| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 3.0, | |
| "eval_steps": 500, | |
| "global_step": 309, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.009708737864077669, | |
| "grad_norm": 6.576948642730713, | |
| "learning_rate": 3.2258064516129035e-07, | |
| "loss": 1.2227, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.019417475728155338, | |
| "grad_norm": 6.762261867523193, | |
| "learning_rate": 6.451612903225807e-07, | |
| "loss": 1.2675, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.02912621359223301, | |
| "grad_norm": 6.975041389465332, | |
| "learning_rate": 9.67741935483871e-07, | |
| "loss": 1.2517, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.038834951456310676, | |
| "grad_norm": 6.626192569732666, | |
| "learning_rate": 1.2903225806451614e-06, | |
| "loss": 1.2515, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.04854368932038835, | |
| "grad_norm": 6.378674030303955, | |
| "learning_rate": 1.6129032258064516e-06, | |
| "loss": 1.2498, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.05825242718446602, | |
| "grad_norm": 6.095025539398193, | |
| "learning_rate": 1.935483870967742e-06, | |
| "loss": 1.2283, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.06796116504854369, | |
| "grad_norm": 4.670423984527588, | |
| "learning_rate": 2.2580645161290324e-06, | |
| "loss": 1.1489, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.07766990291262135, | |
| "grad_norm": 4.401129245758057, | |
| "learning_rate": 2.580645161290323e-06, | |
| "loss": 1.1759, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.08737864077669903, | |
| "grad_norm": 2.814140796661377, | |
| "learning_rate": 2.903225806451613e-06, | |
| "loss": 1.128, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.0970873786407767, | |
| "grad_norm": 2.6264705657958984, | |
| "learning_rate": 3.225806451612903e-06, | |
| "loss": 1.1088, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.10679611650485436, | |
| "grad_norm": 2.619597911834717, | |
| "learning_rate": 3.548387096774194e-06, | |
| "loss": 1.1275, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.11650485436893204, | |
| "grad_norm": 4.305774688720703, | |
| "learning_rate": 3.870967741935484e-06, | |
| "loss": 1.1287, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.1262135922330097, | |
| "grad_norm": 4.335093975067139, | |
| "learning_rate": 4.193548387096774e-06, | |
| "loss": 1.1011, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.13592233009708737, | |
| "grad_norm": 4.46498966217041, | |
| "learning_rate": 4.516129032258065e-06, | |
| "loss": 1.1392, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.14563106796116504, | |
| "grad_norm": 3.7746875286102295, | |
| "learning_rate": 4.838709677419355e-06, | |
| "loss": 1.0678, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.1553398058252427, | |
| "grad_norm": 2.9575371742248535, | |
| "learning_rate": 5.161290322580646e-06, | |
| "loss": 1.0683, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.1650485436893204, | |
| "grad_norm": 2.7806265354156494, | |
| "learning_rate": 5.483870967741935e-06, | |
| "loss": 1.0367, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.17475728155339806, | |
| "grad_norm": 2.41747784614563, | |
| "learning_rate": 5.806451612903226e-06, | |
| "loss": 1.0401, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.18446601941747573, | |
| "grad_norm": 2.0261120796203613, | |
| "learning_rate": 6.129032258064517e-06, | |
| "loss": 1.0364, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.1941747572815534, | |
| "grad_norm": 1.6902999877929688, | |
| "learning_rate": 6.451612903225806e-06, | |
| "loss": 0.9807, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.20388349514563106, | |
| "grad_norm": 1.6579384803771973, | |
| "learning_rate": 6.774193548387097e-06, | |
| "loss": 1.0094, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.21359223300970873, | |
| "grad_norm": 1.3350183963775635, | |
| "learning_rate": 7.096774193548388e-06, | |
| "loss": 0.9756, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.22330097087378642, | |
| "grad_norm": 1.4252748489379883, | |
| "learning_rate": 7.4193548387096784e-06, | |
| "loss": 1.0005, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.23300970873786409, | |
| "grad_norm": 1.2831058502197266, | |
| "learning_rate": 7.741935483870968e-06, | |
| "loss": 0.9875, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.24271844660194175, | |
| "grad_norm": 1.1137187480926514, | |
| "learning_rate": 8.064516129032258e-06, | |
| "loss": 0.9444, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.2524271844660194, | |
| "grad_norm": 1.186092495918274, | |
| "learning_rate": 8.387096774193549e-06, | |
| "loss": 0.9467, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.2621359223300971, | |
| "grad_norm": 0.910400390625, | |
| "learning_rate": 8.70967741935484e-06, | |
| "loss": 0.9549, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.27184466019417475, | |
| "grad_norm": 1.0872230529785156, | |
| "learning_rate": 9.03225806451613e-06, | |
| "loss": 0.9433, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.2815533980582524, | |
| "grad_norm": 1.0862929821014404, | |
| "learning_rate": 9.35483870967742e-06, | |
| "loss": 0.9353, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.2912621359223301, | |
| "grad_norm": 1.1340218782424927, | |
| "learning_rate": 9.67741935483871e-06, | |
| "loss": 0.9254, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.30097087378640774, | |
| "grad_norm": 1.1494905948638916, | |
| "learning_rate": 1e-05, | |
| "loss": 0.9231, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.3106796116504854, | |
| "grad_norm": 0.9529262781143188, | |
| "learning_rate": 9.999680739242022e-06, | |
| "loss": 0.9367, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.32038834951456313, | |
| "grad_norm": 0.9129893183708191, | |
| "learning_rate": 9.99872299773906e-06, | |
| "loss": 0.9194, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.3300970873786408, | |
| "grad_norm": 1.1129016876220703, | |
| "learning_rate": 9.997126897798826e-06, | |
| "loss": 0.8978, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.33980582524271846, | |
| "grad_norm": 1.0057542324066162, | |
| "learning_rate": 9.994892643250147e-06, | |
| "loss": 0.9046, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.34951456310679613, | |
| "grad_norm": 0.8979021906852722, | |
| "learning_rate": 9.99202051941695e-06, | |
| "loss": 0.9191, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.3592233009708738, | |
| "grad_norm": 0.8347776532173157, | |
| "learning_rate": 9.9885108930818e-06, | |
| "loss": 0.8796, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.36893203883495146, | |
| "grad_norm": 0.8598542213439941, | |
| "learning_rate": 9.984364212439089e-06, | |
| "loss": 0.9132, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.3786407766990291, | |
| "grad_norm": 0.8244529962539673, | |
