": 1115, |
| "valid_targets_mean": 2659.2, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 2.036363636363636, |
| "grad_norm": 0.5465980940064771, |
| "learning_rate": 3.573226468838289e-05, |
| "loss": 0.2622, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2518404722213745, |
| "step": 1120, |
| "valid_targets_mean": 2707.6, |
| "valid_targets_min": 800 |
| }, |
| { |
| "epoch": 2.0454545454545454, |
| "grad_norm": 0.5733257551739418, |
| "learning_rate": 3.567612157949536e-05, |
| "loss": 0.2895, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27450916171073914, |
| "step": 1125, |
| "valid_targets_mean": 2856.4, |
| "valid_targets_min": 789 |
| }, |
| { |
| "epoch": 2.0545454545454547, |
| "grad_norm": 0.5681140687556558, |
| "learning_rate": 3.561965631094988e-05, |
| "loss": 0.2464, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23563730716705322, |
| "step": 1130, |
| "valid_targets_mean": 2212.8, |
| "valid_targets_min": 767 |
| }, |
| { |
| "epoch": 2.0636363636363635, |
| "grad_norm": 0.5556954904745599, |
| "learning_rate": 3.556287004316305e-05, |
| "loss": 0.2762, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2981184124946594, |
| "step": 1135, |
| "valid_targets_mean": 2950.0, |
| "valid_targets_min": 700 |
| }, |
| { |
| "epoch": 2.0727272727272728, |
| "grad_norm": 0.5605759813195871, |
| "learning_rate": 3.5505763943148324e-05, |
| "loss": 0.2757, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26870912313461304, |
| "step": 1140, |
| "valid_targets_mean": 2711.8, |
| "valid_targets_min": 1144 |
| }, |
| { |
| "epoch": 2.081818181818182, |
| "grad_norm": 0.5216825087835475, |
| "learning_rate": 3.544833918449199e-05, |
| "loss": 0.2937, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3655780255794525, |
| "step": 1145, |
| "valid_targets_mean": 3950.1, |
| "valid_targets_min": 856 |
| }, |
| { |
| "epoch": 2.090909090909091, |
| "grad_norm": 0.5613165558947877, |
| "learning_rate": 3.5390596947329124e-05, |
| "loss": 0.2991, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2934229373931885, |
| "step": 1150, |
| "valid_targets_mean": 3625.8, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 2.1, |
| "grad_norm": 0.6361732770074459, |
| "learning_rate": 3.5332538418319254e-05, |
| "loss": 0.2788, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31974083185195923, |
| "step": 1155, |
| "valid_targets_mean": 2393.6, |
| "valid_targets_min": 962 |
| }, |
| { |
| "epoch": 2.109090909090909, |
| "grad_norm": 0.6027622062417266, |
| "learning_rate": 3.527416479062205e-05, |
| "loss": 0.2807, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24740280210971832, |
| "step": 1160, |
| "valid_targets_mean": 2359.5, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 2.118181818181818, |
| "grad_norm": 0.5130832544460253, |
| "learning_rate": 3.521547726387275e-05, |
| "loss": 0.2873, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22652898728847504, |
| "step": 1165, |
| "valid_targets_mean": 2752.1, |
| "valid_targets_min": 1490 |
| }, |
| { |
| "epoch": 2.1272727272727274, |
| "grad_norm": 0.5486053889891421, |
| "learning_rate": 3.515647704415754e-05, |
| "loss": 0.2726, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2480347454547882, |
| "step": 1170, |
| "valid_targets_mean": 2579.8, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 2.1363636363636362, |
| "grad_norm": 0.5188330941180774, |
| "learning_rate": 3.509716534398873e-05, |
| "loss": 0.2801, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2117992341518402, |
| "step": 1175, |
| "valid_targets_mean": 2542.3, |
| "valid_targets_min": 1057 |
| }, |
| { |
| "epoch": 2.1454545454545455, |
| "grad_norm": 0.49929458454016185, |
| "learning_rate": 3.503754338227989e-05, |
| "loss": 0.2878, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27674928307533264, |
| "step": 1180, |
| "valid_targets_mean": 3255.7, |
| "valid_targets_min": 1464 |
| }, |
| { |
| "epoch": 2.1545454545454543, |
| "grad_norm": 0.5699600398030361, |
| "learning_rate": 3.497761238432073e-05, |
| "loss": 0.2552, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24738357961177826, |
| "step": 1185, |
| "valid_targets_mean": 2535.2, |
| "valid_targets_min": 954 |
| }, |
| { |
| "epoch": 2.1636363636363636, |
| "grad_norm": 0.5402480392886817, |
| "learning_rate": 3.4917373581752e-05, |
| "loss": 0.2901, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3008500933647156, |
| "step": 1190, |
| "valid_targets_mean": 3093.9, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 2.172727272727273, |
| "grad_norm": 0.6485393078998691, |
| "learning_rate": 3.4856828212540094e-05, |
| "loss": 0.2892, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3143201768398285, |
| "step": 1195, |
| "valid_targets_mean": 2453.3, |
| "valid_targets_min": 600 |
| }, |
| { |
| "epoch": 2.1818181818181817, |
| "grad_norm": 0.6195348387664918, |
| "learning_rate": 3.4795977520951684e-05, |
| "loss": 0.2745, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3513163626194, |
| "step": 1200, |
| "valid_targets_mean": 2938.9, |
| "valid_targets_min": 1226 |
| }, |
| { |
| "epoch": 2.190909090909091, |
| "grad_norm": 0.5549330178340143, |
| "learning_rate": 3.47348227575281e-05, |
| "loss": 0.261, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27062830328941345, |
| "step": 1205, |
| "valid_targets_mean": 3571.6, |
| "valid_targets_min": 655 |
| }, |
| { |
| "epoch": 2.2, |
| "grad_norm": 0.7062538572938993, |
| "learning_rate": 3.467336517905966e-05, |
| "loss": 0.266, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28915590047836304, |
| "step": 1210, |
| "valid_targets_mean": 2486.9, |
| "valid_targets_min": 695 |
| }, |
| { |
| "epoch": 2.209090909090909, |
| "grad_norm": 0.5619405166344583, |
| "learning_rate": 3.46116060485598e-05, |
| "loss": 0.2642, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3145231604576111, |
| "step": 1215, |
| "valid_targets_mean": 2945.8, |
| "valid_targets_min": 664 |
| }, |
| { |
| "epoch": 2.2181818181818183, |
| "grad_norm": 0.5371710217063721, |
| "learning_rate": 3.4549546635239167e-05, |
| "loss": 0.2721, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2760411202907562, |
| "step": 1220, |
| "valid_targets_mean": 3207.3, |
| "valid_targets_min": 1323 |
| }, |
| { |
| "epoch": 2.227272727272727, |
| "grad_norm": 0.5888575885460174, |
| "learning_rate": 3.448718821447953e-05, |
| "loss": 0.2994, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2726590931415558, |
| "step": 1225, |
| "valid_targets_mean": 2388.1, |
| "valid_targets_min": 529 |
| }, |
| { |
| "epoch": 2.2363636363636363, |
| "grad_norm": 0.5202633657149802, |
| "learning_rate": 3.442453206780751e-05, |
| "loss": 0.2745, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2371698021888733, |
| "step": 1230, |
| "valid_targets_mean": 2776.1, |
| "valid_targets_min": 826 |
| }, |
| { |
| "epoch": 2.2454545454545456, |
| "grad_norm": 0.5190998588011124, |
| "learning_rate": 3.4361579482868325e-05, |
| "loss": 0.2182, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19271951913833618, |
| "step": 1235, |
| "valid_targets_mean": 2445.1, |
| "valid_targets_min": 1178 |
| }, |
| { |
| "epoch": 2.2545454545454544, |
| "grad_norm": 0.6415806356522294, |
| "learning_rate": 3.429833175339927e-05, |
| "loss": 0.2732, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3027918338775635, |
| "step": 1240, |
| "valid_targets_mean": 2446.1, |
| "valid_targets_min": 562 |
| }, |
| { |
| "epoch": 2.2636363636363637, |
| "grad_norm": 0.5859047596126794, |
| "learning_rate": 3.423479017920317e-05, |
| "loss": 0.2852, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3389890789985657, |
| "step": 1245, |
| "valid_targets_mean": 2799.6, |
| "valid_targets_min": 1142 |
| }, |
| { |
| "epoch": 2.2727272727272725, |
| "grad_norm": 0.8035925762795723, |
| "learning_rate": 3.4170956066121616e-05, |
| "loss": 0.2684, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2345964014530182, |
| "step": 1250, |
| "valid_targets_mean": 2630.6, |
| "valid_targets_min": 1152 |
| }, |
| { |
| "epoch": 2.2818181818181817, |
| "grad_norm": 0.578674528171854, |
| "learning_rate": 3.410683072600818e-05, |
| "loss": 0.2727, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2537236511707306, |
| "step": 1255, |
| "valid_targets_mean": 2365.8, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 2.290909090909091, |
| "grad_norm": 0.4934272429740994, |
| "learning_rate": 3.4042415476701434e-05, |
| "loss": 0.3008, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2870427966117859, |
| "step": 1260, |
| "valid_targets_mean": 3357.2, |
| "valid_targets_min": 1313 |
| }, |
| { |
| "epoch": 2.3, |
| "grad_norm": 0.5576814977535322, |
| "learning_rate": 3.397771164199787e-05, |
| "loss": 0.2747, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2425975650548935, |
| "step": 1265, |
| "valid_targets_mean": 2477.3, |
| "valid_targets_min": 1004 |
| }, |
| { |
| "epoch": 2.309090909090909, |
| "grad_norm": 0.6455970996280729, |
| "learning_rate": 3.3912720551624684e-05, |
| "loss": 0.2811, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24581097066402435, |
| "step": 1270, |
| "valid_targets_mean": 2579.8, |
| "valid_targets_min": 1020 |
| }, |
| { |
| "epoch": 2.3181818181818183, |
| "grad_norm": 0.5317314295046272, |
| "learning_rate": 3.384744354121246e-05, |
| "loss": 0.2481, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22386619448661804, |
| "step": 1275, |
| "valid_targets_mean": 2630.9, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 2.327272727272727, |
| "grad_norm": 0.6193275788990439, |
| "learning_rate": 3.3781881952267715e-05, |
| "loss": 0.2725, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22538137435913086, |
| "step": 1280, |
| "valid_targets_mean": 2205.0, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 2.3363636363636364, |
| "grad_norm": 0.5519637684834255, |
| "learning_rate": 3.3716037132145354e-05, |
| "loss": 0.292, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32305043935775757, |
| "step": 1285, |
| "valid_targets_mean": 2939.0, |
| "valid_targets_min": 1007 |
| }, |
| { |
| "epoch": 2.3454545454545457, |
| "grad_norm": 0.5178998925686475, |
| "learning_rate": 3.3649910434020934e-05, |
| "loss": 0.2573, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2358517199754715, |
| "step": 1290, |
| "valid_targets_mean": 2688.2, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 2.3545454545454545, |
| "grad_norm": 0.8149568552846596, |
| "learning_rate": 3.35835032168629e-05, |
| "loss": 0.2781, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.35959750413894653, |
| "step": 1295, |
| "valid_targets_mean": 3071.2, |
| "valid_targets_min": 1035 |
| }, |
| { |
| "epoch": 2.3636363636363638, |
| "grad_norm": 1.0108335562213615, |
| "learning_rate": 3.351681684540462e-05, |
| "loss": 0.2922, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3282524347305298, |
| "step": 1300, |
| "valid_targets_mean": 2302.9, |
| "valid_targets_min": 633 |
| }, |
| { |
| "epoch": 2.3727272727272726, |
| "grad_norm": 0.5275055282359982, |
| "learning_rate": 3.3449852690116375e-05, |
| "loss": 0.2849, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2222471833229065, |
| "step": 1305, |
| "valid_targets_mean": 2474.3, |
| "valid_targets_min": 992 |
| }, |
| { |
| "epoch": 2.381818181818182, |
| "grad_norm": 0.5108776081340932, |
| "learning_rate": 3.3382612127177166e-05, |
| "loss": 0.2775, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22455942630767822, |
| "step": 1310, |
| "valid_targets_mean": 2848.2, |
| "valid_targets_min": 1019 |
| }, |
| { |
| "epoch": 2.390909090909091, |
| "grad_norm": 0.540702773277452, |
| "learning_rate": 3.331509653844644e-05, |
| "loss": 0.2676, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2385614961385727, |
| "step": 1315, |
| "valid_targets_mean": 2398.2, |
| "valid_targets_min": 1156 |
| }, |
| { |
| "epoch": 2.4, |
| "grad_norm": 0.5580702423301868, |
| "learning_rate": 3.324730731143571e-05, |
| "loss": 0.2745, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28066539764404297, |
| "step": 1320, |
| "valid_targets_mean": 2742.6, |
| "valid_targets_min": 892 |
| }, |
| { |
| "epoch": 2.409090909090909, |
| "grad_norm": 0.532467694697259, |
| "learning_rate": 3.317924583927999e-05, |
| "loss": 0.263, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26546695828437805, |
| "step": 1325, |
| "valid_targets_mean": 2783.5, |
| "valid_targets_min": 739 |
| }, |
| { |
| "epoch": 2.418181818181818, |
| "grad_norm": 0.5949400757150429, |
| "learning_rate": 3.311091352070924e-05, |
| "loss": 0.2714, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30344507098197937, |
| "step": 1330, |
| "valid_targets_mean": 2494.1, |
| "valid_targets_min": 1160 |
| }, |
| { |
| "epoch": 2.4272727272727272, |
| "grad_norm": 0.6525136381200096, |
| "learning_rate": 3.3042311760019554e-05, |
| "loss": 0.2814, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2595057487487793, |
| "step": 1335, |
| "valid_targets_mean": 3365.3, |
| "valid_targets_min": 1156 |
| }, |
| { |
| "epoch": 2.4363636363636365, |
| "grad_norm": 0.5306637930795548, |
| "learning_rate": 3.297344196704431e-05, |
| "loss": 0.2815, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29292032122612, |
| "step": 1340, |
| "valid_targets_mean": 3093.2, |
| "valid_targets_min": 1225 |
| }, |
| { |
| "epoch": 2.4454545454545453, |
| "grad_norm": 0.5331840161384932, |
| "learning_rate": 3.2904305557125265e-05, |
| "loss": 0.2463, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2697639465332031, |
| "step": 1345, |
| "valid_targets_mean": 3192.3, |
| "valid_targets_min": 1173 |
| }, |
| { |
| "epoch": 2.4545454545454546, |
| "grad_norm": 0.5465573713315527, |
| "learning_rate": 3.2834903951083363e-05, |
| "loss": 0.2679, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28745096921920776, |
| "step": 1350, |
| "valid_targets_mean": 2783.4, |
| "valid_targets_min": 1422 |
| }, |
| { |
| "epoch": 2.463636363636364, |
| "grad_norm": 0.6761778819959714, |
| "learning_rate": 3.27652385751896e-05, |
| "loss": 0.2941, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2797173261642456, |
| "step": 1355, |
| "valid_targets_mean": 2376.6, |
| "valid_targets_min": 1587 |
| }, |
| { |
| "epoch": 2.4727272727272727, |
| "grad_norm": 0.5172807517065136, |
| "learning_rate": 3.269531086113573e-05, |
| "loss": 0.2869, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25817084312438965, |
| "step": 1360, |
| "valid_targets_mean": 3498.8, |
| "valid_targets_min": 1199 |
| }, |
| { |
| "epoch": 2.481818181818182, |
| "grad_norm": 0.4945389497424476, |
| "learning_rate": 3.262512224600478e-05, |
| "loss": 0.2569, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23338064551353455, |
| "step": 1365, |
| "valid_targets_mean": 2960.8, |
| "valid_targets_min": 1387 |
| }, |
| { |
| "epoch": 2.4909090909090907, |
| "grad_norm": 0.46663492782118515, |
| "learning_rate": 3.2554674172241565e-05, |
| "loss": 0.2643, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2113073766231537, |
| "step": 1370, |
| "valid_targets_mean": 2834.1, |
| "valid_targets_min": 1502 |
| }, |
| { |
| "epoch": 2.5, |
| "grad_norm": 0.5271452557335616, |
| "learning_rate": 3.2483968087623026e-05, |
| "loss": 0.289, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2747431993484497, |
| "step": 1375, |
| "valid_targets_mean": 3373.9, |
| "valid_targets_min": 802 |
| }, |
| { |
| "epoch": 2.509090909090909, |
| "grad_norm": 0.5903572935792529, |
| "learning_rate": 3.241300544522848e-05, |
| "loss": 0.2684, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29771339893341064, |
| "step": 1380, |
| "valid_targets_mean": 2850.2, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 2.518181818181818, |
| "grad_norm": 0.48296367582709726, |
| "learning_rate": 3.234178770340975e-05, |
| "loss": 0.2496, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2282666265964508, |
| "step": 1385, |
| "valid_targets_mean": 3135.6, |
| "valid_targets_min": 1104 |
| }, |
| { |
| "epoch": 2.5272727272727273, |
| "grad_norm": 0.575970639650917, |
| "learning_rate": 3.227031632576122e-05, |
| "loss": 0.2749, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2689974904060364, |
| "step": 1390, |
| "valid_targets_mean": 2390.9, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 2.536363636363636, |
| "grad_norm": 0.5132386498577188, |
| "learning_rate": 3.219859278108972e-05, |
| "loss": 0.2619, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2247665524482727, |
| "step": 1395, |
| "valid_targets_mean": 2708.2, |
| "valid_targets_min": 824 |
| }, |
| { |
| "epoch": 2.5454545454545454, |
| "grad_norm": 0.5374983895142559, |
| "learning_rate": 3.212661854338438e-05, |
| "loss": 0.2555, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2777082920074463, |
| "step": 1400, |
| "valid_targets_mean": 3208.6, |
| "valid_targets_min": 1493 |
| }, |
| { |
| "epoch": 2.5545454545454547, |
| "grad_norm": 0.4986505149332797, |
| "learning_rate": 3.20543950917863e-05, |
| "loss": 0.24, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17187553644180298, |
| "step": 1405, |
| "valid_targets_mean": 2338.7, |
| "valid_targets_min": 1317 |
| }, |
| { |
| "epoch": 2.5636363636363635, |
| "grad_norm": 0.5088261424244129, |
| "learning_rate": 3.1981923910558164e-05, |
| "loss": 0.2827, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2423439472913742, |
| "step": 1410, |
| "valid_targets_mean": 2774.1, |
| "valid_targets_min": 1221 |
| }, |
| { |
| "epoch": 2.5727272727272728, |
| "grad_norm": 0.5190336610600647, |
| "learning_rate": 3.190920648905376e-05, |
| "loss": 0.2954, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3125911056995392, |
| "step": 1415, |
| "valid_targets_mean": 3351.1, |
| "valid_targets_min": 825 |
| }, |
| { |
| "epoch": 2.581818181818182, |
| "grad_norm": 0.5811156813435989, |
| "learning_rate": 3.183624432168736e-05, |
| "loss": 0.2621, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2840062081813812, |
| "step": 1420, |
| "valid_targets_mean": 2648.9, |
| "valid_targets_min": 1198 |
| }, |
| { |
| "epoch": 2.590909090909091, |
| "grad_norm": 0.44510133678522557, |
| "learning_rate": 3.1763038907902976e-05, |
| "loss": 0.269, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22076770663261414, |
| "step": 1425, |
| "valid_targets_mean": 3170.4, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 2.6, |
| "grad_norm": 0.5423101344043437, |
| "learning_rate": 3.16895917521436e-05, |
| "loss": 0.2592, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2548125386238098, |
| "step": 1430, |
| "valid_targets_mean": 2706.8, |
| "valid_targets_min": 957 |
| }, |
| { |
| "epoch": 2.6090909090909093, |
| "grad_norm": 0.5415408039241957, |
| "learning_rate": 3.161590436382023e-05, |
| "loss": 0.2596, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23154941201210022, |
| "step": 1435, |
| "valid_targets_mean": 2627.1, |
| "valid_targets_min": 1817 |
| }, |
| { |
| "epoch": 2.618181818181818, |
| "grad_norm": 0.5442786567017693, |
| "learning_rate": 3.1541978257280915e-05, |
| "loss": 0.2451, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2088831067085266, |
| "step": 1440, |
| "valid_targets_mean": 2507.2, |
| "valid_targets_min": 1167 |
| }, |
| { |
| "epoch": 2.6272727272727274, |
| "grad_norm": 0.507250652617475, |
| "learning_rate": 3.1467814951779564e-05, |
| "loss": 0.2837, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3032193183898926, |
| "step": 1445, |
| "valid_targets_mean": 2987.0, |
| "valid_targets_min": 855 |
| }, |
| { |
| "epoch": 2.6363636363636362, |
| "grad_norm": 0.5572879239252921, |
| "learning_rate": 3.139341597144478e-05, |
| "loss": 0.2806, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2529670000076294, |
| "step": 1450, |
| "valid_targets_mean": 2529.0, |
| "valid_targets_min": 1000 |
| }, |
| { |
| "epoch": 2.6454545454545455, |
| "grad_norm": 0.5153350157095853, |
| "learning_rate": 3.13187828452485e-05, |
| "loss": 0.2446, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21580073237419128, |
| "step": 1455, |
| "valid_targets_mean": 2435.6, |
| "valid_targets_min": 1103 |
| }, |
| { |
| "epoch": 2.6545454545454543, |
| "grad_norm": 0.5170627803352125, |
| "learning_rate": 3.1243917106974583e-05, |
| "loss": 0.2974, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26863837242126465, |