| "learning_rate": 9.979581007037776e-06, | |
| "loss": 0.8933, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.3883495145631068, | |
| "grad_norm": 0.733172595500946, | |
| "learning_rate": 9.974161887713775e-06, | |
| "loss": 0.9021, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.39805825242718446, | |
| "grad_norm": 0.9536871910095215, | |
| "learning_rate": 9.968107546511942e-06, | |
| "loss": 0.9121, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.4077669902912621, | |
| "grad_norm": 0.7781872153282166, | |
| "learning_rate": 9.961418756597703e-06, | |
| "loss": 0.9045, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.4174757281553398, | |
| "grad_norm": 0.8622099757194519, | |
| "learning_rate": 9.95409637215831e-06, | |
| "loss": 0.9019, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.42718446601941745, | |
| "grad_norm": 0.694644570350647, | |
| "learning_rate": 9.94614132829377e-06, | |
| "loss": 0.8502, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.4368932038834951, | |
| "grad_norm": 0.7023694515228271, | |
| "learning_rate": 9.937554640897414e-06, | |
| "loss": 0.9056, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.44660194174757284, | |
| "grad_norm": 0.7865616679191589, | |
| "learning_rate": 9.928337406526172e-06, | |
| "loss": 0.8951, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.4563106796116505, | |
| "grad_norm": 0.8198397159576416, | |
| "learning_rate": 9.918490802260538e-06, | |
| "loss": 0.919, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.46601941747572817, | |
| "grad_norm": 0.6332915425300598, | |
| "learning_rate": 9.90801608555425e-06, | |
| "loss": 0.9086, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.47572815533980584, | |
| "grad_norm": 0.665955662727356, | |
| "learning_rate": 9.896914594073703e-06, | |
| "loss": 0.8707, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.4854368932038835, | |
| "grad_norm": 0.8108275532722473, | |
| "learning_rate": 9.885187745527132e-06, | |
| "loss": 0.8903, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.49514563106796117, | |
| "grad_norm": 0.7815718054771423, | |
| "learning_rate": 9.87283703748356e-06, | |
| "loss": 0.8803, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.5048543689320388, | |
| "grad_norm": 0.7477515935897827, | |
| "learning_rate": 9.859864047181551e-06, | |
| "loss": 0.9034, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.5145631067961165, | |
| "grad_norm": 0.8046391606330872, | |
| "learning_rate": 9.846270431327793e-06, | |
| "loss": 0.8962, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.5242718446601942, | |
| "grad_norm": 0.6419571042060852, | |
| "learning_rate": 9.832057925885526e-06, | |
| "loss": 0.8834, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.5339805825242718, | |
| "grad_norm": 0.885932981967926, | |
| "learning_rate": 9.817228345852853e-06, | |
| "loss": 0.9023, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.5436893203883495, | |
| "grad_norm": 0.5639209747314453, | |
| "learning_rate": 9.801783585030959e-06, | |
| "loss": 0.9119, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.5533980582524272, | |
| "grad_norm": 0.8406121730804443, | |
| "learning_rate": 9.785725615782262e-06, | |
| "loss": 0.8605, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.5631067961165048, | |
| "grad_norm": 0.6397046446800232, | |
| "learning_rate": 9.769056488778538e-06, | |
| "loss": 0.87, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.5728155339805825, | |
| "grad_norm": 0.6496418714523315, | |
| "learning_rate": 9.751778332739033e-06, | |
| "loss": 0.8863, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.5825242718446602, | |
| "grad_norm": 0.8817011713981628, | |
| "learning_rate": 9.733893354158628e-06, | |
| "loss": 0.8757, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.5922330097087378, | |
| "grad_norm": 0.6433860063552856, | |
| "learning_rate": 9.715403837026046e-06, | |
| "loss": 0.8613, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.6019417475728155, | |
| "grad_norm": 0.9720786213874817, | |
| "learning_rate": 9.69631214253219e-06, | |
| "loss": 0.8831, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.6116504854368932, | |
| "grad_norm": 0.6686399579048157, | |
| "learning_rate": 9.676620708768608e-06, | |
| "loss": 0.8823, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.6213592233009708, | |
| "grad_norm": 0.6597081422805786, | |
| "learning_rate": 9.656332050416118e-06, | |
| "loss": 0.8351, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.6310679611650486, | |
| "grad_norm": 0.781461775302887, | |
| "learning_rate": 9.635448758423703e-06, | |
| "loss": 0.8792, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.6407766990291263, | |
| "grad_norm": 0.5933522582054138, | |
| "learning_rate": 9.613973499677613e-06, | |
| "loss": 0.912, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.6504854368932039, | |
| "grad_norm": 0.7272161841392517, | |
| "learning_rate": 9.591909016660806e-06, | |
| "loss": 0.8788, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.6601941747572816, | |
| "grad_norm": 0.6520808339118958, | |
| "learning_rate": 9.569258127102708e-06, | |
| "loss": 0.8965, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.6699029126213593, | |
| "grad_norm": 0.7806766033172607, | |
| "learning_rate": 9.546023723619387e-06, | |
| "loss": 0.8917, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.6796116504854369, | |
| "grad_norm": 0.7100508213043213, | |
| "learning_rate": 9.522208773344147e-06, | |
| "loss": 0.8872, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.6893203883495146, | |
| "grad_norm": 0.7099691033363342, | |
| "learning_rate": 9.497816317548625e-06, | |
| "loss": 0.8793, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.6990291262135923, | |
| "grad_norm": 0.8257980346679688, | |
| "learning_rate": 9.472849471254386e-06, | |
| "loss": 0.8539, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.7087378640776699, | |
| "grad_norm": 0.6728125214576721, | |
| "learning_rate": 9.447311422835141e-06, | |
| "loss": 0.8669, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.7184466019417476, | |
| "grad_norm": 0.7229968905448914, | |
| "learning_rate": 9.421205433609568e-06, | |
| "loss": 0.8991, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.7281553398058253, | |
| "grad_norm": 0.7105078101158142, | |
| "learning_rate": 9.39453483742483e-06, | |
| "loss": 0.8704, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.7378640776699029, | |
| "grad_norm": 0.7349528670310974, | |
| "learning_rate": 9.367303040230828e-06, | |
| "loss": 0.8866, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.7475728155339806, | |
| "grad_norm": 0.5864897966384888, | |
| "learning_rate": 9.33951351964525e-06, | |
| "loss": 0.8671, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.7572815533980582, | |
| "grad_norm": 0.6392523646354675, | |
| "learning_rate": 9.311169824509454e-06, | |
| "loss": 0.8435, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.7669902912621359, | |
| "grad_norm": 0.5775032639503479, | |
| "learning_rate": 9.28227557443528e-06, | |
| "loss": 0.8518, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.7766990291262136, | |
| "grad_norm": 0.5982815623283386, | |
| "learning_rate": 9.252834459342801e-06, | |
| "loss": 0.9034, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.7864077669902912, | |
| "grad_norm": 0.6583596467971802, | |
| "learning_rate": 9.222850238989104e-06, | |
| "loss": 0.8638, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.7961165048543689, | |
| "grad_norm": 0.621867299079895, | |
| "learning_rate": 9.192326742488153e-06, | |
| "loss": 0.8658, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.8058252427184466, | |
| "grad_norm": 0.6219890117645264, | |
| "learning_rate": 9.161267867821802e-06, | |
| "loss": 0.8769, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.8155339805825242, | |
| "grad_norm": 0.6069071292877197, | |
| "learning_rate": 9.129677581342e-06, | |
| "loss": 0.8677, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.8252427184466019, | |
| "grad_norm": 0.660553514957428, | |
| "learning_rate": 9.097559917264268e-06, | |
| "loss": 0.8465, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.8349514563106796, | |
| "grad_norm": 0.5944454669952393, | |
| "learning_rate": 9.064918977152517e-06, | |
| "loss": 0.8741, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.8446601941747572, | |
| "grad_norm": 0.6104128360748291, | |
| "learning_rate": 9.031758929395259e-06, | |
| "loss": 0.8248, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.8543689320388349, | |
| "grad_norm": 0.5879605412483215, | |
| "learning_rate": 8.998084008673284e-06, | |
| "loss": 0.8398, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.8640776699029126, | |
| "grad_norm": 0.6651065349578857, | |
| "learning_rate": 8.963898515418885e-06, | |
| "loss": 0.8687, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.8737864077669902, | |
| "grad_norm": 0.6086816191673279, | |
| "learning_rate": 8.929206815266653e-06, | |
| "loss": 0.8639, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.883495145631068, | |
| "grad_norm": 0.6449173092842102, | |
| "learning_rate": 8.89401333849598e-06, | |
| "loss": 0.8483, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.8932038834951457, | |
| "grad_norm": 0.6413861513137817, | |
| "learning_rate": 8.8583225794653e-06, | |
| "loss": 0.9039, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.9029126213592233, | |
| "grad_norm": 0.5411512851715088, | |
| "learning_rate": 8.82213909603812e-06, | |
| "loss": 0.8814, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.912621359223301, | |
| "grad_norm": 0.6556549072265625, | |
| "learning_rate": 8.78546750900098e-06, | |
| "loss": 0.8944, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.9223300970873787, | |
| "grad_norm": 0.6550806760787964, | |
| "learning_rate": 8.748312501473351e-06, | |
| "loss": 0.8819, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.9320388349514563, | |
| "grad_norm": 0.6445295810699463, | |
| "learning_rate": 8.710678818309576e-06, | |
| "loss": 0.8582, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.941747572815534, | |
| "grad_norm": 0.7256100177764893, | |
| "learning_rate": 8.672571265492944e-06, | |
| "loss": 0.8361, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.9514563106796117, | |
| "grad_norm": 0.599454402923584, | |
| "learning_rate": 8.63399470952193e-06, | |
| "loss": 0.8679, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.9611650485436893, | |
| "grad_norm": 0.7651983499526978, | |
| "learning_rate": 8.594954076788736e-06, | |
| "loss": 0.8182, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.970873786407767, | |
| "grad_norm": 0.7521474361419678, | |
| "learning_rate": 8.555454352950161e-06, | |
| "loss": 0.8622, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.9805825242718447, | |
| "grad_norm": 0.6626675724983215, | |
| "learning_rate": 8.515500582290914e-06, | |
| "loss": 0.845, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.9902912621359223, | |
| "grad_norm": 0.7361329793930054, | |
| "learning_rate": 8.475097867079437e-06, | |
| "loss": 0.8672, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "grad_norm": 0.5690233707427979, | |
| "learning_rate": 8.434251366916323e-06, | |
| "loss": 0.8989, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 1.0097087378640777, | |
| "grad_norm": 0.778568685054779, | |
| "learning_rate": 8.392966298075413e-06, | |
| "loss": 0.8642, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 1.0194174757281553, | |
| "grad_norm": 0.5900771021842957, | |
| "learning_rate": 8.351247932837655e-06, | |
| "loss": 0.8001, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 1.029126213592233, | |
| "grad_norm": 0.628034770488739, | |
| "learning_rate": 8.309101598817812e-06, | |
| "loss": 0.8101, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 1.0388349514563107, | |
| "grad_norm": 0.860468864440918, | |
| "learning_rate": 8.266532678284103e-06, | |
| "loss": 0.7936, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 1.0485436893203883, | |
| "grad_norm": 0.7311742901802063, | |
| "learning_rate": 8.223546607470863e-06, | |
| "loss": 0.8278, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 1.058252427184466, | |
| "grad_norm": 0.6832053661346436, | |
| "learning_rate": 8.18014887588431e-06, | |
| "loss": 0.8272, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 1.0679611650485437, | |
| "grad_norm": 0.8355731964111328, | |
| "learning_rate": 8.13634502560152e-06, | |
| "loss": 0.8485, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 1.0776699029126213, | |
| "grad_norm": 0.6029130816459656, | |
| "learning_rate": 8.092140650562665e-06, | |
| "loss": 0.7722, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 1.087378640776699, | |
| "grad_norm": 0.8511623740196228, | |
| "learning_rate": 8.047541395856661e-06, | |
| "loss": 0.8361, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 1.0970873786407767, | |
| "grad_norm": 0.6003599166870117, | |
| "learning_rate": 8.002552957000254e-06, | |
| "loss": 0.8157, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 1.1067961165048543, | |
| "grad_norm": 0.8679140210151672, | |
| "learning_rate": 7.957181079210676e-06, | |
| "loss": 0.819, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 1.116504854368932, | |
| "grad_norm": 0.65385901927948, | |
| "learning_rate": 7.911431556671967e-06, | |
| "loss": 0.8191, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 1.1262135922330097, | |
| "grad_norm": 0.5810941457748413, | |
| "learning_rate": 7.865310231795026e-06, | |
| "loss": 0.8084, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 1.1359223300970873, | |
| "grad_norm": 0.7386210560798645, | |
| "learning_rate": 7.818822994471504e-06, | |
| "loss": 0.8088, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 1.145631067961165, | |