| "step": 1460, |
| "valid_targets_mean": 2920.1, |
| "valid_targets_min": 1100 |
| }, |
| { |
| "epoch": 2.6636363636363636, |
| "grad_norm": 0.5155469818430718, |
| "learning_rate": 3.116882029518732e-05, |
| "loss": 0.2693, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22775819897651672, |
| "step": 1465, |
| "valid_targets_mean": 2753.8, |
| "valid_targets_min": 1362 |
| }, |
| { |
| "epoch": 2.672727272727273, |
| "grad_norm": 0.6111239303899411, |
| "learning_rate": 3.109349395319976e-05, |
| "loss": 0.2491, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28220629692077637, |
| "step": 1470, |
| "valid_targets_mean": 2454.9, |
| "valid_targets_min": 760 |
| }, |
| { |
| "epoch": 2.6818181818181817, |
| "grad_norm": 0.584045217206336, |
| "learning_rate": 3.101793962904205e-05, |
| "loss": 0.2742, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27724209427833557, |
| "step": 1475, |
| "valid_targets_mean": 2150.0, |
| "valid_targets_min": 733 |
| }, |
| { |
| "epoch": 2.690909090909091, |
| "grad_norm": 0.5593938542064445, |
| "learning_rate": 3.094215887542957e-05, |
| "loss": 0.2819, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26411521434783936, |
| "step": 1480, |
| "valid_targets_mean": 2727.2, |
| "valid_targets_min": 1089 |
| }, |
| { |
| "epoch": 2.7, |
| "grad_norm": 0.6659207240617939, |
| "learning_rate": 3.086615324973107e-05, |
| "loss": 0.3038, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2924853563308716, |
| "step": 1485, |
| "valid_targets_mean": 2503.4, |
| "valid_targets_min": 916 |
| }, |
| { |
| "epoch": 2.709090909090909, |
| "grad_norm": 0.5463125878243501, |
| "learning_rate": 3.07899243139366e-05, |
| "loss": 0.2911, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.229422464966774, |
| "step": 1490, |
| "valid_targets_mean": 2213.2, |
| "valid_targets_min": 1103 |
| }, |
| { |
| "epoch": 2.7181818181818183, |
| "grad_norm": 0.5292840040904653, |
| "learning_rate": 3.0713473634625507e-05, |
| "loss": 0.2704, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25266051292419434, |
| "step": 1495, |
| "valid_targets_mean": 2959.2, |
| "valid_targets_min": 1173 |
| }, |
| { |
| "epoch": 2.7272727272727275, |
| "grad_norm": 0.6300182495568534, |
| "learning_rate": 3.0636802782934146e-05, |
| "loss": 0.2662, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2884886860847473, |
| "step": 1500, |
| "valid_targets_mean": 2169.2, |
| "valid_targets_min": 637 |
| }, |
| { |
| "epoch": 2.7363636363636363, |
| "grad_norm": 0.5273137367601014, |
| "learning_rate": 3.055991333452364e-05, |
| "loss": 0.2569, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2575647234916687, |
| "step": 1505, |
| "valid_targets_mean": 2934.0, |
| "valid_targets_min": 1469 |
| }, |
| { |
| "epoch": 2.7454545454545456, |
| "grad_norm": 0.48641356369382016, |
| "learning_rate": 3.0482806869547495e-05, |
| "loss": 0.2593, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2414964735507965, |
| "step": 1510, |
| "valid_targets_mean": 3465.9, |
| "valid_targets_min": 1062 |
| }, |
| { |
| "epoch": 2.7545454545454544, |
| "grad_norm": 0.5613437872799717, |
| "learning_rate": 3.0405484972619116e-05, |
| "loss": 0.2689, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.329323410987854, |
| "step": 1515, |
| "valid_targets_mean": 3002.0, |
| "valid_targets_min": 1267 |
| }, |
| { |
| "epoch": 2.7636363636363637, |
| "grad_norm": 0.5178590049360541, |
| "learning_rate": 3.0327949232779242e-05, |
| "loss": 0.2556, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24617408215999603, |
| "step": 1520, |
| "valid_targets_mean": 2753.1, |
| "valid_targets_min": 384 |
| }, |
| { |
| "epoch": 2.7727272727272725, |
| "grad_norm": 0.5209426517091212, |
| "learning_rate": 3.0250201243463297e-05, |
| "loss": 0.2737, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2629636824131012, |
| "step": 1525, |
| "valid_targets_mean": 2723.1, |
| "valid_targets_min": 983 |
| }, |
| { |
| "epoch": 2.7818181818181817, |
| "grad_norm": 0.48623271439168053, |
| "learning_rate": 3.0172242602468637e-05, |
| "loss": 0.2595, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25595349073410034, |
| "step": 1530, |
| "valid_targets_mean": 3133.2, |
| "valid_targets_min": 2159 |
| }, |
| { |
| "epoch": 2.790909090909091, |
| "grad_norm": 0.524320685222588, |
| "learning_rate": 3.009407491192172e-05, |
| "loss": 0.3067, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3319476246833801, |
| "step": 1535, |
| "valid_targets_mean": 3539.6, |
| "valid_targets_min": 1160 |
| }, |
| { |
| "epoch": 2.8, |
| "grad_norm": 0.5295558188317099, |
| "learning_rate": 3.0015699778245177e-05, |
| "loss": 0.2797, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2332993447780609, |
| "step": 1540, |
| "valid_targets_mean": 2543.6, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 2.809090909090909, |
| "grad_norm": 0.5827710149024911, |
| "learning_rate": 2.9937118812124796e-05, |
| "loss": 0.2836, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3239750266075134, |
| "step": 1545, |
| "valid_targets_mean": 3131.9, |
| "valid_targets_min": 1657 |
| }, |
| { |
| "epoch": 2.8181818181818183, |
| "grad_norm": 0.5971029640504223, |
| "learning_rate": 2.9858333628476423e-05, |
| "loss": 0.2944, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.312186062335968, |
| "step": 1550, |
| "valid_targets_mean": 2550.8, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 2.827272727272727, |
| "grad_norm": 0.6142176132505575, |
| "learning_rate": 2.977934584641278e-05, |
| "loss": 0.3037, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34188252687454224, |
| "step": 1555, |
| "valid_targets_mean": 3110.6, |
| "valid_targets_min": 1112 |
| }, |
| { |
| "epoch": 2.8363636363636364, |
| "grad_norm": 0.6646128032944204, |
| "learning_rate": 2.9700157089210174e-05, |
| "loss": 0.2956, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31395840644836426, |
| "step": 1560, |
| "valid_targets_mean": 2368.2, |
| "valid_targets_min": 659 |
| }, |
| { |
| "epoch": 2.8454545454545457, |
| "grad_norm": 0.8465812739684272, |
| "learning_rate": 2.9620768984275163e-05, |
| "loss": 0.2571, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2895926237106323, |
| "step": 1565, |
| "valid_targets_mean": 3289.6, |
| "valid_targets_min": 1496 |
| }, |
| { |
| "epoch": 2.8545454545454545, |
| "grad_norm": 0.5724604891224522, |
| "learning_rate": 2.9541183163111076e-05, |
| "loss": 0.2794, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2987874746322632, |
| "step": 1570, |
| "valid_targets_mean": 3033.1, |
| "valid_targets_min": 878 |
| }, |
| { |
| "epoch": 2.8636363636363638, |
| "grad_norm": 0.5220424169586242, |
| "learning_rate": 2.9461401261284536e-05, |
| "loss": 0.2875, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28746628761291504, |
| "step": 1575, |
| "valid_targets_mean": 3495.4, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 2.8727272727272726, |
| "grad_norm": 0.540257485034313, |
| "learning_rate": 2.9381424918391775e-05, |
| "loss": 0.2746, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2522783875465393, |
| "step": 1580, |
| "valid_targets_mean": 2805.1, |
| "valid_targets_min": 1328 |
| }, |
| { |
| "epoch": 2.881818181818182, |
| "grad_norm": 0.6350255065226401, |
| "learning_rate": 2.9301255778025014e-05, |
| "loss": 0.2602, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24785231053829193, |
| "step": 1585, |
| "valid_targets_mean": 3572.1, |
| "valid_targets_min": 846 |
| }, |
| { |
| "epoch": 2.8909090909090907, |
| "grad_norm": 0.6254327695227894, |
| "learning_rate": 2.9220895487738627e-05, |
| "loss": 0.267, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3446647524833679, |
| "step": 1590, |
| "valid_targets_mean": 2484.1, |
| "valid_targets_min": 1358 |
| }, |
| { |
| "epoch": 2.9, |
| "grad_norm": 0.5107669205847477, |
| "learning_rate": 2.9140345699015328e-05, |
| "loss": 0.2796, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2758895754814148, |
| "step": 1595, |
| "valid_targets_mean": 2978.6, |
| "valid_targets_min": 1163 |
| }, |
| { |
| "epoch": 2.909090909090909, |
| "grad_norm": 0.5591042375626212, |
| "learning_rate": 2.905960806723219e-05, |
| "loss": 0.2731, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.37159761786460876, |
| "step": 1600, |
| "valid_targets_mean": 3184.2, |
| "valid_targets_min": 509 |
| }, |
| { |
| "epoch": 2.918181818181818, |
| "grad_norm": 0.49905659297103566, |
| "learning_rate": 2.8978684251626652e-05, |
| "loss": 0.2982, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26265135407447815, |
| "step": 1605, |
| "valid_targets_mean": 3049.9, |
| "valid_targets_min": 845 |
| }, |
| { |
| "epoch": 2.9272727272727272, |
| "grad_norm": 0.507734740657102, |
| "learning_rate": 2.8897575915262418e-05, |
| "loss": 0.2519, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19236379861831665, |
| "step": 1610, |
| "valid_targets_mean": 2568.4, |
| "valid_targets_min": 1013 |
| }, |
| { |
| "epoch": 2.9363636363636365, |
| "grad_norm": 0.6471719002391192, |
| "learning_rate": 2.8816284724995273e-05, |
| "loss": 0.2911, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29607704281806946, |
| "step": 1615, |
| "valid_targets_mean": 2066.0, |
| "valid_targets_min": 1141 |
| }, |
| { |
| "epoch": 2.9454545454545453, |
| "grad_norm": 0.5250874195229503, |
| "learning_rate": 2.8734812351438823e-05, |
| "loss": 0.264, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2234722077846527, |
| "step": 1620, |
| "valid_targets_mean": 2377.8, |
| "valid_targets_min": 1175 |
| }, |
| { |
| "epoch": 2.9545454545454546, |
| "grad_norm": 0.5143467837243642, |
| "learning_rate": 2.8653160468930168e-05, |
| "loss": 0.27, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28245672583580017, |
| "step": 1625, |
| "valid_targets_mean": 3324.6, |
| "valid_targets_min": 890 |
| }, |
| { |
| "epoch": 2.963636363636364, |
| "grad_norm": 0.5154121964083332, |
| "learning_rate": 2.85713307554955e-05, |
| "loss": 0.2745, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2525605261325836, |
| "step": 1630, |
| "valid_targets_mean": 2891.5, |
| "valid_targets_min": 1136 |
| }, |
| { |
| "epoch": 2.9727272727272727, |
| "grad_norm": 0.6206896766766579, |
| "learning_rate": 2.8489324892815604e-05, |
| "loss": 0.2919, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28195661306381226, |
| "step": 1635, |
| "valid_targets_mean": 2282.6, |
| "valid_targets_min": 1306 |
| }, |
| { |
| "epoch": 2.981818181818182, |
| "grad_norm": 0.5700730166353059, |
| "learning_rate": 2.8407144566191315e-05, |
| "loss": 0.2677, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27205920219421387, |
| "step": 1640, |
| "valid_targets_mean": 2825.0, |
| "valid_targets_min": 929 |
| }, |
| { |
| "epoch": 2.990909090909091, |
| "grad_norm": 0.6002367822037568, |
| "learning_rate": 2.8324791464508856e-05, |
| "loss": 0.285, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3318069577217102, |
| "step": 1645, |
| "valid_targets_mean": 2618.8, |
| "valid_targets_min": 1390 |
| }, |
| { |
| "epoch": 3.0, |
| "grad_norm": 0.4837522767492499, |
| "learning_rate": 2.824226728020516e-05, |
| "loss": 0.255, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20618417859077454, |
| "step": 1650, |
| "valid_targets_mean": 2752.9, |
| "valid_targets_min": 732 |
| }, |
| { |
| "epoch": 3.0090909090909093, |
| "grad_norm": 0.5814855327909163, |
| "learning_rate": 2.8159573709233074e-05, |
| "loss": 0.2716, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2657179832458496, |
| "step": 1655, |
| "valid_targets_mean": 2486.8, |
| "valid_targets_min": 860 |
| }, |
| { |
| "epoch": 3.018181818181818, |
| "grad_norm": 0.5673715077261331, |
| "learning_rate": 2.80767124510265e-05, |
| "loss": 0.2476, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2121908962726593, |
| "step": 1660, |
| "valid_targets_mean": 2472.4, |
| "valid_targets_min": 1184 |
| }, |
| { |
| "epoch": 3.0272727272727273, |
| "grad_norm": 0.5580763685369373, |
| "learning_rate": 2.7993685208465483e-05, |
| "loss": 0.2784, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2324039787054062, |
| "step": 1665, |
| "valid_targets_mean": 2844.7, |
| "valid_targets_min": 1310 |
| }, |
| { |
| "epoch": 3.036363636363636, |
| "grad_norm": 0.6610751547581967, |
| "learning_rate": 2.7910493687841213e-05, |
| "loss": 0.2578, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27891236543655396, |
| "step": 1670, |
| "valid_targets_mean": 2515.8, |
| "valid_targets_min": 1426 |
| }, |
| { |
| "epoch": 3.0454545454545454, |
| "grad_norm": 0.592826560091231, |
| "learning_rate": 2.7827139598820947e-05, |
| "loss": 0.2634, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23966637253761292, |
| "step": 1675, |
| "valid_targets_mean": 2399.9, |
| "valid_targets_min": 1072 |
| }, |
| { |
| "epoch": 3.0545454545454547, |
| "grad_norm": 0.5627543784994, |
| "learning_rate": 2.774362465441288e-05, |
| "loss": 0.2551, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2496371567249298, |
| "step": 1680, |
| "valid_targets_mean": 2805.5, |
| "valid_targets_min": 953 |
| }, |
| { |
| "epoch": 3.0636363636363635, |
| "grad_norm": 0.5536164197921297, |
| "learning_rate": 2.7659950570930956e-05, |
| "loss": 0.2726, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27527642250061035, |
| "step": 1685, |
| "valid_targets_mean": 2894.5, |
| "valid_targets_min": 1135 |
| }, |
| { |
| "epoch": 3.0727272727272728, |
| "grad_norm": 0.7095322622012897, |
| "learning_rate": 2.7576119067959565e-05, |
| "loss": 0.2634, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30722740292549133, |
| "step": 1690, |
| "valid_targets_mean": 3261.9, |
| "valid_targets_min": 1141 |
| }, |
| { |
| "epoch": 3.081818181818182, |
| "grad_norm": 0.4984100254988549, |
| "learning_rate": 2.7492131868318247e-05, |
| "loss": 0.2476, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20504172146320343, |
| "step": 1695, |
| "valid_targets_mean": 3124.6, |
| "valid_targets_min": 1300 |
| }, |
| { |
| "epoch": 3.090909090909091, |
| "grad_norm": 0.5730280754927857, |
| "learning_rate": 2.7407990698026227e-05, |
| "loss": 0.2431, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2182716727256775, |
| "step": 1700, |
| "valid_targets_mean": 2661.8, |
| "valid_targets_min": 839 |
| }, |
| { |
| "epoch": 3.1, |
| "grad_norm": 0.5750248943907733, |
| "learning_rate": 2.7323697286266998e-05, |
| "loss": 0.2766, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24978840351104736, |
| "step": 1705, |
| "valid_targets_mean": 3135.4, |
| "valid_targets_min": 1387 |
| }, |
| { |
| "epoch": 3.109090909090909, |
| "grad_norm": 0.5049945154865161, |
| "learning_rate": 2.7239253365352774e-05, |
| "loss": 0.2345, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19322291016578674, |
| "step": 1710, |
| "valid_targets_mean": 2814.9, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 3.118181818181818, |
| "grad_norm": 0.6782663119949206, |
| "learning_rate": 2.7154660670688867e-05, |
| "loss": 0.2386, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19753289222717285, |
| "step": 1715, |
| "valid_targets_mean": 2069.5, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 3.1272727272727274, |
| "grad_norm": 0.5456342685836792, |
| "learning_rate": 2.706992094073803e-05, |
| "loss": 0.2509, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2218984067440033, |
| "step": 1720, |
| "valid_targets_mean": 2546.0, |
| "valid_targets_min": 452 |
| }, |
| { |
| "epoch": 3.1363636363636362, |
| "grad_norm": 0.5152857788301313, |
| "learning_rate": 2.6985035916984746e-05, |
| "loss": 0.2721, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30218735337257385, |
| "step": 1725, |
| "valid_targets_mean": 3570.6, |
| "valid_targets_min": 1913 |
| }, |
| { |
| "epoch": 3.1454545454545455, |
| "grad_norm": 0.6288155708565155, |
| "learning_rate": 2.6900007343899414e-05, |
| "loss": 0.2172, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20587828755378723, |
| "step": 1730, |
| "valid_targets_mean": 2014.5, |
| "valid_targets_min": 944 |
| }, |
| { |
| "epoch": 3.1545454545454543, |
| "grad_norm": 0.5998489889629469, |
| "learning_rate": 2.6814836968902535e-05, |
| "loss": 0.2578, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24592693150043488, |
| "step": 1735, |
| "valid_targets_mean": 2606.9, |
| "valid_targets_min": 1093 |
| }, |
| { |
| "epoch": 3.1636363636363636, |
| "grad_norm": 0.6757102777347078, |
| "learning_rate": 2.6729526542328755e-05, |
| "loss": 0.2352, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29409754276275635, |
| "step": 1740, |
| "valid_targets_mean": 2323.0, |
| "valid_targets_min": 862 |
| }, |
| { |
| "epoch": 3.172727272727273, |
| "grad_norm": 0.5086126031637096, |
| "learning_rate": 2.6644077817390933e-05, |
| "loss": 0.2697, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2759111523628235, |
| "step": 1745, |
| "valid_targets_mean": 3899.1, |
| "valid_targets_min": 1370 |
| }, |
| { |
| "epoch": 3.1818181818181817, |
| "grad_norm": 0.9238879483582045, |
| "learning_rate": 2.6558492550144092e-05, |
| "loss": 0.2688, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2706644535064697, |
| "step": 1750, |
| "valid_targets_mean": 2931.5, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 3.190909090909091, |
| "grad_norm": 0.6240061533266604, |
| "learning_rate": 2.6472772499449323e-05, |
| "loss": 0.2734, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29832810163497925, |
| "step": 1755, |
| "valid_targets_mean": 2611.0, |
| "valid_targets_min": 1297 |
| }, |
| { |
| "epoch": 3.2, |
| "grad_norm": 0.6170619826973041, |
| "learning_rate": 2.6386919426937655e-05, |
| "loss": 0.2465, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23452681303024292, |
| "step": 1760, |
| "valid_targets_mean": 2523.8, |
| "valid_targets_min": 694 |
| }, |
| { |
| "epoch": 3.209090909090909, |
| "grad_norm": 0.5500235152257905, |
| "learning_rate": 2.6300935096973858e-05, |
| "loss": 0.2421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2703178822994232, |
| "step": 1765, |
| "valid_targets_mean": 3243.7, |
| "valid_targets_min": 1548 |
| }, |
| { |
| "epoch": 3.2181818181818183, |
| "grad_norm": 0.5775975150590328, |
| "learning_rate": 2.6214821276620157e-05, |
| "loss": 0.2187, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20802685618400574, |
| "step": 1770, |
| "valid_targets_mean": 2234.9, |
| "valid_targets_min": 856 |
| }, |
| { |
| "epoch": 3.227272727272727, |
| "grad_norm": 0.513453903766835, |
| "learning_rate": 2.6128579735599924e-05, |
| "loss": 0.247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23952774703502655, |
| "step": 1775, |
| "valid_targets_mean": 2953.9, |
| "valid_targets_min": 1459 |
| }, |
| { |
| "epoch": 3.2363636363636363, |
| "grad_norm": 0.6551665130118128, |
| "learning_rate": 2.6042212246261337e-05, |
| "loss": 0.2529, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26932787895202637, |
| "step": 1780, |
| "valid_targets_mean": 2470.1, |
| "valid_targets_min": 637 |
| }, |
| { |
| "epoch": 3.2454545454545456, |
| "grad_norm": 0.5737639500767836, |
| "learning_rate": 2.595572058354092e-05, |
| "loss": 0.2277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23401257395744324, |
| "step": 1785, |
| "valid_targets_mean": 2675.4, |
| "valid_targets_min": 825 |
| }, |
| { |
| "epoch": 3.2545454545454544, |
| "grad_norm": 0.6311746143624856, |
| "learning_rate": 2.5869106524927096e-05, |
| "loss": 0.2365, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26091697812080383, |
| "step": 1790, |
| "valid_targets_mean": 2271.6, |
| "valid_targets_min": 963 |
| }, |
| { |
| "epoch": 3.2636363636363637, |
| "grad_norm": 0.6612760745498111, |
| "learning_rate": 2.5782371850423627e-05, |
| "loss": 0.2361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18565328419208527, |
| "step": 1795, |
| "valid_targets_mean": 2437.2, |
| "valid_targets_min": 1132 |
| }, |
| { |
| "epoch": 3.2727272727272725, |
| "grad_norm": 0.5384296217759248, |
| "learning_rate": 2.5695518342513047e-05, |
| "loss": 0.2347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22534595429897308, |
| "step": 1800, |
| "valid_targets_mean": 2859.2, |
| "valid_targets_min": 1080 |
| }, |
| { |
| "epoch": 3.2818181818181817, |
| "grad_norm": 0.6361683766746276, |
| "learning_rate": 2.5608547786120056e-05, |
| "loss": 0.2424, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32746827602386475, |
| "step": 1805, |
| "valid_targets_mean": 2984.2, |
| "valid_targets_min": 796 |
| }, |
| { |
| "epoch": 3.290909090909091, |
| "grad_norm": 0.617580783601667, |
| "learning_rate": 2.55214619685748e-05, |
| "loss": 0.2612, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22853900492191315, |
| "step": 1810, |
| "valid_targets_mean": 2373.4, |
| "valid_targets_min": 676 |
| }, |
| { |
| "epoch": 3.3, |
| "grad_norm": 0.534570867677942, |
| "learning_rate": 2.5434262679576157e-05, |
| "loss": 0.2632, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18518221378326416, |
| "step": 1815, |
| "valid_targets_mean": 2618.7, |
| "valid_targets_min": 1816 |
| }, |
| { |
| "epoch": 3.309090909090909, |
| "grad_norm": 0.6257866478906633, |
| "learning_rate": 2.5346951711154946e-05, |
| "loss": 0.2145, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2663920521736145, |
| "step": 1820, |
| "valid_targets_mean": 2585.3, |
| "valid_targets_min": 1607 |
| }, |
| { |
| "epoch": 3.3181818181818183, |
| "grad_norm": 0.6910737869035166, |
| "learning_rate": 2.5259530857637125e-05, |
| "loss": 0.2338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2623439431190491, |
| "step": 1825, |
| "valid_targets_mean": 3113.5, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 3.327272727272727, |
| "grad_norm": 0.6127138768541329, |
| "learning_rate": 2.5172001915606883e-05, |
| "loss": 0.2576, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32933688163757324, |
| "step": 1830, |
| "valid_targets_mean": 3152.8, |
| "valid_targets_min": 987 |
| }, |
| { |
| "epoch": 3.3363636363636364, |
| "grad_norm": 0.6045062207311135, |
| "learning_rate": 2.5084366683869746e-05, |
| "loss": 0.2338, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22119587659835815, |
| "step": 1835, |
| "valid_targets_mean": 2336.5, |
| "valid_targets_min": 1033 |
| }, |
| { |
| "epoch": 3.3454545454545457, |
| "grad_norm": 0.559110510072633, |
| "learning_rate": 2.4996626963415577e-05, |
| "loss": 0.2568, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2486972063779831, |
| "step": 1840, |
| "valid_targets_mean": 3022.8, |
| "valid_targets_min": 1146 |
| }, |
| { |
| "epoch": 3.3545454545454545, |
| "grad_norm": 0.6611782439671358, |
| "learning_rate": 2.4908784557381616e-05, |
| "loss": 0.2635, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23325376212596893, |
| "step": 1845, |
| "valid_targets_mean": 2603.2, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 3.3636363636363638, |
| "grad_norm": 0.592192383324759, |
| "learning_rate": 2.4820841271015364e-05, |
| "loss": 0.2374, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2461732029914856, |
| "step": 1850, |
| "valid_targets_mean": 2866.9, |
| "valid_targets_min": 602 |
| }, |
| { |
| "epoch": 3.3727272727272726, |
| "grad_norm": 0.6205836521900234, |
| "learning_rate": 2.4732798911637525e-05, |
| "loss": 0.2649, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2674775719642639, |
| "step": 1855, |
| "valid_targets_mean": 2762.9, |
| "valid_targets_min": 1064 |
| }, |
| { |
| "epoch": 3.381818181818182, |
| "grad_norm": 0.5875967584148674, |
| "learning_rate": 2.4644659288604853e-05, |
| "loss": 0.2391, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20846202969551086, |
| "step": 1860, |
| "valid_targets_mean": 2467.1, |
| "valid_targets_min": 1170 |
| }, |
| { |
| "epoch": 3.390909090909091, |
| "grad_norm": 0.5459079349622694, |
| "learning_rate": 2.4556424213272955e-05, |
| "loss": 0.2392, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18211738765239716, |
| "step": 1865, |
| "valid_targets_mean": 2467.2, |
| "valid_targets_min": 1189 |
| }, |
| { |
| "epoch": 3.4, |
| "grad_norm": 0.6271200037020991, |
| "learning_rate": 2.4468095498959086e-05, |
| "loss": 0.2521, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26200708746910095, |
| "step": 1870, |
| "valid_targets_mean": 2224.4, |
| "valid_targets_min": 789 |
| }, |
| { |
| "epoch": 3.409090909090909, |
| "grad_norm": 0.6109998547395447, |
| "learning_rate": 2.4379674960904867e-05, |
| "loss": 0.2733, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2885187268257141, |
| "step": 1875, |
| "valid_targets_mean": 2711.4, |
| "valid_targets_min": 1059 |
| }, |
| { |
| "epoch": 3.418181818181818, |
| "grad_norm": 0.6505678700735392, |
| "learning_rate": 2.4291164416238994e-05, |
| "loss": 0.252, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2580758333206177, |
| "step": 1880, |
| "valid_targets_mean": 2137.9, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 3.4272727272727272, |
| "grad_norm": 0.5212654882972313, |
| "learning_rate": 2.4202565683939872e-05, |
| "loss": 0.2328, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22886228561401367, |
| "step": 1885, |
| "valid_targets_mean": 3005.4, |
| "valid_targets_min": 1378 |
| }, |
| { |
| "epoch": 3.4363636363636365, |
| "grad_norm": 0.5353927980487402, |
| "learning_rate": 2.411388058479827e-05, |
| "loss": 0.245, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27620387077331543, |
| "step": 1890, |
| "valid_targets_mean": 3240.9, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 3.4454545454545453, |
| "grad_norm": 0.6708922860894471, |
| "learning_rate": 2.402511094137987e-05, |
| "loss": 0.2378, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2250135838985443, |
| "step": 1895, |
| "valid_targets_mean": 2237.9, |
| "valid_targets_min": 1056 |
| }, |
| { |
| "epoch": 3.4545454545454546, |
| "grad_norm": 0.574944483109974, |
| "learning_rate": 2.3936258577987807e-05, |
| "loss": 0.25, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23620785772800446, |
| "step": 1900, |
| "valid_targets_mean": 2808.0, |
| "valid_targets_min": 1004 |
| }, |
| { |
| "epoch": 3.463636363636364, |
| "grad_norm": 0.6494130584639681, |
| "learning_rate": 2.3847325320625223e-05, |
| "loss": 0.2447, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23063689470291138, |
| "step": 1905, |
| "valid_targets_mean": 2713.1, |
| "valid_targets_min": 705 |
| }, |
| { |
| "epoch": 3.4727272727272727, |
| "grad_norm": 0.5567043968943757, |
| "learning_rate": 2.3758312996957676e-05, |
| "loss": 0.2347, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19316011667251587, |
| "step": 1910, |
| "valid_targets_mean": 2498.1, |
| "valid_targets_min": 1214 |
| }, |
| { |
| "epoch": 3.481818181818182, |
| "grad_norm": 0.6610141845828834, |
| "learning_rate": 2.366922343627565e-05, |
| "loss": 0.2653, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27701085805892944, |
| "step": 1915, |
| "valid_targets_mean": 2537.4, |
| "valid_targets_min": 1042 |
| }, |
| { |
| "epoch": 3.4909090909090907, |
| "grad_norm": 0.6783407015279428, |
| "learning_rate": 2.358005846945689e-05, |
| "loss": 0.2613, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.30464065074920654, |
| "step": 1920, |
| "valid_targets_mean": 3082.9, |
| "valid_targets_min": 1161 |
| }, |
| { |
| "epoch": 3.5, |
| "grad_norm": 0.6172433110892577, |
| "learning_rate": 2.349081992892885e-05, |
| "loss": 0.2377, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22736988961696625, |
| "step": 1925, |
| "valid_targets_mean": 2372.8, |
| "valid_targets_min": 1144 |
| }, |
| { |
| "epoch": 3.509090909090909, |
| "grad_norm": 0.6146777979649554, |
| "learning_rate": 2.3401509648630954e-05, |
| "loss": 0.2244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2371920943260193, |
| "step": 1930, |
| "valid_targets_mean": 2275.1, |
| "valid_targets_min": 1084 |
| }, |
| { |
| "epoch": 3.518181818181818, |
| "grad_norm": 0.6145076071221921, |
| "learning_rate": 2.331212946397698e-05, |
| "loss": 0.2315, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27125149965286255, |
| "step": 1935, |
| "valid_targets_mean": 2734.1, |
| "valid_targets_min": 1381 |
| }, |
| { |
| "epoch": 3.5272727272727273, |
| "grad_norm": 0.6210041121322781, |
| "learning_rate": 2.3222681211817287e-05, |
| "loss": 0.2387, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2806406319141388, |
| "step": 1940, |
| "valid_targets_mean": 2847.0, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 3.536363636363636, |
| "grad_norm": 0.6803398191009827, |
| "learning_rate": 2.31331667304011e-05, |
| "loss": 0.2648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2533523440361023, |
| "step": 1945, |
| "valid_targets_mean": 2696.0, |
| "valid_targets_min": 1044 |
| }, |
| { |
| "epoch": 3.5454545454545454, |
| "grad_norm": 0.5679755520761927, |
| "learning_rate": 2.3043587859338735e-05, |
| "loss": 0.2282, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19747194647789001, |
| "step": 1950, |
| "valid_targets_mean": 2299.3, |
| "valid_targets_min": 1356 |
| }, |
| { |
| "epoch": 3.5545454545454547, |
| "grad_norm": 0.5612673714925396, |
| "learning_rate": 2.2953946439563736e-05, |
| "loss": 0.2566, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24574324488639832, |
| "step": 1955, |
| "valid_targets_mean": 3264.1, |
| "valid_targets_min": 854 |
| }, |
| { |
| "epoch": 3.5636363636363635, |
| "grad_norm": 0.41355474011693366, |
| "learning_rate": 2.286424431329513e-05, |
| "loss": 0.2467, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20943892002105713, |
| "step": 1960, |
| "valid_targets_mean": 4555.5, |
| "valid_targets_min": 1176 |
| }, |
| { |
| "epoch": 3.5727272727272728, |
| "grad_norm": 0.6210477048431411, |
| "learning_rate": 2.277448332399949e-05, |
| "loss": 0.2421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.31793659925460815, |
| "step": 1965, |
| "valid_targets_mean": 3210.2, |
| "valid_targets_min": 1783 |
| }, |
| { |
| "epoch": 3.581818181818182, |
| "grad_norm": 0.6810529070723943, |
| "learning_rate": 2.2684665316353112e-05, |
| "loss": 0.2648, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29948270320892334, |
| "step": 1970, |
| "valid_targets_mean": 2187.7, |
| "valid_targets_min": 996 |
| }, |
| { |
| "epoch": 3.590909090909091, |
| "grad_norm": 0.6515936749324243, |
| "learning_rate": 2.2594792136204037e-05, |
| "loss": 0.241, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.222039595246315, |
| "step": 1975, |
| "valid_targets_mean": 1998.1, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 3.6, |
| "grad_norm": 0.5732736942837675, |
| "learning_rate": 2.250486563053419e-05, |
| "loss": 0.2346, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20268699526786804, |
| "step": 1980, |
| "valid_targets_mean": 2295.4, |
| "valid_targets_min": 384 |
| }, |
| { |
| "epoch": 3.6090909090909093, |
| "grad_norm": 0.5688575835690776, |
| "learning_rate": 2.241488764742135e-05, |
| "loss": 0.2541, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25517329573631287, |
| "step": 1985, |
| "valid_targets_mean": 3159.9, |
| "valid_targets_min": 930 |
| }, |
| { |
| "epoch": 3.618181818181818, |
| "grad_norm": 0.5306274568561311, |
| "learning_rate": 2.232486003600126e-05, |
| "loss": 0.2328, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22559969127178192, |
| "step": 1990, |
| "valid_targets_mean": 2713.6, |
| "valid_targets_min": 1392 |
| }, |
| { |
| "epoch": 3.6272727272727274, |
| "grad_norm": 0.5048301936353792, |
| "learning_rate": 2.223478464642952e-05, |
| "loss": 0.2398, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18586412072181702, |
| "step": 1995, |
| "valid_targets_mean": 2830.0, |
| "valid_targets_min": 1063 |
| }, |
| { |
| "epoch": 3.6363636363636362, |
| "grad_norm": 0.6212436631613655, |
| "learning_rate": 2.2144663329843653e-05, |
| "loss": 0.2491, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27123504877090454, |
| "step": 2000, |
| "valid_targets_mean": 2453.6, |
| "valid_targets_min": 933 |
| }, |
| { |
| "epoch": 3.6454545454545455, |
| "grad_norm": 0.590331837924596, |
| "learning_rate": 2.205449793832502e-05, |
| "loss": 0.2489, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2864503264427185, |
| "step": 2005, |
| "valid_targets_mean": 2854.1, |
| "valid_targets_min": 588 |
| }, |
| { |
| "epoch": 3.6545454545454543, |
| "grad_norm": 0.5585937656084462, |
| "learning_rate": 2.1964290324860746e-05, |
| "loss": 0.2549, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23903991281986237, |
| "step": 2010, |
| "valid_targets_mean": 2668.5, |
| "valid_targets_min": 907 |
| }, |
| { |
| "epoch": 3.6636363636363636, |
| "grad_norm": 0.5514489995806369, |
| "learning_rate": 2.1874042343305685e-05, |
| "loss": 0.2517, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22939041256904602, |
| "step": 2015, |
| "valid_targets_mean": 2583.8, |
| "valid_targets_min": 1316 |
| }, |
| { |
| "epoch": 3.672727272727273, |
| "grad_norm": 0.5978802005083812, |
| "learning_rate": 2.1783755848344276e-05, |
| "loss": 0.2305, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24126064777374268, |
| "step": 2020, |
| "valid_targets_mean": 2730.9, |
| "valid_targets_min": 1473 |
| }, |
| { |
| "epoch": 3.6818181818181817, |
| "grad_norm": 0.53696951487926, |
| "learning_rate": 2.1693432695452467e-05, |
| "loss": 0.2439, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22672604024410248, |
| "step": 2025, |
| "valid_targets_mean": 2962.2, |
| "valid_targets_min": 875 |
| }, |
| { |
| "epoch": 3.690909090909091, |
| "grad_norm": 0.7675292875811535, |
| "learning_rate": 2.1603074740859534e-05, |
| "loss": 0.2312, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21047256886959076, |
| "step": 2030, |
| "valid_targets_mean": 2160.3, |
| "valid_targets_min": 1081 |
| }, |
| { |
| "epoch": 3.7, |
| "grad_norm": 0.5006566376336993, |
| "learning_rate": 2.1512683841509982e-05, |
| "loss": 0.2435, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21263140439987183, |
| "step": 2035, |
| "valid_targets_mean": 3089.0, |
| "valid_targets_min": 818 |
| }, |
| { |
| "epoch": 3.709090909090909, |
| "grad_norm": 0.56714645986811, |
| "learning_rate": 2.1422261855025357e-05, |
| "loss": 0.2496, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21199214458465576, |
| "step": 2040, |
| "valid_targets_mean": 2500.3, |
| "valid_targets_min": 1225 |
| }, |
| { |
| "epoch": 3.7181818181818183, |
| "grad_norm": 0.6494410349611806, |
| "learning_rate": 2.133181063966608e-05, |
| "loss": 0.26, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.281173974275589, |
| "step": 2045, |
| "valid_targets_mean": 3436.6, |
| "valid_targets_min": 1026 |
| }, |
| { |
| "epoch": 3.7272727272727275, |
| "grad_norm": 0.612579658009119, |
| "learning_rate": 2.1241332054293243e-05, |
| "loss": 0.24, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26280784606933594, |
| "step": 2050, |
| "valid_targets_mean": 2701.9, |
| "valid_targets_min": 1491 |
| }, |
| { |
| "epoch": 3.7363636363636363, |
| "grad_norm": 0.6118786743650745, |
| "learning_rate": 2.115082795833044e-05, |
| "loss": 0.2423, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23389646410942078, |
| "step": 2055, |
| "valid_targets_mean": 2522.9, |
| "valid_targets_min": 1260 |
| }, |
| { |
| "epoch": 3.7454545454545456, |
| "grad_norm": 0.6062010706814313, |
| "learning_rate": 2.1060300211725496e-05, |
| "loss": 0.233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2481895238161087, |
| "step": 2060, |
| "valid_targets_mean": 3095.2, |
| "valid_targets_min": 1285 |
| }, |
| { |
| "epoch": 3.7545454545454544, |
| "grad_norm": 0.5047024357662666, |
| "learning_rate": 2.096975067491233e-05, |
| "loss": 0.264, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.257171630859375, |
| "step": 2065, |
| "valid_targets_mean": 3419.8, |
| "valid_targets_min": 600 |
| }, |
| { |
| "epoch": 3.7636363636363637, |
| "grad_norm": 0.7895766683015748, |
| "learning_rate": 2.087918120877263e-05, |
| "loss": 0.2674, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3457374572753906, |
| "step": 2070, |
| "valid_targets_mean": 3181.0, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 3.7727272727272725, |
| "grad_norm": 0.5562767734285217, |
| "learning_rate": 2.0788593674597663e-05, |
| "loss": 0.2401, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25369274616241455, |
| "step": 2075, |
| "valid_targets_mean": 3111.6, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 3.7818181818181817, |
| "grad_norm": 0.6009442711262772, |
| "learning_rate": 2.0697989934050025e-05, |
| "loss": 0.2387, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23003512620925903, |
| "step": 2080, |
| "valid_targets_mean": 2796.6, |
| "valid_targets_min": 1178 |
| }, |
| { |
| "epoch": 3.790909090909091, |
| "grad_norm": 0.5951511305351987, |
| "learning_rate": 2.0607371849125345e-05, |
| "loss": 0.2625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3040267825126648, |
| "step": 2085, |
| "valid_targets_mean": 2972.1, |
| "valid_targets_min": 1319 |
| }, |
| { |
| "epoch": 3.8, |
| "grad_norm": 0.5979813146299011, |
| "learning_rate": 2.0516741282114062e-05, |
| "loss": 0.2421, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23099355399608612, |
| "step": 2090, |
| "valid_targets_mean": 2213.4, |
| "valid_targets_min": 779 |
| }, |
| { |
| "epoch": 3.809090909090909, |
| "grad_norm": 0.6242364839677382, |
| "learning_rate": 2.0426100095563132e-05, |
| "loss": 0.2593, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22831767797470093, |
| "step": 2095, |
| "valid_targets_mean": 2692.9, |
| "valid_targets_min": 393 |
| }, |
| { |
| "epoch": 3.8181818181818183, |
| "grad_norm": 0.7299492478429815, |
| "learning_rate": 2.0335450152237742e-05, |
| "loss": 0.2376, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2542460858821869, |
| "step": 2100, |
| "valid_targets_mean": 2546.2, |
| "valid_targets_min": 990 |
| }, |
| { |
| "epoch": 3.827272727272727, |
| "grad_norm": 0.6774426739953928, |
| "learning_rate": 2.0244793315083043e-05, |
| "loss": 0.2562, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23378236591815948, |
| "step": 2105, |
| "valid_targets_mean": 3035.1, |
| "valid_targets_min": 1604 |
| }, |
| { |
| "epoch": 3.8363636363636364, |
| "grad_norm": 0.6463334964796907, |
| "learning_rate": 2.0154131447185876e-05, |
| "loss": 0.25, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21351692080497742, |
| "step": 2110, |
| "valid_targets_mean": 2355.0, |
| "valid_targets_min": 1037 |
| }, |
| { |
| "epoch": 3.8454545454545457, |
| "grad_norm": 1.0600862632862642, |
| "learning_rate": 2.0063466411736447e-05, |
| "loss": 0.2363, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21131891012191772, |
| "step": 2115, |
| "valid_targets_mean": 2814.6, |
| "valid_targets_min": 1024 |
| }, |
| { |
| "epoch": 3.8545454545454545, |
| "grad_norm": 0.5765173759608054, |
| "learning_rate": 1.997280007199008e-05, |
| "loss": 0.2361, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2817190885543823, |
| "step": 2120, |
| "valid_targets_mean": 2889.8, |
| "valid_targets_min": 1483 |
| }, |
| { |
| "epoch": 3.8636363636363638, |
| "grad_norm": 0.5816941751168061, |
| "learning_rate": 1.9882134291228877e-05, |
| "loss": 0.2444, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18958780169487, |
| "step": 2125, |
| "valid_targets_mean": 2203.9, |
| "valid_targets_min": 1012 |
| }, |
| { |
| "epoch": 3.8727272727272726, |
| "grad_norm": 0.594692421630156, |
| "learning_rate": 1.9791470932723486e-05, |
| "loss": 0.2777, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2956348657608032, |
| "step": 2130, |
| "valid_targets_mean": 2760.8, |
| "valid_targets_min": 890 |
| }, |
| { |
| "epoch": 3.881818181818182, |
| "grad_norm": 0.5323715954369375, |
| "learning_rate": 1.9700811859694734e-05, |
| "loss": 0.246, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22484076023101807, |
| "step": 2135, |
| "valid_targets_mean": 3079.9, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 3.8909090909090907, |
| "grad_norm": 0.6003638416089673, |
| "learning_rate": 1.961015893527541e-05, |
| "loss": 0.2355, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22621342539787292, |
| "step": 2140, |
| "valid_targets_mean": 2506.4, |
| "valid_targets_min": 639 |
| }, |
| { |
| "epoch": 3.9, |
| "grad_norm": 0.6244575549178024, |
| "learning_rate": 1.9519514022471933e-05, |
| "loss": 0.2311, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24318262934684753, |
| "step": 2145, |
| "valid_targets_mean": 2432.9, |
| "valid_targets_min": 985 |