| "grad_norm": 0.6052452325820923, | |
| "learning_rate": 7.771975781321655e-06, | |
| "loss": 0.8236, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 1.1553398058252426, | |
| "grad_norm": 0.6413220167160034, | |
| "learning_rate": 7.72477457493619e-06, | |
| "loss": 0.8427, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 1.1650485436893203, | |
| "grad_norm": 0.6168550848960876, | |
| "learning_rate": 7.677225403112277e-06, | |
| "loss": 0.8204, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 1.174757281553398, | |
| "grad_norm": 0.6849708557128906, | |
| "learning_rate": 7.629334338083774e-06, | |
| "loss": 0.7887, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 1.1844660194174756, | |
| "grad_norm": 0.7301358580589294, | |
| "learning_rate": 7.58110749574577e-06, | |
| "loss": 0.8137, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 1.1941747572815533, | |
| "grad_norm": 0.6189250349998474, | |
| "learning_rate": 7.532551034873558e-06, | |
| "loss": 0.7985, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 1.203883495145631, | |
| "grad_norm": 0.70215904712677, | |
| "learning_rate": 7.483671156336142e-06, | |
| "loss": 0.8108, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 1.2135922330097086, | |
| "grad_norm": 0.642704963684082, | |
| "learning_rate": 7.43447410230435e-06, | |
| "loss": 0.8161, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 1.2233009708737863, | |
| "grad_norm": 0.6741234660148621, | |
| "learning_rate": 7.384966155453686e-06, | |
| "loss": 0.7965, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 1.233009708737864, | |
| "grad_norm": 0.577409029006958, | |
| "learning_rate": 7.335153638162005e-06, | |
| "loss": 0.7989, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 1.2427184466019416, | |
| "grad_norm": 0.5523732900619507, | |
| "learning_rate": 7.285042911702116e-06, | |
| "loss": 0.7812, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 1.2524271844660193, | |
| "grad_norm": 0.6712595224380493, | |
| "learning_rate": 7.234640375429427e-06, | |
| "loss": 0.8034, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 1.262135922330097, | |
| "grad_norm": 0.6166908144950867, | |
| "learning_rate": 7.183952465964711e-06, | |
| "loss": 0.8004, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 1.2718446601941746, | |
| "grad_norm": 0.7181272506713867, | |
| "learning_rate": 7.132985656372126e-06, | |
| "loss": 0.8122, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 1.2815533980582523, | |
| "grad_norm": 0.5397993922233582, | |
| "learning_rate": 7.0817464553325764e-06, | |
| "loss": 0.8256, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 1.29126213592233, | |
| "grad_norm": 0.599078357219696, | |
| "learning_rate": 7.030241406312528e-06, | |
| "loss": 0.8012, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 1.3009708737864076, | |
| "grad_norm": 0.6068196296691895, | |
| "learning_rate": 6.978477086728375e-06, | |
| "loss": 0.7475, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 1.3106796116504853, | |
| "grad_norm": 0.6208373308181763, | |
| "learning_rate": 6.926460107106483e-06, | |
| "loss": 0.7631, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 1.3203883495145632, | |
| "grad_norm": 0.62087482213974, | |
| "learning_rate": 6.874197110238986e-06, | |
| "loss": 0.7626, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 1.3300970873786409, | |
| "grad_norm": 0.6587031483650208, | |
| "learning_rate": 6.8216947703354815e-06, | |
| "loss": 0.809, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 1.3398058252427185, | |
| "grad_norm": 0.7958360314369202, | |
| "learning_rate": 6.7689597921707065e-06, | |
| "loss": 0.8311, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 1.3495145631067962, | |
| "grad_norm": 0.6715186834335327, | |
| "learning_rate": 6.715998910228296e-06, | |
| "loss": 0.7949, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 1.3592233009708738, | |
| "grad_norm": 0.6351740956306458, | |
| "learning_rate": 6.6628188878407806e-06, | |
| "loss": 0.794, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 1.3689320388349515, | |
| "grad_norm": 0.8707325458526611, | |
| "learning_rate": 6.609426516325859e-06, | |
| "loss": 0.8329, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 1.3786407766990292, | |
| "grad_norm": 0.7321479320526123, | |
| "learning_rate": 6.555828614119132e-06, | |
| "loss": 0.8136, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 1.3883495145631068, | |
| "grad_norm": 0.5844182968139648, | |
| "learning_rate": 6.502032025903356e-06, | |
| "loss": 0.7576, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 1.3980582524271845, | |
| "grad_norm": 0.6121401786804199, | |
| "learning_rate": 6.4480436217343366e-06, | |
| "loss": 0.8205, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 1.4077669902912622, | |
| "grad_norm": 0.6373964548110962, | |
| "learning_rate": 6.393870296163616e-06, | |
| "loss": 0.8104, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 1.4174757281553398, | |
| "grad_norm": 0.5534393787384033, | |
| "learning_rate": 6.339518967357985e-06, | |
| "loss": 0.7822, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 1.4271844660194175, | |
| "grad_norm": 0.593232274055481, | |
| "learning_rate": 6.284996576216014e-06, | |
| "loss": 0.8227, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 1.4368932038834952, | |
| "grad_norm": 0.5890830755233765, | |
| "learning_rate": 6.230310085481677e-06, | |
| "loss": 0.8005, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 1.4466019417475728, | |
| "grad_norm": 0.5876540541648865, | |
| "learning_rate": 6.175466478855161e-06, | |
| "loss": 0.7936, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 1.4563106796116505, | |
| "grad_norm": 0.5593201518058777, | |
| "learning_rate": 6.1204727601010396e-06, | |
| "loss": 0.7473, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 1.4660194174757282, | |
| "grad_norm": 0.5841484069824219, | |
| "learning_rate": 6.065335952153846e-06, | |
| "loss": 0.7974, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 1.4757281553398058, | |
| "grad_norm": 0.49436667561531067, | |
| "learning_rate": 6.010063096221215e-06, | |
| "loss": 0.7921, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 1.4854368932038835, | |
| "grad_norm": 0.6187512874603271, | |
| "learning_rate": 5.954661250884704e-06, | |
| "loss": 0.7822, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 1.4951456310679612, | |
| "grad_norm": 0.6759915351867676, | |
| "learning_rate": 5.899137491198364e-06, | |
| "loss": 0.8273, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 1.5048543689320388, | |
| "grad_norm": 0.5890839099884033, | |
| "learning_rate": 5.843498907785236e-06, | |
| "loss": 0.7979, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 1.5145631067961165, | |
| "grad_norm": 0.5246779322624207, | |
| "learning_rate": 5.78775260593185e-06, | |
| "loss": 0.8198, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 1.5242718446601942, | |
| "grad_norm": 0.5603223443031311, | |
| "learning_rate": 5.731905704680834e-06, | |
| "loss": 0.8023, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 1.5339805825242718, | |
| "grad_norm": 0.5730907917022705, | |
| "learning_rate": 5.6759653359218e-06, | |
| "loss": 0.8231, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 1.5436893203883495, | |
| "grad_norm": 0.5317082405090332, | |
| "learning_rate": 5.6199386434805615e-06, | |
| "loss": 0.7984, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 1.5533980582524272, | |
| "grad_norm": 0.5330399870872498, | |
| "learning_rate": 5.563832782206835e-06, | |
| "loss": 0.7975, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 1.5631067961165048, | |
| "grad_norm": 0.6568978428840637, | |
| "learning_rate": 5.507654917060541e-06, | |
| "loss": 0.824, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 1.5728155339805825, | |
| "grad_norm": 0.6758802533149719, | |
| "learning_rate": 5.451412222196801e-06, | |
| "loss": 0.7844, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 1.5825242718446602, | |
| "grad_norm": 0.5091277360916138, | |
| "learning_rate": 5.395111880049775e-06, | |
| "loss": 0.7946, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 1.5922330097087378, | |
| "grad_norm": 0.5247997045516968, | |
| "learning_rate": 5.338761080415425e-06, | |
| "loss": 0.8035, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 1.6019417475728155, | |
| "grad_norm": 0.5019735097885132, | |
| "learning_rate": 5.28236701953335e-06, | |
| "loss": 0.7802, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 1.6116504854368932, | |
| "grad_norm": 0.5382864475250244, | |
| "learning_rate": 5.225936899167803e-06, | |
| "loss": 0.8301, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 1.6213592233009708, | |
| "grad_norm": 0.503472089767456, | |
| "learning_rate": 5.169477925687981e-06, | |
| "loss": 0.7875, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 1.6310679611650487, | |
| "grad_norm": 0.5378626585006714, | |
| "learning_rate": 5.112997309147753e-06, | |
| "loss": 0.8003, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 1.6407766990291264, | |
| "grad_norm": 0.4998520314693451, | |
| "learning_rate": 5.0565022623649e-06, | |
| "loss": 0.8094, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 1.650485436893204, | |
| "grad_norm": 0.5108544826507568, | |
| "learning_rate": 5e-06, | |
| "loss": 0.7756, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 1.6601941747572817, | |
| "grad_norm": 0.4933609962463379, | |
| "learning_rate": 4.943497737635103e-06, | |
| "loss": 0.8081, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 1.6699029126213594, | |
| "grad_norm": 0.5090092420578003, | |
| "learning_rate": 4.887002690852249e-06, | |
| "loss": 0.8025, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 1.679611650485437, | |
| "grad_norm": 0.5769246816635132, | |
| "learning_rate": 4.830522074312019e-06, | |
| "loss": 0.8068, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 1.6893203883495147, | |
| "grad_norm": 0.5017176866531372, | |
| "learning_rate": 4.774063100832199e-06, | |
| "loss": 0.8078, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 1.6990291262135924, | |
| "grad_norm": 0.5930182337760925, | |
| "learning_rate": 4.717632980466652e-06, | |
| "loss": 0.8529, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 1.70873786407767, | |
| "grad_norm": 0.5795880556106567, | |
| "learning_rate": 4.661238919584578e-06, | |
| "loss": 0.7914, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 1.7184466019417477, | |
| "grad_norm": 0.4433108866214752, | |
| "learning_rate": 4.6048881199502265e-06, | |
| "loss": 0.8377, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 1.7281553398058254, | |
| "grad_norm": 0.5359524488449097, | |
| "learning_rate": 4.548587777803198e-06, | |
| "loss": 0.841, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 1.737864077669903, | |
| "grad_norm": 0.5342350006103516, | |
| "learning_rate": 4.49234508293946e-06, | |
| "loss": 0.8118, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 1.7475728155339807, | |
| "grad_norm": 0.4875918924808502, | |
| "learning_rate": 4.436167217793167e-06, | |
| "loss": 0.8091, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 1.7572815533980584, | |
| "grad_norm": 0.44451451301574707, | |
| "learning_rate": 4.38006135651944e-06, | |
| "loss": 0.8012, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 1.766990291262136, | |
| "grad_norm": 0.6147545576095581, | |
| "learning_rate": 4.3240346640782014e-06, | |
| "loss": 0.7779, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 1.7766990291262137, | |
| "grad_norm": 0.4527565538883209, | |
| "learning_rate": 4.268094295319167e-06, | |
| "loss": 0.8184, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 1.7864077669902914, | |
| "grad_norm": 0.48011067509651184, | |
| "learning_rate": 4.212247394068151e-06, | |
| "loss": 0.8109, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 1.796116504854369, | |
| "grad_norm": 0.5363189578056335, | |
| "learning_rate": 4.1565010922147644e-06, | |
| "loss": 0.8068, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 1.8058252427184467, | |
| "grad_norm": 0.5342061519622803, | |
| "learning_rate": 4.100862508801639e-06, | |
| "loss": 0.8158, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 1.8155339805825244, | |
| "grad_norm": 0.4906405508518219, | |
| "learning_rate": 4.045338749115299e-06, | |
| "loss": 0.8012, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 1.825242718446602, | |
| "grad_norm": 0.4942418038845062, | |
| "learning_rate": 3.989936903778785e-06, | |
| "loss": 0.8276, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 1.8349514563106797, | |
| "grad_norm": 0.47334739565849304, | |
| "learning_rate": 3.934664047846157e-06, | |
| "loss": 0.7817, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 1.8446601941747574, | |
| "grad_norm": 0.4942063093185425, | |
| "learning_rate": 3.879527239898962e-06, | |
| "loss": 0.8325, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 1.854368932038835, | |
| "grad_norm": 0.5283282995223999, | |
| "learning_rate": 3.8245335211448404e-06, | |
| "loss": 0.7786, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 1.8640776699029127, | |
| "grad_norm": 0.5127359628677368, | |
| "learning_rate": 3.769689914518326e-06, | |
| "loss": 0.8142, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 1.8737864077669903, | |
| "grad_norm": 0.49381014704704285, | |
| "learning_rate": 3.715003423783986e-06, | |
| "loss": 0.7969, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 1.883495145631068, | |
| "grad_norm": 0.46980249881744385, | |
| "learning_rate": 3.660481032642016e-06, | |
| "loss": 0.7949, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 1.8932038834951457, | |
| "grad_norm": 0.5440580248832703, | |
| "learning_rate": 3.6061297038363853e-06, | |
| "loss": 0.7786, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 1.9029126213592233, | |