| }, |
| { |
| "epoch": 3.909090909090909, |
| "grad_norm": 0.591385714850753, |
| "learning_rate": 1.942887898412608e-05, |
| "loss": 0.2367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2611646056175232, |
| "step": 2150, |
| "valid_targets_mean": 2616.4, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 3.918181818181818, |
| "grad_norm": 0.5568234541673653, |
| "learning_rate": 1.9338255682876682e-05, |
| "loss": 0.2428, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2828320264816284, |
| "step": 2155, |
| "valid_targets_mean": 3327.7, |
| "valid_targets_min": 1790 |
| }, |
| { |
| "epoch": 3.9272727272727272, |
| "grad_norm": 0.6157245144679648, |
| "learning_rate": 1.924764598112138e-05, |
| "loss": 0.2593, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3046063780784607, |
| "step": 2160, |
| "valid_targets_mean": 2769.7, |
| "valid_targets_min": 1060 |
| }, |
| { |
| "epoch": 3.9363636363636365, |
| "grad_norm": 0.5875934472439608, |
| "learning_rate": 1.9157051740978326e-05, |
| "loss": 0.2509, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.32576432824134827, |
| "step": 2165, |
| "valid_targets_mean": 2968.4, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 3.9454545454545453, |
| "grad_norm": 0.5789081321587037, |
| "learning_rate": 1.9066474824247913e-05, |
| "loss": 0.2396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2773364186286926, |
| "step": 2170, |
| "valid_targets_mean": 2954.2, |
| "valid_targets_min": 864 |
| }, |
| { |
| "epoch": 3.9545454545454546, |
| "grad_norm": 0.5559791711562166, |
| "learning_rate": 1.8975917092374542e-05, |
| "loss": 0.235, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26809555292129517, |
| "step": 2175, |
| "valid_targets_mean": 3171.6, |
| "valid_targets_min": 1244 |
| }, |
| { |
| "epoch": 3.963636363636364, |
| "grad_norm": 0.5597868437982705, |
| "learning_rate": 1.888538040640831e-05, |
| "loss": 0.2554, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27784445881843567, |
| "step": 2180, |
| "valid_targets_mean": 2788.2, |
| "valid_targets_min": 1314 |
| }, |
| { |
| "epoch": 3.9727272727272727, |
| "grad_norm": 0.5351246359697984, |
| "learning_rate": 1.8794866626966834e-05, |
| "loss": 0.2269, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21700698137283325, |
| "step": 2185, |
| "valid_targets_mean": 3254.6, |
| "valid_targets_min": 1469 |
| }, |
| { |
| "epoch": 3.981818181818182, |
| "grad_norm": 0.5965100509204639, |
| "learning_rate": 1.8704377614196963e-05, |
| "loss": 0.2584, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27556589245796204, |
| "step": 2190, |
| "valid_targets_mean": 3114.7, |
| "valid_targets_min": 1228 |
| }, |
| { |
| "epoch": 3.990909090909091, |
| "grad_norm": 0.6058379024886905, |
| "learning_rate": 1.8613915227736584e-05, |
| "loss": 0.2358, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28704237937927246, |
| "step": 2195, |
| "valid_targets_mean": 2558.3, |
| "valid_targets_min": 1185 |
| }, |
| { |
| "epoch": 4.0, |
| "grad_norm": 0.5354214291936074, |
| "learning_rate": 1.852348132667635e-05, |
| "loss": 0.2383, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21756961941719055, |
| "step": 2200, |
| "valid_targets_mean": 2745.4, |
| "valid_targets_min": 1410 |
| }, |
| { |
| "epoch": 4.009090909090909, |
| "grad_norm": 0.63811517358803, |
| "learning_rate": 1.843307776952155e-05, |
| "loss": 0.2401, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24117009341716766, |
| "step": 2205, |
| "valid_targets_mean": 2653.8, |
| "valid_targets_min": 873 |
| }, |
| { |
| "epoch": 4.0181818181818185, |
| "grad_norm": 0.5976703740808316, |
| "learning_rate": 1.834270641415386e-05, |
| "loss": 0.2432, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2388264387845993, |
| "step": 2210, |
| "valid_targets_mean": 3296.8, |
| "valid_targets_min": 723 |
| }, |
| { |
| "epoch": 4.027272727272727, |
| "grad_norm": 0.6301202500239508, |
| "learning_rate": 1.8252369117793172e-05, |
| "loss": 0.2316, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25437045097351074, |
| "step": 2215, |
| "valid_targets_mean": 2990.8, |
| "valid_targets_min": 963 |
| }, |
| { |
| "epoch": 4.036363636363636, |
| "grad_norm": 0.6135121353685487, |
| "learning_rate": 1.8162067736959454e-05, |
| "loss": 0.261, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2367469221353531, |
| "step": 2220, |
| "valid_targets_mean": 2717.5, |
| "valid_targets_min": 1285 |
| }, |
| { |
| "epoch": 4.045454545454546, |
| "grad_norm": 0.5929711801863091, |
| "learning_rate": 1.8071804127434545e-05, |
| "loss": 0.2188, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22404639422893524, |
| "step": 2225, |
| "valid_targets_mean": 2649.3, |
| "valid_targets_min": 1056 |
| }, |
| { |
| "epoch": 4.054545454545455, |
| "grad_norm": 0.5604357256623951, |
| "learning_rate": 1.7981580144224066e-05, |
| "loss": 0.2318, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21794027090072632, |
| "step": 2230, |
| "valid_targets_mean": 3121.7, |
| "valid_targets_min": 1358 |
| }, |
| { |
| "epoch": 4.0636363636363635, |
| "grad_norm": 0.602733331600872, |
| "learning_rate": 1.7891397641519272e-05, |
| "loss": 0.211, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22328242659568787, |
| "step": 2235, |
| "valid_targets_mean": 2848.9, |
| "valid_targets_min": 1513 |
| }, |
| { |
| "epoch": 4.072727272727272, |
| "grad_norm": 0.6114398983536747, |
| "learning_rate": 1.7801258472658964e-05, |
| "loss": 0.2193, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22472426295280457, |
| "step": 2240, |
| "valid_targets_mean": 2859.0, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 4.081818181818182, |
| "grad_norm": 0.7949075687643492, |
| "learning_rate": 1.7711164490091365e-05, |
| "loss": 0.21, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22244544327259064, |
| "step": 2245, |
| "valid_targets_mean": 2770.4, |
| "valid_targets_min": 1027 |
| }, |
| { |
| "epoch": 4.090909090909091, |
| "grad_norm": 0.5050344354024722, |
| "learning_rate": 1.7621117545336098e-05, |
| "loss": 0.2201, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22612264752388, |
| "step": 2250, |
| "valid_targets_mean": 3871.1, |
| "valid_targets_min": 719 |
| }, |
| { |
| "epoch": 4.1, |
| "grad_norm": 0.5867166500193305, |
| "learning_rate": 1.7531119488946107e-05, |
| "loss": 0.2244, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21166576445102692, |
| "step": 2255, |
| "valid_targets_mean": 3124.7, |
| "valid_targets_min": 1342 |
| }, |
| { |
| "epoch": 4.109090909090909, |
| "grad_norm": 0.7449351531483703, |
| "learning_rate": 1.7441172170469634e-05, |
| "loss": 0.2316, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22294163703918457, |
| "step": 2260, |
| "valid_targets_mean": 2283.2, |
| "valid_targets_min": 938 |
| }, |
| { |
| "epoch": 4.118181818181818, |
| "grad_norm": 0.5806733663439438, |
| "learning_rate": 1.7351277438412197e-05, |
| "loss": 0.2336, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16949409246444702, |
| "step": 2265, |
| "valid_targets_mean": 2851.5, |
| "valid_targets_min": 1303 |
| }, |
| { |
| "epoch": 4.127272727272727, |
| "grad_norm": 0.6484255092516409, |
| "learning_rate": 1.726143714019862e-05, |
| "loss": 0.2262, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22443990409374237, |
| "step": 2270, |
| "valid_targets_mean": 2461.2, |
| "valid_targets_min": 886 |
| }, |
| { |
| "epoch": 4.136363636363637, |
| "grad_norm": 0.7029797028928565, |
| "learning_rate": 1.7171653122135065e-05, |
| "loss": 0.2386, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23379763960838318, |
| "step": 2275, |
| "valid_targets_mean": 2964.6, |
| "valid_targets_min": 1116 |
| }, |
| { |
| "epoch": 4.1454545454545455, |
| "grad_norm": 0.6487919761685919, |
| "learning_rate": 1.708192722937106e-05, |
| "loss": 0.2327, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24210688471794128, |
| "step": 2280, |
| "valid_targets_mean": 2856.1, |
| "valid_targets_min": 982 |
| }, |
| { |
| "epoch": 4.154545454545454, |
| "grad_norm": 0.5408810770304529, |
| "learning_rate": 1.6992261305861635e-05, |
| "loss": 0.22, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19399434328079224, |
| "step": 2285, |
| "valid_targets_mean": 3258.6, |
| "valid_targets_min": 990 |
| }, |
| { |
| "epoch": 4.163636363636364, |
| "grad_norm": 0.542802352358838, |
| "learning_rate": 1.6902657194329357e-05, |
| "loss": 0.239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.261577308177948, |
| "step": 2290, |
| "valid_targets_mean": 3440.7, |
| "valid_targets_min": 1626 |
| }, |
| { |
| "epoch": 4.172727272727273, |
| "grad_norm": 0.5891703591156177, |
| "learning_rate": 1.681311673622651e-05, |
| "loss": 0.2367, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18795019388198853, |
| "step": 2295, |
| "valid_targets_mean": 2520.1, |
| "valid_targets_min": 1166 |
| }, |
| { |
| "epoch": 4.181818181818182, |
| "grad_norm": 0.6252714768421596, |
| "learning_rate": 1.6723641771697246e-05, |
| "loss": 0.2308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23848536610603333, |
| "step": 2300, |
| "valid_targets_mean": 2629.7, |
| "valid_targets_min": 1208 |
| }, |
| { |
| "epoch": 4.190909090909091, |
| "grad_norm": 0.6719808729681732, |
| "learning_rate": 1.663423413953976e-05, |
| "loss": 0.2227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22473207116127014, |
| "step": 2305, |
| "valid_targets_mean": 2721.4, |
| "valid_targets_min": 1066 |
| }, |
| { |
| "epoch": 4.2, |
| "grad_norm": 0.6968029217292044, |
| "learning_rate": 1.6544895677168483e-05, |
| "loss": 0.2203, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23242871463298798, |
| "step": 2310, |
| "valid_targets_mean": 2373.2, |
| "valid_targets_min": 1193 |
| }, |
| { |
| "epoch": 4.209090909090909, |
| "grad_norm": 0.5804420649431439, |
| "learning_rate": 1.6455628220576357e-05, |
| "loss": 0.2122, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23284761607646942, |
| "step": 2315, |
| "valid_targets_mean": 3136.9, |
| "valid_targets_min": 1284 |
| }, |
| { |
| "epoch": 4.218181818181818, |
| "grad_norm": 0.6666888014451425, |
| "learning_rate": 1.6366433604297072e-05, |
| "loss": 0.23, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26939576864242554, |
| "step": 2320, |
| "valid_targets_mean": 2871.4, |
| "valid_targets_min": 795 |
| }, |
| { |
| "epoch": 4.2272727272727275, |
| "grad_norm": 0.5785967965305059, |
| "learning_rate": 1.62773136613674e-05, |
| "loss": 0.2232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24148644506931305, |
| "step": 2325, |
| "valid_targets_mean": 2955.4, |
| "valid_targets_min": 1222 |
| }, |
| { |
| "epoch": 4.236363636363636, |
| "grad_norm": 0.5265728273630983, |
| "learning_rate": 1.6188270223289483e-05, |
| "loss": 0.204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19100099802017212, |
| "step": 2330, |
| "valid_targets_mean": 3084.8, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 4.245454545454545, |
| "grad_norm": 0.644345992130894, |
| "learning_rate": 1.609930511999321e-05, |
| "loss": 0.2379, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2727709412574768, |
| "step": 2335, |
| "valid_targets_mean": 3700.1, |
| "valid_targets_min": 1440 |
| }, |
| { |
| "epoch": 4.254545454545455, |
| "grad_norm": 0.5736669297714321, |
| "learning_rate": 1.6010420179798623e-05, |
| "loss": 0.2166, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2274719476699829, |
| "step": 2340, |
| "valid_targets_mean": 2947.7, |
| "valid_targets_min": 1051 |
| }, |
| { |
| "epoch": 4.263636363636364, |
| "grad_norm": 0.4854164691385361, |
| "learning_rate": 1.5921617229378338e-05, |
| "loss": 0.2429, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25527408719062805, |
| "step": 2345, |
| "valid_targets_mean": 4699.1, |
| "valid_targets_min": 927 |
| }, |
| { |
| "epoch": 4.2727272727272725, |
| "grad_norm": 0.6550718663752602, |
| "learning_rate": 1.583289809372e-05, |
| "loss": 0.2299, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19673293828964233, |
| "step": 2350, |
| "valid_targets_mean": 2131.5, |
| "valid_targets_min": 993 |
| }, |
| { |
| "epoch": 4.281818181818182, |
| "grad_norm": 0.6070306508847103, |
| "learning_rate": 1.5744264596088763e-05, |
| "loss": 0.2277, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2816806435585022, |
| "step": 2355, |
| "valid_targets_mean": 3366.7, |
| "valid_targets_min": 818 |
| }, |
| { |
| "epoch": 4.290909090909091, |
| "grad_norm": 0.6690799082664834, |
| "learning_rate": 1.5655718557989848e-05, |
| "loss": 0.2393, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2496490776538849, |
| "step": 2360, |
| "valid_targets_mean": 2725.5, |
| "valid_targets_min": 1070 |
| }, |
| { |
| "epoch": 4.3, |
| "grad_norm": 0.5695334533687919, |
| "learning_rate": 1.5567261799131102e-05, |
| "loss": 0.2307, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25420039892196655, |
| "step": 2365, |
| "valid_targets_mean": 3258.9, |
| "valid_targets_min": 1104 |
| }, |
| { |
| "epoch": 4.309090909090909, |
| "grad_norm": 0.6093414327334828, |
| "learning_rate": 1.5478896137385584e-05, |
| "loss": 0.2182, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2125495821237564, |
| "step": 2370, |
| "valid_targets_mean": 2488.6, |
| "valid_targets_min": 898 |
| }, |
| { |
| "epoch": 4.318181818181818, |
| "grad_norm": 0.6250240012975116, |
| "learning_rate": 1.5390623388754232e-05, |
| "loss": 0.2196, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2510995864868164, |
| "step": 2375, |
| "valid_targets_mean": 3127.1, |
| "valid_targets_min": 1117 |
| }, |
| { |
| "epoch": 4.327272727272727, |
| "grad_norm": 0.5186773991836232, |
| "learning_rate": 1.5302445367328507e-05, |
| "loss": 0.2063, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19779804348945618, |
| "step": 2380, |
| "valid_targets_mean": 3112.1, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 4.336363636363636, |
| "grad_norm": 0.6104190805746833, |
| "learning_rate": 1.5214363885253156e-05, |
| "loss": 0.2564, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2571101784706116, |
| "step": 2385, |
| "valid_targets_mean": 3122.3, |
| "valid_targets_min": 1132 |
| }, |
| { |
| "epoch": 4.345454545454546, |
| "grad_norm": 0.642267748594081, |
| "learning_rate": 1.5126380752688934e-05, |
| "loss": 0.2371, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21655862033367157, |
| "step": 2390, |
| "valid_targets_mean": 2603.0, |
| "valid_targets_min": 695 |
| }, |
| { |
| "epoch": 4.3545454545454545, |
| "grad_norm": 0.9587976914904419, |
| "learning_rate": 1.503849777777543e-05, |
| "loss": 0.2337, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2582545876502991, |
| "step": 2395, |
| "valid_targets_mean": 1959.4, |
| "valid_targets_min": 660 |
| }, |
| { |
| "epoch": 4.363636363636363, |
| "grad_norm": 0.6358892336631516, |
| "learning_rate": 1.4950716766593872e-05, |
| "loss": 0.2469, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.34545302391052246, |
| "step": 2400, |
| "valid_targets_mean": 3421.1, |
| "valid_targets_min": 1122 |
| }, |
| { |
| "epoch": 4.372727272727273, |
| "grad_norm": 0.5952265307192156, |
| "learning_rate": 1.4863039523130054e-05, |
| "loss": 0.2322, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2792406678199768, |
| "step": 2405, |
| "valid_targets_mean": 3371.1, |
| "valid_targets_min": 1372 |
| }, |
| { |
| "epoch": 4.381818181818182, |
| "grad_norm": 0.5570515523505959, |
| "learning_rate": 1.4775467849237234e-05, |
| "loss": 0.2185, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1831461489200592, |
| "step": 2410, |
| "valid_targets_mean": 2825.3, |
| "valid_targets_min": 681 |
| }, |
| { |
| "epoch": 4.390909090909091, |
| "grad_norm": 0.5578379745764596, |
| "learning_rate": 1.468800354459912e-05, |
| "loss": 0.2505, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2448475956916809, |
| "step": 2415, |
| "valid_targets_mean": 3258.2, |
| "valid_targets_min": 993 |
| }, |
| { |
| "epoch": 4.4, |
| "grad_norm": 0.6261795588390912, |
| "learning_rate": 1.4600648406692863e-05, |
| "loss": 0.2201, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2323095053434372, |
| "step": 2420, |
| "valid_targets_mean": 2830.1, |
| "valid_targets_min": 986 |
| }, |
| { |
| "epoch": 4.409090909090909, |
| "grad_norm": 0.6684524920292939, |
| "learning_rate": 1.451340423075214e-05, |
| "loss": 0.2191, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24161775410175323, |
| "step": 2425, |
| "valid_targets_mean": 2482.1, |
| "valid_targets_min": 934 |
| }, |
| { |
| "epoch": 4.418181818181818, |
| "grad_norm": 0.7897405656539278, |
| "learning_rate": 1.4426272809730248e-05, |
| "loss": 0.2275, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24546557664871216, |
| "step": 2430, |
| "valid_targets_mean": 3149.2, |
| "valid_targets_min": 1224 |
| }, |
| { |
| "epoch": 4.427272727272728, |
| "grad_norm": 0.6417621177432642, |
| "learning_rate": 1.433925593426326e-05, |
| "loss": 0.1884, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18926918506622314, |
| "step": 2435, |
| "valid_targets_mean": 2471.9, |
| "valid_targets_min": 1215 |
| }, |
| { |
| "epoch": 4.4363636363636365, |
| "grad_norm": 0.5720794355531269, |
| "learning_rate": 1.4252355392633237e-05, |
| "loss": 0.2118, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2298586070537567, |
| "step": 2440, |
| "valid_targets_mean": 2885.1, |
| "valid_targets_min": 1001 |
| }, |
| { |
| "epoch": 4.445454545454545, |
| "grad_norm": 0.6127734036606598, |
| "learning_rate": 1.4165572970731435e-05, |
| "loss": 0.232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25842714309692383, |
| "step": 2445, |
| "valid_targets_mean": 3069.7, |
| "valid_targets_min": 1390 |
| }, |
| { |
| "epoch": 4.454545454545454, |
| "grad_norm": 0.6598183505433317, |
| "learning_rate": 1.4078910452021664e-05, |
| "loss": 0.2309, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26327598094940186, |
| "step": 2450, |
| "valid_targets_mean": 3190.2, |
| "valid_targets_min": 1526 |
| }, |
| { |
| "epoch": 4.463636363636364, |
| "grad_norm": 0.5855757963676725, |
| "learning_rate": 1.3992369617503594e-05, |
| "loss": 0.2145, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2034580111503601, |
| "step": 2455, |
| "valid_targets_mean": 2694.4, |
| "valid_targets_min": 587 |
| }, |
| { |
| "epoch": 4.472727272727273, |
| "grad_norm": 0.7452063256908972, |
| "learning_rate": 1.3905952245676173e-05, |
| "loss": 0.2053, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20218415558338165, |
| "step": 2460, |
| "valid_targets_mean": 2034.1, |
| "valid_targets_min": 969 |
| }, |
| { |
| "epoch": 4.4818181818181815, |
| "grad_norm": 0.595742074848808, |
| "learning_rate": 1.3819660112501054e-05, |
| "loss": 0.2341, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20664814114570618, |
| "step": 2465, |
| "valid_targets_mean": 2481.1, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 4.490909090909091, |
| "grad_norm": 0.6514179010516953, |
| "learning_rate": 1.3733494991366128e-05, |
| "loss": 0.2263, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2232770323753357, |
| "step": 2470, |
| "valid_targets_mean": 2275.2, |
| "valid_targets_min": 1087 |
| }, |
| { |
| "epoch": 4.5, |
| "grad_norm": 0.6180164291435978, |
| "learning_rate": 1.364745865304906e-05, |
| "loss": 0.1951, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2033901810646057, |
| "step": 2475, |
| "valid_targets_mean": 2587.8, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 4.509090909090909, |
| "grad_norm": 0.5871826019069449, |
| "learning_rate": 1.3561552865680899e-05, |
| "loss": 0.2282, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2353324145078659, |
| "step": 2480, |
| "valid_targets_mean": 2887.1, |
| "valid_targets_min": 840 |
| }, |
| { |
| "epoch": 4.5181818181818185, |
| "grad_norm": 0.6644157928804563, |
| "learning_rate": 1.3475779394709754e-05, |
| "loss": 0.2183, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20248183608055115, |
| "step": 2485, |
| "valid_targets_mean": 2442.0, |
| "valid_targets_min": 1103 |
| }, |
| { |
| "epoch": 4.527272727272727, |
| "grad_norm": 0.6995359046838049, |
| "learning_rate": 1.3390140002864481e-05, |
| "loss": 0.2474, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2637696862220764, |
| "step": 2490, |