| "grad_norm": 0.5657265782356262, | |
| "learning_rate": 3.5519563782656642e-06, | |
| "loss": 0.8131, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 1.912621359223301, | |
| "grad_norm": 0.560371458530426, | |
| "learning_rate": 3.497967974096647e-06, | |
| "loss": 0.7723, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 1.9223300970873787, | |
| "grad_norm": 0.4916374087333679, | |
| "learning_rate": 3.4441713858808684e-06, | |
| "loss": 0.793, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 1.9320388349514563, | |
| "grad_norm": 0.5217000246047974, | |
| "learning_rate": 3.3905734836741415e-06, | |
| "loss": 0.8259, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 1.941747572815534, | |
| "grad_norm": 0.6433777213096619, | |
| "learning_rate": 3.3371811121592203e-06, | |
| "loss": 0.7691, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.9514563106796117, | |
| "grad_norm": 0.5687490701675415, | |
| "learning_rate": 3.2840010897717045e-06, | |
| "loss": 0.7592, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 1.9611650485436893, | |
| "grad_norm": 0.5231119990348816, | |
| "learning_rate": 3.2310402078292956e-06, | |
| "loss": 0.806, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 1.970873786407767, | |
| "grad_norm": 0.5205345153808594, | |
| "learning_rate": 3.178305229664519e-06, | |
| "loss": 0.8302, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 1.9805825242718447, | |
| "grad_norm": 0.5407435894012451, | |
| "learning_rate": 3.125802889761016e-06, | |
| "loss": 0.7726, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 1.9902912621359223, | |
| "grad_norm": 0.628868818283081, | |
| "learning_rate": 3.073539892893519e-06, | |
| "loss": 0.8098, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "grad_norm": 0.5963810682296753, | |
| "learning_rate": 3.021522913271627e-06, | |
| "loss": 0.7834, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 2.0097087378640777, | |
| "grad_norm": 0.5925841927528381, | |
| "learning_rate": 2.969758593687475e-06, | |
| "loss": 0.7591, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 2.0194174757281553, | |
| "grad_norm": 0.47755351662635803, | |
| "learning_rate": 2.9182535446674244e-06, | |
| "loss": 0.7803, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 2.029126213592233, | |
| "grad_norm": 0.5342966914176941, | |
| "learning_rate": 2.8670143436278757e-06, | |
| "loss": 0.7553, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 2.0388349514563107, | |
| "grad_norm": 0.5112743377685547, | |
| "learning_rate": 2.8160475340352913e-06, | |
| "loss": 0.7849, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 2.0485436893203883, | |
| "grad_norm": 0.5908800363540649, | |
| "learning_rate": 2.765359624570574e-06, | |
| "loss": 0.7618, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 2.058252427184466, | |
| "grad_norm": 0.4771662652492523, | |
| "learning_rate": 2.714957088297886e-06, | |
| "loss": 0.6928, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 2.0679611650485437, | |
| "grad_norm": 0.5731369853019714, | |
| "learning_rate": 2.664846361837997e-06, | |
| "loss": 0.7647, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 2.0776699029126213, | |
| "grad_norm": 0.4691663384437561, | |
| "learning_rate": 2.6150338445463146e-06, | |
| "loss": 0.7537, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 2.087378640776699, | |
| "grad_norm": 0.5235918760299683, | |
| "learning_rate": 2.565525897695651e-06, | |
| "loss": 0.7775, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 2.0970873786407767, | |
| "grad_norm": 0.49530091881752014, | |
| "learning_rate": 2.51632884366386e-06, | |
| "loss": 0.7784, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 2.1067961165048543, | |
| "grad_norm": 0.5703685283660889, | |
| "learning_rate": 2.4674489651264433e-06, | |
| "loss": 0.7207, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 2.116504854368932, | |
| "grad_norm": 0.45397961139678955, | |
| "learning_rate": 2.418892504254231e-06, | |
| "loss": 0.7493, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 2.1262135922330097, | |
| "grad_norm": 0.512279212474823, | |
| "learning_rate": 2.3706656619162278e-06, | |
| "loss": 0.7775, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 2.1359223300970873, | |
| "grad_norm": 0.5071080327033997, | |
| "learning_rate": 2.322774596887726e-06, | |
| "loss": 0.7594, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 2.145631067961165, | |
| "grad_norm": 0.5117892026901245, | |
| "learning_rate": 2.275225425063813e-06, | |
| "loss": 0.7265, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 2.1553398058252426, | |
| "grad_norm": 0.4993806779384613, | |
| "learning_rate": 2.2280242186783473e-06, | |
| "loss": 0.7149, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 2.1650485436893203, | |
| "grad_norm": 0.4563067853450775, | |
| "learning_rate": 2.1811770055284968e-06, | |
| "loss": 0.7902, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 2.174757281553398, | |
| "grad_norm": 0.47798827290534973, | |
| "learning_rate": 2.134689768204975e-06, | |
| "loss": 0.7583, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 2.1844660194174756, | |
| "grad_norm": 0.46698057651519775, | |
| "learning_rate": 2.0885684433280336e-06, | |
| "loss": 0.7717, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 2.1941747572815533, | |
| "grad_norm": 0.46322739124298096, | |
| "learning_rate": 2.042818920789326e-06, | |
| "loss": 0.7622, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 2.203883495145631, | |
| "grad_norm": 0.4613381028175354, | |
| "learning_rate": 1.9974470429997482e-06, | |
| "loss": 0.7508, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 2.2135922330097086, | |
| "grad_norm": 0.4670693278312683, | |
| "learning_rate": 1.9524586041433393e-06, | |
| "loss": 0.7132, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 2.2233009708737863, | |
| "grad_norm": 0.45155465602874756, | |
| "learning_rate": 1.907859349437336e-06, | |
| "loss": 0.7544, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 2.233009708737864, | |
| "grad_norm": 0.4583199620246887, | |
| "learning_rate": 1.8636549743984815e-06, | |
| "loss": 0.7485, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 2.2427184466019416, | |
| "grad_norm": 0.4572857618331909, | |
| "learning_rate": 1.8198511241156902e-06, | |
| "loss": 0.7552, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 2.2524271844660193, | |
| "grad_norm": 0.47898781299591064, | |
| "learning_rate": 1.776453392529139e-06, | |
| "loss": 0.7683, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 2.262135922330097, | |
| "grad_norm": 0.47700902819633484, | |
| "learning_rate": 1.7334673217158976e-06, | |
| "loss": 0.7444, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 2.2718446601941746, | |
| "grad_norm": 0.45447424054145813, | |
| "learning_rate": 1.6908984011821883e-06, | |
| "loss": 0.7495, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 2.2815533980582523, | |
| "grad_norm": 0.4574134647846222, | |
| "learning_rate": 1.6487520671623469e-06, | |
| "loss": 0.7545, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 2.29126213592233, | |
| "grad_norm": 0.4218170642852783, | |
| "learning_rate": 1.6070337019245896e-06, | |
| "loss": 0.7521, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 2.3009708737864076, | |
| "grad_norm": 0.41513901948928833, | |
| "learning_rate": 1.5657486330836786e-06, | |
| "loss": 0.759, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 2.3106796116504853, | |
| "grad_norm": 0.4804514944553375, | |
| "learning_rate": 1.5249021329205638e-06, | |
| "loss": 0.757, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 2.320388349514563, | |
| "grad_norm": 0.48271673917770386, | |
| "learning_rate": 1.4844994177090871e-06, | |
| "loss": 0.7533, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 2.3300970873786406, | |
| "grad_norm": 0.4723464846611023, | |
| "learning_rate": 1.4445456470498392e-06, | |
| "loss": 0.7859, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 2.3398058252427183, | |
| "grad_norm": 0.4236687123775482, | |
| "learning_rate": 1.4050459232112652e-06, | |
| "loss": 0.7873, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 2.349514563106796, | |
| "grad_norm": 0.4499347507953644, | |
| "learning_rate": 1.3660052904780707e-06, | |
| "loss": 0.7664, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 2.3592233009708736, | |
| "grad_norm": 0.4442787170410156, | |
| "learning_rate": 1.3274287345070564e-06, | |
| "loss": 0.7408, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 2.3689320388349513, | |
| "grad_norm": 0.45914286375045776, | |
| "learning_rate": 1.2893211816904243e-06, | |
| "loss": 0.7218, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 2.378640776699029, | |
| "grad_norm": 0.42868661880493164, | |
| "learning_rate": 1.2516874985266508e-06, | |
| "loss": 0.7466, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 2.3883495145631066, | |
| "grad_norm": 0.46668022871017456, | |
| "learning_rate": 1.2145324909990202e-06, | |
| "loss": 0.7116, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 2.3980582524271843, | |
| "grad_norm": 0.47152259945869446, | |
| "learning_rate": 1.1778609039618804e-06, | |
| "loss": 0.7446, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 2.407766990291262, | |
| "grad_norm": 0.4311952292919159, | |
| "learning_rate": 1.1416774205347015e-06, | |
| "loss": 0.7546, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 2.4174757281553396, | |
| "grad_norm": 0.47215738892555237, | |
| "learning_rate": 1.1059866615040205e-06, | |
| "loss": 0.7843, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 2.4271844660194173, | |
| "grad_norm": 0.42326658964157104, | |
| "learning_rate": 1.0707931847333487e-06, | |
| "loss": 0.7519, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 2.436893203883495, | |
| "grad_norm": 0.43430864810943604, | |
| "learning_rate": 1.036101484581117e-06, | |
| "loss": 0.7283, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 2.4466019417475726, | |
| "grad_norm": 0.45083683729171753, | |
| "learning_rate": 1.0019159913267156e-06, | |
| "loss": 0.7529, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 2.4563106796116507, | |
| "grad_norm": 0.49279189109802246, | |
| "learning_rate": 9.68241070604743e-07, | |
| "loss": 0.763, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 2.466019417475728, | |
| "grad_norm": 0.44121843576431274, | |
| "learning_rate": 9.350810228474855e-07, | |
| "loss": 0.7093, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 2.475728155339806, | |
| "grad_norm": 0.46301037073135376, | |
| "learning_rate": 9.024400827357344e-07, | |
| "loss": 0.7699, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 2.4854368932038833, | |
| "grad_norm": 0.4788742661476135, | |
| "learning_rate": 8.703224186580012e-07, | |
| "loss": 0.7811, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 2.4951456310679614, | |
| "grad_norm": 0.4840719401836395, | |
| "learning_rate": 8.387321321781977e-07, | |
| "loss": 0.7633, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 2.5048543689320386, | |
| "grad_norm": 0.4576054811477661, | |
| "learning_rate": 8.07673257511849e-07, | |
| "loss": 0.7755, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 2.5145631067961167, | |
| "grad_norm": 0.43629974126815796, | |
| "learning_rate": 7.771497610108981e-07, | |
| "loss": 0.7574, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 2.524271844660194, | |
| "grad_norm": 0.4582881033420563, | |
| "learning_rate": 7.471655406572003e-07, | |
| "loss": 0.7703, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 2.533980582524272, | |
| "grad_norm": 0.4240984618663788, | |
| "learning_rate": 7.177244255647209e-07, | |
| "loss": 0.7567, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 2.5436893203883493, | |
| "grad_norm": 0.4236100912094116, | |
| "learning_rate": 6.888301754905469e-07, | |
| "loss": 0.7912, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 2.5533980582524274, | |
| "grad_norm": 0.40091320872306824, | |
| "learning_rate": 6.604864803547511e-07, | |
| "loss": 0.7474, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 2.5631067961165046, | |
| "grad_norm": 0.44925597310066223, | |
| "learning_rate": 6.326969597691724e-07, | |
| "loss": 0.7381, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 2.5728155339805827, | |
| "grad_norm": 0.4370911121368408, | |
| "learning_rate": 6.054651625751717e-07, | |
| "loss": 0.7587, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 2.58252427184466, | |
| "grad_norm": 0.43402257561683655, | |
| "learning_rate": 5.787945663904332e-07, | |
| "loss": 0.7715, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 2.592233009708738, | |
| "grad_norm": 0.4456167221069336, | |
| "learning_rate": 5.526885771648599e-07, | |
| "loss": 0.7469, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 2.6019417475728153, | |
| "grad_norm": 0.44401729106903076, | |
| "learning_rate": 5.271505287456153e-07, | |
| "loss": 0.7223, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 2.6116504854368934, | |
| "grad_norm": 0.4610472023487091, | |
| "learning_rate": 5.021836824513759e-07, | |
| "loss": 0.7802, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 2.6213592233009706, | |
| "grad_norm": 0.4633938670158386, | |
| "learning_rate": 4.777912266558532e-07, | |
| "loss": 0.7417, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 2.6310679611650487, | |
| "grad_norm": 0.4150261878967285, | |
| "learning_rate": 4.5397627638061604e-07, | |
| "loss": 0.7705, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 2.6407766990291264, | |
| "grad_norm": 0.45528119802474976, | |
| "learning_rate": 4.307418728972934e-07, | |
| "loss": 0.7391, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 2.650485436893204, | |
| "grad_norm": 0.4313214123249054, | |
| "learning_rate": 4.080909833391944e-07, | |