| "valid_targets_mean": 2592.6, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 4.536363636363636, |
| "grad_norm": 0.6402215815007072, |
| "learning_rate": 1.3304636450118495e-05, |
| "loss": 0.2061, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21194347739219666, |
| "step": 2495, |
| "valid_targets_mean": 2424.1, |
| "valid_targets_min": 809 |
| }, |
| { |
| "epoch": 4.545454545454545, |
| "grad_norm": 0.527751097417042, |
| "learning_rate": 1.3219270493653587e-05, |
| "loss": 0.2178, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2287774384021759, |
| "step": 2500, |
| "valid_targets_mean": 3634.9, |
| "valid_targets_min": 1182 |
| }, |
| { |
| "epoch": 4.554545454545455, |
| "grad_norm": 0.6224833314845171, |
| "learning_rate": 1.3134043887823807e-05, |
| "loss": 0.2242, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17964226007461548, |
| "step": 2505, |
| "valid_targets_mean": 2677.7, |
| "valid_targets_min": 1474 |
| }, |
| { |
| "epoch": 4.5636363636363635, |
| "grad_norm": 0.4945668374502826, |
| "learning_rate": 1.3048958384119397e-05, |
| "loss": 0.224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19316023588180542, |
| "step": 2510, |
| "valid_targets_mean": 3429.4, |
| "valid_targets_min": 1536 |
| }, |
| { |
| "epoch": 4.572727272727272, |
| "grad_norm": 0.6840328864843762, |
| "learning_rate": 1.2964015731130836e-05, |
| "loss": 0.2362, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27236825227737427, |
| "step": 2515, |
| "valid_targets_mean": 2695.2, |
| "valid_targets_min": 635 |
| }, |
| { |
| "epoch": 4.581818181818182, |
| "grad_norm": 0.7745052573395034, |
| "learning_rate": 1.2879217674512865e-05, |
| "loss": 0.2227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22089771926403046, |
| "step": 2520, |
| "valid_targets_mean": 2492.4, |
| "valid_targets_min": 957 |
| }, |
| { |
| "epoch": 4.590909090909091, |
| "grad_norm": 0.5821037352524397, |
| "learning_rate": 1.279456595694864e-05, |
| "loss": 0.2276, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2236514389514923, |
| "step": 2525, |
| "valid_targets_mean": 2995.3, |
| "valid_targets_min": 1082 |
| }, |
| { |
| "epoch": 4.6, |
| "grad_norm": 0.6382341766426172, |
| "learning_rate": 1.2710062318113887e-05, |
| "loss": 0.2571, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3152768611907959, |
| "step": 2530, |
| "valid_targets_mean": 2937.4, |
| "valid_targets_min": 1238 |
| }, |
| { |
| "epoch": 4.609090909090909, |
| "grad_norm": 0.8458728089172884, |
| "learning_rate": 1.2625708494641188e-05, |
| "loss": 0.206, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21248075366020203, |
| "step": 2535, |
| "valid_targets_mean": 3374.9, |
| "valid_targets_min": 1280 |
| }, |
| { |
| "epoch": 4.618181818181818, |
| "grad_norm": 0.6472939984619872, |
| "learning_rate": 1.2541506220084262e-05, |
| "loss": 0.2235, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2410563975572586, |
| "step": 2540, |
| "valid_targets_mean": 2703.0, |
| "valid_targets_min": 862 |
| }, |
| { |
| "epoch": 4.627272727272727, |
| "grad_norm": 0.615416117105437, |
| "learning_rate": 1.2457457224882356e-05, |
| "loss": 0.2324, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22178882360458374, |
| "step": 2545, |
| "valid_targets_mean": 2794.3, |
| "valid_targets_min": 867 |
| }, |
| { |
| "epoch": 4.636363636363637, |
| "grad_norm": 0.5951692823267178, |
| "learning_rate": 1.237356323632468e-05, |
| "loss": 0.2354, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2083178460597992, |
| "step": 2550, |
| "valid_targets_mean": 2768.1, |
| "valid_targets_min": 1038 |
| }, |
| { |
| "epoch": 4.6454545454545455, |
| "grad_norm": 0.6261777189553865, |
| "learning_rate": 1.2289825978514882e-05, |
| "loss": 0.2201, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2828991711139679, |
| "step": 2555, |
| "valid_targets_mean": 2898.4, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 4.654545454545454, |
| "grad_norm": 0.6988710752962709, |
| "learning_rate": 1.2206247172335662e-05, |
| "loss": 0.219, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2878910005092621, |
| "step": 2560, |
| "valid_targets_mean": 2671.6, |
| "valid_targets_min": 788 |
| }, |
| { |
| "epoch": 4.663636363636364, |
| "grad_norm": 0.6660297859275669, |
| "learning_rate": 1.2122828535413378e-05, |
| "loss": 0.2282, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2508106231689453, |
| "step": 2565, |
| "valid_targets_mean": 2528.4, |
| "valid_targets_min": 892 |
| }, |
| { |
| "epoch": 4.672727272727273, |
| "grad_norm": 0.6306980294197534, |
| "learning_rate": 1.2039571782082762e-05, |
| "loss": 0.1944, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19929194450378418, |
| "step": 2570, |
| "valid_targets_mean": 2654.8, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 4.681818181818182, |
| "grad_norm": 0.6979301876844083, |
| "learning_rate": 1.1956478623351652e-05, |
| "loss": 0.2106, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20347945392131805, |
| "step": 2575, |
| "valid_targets_mean": 2358.7, |
| "valid_targets_min": 1167 |
| }, |
| { |
| "epoch": 4.690909090909091, |
| "grad_norm": 0.6279429718021498, |
| "learning_rate": 1.187355076686589e-05, |
| "loss": 0.2212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21126504242420197, |
| "step": 2580, |
| "valid_targets_mean": 2410.6, |
| "valid_targets_min": 1279 |
| }, |
| { |
| "epoch": 4.7, |
| "grad_norm": 0.5972306952510262, |
| "learning_rate": 1.1790789916874172e-05, |
| "loss": 0.2292, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2583068013191223, |
| "step": 2585, |
| "valid_targets_mean": 2919.1, |
| "valid_targets_min": 846 |
| }, |
| { |
| "epoch": 4.709090909090909, |
| "grad_norm": 0.5300072603352336, |
| "learning_rate": 1.1708197774193055e-05, |
| "loss": 0.2237, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21120886504650116, |
| "step": 2590, |
| "valid_targets_mean": 3532.8, |
| "valid_targets_min": 1134 |
| }, |
| { |
| "epoch": 4.718181818181818, |
| "grad_norm": 0.6386463342420782, |
| "learning_rate": 1.1625776036172006e-05, |
| "loss": 0.2303, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2677278518676758, |
| "step": 2595, |
| "valid_targets_mean": 2936.8, |
| "valid_targets_min": 1543 |
| }, |
| { |
| "epoch": 4.7272727272727275, |
| "grad_norm": 0.5972696678353682, |
| "learning_rate": 1.1543526396658475e-05, |
| "loss": 0.2308, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22586411237716675, |
| "step": 2600, |
| "valid_targets_mean": 2978.4, |
| "valid_targets_min": 1216 |
| }, |
| { |
| "epoch": 4.736363636363636, |
| "grad_norm": 0.615542151623342, |
| "learning_rate": 1.1461450545963167e-05, |
| "loss": 0.243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2534103989601135, |
| "step": 2605, |
| "valid_targets_mean": 2970.2, |
| "valid_targets_min": 919 |
| }, |
| { |
| "epoch": 4.745454545454545, |
| "grad_norm": 0.6578937951778845, |
| "learning_rate": 1.137955017082521e-05, |
| "loss": 0.2202, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22950537502765656, |
| "step": 2610, |
| "valid_targets_mean": 3471.8, |
| "valid_targets_min": 719 |
| }, |
| { |
| "epoch": 4.754545454545455, |
| "grad_norm": 0.6408798429955752, |
| "learning_rate": 1.1297826954377587e-05, |
| "loss": 0.224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2171386182308197, |
| "step": 2615, |
| "valid_targets_mean": 2560.1, |
| "valid_targets_min": 1227 |
| }, |
| { |
| "epoch": 4.763636363636364, |
| "grad_norm": 0.5415984276640066, |
| "learning_rate": 1.1216282576112436e-05, |
| "loss": 0.2262, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20664772391319275, |
| "step": 2620, |
| "valid_targets_mean": 3412.6, |
| "valid_targets_min": 1017 |
| }, |
| { |
| "epoch": 4.7727272727272725, |
| "grad_norm": 0.5251749491334398, |
| "learning_rate": 1.1134918711846651e-05, |
| "loss": 0.2224, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19899095594882965, |
| "step": 2625, |
| "valid_targets_mean": 2995.6, |
| "valid_targets_min": 1573 |
| }, |
| { |
| "epoch": 4.781818181818182, |
| "grad_norm": 0.6290363031722594, |
| "learning_rate": 1.1053737033687346e-05, |
| "loss": 0.2061, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21700023114681244, |
| "step": 2630, |
| "valid_targets_mean": 2623.8, |
| "valid_targets_min": 1085 |
| }, |
| { |
| "epoch": 4.790909090909091, |
| "grad_norm": 0.6849123593454419, |
| "learning_rate": 1.097273920999757e-05, |
| "loss": 0.2407, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23019665479660034, |
| "step": 2635, |
| "valid_targets_mean": 3421.1, |
| "valid_targets_min": 1358 |
| }, |
| { |
| "epoch": 4.8, |
| "grad_norm": 0.5771402980273382, |
| "learning_rate": 1.0891926905361948e-05, |
| "loss": 0.2223, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18821823596954346, |
| "step": 2640, |
| "valid_targets_mean": 2784.2, |
| "valid_targets_min": 1082 |
| }, |
| { |
| "epoch": 4.809090909090909, |
| "grad_norm": 0.6207164760514525, |
| "learning_rate": 1.081130178055251e-05, |
| "loss": 0.2271, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23473981022834778, |
| "step": 2645, |
| "valid_targets_mean": 2909.9, |
| "valid_targets_min": 1405 |
| }, |
| { |
| "epoch": 4.818181818181818, |
| "grad_norm": 0.5918362524255482, |
| "learning_rate": 1.0730865492494593e-05, |
| "loss": 0.2198, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24564619362354279, |
| "step": 2650, |
| "valid_targets_mean": 3247.2, |
| "valid_targets_min": 723 |
| }, |
| { |
| "epoch": 4.827272727272727, |
| "grad_norm": 0.5917360630872475, |
| "learning_rate": 1.0650619694232704e-05, |
| "loss": 0.2356, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2355796992778778, |
| "step": 2655, |
| "valid_targets_mean": 2881.9, |
| "valid_targets_min": 1472 |
| }, |
| { |
| "epoch": 4.836363636363636, |
| "grad_norm": 0.6108234295069259, |
| "learning_rate": 1.057056603489665e-05, |
| "loss": 0.2148, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25410670042037964, |
| "step": 2660, |
| "valid_targets_mean": 2952.7, |
| "valid_targets_min": 917 |
| }, |
| { |
| "epoch": 4.845454545454546, |
| "grad_norm": 0.6401240752548848, |
| "learning_rate": 1.0490706159667534e-05, |
| "loss": 0.2231, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23201113939285278, |
| "step": 2665, |
| "valid_targets_mean": 2641.9, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 4.8545454545454545, |
| "grad_norm": 0.6225393374592852, |
| "learning_rate": 1.0411041709744063e-05, |
| "loss": 0.2318, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24733442068099976, |
| "step": 2670, |
| "valid_targets_mean": 2914.0, |
| "valid_targets_min": 797 |
| }, |
| { |
| "epoch": 4.863636363636363, |
| "grad_norm": 0.6863647029544806, |
| "learning_rate": 1.0331574322308722e-05, |
| "loss": 0.2409, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22286084294319153, |
| "step": 2675, |
| "valid_targets_mean": 2257.1, |
| "valid_targets_min": 588 |
| }, |
| { |
| "epoch": 4.872727272727273, |
| "grad_norm": 0.7067563529480779, |
| "learning_rate": 1.0252305630494201e-05, |
| "loss": 0.2072, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17897702753543854, |
| "step": 2680, |
| "valid_targets_mean": 2485.5, |
| "valid_targets_min": 1366 |
| }, |
| { |
| "epoch": 4.881818181818182, |
| "grad_norm": 0.6697464783481224, |
| "learning_rate": 1.0173237263349776e-05, |
| "loss": 0.2516, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2502247393131256, |
| "step": 2685, |
| "valid_targets_mean": 2746.4, |
| "valid_targets_min": 672 |
| }, |
| { |
| "epoch": 4.890909090909091, |
| "grad_norm": 0.6173196084339941, |
| "learning_rate": 1.0094370845807857e-05, |
| "loss": 0.2335, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19461512565612793, |
| "step": 2690, |
| "valid_targets_mean": 2824.4, |
| "valid_targets_min": 761 |
| }, |
| { |
| "epoch": 4.9, |
| "grad_norm": 0.6532101802559712, |
| "learning_rate": 1.001570799865061e-05, |
| "loss": 0.2125, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2121875137090683, |
| "step": 2695, |
| "valid_targets_mean": 2877.6, |
| "valid_targets_min": 1099 |
| }, |
| { |
| "epoch": 4.909090909090909, |
| "grad_norm": 0.5488728499968432, |
| "learning_rate": 9.937250338476607e-06, |
| "loss": 0.2294, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.29637664556503296, |
| "step": 2700, |
| "valid_targets_mean": 4819.6, |
| "valid_targets_min": 1300 |
| }, |
| { |
| "epoch": 4.918181818181818, |
| "grad_norm": 0.6606645400766106, |
| "learning_rate": 9.858999477667656e-06, |
| "loss": 0.2025, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23928359150886536, |
| "step": 2705, |
| "valid_targets_mean": 2634.3, |
| "valid_targets_min": 1170 |
| }, |
| { |
| "epoch": 4.927272727272728, |
| "grad_norm": 0.544686799589651, |
| "learning_rate": 9.780957024355591e-06, |
| "loss": 0.2217, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2297343909740448, |
| "step": 2710, |
| "valid_targets_mean": 3213.6, |
| "valid_targets_min": 1323 |
| }, |
| { |
| "epoch": 4.9363636363636365, |
| "grad_norm": 0.5949988577753448, |
| "learning_rate": 9.703124582389312e-06, |
| "loss": 0.2411, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2422581911087036, |
| "step": 2715, |
| "valid_targets_mean": 3500.9, |
| "valid_targets_min": 1391 |
| }, |
| { |
| "epoch": 4.945454545454545, |
| "grad_norm": 0.6392651426837851, |
| "learning_rate": 9.62550375130175e-06, |
| "loss": 0.241, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23042556643486023, |
| "step": 2720, |
| "valid_targets_mean": 3213.1, |
| "valid_targets_min": 1433 |
| }, |
| { |
| "epoch": 4.954545454545455, |
| "grad_norm": 0.6966050262955494, |
| "learning_rate": 9.548096126277058e-06, |
| "loss": 0.2243, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25255775451660156, |
| "step": 2725, |
| "valid_targets_mean": 3336.6, |
| "valid_targets_min": 1244 |
| }, |
| { |
| "epoch": 4.963636363636364, |
| "grad_norm": 0.5928575852962964, |
| "learning_rate": 9.470903298117744e-06, |
| "loss": 0.2266, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2821640074253082, |
| "step": 2730, |
| "valid_targets_mean": 3400.3, |
| "valid_targets_min": 1184 |
| }, |
| { |
| "epoch": 4.972727272727273, |
| "grad_norm": 0.6486168657443957, |
| "learning_rate": 9.393926853212083e-06, |
| "loss": 0.2344, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19834952056407928, |
| "step": 2735, |
| "valid_targets_mean": 2398.2, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 4.9818181818181815, |
| "grad_norm": 0.48416111164674525, |
| "learning_rate": 9.317168373501426e-06, |
| "loss": 0.2, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21909256279468536, |
| "step": 2740, |
| "valid_targets_mean": 3984.8, |
| "valid_targets_min": 633 |
| }, |
| { |
| "epoch": 4.990909090909091, |
| "grad_norm": 0.5590944867565344, |
| "learning_rate": 9.240629436447752e-06, |
| "loss": 0.2292, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2566272020339966, |
| "step": 2745, |
| "valid_targets_mean": 3417.1, |
| "valid_targets_min": 1627 |
| }, |
| { |
| "epoch": 5.0, |
| "grad_norm": 0.6105837422351534, |
| "learning_rate": 9.164311615001202e-06, |
| "loss": 0.2285, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22403398156166077, |
| "step": 2750, |
| "valid_targets_mean": 2896.4, |
| "valid_targets_min": 1010 |
| }, |
| { |
| "epoch": 5.009090909090909, |
| "grad_norm": 0.6206679943666114, |
| "learning_rate": 9.08821647756778e-06, |
| "loss": 0.2022, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2819976210594177, |
| "step": 2755, |
| "valid_targets_mean": 3098.4, |
| "valid_targets_min": 531 |
| }, |
| { |
| "epoch": 5.0181818181818185, |
| "grad_norm": 0.5869668965068552, |
| "learning_rate": 9.012345587977129e-06, |
| "loss": 0.238, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18100440502166748, |
| "step": 2760, |
| "valid_targets_mean": 2610.1, |
| "valid_targets_min": 894 |
| }, |
| { |
| "epoch": 5.027272727272727, |
| "grad_norm": 0.6800086014403005, |
| "learning_rate": 8.936700505450356e-06, |
| "loss": 0.2233, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2488502562046051, |
| "step": 2765, |
| "valid_targets_mean": 2924.6, |
| "valid_targets_min": 932 |
| }, |
| { |
| "epoch": 5.036363636363636, |
| "grad_norm": 0.5733287013285376, |
| "learning_rate": 8.861282784568045e-06, |
| "loss": 0.2009, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20497402548789978, |
| "step": 2770, |
| "valid_targets_mean": 3032.6, |
| "valid_targets_min": 1401 |
| }, |
| { |
| "epoch": 5.045454545454546, |
| "grad_norm": 0.5579965277111765, |
| "learning_rate": 8.786093975238226e-06, |
| "loss": 0.2287, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19170859456062317, |
| "step": 2775, |
| "valid_targets_mean": 3130.4, |
| "valid_targets_min": 1230 |
| }, |
| { |
| "epoch": 5.054545454545455, |
| "grad_norm": 0.5424604659652602, |
| "learning_rate": 8.711135622664622e-06, |
| "loss": 0.2048, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1387612372636795, |
| "step": 2780, |
| "valid_targets_mean": 3084.8, |
| "valid_targets_min": 1151 |
| }, |
| { |
| "epoch": 5.0636363636363635, |
| "grad_norm": 0.6300869155028218, |
| "learning_rate": 8.636409267314806e-06, |
| "loss": 0.1891, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1826111376285553, |
| "step": 2785, |
| "valid_targets_mean": 2739.9, |
| "valid_targets_min": 1336 |
| }, |
| { |
| "epoch": 5.072727272727272, |
| "grad_norm": 0.6418965447840044, |
| "learning_rate": 8.561916444888618e-06, |
| "loss": 0.219, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18538132309913635, |
| "step": 2790, |
| "valid_targets_mean": 3084.9, |
| "valid_targets_min": 1446 |
| }, |
| { |
| "epoch": 5.081818181818182, |
| "grad_norm": 1.0767392353239975, |
| "learning_rate": 8.487658686286533e-06, |
| "loss": 0.235, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1957949846982956, |
| "step": 2795, |
| "valid_targets_mean": 2874.3, |
| "valid_targets_min": 1079 |
| }, |
| { |
| "epoch": 5.090909090909091, |
| "grad_norm": 0.8212955769888806, |
| "learning_rate": 8.413637517578246e-06, |
| "loss": 0.2178, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21617895364761353, |
| "step": 2800, |
| "valid_targets_mean": 2496.1, |
| "valid_targets_min": 1155 |
| }, |
| { |
| "epoch": 5.1, |
| "grad_norm": 1.2956418315388283, |
| "learning_rate": 8.339854459971313e-06, |
| "loss": 0.2267, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.249013289809227, |
| "step": 2805, |
| "valid_targets_mean": 3000.8, |
| "valid_targets_min": 1115 |
| }, |
| { |
| "epoch": 5.109090909090909, |
| "grad_norm": 0.6165988995663725, |
| "learning_rate": 8.266311029779843e-06, |
| "loss": 0.2227, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.224916011095047, |
| "step": 2810, |
| "valid_targets_mean": 3321.6, |
| "valid_targets_min": 904 |
| }, |
| { |
| "epoch": 5.118181818181818, |
| "grad_norm": 0.7398245215368483, |
| "learning_rate": 8.193008738393409e-06, |
| "loss": 0.2099, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24344930052757263, |
| "step": 2815, |
| "valid_targets_mean": 2388.0, |
| "valid_targets_min": 811 |
| }, |
| { |
| "epoch": 5.127272727272727, |
| "grad_norm": 0.6021280092554041, |
| "learning_rate": 8.119949092245893e-06, |
| "loss": 0.1942, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.170393168926239, |
| "step": 2820, |
| "valid_targets_mean": 2751.1, |
| "valid_targets_min": 1155 |
| }, |
| { |
| "epoch": 5.136363636363637, |
| "grad_norm": 0.5909261979045841, |
| "learning_rate": 8.047133592784626e-06, |
| "loss": 0.2016, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1882145255804062, |
| "step": 2825, |
| "valid_targets_mean": 3168.0, |
| "valid_targets_min": 1053 |
| }, |
| { |
| "epoch": 5.1454545454545455, |
| "grad_norm": 2.4155979064615596, |
| "learning_rate": 7.974563736439454e-06, |
| "loss": 0.1791, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.15529166162014008, |
| "step": 2830, |
| "valid_targets_mean": 2921.1, |
| "valid_targets_min": 1387 |
| }, |
| { |
| "epoch": 5.154545454545454, |
| "grad_norm": 0.6086422217400963, |
| "learning_rate": 7.902241014592042e-06, |
| "loss": 0.2193, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20463624596595764, |
| "step": 2835, |
| "valid_targets_mean": 2895.8, |
| "valid_targets_min": 1560 |
| }, |
| { |
| "epoch": 5.163636363636364, |
| "grad_norm": 0.6025680214907746, |
| "learning_rate": 7.830166913545181e-06, |
| "loss": 0.2184, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20843620598316193, |
| "step": 2840, |
| "valid_targets_mean": 3195.0, |
| "valid_targets_min": 1078 |
| }, |
| { |
| "epoch": 5.172727272727273, |
| "grad_norm": 0.5849462224788665, |
| "learning_rate": 7.758342914492257e-06, |
| "loss": 0.1822, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1787065714597702, |
| "step": 2845, |
| "valid_targets_mean": 3042.2, |
| "valid_targets_min": 1258 |
| }, |
| { |
| "epoch": 5.181818181818182, |
| "grad_norm": 0.7462903199250011, |
| "learning_rate": 7.686770493486835e-06, |
| "loss": 0.1946, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18932729959487915, |
| "step": 2850, |
| "valid_targets_mean": 2514.8, |
| "valid_targets_min": 700 |
| }, |
| { |
| "epoch": 5.190909090909091, |
| "grad_norm": 0.6088941331030795, |
| "learning_rate": 7.615451121412285e-06, |
| "loss": 0.2135, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23179104924201965, |
| "step": 2855, |
| "valid_targets_mean": 2957.9, |
| "valid_targets_min": 998 |
| }, |
| { |
| "epoch": 5.2, |
| "grad_norm": 0.6362339563375317, |
| "learning_rate": 7.5443862639516e-06, |
| "loss": 0.2127, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23404110968112946, |
| "step": 2860, |
| "valid_targets_mean": 2948.4, |
| "valid_targets_min": 560 |
| }, |
| { |
| "epoch": 5.209090909090909, |
| "grad_norm": 0.7425390338195141, |
| "learning_rate": 7.4735773815572044e-06, |
| "loss": 0.2001, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2154429703950882, |
| "step": 2865, |
| "valid_targets_mean": 3096.9, |
| "valid_targets_min": 839 |
| }, |
| { |
| "epoch": 5.218181818181818, |
| "grad_norm": 0.6732004978716041, |
| "learning_rate": 7.403025929421026e-06, |
| "loss": 0.1947, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19037508964538574, |
| "step": 2870, |
| "valid_targets_mean": 2717.5, |
| "valid_targets_min": 1116 |
| }, |
| { |
| "epoch": 5.2272727272727275, |
| "grad_norm": 0.5864918250329665, |
| "learning_rate": 7.332733357444524e-06, |
| "loss": 0.2151, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22506625950336456, |
| "step": 2875, |
| "valid_targets_mean": 3214.4, |
| "valid_targets_min": 1316 |
| }, |
| { |
| "epoch": 5.236363636363636, |
| "grad_norm": 0.6619039179552094, |
| "learning_rate": 7.262701110208936e-06, |
| "loss": 0.2218, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2768268883228302, |
| "step": 2880, |
| "valid_targets_mean": 3115.4, |
| "valid_targets_min": 1317 |
| }, |
| { |
| "epoch": 5.245454545454545, |
| "grad_norm": 0.8268864892205662, |
| "learning_rate": 7.192930626945556e-06, |
| "loss": 0.197, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22147375345230103, |
| "step": 2885, |
| "valid_targets_mean": 2782.2, |
| "valid_targets_min": 843 |
| }, |
| { |
| "epoch": 5.254545454545455, |
| "grad_norm": 0.6740820931397506, |
| "learning_rate": 7.123423341506168e-06, |
| "loss": 0.2014, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20615090429782867, |
| "step": 2890, |
| "valid_targets_mean": 2661.6, |
| "valid_targets_min": 1062 |
| }, |
| { |
| "epoch": 5.263636363636364, |
| "grad_norm": 0.6254507861499846, |
| "learning_rate": 7.054180682333602e-06, |
| "loss": 0.1946, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18174424767494202, |
| "step": 2895, |
| "valid_targets_mean": 2556.6, |
| "valid_targets_min": 1177 |
| }, |
| { |
| "epoch": 5.2727272727272725, |
| "grad_norm": 0.6507303440906587, |
| "learning_rate": 6.985204072432348e-06, |
| "loss": 0.212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17763866484165192, |
| "step": 2900, |
| "valid_targets_mean": 2767.8, |
| "valid_targets_min": 1254 |
| }, |
| { |
| "epoch": 5.281818181818182, |
| "grad_norm": 0.615476352490313, |
| "learning_rate": 6.916494929339315e-06, |
| "loss": 0.2284, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19773416221141815, |
| "step": 2905, |
| "valid_targets_mean": 3307.8, |
| "valid_targets_min": 1636 |
| }, |
| { |
| "epoch": 5.290909090909091, |
| "grad_norm": 0.6078720402135201, |
| "learning_rate": 6.848054665094714e-06, |
| "loss": 0.2094, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18933966755867004, |
| "step": 2910, |
| "valid_targets_mean": 2963.9, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 5.3, |
| "grad_norm": 0.6578005853230192, |
| "learning_rate": 6.779884686213043e-06, |
| "loss": 0.2041, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23294946551322937, |
| "step": 2915, |
| "valid_targets_mean": 3041.9, |
| "valid_targets_min": 1223 |
| }, |
| { |
| "epoch": 5.309090909090909, |
| "grad_norm": 0.6035763320821946, |
| "learning_rate": 6.71198639365415e-06, |
| "loss": 0.2269, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18072904646396637, |
| "step": 2920, |
| "valid_targets_mean": 3273.4, |
| "valid_targets_min": 1488 |
| }, |
| { |
| "epoch": 5.318181818181818, |
| "grad_norm": 0.6502916000400304, |
| "learning_rate": 6.644361182794494e-06, |
| "loss": 0.1995, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22111941874027252, |
| "step": 2925, |
| "valid_targets_mean": 2933.1, |
| "valid_targets_min": 1206 |
| }, |
| { |
| "epoch": 5.327272727272727, |
| "grad_norm": 0.4745748691092994, |
| "learning_rate": 6.577010443398388e-06, |
| "loss": 0.1949, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.15541119873523712, |
| "step": 2930, |
| "valid_targets_mean": 3669.4, |
| "valid_targets_min": 393 |
| }, |
| { |
| "epoch": 5.336363636363636, |
| "grad_norm": 0.6421309510997331, |
| "learning_rate": 6.50993555958954e-06, |
| "loss": 0.204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20444971323013306, |
| "step": 2935, |
| "valid_targets_mean": 2923.8, |
| "valid_targets_min": 764 |
| }, |
| { |
| "epoch": 5.345454545454546, |
| "grad_norm": 0.6744054487187182, |
| "learning_rate": 6.4431379098225185e-06, |
| "loss": 0.1947, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18364882469177246, |
| "step": 2940, |
| "valid_targets_mean": 2425.6, |
| "valid_targets_min": 1304 |
| }, |
| { |
| "epoch": 5.3545454545454545, |
| "grad_norm": 0.5763112027918369, |
| "learning_rate": 6.376618866854485e-06, |
| "loss": 0.2181, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2389511615037918, |
| "step": 2945, |
| "valid_targets_mean": 3468.6, |
| "valid_targets_min": 1635 |
| }, |
| { |
| "epoch": 5.363636363636363, |
| "grad_norm": 0.6042546105394436, |
| "learning_rate": 6.310379797716946e-06, |
| "loss": 0.2075, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18797212839126587, |
| "step": 2950, |
| "valid_targets_mean": 2697.3, |
| "valid_targets_min": 1013 |
| }, |
| { |
| "epoch": 5.372727272727273, |
| "grad_norm": 0.6206834860599623, |
| "learning_rate": 6.24442206368766e-06, |
| "loss": 0.1938, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17718634009361267, |
| "step": 2955, |
| "valid_targets_mean": 2515.1, |
| "valid_targets_min": 873 |
| }, |
| { |
| "epoch": 5.381818181818182, |
| "grad_norm": 0.7073609169477288, |
| "learning_rate": 6.178747020262708e-06, |
| "loss": 0.2249, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21706639230251312, |
| "step": 2960, |
| "valid_targets_mean": 2701.6, |
| "valid_targets_min": 1219 |
| }, |
| { |
| "epoch": 5.390909090909091, |
| "grad_norm": 0.6983928267652495, |
| "learning_rate": 6.1133560171285625e-06, |
| "loss": 0.2115, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1814163625240326, |
| "step": 2965, |
| "valid_targets_mean": 2054.0, |
| "valid_targets_min": 688 |
| }, |
| { |
| "epoch": 5.4, |
| "grad_norm": 0.9371617766106026, |
| "learning_rate": 6.04825039813443e-06, |
| "loss": 0.206, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23152729868888855, |
| "step": 2970, |
| "valid_targets_mean": 3087.6, |
| "valid_targets_min": 1067 |
| }, |
| { |
| "epoch": 5.409090909090909, |
| "grad_norm": 0.6439823648754108, |
| "learning_rate": 5.983431501264545e-06, |
| "loss": 0.1861, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20572486519813538, |
| "step": 2975, |
| "valid_targets_mean": 2651.6, |
| "valid_targets_min": 1131 |
| }, |
| { |
| "epoch": 5.418181818181818, |
| "grad_norm": 0.6292024202102485, |
| "learning_rate": 5.918900658610765e-06, |
| "loss": 0.1932, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22478453814983368, |
| "step": 2980, |
| "valid_targets_mean": 3044.0, |
| "valid_targets_min": 1557 |
| }, |
| { |
| "epoch": 5.427272727272728, |
| "grad_norm": 0.7004939743631798, |
| "learning_rate": 5.8546591963451226e-06, |
| "loss": 0.2288, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.33600202202796936, |
| "step": 2985, |
| "valid_targets_mean": 2969.0, |
| "valid_targets_min": 1061 |
| }, |
| { |
| "epoch": 5.4363636363636365, |
| "grad_norm": 0.7212119858500032, |
| "learning_rate": 5.790708434692627e-06, |
| "loss": 0.2041, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24133487045764923, |
| "step": 2990, |
| "valid_targets_mean": 2493.2, |
| "valid_targets_min": 1084 |
| }, |
| { |
| "epoch": 5.445454545454545, |
| "grad_norm": 0.5878479725042247, |
| "learning_rate": 5.727049687904076e-06, |
| "loss": 0.2398, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27190083265304565, |
| "step": 2995, |
| "valid_targets_mean": 3558.1, |
| "valid_targets_min": 1027 |
| }, |
| { |
| "epoch": 5.454545454545454, |
| "grad_norm": 0.7210660143522136, |
| "learning_rate": 5.66368426422909e-06, |
| "loss": 0.2017, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24498572945594788, |
| "step": 3000, |
| "valid_targets_mean": 2708.6, |
| "valid_targets_min": 893 |
| }, |
| { |
| "epoch": 5.463636363636364, |
| "grad_norm": 1.0935952651981136, |
| "learning_rate": 5.60061346588922e-06, |
| "loss": 0.2147, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2006685733795166, |
| "step": 3005, |
| "valid_targets_mean": 2522.6, |
| "valid_targets_min": 1501 |
| }, |
| { |
| "epoch": 5.472727272727273, |
| "grad_norm": 0.7427392414896691, |
| "learning_rate": 5.537838589051155e-06, |
| "loss": 0.2138, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21275198459625244, |
| "step": 3010, |
| "valid_targets_mean": 2713.2, |
| "valid_targets_min": 1302 |
| }, |
| { |
| "epoch": 5.4818181818181815, |
| "grad_norm": 0.6797980800187736, |
| "learning_rate": 5.475360923800141e-06, |
| "loss": 0.2157, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28571492433547974, |
| "step": 3015, |
| "valid_targets_mean": 3031.4, |
| "valid_targets_min": 1583 |
| }, |
| { |
| "epoch": 5.490909090909091, |
| "grad_norm": 0.5982952442155729, |
| "learning_rate": 5.413181754113392e-06, |
| "loss": 0.188, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.15068712830543518, |
| "step": 3020, |
| "valid_targets_mean": 2895.8, |
| "valid_targets_min": 1189 |
| }, |
| { |
| "epoch": 5.5, |
| "grad_norm": 0.8202460368123184, |
| "learning_rate": 5.351302357833785e-06, |
| "loss": 0.218, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18848514556884766, |
| "step": 3025, |
| "valid_targets_mean": 2772.6, |
| "valid_targets_min": 748 |
| }, |
| { |
| "epoch": 5.509090909090909, |
| "grad_norm": 0.7535260717612179, |
| "learning_rate": 5.289724006643529e-06, |
| "loss": 0.2057, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.233030304312706, |
| "step": 3030, |
| "valid_targets_mean": 2721.9, |
| "valid_targets_min": 753 |
| }, |
| { |
| "epoch": 5.5181818181818185, |
| "grad_norm": 0.6261585733081678, |
| "learning_rate": 5.2284479660380906e-06, |
| "loss": 0.2098, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2688322365283966, |
| "step": 3035, |
| "valid_targets_mean": 3717.1, |
| "valid_targets_min": 1169 |
| }, |
| { |
| "epoch": 5.527272727272727, |
| "grad_norm": 0.9190797360368729, |
| "learning_rate": 5.167475495300134e-06, |
| "loss": 0.2059, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1784450113773346, |
| "step": 3040, |
| "valid_targets_mean": 2067.2, |
| "valid_targets_min": 933 |
| }, |
| { |
| "epoch": 5.536363636363636, |
| "grad_norm": 0.8291481178474116, |
| "learning_rate": 5.1068078474736695e-06, |
| "loss": 0.2198, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1946008801460266, |
| "step": 3045, |
| "valid_targets_mean": 2897.1, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 5.545454545454545, |
| "grad_norm": 0.7189873688072802, |
| "learning_rate": 5.046446269338314e-06, |
| "loss": 0.2177, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24384161829948425, |
| "step": 3050, |
| "valid_targets_mean": 2461.3, |
| "valid_targets_min": 588 |
| }, |
| { |
| "epoch": 5.554545454545455, |
| "grad_norm": 0.730651619011706, |
| "learning_rate": 4.986392001383633e-06, |
| "loss": 0.195, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19608566164970398, |
| "step": 3055, |
| "valid_targets_mean": 2166.1, |
| "valid_targets_min": 1070 |
| }, |
| { |
| "epoch": 5.5636363636363635, |
| "grad_norm": 0.5985262442762385, |
| "learning_rate": 4.926646277783675e-06, |
| "loss": 0.2396, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.28461572527885437, |
| "step": 3060, |
| "valid_targets_mean": 3714.2, |
| "valid_targets_min": 978 |
| }, |
| { |
| "epoch": 5.572727272727272, |
| "grad_norm": 0.6643144693677729, |
| "learning_rate": 4.867210326371596e-06, |
| "loss": 0.2388, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25610047578811646, |
| "step": 3065, |
| "valid_targets_mean": 3391.6, |
| "valid_targets_min": 883 |
| }, |
| { |
| "epoch": 5.581818181818182, |
| "grad_norm": 0.6203488079735966, |
| "learning_rate": 4.808085368614441e-06, |
| "loss": 0.2193, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19354897737503052, |
| "step": 3070, |
| "valid_targets_mean": 2905.4, |
| "valid_targets_min": 1005 |
| }, |
| { |
| "epoch": 5.590909090909091, |
| "grad_norm": 0.6419618861722692, |
| "learning_rate": 4.74927261958801e-06, |
| "loss": 0.2114, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23560906946659088, |
| "step": 3075, |
| "valid_targets_mean": 3003.8, |
| "valid_targets_min": 1333 |
| }, |
| { |
| "epoch": 5.6, |
| "grad_norm": 0.6991571121030598, |
| "learning_rate": 4.690773287951942e-06, |
| "loss": 0.2316, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2553230822086334, |
| "step": 3080, |
| "valid_targets_mean": 3061.7, |
| "valid_targets_min": 949 |
| }, |
| { |
| "epoch": 5.609090909090909, |
| "grad_norm": 0.5994102701086738, |
| "learning_rate": 4.632588575924795e-06, |
| "loss": 0.208, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.205175518989563, |
| "step": 3085, |
| "valid_targets_mean": 3021.4, |
| "valid_targets_min": 726 |
| }, |
| { |
| "epoch": 5.618181818181818, |
| "grad_norm": 0.7723586734348693, |
| "learning_rate": 4.574719679259425e-06, |
| "loss": 0.2213, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2598193287849426, |
| "step": 3090, |
| "valid_targets_mean": 2370.8, |
| "valid_targets_min": 876 |
| }, |
| { |
| "epoch": 5.627272727272727, |
| "grad_norm": 0.6452905331420319, |
| "learning_rate": 4.5171677872183506e-06, |
| "loss": 0.2359, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2051079422235489, |
| "step": 3095, |
| "valid_targets_mean": 2581.9, |
| "valid_targets_min": 1404 |
| }, |
| { |
| "epoch": 5.636363636363637, |
| "grad_norm": 0.8950147725188888, |
| "learning_rate": 4.459934082549353e-06, |
| "loss": 0.1942, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2329583317041397, |
| "step": 3100, |
| "valid_targets_mean": 2573.9, |
| "valid_targets_min": 384 |
| }, |
| { |
| "epoch": 5.6454545454545455, |
| "grad_norm": 0.7301637246838704, |
| "learning_rate": 4.4030197414611344e-06, |
| "loss": 0.1995, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20358063280582428, |
| "step": 3105, |
| "valid_targets_mean": 2648.5, |
| "valid_targets_min": 787 |
| }, |
| { |
| "epoch": 5.654545454545454, |
| "grad_norm": 0.6920666316105244, |
| "learning_rate": 4.346425933599165e-06, |
| "loss": 0.2329, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23128581047058105, |
| "step": 3110, |
| "valid_targets_mean": 2874.4, |
| "valid_targets_min": 951 |
| }, |
| { |
| "epoch": 5.663636363636364, |
| "grad_norm": 0.6614747524469519, |
| "learning_rate": 4.2901538220216565e-06, |
| "loss": 0.1998, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23644432425498962, |
| "step": 3115, |
| "valid_targets_mean": 2487.2, |
| "valid_targets_min": 752 |
| }, |
| { |
| "epoch": 5.672727272727273, |
| "grad_norm": 0.5958281168548293, |
| "learning_rate": 4.234204563175625e-06, |
| "loss": 0.2067, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18708546459674835, |
| "step": 3120, |
| "valid_targets_mean": 2758.8, |
| "valid_targets_min": 1276 |
| }, |
| { |
| "epoch": 5.681818181818182, |
| "grad_norm": 0.6424913341469746, |
| "learning_rate": 4.17857930687318e-06, |
| "loss": 0.2037, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19697518646717072, |
| "step": 3125, |
| "valid_targets_mean": 2944.0, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 5.690909090909091, |
| "grad_norm": 0.7458618870765347, |
| "learning_rate": 4.123279196267815e-06, |
| "loss": 0.2095, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1647173911333084, |
| "step": 3130, |
| "valid_targets_mean": 2270.9, |
| "valid_targets_min": 1383 |
| }, |
| { |
| "epoch": 5.7, |
| "grad_norm": 0.599433937279404, |
| "learning_rate": 4.068305367831002e-06, |
| "loss": 0.2014, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18223410844802856, |
| "step": 3135, |
| "valid_targets_mean": 2552.9, |
| "valid_targets_min": 664 |
| }, |
| { |
| "epoch": 5.709090909090909, |
| "grad_norm": 0.730601336937522, |
| "learning_rate": 4.013658951328769e-06, |
| "loss": 0.2042, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2571181654930115, |
| "step": 3140, |
| "valid_targets_mean": 2819.1, |
| "valid_targets_min": 1035 |
| }, |
| { |
| "epoch": 5.718181818181818, |
| "grad_norm": 0.7330713546666404, |
| "learning_rate": 3.95934106979853e-06, |
| "loss": 0.2132, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22695882618427277, |
| "step": 3145, |
| "valid_targets_mean": 3382.2, |
| "valid_targets_min": 1015 |
| }, |
| { |
| "epoch": 5.7272727272727275, |
| "grad_norm": 0.6755758105804183, |
| "learning_rate": 3.905352839525962e-06, |
| "loss": 0.1898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1930406093597412, |
| "step": 3150, |
| "valid_targets_mean": 2554.9, |
| "valid_targets_min": 661 |
| }, |
| { |
| "epoch": 5.736363636363636, |
| "grad_norm": 0.6885232091114, |
| "learning_rate": 3.851695370022093e-06, |
| "loss": 0.206, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2078118473291397, |
| "step": 3155, |
| "valid_targets_mean": 2820.8, |
| "valid_targets_min": 1041 |
| }, |
| { |
| "epoch": 5.745454545454545, |
| "grad_norm": 0.6170351845839065, |
| "learning_rate": 3.7983697640005048e-06, |
| "loss": 0.203, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19898808002471924, |
| "step": 3160, |
| "valid_targets_mean": 2825.7, |
| "valid_targets_min": 1201 |
| }, |
| { |
| "epoch": 5.754545454545455, |
| "grad_norm": 0.6069878106397814, |
| "learning_rate": 3.7453771173546426e-06, |
| "loss": 0.2057, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20460426807403564, |
| "step": 3165, |
| "valid_targets_mean": 3422.9, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 5.763636363636364, |
| "grad_norm": 0.9297530833632488, |
| "learning_rate": 3.6927185191353188e-06, |
| "loss": 0.1754, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1826719045639038, |
| "step": 3170, |
| "valid_targets_mean": 2704.4, |
| "valid_targets_min": 1095 |
| }, |
| { |
| "epoch": 5.7727272727272725, |
| "grad_norm": 0.7803037852084785, |
| "learning_rate": 3.640395051528316e-06, |
| "loss": 0.2387, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20684343576431274, |
| "step": 3175, |
| "valid_targets_mean": 2536.1, |
| "valid_targets_min": 990 |
| }, |
| { |
| "epoch": 5.781818181818182, |
| "grad_norm": 0.6068693039253785, |
| "learning_rate": 3.5884077898321713e-06, |
| "loss": 0.2106, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16533026099205017, |
| "step": 3180, |
| "valid_targets_mean": 2962.0, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 5.790909090909091, |
| "grad_norm": 0.5920662855128672, |
| "learning_rate": 3.536757802436039e-06, |
| "loss": 0.203, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21043473482131958, |
| "step": 3185, |
| "valid_targets_mean": 3233.5, |
| "valid_targets_min": 1678 |
| }, |
| { |
| "epoch": 5.8, |
| "grad_norm": 0.6175569755361424, |
| "learning_rate": 3.4854461507977776e-06, |
| "loss": 0.2042, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1744406521320343, |
| "step": 3190, |
| "valid_targets_mean": 2969.8, |
| "valid_targets_min": 1746 |
| }, |
| { |
| "epoch": 5.809090909090909, |
| "grad_norm": 0.6499549912985492, |
| "learning_rate": 3.4344738894220964e-06, |
| "loss": 0.2186, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18466490507125854, |
| "step": 3195, |
| "valid_targets_mean": 2702.5, |
| "valid_targets_min": 1070 |
| }, |
| { |
| "epoch": 5.818181818181818, |
| "grad_norm": 0.6417333006893297, |
| "learning_rate": 3.383842065838907e-06, |
| "loss": 0.191, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1893327832221985, |
| "step": 3200, |
| "valid_targets_mean": 2411.5, |
| "valid_targets_min": 1231 |
| }, |
| { |
| "epoch": 5.827272727272727, |
| "grad_norm": 0.573769960160142, |
| "learning_rate": 3.3335517205818e-06, |
| "loss": 0.188, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21150127053260803, |
| "step": 3205, |
| "valid_targets_mean": 3080.4, |
| "valid_targets_min": 1333 |
| }, |
| { |
| "epoch": 5.836363636363636, |
| "grad_norm": 0.6773897146071632, |
| "learning_rate": 3.2836038871666444e-06, |
| "loss": 0.2023, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20320796966552734, |
| "step": 3210, |
| "valid_targets_mean": 2815.0, |
| "valid_targets_min": 1075 |
| }, |
| { |
| "epoch": 5.845454545454546, |
| "grad_norm": 0.6795811187617684, |
| "learning_rate": 3.2339995920703517e-06, |
| "loss": 0.2221, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2739708423614502, |
| "step": 3215, |
| "valid_targets_mean": 3073.7, |
| "valid_targets_min": 1029 |
| }, |
| { |
| "epoch": 5.8545454545454545, |
| "grad_norm": 0.633708555899195, |
| "learning_rate": 3.184739854709784e-06, |
| "loss": 0.2047, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25379064679145813, |
| "step": 3220, |
| "valid_targets_mean": 3036.8, |
| "valid_targets_min": 1249 |
| }, |
| { |
| "epoch": 5.863636363636363, |
| "grad_norm": 0.742362960713796, |
| "learning_rate": 3.1358256874208214e-06, |
| "loss": 0.2232, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1899518221616745, |
| "step": 3225, |
| "valid_targets_mean": 2378.8, |
| "valid_targets_min": 873 |
| }, |
| { |
| "epoch": 5.872727272727273, |
| "grad_norm": 0.6523582818519996, |
| "learning_rate": 3.0872580954375177e-06, |
| "loss": 0.203, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21299681067466736, |
| "step": 3230, |
| "valid_targets_mean": 3180.2, |
| "valid_targets_min": 1310 |
| }, |
| { |
| "epoch": 5.881818181818182, |
| "grad_norm": 0.6231922205998752, |
| "learning_rate": 3.039038076871481e-06, |
| "loss": 0.2108, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21740776300430298, |
| "step": 3235, |
| "valid_targets_mean": 2775.8, |
| "valid_targets_min": 1065 |
| }, |
| { |
| "epoch": 5.890909090909091, |
| "grad_norm": 0.7527402910704271, |
| "learning_rate": 2.9911666226913374e-06, |
| "loss": 0.2038, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2521063983440399, |
| "step": 3240, |
| "valid_targets_mean": 3143.5, |
| "valid_targets_min": 1289 |
| }, |
| { |
| "epoch": 5.9, |
| "grad_norm": 0.7085552104018334, |
| "learning_rate": 2.9436447167023674e-06, |
| "loss": 0.2313, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.27146613597869873, |
| "step": 3245, |
| "valid_targets_mean": 2884.2, |
| "valid_targets_min": 655 |
| }, |
| { |
| "epoch": 5.909090909090909, |
| "grad_norm": 0.6308634682071596, |
| "learning_rate": 2.896473335526313e-06, |
| "loss": 0.2239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23091664910316467, |
| "step": 3250, |
| "valid_targets_mean": 3103.6, |
| "valid_targets_min": 1442 |
| }, |
| { |
| "epoch": 5.918181818181818, |
| "grad_norm": 0.6557102602080491, |
| "learning_rate": 2.849653448581271e-06, |
| "loss": 0.1925, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.193063884973526, |
| "step": 3255, |
| "valid_targets_mean": 2485.2, |
| "valid_targets_min": 885 |
| }, |
| { |
| "epoch": 5.927272727272728, |
| "grad_norm": 0.6376956968805698, |
| "learning_rate": 2.8031860180617898e-06, |
| "loss": 0.2473, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25527384877204895, |
| "step": 3260, |
| "valid_targets_mean": 3161.6, |
| "valid_targets_min": 1225 |
| }, |
| { |
| "epoch": 5.9363636363636365, |
| "grad_norm": 0.6524169940045439, |
| "learning_rate": 2.757071998919094e-06, |
| "loss": 0.2084, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22266694903373718, |
| "step": 3265, |
| "valid_targets_mean": 3114.8, |
| "valid_targets_min": 1017 |
| }, |
| { |
| "epoch": 5.945454545454545, |
| "grad_norm": 0.6834100735609628, |
| "learning_rate": 2.7113123388414674e-06, |
| "loss": 0.2272, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22942128777503967, |
| "step": 3270, |
| "valid_targets_mean": 2606.1, |
| "valid_targets_min": 692 |
| }, |
| { |
| "epoch": 5.954545454545455, |
| "grad_norm": 0.6232577902181976, |
| "learning_rate": 2.665907978234754e-06, |
| "loss": 0.2212, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2783876061439514, |
| "step": 3275, |
| "valid_targets_mean": 3247.4, |
| "valid_targets_min": 1387 |
| }, |
| { |
| "epoch": 5.963636363636364, |
| "grad_norm": 0.6630897162275445, |
| "learning_rate": 2.6208598502030546e-06, |
| "loss": 0.2265, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2094637155532837, |
| "step": 3280, |
| "valid_targets_mean": 2547.4, |
| "valid_targets_min": 1126 |
| }, |
| { |
| "epoch": 5.972727272727273, |
| "grad_norm": 0.7077469988970768, |
| "learning_rate": 2.5761688805295305e-06, |
| "loss": 0.1933, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22009465098381042, |
| "step": 3285, |
| "valid_targets_mean": 2548.6, |
| "valid_targets_min": 1294 |
| }, |
| { |
| "epoch": 5.9818181818181815, |
| "grad_norm": 0.6715778542941565, |
| "learning_rate": 2.531835987657407e-06, |
| "loss": 0.2145, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19744186103343964, |
| "step": 3290, |
| "valid_targets_mean": 2664.1, |
| "valid_targets_min": 602 |
| }, |
| { |
| "epoch": 5.990909090909091, |
| "grad_norm": 0.7233321904636786, |
| "learning_rate": 2.487862082671064e-06, |
| "loss": 0.2175, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24416381120681763, |
| "step": 3295, |
| "valid_targets_mean": 2344.6, |
| "valid_targets_min": 600 |
| }, |
| { |
| "epoch": 6.0, |
| "grad_norm": 0.6660871787495303, |
| "learning_rate": 2.4442480692773398e-06, |
| "loss": 0.2042, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18832653760910034, |
| "step": 3300, |
| "valid_targets_mean": 2622.1, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 6.009090909090909, |
| "grad_norm": 0.7375102372730574, |
| "learning_rate": 2.400994843786939e-06, |
| "loss": 0.2019, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2507600784301758, |
| "step": 3305, |
| "valid_targets_mean": 2900.4, |
| "valid_targets_min": 672 |
| }, |
| { |
| "epoch": 6.0181818181818185, |
| "grad_norm": 0.6661010728917941, |
| "learning_rate": 2.3581032950960215e-06, |
| "loss": 0.2104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2171628773212433, |
| "step": 3310, |
| "valid_targets_mean": 2410.8, |
| "valid_targets_min": 1208 |
| }, |
| { |
| "epoch": 6.027272727272727, |
| "grad_norm": 0.6587778839047856, |
| "learning_rate": 2.3155743046679468e-06, |
| "loss": 0.1866, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18798887729644775, |
| "step": 3315, |
| "valid_targets_mean": 2753.6, |
| "valid_targets_min": 923 |
| }, |
| { |
| "epoch": 6.036363636363636, |
| "grad_norm": 0.6565679632927076, |
| "learning_rate": 2.273408746515133e-06, |
| "loss": 0.2033, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18916954100131989, |
| "step": 3320, |
| "valid_targets_mean": 2660.1, |
| "valid_targets_min": 839 |
| }, |
| { |
| "epoch": 6.045454545454546, |
| "grad_norm": 0.668772335037644, |
| "learning_rate": 2.2316074871811157e-06, |
| "loss": 0.1858, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17284397780895233, |
| "step": 3325, |
| "valid_targets_mean": 2571.2, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 6.054545454545455, |
| "grad_norm": 0.7117814748250814, |
| "learning_rate": 2.190171385722726e-06, |
| "loss": 0.2067, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22605367004871368, |
| "step": 3330, |
| "valid_targets_mean": 2528.2, |
| "valid_targets_min": 1316 |
| }, |
| { |
| "epoch": 6.0636363636363635, |
| "grad_norm": 0.6894262085614624, |
| "learning_rate": 2.1491012936924548e-06, |
| "loss": 0.2192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20072168111801147, |
| "step": 3335, |
| "valid_targets_mean": 2462.0, |
| "valid_targets_min": 1177 |
| }, |
| { |
| "epoch": 6.072727272727272, |
| "grad_norm": 0.6221932962788573, |
| "learning_rate": 2.108398055120926e-06, |
| "loss": 0.1981, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18132944405078888, |
| "step": 3340, |
| "valid_targets_mean": 2793.1, |
| "valid_targets_min": 1665 |
| }, |
| { |
| "epoch": 6.081818181818182, |
| "grad_norm": 0.5952761142844478, |
| "learning_rate": 2.068062506499584e-06, |
| "loss": 0.1902, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.15214930474758148, |
| "step": 3345, |
| "valid_targets_mean": 2650.2, |
| "valid_targets_min": 1429 |
| }, |
| { |
| "epoch": 6.090909090909091, |
| "grad_norm": 0.6486703925050182, |
| "learning_rate": 2.0280954767634674e-06, |
| "loss": 0.2117, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19928964972496033, |
| "step": 3350, |
| "valid_targets_mean": 2931.4, |
| "valid_targets_min": 1410 |
| }, |
| { |
| "epoch": 6.1, |
| "grad_norm": 0.6431480925390909, |
| "learning_rate": 1.988497787274195e-06, |
| "loss": 0.1851, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2235788255929947, |
| "step": 3355, |
| "valid_targets_mean": 3253.4, |
| "valid_targets_min": 682 |
| }, |
| { |
| "epoch": 6.109090909090909, |
| "grad_norm": 0.669316954228609, |
| "learning_rate": 1.9492702518030905e-06, |
| "loss": 0.2015, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20945610105991364, |
| "step": 3360, |
| "valid_targets_mean": 3001.4, |
| "valid_targets_min": 1636 |
| }, |
| { |
| "epoch": 6.118181818181818, |
| "grad_norm": 0.6446973463301628, |
| "learning_rate": 1.910413676514438e-06, |
| "loss": 0.186, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1692398488521576, |
| "step": 3365, |
| "valid_targets_mean": 2664.7, |
| "valid_targets_min": 778 |
| }, |
| { |
| "epoch": 6.127272727272727, |
| "grad_norm": 0.7129875180586859, |
| "learning_rate": 1.8719288599489304e-06, |
| "loss": 0.2172, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24184544384479523, |
| "step": 3370, |
| "valid_targets_mean": 2537.6, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 6.136363636363637, |
| "grad_norm": 0.6831168369791281, |
| "learning_rate": 1.833816593007256e-06, |
| "loss": 0.2118, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1838299185037613, |
| "step": 3375, |
| "valid_targets_mean": 2457.5, |
| "valid_targets_min": 1541 |
| }, |
| { |
| "epoch": 6.1454545454545455, |
| "grad_norm": 0.6179852974797505, |
| "learning_rate": 1.796077658933848e-06, |
| "loss": 0.1992, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19142737984657288, |
| "step": 3380, |
| "valid_targets_mean": 3180.6, |
| "valid_targets_min": 1035 |
| }, |
| { |
| "epoch": 6.154545454545454, |
| "grad_norm": 0.7174728400905155, |
| "learning_rate": 1.7587128333007709e-06, |
| "loss": 0.1876, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1669439673423767, |
| "step": 3385, |
| "valid_targets_mean": 2359.4, |
| "valid_targets_min": 1171 |
| }, |
| { |
| "epoch": 6.163636363636364, |
| "grad_norm": 0.7075411705418458, |
| "learning_rate": 1.7217228839918098e-06, |
| "loss": 0.2152, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20773687958717346, |
| "step": 3390, |
| "valid_targets_mean": 2524.9, |
| "valid_targets_min": 1178 |
| }, |
| { |
| "epoch": 6.172727272727273, |
| "grad_norm": 0.6514407325688949, |
| "learning_rate": 1.6851085711866598e-06, |
| "loss": 0.1921, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18543276190757751, |
| "step": 3395, |
| "valid_targets_mean": 2648.4, |
| "valid_targets_min": 968 |
| }, |
| { |
| "epoch": 6.181818181818182, |
| "grad_norm": 0.6901703887890732, |
| "learning_rate": 1.648870647345322e-06, |
| "loss": 0.1766, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1513475477695465, |
| "step": 3400, |
| "valid_targets_mean": 2702.0, |
| "valid_targets_min": 1100 |
| }, |
| { |
| "epoch": 6.190909090909091, |
| "grad_norm": 0.6936967662817098, |
| "learning_rate": 1.6130098571926468e-06, |
| "loss": 0.2055, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23872001469135284, |
| "step": 3405, |
| "valid_targets_mean": 2802.6, |
| "valid_targets_min": 1819 |
| }, |
| { |
| "epoch": 6.2, |
| "grad_norm": 0.954901470393649, |
| "learning_rate": 1.577526937703e-06, |
| "loss": 0.2068, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23395466804504395, |
| "step": 3410, |
| "valid_targets_mean": 2999.4, |
| "valid_targets_min": 1309 |
| }, |
| { |
| "epoch": 6.209090909090909, |
| "grad_norm": 0.7301417694762667, |
| "learning_rate": 1.5424226180851443e-06, |
| "loss": 0.1896, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22834432125091553, |
| "step": 3415, |
| "valid_targets_mean": 2777.3, |
| "valid_targets_min": 1019 |
| }, |
| { |
| "epoch": 6.218181818181818, |
| "grad_norm": 0.722443584662105, |
| "learning_rate": 1.5076976197672432e-06, |
| "loss": 0.2091, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21865253150463104, |
| "step": 3420, |
| "valid_targets_mean": 2611.9, |
| "valid_targets_min": 859 |
| }, |
| { |
| "epoch": 6.2272727272727275, |
| "grad_norm": 0.6718376412603478, |
| "learning_rate": 1.473352656382039e-06, |
| "loss": 0.1625, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1492437869310379, |
| "step": 3425, |
| "valid_targets_mean": 2268.1, |
| "valid_targets_min": 1306 |
| }, |
| { |
| "epoch": 6.236363636363636, |
| "grad_norm": 0.6385879578869108, |
| "learning_rate": 1.439388433752178e-06, |
| "loss": 0.1971, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21833261847496033, |
| "step": 3430, |
| "valid_targets_mean": 3357.3, |
| "valid_targets_min": 1110 |
| }, |
| { |
| "epoch": 6.245454545454545, |
| "grad_norm": 0.6898363347432659, |
| "learning_rate": 1.4058056498757112e-06, |
| "loss": 0.2109, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21196958422660828, |
| "step": 3435, |
| "valid_targets_mean": 2529.1, |
| "valid_targets_min": 884 |
| }, |
| { |
| "epoch": 6.254545454545455, |
| "grad_norm": 0.6717448075871523, |
| "learning_rate": 1.372604994911757e-06, |
| "loss": 0.2076, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18704624474048615, |
| "step": 3440, |
| "valid_targets_mean": 2528.5, |
| "valid_targets_min": 811 |
| }, |
| { |
| "epoch": 6.263636363636364, |
| "grad_norm": 0.6170083596984325, |
| "learning_rate": 1.3397871511662986e-06, |
| "loss": 0.2132, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17854154109954834, |
| "step": 3445, |
| "valid_targets_mean": 2765.4, |
| "valid_targets_min": 1422 |
| }, |
| { |
| "epoch": 6.2727272727272725, |
| "grad_norm": 0.5945835881991018, |
| "learning_rate": 1.307352793078187e-06, |
| "loss": 0.1855, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20941226184368134, |
| "step": 3450, |
| "valid_targets_mean": 3193.4, |
| "valid_targets_min": 973 |
| }, |
| { |
| "epoch": 6.281818181818182, |
| "grad_norm": 0.7318339507251856, |
| "learning_rate": 1.275302587205256e-06, |
| "loss": 0.2191, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22563597559928894, |
| "step": 3455, |
| "valid_targets_mean": 2949.8, |
| "valid_targets_min": 706 |
| }, |
| { |
| "epoch": 6.290909090909091, |
| "grad_norm": 0.5851474607507773, |
| "learning_rate": 1.2436371922106404e-06, |
| "loss": 0.209, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22322428226470947, |
| "step": 3460, |
| "valid_targets_mean": 3700.2, |
| "valid_targets_min": 1437 |
| }, |
| { |
| "epoch": 6.3, |
| "grad_norm": 0.5554704922918074, |
| "learning_rate": 1.2123572588492306e-06, |
| "loss": 0.1906, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18534304201602936, |
| "step": 3465, |
| "valid_targets_mean": 3815.0, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 6.309090909090909, |
| "grad_norm": 0.6342467907091855, |
| "learning_rate": 1.1814634299543103e-06, |
| "loss": 0.1845, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20048777759075165, |
| "step": 3470, |
| "valid_targets_mean": 2843.0, |
| "valid_targets_min": 1148 |
| }, |
| { |
| "epoch": 6.318181818181818, |
| "grad_norm": 0.6654750037522876, |
| "learning_rate": 1.1509563404243274e-06, |
| "loss": 0.2293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21820786595344543, |
| "step": 3475, |
| "valid_targets_mean": 2823.8, |
| "valid_targets_min": 1081 |
| }, |
| { |
| "epoch": 6.327272727272727, |
| "grad_norm": 0.6803497267833459, |
| "learning_rate": 1.1208366172098684e-06, |
| "loss": 0.1835, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21640288829803467, |
| "step": 3480, |
| "valid_targets_mean": 2524.6, |
| "valid_targets_min": 1275 |
| }, |
| { |
| "epoch": 6.336363636363636, |
| "grad_norm": 0.68657482835238, |
| "learning_rate": 1.0911048793007484e-06, |
| "loss": 0.2174, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23357044160366058, |
| "step": 3485, |
| "valid_targets_mean": 3060.0, |
| "valid_targets_min": 1442 |
| }, |
| { |
| "epoch": 6.345454545454546, |
| "grad_norm": 0.6951836522091235, |
| "learning_rate": 1.0617617377133205e-06, |
| "loss": 0.2264, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21933671832084656, |
| "step": 3490, |
| "valid_targets_mean": 3150.7, |
| "valid_targets_min": 1106 |
| }, |
| { |
| "epoch": 6.3545454545454545, |
| "grad_norm": 0.9540180459343338, |
| "learning_rate": 1.0328077954778904e-06, |
| "loss": 0.2126, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21296435594558716, |
| "step": 3495, |
| "valid_targets_mean": 2526.2, |
| "valid_targets_min": 1252 |
| }, |
| { |
| "epoch": 6.363636363636363, |
| "grad_norm": 0.7215486794128292, |
| "learning_rate": 1.004243647626344e-06, |
| "loss": 0.2293, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2184891402721405, |
| "step": 3500, |
| "valid_targets_mean": 2418.1, |
| "valid_targets_min": 782 |
| }, |
| { |
| "epoch": 6.372727272727273, |
| "grad_norm": 0.7230020636540598, |
| "learning_rate": 9.760698811799064e-07, |
| "loss": 0.204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2301444411277771, |
| "step": 3505, |
| "valid_targets_mean": 2908.6, |
| "valid_targets_min": 866 |
| }, |
| { |
| "epoch": 6.381818181818182, |
| "grad_norm": 0.6685846433200956, |
| "learning_rate": 9.482870751370755e-07, |
| "loss": 0.1975, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17493730783462524, |
| "step": 3510, |
| "valid_targets_mean": 2570.5, |
| "valid_targets_min": 846 |
| }, |
| { |
| "epoch": 6.390909090909091, |
| "grad_norm": 0.6689691547111833, |
| "learning_rate": 9.208958004617475e-07, |
| "loss": 0.2175, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24776533246040344, |
| "step": 3515, |
| "valid_targets_mean": 2873.2, |
| "valid_targets_min": 1244 |
| }, |
| { |
| "epoch": 6.4, |
| "grad_norm": 0.7556459012137372, |
| "learning_rate": 8.938966200714482e-07, |
| "loss": 0.1883, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2316136360168457, |
| "step": 3520, |
| "valid_targets_mean": 2475.9, |
| "valid_targets_min": 1078 |
| }, |
| { |
| "epoch": 6.409090909090909, |
| "grad_norm": 0.6585887436395376, |
| "learning_rate": 8.672900888257918e-07, |
| "loss": 0.2191, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.247239351272583, |
| "step": 3525, |
| "valid_targets_mean": 3246.8, |
| "valid_targets_min": 1413 |
| }, |
| { |
| "epoch": 6.418181818181818, |
| "grad_norm": 0.6424132876151116, |
| "learning_rate": 8.410767535150599e-07, |
| "loss": 0.2079, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1996202915906906, |
| "step": 3530, |
| "valid_targets_mean": 2780.2, |
| "valid_targets_min": 513 |
| }, |
| { |
| "epoch": 6.427272727272728, |
| "grad_norm": 0.6848931844740308, |
| "learning_rate": 8.152571528489828e-07, |
| "loss": 0.1997, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19339044392108917, |
| "step": 3535, |
| "valid_targets_mean": 2968.0, |
| "valid_targets_min": 962 |
| }, |
| { |
| "epoch": 6.4363636363636365, |
| "grad_norm": 0.7963468724693825, |
| "learning_rate": 7.898318174456498e-07, |
| "loss": 0.2011, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.25005707144737244, |
| "step": 3540, |
| "valid_targets_mean": 2918.1, |
| "valid_targets_min": 1036 |
| }, |
| { |
| "epoch": 6.445454545454545, |
| "grad_norm": 0.6564266424153401, |
| "learning_rate": 7.64801269820612e-07, |
| "loss": 0.2247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17174042761325836, |
| "step": 3545, |
| "valid_targets_mean": 2493.9, |
| "valid_targets_min": 796 |
| }, |
| { |
| "epoch": 6.454545454545454, |
| "grad_norm": 0.7142428582801825, |
| "learning_rate": 7.401660243761543e-07, |
| "loss": 0.1944, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23733559250831604, |
| "step": 3550, |
| "valid_targets_mean": 2868.8, |
| "valid_targets_min": 1479 |
| }, |
| { |
| "epoch": 6.463636363636364, |
| "grad_norm": 0.7267944382985336, |
| "learning_rate": 7.159265873907006e-07, |
| "loss": 0.187, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23389649391174316, |
| "step": 3555, |
| "valid_targets_mean": 2494.9, |
| "valid_targets_min": 1107 |
| }, |
| { |
| "epoch": 6.472727272727273, |
| "grad_norm": 0.6732130872138381, |
| "learning_rate": 6.920834570084389e-07, |
| "loss": 0.2185, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21910206973552704, |
| "step": 3560, |
| "valid_targets_mean": 2901.8, |
| "valid_targets_min": 1316 |
| }, |
| { |
| "epoch": 6.4818181818181815, |
| "grad_norm": 0.6204104076430839, |
| "learning_rate": 6.686371232290567e-07, |
| "loss": 0.1889, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19802677631378174, |
| "step": 3565, |
| "valid_targets_mean": 3183.4, |
| "valid_targets_min": 1297 |
| }, |
| { |
| "epoch": 6.490909090909091, |
| "grad_norm": 1.0729417915230997, |
| "learning_rate": 6.455880678976845e-07, |
| "loss": 0.2095, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2234289050102234, |
| "step": 3570, |
| "valid_targets_mean": 3047.2, |
| "valid_targets_min": 1492 |
| }, |
| { |
| "epoch": 6.5, |
| "grad_norm": 0.6170422752355285, |
| "learning_rate": 6.229367646949924e-07, |
| "loss": 0.1958, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1874658465385437, |
| "step": 3575, |
| "valid_targets_mean": 2822.6, |
| "valid_targets_min": 871 |
| }, |
| { |
| "epoch": 6.509090909090909, |
| "grad_norm": 0.665717657809791, |
| "learning_rate": 6.006836791274606e-07, |
| "loss": 0.1916, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22099323570728302, |
| "step": 3580, |
| "valid_targets_mean": 2710.2, |
| "valid_targets_min": 1198 |
| }, |
| { |
| "epoch": 6.5181818181818185, |
| "grad_norm": 0.6951502959945852, |
| "learning_rate": 5.788292685177954e-07, |
| "loss": 0.2122, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18100181221961975, |
| "step": 3585, |
| "valid_targets_mean": 2566.7, |
| "valid_targets_min": 1336 |
| }, |
| { |
| "epoch": 6.527272727272727, |
| "grad_norm": 0.6090897306366719, |
| "learning_rate": 5.573739819955459e-07, |
| "loss": 0.2107, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18592379987239838, |
| "step": 3590, |
| "valid_targets_mean": 2851.6, |
| "valid_targets_min": 1019 |
| }, |
| { |
| "epoch": 6.536363636363636, |
| "grad_norm": 0.6839505277779282, |
| "learning_rate": 5.363182604878803e-07, |
| "loss": 0.2159, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22848966717720032, |
| "step": 3595, |
| "valid_targets_mean": 2868.8, |
| "valid_targets_min": 509 |
| }, |
| { |
| "epoch": 6.545454545454545, |
| "grad_norm": 0.5902724277197199, |
| "learning_rate": 5.156625367104973e-07, |
| "loss": 0.1879, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17370611429214478, |
| "step": 3600, |
| "valid_targets_mean": 3350.9, |
| "valid_targets_min": 1023 |
| }, |
| { |
| "epoch": 6.554545454545455, |
| "grad_norm": 0.5824926455184319, |
| "learning_rate": 4.954072351587646e-07, |
| "loss": 0.2004, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2094963937997818, |
| "step": 3605, |
| "valid_targets_mean": 3443.2, |
| "valid_targets_min": 1756 |
| }, |
| { |
| "epoch": 6.5636363636363635, |
| "grad_norm": 0.7223293649275785, |
| "learning_rate": 4.75552772098975e-07, |
| "loss": 0.1998, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22199690341949463, |
| "step": 3610, |
| "valid_targets_mean": 2564.2, |
| "valid_targets_min": 1198 |
| }, |
| { |
| "epoch": 6.572727272727272, |
| "grad_norm": 0.6704296286028131, |
| "learning_rate": 4.560995555597969e-07, |
| "loss": 0.2145, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20293614268302917, |
| "step": 3615, |
| "valid_targets_mean": 2849.2, |
| "valid_targets_min": 925 |
| }, |
| { |
| "epoch": 6.581818181818182, |
| "grad_norm": 0.7454734458543368, |
| "learning_rate": 4.3704798532388624e-07, |
| "loss": 0.247, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.3295115828514099, |
| "step": 3620, |
| "valid_targets_mean": 2756.9, |
| "valid_targets_min": 839 |
| }, |
| { |
| "epoch": 6.590909090909091, |
| "grad_norm": 0.7071629110400431, |
| "learning_rate": 4.1839845291968607e-07, |
| "loss": 0.2165, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23328623175621033, |
| "step": 3625, |
| "valid_targets_mean": 3020.6, |
| "valid_targets_min": 976 |
| }, |
| { |
| "epoch": 6.6, |
| "grad_norm": 0.6733228752450922, |
| "learning_rate": 4.001513416133551e-07, |
| "loss": 0.1855, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1846185326576233, |
| "step": 3630, |
| "valid_targets_mean": 2510.1, |
| "valid_targets_min": 1328 |
| }, |
| { |
| "epoch": 6.609090909090909, |
| "grad_norm": 0.7503433569098521, |
| "learning_rate": 3.823070264009099e-07, |
| "loss": 0.1863, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16027329862117767, |
| "step": 3635, |
| "valid_targets_mean": 2422.6, |
| "valid_targets_min": 1004 |
| }, |
| { |
| "epoch": 6.618181818181818, |
| "grad_norm": 0.7548043438842913, |
| "learning_rate": 3.648658740005107e-07, |
| "loss": 0.1785, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1915736049413681, |
| "step": 3640, |
| "valid_targets_mean": 2116.1, |
| "valid_targets_min": 899 |
| }, |
| { |
| "epoch": 6.627272727272727, |
| "grad_norm": 0.6091929259291805, |
| "learning_rate": 3.4782824284492975e-07, |
| "loss": 0.1998, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2477787882089615, |
| "step": 3645, |
| "valid_targets_mean": 3713.2, |
| "valid_targets_min": 1872 |
| }, |
| { |
| "epoch": 6.636363636363637, |
| "grad_norm": 0.6229880345830455, |
| "learning_rate": 3.31194483074182e-07, |
| "loss": 0.1963, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1972864866256714, |
| "step": 3650, |
| "valid_targets_mean": 2992.7, |
| "valid_targets_min": 1513 |
| }, |
| { |
| "epoch": 6.6454545454545455, |
| "grad_norm": 0.628801173843993, |
| "learning_rate": 3.149649365283258e-07, |
| "loss": 0.2056, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2724452018737793, |
| "step": 3655, |
| "valid_targets_mean": 3391.6, |
| "valid_targets_min": 1819 |
| }, |
| { |
| "epoch": 6.654545454545454, |
| "grad_norm": 0.6261436041334418, |
| "learning_rate": 2.9913993674044904e-07, |
| "loss": 0.1983, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18547657132148743, |
| "step": 3660, |
| "valid_targets_mean": 2915.8, |
| "valid_targets_min": 1197 |
| }, |
| { |
| "epoch": 6.663636363636364, |
| "grad_norm": 0.7463194929612665, |
| "learning_rate": 2.8371980892979436e-07, |
| "loss": 0.2333, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22626563906669617, |
| "step": 3665, |
| "valid_targets_mean": 2315.8, |
| "valid_targets_min": 600 |
| }, |
| { |
| "epoch": 6.672727272727273, |
| "grad_norm": 0.6492852106251501, |
| "learning_rate": 2.687048699951067e-07, |
| "loss": 0.1793, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19556567072868347, |
| "step": 3670, |
| "valid_targets_mean": 3156.1, |
| "valid_targets_min": 1442 |
| }, |
| { |
| "epoch": 6.681818181818182, |
| "grad_norm": 0.7197581196148153, |
| "learning_rate": 2.5409542850808765e-07, |
| "loss": 0.195, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2314002513885498, |
| "step": 3675, |
| "valid_targets_mean": 2628.3, |
| "valid_targets_min": 1269 |
| }, |
| { |
| "epoch": 6.690909090909091, |
| "grad_norm": 0.6782253107747308, |
| "learning_rate": 2.3989178470707364e-07, |
| "loss": 0.204, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21835069358348846, |
| "step": 3680, |
| "valid_targets_mean": 2813.1, |
| "valid_targets_min": 776 |
| }, |
| { |
| "epoch": 6.7, |
| "grad_norm": 0.738040860037262, |
| "learning_rate": 2.260942304908609e-07, |
| "loss": 0.1898, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2219957709312439, |
| "step": 3685, |
| "valid_targets_mean": 2700.1, |
| "valid_targets_min": 1276 |
| }, |
| { |
| "epoch": 6.709090909090909, |
| "grad_norm": 0.6134241635817874, |
| "learning_rate": 2.1270304941271025e-07, |
| "loss": 0.1914, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.14740461111068726, |
| "step": 3690, |
| "valid_targets_mean": 2586.2, |
| "valid_targets_min": 1251 |
| }, |
| { |
| "epoch": 6.718181818181818, |
| "grad_norm": 0.693132393478107, |
| "learning_rate": 1.9971851667451413e-07, |
| "loss": 0.2047, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23066049814224243, |
| "step": 3695, |
| "valid_targets_mean": 2906.5, |
| "valid_targets_min": 700 |
| }, |
| { |
| "epoch": 6.7272727272727275, |
| "grad_norm": 0.6619328999399736, |
| "learning_rate": 1.8714089912113876e-07, |
| "loss": 0.2192, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1960105299949646, |
| "step": 3700, |
| "valid_targets_mean": 3064.5, |
| "valid_targets_min": 1029 |
| }, |
| { |
| "epoch": 6.736363636363636, |
| "grad_norm": 0.6829964172452905, |
| "learning_rate": 1.749704552349507e-07, |
| "loss": 0.1859, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1751609444618225, |
| "step": 3705, |
| "valid_targets_mean": 2536.2, |
| "valid_targets_min": 849 |
| }, |
| { |
| "epoch": 6.745454545454545, |
| "grad_norm": 0.5770159628606195, |
| "learning_rate": 1.6320743513049686e-07, |
| "loss": 0.2, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.15475375950336456, |
| "step": 3710, |
| "valid_targets_mean": 2827.8, |
| "valid_targets_min": 910 |
| }, |
| { |
| "epoch": 6.754545454545455, |
| "grad_norm": 0.6445532521114699, |
| "learning_rate": 1.5185208054936394e-07, |
| "loss": 0.221, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16979286074638367, |
| "step": 3715, |
| "valid_targets_mean": 2648.4, |
| "valid_targets_min": 1291 |
| }, |
| { |
| "epoch": 6.763636363636364, |
| "grad_norm": 0.5943719803037179, |
| "learning_rate": 1.4090462485521816e-07, |
| "loss": 0.2099, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18774190545082092, |
| "step": 3720, |
| "valid_targets_mean": 2772.4, |
| "valid_targets_min": 1199 |
| }, |
| { |
| "epoch": 6.7727272727272725, |
| "grad_norm": 0.6756921790231029, |
| "learning_rate": 1.303652930289956e-07, |
| "loss": 0.1938, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21918311715126038, |
| "step": 3725, |
| "valid_targets_mean": 2937.1, |
| "valid_targets_min": 1241 |
| }, |
| { |
| "epoch": 6.781818181818182, |
| "grad_norm": 0.6616175262354211, |
| "learning_rate": 1.2023430166429485e-07, |
| "loss": 0.1735, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16985666751861572, |
| "step": 3730, |
| "valid_targets_mean": 2460.4, |
| "valid_targets_min": 987 |
| }, |
| { |
| "epoch": 6.790909090909091, |
| "grad_norm": 0.7910618205562641, |
| "learning_rate": 1.1051185896291616e-07, |
| "loss": 0.1924, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21274155378341675, |
| "step": 3735, |
| "valid_targets_mean": 2476.2, |
| "valid_targets_min": 886 |
| }, |
| { |
| "epoch": 6.8, |
| "grad_norm": 0.7154269068956903, |
| "learning_rate": 1.011981647305782e-07, |
| "loss": 0.2056, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.21139511466026306, |
| "step": 3740, |
| "valid_targets_mean": 2405.8, |
| "valid_targets_min": 1248 |
| }, |
| { |
| "epoch": 6.809090909090909, |
| "grad_norm": 0.7284484849008579, |
| "learning_rate": 9.22934103728279e-08, |
| "loss": 0.2117, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.24537286162376404, |
| "step": 3745, |
| "valid_targets_mean": 2740.6, |
| "valid_targets_min": 870 |
| }, |
| { |
| "epoch": 6.818181818181818, |
| "grad_norm": 0.6847980903768206, |
| "learning_rate": 8.37977788910882e-08, |
| "loss": 0.1934, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2240789234638214, |
| "step": 3750, |
| "valid_targets_mean": 2620.1, |
| "valid_targets_min": 1441 |
| }, |
| { |
| "epoch": 6.827272727272727, |
| "grad_norm": 0.6915252822860436, |
| "learning_rate": 7.571144487891202e-08, |
| "loss": 0.2226, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2609437108039856, |
| "step": 3755, |
| "valid_targets_mean": 3013.9, |
| "valid_targets_min": 802 |
| }, |
| { |
| "epoch": 6.836363636363636, |
| "grad_norm": 0.7415434992825277, |
| "learning_rate": 6.803457451838746e-08, |
| "loss": 0.2094, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19083890318870544, |
| "step": 3760, |
| "valid_targets_mean": 2380.1, |
| "valid_targets_min": 1347 |
| }, |
| { |
| "epoch": 6.845454545454546, |
| "grad_norm": 0.6739299134659402, |
| "learning_rate": 6.076732557672272e-08, |
| "loss": 0.1937, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1840040683746338, |
| "step": 3765, |
| "valid_targets_mean": 2367.7, |
| "valid_targets_min": 1454 |
| }, |
| { |
| "epoch": 6.8545454545454545, |
| "grad_norm": 0.7509615032546884, |
| "learning_rate": 5.390984740299976e-08, |
| "loss": 0.2091, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.23984074592590332, |
| "step": 3770, |
| "valid_targets_mean": 2647.2, |
| "valid_targets_min": 752 |
| }, |
| { |
| "epoch": 6.863636363636363, |
| "grad_norm": 0.6790819350475567, |
| "learning_rate": 4.7462280925116847e-08, |
| "loss": 0.1877, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16014784574508667, |
| "step": 3775, |
| "valid_targets_mean": 2598.4, |
| "valid_targets_min": 1736 |
| }, |
| { |
| "epoch": 6.872727272727273, |
| "grad_norm": 0.650157897278632, |
| "learning_rate": 4.142475864688411e-08, |
| "loss": 0.1967, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17384964227676392, |
| "step": 3780, |
| "valid_targets_mean": 2974.6, |
| "valid_targets_min": 393 |
| }, |
| { |
| "epoch": 6.881818181818182, |
| "grad_norm": 0.6659622954454439, |
| "learning_rate": 3.5797404645296906e-08, |
| "loss": 0.2172, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26057201623916626, |
| "step": 3785, |
| "valid_targets_mean": 2915.7, |
| "valid_targets_min": 986 |
| }, |
| { |
| "epoch": 6.890909090909091, |
| "grad_norm": 0.6656737432362372, |
| "learning_rate": 3.0580334567995585e-08, |
| "loss": 0.2104, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.1925896406173706, |
| "step": 3790, |
| "valid_targets_mean": 2642.1, |
| "valid_targets_min": 1109 |
| }, |
| { |
| "epoch": 6.9, |
| "grad_norm": 0.6694609275277352, |
| "learning_rate": 2.5773655630880746e-08, |
| "loss": 0.1896, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16200029850006104, |
| "step": 3795, |
| "valid_targets_mean": 2400.2, |
| "valid_targets_min": 830 |
| }, |
| { |
| "epoch": 6.909090909090909, |
| "grad_norm": 0.6807347889354627, |
| "learning_rate": 2.1377466615912778e-08, |
| "loss": 0.2413, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.26320990920066833, |
| "step": 3800, |
| "valid_targets_mean": 2810.3, |
| "valid_targets_min": 1130 |
| }, |
| { |
| "epoch": 6.918181818181818, |
| "grad_norm": 0.7059977237079785, |
| "learning_rate": 1.7391857869086815e-08, |
| "loss": 0.2071, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2371351718902588, |
| "step": 3805, |
| "valid_targets_mean": 2632.0, |
| "valid_targets_min": 888 |
| }, |
| { |
| "epoch": 6.927272727272728, |
| "grad_norm": 0.740272454756386, |
| "learning_rate": 1.3816911298565327e-08, |
| "loss": 0.2239, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19789424538612366, |
| "step": 3810, |
| "valid_targets_mean": 2359.9, |
| "valid_targets_min": 939 |
| }, |
| { |
| "epoch": 6.9363636363636365, |
| "grad_norm": 0.5823718609644692, |
| "learning_rate": 1.0652700373006142e-08, |
| "loss": 0.2081, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17436788976192474, |
| "step": 3815, |
| "valid_targets_mean": 2827.7, |
| "valid_targets_min": 1378 |
| }, |
| { |
| "epoch": 6.945454545454545, |
| "grad_norm": 0.6873846620440276, |
| "learning_rate": 7.899290120039205e-09, |
| "loss": 0.1954, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.2165110558271408, |
| "step": 3820, |
| "valid_targets_mean": 2476.6, |
| "valid_targets_min": 1096 |
| }, |
| { |
| "epoch": 6.954545454545455, |
| "grad_norm": 0.7150954695545816, |
| "learning_rate": 5.556737124945422e-09, |
| "loss": 0.1973, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.20282520353794098, |
| "step": 3825, |
| "valid_targets_mean": 2931.9, |
| "valid_targets_min": 1063 |
| }, |
| { |
| "epoch": 6.963636363636364, |
| "grad_norm": 0.6049679393727518, |
| "learning_rate": 3.6250895294842605e-09, |
| "loss": 0.2149, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.19065088033676147, |
| "step": 3830, |
| "valid_targets_mean": 2879.9, |
| "valid_targets_min": 878 |
| }, |
| { |
| "epoch": 6.972727272727273, |
| "grad_norm": 0.6292861196115748, |
| "learning_rate": 2.1043870309078727e-09, |
| "loss": 0.1856, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.22311565279960632, |
| "step": 3835, |
| "valid_targets_mean": 2907.8, |
| "valid_targets_min": 1273 |
| }, |
| { |
| "epoch": 6.9818181818181815, |
| "grad_norm": 0.6456495609565179, |
| "learning_rate": 9.946608811395308e-10, |
| "loss": 0.1887, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.17371340095996857, |
| "step": 3840, |
| "valid_targets_mean": 2698.8, |
| "valid_targets_min": 1180 |
| }, |
| { |
| "epoch": 6.990909090909091, |
| "grad_norm": 0.6481073142467838, |
| "learning_rate": 2.959338861407979e-10, |
| "loss": 0.2125, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.18638712167739868, |
| "step": 3845, |
| "valid_targets_mean": 2776.0, |
| "valid_targets_min": 1351 |
| }, |
| { |
| "epoch": 7.0, |
| "grad_norm": 0.6608342332932761, |
| "learning_rate": 8.220405436354384e-12, |
| "loss": 0.2038, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16949467360973358, |
| "step": 3850, |
| "valid_targets_mean": 2481.9, |
| "valid_targets_min": 1022 |
| }, |
| { |
| "epoch": 7.0, |
| "loss_nan_ranks": 0, |
| "loss_rank_avg": 0.16949467360973358, |
| "step": 3850, |
| "total_flos": 488803397074944.0, |
| "train_loss": 0.26370995024582006, |
| "train_runtime": 7874.4592, |
| "train_samples_per_second": 7.812, |
| "train_steps_per_second": 0.489, |
| "valid_targets_mean": 2481.9, |
| "valid_targets_min": 1022 |
| } |
| ], |
| "logging_steps": 5, |
| "max_steps": 3850, |
| "num_input_tokens_seen": 0, |
| "num_train_epochs": 7, |
| "save_steps": 1500, |
| "stateful_callbacks": { |
| "TrainerControl": { |
| "args": { |
| "should_epoch_stop": false, |
| "should_evaluate": false, |
| "should_log": false, |
| "should_save": true, |
| "should_training_stop": true |
| }, |
| "attributes": {} |
| } |
| }, |
| "total_flos": 488803397074944.0, |
| "train_batch_size": 1, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