| "loss": 0.7624, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 2.6601941747572817, | |
| "grad_norm": 0.403175950050354, | |
| "learning_rate": 3.8602650032238675e-07, | |
| "loss": 0.7797, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 2.6699029126213594, | |
| "grad_norm": 0.4482267498970032, | |
| "learning_rate": 3.6455124157629805e-07, | |
| "loss": 0.7373, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 2.679611650485437, | |
| "grad_norm": 0.42840349674224854, | |
| "learning_rate": 3.436679495838835e-07, | |
| "loss": 0.7349, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 2.6893203883495147, | |
| "grad_norm": 0.41052892804145813, | |
| "learning_rate": 3.233792912313943e-07, | |
| "loss": 0.754, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 2.6990291262135924, | |
| "grad_norm": 0.4851820170879364, | |
| "learning_rate": 3.0368785746780925e-07, | |
| "loss": 0.7887, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 2.70873786407767, | |
| "grad_norm": 0.43496590852737427, | |
| "learning_rate": 2.8459616297395464e-07, | |
| "loss": 0.7558, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 2.7184466019417477, | |
| "grad_norm": 0.44541043043136597, | |
| "learning_rate": 2.6610664584137413e-07, | |
| "loss": 0.7088, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 2.7281553398058254, | |
| "grad_norm": 0.45385488867759705, | |
| "learning_rate": 2.482216672609677e-07, | |
| "loss": 0.7279, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 2.737864077669903, | |
| "grad_norm": 0.4380588233470917, | |
| "learning_rate": 2.3094351122146307e-07, | |
| "loss": 0.7354, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 2.7475728155339807, | |
| "grad_norm": 0.4383409321308136, | |
| "learning_rate": 2.142743842177386e-07, | |
| "loss": 0.7427, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 2.7572815533980584, | |
| "grad_norm": 0.4154166579246521, | |
| "learning_rate": 1.98216414969043e-07, | |
| "loss": 0.7533, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 2.766990291262136, | |
| "grad_norm": 0.40241143107414246, | |
| "learning_rate": 1.8277165414714858e-07, | |
| "loss": 0.7471, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 2.7766990291262137, | |
| "grad_norm": 0.4261252284049988, | |
| "learning_rate": 1.6794207411447548e-07, | |
| "loss": 0.7628, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 2.7864077669902914, | |
| "grad_norm": 0.39932355284690857, | |
| "learning_rate": 1.5372956867220678e-07, | |
| "loss": 0.7747, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 2.796116504854369, | |
| "grad_norm": 0.40878960490226746, | |
| "learning_rate": 1.4013595281844872e-07, | |
| "loss": 0.7583, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 2.8058252427184467, | |
| "grad_norm": 0.42195621132850647, | |
| "learning_rate": 1.2716296251644e-07, | |
| "loss": 0.7675, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 2.8155339805825244, | |
| "grad_norm": 0.41512733697891235, | |
| "learning_rate": 1.1481225447286803e-07, | |
| "loss": 0.7228, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 2.825242718446602, | |
| "grad_norm": 0.42841315269470215, | |
| "learning_rate": 1.0308540592629756e-07, | |
| "loss": 0.7937, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 2.8349514563106797, | |
| "grad_norm": 0.445648193359375, | |
| "learning_rate": 9.198391444575072e-08, | |
| "loss": 0.7037, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 2.8446601941747574, | |
| "grad_norm": 0.4242344796657562, | |
| "learning_rate": 8.150919773946165e-08, | |
| "loss": 0.7795, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 2.854368932038835, | |
| "grad_norm": 0.4234623610973358, | |
| "learning_rate": 7.166259347382854e-08, | |
| "loss": 0.7292, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 2.8640776699029127, | |
| "grad_norm": 0.3901122808456421, | |
| "learning_rate": 6.244535910258697e-08, | |
| "loss": 0.776, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 2.8737864077669903, | |
| "grad_norm": 0.41637924313545227, | |
| "learning_rate": 5.3858671706230605e-08, | |
| "loss": 0.7917, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 2.883495145631068, | |
| "grad_norm": 0.39965224266052246, | |
| "learning_rate": 4.590362784169022e-08, | |
| "loss": 0.8002, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 2.8932038834951457, | |
| "grad_norm": 0.48560866713523865, | |
| "learning_rate": 3.858124340229863e-08, | |
| "loss": 0.7326, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 2.9029126213592233, | |
| "grad_norm": 0.41089245676994324, | |
| "learning_rate": 3.1892453488058803e-08, | |
| "loss": 0.7604, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 2.912621359223301, | |
| "grad_norm": 0.4088441729545593, | |
| "learning_rate": 2.5838112286226123e-08, | |
| "loss": 0.7339, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 2.9223300970873787, | |
| "grad_norm": 0.43762901425361633, | |
| "learning_rate": 2.0418992962224495e-08, | |
| "loss": 0.7299, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 2.9320388349514563, | |
| "grad_norm": 0.41060447692871094, | |
| "learning_rate": 1.563578756091144e-08, | |
| "loss": 0.7919, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 2.941747572815534, | |
| "grad_norm": 0.4254586398601532, | |
| "learning_rate": 1.1489106918200487e-08, | |
| "loss": 0.6998, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 2.9514563106796117, | |
| "grad_norm": 0.4328642785549164, | |
| "learning_rate": 7.979480583052423e-09, | |
| "loss": 0.7362, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 2.9611650485436893, | |
| "grad_norm": 0.40661028027534485, | |
| "learning_rate": 5.107356749853298e-09, | |
| "loss": 0.7416, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 2.970873786407767, | |
| "grad_norm": 0.4029662609100342, | |
| "learning_rate": 2.8731022011757593e-09, | |
| "loss": 0.7835, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 2.9805825242718447, | |
| "grad_norm": 0.4190027117729187, | |
| "learning_rate": 1.2770022609409628e-09, | |
| "loss": 0.7464, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 2.9902912621359223, | |
| "grad_norm": 0.40702641010284424, | |
| "learning_rate": 3.1926075797827914e-10, | |
| "loss": 0.7806, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "grad_norm": 0.40051835775375366, | |
| "learning_rate": 0.0, | |
| "loss": 0.7573, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "step": 309, | |
| "total_flos": 266439779352576.0, | |
| "train_loss": 0.8324689010586168, | |
| "train_runtime": 15120.0259, | |
| "train_samples_per_second": 1.962, | |
| "train_steps_per_second": 0.02 | |
| } | |
| ], | |
| "logging_steps": 1.0, | |
| "max_steps": 309, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 3, | |
| "save_steps": 500, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 266439779352576.0, | |
| "train_batch_size": 1, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |