{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.989955617846298, "global_step": 8000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0.0, "loss": 2.6426, "step": 1 }, { "epoch": 0.0, "learning_rate": 2.5275217623009062e-06, "loss": 2.1432, "step": 2 }, { "epoch": 0.0, "learning_rate": 4.006027213003589e-06, "loss": 3.0772, "step": 3 }, { "epoch": 0.0, "learning_rate": 5.0550435246018125e-06, "loss": 1.8552, "step": 4 }, { "epoch": 0.0, "learning_rate": 5.868723790325692e-06, "loss": 2.3031, "step": 5 }, { "epoch": 0.0, "learning_rate": 6.5335489753044945e-06, "loss": 2.7613, "step": 6 }, { "epoch": 0.0, "learning_rate": 7.095650660003159e-06, "loss": 1.9173, "step": 7 }, { "epoch": 0.0, "learning_rate": 7.582565286902719e-06, "loss": 2.2674, "step": 8 }, { "epoch": 0.0, "learning_rate": 8.012054426007177e-06, "loss": 1.8643, "step": 9 }, { "epoch": 0.0, "learning_rate": 8.3962455526266e-06, "loss": 1.9096, "step": 10 }, { "epoch": 0.0, "learning_rate": 8.74378870129762e-06, "loss": 2.3473, "step": 11 }, { "epoch": 0.0, "learning_rate": 9.061070737605401e-06, "loss": 2.0811, "step": 12 }, { "epoch": 0.0, "learning_rate": 9.352941917684244e-06, "loss": 2.3357, "step": 13 }, { "epoch": 0.01, "learning_rate": 9.623172422304064e-06, "loss": 2.7802, "step": 14 }, { "epoch": 0.01, "learning_rate": 9.87475100332928e-06, "loss": 3.3236, "step": 15 }, { "epoch": 0.01, "learning_rate": 1.0110087049203625e-05, "loss": 2.1941, "step": 16 }, { "epoch": 0.01, "learning_rate": 1.0331151283856529e-05, "loss": 2.2023, "step": 17 }, { "epoch": 0.01, "learning_rate": 1.0539576188308083e-05, "loss": 1.7227, "step": 18 }, { "epoch": 0.01, "learning_rate": 1.073672923490544e-05, "loss": 2.9091, "step": 19 }, { "epoch": 0.01, "learning_rate": 1.0923767314927505e-05, "loss": 2.3993, "step": 20 }, { "epoch": 0.01, "learning_rate": 1.1101677873006746e-05, "loss": 2.3862, "step": 21 }, { "epoch": 0.01, "learning_rate": 1.1271310463598526e-05, "loss": 2.5122, "step": 22 }, { "epoch": 0.01, "learning_rate": 1.1433401287050556e-05, "loss": 1.9362, "step": 23 }, { "epoch": 0.01, "learning_rate": 1.1588592499906309e-05, "loss": 2.1045, "step": 24 }, { "epoch": 0.01, "learning_rate": 1.1737447580651384e-05, "loss": 1.7602, "step": 25 }, { "epoch": 0.01, "learning_rate": 1.1880463679985151e-05, "loss": 2.0888, "step": 26 }, { "epoch": 0.01, "learning_rate": 1.2018081639010765e-05, "loss": 2.1335, "step": 27 }, { "epoch": 0.01, "learning_rate": 1.2150694184604972e-05, "loss": 1.7835, "step": 28 }, { "epoch": 0.01, "learning_rate": 1.2278652686029154e-05, "loss": 1.7154, "step": 29 }, { "epoch": 0.01, "learning_rate": 1.2402272765630187e-05, "loss": 1.6279, "step": 30 }, { "epoch": 0.01, "learning_rate": 1.2521838989213684e-05, "loss": 1.9767, "step": 31 }, { "epoch": 0.01, "learning_rate": 1.263760881150453e-05, "loss": 1.8594, "step": 32 }, { "epoch": 0.01, "learning_rate": 1.2749815914301206e-05, "loss": 1.7551, "step": 33 }, { "epoch": 0.01, "learning_rate": 1.2858673046157435e-05, "loss": 2.18, "step": 34 }, { "epoch": 0.01, "learning_rate": 1.2964374450328852e-05, "loss": 2.2767, "step": 35 }, { "epoch": 0.01, "learning_rate": 1.3067097950608989e-05, "loss": 2.0238, "step": 36 }, { "epoch": 0.01, "learning_rate": 1.3167006751318872e-05, "loss": 1.7321, "step": 37 }, { "epoch": 0.01, "learning_rate": 1.3264250997206345e-05, "loss": 2.1698, "step": 38 }, { "epoch": 0.01, "learning_rate": 1.3358969130687831e-05, "loss": 3.2648, "step": 39 }, { "epoch": 0.01, "learning_rate": 1.3451289077228411e-05, "loss": 2.1091, "step": 40 }, { "epoch": 0.02, "learning_rate": 1.3541329284331054e-05, "loss": 1.7113, "step": 41 }, { "epoch": 0.02, "learning_rate": 1.3629199635307655e-05, "loss": 1.7151, "step": 42 }, { "epoch": 0.02, "learning_rate": 1.371500225551594e-05, "loss": 1.7941, "step": 43 }, { "epoch": 0.02, "learning_rate": 1.379883222589943e-05, "loss": 2.7576, "step": 44 }, { "epoch": 0.02, "learning_rate": 1.3880778216332868e-05, "loss": 1.9716, "step": 45 }, { "epoch": 0.02, "learning_rate": 1.3960923049351463e-05, "loss": 1.7068, "step": 46 }, { "epoch": 0.02, "learning_rate": 1.403934420324923e-05, "loss": 2.9255, "step": 47 }, { "epoch": 0.02, "learning_rate": 1.4116114262207214e-05, "loss": 2.4605, "step": 48 }, { "epoch": 0.02, "learning_rate": 1.4191301320006319e-05, "loss": 1.6237, "step": 49 }, { "epoch": 0.02, "learning_rate": 1.426496934295229e-05, "loss": 2.0271, "step": 50 }, { "epoch": 0.02, "learning_rate": 1.4337178496860115e-05, "loss": 1.8699, "step": 51 }, { "epoch": 0.02, "learning_rate": 1.4407985442286057e-05, "loss": 2.3521, "step": 52 }, { "epoch": 0.02, "learning_rate": 1.4477443601636986e-05, "loss": 1.6885, "step": 53 }, { "epoch": 0.02, "learning_rate": 1.4545603401311673e-05, "loss": 1.9233, "step": 54 }, { "epoch": 0.02, "learning_rate": 1.4612512491623312e-05, "loss": 1.7288, "step": 55 }, { "epoch": 0.02, "learning_rate": 1.4678215946905881e-05, "loss": 3.0549, "step": 56 }, { "epoch": 0.02, "learning_rate": 1.4742756447909029e-05, "loss": 1.8896, "step": 57 }, { "epoch": 0.02, "learning_rate": 1.4806174448330058e-05, "loss": 2.0596, "step": 58 }, { "epoch": 0.02, "learning_rate": 1.486850832711022e-05, "loss": 1.6026, "step": 59 }, { "epoch": 0.02, "learning_rate": 1.4929794527931092e-05, "loss": 1.5923, "step": 60 }, { "epoch": 0.02, "learning_rate": 1.4990067687180732e-05, "loss": 1.8113, "step": 61 }, { "epoch": 0.02, "learning_rate": 1.504936075151459e-05, "loss": 1.7069, "step": 62 }, { "epoch": 0.02, "learning_rate": 1.5107705086010336e-05, "loss": 1.8804, "step": 63 }, { "epoch": 0.02, "learning_rate": 1.5165130573805438e-05, "loss": 2.6453, "step": 64 }, { "epoch": 0.02, "learning_rate": 1.5221665708009934e-05, "loss": 1.5752, "step": 65 }, { "epoch": 0.02, "learning_rate": 1.5277337676602113e-05, "loss": 1.6977, "step": 66 }, { "epoch": 0.03, "learning_rate": 1.5332172440940306e-05, "loss": 2.4345, "step": 67 }, { "epoch": 0.03, "learning_rate": 1.538619480845834e-05, "loss": 1.8328, "step": 68 }, { "epoch": 0.03, "learning_rate": 1.5439428500054147e-05, "loss": 2.1718, "step": 69 }, { "epoch": 0.03, "learning_rate": 1.549189621262976e-05, "loss": 1.7481, "step": 70 }, { "epoch": 0.03, "learning_rate": 1.5543619677195398e-05, "loss": 1.8838, "step": 71 }, { "epoch": 0.03, "learning_rate": 1.5594619712909895e-05, "loss": 1.8175, "step": 72 }, { "epoch": 0.03, "learning_rate": 1.564491627739385e-05, "loss": 1.7094, "step": 73 }, { "epoch": 0.03, "learning_rate": 1.569452851361978e-05, "loss": 2.3546, "step": 74 }, { "epoch": 0.03, "learning_rate": 1.5743474793654972e-05, "loss": 1.5372, "step": 75 }, { "epoch": 0.03, "learning_rate": 1.579177275950725e-05, "loss": 2.1604, "step": 76 }, { "epoch": 0.03, "learning_rate": 1.5839439361300778e-05, "loss": 1.6259, "step": 77 }, { "epoch": 0.03, "learning_rate": 1.5886490892988737e-05, "loss": 2.2185, "step": 78 }, { "epoch": 0.03, "learning_rate": 1.5932943025791115e-05, "loss": 1.6338, "step": 79 }, { "epoch": 0.03, "learning_rate": 1.5978810839529317e-05, "loss": 2.0721, "step": 80 }, { "epoch": 0.03, "learning_rate": 1.6024108852014355e-05, "loss": 2.0749, "step": 81 }, { "epoch": 0.03, "learning_rate": 1.606885104663196e-05, "loss": 2.1576, "step": 82 }, { "epoch": 0.03, "learning_rate": 1.611305089825575e-05, "loss": 2.0201, "step": 83 }, { "epoch": 0.03, "learning_rate": 1.615672139760856e-05, "loss": 2.0666, "step": 84 }, { "epoch": 0.03, "learning_rate": 1.619987507418222e-05, "loss": 1.9993, "step": 85 }, { "epoch": 0.03, "learning_rate": 1.6242524017816848e-05, "loss": 1.5733, "step": 86 }, { "epoch": 0.03, "learning_rate": 1.628467989903274e-05, "loss": 1.6697, "step": 87 }, { "epoch": 0.03, "learning_rate": 1.6326353988200337e-05, "loss": 1.8539, "step": 88 }, { "epoch": 0.03, "learning_rate": 1.6367557173627085e-05, "loss": 1.6318, "step": 89 }, { "epoch": 0.03, "learning_rate": 1.6408299978633774e-05, "loss": 2.3944, "step": 90 }, { "epoch": 0.03, "learning_rate": 1.6448592577687402e-05, "loss": 2.0548, "step": 91 }, { "epoch": 0.03, "learning_rate": 1.6488444811652367e-05, "loss": 1.9445, "step": 92 }, { "epoch": 0.03, "learning_rate": 1.6527866202217273e-05, "loss": 1.7664, "step": 93 }, { "epoch": 0.04, "learning_rate": 1.6566865965550136e-05, "loss": 1.4801, "step": 94 }, { "epoch": 0.04, "learning_rate": 1.660545302523113e-05, "loss": 1.7899, "step": 95 }, { "epoch": 0.04, "learning_rate": 1.6643636024508122e-05, "loss": 1.7782, "step": 96 }, { "epoch": 0.04, "learning_rate": 1.6681423337917193e-05, "loss": 1.6831, "step": 97 }, { "epoch": 0.04, "learning_rate": 1.6718823082307228e-05, "loss": 2.0187, "step": 98 }, { "epoch": 0.04, "learning_rate": 1.6755843127304794e-05, "loss": 1.8194, "step": 99 }, { "epoch": 0.04, "learning_rate": 1.67924911052532e-05, "loss": 2.9002, "step": 100 }, { "epoch": 0.04, "learning_rate": 1.6828774420656947e-05, "loss": 1.6493, "step": 101 }, { "epoch": 0.04, "learning_rate": 1.686470025916102e-05, "loss": 2.2883, "step": 102 }, { "epoch": 0.04, "learning_rate": 1.6900275596092067e-05, "loss": 1.9676, "step": 103 }, { "epoch": 0.04, "learning_rate": 1.693550720458696e-05, "loss": 1.816, "step": 104 }, { "epoch": 0.04, "learning_rate": 1.697040166333244e-05, "loss": 2.9702, "step": 105 }, { "epoch": 0.04, "learning_rate": 1.7004965363937892e-05, "loss": 1.62, "step": 106 }, { "epoch": 0.04, "learning_rate": 1.7039204517962008e-05, "loss": 1.608, "step": 107 }, { "epoch": 0.04, "learning_rate": 1.707312516361258e-05, "loss": 3.1899, "step": 108 }, { "epoch": 0.04, "learning_rate": 1.7106733172137546e-05, "loss": 2.2281, "step": 109 }, { "epoch": 0.04, "learning_rate": 1.714003425392422e-05, "loss": 2.2919, "step": 110 }, { "epoch": 0.04, "learning_rate": 1.717303396432246e-05, "loss": 1.6086, "step": 111 }, { "epoch": 0.04, "learning_rate": 1.7205737709206783e-05, "loss": 2.1789, "step": 112 }, { "epoch": 0.04, "learning_rate": 1.7238150750291204e-05, "loss": 1.7137, "step": 113 }, { "epoch": 0.04, "learning_rate": 1.7270278210209935e-05, "loss": 2.1611, "step": 114 }, { "epoch": 0.04, "learning_rate": 1.730212507737625e-05, "loss": 2.0386, "step": 115 }, { "epoch": 0.04, "learning_rate": 1.7333696210630963e-05, "loss": 1.9105, "step": 116 }, { "epoch": 0.04, "learning_rate": 1.736499634369142e-05, "loss": 1.7788, "step": 117 }, { "epoch": 0.04, "learning_rate": 1.7396030089411127e-05, "loss": 1.9972, "step": 118 }, { "epoch": 0.04, "learning_rate": 1.742680194385969e-05, "loss": 1.9597, "step": 119 }, { "epoch": 0.04, "learning_rate": 1.7457316290231997e-05, "loss": 1.6824, "step": 120 }, { "epoch": 0.05, "learning_rate": 1.748757740259524e-05, "loss": 1.647, "step": 121 }, { "epoch": 0.05, "learning_rate": 1.751758944948164e-05, "loss": 2.0926, "step": 122 }, { "epoch": 0.05, "learning_rate": 1.754735649733464e-05, "loss": 1.6035, "step": 123 }, { "epoch": 0.05, "learning_rate": 1.7576882513815497e-05, "loss": 1.7259, "step": 124 }, { "epoch": 0.05, "learning_rate": 1.7606171370977078e-05, "loss": 2.0771, "step": 125 }, { "epoch": 0.05, "learning_rate": 1.7635226848311243e-05, "loss": 2.8924, "step": 126 }, { "epoch": 0.05, "learning_rate": 1.766405263567574e-05, "loss": 1.4472, "step": 127 }, { "epoch": 0.05, "learning_rate": 1.7692652336106342e-05, "loss": 1.8262, "step": 128 }, { "epoch": 0.05, "learning_rate": 1.7721029468519532e-05, "loss": 1.6181, "step": 129 }, { "epoch": 0.05, "learning_rate": 1.774918747031084e-05, "loss": 2.4223, "step": 130 }, { "epoch": 0.05, "learning_rate": 1.7777129699853667e-05, "loss": 1.8272, "step": 131 }, { "epoch": 0.05, "learning_rate": 1.780485943890302e-05, "loss": 1.6758, "step": 132 }, { "epoch": 0.05, "learning_rate": 1.78323798949086e-05, "loss": 2.192, "step": 133 }, { "epoch": 0.05, "learning_rate": 1.7859694203241214e-05, "loss": 2.2492, "step": 134 }, { "epoch": 0.05, "learning_rate": 1.7886805429336457e-05, "loss": 1.495, "step": 135 }, { "epoch": 0.05, "learning_rate": 1.7913716570759248e-05, "loss": 1.9457, "step": 136 }, { "epoch": 0.05, "learning_rate": 1.7940430559192765e-05, "loss": 2.5729, "step": 137 }, { "epoch": 0.05, "learning_rate": 1.796695026235505e-05, "loss": 1.7335, "step": 138 }, { "epoch": 0.05, "learning_rate": 1.7993278485846422e-05, "loss": 1.6763, "step": 139 }, { "epoch": 0.05, "learning_rate": 1.8019417974930662e-05, "loss": 2.1221, "step": 140 }, { "epoch": 0.05, "learning_rate": 1.804537141625282e-05, "loss": 2.2923, "step": 141 }, { "epoch": 0.05, "learning_rate": 1.8071141439496305e-05, "loss": 1.8324, "step": 142 }, { "epoch": 0.05, "learning_rate": 1.809673061898186e-05, "loss": 1.5252, "step": 143 }, { "epoch": 0.05, "learning_rate": 1.8122141475210802e-05, "loss": 1.6176, "step": 144 }, { "epoch": 0.05, "learning_rate": 1.8147376476354842e-05, "loss": 2.1793, "step": 145 }, { "epoch": 0.05, "learning_rate": 1.817243803969476e-05, "loss": 1.7846, "step": 146 }, { "epoch": 0.05, "learning_rate": 1.8197328533009905e-05, "loss": 1.9344, "step": 147 }, { "epoch": 0.06, "learning_rate": 1.822205027592068e-05, "loss": 1.7784, "step": 148 }, { "epoch": 0.06, "learning_rate": 1.8246605541185748e-05, "loss": 1.6667, "step": 149 }, { "epoch": 0.06, "learning_rate": 1.8270996555955876e-05, "loss": 1.7487, "step": 150 }, { "epoch": 0.06, "learning_rate": 1.8295225502986157e-05, "loss": 1.4201, "step": 151 }, { "epoch": 0.06, "learning_rate": 1.831929452180816e-05, "loss": 1.8313, "step": 152 }, { "epoch": 0.06, "learning_rate": 1.8343205709863705e-05, "loss": 1.7357, "step": 153 }, { "epoch": 0.06, "learning_rate": 1.8366961123601686e-05, "loss": 2.1689, "step": 154 }, { "epoch": 0.06, "learning_rate": 1.8390562779539376e-05, "loss": 1.7823, "step": 155 }, { "epoch": 0.06, "learning_rate": 1.8414012655289645e-05, "loss": 1.5239, "step": 156 }, { "epoch": 0.06, "learning_rate": 1.8437312690555323e-05, "loss": 2.3732, "step": 157 }, { "epoch": 0.06, "learning_rate": 1.8460464788092026e-05, "loss": 2.1957, "step": 158 }, { "epoch": 0.06, "learning_rate": 1.8483470814640572e-05, "loss": 2.2766, "step": 159 }, { "epoch": 0.06, "learning_rate": 1.8506332601830225e-05, "loss": 2.0382, "step": 160 }, { "epoch": 0.06, "learning_rate": 1.8529051947053716e-05, "loss": 1.751, "step": 161 }, { "epoch": 0.06, "learning_rate": 1.8551630614315255e-05, "loss": 1.8844, "step": 162 }, { "epoch": 0.06, "learning_rate": 1.857407033505242e-05, "loss": 1.8671, "step": 163 }, { "epoch": 0.06, "learning_rate": 1.8596372808932867e-05, "loss": 1.8401, "step": 164 }, { "epoch": 0.06, "learning_rate": 1.86185397046269e-05, "loss": 1.7625, "step": 165 }, { "epoch": 0.06, "learning_rate": 1.8640572660556654e-05, "loss": 2.4511, "step": 166 }, { "epoch": 0.06, "learning_rate": 1.8662473285622785e-05, "loss": 1.7486, "step": 167 }, { "epoch": 0.06, "learning_rate": 1.8684243159909464e-05, "loss": 2.5572, "step": 168 }, { "epoch": 0.06, "learning_rate": 1.8705883835368487e-05, "loss": 1.8903, "step": 169 }, { "epoch": 0.06, "learning_rate": 1.8727396836483127e-05, "loss": 2.3876, "step": 170 }, { "epoch": 0.06, "learning_rate": 1.8748783660912618e-05, "loss": 2.1449, "step": 171 }, { "epoch": 0.06, "learning_rate": 1.8770045780117752e-05, "loss": 1.7251, "step": 172 }, { "epoch": 0.06, "learning_rate": 1.8791184639968442e-05, "loss": 2.8774, "step": 173 }, { "epoch": 0.07, "learning_rate": 1.8812201661333647e-05, "loss": 1.8219, "step": 174 }, { "epoch": 0.07, "learning_rate": 1.8833098240654544e-05, "loss": 2.0163, "step": 175 }, { "epoch": 0.07, "learning_rate": 1.885387575050124e-05, "loss": 1.565, "step": 176 }, { "epoch": 0.07, "learning_rate": 1.8874535540113804e-05, "loss": 2.0268, "step": 177 }, { "epoch": 0.07, "learning_rate": 1.889507893592799e-05, "loss": 2.2784, "step": 178 }, { "epoch": 0.07, "learning_rate": 1.8915507242086283e-05, "loss": 2.258, "step": 179 }, { "epoch": 0.07, "learning_rate": 1.893582174093468e-05, "loss": 2.1834, "step": 180 }, { "epoch": 0.07, "learning_rate": 1.895602369350575e-05, "loss": 2.2412, "step": 181 }, { "epoch": 0.07, "learning_rate": 1.8976114339988306e-05, "loss": 1.8971, "step": 182 }, { "epoch": 0.07, "learning_rate": 1.8996094900184322e-05, "loss": 1.9897, "step": 183 }, { "epoch": 0.07, "learning_rate": 1.9015966573953278e-05, "loss": 1.5837, "step": 184 }, { "epoch": 0.07, "learning_rate": 1.9035730541644563e-05, "loss": 1.5574, "step": 185 }, { "epoch": 0.07, "learning_rate": 1.905538796451818e-05, "loss": 2.625, "step": 186 }, { "epoch": 0.07, "learning_rate": 1.9074939985154147e-05, "loss": 1.7953, "step": 187 }, { "epoch": 0.07, "learning_rate": 1.9094387727851043e-05, "loss": 2.1029, "step": 188 }, { "epoch": 0.07, "learning_rate": 1.9113732299013924e-05, "loss": 2.1656, "step": 189 }, { "epoch": 0.07, "learning_rate": 1.9132974787532037e-05, "loss": 2.472, "step": 190 }, { "epoch": 0.07, "learning_rate": 1.9152116265146608e-05, "loss": 1.5024, "step": 191 }, { "epoch": 0.07, "learning_rate": 1.9171157786809026e-05, "loss": 1.5383, "step": 192 }, { "epoch": 0.07, "learning_rate": 1.919010039102974e-05, "loss": 1.5722, "step": 193 }, { "epoch": 0.07, "learning_rate": 1.92089451002181e-05, "loss": 1.7659, "step": 194 }, { "epoch": 0.07, "learning_rate": 1.9227692921013524e-05, "loss": 2.8222, "step": 195 }, { "epoch": 0.07, "learning_rate": 1.924634484460813e-05, "loss": 1.5663, "step": 196 }, { "epoch": 0.07, "learning_rate": 1.926490184706121e-05, "loss": 1.7263, "step": 197 }, { "epoch": 0.07, "learning_rate": 1.92833648896057e-05, "loss": 2.0352, "step": 198 }, { "epoch": 0.07, "learning_rate": 1.9301734918946974e-05, "loss": 2.2065, "step": 199 }, { "epoch": 0.07, "learning_rate": 1.9320012867554103e-05, "loss": 2.1599, "step": 200 }, { "epoch": 0.08, "learning_rate": 1.9338199653943897e-05, "loss": 1.7651, "step": 201 }, { "epoch": 0.08, "learning_rate": 1.9356296182957855e-05, "loss": 1.8416, "step": 202 }, { "epoch": 0.08, "learning_rate": 1.9374303346032312e-05, "loss": 2.1721, "step": 203 }, { "epoch": 0.08, "learning_rate": 1.939222202146193e-05, "loss": 2.2034, "step": 204 }, { "epoch": 0.08, "learning_rate": 1.9410053074656746e-05, "loss": 1.6954, "step": 205 }, { "epoch": 0.08, "learning_rate": 1.9427797358392975e-05, "loss": 1.6021, "step": 206 }, { "epoch": 0.08, "learning_rate": 1.944545571305773e-05, "loss": 1.5624, "step": 207 }, { "epoch": 0.08, "learning_rate": 1.946302896688787e-05, "loss": 3.3599, "step": 208 }, { "epoch": 0.08, "learning_rate": 1.9480517936203057e-05, "loss": 1.6392, "step": 209 }, { "epoch": 0.08, "learning_rate": 1.9497923425633346e-05, "loss": 1.4861, "step": 210 }, { "epoch": 0.08, "learning_rate": 1.9515246228341283e-05, "loss": 1.9002, "step": 211 }, { "epoch": 0.08, "learning_rate": 1.95324871262388e-05, "loss": 2.6061, "step": 212 }, { "epoch": 0.08, "learning_rate": 1.9549646890198985e-05, "loss": 2.1217, "step": 213 }, { "epoch": 0.08, "learning_rate": 1.9566726280262912e-05, "loss": 1.6178, "step": 214 }, { "epoch": 0.08, "learning_rate": 1.9583726045841635e-05, "loss": 2.086, "step": 215 }, { "epoch": 0.08, "learning_rate": 1.9600646925913486e-05, "loss": 1.8507, "step": 216 }, { "epoch": 0.08, "learning_rate": 1.9617489649216842e-05, "loss": 2.2009, "step": 217 }, { "epoch": 0.08, "learning_rate": 1.9634254934438453e-05, "loss": 1.9894, "step": 218 }, { "epoch": 0.08, "learning_rate": 1.965094349039744e-05, "loss": 1.5186, "step": 219 }, { "epoch": 0.08, "learning_rate": 1.9667556016225124e-05, "loss": 1.6066, "step": 220 }, { "epoch": 0.08, "learning_rate": 1.9684093201540767e-05, "loss": 1.8685, "step": 221 }, { "epoch": 0.08, "learning_rate": 1.9700555726623364e-05, "loss": 1.4328, "step": 222 }, { "epoch": 0.08, "learning_rate": 1.9716944262579534e-05, "loss": 1.8252, "step": 223 }, { "epoch": 0.08, "learning_rate": 1.973325947150769e-05, "loss": 1.9606, "step": 224 }, { "epoch": 0.08, "learning_rate": 1.974950200665856e-05, "loss": 1.8818, "step": 225 }, { "epoch": 0.08, "learning_rate": 1.976567251259211e-05, "loss": 1.9281, "step": 226 }, { "epoch": 0.08, "learning_rate": 1.9781771625331026e-05, "loss": 2.0567, "step": 227 }, { "epoch": 0.09, "learning_rate": 1.9797799972510842e-05, "loss": 1.9799, "step": 228 }, { "epoch": 0.09, "learning_rate": 1.981375817352673e-05, "loss": 1.702, "step": 229 }, { "epoch": 0.09, "learning_rate": 1.9829646839677157e-05, "loss": 2.0745, "step": 230 }, { "epoch": 0.09, "learning_rate": 1.9845466574304366e-05, "loss": 1.8447, "step": 231 }, { "epoch": 0.09, "learning_rate": 1.986121797293187e-05, "loss": 2.7649, "step": 232 }, { "epoch": 0.09, "learning_rate": 1.9876901623398955e-05, "loss": 1.7408, "step": 233 }, { "epoch": 0.09, "learning_rate": 1.989251810599233e-05, "loss": 2.1156, "step": 234 }, { "epoch": 0.09, "learning_rate": 1.9908067993574922e-05, "loss": 1.7571, "step": 235 }, { "epoch": 0.09, "learning_rate": 1.9923551851712035e-05, "loss": 1.901, "step": 236 }, { "epoch": 0.09, "learning_rate": 1.9938970238794706e-05, "loss": 1.6641, "step": 237 }, { "epoch": 0.09, "learning_rate": 1.9954323706160593e-05, "loss": 1.9086, "step": 238 }, { "epoch": 0.09, "learning_rate": 1.996961279821224e-05, "loss": 3.2042, "step": 239 }, { "epoch": 0.09, "learning_rate": 1.9984838052532908e-05, "loss": 1.7143, "step": 240 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 2.3531, "step": 241 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.7472, "step": 242 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 2.7393, "step": 243 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 2.2857, "step": 244 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.6813, "step": 245 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.7284, "step": 246 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.9587, "step": 247 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.6377, "step": 248 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 2.0021, "step": 249 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 2.184, "step": 250 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.5867, "step": 251 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.7226, "step": 252 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.9286, "step": 253 }, { "epoch": 0.09, "learning_rate": 2e-05, "loss": 1.5381, "step": 254 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.9171, "step": 255 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.9027, "step": 256 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 2.2408, "step": 257 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.6926, "step": 258 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 3.6121, "step": 259 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 2.6241, "step": 260 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.542, "step": 261 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.5244, "step": 262 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.9413, "step": 263 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.7372, "step": 264 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.9097, "step": 265 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.574, "step": 266 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.6404, "step": 267 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.4424, "step": 268 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 2.1742, "step": 269 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 2.3844, "step": 270 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.7161, "step": 271 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 2.1741, "step": 272 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.6273, "step": 273 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.975, "step": 274 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 2.3553, "step": 275 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 2.6571, "step": 276 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.861, "step": 277 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.5983, "step": 278 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.9935, "step": 279 }, { "epoch": 0.1, "learning_rate": 2e-05, "loss": 1.6852, "step": 280 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.3953, "step": 281 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.5885, "step": 282 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.8756, "step": 283 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.7659, "step": 284 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.0443, "step": 285 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.6795, "step": 286 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.4615, "step": 287 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.6409, "step": 288 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.6848, "step": 289 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.5496, "step": 290 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.7684, "step": 291 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.0201, "step": 292 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.9063, "step": 293 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.1326, "step": 294 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.5165, "step": 295 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.6364, "step": 296 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.4971, "step": 297 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.6709, "step": 298 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.5658, "step": 299 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.8525, "step": 300 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.0225, "step": 301 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.0197, "step": 302 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.0148, "step": 303 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 2.0071, "step": 304 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.8622, "step": 305 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.5113, "step": 306 }, { "epoch": 0.11, "learning_rate": 2e-05, "loss": 1.6686, "step": 307 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.7636, "step": 308 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.8152, "step": 309 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.5611, "step": 310 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.3336, "step": 311 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.0939, "step": 312 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.1763, "step": 313 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.9391, "step": 314 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.3628, "step": 315 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.39, "step": 316 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.5587, "step": 317 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.5259, "step": 318 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.5679, "step": 319 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.7421, "step": 320 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.5828, "step": 321 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.7818, "step": 322 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.2803, "step": 323 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.6553, "step": 324 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.6656, "step": 325 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.7929, "step": 326 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.9603, "step": 327 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.0342, "step": 328 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.6577, "step": 329 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.8132, "step": 330 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.3666, "step": 331 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.3577, "step": 332 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 2.5034, "step": 333 }, { "epoch": 0.12, "learning_rate": 2e-05, "loss": 1.9763, "step": 334 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.0754, "step": 335 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.7882, "step": 336 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.0016, "step": 337 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.8509, "step": 338 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.825, "step": 339 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.4503, "step": 340 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.5286, "step": 341 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.1777, "step": 342 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.8353, "step": 343 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.6891, "step": 344 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.5987, "step": 345 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.4378, "step": 346 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.8246, "step": 347 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.5212, "step": 348 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.5818, "step": 349 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.6003, "step": 350 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.6685, "step": 351 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.3171, "step": 352 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.388, "step": 353 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.6642, "step": 354 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.9851, "step": 355 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.0377, "step": 356 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.664, "step": 357 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.6161, "step": 358 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.3906, "step": 359 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 1.5749, "step": 360 }, { "epoch": 0.13, "learning_rate": 2e-05, "loss": 2.3155, "step": 361 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.9047, "step": 362 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.6392, "step": 363 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.2611, "step": 364 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.8287, "step": 365 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.3524, "step": 366 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.6099, "step": 367 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.8831, "step": 368 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.7487, "step": 369 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.2914, "step": 370 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.6012, "step": 371 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.5518, "step": 372 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.7825, "step": 373 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.9791, "step": 374 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.5105, "step": 375 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.5763, "step": 376 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.4131, "step": 377 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.4956, "step": 378 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.0049, "step": 379 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.7213, "step": 380 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.1581, "step": 381 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.6846, "step": 382 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.6508, "step": 383 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.6833, "step": 384 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.3529, "step": 385 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 1.7153, "step": 386 }, { "epoch": 0.14, "learning_rate": 2e-05, "loss": 2.1558, "step": 387 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.4667, "step": 388 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.9089, "step": 389 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.736, "step": 390 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.2257, "step": 391 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.6381, "step": 392 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.6005, "step": 393 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.0476, "step": 394 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.5129, "step": 395 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.371, "step": 396 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.5201, "step": 397 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.7748, "step": 398 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.3476, "step": 399 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.7525, "step": 400 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.4762, "step": 401 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.5774, "step": 402 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.5272, "step": 403 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.9071, "step": 404 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.4695, "step": 405 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.0196, "step": 406 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 2.1932, "step": 407 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.5739, "step": 408 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.7021, "step": 409 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.2952, "step": 410 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.7524, "step": 411 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.8419, "step": 412 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.6079, "step": 413 }, { "epoch": 0.15, "learning_rate": 2e-05, "loss": 1.8207, "step": 414 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.812, "step": 415 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.7005, "step": 416 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.4748, "step": 417 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.9244, "step": 418 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.1323, "step": 419 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.5889, "step": 420 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.4321, "step": 421 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.498, "step": 422 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.6541, "step": 423 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.6255, "step": 424 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.1063, "step": 425 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.6717, "step": 426 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.8489, "step": 427 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.9011, "step": 428 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.056, "step": 429 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.772, "step": 430 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.5775, "step": 431 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.1235, "step": 432 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.6159, "step": 433 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.8361, "step": 434 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.2248, "step": 435 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.809, "step": 436 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.7378, "step": 437 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.5916, "step": 438 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.4958, "step": 439 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 1.6946, "step": 440 }, { "epoch": 0.16, "learning_rate": 2e-05, "loss": 2.2082, "step": 441 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.648, "step": 442 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5769, "step": 443 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.7213, "step": 444 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 2.5261, "step": 445 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5235, "step": 446 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.49, "step": 447 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.916, "step": 448 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5436, "step": 449 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5002, "step": 450 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 2.1652, "step": 451 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5829, "step": 452 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5892, "step": 453 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5975, "step": 454 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 3.5801, "step": 455 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5526, "step": 456 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 2.3531, "step": 457 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.5261, "step": 458 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 2.0296, "step": 459 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.729, "step": 460 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.4886, "step": 461 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.4129, "step": 462 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.9116, "step": 463 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.7786, "step": 464 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.8222, "step": 465 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.8767, "step": 466 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 2.0499, "step": 467 }, { "epoch": 0.17, "learning_rate": 2e-05, "loss": 1.4492, "step": 468 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.8993, "step": 469 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.8031, "step": 470 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.5636, "step": 471 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.5819, "step": 472 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 2.9642, "step": 473 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 2.1877, "step": 474 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 2.0724, "step": 475 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 2.0779, "step": 476 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.4377, "step": 477 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.928, "step": 478 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 2.4608, "step": 479 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.8081, "step": 480 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.8256, "step": 481 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.5102, "step": 482 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.7066, "step": 483 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.7979, "step": 484 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.8199, "step": 485 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.6691, "step": 486 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.72, "step": 487 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.6178, "step": 488 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.6891, "step": 489 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 2.0369, "step": 490 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.8376, "step": 491 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.4432, "step": 492 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.5108, "step": 493 }, { "epoch": 0.18, "learning_rate": 2e-05, "loss": 1.723, "step": 494 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.506, "step": 495 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.6638, "step": 496 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.7543, "step": 497 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.2002, "step": 498 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.6944, "step": 499 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.8226, "step": 500 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.0406, "step": 501 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.7988, "step": 502 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.0339, "step": 503 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.0903, "step": 504 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.1106, "step": 505 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.4248, "step": 506 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.0909, "step": 507 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.8174, "step": 508 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.8184, "step": 509 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.8973, "step": 510 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.8142, "step": 511 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.0682, "step": 512 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.1716, "step": 513 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.2174, "step": 514 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.9621, "step": 515 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.602, "step": 516 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.6271, "step": 517 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.825, "step": 518 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.9609, "step": 519 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 1.3973, "step": 520 }, { "epoch": 0.19, "learning_rate": 2e-05, "loss": 2.0571, "step": 521 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.7373, "step": 522 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.8226, "step": 523 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.6731, "step": 524 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.8957, "step": 525 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 2.0815, "step": 526 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 2.1208, "step": 527 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.5355, "step": 528 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.5498, "step": 529 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.5178, "step": 530 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.5751, "step": 531 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.5022, "step": 532 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.7809, "step": 533 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.4953, "step": 534 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.6288, "step": 535 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.6875, "step": 536 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 2.6228, "step": 537 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 2.036, "step": 538 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.5156, "step": 539 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 2.045, "step": 540 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.7949, "step": 541 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.4789, "step": 542 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 3.095, "step": 543 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.8496, "step": 544 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.9156, "step": 545 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.9493, "step": 546 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.6735, "step": 547 }, { "epoch": 0.2, "learning_rate": 2e-05, "loss": 1.5374, "step": 548 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.5516, "step": 549 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4139, "step": 550 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4687, "step": 551 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4564, "step": 552 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.0734, "step": 553 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.5033, "step": 554 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.0883, "step": 555 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.5602, "step": 556 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.0761, "step": 557 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.8671, "step": 558 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.639, "step": 559 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.795, "step": 560 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4592, "step": 561 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4944, "step": 562 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.8873, "step": 563 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 3.4738, "step": 564 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.3862, "step": 565 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4829, "step": 566 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.6151, "step": 567 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4507, "step": 568 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.5708, "step": 569 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 2.0362, "step": 570 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.5099, "step": 571 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.9209, "step": 572 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.4007, "step": 573 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.5702, "step": 574 }, { "epoch": 0.21, "learning_rate": 2e-05, "loss": 1.5271, "step": 575 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.5788, "step": 576 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.4685, "step": 577 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.7763, "step": 578 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.6244, "step": 579 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.9173, "step": 580 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.9769, "step": 581 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.5702, "step": 582 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.6108, "step": 583 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.3447, "step": 584 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.6518, "step": 585 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.9364, "step": 586 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.9251, "step": 587 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.8371, "step": 588 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.8775, "step": 589 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.9409, "step": 590 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.6387, "step": 591 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.715, "step": 592 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 2.061, "step": 593 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 2.4085, "step": 594 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 2.5026, "step": 595 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.6605, "step": 596 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 2.0384, "step": 597 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.7342, "step": 598 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.6327, "step": 599 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 2.3445, "step": 600 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.4931, "step": 601 }, { "epoch": 0.22, "learning_rate": 2e-05, "loss": 1.6974, "step": 602 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.5977, "step": 603 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.8938, "step": 604 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 2.077, "step": 605 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.7221, "step": 606 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.8218, "step": 607 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.3688, "step": 608 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 2.2238, "step": 609 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.6617, "step": 610 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.6247, "step": 611 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.917, "step": 612 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.5813, "step": 613 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.587, "step": 614 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.6096, "step": 615 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.7352, "step": 616 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.6765, "step": 617 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.6761, "step": 618 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.586, "step": 619 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 2.2945, "step": 620 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.4499, "step": 621 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.8824, "step": 622 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.9487, "step": 623 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.6092, "step": 624 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 2.1745, "step": 625 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.5169, "step": 626 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.8492, "step": 627 }, { "epoch": 0.23, "learning_rate": 2e-05, "loss": 1.3554, "step": 628 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.9643, "step": 629 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.5421, "step": 630 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.6254, "step": 631 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.5062, "step": 632 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.8142, "step": 633 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.0032, "step": 634 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.4378, "step": 635 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.9647, "step": 636 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.5972, "step": 637 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.6318, "step": 638 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.7093, "step": 639 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.4535, "step": 640 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.1349, "step": 641 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.2957, "step": 642 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.5533, "step": 643 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.8759, "step": 644 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.4045, "step": 645 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.3177, "step": 646 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.1266, "step": 647 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.7482, "step": 648 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.0209, "step": 649 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.9525, "step": 650 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.4767, "step": 651 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.0655, "step": 652 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 2.5588, "step": 653 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.7254, "step": 654 }, { "epoch": 0.24, "learning_rate": 2e-05, "loss": 1.5964, "step": 655 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.6383, "step": 656 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.6446, "step": 657 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.2439, "step": 658 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.5628, "step": 659 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.5899, "step": 660 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.7777, "step": 661 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.8932, "step": 662 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.508, "step": 663 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.0793, "step": 664 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.8367, "step": 665 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.609, "step": 666 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.2323, "step": 667 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.0026, "step": 668 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.3825, "step": 669 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.8817, "step": 670 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.8535, "step": 671 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.6879, "step": 672 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.7014, "step": 673 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.3204, "step": 674 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.8221, "step": 675 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.5395, "step": 676 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.7874, "step": 677 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.4346, "step": 678 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 1.6278, "step": 679 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 3.057, "step": 680 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.6561, "step": 681 }, { "epoch": 0.25, "learning_rate": 2e-05, "loss": 2.3093, "step": 682 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.5774, "step": 683 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.6376, "step": 684 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.0877, "step": 685 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.5623, "step": 686 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.5157, "step": 687 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.5601, "step": 688 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.3382, "step": 689 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.9512, "step": 690 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.7347, "step": 691 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.8252, "step": 692 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.9419, "step": 693 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.7935, "step": 694 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.3956, "step": 695 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.4052, "step": 696 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.2299, "step": 697 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.7798, "step": 698 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.1505, "step": 699 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.6491, "step": 700 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.6892, "step": 701 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.2204, "step": 702 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 2.1628, "step": 703 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.9972, "step": 704 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.8634, "step": 705 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.8514, "step": 706 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.6892, "step": 707 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.6355, "step": 708 }, { "epoch": 0.26, "learning_rate": 2e-05, "loss": 1.5195, "step": 709 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.3958, "step": 710 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.8441, "step": 711 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.8787, "step": 712 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.4317, "step": 713 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.7979, "step": 714 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.5799, "step": 715 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.8465, "step": 716 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.4687, "step": 717 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 2.0722, "step": 718 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.5615, "step": 719 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.5201, "step": 720 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.875, "step": 721 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.8253, "step": 722 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.5788, "step": 723 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.6062, "step": 724 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.8491, "step": 725 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.4355, "step": 726 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.4587, "step": 727 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.4501, "step": 728 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.8153, "step": 729 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.7286, "step": 730 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.487, "step": 731 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 2.0948, "step": 732 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 2.1334, "step": 733 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.8207, "step": 734 }, { "epoch": 0.27, "learning_rate": 2e-05, "loss": 1.5271, "step": 735 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.6884, "step": 736 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.5589, "step": 737 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.6545, "step": 738 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.5766, "step": 739 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.5348, "step": 740 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.4435, "step": 741 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.7061, "step": 742 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.6083, "step": 743 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.6627, "step": 744 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 2.1097, "step": 745 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.7156, "step": 746 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.5944, "step": 747 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 2.2239, "step": 748 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.6477, "step": 749 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.987, "step": 750 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.5295, "step": 751 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.7906, "step": 752 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.7906, "step": 753 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 3.6881, "step": 754 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.4764, "step": 755 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.6436, "step": 756 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.7328, "step": 757 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.8706, "step": 758 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 2.2326, "step": 759 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.3992, "step": 760 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.3581, "step": 761 }, { "epoch": 0.28, "learning_rate": 2e-05, "loss": 1.6722, "step": 762 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.8405, "step": 763 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.8567, "step": 764 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.9563, "step": 765 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 2.4572, "step": 766 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.6714, "step": 767 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.4369, "step": 768 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.8769, "step": 769 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.6436, "step": 770 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.7417, "step": 771 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.5206, "step": 772 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.5447, "step": 773 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.7899, "step": 774 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.2825, "step": 775 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 2.7144, "step": 776 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 2.0266, "step": 777 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.5898, "step": 778 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.4428, "step": 779 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.6207, "step": 780 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.5125, "step": 781 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.9827, "step": 782 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.5161, "step": 783 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 2.0577, "step": 784 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.966, "step": 785 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 2.4179, "step": 786 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.4965, "step": 787 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 1.7884, "step": 788 }, { "epoch": 0.29, "learning_rate": 2e-05, "loss": 2.2702, "step": 789 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.8465, "step": 790 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.8145, "step": 791 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.7795, "step": 792 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.7829, "step": 793 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.9039, "step": 794 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.7393, "step": 795 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.2893, "step": 796 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.9096, "step": 797 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.4102, "step": 798 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.6492, "step": 799 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.5917, "step": 800 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 2.0997, "step": 801 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 2.4327, "step": 802 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.3769, "step": 803 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.5113, "step": 804 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.9964, "step": 805 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.6749, "step": 806 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.5081, "step": 807 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.7766, "step": 808 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.6078, "step": 809 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.6632, "step": 810 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.553, "step": 811 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.2439, "step": 812 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.6738, "step": 813 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.1895, "step": 814 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 1.6099, "step": 815 }, { "epoch": 0.3, "learning_rate": 2e-05, "loss": 2.1326, "step": 816 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.5352, "step": 817 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.8948, "step": 818 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.6495, "step": 819 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.5012, "step": 820 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.7545, "step": 821 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.469, "step": 822 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.4433, "step": 823 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.478, "step": 824 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.6032, "step": 825 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.3213, "step": 826 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.9365, "step": 827 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.7652, "step": 828 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.6464, "step": 829 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.4451, "step": 830 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 2.4611, "step": 831 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.1879, "step": 832 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.7898, "step": 833 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 2.7522, "step": 834 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.7038, "step": 835 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.9375, "step": 836 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.5881, "step": 837 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.772, "step": 838 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.497, "step": 839 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 2.0743, "step": 840 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.9872, "step": 841 }, { "epoch": 0.31, "learning_rate": 2e-05, "loss": 1.7047, "step": 842 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.9707, "step": 843 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 2.0615, "step": 844 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.7956, "step": 845 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.5934, "step": 846 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.4842, "step": 847 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.9689, "step": 848 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.4864, "step": 849 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.4982, "step": 850 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.6386, "step": 851 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.4347, "step": 852 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.8897, "step": 853 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.6582, "step": 854 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.741, "step": 855 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.5276, "step": 856 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.478, "step": 857 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.711, "step": 858 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.2876, "step": 859 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.9817, "step": 860 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.7538, "step": 861 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 2.5501, "step": 862 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 2.0698, "step": 863 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.593, "step": 864 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.8321, "step": 865 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.8851, "step": 866 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.4257, "step": 867 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 2.2509, "step": 868 }, { "epoch": 0.32, "learning_rate": 2e-05, "loss": 1.5088, "step": 869 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.5322, "step": 870 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.5028, "step": 871 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.5852, "step": 872 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.6745, "step": 873 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.4878, "step": 874 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.8096, "step": 875 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 2.713, "step": 876 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.4492, "step": 877 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 2.6804, "step": 878 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.722, "step": 879 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.8867, "step": 880 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.4435, "step": 881 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.9356, "step": 882 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 2.3085, "step": 883 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 2.0268, "step": 884 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.9494, "step": 885 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.5602, "step": 886 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.7802, "step": 887 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.7995, "step": 888 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.9498, "step": 889 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.4899, "step": 890 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.5866, "step": 891 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.5154, "step": 892 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.4376, "step": 893 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.9403, "step": 894 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 1.6659, "step": 895 }, { "epoch": 0.33, "learning_rate": 2e-05, "loss": 2.9111, "step": 896 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.4505, "step": 897 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.606, "step": 898 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.5247, "step": 899 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.2814, "step": 900 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.659, "step": 901 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 2.1309, "step": 902 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 2.0639, "step": 903 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.187, "step": 904 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.9303, "step": 905 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.7262, "step": 906 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.3869, "step": 907 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.8738, "step": 908 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.5912, "step": 909 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.7191, "step": 910 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.614, "step": 911 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 2.3484, "step": 912 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.6876, "step": 913 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.63, "step": 914 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 2.5607, "step": 915 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 2.0209, "step": 916 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.3258, "step": 917 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 2.444, "step": 918 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.7102, "step": 919 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.5413, "step": 920 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.4057, "step": 921 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 1.4291, "step": 922 }, { "epoch": 0.34, "learning_rate": 2e-05, "loss": 2.1571, "step": 923 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.6949, "step": 924 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.7771, "step": 925 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.2977, "step": 926 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4505, "step": 927 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.5475, "step": 928 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 2.9381, "step": 929 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4272, "step": 930 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4238, "step": 931 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.7175, "step": 932 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.5091, "step": 933 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4202, "step": 934 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.9427, "step": 935 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 2.0026, "step": 936 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.8063, "step": 937 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.8774, "step": 938 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.7602, "step": 939 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4604, "step": 940 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4902, "step": 941 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.9359, "step": 942 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4005, "step": 943 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4053, "step": 944 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.6944, "step": 945 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.3819, "step": 946 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.4929, "step": 947 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.3946, "step": 948 }, { "epoch": 0.35, "learning_rate": 2e-05, "loss": 1.2182, "step": 949 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.3681, "step": 950 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.3273, "step": 951 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.9816, "step": 952 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.4487, "step": 953 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.9943, "step": 954 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.5602, "step": 955 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 2.646, "step": 956 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.5005, "step": 957 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.5472, "step": 958 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.7534, "step": 959 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.6741, "step": 960 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.649, "step": 961 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.7113, "step": 962 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.6238, "step": 963 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.444, "step": 964 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 2.2359, "step": 965 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.7458, "step": 966 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.8945, "step": 967 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.6348, "step": 968 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.3328, "step": 969 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.8161, "step": 970 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 2.0598, "step": 971 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.4149, "step": 972 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.8196, "step": 973 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.5632, "step": 974 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.6117, "step": 975 }, { "epoch": 0.36, "learning_rate": 2e-05, "loss": 1.4214, "step": 976 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.8822, "step": 977 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.8403, "step": 978 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.7729, "step": 979 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.8613, "step": 980 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.6065, "step": 981 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.7172, "step": 982 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.7552, "step": 983 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.3631, "step": 984 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.7286, "step": 985 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.6131, "step": 986 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 2.0016, "step": 987 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.5411, "step": 988 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.9132, "step": 989 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.8417, "step": 990 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.4933, "step": 991 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.4303, "step": 992 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.3712, "step": 993 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.8224, "step": 994 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.5663, "step": 995 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.5784, "step": 996 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.4767, "step": 997 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.6085, "step": 998 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.9144, "step": 999 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.5628, "step": 1000 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.6224, "step": 1001 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.7548, "step": 1002 }, { "epoch": 0.37, "learning_rate": 2e-05, "loss": 1.5985, "step": 1003 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.3909, "step": 1004 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.9022, "step": 1005 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.3447, "step": 1006 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 2.0048, "step": 1007 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.8514, "step": 1008 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.4989, "step": 1009 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.9548, "step": 1010 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.4986, "step": 1011 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.6737, "step": 1012 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.7765, "step": 1013 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 2.6276, "step": 1014 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.3523, "step": 1015 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.4918, "step": 1016 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.2913, "step": 1017 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.5333, "step": 1018 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.8209, "step": 1019 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.785, "step": 1020 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.5426, "step": 1021 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.7187, "step": 1022 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.3803, "step": 1023 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.8314, "step": 1024 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.6576, "step": 1025 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.9164, "step": 1026 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.4445, "step": 1027 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.3407, "step": 1028 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 1.8844, "step": 1029 }, { "epoch": 0.38, "learning_rate": 2e-05, "loss": 2.1651, "step": 1030 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 2.0773, "step": 1031 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.6243, "step": 1032 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.5412, "step": 1033 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.5811, "step": 1034 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.5684, "step": 1035 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.8248, "step": 1036 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.3055, "step": 1037 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.4955, "step": 1038 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 2.5075, "step": 1039 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 2.1492, "step": 1040 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 2.1007, "step": 1041 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.9225, "step": 1042 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 2.1127, "step": 1043 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.5687, "step": 1044 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.5819, "step": 1045 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.8287, "step": 1046 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.5231, "step": 1047 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.6052, "step": 1048 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 2.1585, "step": 1049 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.4768, "step": 1050 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.9845, "step": 1051 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.7128, "step": 1052 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.5137, "step": 1053 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.9098, "step": 1054 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.3526, "step": 1055 }, { "epoch": 0.39, "learning_rate": 2e-05, "loss": 1.7865, "step": 1056 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.3144, "step": 1057 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7511, "step": 1058 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.8867, "step": 1059 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 2.1523, "step": 1060 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7507, "step": 1061 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.5401, "step": 1062 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7681, "step": 1063 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7431, "step": 1064 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.4375, "step": 1065 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7002, "step": 1066 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.6208, "step": 1067 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7552, "step": 1068 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7801, "step": 1069 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.789, "step": 1070 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 2.3817, "step": 1071 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.6518, "step": 1072 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.5654, "step": 1073 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.6511, "step": 1074 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.7186, "step": 1075 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.4987, "step": 1076 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.6181, "step": 1077 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.6514, "step": 1078 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 2.0842, "step": 1079 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.2856, "step": 1080 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.6423, "step": 1081 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.8219, "step": 1082 }, { "epoch": 0.4, "learning_rate": 2e-05, "loss": 1.833, "step": 1083 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.5211, "step": 1084 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.9472, "step": 1085 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.9548, "step": 1086 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.5452, "step": 1087 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.8959, "step": 1088 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.3963, "step": 1089 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 2.2558, "step": 1090 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.526, "step": 1091 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.4747, "step": 1092 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.6489, "step": 1093 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 2.1231, "step": 1094 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 2.1064, "step": 1095 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.6718, "step": 1096 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.6053, "step": 1097 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.6435, "step": 1098 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.8912, "step": 1099 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.5842, "step": 1100 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.2527, "step": 1101 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.7884, "step": 1102 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.8226, "step": 1103 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.4755, "step": 1104 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.8175, "step": 1105 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 2.3868, "step": 1106 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.6238, "step": 1107 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 1.9308, "step": 1108 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 2.1983, "step": 1109 }, { "epoch": 0.41, "learning_rate": 2e-05, "loss": 2.3037, "step": 1110 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.6852, "step": 1111 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 2.2465, "step": 1112 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.4685, "step": 1113 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.8147, "step": 1114 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.6771, "step": 1115 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.3973, "step": 1116 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 2.1051, "step": 1117 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.3562, "step": 1118 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.986, "step": 1119 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.4837, "step": 1120 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.5854, "step": 1121 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.8012, "step": 1122 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.9932, "step": 1123 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.3788, "step": 1124 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.5206, "step": 1125 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.4281, "step": 1126 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.325, "step": 1127 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.501, "step": 1128 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.5319, "step": 1129 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.3917, "step": 1130 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.6118, "step": 1131 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.3952, "step": 1132 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.2959, "step": 1133 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 2.4375, "step": 1134 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.5644, "step": 1135 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.5384, "step": 1136 }, { "epoch": 0.42, "learning_rate": 2e-05, "loss": 1.4432, "step": 1137 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 2.5861, "step": 1138 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.7049, "step": 1139 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.619, "step": 1140 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.3808, "step": 1141 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.481, "step": 1142 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.9082, "step": 1143 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.3983, "step": 1144 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.6484, "step": 1145 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.7219, "step": 1146 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.2544, "step": 1147 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.9587, "step": 1148 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.4783, "step": 1149 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.9148, "step": 1150 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.6251, "step": 1151 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 3.5281, "step": 1152 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.5084, "step": 1153 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.7545, "step": 1154 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 2.3253, "step": 1155 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.6452, "step": 1156 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.8801, "step": 1157 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.663, "step": 1158 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 2.1562, "step": 1159 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.6904, "step": 1160 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.1873, "step": 1161 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.713, "step": 1162 }, { "epoch": 0.43, "learning_rate": 2e-05, "loss": 1.7697, "step": 1163 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.2705, "step": 1164 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.5913, "step": 1165 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.4611, "step": 1166 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.5884, "step": 1167 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.4207, "step": 1168 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.3926, "step": 1169 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.0447, "step": 1170 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.7219, "step": 1171 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.1341, "step": 1172 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.3137, "step": 1173 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.9762, "step": 1174 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.4525, "step": 1175 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.287, "step": 1176 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.6682, "step": 1177 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.5339, "step": 1178 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.063, "step": 1179 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.0277, "step": 1180 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 2.212, "step": 1181 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.8284, "step": 1182 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.6129, "step": 1183 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.6322, "step": 1184 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.2369, "step": 1185 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.3096, "step": 1186 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.4706, "step": 1187 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.7148, "step": 1188 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.536, "step": 1189 }, { "epoch": 0.44, "learning_rate": 2e-05, "loss": 1.7853, "step": 1190 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.3766, "step": 1191 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.6417, "step": 1192 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.3066, "step": 1193 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.8949, "step": 1194 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.7711, "step": 1195 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.5828, "step": 1196 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.4283, "step": 1197 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 2.2515, "step": 1198 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 2.0316, "step": 1199 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.2461, "step": 1200 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.4582, "step": 1201 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 2.147, "step": 1202 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.2287, "step": 1203 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.6608, "step": 1204 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.5625, "step": 1205 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.3907, "step": 1206 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.5792, "step": 1207 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.3309, "step": 1208 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.6607, "step": 1209 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.6987, "step": 1210 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.554, "step": 1211 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.5969, "step": 1212 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.2044, "step": 1213 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 2.5076, "step": 1214 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 2.076, "step": 1215 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.4814, "step": 1216 }, { "epoch": 0.45, "learning_rate": 2e-05, "loss": 1.4811, "step": 1217 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.8774, "step": 1218 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.6411, "step": 1219 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.3005, "step": 1220 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.7224, "step": 1221 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.2524, "step": 1222 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.4341, "step": 1223 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.5051, "step": 1224 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 2.0344, "step": 1225 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 2.3001, "step": 1226 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.2812, "step": 1227 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.5391, "step": 1228 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.4578, "step": 1229 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.3899, "step": 1230 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.9769, "step": 1231 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.8251, "step": 1232 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 2.7181, "step": 1233 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.6243, "step": 1234 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.5654, "step": 1235 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.6195, "step": 1236 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 2.177, "step": 1237 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.8183, "step": 1238 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.5955, "step": 1239 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.3666, "step": 1240 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.9044, "step": 1241 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.3742, "step": 1242 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.5269, "step": 1243 }, { "epoch": 0.46, "learning_rate": 2e-05, "loss": 1.4488, "step": 1244 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.3642, "step": 1245 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.9219, "step": 1246 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6817, "step": 1247 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.5066, "step": 1248 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 2.1001, "step": 1249 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6121, "step": 1250 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 2.1555, "step": 1251 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6298, "step": 1252 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.4136, "step": 1253 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6363, "step": 1254 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.7189, "step": 1255 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.3455, "step": 1256 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.4356, "step": 1257 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.3853, "step": 1258 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6706, "step": 1259 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.7491, "step": 1260 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6034, "step": 1261 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.7922, "step": 1262 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.4488, "step": 1263 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.4116, "step": 1264 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.612, "step": 1265 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6997, "step": 1266 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.5422, "step": 1267 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.8304, "step": 1268 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.6946, "step": 1269 }, { "epoch": 0.47, "learning_rate": 2e-05, "loss": 1.3655, "step": 1270 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.5128, "step": 1271 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.4819, "step": 1272 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.975, "step": 1273 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.6144, "step": 1274 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.5811, "step": 1275 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.3873, "step": 1276 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.7556, "step": 1277 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 2.154, "step": 1278 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.4096, "step": 1279 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.2516, "step": 1280 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.2246, "step": 1281 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.9521, "step": 1282 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.5806, "step": 1283 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.4936, "step": 1284 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.9058, "step": 1285 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.4478, "step": 1286 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.7734, "step": 1287 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.6738, "step": 1288 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.7372, "step": 1289 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.2466, "step": 1290 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.5913, "step": 1291 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.8675, "step": 1292 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.6808, "step": 1293 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 2.2938, "step": 1294 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.7602, "step": 1295 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.4884, "step": 1296 }, { "epoch": 0.48, "learning_rate": 2e-05, "loss": 1.5179, "step": 1297 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.6515, "step": 1298 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.8326, "step": 1299 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.6533, "step": 1300 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 2.2731, "step": 1301 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 2.0713, "step": 1302 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.2439, "step": 1303 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.6725, "step": 1304 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.6174, "step": 1305 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.3019, "step": 1306 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.2333, "step": 1307 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.5107, "step": 1308 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.1763, "step": 1309 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.3679, "step": 1310 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.2656, "step": 1311 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 2.0889, "step": 1312 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.757, "step": 1313 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.4009, "step": 1314 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.6276, "step": 1315 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.6252, "step": 1316 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.5074, "step": 1317 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.9464, "step": 1318 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.4789, "step": 1319 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.9045, "step": 1320 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.4464, "step": 1321 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.3504, "step": 1322 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 1.6699, "step": 1323 }, { "epoch": 0.49, "learning_rate": 2e-05, "loss": 2.1489, "step": 1324 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.4096, "step": 1325 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.7157, "step": 1326 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.7621, "step": 1327 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.6277, "step": 1328 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.5124, "step": 1329 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.2282, "step": 1330 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.4323, "step": 1331 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.1382, "step": 1332 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.7591, "step": 1333 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.9783, "step": 1334 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.6177, "step": 1335 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.6604, "step": 1336 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 2.0042, "step": 1337 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.6603, "step": 1338 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.2464, "step": 1339 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.5723, "step": 1340 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.4346, "step": 1341 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.762, "step": 1342 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.5215, "step": 1343 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.3502, "step": 1344 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.4455, "step": 1345 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.2855, "step": 1346 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 2.1018, "step": 1347 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.7894, "step": 1348 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.8865, "step": 1349 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.5746, "step": 1350 }, { "epoch": 0.5, "learning_rate": 2e-05, "loss": 1.8782, "step": 1351 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.6508, "step": 1352 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.4494, "step": 1353 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.5627, "step": 1354 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.5618, "step": 1355 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.5462, "step": 1356 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.8697, "step": 1357 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 2.5392, "step": 1358 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.7028, "step": 1359 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.6992, "step": 1360 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.8728, "step": 1361 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.6465, "step": 1362 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.3286, "step": 1363 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.3972, "step": 1364 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.5639, "step": 1365 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.8418, "step": 1366 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.2785, "step": 1367 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.6626, "step": 1368 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.3858, "step": 1369 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.9366, "step": 1370 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.772, "step": 1371 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.9543, "step": 1372 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.8551, "step": 1373 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.6972, "step": 1374 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.1396, "step": 1375 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.3836, "step": 1376 }, { "epoch": 0.51, "learning_rate": 2e-05, "loss": 1.4957, "step": 1377 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.7041, "step": 1378 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.5328, "step": 1379 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.5877, "step": 1380 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 2.521, "step": 1381 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.6264, "step": 1382 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.7323, "step": 1383 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.7564, "step": 1384 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.5015, "step": 1385 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.7647, "step": 1386 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 2.1381, "step": 1387 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.5462, "step": 1388 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.4164, "step": 1389 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.0759, "step": 1390 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.174, "step": 1391 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.8517, "step": 1392 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.5048, "step": 1393 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.4957, "step": 1394 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.4086, "step": 1395 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.7392, "step": 1396 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.2521, "step": 1397 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.4073, "step": 1398 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.6976, "step": 1399 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.5866, "step": 1400 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.4821, "step": 1401 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 1.8881, "step": 1402 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 2.0154, "step": 1403 }, { "epoch": 0.52, "learning_rate": 2e-05, "loss": 3.0804, "step": 1404 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.9807, "step": 1405 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.4781, "step": 1406 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.6082, "step": 1407 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 2.2579, "step": 1408 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.7002, "step": 1409 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.9062, "step": 1410 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 2.0815, "step": 1411 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.8441, "step": 1412 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.4769, "step": 1413 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.3499, "step": 1414 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.3172, "step": 1415 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.521, "step": 1416 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.3929, "step": 1417 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.6244, "step": 1418 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.4707, "step": 1419 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.5086, "step": 1420 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 2.2695, "step": 1421 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 2.3855, "step": 1422 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.7099, "step": 1423 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.8627, "step": 1424 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.3532, "step": 1425 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.6408, "step": 1426 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.9457, "step": 1427 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 2.4009, "step": 1428 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.9341, "step": 1429 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.3105, "step": 1430 }, { "epoch": 0.53, "learning_rate": 2e-05, "loss": 1.7554, "step": 1431 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.8778, "step": 1432 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 2.1513, "step": 1433 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.5177, "step": 1434 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.6101, "step": 1435 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.5057, "step": 1436 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.4885, "step": 1437 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.9696, "step": 1438 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.8649, "step": 1439 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.7304, "step": 1440 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.8764, "step": 1441 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.3251, "step": 1442 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.643, "step": 1443 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.5543, "step": 1444 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.8389, "step": 1445 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.4516, "step": 1446 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.2422, "step": 1447 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.5845, "step": 1448 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.693, "step": 1449 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.6636, "step": 1450 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.9641, "step": 1451 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.4092, "step": 1452 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.2475, "step": 1453 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.985, "step": 1454 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.3484, "step": 1455 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.7386, "step": 1456 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.4029, "step": 1457 }, { "epoch": 0.54, "learning_rate": 2e-05, "loss": 1.8005, "step": 1458 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.6085, "step": 1459 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.5939, "step": 1460 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.5529, "step": 1461 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.6431, "step": 1462 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.5442, "step": 1463 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.9348, "step": 1464 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 2.2976, "step": 1465 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.7594, "step": 1466 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.807, "step": 1467 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 2.4882, "step": 1468 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.4974, "step": 1469 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.3605, "step": 1470 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.9907, "step": 1471 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 3.5683, "step": 1472 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 2.2847, "step": 1473 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.8611, "step": 1474 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.2845, "step": 1475 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.3899, "step": 1476 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.5466, "step": 1477 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.635, "step": 1478 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.4238, "step": 1479 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 2.0741, "step": 1480 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.3414, "step": 1481 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.5041, "step": 1482 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.4057, "step": 1483 }, { "epoch": 0.55, "learning_rate": 2e-05, "loss": 1.532, "step": 1484 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.0304, "step": 1485 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.5456, "step": 1486 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 2.4082, "step": 1487 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.461, "step": 1488 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.5073, "step": 1489 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.8462, "step": 1490 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.7986, "step": 1491 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.1798, "step": 1492 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 2.2073, "step": 1493 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.5786, "step": 1494 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.6986, "step": 1495 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 2.0453, "step": 1496 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 2.0611, "step": 1497 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.3981, "step": 1498 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.9551, "step": 1499 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.6657, "step": 1500 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.2354, "step": 1501 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.3096, "step": 1502 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.4959, "step": 1503 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.7338, "step": 1504 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.6965, "step": 1505 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.301, "step": 1506 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 2.3316, "step": 1507 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.7328, "step": 1508 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.4845, "step": 1509 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.7463, "step": 1510 }, { "epoch": 0.56, "learning_rate": 2e-05, "loss": 1.9308, "step": 1511 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.2811, "step": 1512 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.5879, "step": 1513 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.7588, "step": 1514 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.5479, "step": 1515 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.8111, "step": 1516 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 2.1241, "step": 1517 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.9139, "step": 1518 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.658, "step": 1519 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.2222, "step": 1520 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.8014, "step": 1521 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.7731, "step": 1522 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.284, "step": 1523 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.1379, "step": 1524 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.5034, "step": 1525 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.3167, "step": 1526 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.9443, "step": 1527 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.319, "step": 1528 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.8064, "step": 1529 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.2622, "step": 1530 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.3214, "step": 1531 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.8468, "step": 1532 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.8747, "step": 1533 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.2946, "step": 1534 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.2916, "step": 1535 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 2.3027, "step": 1536 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.406, "step": 1537 }, { "epoch": 0.57, "learning_rate": 2e-05, "loss": 1.5862, "step": 1538 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 2.5164, "step": 1539 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.3232, "step": 1540 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 2.8432, "step": 1541 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 2.6278, "step": 1542 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.4512, "step": 1543 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.4191, "step": 1544 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.375, "step": 1545 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.724, "step": 1546 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.4949, "step": 1547 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 2.3568, "step": 1548 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.8953, "step": 1549 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 2.1226, "step": 1550 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.3594, "step": 1551 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.935, "step": 1552 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.3952, "step": 1553 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.8905, "step": 1554 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.7077, "step": 1555 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.4455, "step": 1556 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 2.0322, "step": 1557 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.7846, "step": 1558 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.5047, "step": 1559 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.2329, "step": 1560 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.646, "step": 1561 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.3324, "step": 1562 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.7138, "step": 1563 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.5764, "step": 1564 }, { "epoch": 0.58, "learning_rate": 2e-05, "loss": 1.4965, "step": 1565 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 2.1971, "step": 1566 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.7019, "step": 1567 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.7506, "step": 1568 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 2.1051, "step": 1569 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.2778, "step": 1570 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.4771, "step": 1571 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 2.0408, "step": 1572 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.8428, "step": 1573 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.3844, "step": 1574 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.5371, "step": 1575 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.441, "step": 1576 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.4278, "step": 1577 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.3442, "step": 1578 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 2.2315, "step": 1579 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 3.2346, "step": 1580 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.8498, "step": 1581 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.989, "step": 1582 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.6041, "step": 1583 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.6245, "step": 1584 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.4035, "step": 1585 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.8192, "step": 1586 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.4926, "step": 1587 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.5149, "step": 1588 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 2.5341, "step": 1589 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.5579, "step": 1590 }, { "epoch": 0.59, "learning_rate": 2e-05, "loss": 1.7486, "step": 1591 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.5869, "step": 1592 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.4454, "step": 1593 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.3763, "step": 1594 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 2.0731, "step": 1595 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.2005, "step": 1596 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.7764, "step": 1597 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.7704, "step": 1598 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.5047, "step": 1599 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.5435, "step": 1600 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.6287, "step": 1601 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.702, "step": 1602 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.6765, "step": 1603 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.5232, "step": 1604 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 2.3438, "step": 1605 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.5044, "step": 1606 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.5632, "step": 1607 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.694, "step": 1608 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 2.1137, "step": 1609 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.3448, "step": 1610 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.3013, "step": 1611 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.7928, "step": 1612 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.6214, "step": 1613 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.3397, "step": 1614 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 2.0504, "step": 1615 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.5043, "step": 1616 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.4385, "step": 1617 }, { "epoch": 0.6, "learning_rate": 2e-05, "loss": 1.4779, "step": 1618 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 2.1981, "step": 1619 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.7674, "step": 1620 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.4841, "step": 1621 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.5356, "step": 1622 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.2819, "step": 1623 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.6767, "step": 1624 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.273, "step": 1625 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.5415, "step": 1626 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.4903, "step": 1627 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.8463, "step": 1628 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.7454, "step": 1629 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 2.3922, "step": 1630 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.4613, "step": 1631 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.7089, "step": 1632 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.2885, "step": 1633 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.7682, "step": 1634 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.4302, "step": 1635 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.4948, "step": 1636 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.8053, "step": 1637 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.4523, "step": 1638 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.3922, "step": 1639 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.6256, "step": 1640 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.5917, "step": 1641 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.5517, "step": 1642 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.8389, "step": 1643 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.5762, "step": 1644 }, { "epoch": 0.61, "learning_rate": 2e-05, "loss": 1.9453, "step": 1645 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.3534, "step": 1646 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.6139, "step": 1647 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.5672, "step": 1648 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4443, "step": 1649 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.3865, "step": 1650 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4541, "step": 1651 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.5204, "step": 1652 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.7784, "step": 1653 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 2.0472, "step": 1654 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4897, "step": 1655 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.7166, "step": 1656 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 2.5525, "step": 1657 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.2719, "step": 1658 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.8147, "step": 1659 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.7133, "step": 1660 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.3873, "step": 1661 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.6597, "step": 1662 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4029, "step": 1663 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4471, "step": 1664 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.5607, "step": 1665 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4138, "step": 1666 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.6275, "step": 1667 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 2.2077, "step": 1668 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4158, "step": 1669 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.4082, "step": 1670 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 1.2576, "step": 1671 }, { "epoch": 0.62, "learning_rate": 2e-05, "loss": 2.0042, "step": 1672 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.2487, "step": 1673 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.291, "step": 1674 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.3492, "step": 1675 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.5192, "step": 1676 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.5343, "step": 1677 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.7237, "step": 1678 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.5759, "step": 1679 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.7773, "step": 1680 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.6371, "step": 1681 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 2.2368, "step": 1682 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.3534, "step": 1683 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.7318, "step": 1684 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.4745, "step": 1685 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.5537, "step": 1686 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.2365, "step": 1687 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 2.4055, "step": 1688 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.3305, "step": 1689 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.4608, "step": 1690 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.3042, "step": 1691 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 2.0235, "step": 1692 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.6772, "step": 1693 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.4948, "step": 1694 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.368, "step": 1695 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.8558, "step": 1696 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.3672, "step": 1697 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.2965, "step": 1698 }, { "epoch": 0.63, "learning_rate": 2e-05, "loss": 1.7203, "step": 1699 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.508, "step": 1700 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.6095, "step": 1701 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.3312, "step": 1702 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.3873, "step": 1703 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.5792, "step": 1704 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.3771, "step": 1705 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 2.0615, "step": 1706 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.2262, "step": 1707 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.6734, "step": 1708 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.9199, "step": 1709 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.6544, "step": 1710 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.6199, "step": 1711 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.5559, "step": 1712 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.593, "step": 1713 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.2848, "step": 1714 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 2.0997, "step": 1715 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.5125, "step": 1716 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 2.064, "step": 1717 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.9526, "step": 1718 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.2652, "step": 1719 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.1626, "step": 1720 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.426, "step": 1721 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.7025, "step": 1722 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.4053, "step": 1723 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.7592, "step": 1724 }, { "epoch": 0.64, "learning_rate": 2e-05, "loss": 1.8363, "step": 1725 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.7837, "step": 1726 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.2489, "step": 1727 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.1034, "step": 1728 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 2.0644, "step": 1729 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.335, "step": 1730 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.3808, "step": 1731 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.7763, "step": 1732 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.61, "step": 1733 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.9062, "step": 1734 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.7708, "step": 1735 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.6392, "step": 1736 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 2.118, "step": 1737 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.1808, "step": 1738 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.8761, "step": 1739 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.3459, "step": 1740 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.6554, "step": 1741 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.5146, "step": 1742 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.7459, "step": 1743 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.4962, "step": 1744 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.6026, "step": 1745 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.4348, "step": 1746 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.4407, "step": 1747 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.1687, "step": 1748 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.4635, "step": 1749 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.9627, "step": 1750 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.3668, "step": 1751 }, { "epoch": 0.65, "learning_rate": 2e-05, "loss": 1.9615, "step": 1752 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 2.1418, "step": 1753 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.5399, "step": 1754 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.8025, "step": 1755 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.6605, "step": 1756 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.7631, "step": 1757 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.6446, "step": 1758 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.8257, "step": 1759 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.1837, "step": 1760 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.5648, "step": 1761 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 2.1357, "step": 1762 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 2.2222, "step": 1763 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.4664, "step": 1764 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.5159, "step": 1765 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.6815, "step": 1766 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.6131, "step": 1767 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.3215, "step": 1768 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.6084, "step": 1769 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 2.0539, "step": 1770 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.8226, "step": 1771 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.378, "step": 1772 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.6084, "step": 1773 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 2.4558, "step": 1774 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.7188, "step": 1775 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.4962, "step": 1776 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.7711, "step": 1777 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.2223, "step": 1778 }, { "epoch": 0.66, "learning_rate": 2e-05, "loss": 1.9612, "step": 1779 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.5893, "step": 1780 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 2.1346, "step": 1781 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.1148, "step": 1782 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.3838, "step": 1783 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 2.2257, "step": 1784 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.064, "step": 1785 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.3792, "step": 1786 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.4059, "step": 1787 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.6211, "step": 1788 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.3741, "step": 1789 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.5335, "step": 1790 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.8719, "step": 1791 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.4019, "step": 1792 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.372, "step": 1793 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.8528, "step": 1794 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.385, "step": 1795 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.758, "step": 1796 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.4455, "step": 1797 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.4226, "step": 1798 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.8066, "step": 1799 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.4942, "step": 1800 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 2.2703, "step": 1801 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.7017, "step": 1802 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.6767, "step": 1803 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.8597, "step": 1804 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.3742, "step": 1805 }, { "epoch": 0.67, "learning_rate": 2e-05, "loss": 1.5126, "step": 1806 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.5907, "step": 1807 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.6824, "step": 1808 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.8431, "step": 1809 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.9725, "step": 1810 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.6167, "step": 1811 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.182, "step": 1812 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.7915, "step": 1813 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.7932, "step": 1814 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.9054, "step": 1815 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.6523, "step": 1816 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.5709, "step": 1817 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.2649, "step": 1818 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.503, "step": 1819 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 2.7035, "step": 1820 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.4858, "step": 1821 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.598, "step": 1822 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.8645, "step": 1823 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.4002, "step": 1824 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.3165, "step": 1825 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.489, "step": 1826 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.3111, "step": 1827 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.5043, "step": 1828 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.3269, "step": 1829 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 2.0715, "step": 1830 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.7515, "step": 1831 }, { "epoch": 0.68, "learning_rate": 2e-05, "loss": 1.311, "step": 1832 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.4404, "step": 1833 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.1834, "step": 1834 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.8403, "step": 1835 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.3106, "step": 1836 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.4601, "step": 1837 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.63, "step": 1838 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.7299, "step": 1839 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.546, "step": 1840 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.3781, "step": 1841 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.7748, "step": 1842 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.5022, "step": 1843 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.6502, "step": 1844 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.4791, "step": 1845 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.6426, "step": 1846 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.4462, "step": 1847 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.4219, "step": 1848 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.8584, "step": 1849 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.6554, "step": 1850 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.4207, "step": 1851 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.3353, "step": 1852 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.675, "step": 1853 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 2.0304, "step": 1854 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.8295, "step": 1855 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.2463, "step": 1856 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.2747, "step": 1857 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 2.1985, "step": 1858 }, { "epoch": 0.69, "learning_rate": 2e-05, "loss": 1.5876, "step": 1859 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.6492, "step": 1860 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.487, "step": 1861 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 2.8392, "step": 1862 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.6453, "step": 1863 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.4141, "step": 1864 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.4949, "step": 1865 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.499, "step": 1866 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.5926, "step": 1867 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.6454, "step": 1868 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.4592, "step": 1869 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.4915, "step": 1870 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 2.5463, "step": 1871 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.5318, "step": 1872 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 2.0111, "step": 1873 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.629, "step": 1874 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.8634, "step": 1875 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 2.2822, "step": 1876 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 2.0064, "step": 1877 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.6262, "step": 1878 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.3689, "step": 1879 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.3876, "step": 1880 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.5219, "step": 1881 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.633, "step": 1882 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.5024, "step": 1883 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 2.4085, "step": 1884 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.2259, "step": 1885 }, { "epoch": 0.7, "learning_rate": 2e-05, "loss": 1.7164, "step": 1886 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 2.3288, "step": 1887 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.3667, "step": 1888 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.5494, "step": 1889 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.4754, "step": 1890 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.1629, "step": 1891 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.3396, "step": 1892 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.5773, "step": 1893 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.8149, "step": 1894 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.8945, "step": 1895 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.2343, "step": 1896 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.6059, "step": 1897 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.5479, "step": 1898 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.5699, "step": 1899 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 2.8699, "step": 1900 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.5889, "step": 1901 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.6493, "step": 1902 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.3617, "step": 1903 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.6215, "step": 1904 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.7111, "step": 1905 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.1727, "step": 1906 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.4261, "step": 1907 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.3931, "step": 1908 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.4983, "step": 1909 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.1697, "step": 1910 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.3803, "step": 1911 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 1.4312, "step": 1912 }, { "epoch": 0.71, "learning_rate": 2e-05, "loss": 2.271, "step": 1913 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.5345, "step": 1914 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.3433, "step": 1915 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.6443, "step": 1916 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.6958, "step": 1917 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.8254, "step": 1918 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.4377, "step": 1919 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.2447, "step": 1920 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.262, "step": 1921 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.4247, "step": 1922 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.481, "step": 1923 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.5775, "step": 1924 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.8027, "step": 1925 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.3967, "step": 1926 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.7209, "step": 1927 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 2.3709, "step": 1928 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.8767, "step": 1929 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.3569, "step": 1930 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.9034, "step": 1931 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 2.4483, "step": 1932 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.4203, "step": 1933 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.504, "step": 1934 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.4528, "step": 1935 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 2.1222, "step": 1936 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.6427, "step": 1937 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.2897, "step": 1938 }, { "epoch": 0.72, "learning_rate": 2e-05, "loss": 1.5023, "step": 1939 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.6469, "step": 1940 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.3215, "step": 1941 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.627, "step": 1942 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.5213, "step": 1943 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.364, "step": 1944 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.3859, "step": 1945 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.883, "step": 1946 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.4225, "step": 1947 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.4581, "step": 1948 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.3981, "step": 1949 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.3111, "step": 1950 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.874, "step": 1951 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.3497, "step": 1952 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.8436, "step": 1953 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.4347, "step": 1954 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.6195, "step": 1955 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.4121, "step": 1956 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.1959, "step": 1957 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.4804, "step": 1958 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.8263, "step": 1959 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.5406, "step": 1960 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.4781, "step": 1961 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.2565, "step": 1962 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.5166, "step": 1963 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.6183, "step": 1964 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.2777, "step": 1965 }, { "epoch": 0.73, "learning_rate": 2e-05, "loss": 1.4089, "step": 1966 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.4843, "step": 1967 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.0954, "step": 1968 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.2762, "step": 1969 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.3551, "step": 1970 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.3578, "step": 1971 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.4362, "step": 1972 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.6086, "step": 1973 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.915, "step": 1974 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.9253, "step": 1975 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.9836, "step": 1976 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.6088, "step": 1977 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.2113, "step": 1978 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.1659, "step": 1979 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.2502, "step": 1980 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 2.1074, "step": 1981 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.5561, "step": 1982 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.3793, "step": 1983 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.4353, "step": 1984 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.222, "step": 1985 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.503, "step": 1986 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.8319, "step": 1987 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.2451, "step": 1988 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.3586, "step": 1989 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 2.4399, "step": 1990 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.3833, "step": 1991 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.2816, "step": 1992 }, { "epoch": 0.74, "learning_rate": 2e-05, "loss": 1.9096, "step": 1993 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.6632, "step": 1994 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.3855, "step": 1995 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.1952, "step": 1996 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.5532, "step": 1997 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.9701, "step": 1998 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.4493, "step": 1999 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.4577, "step": 2000 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.2417, "step": 2001 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.1846, "step": 2002 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.3381, "step": 2003 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.2412, "step": 2004 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.374, "step": 2005 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.505, "step": 2006 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.805, "step": 2007 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.3047, "step": 2008 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 2.3218, "step": 2009 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.7604, "step": 2010 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 2.0234, "step": 2011 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.7774, "step": 2012 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.7681, "step": 2013 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.9197, "step": 2014 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.3265, "step": 2015 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.6637, "step": 2016 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.4493, "step": 2017 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.745, "step": 2018 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.6845, "step": 2019 }, { "epoch": 0.75, "learning_rate": 2e-05, "loss": 1.8258, "step": 2020 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.7147, "step": 2021 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.2545, "step": 2022 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.5002, "step": 2023 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.4363, "step": 2024 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.9667, "step": 2025 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.4917, "step": 2026 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 2.187, "step": 2027 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.396, "step": 2028 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.8992, "step": 2029 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.4183, "step": 2030 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.8049, "step": 2031 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.5636, "step": 2032 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.3711, "step": 2033 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.6735, "step": 2034 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.7509, "step": 2035 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.4927, "step": 2036 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.115, "step": 2037 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.3487, "step": 2038 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.7006, "step": 2039 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.6565, "step": 2040 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.1108, "step": 2041 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.8309, "step": 2042 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.9344, "step": 2043 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.3466, "step": 2044 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.5072, "step": 2045 }, { "epoch": 0.76, "learning_rate": 2e-05, "loss": 1.7308, "step": 2046 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.8974, "step": 2047 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.1996, "step": 2048 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.2269, "step": 2049 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.3121, "step": 2050 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.2618, "step": 2051 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.5267, "step": 2052 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.0887, "step": 2053 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 2.0209, "step": 2054 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.3651, "step": 2055 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.5421, "step": 2056 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.9234, "step": 2057 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.5979, "step": 2058 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.3575, "step": 2059 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.3713, "step": 2060 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 2.1987, "step": 2061 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.7071, "step": 2062 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.6262, "step": 2063 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 2.2332, "step": 2064 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.2566, "step": 2065 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.2673, "step": 2066 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.1479, "step": 2067 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.7207, "step": 2068 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.4574, "step": 2069 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.39, "step": 2070 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 2.3245, "step": 2071 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.3314, "step": 2072 }, { "epoch": 0.77, "learning_rate": 2e-05, "loss": 1.8017, "step": 2073 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.5438, "step": 2074 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 2.7061, "step": 2075 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.8479, "step": 2076 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.6302, "step": 2077 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.5848, "step": 2078 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.7522, "step": 2079 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.6169, "step": 2080 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 2.0056, "step": 2081 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.4178, "step": 2082 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.6564, "step": 2083 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 2.3584, "step": 2084 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.4668, "step": 2085 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.4358, "step": 2086 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.5713, "step": 2087 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.4004, "step": 2088 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.8102, "step": 2089 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 2.066, "step": 2090 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.6083, "step": 2091 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.2107, "step": 2092 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.6313, "step": 2093 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.456, "step": 2094 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.1211, "step": 2095 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.3077, "step": 2096 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.1781, "step": 2097 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.5113, "step": 2098 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 2.2043, "step": 2099 }, { "epoch": 0.78, "learning_rate": 2e-05, "loss": 1.7882, "step": 2100 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.4337, "step": 2101 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.2149, "step": 2102 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.2136, "step": 2103 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.3573, "step": 2104 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.4377, "step": 2105 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.9374, "step": 2106 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 2.3217, "step": 2107 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.8791, "step": 2108 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.5942, "step": 2109 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.241, "step": 2110 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.343, "step": 2111 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.3503, "step": 2112 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.4459, "step": 2113 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 2.3942, "step": 2114 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.3881, "step": 2115 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 2.3094, "step": 2116 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.4745, "step": 2117 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.7948, "step": 2118 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 2.0309, "step": 2119 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.298, "step": 2120 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.49, "step": 2121 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.8878, "step": 2122 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.106, "step": 2123 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.189, "step": 2124 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.3104, "step": 2125 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.1912, "step": 2126 }, { "epoch": 0.79, "learning_rate": 2e-05, "loss": 1.2646, "step": 2127 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.1596, "step": 2128 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.321, "step": 2129 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.7398, "step": 2130 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.2315, "step": 2131 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 2.1291, "step": 2132 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.3949, "step": 2133 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.8328, "step": 2134 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.4929, "step": 2135 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.5054, "step": 2136 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.5833, "step": 2137 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.4222, "step": 2138 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.9537, "step": 2139 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.8338, "step": 2140 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.1937, "step": 2141 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.6106, "step": 2142 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.7817, "step": 2143 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.5023, "step": 2144 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.2402, "step": 2145 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.7466, "step": 2146 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.5621, "step": 2147 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.6348, "step": 2148 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.2946, "step": 2149 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.3914, "step": 2150 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.8733, "step": 2151 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.3673, "step": 2152 }, { "epoch": 0.8, "learning_rate": 2e-05, "loss": 1.6595, "step": 2153 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.418, "step": 2154 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.9029, "step": 2155 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 3.183, "step": 2156 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.5504, "step": 2157 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.5337, "step": 2158 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.3009, "step": 2159 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.4767, "step": 2160 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.4729, "step": 2161 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.599, "step": 2162 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.3884, "step": 2163 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.2051, "step": 2164 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.4376, "step": 2165 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.4875, "step": 2166 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 2.2224, "step": 2167 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.2378, "step": 2168 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.8081, "step": 2169 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.6299, "step": 2170 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.5677, "step": 2171 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.1811, "step": 2172 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.4586, "step": 2173 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.305, "step": 2174 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.7856, "step": 2175 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.5851, "step": 2176 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.609, "step": 2177 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.7514, "step": 2178 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.7656, "step": 2179 }, { "epoch": 0.81, "learning_rate": 2e-05, "loss": 1.688, "step": 2180 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.7798, "step": 2181 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.7866, "step": 2182 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.4963, "step": 2183 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.4198, "step": 2184 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.4633, "step": 2185 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.4675, "step": 2186 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.4825, "step": 2187 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.6424, "step": 2188 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.6173, "step": 2189 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.2976, "step": 2190 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.8499, "step": 2191 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.7301, "step": 2192 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.5439, "step": 2193 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.4085, "step": 2194 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.4019, "step": 2195 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 2.145, "step": 2196 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.5864, "step": 2197 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 3.0551, "step": 2198 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.3097, "step": 2199 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 2.1576, "step": 2200 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.6877, "step": 2201 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.7323, "step": 2202 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.3082, "step": 2203 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.346, "step": 2204 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.6644, "step": 2205 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.6098, "step": 2206 }, { "epoch": 0.82, "learning_rate": 2e-05, "loss": 1.6244, "step": 2207 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.4762, "step": 2208 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.8802, "step": 2209 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.801, "step": 2210 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.548, "step": 2211 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.3714, "step": 2212 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.1777, "step": 2213 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.2967, "step": 2214 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.4406, "step": 2215 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.8765, "step": 2216 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.4515, "step": 2217 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.3733, "step": 2218 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.3557, "step": 2219 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.7863, "step": 2220 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.9425, "step": 2221 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.4841, "step": 2222 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.9595, "step": 2223 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 2.083, "step": 2224 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.4573, "step": 2225 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.5167, "step": 2226 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.7274, "step": 2227 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 2.1654, "step": 2228 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.2797, "step": 2229 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.6335, "step": 2230 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.5768, "step": 2231 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.7959, "step": 2232 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.6903, "step": 2233 }, { "epoch": 0.83, "learning_rate": 2e-05, "loss": 1.7664, "step": 2234 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.2526, "step": 2235 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.2276, "step": 2236 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.7148, "step": 2237 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 2.0154, "step": 2238 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.3584, "step": 2239 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.416, "step": 2240 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.2641, "step": 2241 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.1415, "step": 2242 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.5178, "step": 2243 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 2.039, "step": 2244 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.5542, "step": 2245 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.4198, "step": 2246 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.9184, "step": 2247 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.4777, "step": 2248 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.3824, "step": 2249 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.5584, "step": 2250 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.6927, "step": 2251 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.3671, "step": 2252 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 2.1204, "step": 2253 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.5057, "step": 2254 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.1794, "step": 2255 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.7581, "step": 2256 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.9668, "step": 2257 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.8053, "step": 2258 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.3461, "step": 2259 }, { "epoch": 0.84, "learning_rate": 2e-05, "loss": 1.6504, "step": 2260 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.3744, "step": 2261 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.7908, "step": 2262 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.4163, "step": 2263 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.6212, "step": 2264 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.5147, "step": 2265 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.2556, "step": 2266 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.8321, "step": 2267 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.4733, "step": 2268 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.3824, "step": 2269 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.4187, "step": 2270 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.4075, "step": 2271 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.6744, "step": 2272 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.2255, "step": 2273 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.5499, "step": 2274 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.9502, "step": 2275 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 2.4821, "step": 2276 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.6168, "step": 2277 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.568, "step": 2278 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.3036, "step": 2279 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.7411, "step": 2280 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.4559, "step": 2281 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.4669, "step": 2282 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.8398, "step": 2283 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.1492, "step": 2284 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.377, "step": 2285 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 1.6339, "step": 2286 }, { "epoch": 0.85, "learning_rate": 2e-05, "loss": 2.3689, "step": 2287 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.4694, "step": 2288 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.2938, "step": 2289 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.7789, "step": 2290 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.7353, "step": 2291 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.8859, "step": 2292 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.4439, "step": 2293 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.5697, "step": 2294 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.6987, "step": 2295 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.7926, "step": 2296 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.839, "step": 2297 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.5176, "step": 2298 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.8216, "step": 2299 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.2925, "step": 2300 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.5585, "step": 2301 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 2.3262, "step": 2302 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.159, "step": 2303 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.4533, "step": 2304 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.3607, "step": 2305 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.3489, "step": 2306 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 2.0503, "step": 2307 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.4335, "step": 2308 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.6702, "step": 2309 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.318, "step": 2310 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.383, "step": 2311 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.3672, "step": 2312 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.7073, "step": 2313 }, { "epoch": 0.86, "learning_rate": 2e-05, "loss": 1.856, "step": 2314 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.3213, "step": 2315 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.4675, "step": 2316 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.1682, "step": 2317 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 2.0739, "step": 2318 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.5309, "step": 2319 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.4758, "step": 2320 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.9415, "step": 2321 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.8585, "step": 2322 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.7869, "step": 2323 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.4483, "step": 2324 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.9181, "step": 2325 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.186, "step": 2326 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.6602, "step": 2327 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.6901, "step": 2328 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.5709, "step": 2329 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.477, "step": 2330 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.7383, "step": 2331 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.7962, "step": 2332 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.5659, "step": 2333 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.3583, "step": 2334 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.1308, "step": 2335 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.5665, "step": 2336 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.8323, "step": 2337 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 2.0134, "step": 2338 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.5842, "step": 2339 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.2709, "step": 2340 }, { "epoch": 0.87, "learning_rate": 2e-05, "loss": 1.8723, "step": 2341 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.9387, "step": 2342 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.342, "step": 2343 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.6999, "step": 2344 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.4421, "step": 2345 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.6898, "step": 2346 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.2247, "step": 2347 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.542, "step": 2348 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.2356, "step": 2349 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.2116, "step": 2350 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.5147, "step": 2351 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.2345, "step": 2352 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.668, "step": 2353 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.3817, "step": 2354 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.6282, "step": 2355 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.8462, "step": 2356 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 2.1455, "step": 2357 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.4427, "step": 2358 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.3702, "step": 2359 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.0751, "step": 2360 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.4693, "step": 2361 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 2.3856, "step": 2362 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.7808, "step": 2363 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.4586, "step": 2364 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.7519, "step": 2365 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 2.1742, "step": 2366 }, { "epoch": 0.88, "learning_rate": 2e-05, "loss": 1.3783, "step": 2367 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 2.8484, "step": 2368 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.5953, "step": 2369 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.571, "step": 2370 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.6644, "step": 2371 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 2.5511, "step": 2372 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.5644, "step": 2373 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.3993, "step": 2374 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.775, "step": 2375 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.2206, "step": 2376 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.8217, "step": 2377 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 2.1104, "step": 2378 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.3068, "step": 2379 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.6765, "step": 2380 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.669, "step": 2381 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.9447, "step": 2382 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 2.2206, "step": 2383 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.9217, "step": 2384 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.4393, "step": 2385 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.7475, "step": 2386 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.4357, "step": 2387 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.9629, "step": 2388 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.3201, "step": 2389 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.8236, "step": 2390 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.2141, "step": 2391 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.5291, "step": 2392 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.8764, "step": 2393 }, { "epoch": 0.89, "learning_rate": 2e-05, "loss": 1.3182, "step": 2394 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.7401, "step": 2395 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.3135, "step": 2396 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.7902, "step": 2397 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.7237, "step": 2398 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.1723, "step": 2399 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.2595, "step": 2400 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 2.0641, "step": 2401 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.286, "step": 2402 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.5638, "step": 2403 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.7432, "step": 2404 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.3601, "step": 2405 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.4836, "step": 2406 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.2276, "step": 2407 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.6386, "step": 2408 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.2818, "step": 2409 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.6697, "step": 2410 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.2644, "step": 2411 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.6838, "step": 2412 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.8646, "step": 2413 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.4597, "step": 2414 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.7644, "step": 2415 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.5379, "step": 2416 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.6939, "step": 2417 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.6024, "step": 2418 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.713, "step": 2419 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.5282, "step": 2420 }, { "epoch": 0.9, "learning_rate": 2e-05, "loss": 1.7409, "step": 2421 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.3806, "step": 2422 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 2.1162, "step": 2423 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 2.3082, "step": 2424 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 2.2003, "step": 2425 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.6006, "step": 2426 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.2593, "step": 2427 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.776, "step": 2428 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.3255, "step": 2429 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.3188, "step": 2430 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.5873, "step": 2431 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.5529, "step": 2432 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.7953, "step": 2433 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.451, "step": 2434 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.3724, "step": 2435 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.3948, "step": 2436 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.8993, "step": 2437 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.8255, "step": 2438 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.5792, "step": 2439 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.3885, "step": 2440 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.585, "step": 2441 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.258, "step": 2442 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.9237, "step": 2443 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.4129, "step": 2444 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.079, "step": 2445 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 2.1225, "step": 2446 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.0873, "step": 2447 }, { "epoch": 0.91, "learning_rate": 2e-05, "loss": 1.5572, "step": 2448 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.4045, "step": 2449 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.3412, "step": 2450 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.1838, "step": 2451 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.4954, "step": 2452 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.4201, "step": 2453 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.5697, "step": 2454 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.2173, "step": 2455 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.0303, "step": 2456 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.7773, "step": 2457 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.2004, "step": 2458 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.2709, "step": 2459 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.5086, "step": 2460 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.7086, "step": 2461 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 2.0464, "step": 2462 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.325, "step": 2463 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.1554, "step": 2464 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 2.2234, "step": 2465 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.4564, "step": 2466 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.2627, "step": 2467 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 2.2406, "step": 2468 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.3149, "step": 2469 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 2.3119, "step": 2470 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.2651, "step": 2471 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.1988, "step": 2472 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.3555, "step": 2473 }, { "epoch": 0.92, "learning_rate": 2e-05, "loss": 1.9766, "step": 2474 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 2.0532, "step": 2475 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3453, "step": 2476 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3258, "step": 2477 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3988, "step": 2478 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3557, "step": 2479 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.2595, "step": 2480 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.5993, "step": 2481 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 2.0498, "step": 2482 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3872, "step": 2483 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.549, "step": 2484 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3871, "step": 2485 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.0607, "step": 2486 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.4682, "step": 2487 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.407, "step": 2488 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3828, "step": 2489 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.5485, "step": 2490 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.2385, "step": 2491 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.4199, "step": 2492 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.6124, "step": 2493 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3663, "step": 2494 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.2359, "step": 2495 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.597, "step": 2496 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.2575, "step": 2497 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 2.65, "step": 2498 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.3635, "step": 2499 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 1.5087, "step": 2500 }, { "epoch": 0.93, "learning_rate": 2e-05, "loss": 2.7118, "step": 2501 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.9555, "step": 2502 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.5927, "step": 2503 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.6219, "step": 2504 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.5601, "step": 2505 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.4474, "step": 2506 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.3417, "step": 2507 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.2486, "step": 2508 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.6534, "step": 2509 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.436, "step": 2510 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.3497, "step": 2511 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.6996, "step": 2512 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.2151, "step": 2513 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.7666, "step": 2514 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.6013, "step": 2515 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.2708, "step": 2516 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.7834, "step": 2517 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.306, "step": 2518 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.2883, "step": 2519 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.3835, "step": 2520 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.4207, "step": 2521 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.2731, "step": 2522 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 2.1373, "step": 2523 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.1806, "step": 2524 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.4144, "step": 2525 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.699, "step": 2526 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 2.0122, "step": 2527 }, { "epoch": 0.94, "learning_rate": 2e-05, "loss": 1.7157, "step": 2528 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.1691, "step": 2529 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.2002, "step": 2530 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.9996, "step": 2531 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.1598, "step": 2532 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 2.3683, "step": 2533 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.7042, "step": 2534 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.515, "step": 2535 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.0549, "step": 2536 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.8395, "step": 2537 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.8226, "step": 2538 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.4508, "step": 2539 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.718, "step": 2540 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.2089, "step": 2541 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.865, "step": 2542 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.8283, "step": 2543 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.5611, "step": 2544 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.7024, "step": 2545 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.5123, "step": 2546 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.4754, "step": 2547 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.1212, "step": 2548 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.6419, "step": 2549 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.0555, "step": 2550 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 3.5907, "step": 2551 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.2743, "step": 2552 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.4126, "step": 2553 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.9177, "step": 2554 }, { "epoch": 0.95, "learning_rate": 2e-05, "loss": 1.5119, "step": 2555 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.4339, "step": 2556 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.7811, "step": 2557 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.5351, "step": 2558 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.1555, "step": 2559 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.4153, "step": 2560 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.469, "step": 2561 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.4258, "step": 2562 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.2532, "step": 2563 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.1288, "step": 2564 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.4024, "step": 2565 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.8174, "step": 2566 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.3802, "step": 2567 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.619, "step": 2568 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.5091, "step": 2569 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.8934, "step": 2570 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.1032, "step": 2571 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.3129, "step": 2572 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.1548, "step": 2573 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.2779, "step": 2574 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 2.2639, "step": 2575 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.3374, "step": 2576 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.8931, "step": 2577 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.3795, "step": 2578 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.4363, "step": 2579 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 2.2632, "step": 2580 }, { "epoch": 0.96, "learning_rate": 2e-05, "loss": 1.6221, "step": 2581 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.8264, "step": 2582 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.7944, "step": 2583 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.8189, "step": 2584 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.166, "step": 2585 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.4291, "step": 2586 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.3413, "step": 2587 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.3627, "step": 2588 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 2.3757, "step": 2589 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.2388, "step": 2590 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.4553, "step": 2591 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.281, "step": 2592 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.601, "step": 2593 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.3046, "step": 2594 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.9662, "step": 2595 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.5179, "step": 2596 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.3427, "step": 2597 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.5002, "step": 2598 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.7065, "step": 2599 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.993, "step": 2600 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.5632, "step": 2601 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.7949, "step": 2602 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 2.1607, "step": 2603 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.3311, "step": 2604 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 2.5285, "step": 2605 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.734, "step": 2606 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 2.1876, "step": 2607 }, { "epoch": 0.97, "learning_rate": 2e-05, "loss": 1.6215, "step": 2608 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.4027, "step": 2609 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.2532, "step": 2610 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.1372, "step": 2611 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.6261, "step": 2612 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.7534, "step": 2613 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.6698, "step": 2614 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.8655, "step": 2615 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.5975, "step": 2616 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.2074, "step": 2617 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.5871, "step": 2618 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.5563, "step": 2619 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 2.3665, "step": 2620 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.2221, "step": 2621 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.5773, "step": 2622 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.537, "step": 2623 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.351, "step": 2624 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 2.3814, "step": 2625 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.2293, "step": 2626 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.5551, "step": 2627 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.3611, "step": 2628 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.4901, "step": 2629 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.8713, "step": 2630 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.7139, "step": 2631 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.6279, "step": 2632 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.2752, "step": 2633 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.4165, "step": 2634 }, { "epoch": 0.98, "learning_rate": 2e-05, "loss": 1.7105, "step": 2635 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.9619, "step": 2636 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.9869, "step": 2637 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.3785, "step": 2638 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 2.2022, "step": 2639 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.6332, "step": 2640 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 2.4525, "step": 2641 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.5815, "step": 2642 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.4574, "step": 2643 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.3569, "step": 2644 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.4156, "step": 2645 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.3883, "step": 2646 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.2817, "step": 2647 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.3384, "step": 2648 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.4771, "step": 2649 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.3317, "step": 2650 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.6706, "step": 2651 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.4153, "step": 2652 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.3989, "step": 2653 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.0488, "step": 2654 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 2.3942, "step": 2655 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.4822, "step": 2656 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.6888, "step": 2657 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.4965, "step": 2658 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.494, "step": 2659 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.2092, "step": 2660 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.3115, "step": 2661 }, { "epoch": 0.99, "learning_rate": 2e-05, "loss": 1.9762, "step": 2662 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.7571, "step": 2663 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.7509, "step": 2664 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.5168, "step": 2665 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.3897, "step": 2666 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.3139, "step": 2667 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.3696, "step": 2668 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.8737, "step": 2669 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 2.2141, "step": 2670 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.6401, "step": 2671 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.4046, "step": 2672 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.5658, "step": 2673 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.5152, "step": 2674 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.8003, "step": 2675 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 2.223, "step": 2676 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.519, "step": 2677 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.4477, "step": 2678 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.0143, "step": 2679 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.2429, "step": 2680 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.0227, "step": 2681 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 0.9606, "step": 2682 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.0687, "step": 2683 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 0.7102, "step": 2684 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.588, "step": 2685 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 0.6557, "step": 2686 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.0667, "step": 2687 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.3252, "step": 2688 }, { "epoch": 1.0, "learning_rate": 2e-05, "loss": 1.0521, "step": 2689 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.9522, "step": 2690 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.1598, "step": 2691 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.9103, "step": 2692 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.13, "step": 2693 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.8708, "step": 2694 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.102, "step": 2695 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.868, "step": 2696 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.1812, "step": 2697 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.65, "step": 2698 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.7859, "step": 2699 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.3484, "step": 2700 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.1641, "step": 2701 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.1549, "step": 2702 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.2583, "step": 2703 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.0347, "step": 2704 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.0245, "step": 2705 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.9544, "step": 2706 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.0192, "step": 2707 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.919, "step": 2708 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.0935, "step": 2709 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.8229, "step": 2710 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.8188, "step": 2711 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.2864, "step": 2712 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.9044, "step": 2713 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 1.0752, "step": 2714 }, { "epoch": 1.01, "learning_rate": 2e-05, "loss": 0.817, "step": 2715 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.8169, "step": 2716 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.4542, "step": 2717 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.8168, "step": 2718 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.0622, "step": 2719 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.9127, "step": 2720 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.1511, "step": 2721 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.2098, "step": 2722 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.9875, "step": 2723 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.0889, "step": 2724 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.2546, "step": 2725 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.8123, "step": 2726 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.7818, "step": 2727 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.1889, "step": 2728 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.1597, "step": 2729 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.0978, "step": 2730 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.7165, "step": 2731 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.2616, "step": 2732 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.0618, "step": 2733 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.3416, "step": 2734 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.178, "step": 2735 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.329, "step": 2736 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.4869, "step": 2737 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 1.2119, "step": 2738 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.8927, "step": 2739 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.8367, "step": 2740 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.9814, "step": 2741 }, { "epoch": 1.02, "learning_rate": 2e-05, "loss": 0.8869, "step": 2742 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.7004, "step": 2743 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.6881, "step": 2744 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.9215, "step": 2745 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.9172, "step": 2746 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.6754, "step": 2747 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.12, "step": 2748 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.1008, "step": 2749 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.0225, "step": 2750 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.0016, "step": 2751 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.5993, "step": 2752 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.8628, "step": 2753 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.7805, "step": 2754 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.37, "step": 2755 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.9331, "step": 2756 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.0843, "step": 2757 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.8575, "step": 2758 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.1181, "step": 2759 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.2466, "step": 2760 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.9127, "step": 2761 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.3005, "step": 2762 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.1873, "step": 2763 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.9325, "step": 2764 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.8886, "step": 2765 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 1.136, "step": 2766 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.8839, "step": 2767 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.8079, "step": 2768 }, { "epoch": 1.03, "learning_rate": 2e-05, "loss": 0.8638, "step": 2769 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.1048, "step": 2770 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8697, "step": 2771 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.1597, "step": 2772 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.0485, "step": 2773 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8108, "step": 2774 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8414, "step": 2775 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8183, "step": 2776 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.0302, "step": 2777 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.3683, "step": 2778 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.1882, "step": 2779 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8687, "step": 2780 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.5704, "step": 2781 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8109, "step": 2782 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.6453, "step": 2783 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.4996, "step": 2784 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.192, "step": 2785 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.6577, "step": 2786 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8167, "step": 2787 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.1304, "step": 2788 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.152, "step": 2789 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.0457, "step": 2790 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.7068, "step": 2791 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.1746, "step": 2792 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 1.0835, "step": 2793 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.7834, "step": 2794 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.8981, "step": 2795 }, { "epoch": 1.04, "learning_rate": 2e-05, "loss": 0.9261, "step": 2796 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.3641, "step": 2797 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.0976, "step": 2798 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.9718, "step": 2799 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.8482, "step": 2800 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.865, "step": 2801 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.9938, "step": 2802 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.7702, "step": 2803 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.8721, "step": 2804 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.7644, "step": 2805 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.2648, "step": 2806 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.1359, "step": 2807 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.9411, "step": 2808 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.0024, "step": 2809 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.2191, "step": 2810 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.7712, "step": 2811 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.801, "step": 2812 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.852, "step": 2813 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.8661, "step": 2814 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.9381, "step": 2815 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.7406, "step": 2816 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.1565, "step": 2817 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.3131, "step": 2818 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.46, "step": 2819 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.4156, "step": 2820 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 1.3602, "step": 2821 }, { "epoch": 1.05, "learning_rate": 2e-05, "loss": 0.9227, "step": 2822 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.8911, "step": 2823 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.2828, "step": 2824 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.8252, "step": 2825 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.0342, "step": 2826 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.9968, "step": 2827 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.2692, "step": 2828 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.8844, "step": 2829 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.8759, "step": 2830 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.8103, "step": 2831 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.7909, "step": 2832 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.7589, "step": 2833 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.1389, "step": 2834 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.8277, "step": 2835 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.809, "step": 2836 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.7135, "step": 2837 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.0393, "step": 2838 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.0951, "step": 2839 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.774, "step": 2840 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.0617, "step": 2841 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.1565, "step": 2842 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.0498, "step": 2843 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.8353, "step": 2844 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 2.0514, "step": 2845 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.9381, "step": 2846 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.083, "step": 2847 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 0.8856, "step": 2848 }, { "epoch": 1.06, "learning_rate": 2e-05, "loss": 1.0841, "step": 2849 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.1977, "step": 2850 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.8372, "step": 2851 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.9784, "step": 2852 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.1301, "step": 2853 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.8844, "step": 2854 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.9562, "step": 2855 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.2933, "step": 2856 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.7311, "step": 2857 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.7672, "step": 2858 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.4133, "step": 2859 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.9511, "step": 2860 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.2537, "step": 2861 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.2654, "step": 2862 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.8416, "step": 2863 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.0019, "step": 2864 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.9097, "step": 2865 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.1306, "step": 2866 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.7048, "step": 2867 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.7257, "step": 2868 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.253, "step": 2869 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.9786, "step": 2870 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.1218, "step": 2871 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.925, "step": 2872 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.1626, "step": 2873 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.3812, "step": 2874 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 0.7578, "step": 2875 }, { "epoch": 1.07, "learning_rate": 2e-05, "loss": 1.047, "step": 2876 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 1.2458, "step": 2877 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.9796, "step": 2878 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8776, "step": 2879 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8822, "step": 2880 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8586, "step": 2881 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8172, "step": 2882 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.922, "step": 2883 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.7032, "step": 2884 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8594, "step": 2885 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.9442, "step": 2886 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8395, "step": 2887 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8895, "step": 2888 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.7545, "step": 2889 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 1.1518, "step": 2890 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 1.3386, "step": 2891 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8813, "step": 2892 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.9199, "step": 2893 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8587, "step": 2894 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.9533, "step": 2895 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.8129, "step": 2896 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.9567, "step": 2897 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 1.0348, "step": 2898 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.7939, "step": 2899 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.5999, "step": 2900 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 1.0952, "step": 2901 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 1.0627, "step": 2902 }, { "epoch": 1.08, "learning_rate": 2e-05, "loss": 0.934, "step": 2903 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.0508, "step": 2904 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.0905, "step": 2905 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.3051, "step": 2906 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.9793, "step": 2907 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.7754, "step": 2908 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.1523, "step": 2909 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.2746, "step": 2910 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.9141, "step": 2911 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.8419, "step": 2912 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.8779, "step": 2913 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.026, "step": 2914 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.9782, "step": 2915 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.1935, "step": 2916 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.9543, "step": 2917 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.9504, "step": 2918 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.8053, "step": 2919 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.8365, "step": 2920 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.0777, "step": 2921 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.8425, "step": 2922 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.3396, "step": 2923 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.1817, "step": 2924 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.3409, "step": 2925 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.7791, "step": 2926 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.8678, "step": 2927 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 1.2612, "step": 2928 }, { "epoch": 1.09, "learning_rate": 2e-05, "loss": 0.7076, "step": 2929 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9229, "step": 2930 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9566, "step": 2931 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.8402, "step": 2932 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.936, "step": 2933 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9085, "step": 2934 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9107, "step": 2935 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 1.3036, "step": 2936 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.8951, "step": 2937 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.8005, "step": 2938 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 1.2535, "step": 2939 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 1.2748, "step": 2940 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.7581, "step": 2941 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 1.0292, "step": 2942 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.7852, "step": 2943 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.8682, "step": 2944 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9714, "step": 2945 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 1.0955, "step": 2946 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9444, "step": 2947 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.7934, "step": 2948 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9861, "step": 2949 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.8515, "step": 2950 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 1.143, "step": 2951 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.7778, "step": 2952 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.8407, "step": 2953 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.9956, "step": 2954 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 1.001, "step": 2955 }, { "epoch": 1.1, "learning_rate": 2e-05, "loss": 0.7742, "step": 2956 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.0396, "step": 2957 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.6944, "step": 2958 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.8838, "step": 2959 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.7939, "step": 2960 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.0855, "step": 2961 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.0113, "step": 2962 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.8859, "step": 2963 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.9576, "step": 2964 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.0346, "step": 2965 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.7187, "step": 2966 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.146, "step": 2967 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.9469, "step": 2968 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.7718, "step": 2969 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.7976, "step": 2970 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.3831, "step": 2971 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.1718, "step": 2972 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.0767, "step": 2973 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.1837, "step": 2974 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.3024, "step": 2975 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.9801, "step": 2976 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.8801, "step": 2977 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.863, "step": 2978 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.4376, "step": 2979 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.9374, "step": 2980 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.7102, "step": 2981 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 0.9576, "step": 2982 }, { "epoch": 1.11, "learning_rate": 2e-05, "loss": 1.3328, "step": 2983 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.9485, "step": 2984 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.3025, "step": 2985 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.8694, "step": 2986 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.8826, "step": 2987 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.8702, "step": 2988 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.294, "step": 2989 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.7659, "step": 2990 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.0029, "step": 2991 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.7488, "step": 2992 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.7495, "step": 2993 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.8791, "step": 2994 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.6889, "step": 2995 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.6373, "step": 2996 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.0664, "step": 2997 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.6399, "step": 2998 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.5099, "step": 2999 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.1431, "step": 3000 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.037, "step": 3001 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.9156, "step": 3002 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.9571, "step": 3003 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.4939, "step": 3004 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.9353, "step": 3005 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.8082, "step": 3006 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.9383, "step": 3007 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.1075, "step": 3008 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 1.0936, "step": 3009 }, { "epoch": 1.12, "learning_rate": 2e-05, "loss": 0.8362, "step": 3010 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.5962, "step": 3011 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.7769, "step": 3012 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.2577, "step": 3013 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.9368, "step": 3014 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.8686, "step": 3015 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.1472, "step": 3016 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.8214, "step": 3017 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.003, "step": 3018 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.1685, "step": 3019 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.8609, "step": 3020 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.0444, "step": 3021 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.9435, "step": 3022 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.8646, "step": 3023 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.6336, "step": 3024 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.9021, "step": 3025 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.9138, "step": 3026 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.1526, "step": 3027 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.9336, "step": 3028 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.7831, "step": 3029 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.8341, "step": 3030 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.8246, "step": 3031 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.0031, "step": 3032 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.7189, "step": 3033 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.0263, "step": 3034 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 0.8669, "step": 3035 }, { "epoch": 1.13, "learning_rate": 2e-05, "loss": 1.1955, "step": 3036 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.8854, "step": 3037 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.7513, "step": 3038 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.0845, "step": 3039 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.7634, "step": 3040 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9359, "step": 3041 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.043, "step": 3042 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.1292, "step": 3043 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.7722, "step": 3044 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9464, "step": 3045 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9087, "step": 3046 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.0352, "step": 3047 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9697, "step": 3048 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.789, "step": 3049 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9933, "step": 3050 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9251, "step": 3051 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.759, "step": 3052 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.0579, "step": 3053 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.3162, "step": 3054 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.8186, "step": 3055 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.7529, "step": 3056 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9271, "step": 3057 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.8729, "step": 3058 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9333, "step": 3059 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.062, "step": 3060 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.2942, "step": 3061 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 1.0517, "step": 3062 }, { "epoch": 1.14, "learning_rate": 2e-05, "loss": 0.9499, "step": 3063 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.0641, "step": 3064 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.2033, "step": 3065 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.4899, "step": 3066 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.1115, "step": 3067 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9066, "step": 3068 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.2329, "step": 3069 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9846, "step": 3070 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.7498, "step": 3071 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.9737, "step": 3072 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.2891, "step": 3073 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.2325, "step": 3074 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9209, "step": 3075 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9005, "step": 3076 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.8571, "step": 3077 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.6689, "step": 3078 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9845, "step": 3079 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.7448, "step": 3080 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.7097, "step": 3081 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9385, "step": 3082 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.2232, "step": 3083 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 1.1082, "step": 3084 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.8787, "step": 3085 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9503, "step": 3086 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.8773, "step": 3087 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.7103, "step": 3088 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.7221, "step": 3089 }, { "epoch": 1.15, "learning_rate": 2e-05, "loss": 0.9104, "step": 3090 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.0201, "step": 3091 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.7665, "step": 3092 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.9471, "step": 3093 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.6516, "step": 3094 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.8215, "step": 3095 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.7494, "step": 3096 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.658, "step": 3097 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.0463, "step": 3098 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.9195, "step": 3099 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.083, "step": 3100 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.2339, "step": 3101 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.7465, "step": 3102 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.1494, "step": 3103 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.2352, "step": 3104 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.0629, "step": 3105 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.9264, "step": 3106 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.8888, "step": 3107 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.1822, "step": 3108 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.0064, "step": 3109 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.9735, "step": 3110 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.7938, "step": 3111 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.0209, "step": 3112 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.9668, "step": 3113 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.9289, "step": 3114 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 1.8127, "step": 3115 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.6573, "step": 3116 }, { "epoch": 1.16, "learning_rate": 2e-05, "loss": 0.8536, "step": 3117 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.5037, "step": 3118 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.7358, "step": 3119 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.1198, "step": 3120 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.9969, "step": 3121 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.6794, "step": 3122 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.0107, "step": 3123 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.4515, "step": 3124 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.2906, "step": 3125 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.1104, "step": 3126 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.1926, "step": 3127 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.0764, "step": 3128 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.8137, "step": 3129 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.3987, "step": 3130 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.947, "step": 3131 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.6654, "step": 3132 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.1632, "step": 3133 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.9218, "step": 3134 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.8191, "step": 3135 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.1604, "step": 3136 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.7671, "step": 3137 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.6962, "step": 3138 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.6824, "step": 3139 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.9501, "step": 3140 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.9402, "step": 3141 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 0.9247, "step": 3142 }, { "epoch": 1.17, "learning_rate": 2e-05, "loss": 1.6484, "step": 3143 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.32, "step": 3144 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.1809, "step": 3145 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.2356, "step": 3146 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.0775, "step": 3147 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.0812, "step": 3148 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.2521, "step": 3149 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.8565, "step": 3150 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.9616, "step": 3151 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.0073, "step": 3152 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.8188, "step": 3153 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.8408, "step": 3154 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.812, "step": 3155 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.7486, "step": 3156 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.9692, "step": 3157 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.3153, "step": 3158 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.065, "step": 3159 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.0934, "step": 3160 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.43, "step": 3161 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.8099, "step": 3162 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.7962, "step": 3163 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.9525, "step": 3164 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.8679, "step": 3165 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 1.1474, "step": 3166 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.931, "step": 3167 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.8189, "step": 3168 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.9813, "step": 3169 }, { "epoch": 1.18, "learning_rate": 2e-05, "loss": 0.9971, "step": 3170 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8935, "step": 3171 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 1.0986, "step": 3172 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8261, "step": 3173 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.7367, "step": 3174 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8953, "step": 3175 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9307, "step": 3176 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9462, "step": 3177 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8649, "step": 3178 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8635, "step": 3179 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9005, "step": 3180 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 1.0809, "step": 3181 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9469, "step": 3182 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8048, "step": 3183 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9992, "step": 3184 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9225, "step": 3185 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9173, "step": 3186 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.7867, "step": 3187 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8908, "step": 3188 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 1.0846, "step": 3189 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.7836, "step": 3190 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.8853, "step": 3191 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 1.1562, "step": 3192 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 1.0684, "step": 3193 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.6586, "step": 3194 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 0.9943, "step": 3195 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 1.2384, "step": 3196 }, { "epoch": 1.19, "learning_rate": 2e-05, "loss": 1.069, "step": 3197 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9004, "step": 3198 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.8718, "step": 3199 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.0595, "step": 3200 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9127, "step": 3201 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9104, "step": 3202 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.8533, "step": 3203 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9259, "step": 3204 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.2587, "step": 3205 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.4573, "step": 3206 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.0039, "step": 3207 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9819, "step": 3208 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.7811, "step": 3209 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.0106, "step": 3210 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.0585, "step": 3211 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.911, "step": 3212 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9084, "step": 3213 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.7161, "step": 3214 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.2029, "step": 3215 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9064, "step": 3216 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.773, "step": 3217 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 1.1988, "step": 3218 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9474, "step": 3219 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9323, "step": 3220 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9008, "step": 3221 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9886, "step": 3222 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.9561, "step": 3223 }, { "epoch": 1.2, "learning_rate": 2e-05, "loss": 0.7055, "step": 3224 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.2444, "step": 3225 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.2579, "step": 3226 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.2035, "step": 3227 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.7161, "step": 3228 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.3166, "step": 3229 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.7399, "step": 3230 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.0215, "step": 3231 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.8484, "step": 3232 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.9989, "step": 3233 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.6493, "step": 3234 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.6924, "step": 3235 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.7924, "step": 3236 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.8226, "step": 3237 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.9698, "step": 3238 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.9515, "step": 3239 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.2094, "step": 3240 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.7346, "step": 3241 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.9336, "step": 3242 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.8895, "step": 3243 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.3868, "step": 3244 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.1721, "step": 3245 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.9553, "step": 3246 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.8918, "step": 3247 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.8116, "step": 3248 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 1.0649, "step": 3249 }, { "epoch": 1.21, "learning_rate": 2e-05, "loss": 0.8726, "step": 3250 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.1346, "step": 3251 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.6666, "step": 3252 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.2768, "step": 3253 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.4126, "step": 3254 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.2018, "step": 3255 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.8676, "step": 3256 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.8736, "step": 3257 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.1487, "step": 3258 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.1018, "step": 3259 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.7409, "step": 3260 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.869, "step": 3261 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.1449, "step": 3262 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.8587, "step": 3263 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.0736, "step": 3264 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.9246, "step": 3265 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.8953, "step": 3266 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.2021, "step": 3267 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.8866, "step": 3268 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.6357, "step": 3269 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.8986, "step": 3270 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.8014, "step": 3271 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.2769, "step": 3272 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.0144, "step": 3273 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.958, "step": 3274 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 0.7015, "step": 3275 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.0607, "step": 3276 }, { "epoch": 1.22, "learning_rate": 2e-05, "loss": 1.3328, "step": 3277 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 1.6984, "step": 3278 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8932, "step": 3279 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 1.0249, "step": 3280 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8521, "step": 3281 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.9525, "step": 3282 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8539, "step": 3283 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.6356, "step": 3284 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.6618, "step": 3285 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.9208, "step": 3286 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 1.2089, "step": 3287 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.7669, "step": 3288 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.7762, "step": 3289 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 1.149, "step": 3290 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8503, "step": 3291 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.9972, "step": 3292 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 1.1053, "step": 3293 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 1.1238, "step": 3294 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.7589, "step": 3295 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.9716, "step": 3296 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8146, "step": 3297 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.9978, "step": 3298 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8406, "step": 3299 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8829, "step": 3300 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.9619, "step": 3301 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.796, "step": 3302 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.9249, "step": 3303 }, { "epoch": 1.23, "learning_rate": 2e-05, "loss": 0.8345, "step": 3304 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.4107, "step": 3305 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.1968, "step": 3306 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.869, "step": 3307 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.0056, "step": 3308 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.2766, "step": 3309 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.7561, "step": 3310 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.2897, "step": 3311 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.0749, "step": 3312 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.8766, "step": 3313 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.75, "step": 3314 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.9047, "step": 3315 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.1242, "step": 3316 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.9982, "step": 3317 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.9762, "step": 3318 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.1993, "step": 3319 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.0841, "step": 3320 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.2119, "step": 3321 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.0901, "step": 3322 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.1494, "step": 3323 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.0673, "step": 3324 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.7669, "step": 3325 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 0.9669, "step": 3326 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.1032, "step": 3327 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.4198, "step": 3328 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.2019, "step": 3329 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.0347, "step": 3330 }, { "epoch": 1.24, "learning_rate": 2e-05, "loss": 1.4002, "step": 3331 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.8439, "step": 3332 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.9728, "step": 3333 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.8232, "step": 3334 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.7591, "step": 3335 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.2508, "step": 3336 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.0883, "step": 3337 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.7566, "step": 3338 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.9019, "step": 3339 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.4708, "step": 3340 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.6371, "step": 3341 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.1767, "step": 3342 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.1926, "step": 3343 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.8078, "step": 3344 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.8163, "step": 3345 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.0101, "step": 3346 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.9232, "step": 3347 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.7534, "step": 3348 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.3041, "step": 3349 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.9183, "step": 3350 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.4512, "step": 3351 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.3552, "step": 3352 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.3718, "step": 3353 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.7367, "step": 3354 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.1753, "step": 3355 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 1.0498, "step": 3356 }, { "epoch": 1.25, "learning_rate": 2e-05, "loss": 0.8107, "step": 3357 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.7945, "step": 3358 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.9864, "step": 3359 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.9972, "step": 3360 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.82, "step": 3361 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.8524, "step": 3362 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.1927, "step": 3363 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.7735, "step": 3364 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.7204, "step": 3365 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.9017, "step": 3366 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.8932, "step": 3367 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.7965, "step": 3368 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.9756, "step": 3369 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.8581, "step": 3370 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.0953, "step": 3371 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.9735, "step": 3372 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.1434, "step": 3373 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.9713, "step": 3374 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.2023, "step": 3375 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.3712, "step": 3376 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.2342, "step": 3377 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.8303, "step": 3378 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.7817, "step": 3379 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.0162, "step": 3380 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 1.0527, "step": 3381 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.9022, "step": 3382 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.7794, "step": 3383 }, { "epoch": 1.26, "learning_rate": 2e-05, "loss": 0.8057, "step": 3384 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.7969, "step": 3385 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9691, "step": 3386 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9021, "step": 3387 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.0491, "step": 3388 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.8763, "step": 3389 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.0557, "step": 3390 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.8796, "step": 3391 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.4543, "step": 3392 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.2512, "step": 3393 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.7926, "step": 3394 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.028, "step": 3395 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9279, "step": 3396 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9966, "step": 3397 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9949, "step": 3398 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9515, "step": 3399 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.8832, "step": 3400 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9517, "step": 3401 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.8233, "step": 3402 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.6729, "step": 3403 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.8912, "step": 3404 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.0682, "step": 3405 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.6889, "step": 3406 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.8034, "step": 3407 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.1874, "step": 3408 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.7025, "step": 3409 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 0.9153, "step": 3410 }, { "epoch": 1.27, "learning_rate": 2e-05, "loss": 1.1345, "step": 3411 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.1965, "step": 3412 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.8398, "step": 3413 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.9833, "step": 3414 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.0068, "step": 3415 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.1356, "step": 3416 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.5211, "step": 3417 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.1797, "step": 3418 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.8015, "step": 3419 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.17, "step": 3420 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.8282, "step": 3421 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.8074, "step": 3422 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.6965, "step": 3423 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.8739, "step": 3424 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.8972, "step": 3425 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.7777, "step": 3426 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.7072, "step": 3427 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.3775, "step": 3428 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.7674, "step": 3429 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.803, "step": 3430 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.6043, "step": 3431 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.7278, "step": 3432 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.2682, "step": 3433 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.9007, "step": 3434 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.9497, "step": 3435 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.9335, "step": 3436 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 0.8622, "step": 3437 }, { "epoch": 1.28, "learning_rate": 2e-05, "loss": 1.0628, "step": 3438 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.7179, "step": 3439 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.786, "step": 3440 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.3993, "step": 3441 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8876, "step": 3442 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.0273, "step": 3443 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8115, "step": 3444 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8445, "step": 3445 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8311, "step": 3446 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8302, "step": 3447 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.933, "step": 3448 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.9346, "step": 3449 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8656, "step": 3450 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.9176, "step": 3451 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.7463, "step": 3452 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8407, "step": 3453 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.9605, "step": 3454 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.0219, "step": 3455 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.0009, "step": 3456 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.2166, "step": 3457 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.1489, "step": 3458 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.1313, "step": 3459 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.769, "step": 3460 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.0734, "step": 3461 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.7392, "step": 3462 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 1.1734, "step": 3463 }, { "epoch": 1.29, "learning_rate": 2e-05, "loss": 0.8943, "step": 3464 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.0523, "step": 3465 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.2392, "step": 3466 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.1515, "step": 3467 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.7664, "step": 3468 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.749, "step": 3469 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.6812, "step": 3470 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.8599, "step": 3471 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.1162, "step": 3472 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9283, "step": 3473 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.8386, "step": 3474 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9179, "step": 3475 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9644, "step": 3476 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.4627, "step": 3477 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.8679, "step": 3478 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.0147, "step": 3479 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.7088, "step": 3480 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9631, "step": 3481 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9974, "step": 3482 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9207, "step": 3483 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.856, "step": 3484 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.8991, "step": 3485 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.1275, "step": 3486 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 1.019, "step": 3487 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.8417, "step": 3488 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9998, "step": 3489 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9882, "step": 3490 }, { "epoch": 1.3, "learning_rate": 2e-05, "loss": 0.9874, "step": 3491 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.8558, "step": 3492 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.9464, "step": 3493 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.7821, "step": 3494 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.8456, "step": 3495 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 1.0561, "step": 3496 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.7617, "step": 3497 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 1.2071, "step": 3498 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.8311, "step": 3499 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.7861, "step": 3500 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.8243, "step": 3501 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.9158, "step": 3502 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.8188, "step": 3503 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.9384, "step": 3504 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 1.5685, "step": 3505 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.9843, "step": 3506 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 1.0794, "step": 3507 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 1.1251, "step": 3508 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.7745, "step": 3509 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 1.0366, "step": 3510 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.7293, "step": 3511 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.9607, "step": 3512 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.798, "step": 3513 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.9043, "step": 3514 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.8439, "step": 3515 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.9913, "step": 3516 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 1.3054, "step": 3517 }, { "epoch": 1.31, "learning_rate": 2e-05, "loss": 0.954, "step": 3518 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8462, "step": 3519 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.3988, "step": 3520 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8029, "step": 3521 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.937, "step": 3522 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8091, "step": 3523 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8797, "step": 3524 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.9373, "step": 3525 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.0735, "step": 3526 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.2059, "step": 3527 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.0168, "step": 3528 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.0603, "step": 3529 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.3089, "step": 3530 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.9824, "step": 3531 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.9372, "step": 3532 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.7768, "step": 3533 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.2137, "step": 3534 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8736, "step": 3535 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.2186, "step": 3536 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8523, "step": 3537 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.0023, "step": 3538 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8684, "step": 3539 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.9895, "step": 3540 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.1449, "step": 3541 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8694, "step": 3542 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.8849, "step": 3543 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 0.9295, "step": 3544 }, { "epoch": 1.32, "learning_rate": 2e-05, "loss": 1.0339, "step": 3545 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.9289, "step": 3546 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.2868, "step": 3547 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.0503, "step": 3548 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.9137, "step": 3549 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.1115, "step": 3550 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.7524, "step": 3551 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.0927, "step": 3552 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.8405, "step": 3553 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.2654, "step": 3554 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.1154, "step": 3555 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.5441, "step": 3556 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.0043, "step": 3557 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.807, "step": 3558 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.859, "step": 3559 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.8312, "step": 3560 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.2787, "step": 3561 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.0076, "step": 3562 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.2002, "step": 3563 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.0359, "step": 3564 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.0511, "step": 3565 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.9332, "step": 3566 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.3608, "step": 3567 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.9257, "step": 3568 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.1199, "step": 3569 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 1.0064, "step": 3570 }, { "epoch": 1.33, "learning_rate": 2e-05, "loss": 0.8823, "step": 3571 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.9633, "step": 3572 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.4545, "step": 3573 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.3605, "step": 3574 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.9811, "step": 3575 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.9154, "step": 3576 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.2263, "step": 3577 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.9329, "step": 3578 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.0164, "step": 3579 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.8314, "step": 3580 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.0382, "step": 3581 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.0374, "step": 3582 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.4583, "step": 3583 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.3935, "step": 3584 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.972, "step": 3585 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.2372, "step": 3586 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.7506, "step": 3587 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.8643, "step": 3588 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.5635, "step": 3589 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.7983, "step": 3590 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.981, "step": 3591 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.9254, "step": 3592 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.8706, "step": 3593 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.0893, "step": 3594 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.7263, "step": 3595 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 0.7752, "step": 3596 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.1859, "step": 3597 }, { "epoch": 1.34, "learning_rate": 2e-05, "loss": 1.1537, "step": 3598 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.036, "step": 3599 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.7459, "step": 3600 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.8435, "step": 3601 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.0063, "step": 3602 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.0501, "step": 3603 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.923, "step": 3604 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.9343, "step": 3605 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.9256, "step": 3606 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.1535, "step": 3607 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.7502, "step": 3608 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.7366, "step": 3609 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.9882, "step": 3610 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.7148, "step": 3611 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.0303, "step": 3612 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.8326, "step": 3613 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.8117, "step": 3614 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.2737, "step": 3615 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.8885, "step": 3616 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.7656, "step": 3617 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.9852, "step": 3618 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.9137, "step": 3619 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.7968, "step": 3620 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.0578, "step": 3621 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.743, "step": 3622 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.7288, "step": 3623 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 1.8303, "step": 3624 }, { "epoch": 1.35, "learning_rate": 2e-05, "loss": 0.9494, "step": 3625 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.7332, "step": 3626 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.9231, "step": 3627 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.0019, "step": 3628 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.1151, "step": 3629 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.9922, "step": 3630 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.9975, "step": 3631 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.8189, "step": 3632 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.6841, "step": 3633 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.0928, "step": 3634 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.8232, "step": 3635 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.8033, "step": 3636 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.1737, "step": 3637 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.8354, "step": 3638 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.9997, "step": 3639 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.685, "step": 3640 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.0363, "step": 3641 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.5638, "step": 3642 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.9968, "step": 3643 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.9069, "step": 3644 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.6587, "step": 3645 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.3621, "step": 3646 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.0087, "step": 3647 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 1.0045, "step": 3648 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.9255, "step": 3649 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.7596, "step": 3650 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.7254, "step": 3651 }, { "epoch": 1.36, "learning_rate": 2e-05, "loss": 0.8629, "step": 3652 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8528, "step": 3653 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 1.1613, "step": 3654 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8555, "step": 3655 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.9768, "step": 3656 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 1.068, "step": 3657 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.9885, "step": 3658 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8175, "step": 3659 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8557, "step": 3660 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.9721, "step": 3661 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 1.1357, "step": 3662 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.9162, "step": 3663 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8777, "step": 3664 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 1.1899, "step": 3665 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 1.0515, "step": 3666 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.964, "step": 3667 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8692, "step": 3668 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 1.1128, "step": 3669 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8098, "step": 3670 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.7986, "step": 3671 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.9458, "step": 3672 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8993, "step": 3673 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.9138, "step": 3674 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.7412, "step": 3675 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8149, "step": 3676 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 1.1956, "step": 3677 }, { "epoch": 1.37, "learning_rate": 2e-05, "loss": 0.8127, "step": 3678 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9154, "step": 3679 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9336, "step": 3680 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.0046, "step": 3681 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9652, "step": 3682 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.8732, "step": 3683 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.8033, "step": 3684 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.3524, "step": 3685 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.407, "step": 3686 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9884, "step": 3687 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.8125, "step": 3688 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.2664, "step": 3689 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9427, "step": 3690 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.4264, "step": 3691 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.654, "step": 3692 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.7862, "step": 3693 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9993, "step": 3694 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.8701, "step": 3695 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9278, "step": 3696 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.0836, "step": 3697 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.8631, "step": 3698 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.8083, "step": 3699 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.0882, "step": 3700 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.8893, "step": 3701 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.0417, "step": 3702 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9986, "step": 3703 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 1.6492, "step": 3704 }, { "epoch": 1.38, "learning_rate": 2e-05, "loss": 0.9698, "step": 3705 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.0026, "step": 3706 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.2017, "step": 3707 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.7671, "step": 3708 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.9885, "step": 3709 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.2427, "step": 3710 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.223, "step": 3711 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.8486, "step": 3712 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.277, "step": 3713 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.9361, "step": 3714 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.0736, "step": 3715 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.8014, "step": 3716 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.7541, "step": 3717 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.8791, "step": 3718 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.0326, "step": 3719 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.7599, "step": 3720 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.8722, "step": 3721 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.2397, "step": 3722 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.991, "step": 3723 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.2703, "step": 3724 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.7936, "step": 3725 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.3986, "step": 3726 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 0.7954, "step": 3727 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.0233, "step": 3728 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.3203, "step": 3729 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.3918, "step": 3730 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.5557, "step": 3731 }, { "epoch": 1.39, "learning_rate": 2e-05, "loss": 1.4935, "step": 3732 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.7877, "step": 3733 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.1186, "step": 3734 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.811, "step": 3735 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.0881, "step": 3736 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.1173, "step": 3737 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.1569, "step": 3738 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.0193, "step": 3739 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.8668, "step": 3740 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 2.0691, "step": 3741 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.2676, "step": 3742 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.7511, "step": 3743 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.0321, "step": 3744 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.8717, "step": 3745 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.008, "step": 3746 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.0302, "step": 3747 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.8299, "step": 3748 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.0578, "step": 3749 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.2457, "step": 3750 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.8156, "step": 3751 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.287, "step": 3752 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.8732, "step": 3753 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.986, "step": 3754 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.9095, "step": 3755 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.7721, "step": 3756 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.9302, "step": 3757 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 0.9676, "step": 3758 }, { "epoch": 1.4, "learning_rate": 2e-05, "loss": 1.1807, "step": 3759 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.944, "step": 3760 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.7452, "step": 3761 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.2634, "step": 3762 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.4166, "step": 3763 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.8217, "step": 3764 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.8595, "step": 3765 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9198, "step": 3766 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.0223, "step": 3767 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.016, "step": 3768 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.7497, "step": 3769 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.822, "step": 3770 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.0332, "step": 3771 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9815, "step": 3772 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.8924, "step": 3773 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.8363, "step": 3774 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9316, "step": 3775 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.1194, "step": 3776 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.0476, "step": 3777 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9992, "step": 3778 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.8048, "step": 3779 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9278, "step": 3780 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9457, "step": 3781 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.0417, "step": 3782 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.7622, "step": 3783 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 1.305, "step": 3784 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9605, "step": 3785 }, { "epoch": 1.41, "learning_rate": 2e-05, "loss": 0.9016, "step": 3786 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.7524, "step": 3787 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.0733, "step": 3788 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.1119, "step": 3789 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.9205, "step": 3790 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.9658, "step": 3791 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.0125, "step": 3792 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.0497, "step": 3793 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.8984, "step": 3794 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.1235, "step": 3795 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.7853, "step": 3796 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.2141, "step": 3797 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.3526, "step": 3798 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.793, "step": 3799 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.1151, "step": 3800 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.4639, "step": 3801 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.9492, "step": 3802 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.1112, "step": 3803 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.6553, "step": 3804 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.872, "step": 3805 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.9472, "step": 3806 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.9061, "step": 3807 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.019, "step": 3808 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.7442, "step": 3809 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 1.2078, "step": 3810 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.7619, "step": 3811 }, { "epoch": 1.42, "learning_rate": 2e-05, "loss": 0.7786, "step": 3812 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.938, "step": 3813 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.8355, "step": 3814 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.8933, "step": 3815 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.23, "step": 3816 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.0497, "step": 3817 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.2217, "step": 3818 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.7076, "step": 3819 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.7069, "step": 3820 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.6603, "step": 3821 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.8557, "step": 3822 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.7883, "step": 3823 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.8694, "step": 3824 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.4393, "step": 3825 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.9596, "step": 3826 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.1108, "step": 3827 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.3428, "step": 3828 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.9031, "step": 3829 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.9315, "step": 3830 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.9982, "step": 3831 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.2006, "step": 3832 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.2628, "step": 3833 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.7819, "step": 3834 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.2336, "step": 3835 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.7763, "step": 3836 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.7399, "step": 3837 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 1.1634, "step": 3838 }, { "epoch": 1.43, "learning_rate": 2e-05, "loss": 0.7662, "step": 3839 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.1563, "step": 3840 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.9622, "step": 3841 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8898, "step": 3842 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.0804, "step": 3843 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8191, "step": 3844 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.0454, "step": 3845 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.988, "step": 3846 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8894, "step": 3847 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8306, "step": 3848 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.5375, "step": 3849 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8369, "step": 3850 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.0073, "step": 3851 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.1027, "step": 3852 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.1508, "step": 3853 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.1721, "step": 3854 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.6996, "step": 3855 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8921, "step": 3856 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.0597, "step": 3857 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.1605, "step": 3858 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8351, "step": 3859 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8046, "step": 3860 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.9718, "step": 3861 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8427, "step": 3862 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.9507, "step": 3863 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 0.8433, "step": 3864 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.1585, "step": 3865 }, { "epoch": 1.44, "learning_rate": 2e-05, "loss": 1.0199, "step": 3866 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.088, "step": 3867 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.9842, "step": 3868 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.7801, "step": 3869 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.9413, "step": 3870 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.9694, "step": 3871 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.8029, "step": 3872 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.8259, "step": 3873 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.1519, "step": 3874 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.0915, "step": 3875 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.7403, "step": 3876 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.8948, "step": 3877 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.059, "step": 3878 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.1912, "step": 3879 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.0144, "step": 3880 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.1758, "step": 3881 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.9877, "step": 3882 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.951, "step": 3883 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.8591, "step": 3884 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.0096, "step": 3885 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.8294, "step": 3886 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.111, "step": 3887 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.3179, "step": 3888 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.3019, "step": 3889 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.3754, "step": 3890 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 1.0161, "step": 3891 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.649, "step": 3892 }, { "epoch": 1.45, "learning_rate": 2e-05, "loss": 0.7255, "step": 3893 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.6197, "step": 3894 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.2291, "step": 3895 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.8937, "step": 3896 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.25, "step": 3897 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.293, "step": 3898 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.2409, "step": 3899 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.8929, "step": 3900 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.984, "step": 3901 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.791, "step": 3902 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.8565, "step": 3903 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.1253, "step": 3904 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.1176, "step": 3905 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.0278, "step": 3906 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.6138, "step": 3907 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.7502, "step": 3908 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.7059, "step": 3909 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.7424, "step": 3910 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.7474, "step": 3911 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.2349, "step": 3912 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.9675, "step": 3913 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.0178, "step": 3914 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.8137, "step": 3915 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.1642, "step": 3916 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.0569, "step": 3917 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 0.7871, "step": 3918 }, { "epoch": 1.46, "learning_rate": 2e-05, "loss": 1.0923, "step": 3919 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.9314, "step": 3920 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.0848, "step": 3921 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.6962, "step": 3922 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.8046, "step": 3923 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.9378, "step": 3924 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.0413, "step": 3925 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.1549, "step": 3926 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.082, "step": 3927 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.6616, "step": 3928 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.9588, "step": 3929 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.9771, "step": 3930 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.9205, "step": 3931 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.1616, "step": 3932 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.0307, "step": 3933 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.0057, "step": 3934 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.2444, "step": 3935 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.9486, "step": 3936 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.2335, "step": 3937 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.7979, "step": 3938 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.9822, "step": 3939 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.37, "step": 3940 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.7825, "step": 3941 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.8988, "step": 3942 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.0231, "step": 3943 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.2484, "step": 3944 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 0.7037, "step": 3945 }, { "epoch": 1.47, "learning_rate": 2e-05, "loss": 1.0952, "step": 3946 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.0474, "step": 3947 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.9514, "step": 3948 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.811, "step": 3949 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.9634, "step": 3950 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.716, "step": 3951 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.3129, "step": 3952 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.2242, "step": 3953 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.8115, "step": 3954 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.7058, "step": 3955 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.796, "step": 3956 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.7368, "step": 3957 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.397, "step": 3958 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.9358, "step": 3959 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.7561, "step": 3960 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.7898, "step": 3961 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.919, "step": 3962 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.8832, "step": 3963 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.5982, "step": 3964 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.1058, "step": 3965 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.0991, "step": 3966 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.0195, "step": 3967 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.3073, "step": 3968 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.7989, "step": 3969 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.7406, "step": 3970 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.9715, "step": 3971 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 0.7677, "step": 3972 }, { "epoch": 1.48, "learning_rate": 2e-05, "loss": 1.025, "step": 3973 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0132, "step": 3974 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.9199, "step": 3975 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0004, "step": 3976 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.754, "step": 3977 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0056, "step": 3978 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0927, "step": 3979 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.9193, "step": 3980 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.8573, "step": 3981 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0859, "step": 3982 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.9845, "step": 3983 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0687, "step": 3984 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.9044, "step": 3985 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0214, "step": 3986 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.8095, "step": 3987 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.2876, "step": 3988 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.1073, "step": 3989 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.7992, "step": 3990 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.675, "step": 3991 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.9095, "step": 3992 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0148, "step": 3993 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.8053, "step": 3994 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0542, "step": 3995 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.0973, "step": 3996 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.7898, "step": 3997 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.9854, "step": 3998 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 1.119, "step": 3999 }, { "epoch": 1.49, "learning_rate": 2e-05, "loss": 0.8883, "step": 4000 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.0853, "step": 4001 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.2439, "step": 4002 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.6675, "step": 4003 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.678, "step": 4004 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.0431, "step": 4005 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.8738, "step": 4006 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.7992, "step": 4007 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.1375, "step": 4008 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.4674, "step": 4009 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.2416, "step": 4010 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.9148, "step": 4011 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.1285, "step": 4012 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.6357, "step": 4013 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.8324, "step": 4014 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.813, "step": 4015 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.6097, "step": 4016 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.775, "step": 4017 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.418, "step": 4018 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.7281, "step": 4019 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.7474, "step": 4020 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.8144, "step": 4021 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.92, "step": 4022 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.9991, "step": 4023 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.0249, "step": 4024 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 0.7345, "step": 4025 }, { "epoch": 1.5, "learning_rate": 2e-05, "loss": 1.2044, "step": 4026 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.1116, "step": 4027 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.9568, "step": 4028 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.9861, "step": 4029 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.849, "step": 4030 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.1012, "step": 4031 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.9503, "step": 4032 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.1253, "step": 4033 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.9713, "step": 4034 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.1467, "step": 4035 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.1159, "step": 4036 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.0999, "step": 4037 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.0507, "step": 4038 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.69, "step": 4039 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.9016, "step": 4040 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.0389, "step": 4041 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.7734, "step": 4042 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.8254, "step": 4043 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.7293, "step": 4044 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.0699, "step": 4045 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.7403, "step": 4046 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.9653, "step": 4047 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.3866, "step": 4048 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.8783, "step": 4049 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.942, "step": 4050 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.9734, "step": 4051 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 1.1048, "step": 4052 }, { "epoch": 1.51, "learning_rate": 2e-05, "loss": 0.8707, "step": 4053 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.8433, "step": 4054 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.1086, "step": 4055 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.9381, "step": 4056 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.1077, "step": 4057 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.896, "step": 4058 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.3391, "step": 4059 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.8147, "step": 4060 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.1736, "step": 4061 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.2219, "step": 4062 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.7777, "step": 4063 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.9044, "step": 4064 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.8574, "step": 4065 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.7718, "step": 4066 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.0412, "step": 4067 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.9749, "step": 4068 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.1372, "step": 4069 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.7201, "step": 4070 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.3835, "step": 4071 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.0731, "step": 4072 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.0739, "step": 4073 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.7358, "step": 4074 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.8833, "step": 4075 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.7594, "step": 4076 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.0652, "step": 4077 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.703, "step": 4078 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 0.8883, "step": 4079 }, { "epoch": 1.52, "learning_rate": 2e-05, "loss": 1.0806, "step": 4080 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.5375, "step": 4081 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.1966, "step": 4082 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.0455, "step": 4083 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.054, "step": 4084 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.9068, "step": 4085 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.9209, "step": 4086 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.8413, "step": 4087 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.3683, "step": 4088 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.8828, "step": 4089 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.9442, "step": 4090 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.7864, "step": 4091 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.922, "step": 4092 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.9184, "step": 4093 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.1796, "step": 4094 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.7989, "step": 4095 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.846, "step": 4096 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.0636, "step": 4097 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.8367, "step": 4098 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.0323, "step": 4099 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.7225, "step": 4100 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.9561, "step": 4101 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.6968, "step": 4102 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.9236, "step": 4103 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 0.9752, "step": 4104 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.2921, "step": 4105 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.1344, "step": 4106 }, { "epoch": 1.53, "learning_rate": 2e-05, "loss": 1.0281, "step": 4107 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.7731, "step": 4108 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.7491, "step": 4109 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.0613, "step": 4110 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.6843, "step": 4111 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.0021, "step": 4112 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.9176, "step": 4113 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.0856, "step": 4114 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.9009, "step": 4115 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.3, "step": 4116 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.0619, "step": 4117 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.9082, "step": 4118 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.9617, "step": 4119 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.8038, "step": 4120 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.2297, "step": 4121 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.7041, "step": 4122 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.8665, "step": 4123 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.0409, "step": 4124 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.1413, "step": 4125 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.2038, "step": 4126 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.7584, "step": 4127 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.3518, "step": 4128 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.7509, "step": 4129 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 1.1209, "step": 4130 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.8295, "step": 4131 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.861, "step": 4132 }, { "epoch": 1.54, "learning_rate": 2e-05, "loss": 0.9016, "step": 4133 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.0019, "step": 4134 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.4733, "step": 4135 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.1833, "step": 4136 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.8962, "step": 4137 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.7497, "step": 4138 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.7143, "step": 4139 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.9898, "step": 4140 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.8084, "step": 4141 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.8997, "step": 4142 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.6905, "step": 4143 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.1238, "step": 4144 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.0584, "step": 4145 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.0843, "step": 4146 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.9254, "step": 4147 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.0457, "step": 4148 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.0309, "step": 4149 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.8574, "step": 4150 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.3859, "step": 4151 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.8863, "step": 4152 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.1271, "step": 4153 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.9166, "step": 4154 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.2134, "step": 4155 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.8327, "step": 4156 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.0972, "step": 4157 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 1.053, "step": 4158 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.7966, "step": 4159 }, { "epoch": 1.55, "learning_rate": 2e-05, "loss": 0.8, "step": 4160 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9313, "step": 4161 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.8049, "step": 4162 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.7928, "step": 4163 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9002, "step": 4164 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1244, "step": 4165 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1957, "step": 4166 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9257, "step": 4167 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.0759, "step": 4168 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9392, "step": 4169 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9299, "step": 4170 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.2029, "step": 4171 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1639, "step": 4172 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.6749, "step": 4173 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9172, "step": 4174 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1969, "step": 4175 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9859, "step": 4176 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1842, "step": 4177 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.7479, "step": 4178 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1361, "step": 4179 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9008, "step": 4180 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9841, "step": 4181 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.9741, "step": 4182 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.7863, "step": 4183 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1942, "step": 4184 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 0.8618, "step": 4185 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.1394, "step": 4186 }, { "epoch": 1.56, "learning_rate": 2e-05, "loss": 1.0715, "step": 4187 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.7977, "step": 4188 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.0832, "step": 4189 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.6859, "step": 4190 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.9222, "step": 4191 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.7176, "step": 4192 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8462, "step": 4193 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8733, "step": 4194 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8266, "step": 4195 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8583, "step": 4196 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.7683, "step": 4197 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.9169, "step": 4198 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.4209, "step": 4199 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.2855, "step": 4200 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.7581, "step": 4201 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.1151, "step": 4202 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.9883, "step": 4203 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8496, "step": 4204 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.9812, "step": 4205 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.7446, "step": 4206 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8589, "step": 4207 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.9119, "step": 4208 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8516, "step": 4209 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.421, "step": 4210 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 0.8931, "step": 4211 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.1571, "step": 4212 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.1534, "step": 4213 }, { "epoch": 1.57, "learning_rate": 2e-05, "loss": 1.0874, "step": 4214 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.0345, "step": 4215 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.8477, "step": 4216 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.0791, "step": 4217 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.7104, "step": 4218 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.0755, "step": 4219 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.8805, "step": 4220 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.1133, "step": 4221 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.9583, "step": 4222 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.0448, "step": 4223 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.1323, "step": 4224 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.9248, "step": 4225 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.7453, "step": 4226 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.9272, "step": 4227 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.3295, "step": 4228 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.6752, "step": 4229 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.9112, "step": 4230 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.8685, "step": 4231 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.975, "step": 4232 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.2077, "step": 4233 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.9289, "step": 4234 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.8942, "step": 4235 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.695, "step": 4236 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.8221, "step": 4237 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.9075, "step": 4238 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 1.1698, "step": 4239 }, { "epoch": 1.58, "learning_rate": 2e-05, "loss": 0.898, "step": 4240 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.0134, "step": 4241 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.0854, "step": 4242 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.8714, "step": 4243 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.7942, "step": 4244 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.9798, "step": 4245 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.1598, "step": 4246 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.0237, "step": 4247 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.9086, "step": 4248 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.8927, "step": 4249 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.2526, "step": 4250 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.7425, "step": 4251 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.806, "step": 4252 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.7782, "step": 4253 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.2793, "step": 4254 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.0578, "step": 4255 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.7523, "step": 4256 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.767, "step": 4257 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.7889, "step": 4258 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.961, "step": 4259 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.3069, "step": 4260 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.7037, "step": 4261 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.9101, "step": 4262 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.253, "step": 4263 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.0884, "step": 4264 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 1.0191, "step": 4265 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.9083, "step": 4266 }, { "epoch": 1.59, "learning_rate": 2e-05, "loss": 0.9925, "step": 4267 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.8474, "step": 4268 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.9062, "step": 4269 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.1722, "step": 4270 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.7048, "step": 4271 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.8449, "step": 4272 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.8388, "step": 4273 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.0761, "step": 4274 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.3106, "step": 4275 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.7551, "step": 4276 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.2472, "step": 4277 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.2711, "step": 4278 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.2503, "step": 4279 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.8499, "step": 4280 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.6469, "step": 4281 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.759, "step": 4282 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.719, "step": 4283 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.0007, "step": 4284 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.7745, "step": 4285 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.3657, "step": 4286 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.8479, "step": 4287 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.3329, "step": 4288 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.8175, "step": 4289 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.0979, "step": 4290 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.742, "step": 4291 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.5499, "step": 4292 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 0.7748, "step": 4293 }, { "epoch": 1.6, "learning_rate": 2e-05, "loss": 1.1315, "step": 4294 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.7536, "step": 4295 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.7194, "step": 4296 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 1.1576, "step": 4297 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8389, "step": 4298 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8518, "step": 4299 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8933, "step": 4300 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 1.111, "step": 4301 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8849, "step": 4302 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9739, "step": 4303 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 1.0157, "step": 4304 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8667, "step": 4305 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9669, "step": 4306 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.7553, "step": 4307 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8124, "step": 4308 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9167, "step": 4309 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8033, "step": 4310 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9591, "step": 4311 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9138, "step": 4312 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9231, "step": 4313 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.8318, "step": 4314 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.7198, "step": 4315 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9132, "step": 4316 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9853, "step": 4317 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.7268, "step": 4318 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 1.0107, "step": 4319 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 0.9527, "step": 4320 }, { "epoch": 1.61, "learning_rate": 2e-05, "loss": 1.2251, "step": 4321 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.9487, "step": 4322 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.0023, "step": 4323 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.9319, "step": 4324 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.9583, "step": 4325 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.163, "step": 4326 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.0492, "step": 4327 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.1489, "step": 4328 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.0151, "step": 4329 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.0439, "step": 4330 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.826, "step": 4331 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.017, "step": 4332 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.197, "step": 4333 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.811, "step": 4334 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.1861, "step": 4335 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.8152, "step": 4336 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.9114, "step": 4337 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.799, "step": 4338 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.3652, "step": 4339 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.1088, "step": 4340 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.4567, "step": 4341 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.0522, "step": 4342 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.7445, "step": 4343 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.8399, "step": 4344 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.857, "step": 4345 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 0.9388, "step": 4346 }, { "epoch": 1.62, "learning_rate": 2e-05, "loss": 1.2173, "step": 4347 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.0204, "step": 4348 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.7679, "step": 4349 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.1052, "step": 4350 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.9004, "step": 4351 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.9344, "step": 4352 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.996, "step": 4353 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.8069, "step": 4354 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.7832, "step": 4355 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.9414, "step": 4356 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.8085, "step": 4357 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.8664, "step": 4358 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.7418, "step": 4359 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.6793, "step": 4360 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.0347, "step": 4361 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.1484, "step": 4362 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.9313, "step": 4363 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.947, "step": 4364 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.0835, "step": 4365 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.0766, "step": 4366 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.99, "step": 4367 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 0.9339, "step": 4368 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.3962, "step": 4369 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.1259, "step": 4370 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.1971, "step": 4371 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.2042, "step": 4372 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.6089, "step": 4373 }, { "epoch": 1.63, "learning_rate": 2e-05, "loss": 1.0762, "step": 4374 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.7686, "step": 4375 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.8045, "step": 4376 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.9141, "step": 4377 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.8813, "step": 4378 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.1847, "step": 4379 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.3697, "step": 4380 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.2563, "step": 4381 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.7128, "step": 4382 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.9119, "step": 4383 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.3093, "step": 4384 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.9206, "step": 4385 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.256, "step": 4386 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.8501, "step": 4387 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.7833, "step": 4388 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.3578, "step": 4389 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.8189, "step": 4390 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.9497, "step": 4391 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.9313, "step": 4392 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.5348, "step": 4393 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.7858, "step": 4394 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.7768, "step": 4395 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.8356, "step": 4396 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.8371, "step": 4397 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.9852, "step": 4398 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.0813, "step": 4399 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 1.3179, "step": 4400 }, { "epoch": 1.64, "learning_rate": 2e-05, "loss": 0.8611, "step": 4401 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.8969, "step": 4402 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.0984, "step": 4403 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.7247, "step": 4404 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.1407, "step": 4405 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.9762, "step": 4406 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.0717, "step": 4407 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.9849, "step": 4408 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.2585, "step": 4409 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.817, "step": 4410 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.8944, "step": 4411 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.2967, "step": 4412 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.8698, "step": 4413 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.0884, "step": 4414 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.9873, "step": 4415 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.0852, "step": 4416 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.7751, "step": 4417 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.8788, "step": 4418 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.9814, "step": 4419 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.375, "step": 4420 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.0857, "step": 4421 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.7847, "step": 4422 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.8144, "step": 4423 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.3894, "step": 4424 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.7949, "step": 4425 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 0.9523, "step": 4426 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.2409, "step": 4427 }, { "epoch": 1.65, "learning_rate": 2e-05, "loss": 1.0719, "step": 4428 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.8325, "step": 4429 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 1.0177, "step": 4430 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.7463, "step": 4431 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.8465, "step": 4432 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 1.3706, "step": 4433 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 1.0043, "step": 4434 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 1.3395, "step": 4435 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.9413, "step": 4436 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.9255, "step": 4437 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 1.0258, "step": 4438 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 1.2828, "step": 4439 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.8641, "step": 4440 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.9255, "step": 4441 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.747, "step": 4442 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.6993, "step": 4443 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.9189, "step": 4444 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.7491, "step": 4445 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 1.1257, "step": 4446 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.8662, "step": 4447 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.7479, "step": 4448 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.7053, "step": 4449 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.9428, "step": 4450 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.923, "step": 4451 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.8663, "step": 4452 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.8638, "step": 4453 }, { "epoch": 1.66, "learning_rate": 2e-05, "loss": 0.8641, "step": 4454 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.8655, "step": 4455 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.8613, "step": 4456 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.9437, "step": 4457 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.0241, "step": 4458 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.6393, "step": 4459 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.4383, "step": 4460 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.8987, "step": 4461 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.9615, "step": 4462 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.4935, "step": 4463 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.7748, "step": 4464 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.008, "step": 4465 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.0376, "step": 4466 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.931, "step": 4467 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.9523, "step": 4468 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.2255, "step": 4469 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.8858, "step": 4470 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.1158, "step": 4471 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.0202, "step": 4472 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.6197, "step": 4473 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.921, "step": 4474 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.7705, "step": 4475 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.237, "step": 4476 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 1.1044, "step": 4477 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.9684, "step": 4478 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.973, "step": 4479 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.5864, "step": 4480 }, { "epoch": 1.67, "learning_rate": 2e-05, "loss": 0.8229, "step": 4481 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.8767, "step": 4482 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.8658, "step": 4483 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.8542, "step": 4484 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.7256, "step": 4485 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.8407, "step": 4486 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.9843, "step": 4487 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.7787, "step": 4488 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.3144, "step": 4489 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.7745, "step": 4490 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.0457, "step": 4491 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.9482, "step": 4492 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.8896, "step": 4493 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.8148, "step": 4494 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.7459, "step": 4495 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.1983, "step": 4496 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.1039, "step": 4497 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.1039, "step": 4498 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.9958, "step": 4499 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.9426, "step": 4500 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.938, "step": 4501 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.6923, "step": 4502 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.1089, "step": 4503 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.7539, "step": 4504 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.9371, "step": 4505 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.0577, "step": 4506 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 0.9483, "step": 4507 }, { "epoch": 1.68, "learning_rate": 2e-05, "loss": 1.03, "step": 4508 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8936, "step": 4509 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.5101, "step": 4510 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8436, "step": 4511 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.7873, "step": 4512 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.0655, "step": 4513 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.6638, "step": 4514 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.1414, "step": 4515 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8598, "step": 4516 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.0744, "step": 4517 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.0394, "step": 4518 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.7502, "step": 4519 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8338, "step": 4520 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.9776, "step": 4521 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.9194, "step": 4522 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.0463, "step": 4523 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8374, "step": 4524 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.9164, "step": 4525 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.7189, "step": 4526 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.7592, "step": 4527 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8499, "step": 4528 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.0124, "step": 4529 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.0219, "step": 4530 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8463, "step": 4531 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 1.0526, "step": 4532 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.6422, "step": 4533 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.9337, "step": 4534 }, { "epoch": 1.69, "learning_rate": 2e-05, "loss": 0.8435, "step": 4535 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.8929, "step": 4536 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.813, "step": 4537 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.0837, "step": 4538 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.137, "step": 4539 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.7215, "step": 4540 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.6577, "step": 4541 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.2063, "step": 4542 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.727, "step": 4543 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.3169, "step": 4544 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.6617, "step": 4545 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.7463, "step": 4546 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.6779, "step": 4547 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.2174, "step": 4548 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.0745, "step": 4549 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.8886, "step": 4550 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.8639, "step": 4551 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.0189, "step": 4552 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.0126, "step": 4553 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.9203, "step": 4554 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.7925, "step": 4555 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.8639, "step": 4556 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.6513, "step": 4557 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.3497, "step": 4558 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.6951, "step": 4559 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 0.9151, "step": 4560 }, { "epoch": 1.7, "learning_rate": 2e-05, "loss": 1.4227, "step": 4561 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.1328, "step": 4562 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.7312, "step": 4563 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.8893, "step": 4564 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.1089, "step": 4565 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.3319, "step": 4566 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.6941, "step": 4567 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.1482, "step": 4568 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.9495, "step": 4569 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.2023, "step": 4570 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.8581, "step": 4571 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.4128, "step": 4572 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 2.4393, "step": 4573 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.5459, "step": 4574 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.9928, "step": 4575 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.9702, "step": 4576 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.674, "step": 4577 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.3018, "step": 4578 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.1861, "step": 4579 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.8902, "step": 4580 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.9187, "step": 4581 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.8939, "step": 4582 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.9374, "step": 4583 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.8602, "step": 4584 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 1.0328, "step": 4585 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.8463, "step": 4586 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.9627, "step": 4587 }, { "epoch": 1.71, "learning_rate": 2e-05, "loss": 0.9206, "step": 4588 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.0307, "step": 4589 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.0724, "step": 4590 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.1682, "step": 4591 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.6835, "step": 4592 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.1105, "step": 4593 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.0668, "step": 4594 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.8526, "step": 4595 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.7121, "step": 4596 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.1428, "step": 4597 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.916, "step": 4598 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.9689, "step": 4599 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.0202, "step": 4600 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.8702, "step": 4601 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.3688, "step": 4602 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.8022, "step": 4603 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.8285, "step": 4604 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.8859, "step": 4605 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.1398, "step": 4606 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.94, "step": 4607 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.8139, "step": 4608 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.9258, "step": 4609 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.0888, "step": 4610 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.4864, "step": 4611 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.8233, "step": 4612 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 1.1398, "step": 4613 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.9894, "step": 4614 }, { "epoch": 1.72, "learning_rate": 2e-05, "loss": 0.9129, "step": 4615 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.3979, "step": 4616 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.764, "step": 4617 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.8748, "step": 4618 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.2818, "step": 4619 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.8671, "step": 4620 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9446, "step": 4621 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9853, "step": 4622 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9611, "step": 4623 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.8542, "step": 4624 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.2368, "step": 4625 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9175, "step": 4626 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9446, "step": 4627 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.1947, "step": 4628 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9026, "step": 4629 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.791, "step": 4630 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.874, "step": 4631 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.8742, "step": 4632 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9481, "step": 4633 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.4392, "step": 4634 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.2012, "step": 4635 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.154, "step": 4636 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9423, "step": 4637 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.0111, "step": 4638 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.1009, "step": 4639 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 0.9529, "step": 4640 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.2209, "step": 4641 }, { "epoch": 1.73, "learning_rate": 2e-05, "loss": 1.0076, "step": 4642 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.0413, "step": 4643 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.1624, "step": 4644 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9885, "step": 4645 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.7337, "step": 4646 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.1713, "step": 4647 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.0929, "step": 4648 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9907, "step": 4649 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.0594, "step": 4650 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.1024, "step": 4651 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.0601, "step": 4652 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9344, "step": 4653 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9104, "step": 4654 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.7704, "step": 4655 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9032, "step": 4656 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9066, "step": 4657 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.8213, "step": 4658 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.1611, "step": 4659 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9199, "step": 4660 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.1282, "step": 4661 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 1.26, "step": 4662 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.7935, "step": 4663 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.8239, "step": 4664 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.8206, "step": 4665 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.8825, "step": 4666 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.999, "step": 4667 }, { "epoch": 1.74, "learning_rate": 2e-05, "loss": 0.9365, "step": 4668 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.1321, "step": 4669 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.354, "step": 4670 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9503, "step": 4671 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.1066, "step": 4672 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.6785, "step": 4673 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.7487, "step": 4674 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9203, "step": 4675 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9224, "step": 4676 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.934, "step": 4677 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.902, "step": 4678 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9104, "step": 4679 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9196, "step": 4680 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.7765, "step": 4681 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9861, "step": 4682 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.6767, "step": 4683 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.0796, "step": 4684 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9719, "step": 4685 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.8581, "step": 4686 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.988, "step": 4687 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.1966, "step": 4688 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.697, "step": 4689 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.6515, "step": 4690 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.1972, "step": 4691 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.0458, "step": 4692 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9667, "step": 4693 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 0.9133, "step": 4694 }, { "epoch": 1.75, "learning_rate": 2e-05, "loss": 1.139, "step": 4695 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7962, "step": 4696 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.8686, "step": 4697 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.4774, "step": 4698 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7675, "step": 4699 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7974, "step": 4700 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.4038, "step": 4701 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.0687, "step": 4702 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.2356, "step": 4703 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7512, "step": 4704 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.879, "step": 4705 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.8192, "step": 4706 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7282, "step": 4707 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7001, "step": 4708 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.4564, "step": 4709 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.0798, "step": 4710 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.8959, "step": 4711 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.0681, "step": 4712 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.046, "step": 4713 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.1039, "step": 4714 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.6733, "step": 4715 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.9705, "step": 4716 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.3283, "step": 4717 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 1.4479, "step": 4718 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7232, "step": 4719 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.7185, "step": 4720 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.8036, "step": 4721 }, { "epoch": 1.76, "learning_rate": 2e-05, "loss": 0.9858, "step": 4722 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.8557, "step": 4723 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.6851, "step": 4724 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.9535, "step": 4725 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.7548, "step": 4726 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 1.4169, "step": 4727 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.868, "step": 4728 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.6367, "step": 4729 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.8391, "step": 4730 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 1.289, "step": 4731 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.7542, "step": 4732 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 1.1356, "step": 4733 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.7844, "step": 4734 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.9894, "step": 4735 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 1.2283, "step": 4736 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.7339, "step": 4737 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 1.2426, "step": 4738 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.9889, "step": 4739 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.9865, "step": 4740 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.9972, "step": 4741 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.9647, "step": 4742 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.9731, "step": 4743 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 1.1522, "step": 4744 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.8861, "step": 4745 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.7223, "step": 4746 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.8544, "step": 4747 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 1.0235, "step": 4748 }, { "epoch": 1.77, "learning_rate": 2e-05, "loss": 0.8929, "step": 4749 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.8414, "step": 4750 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 2.1744, "step": 4751 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.9785, "step": 4752 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.8009, "step": 4753 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.874, "step": 4754 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.0789, "step": 4755 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.8098, "step": 4756 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.1375, "step": 4757 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.7575, "step": 4758 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.1685, "step": 4759 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.846, "step": 4760 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.0576, "step": 4761 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.0116, "step": 4762 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.0664, "step": 4763 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.9882, "step": 4764 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.1212, "step": 4765 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.3293, "step": 4766 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.9037, "step": 4767 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.867, "step": 4768 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.3608, "step": 4769 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.8974, "step": 4770 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.9184, "step": 4771 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.3171, "step": 4772 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.7415, "step": 4773 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 1.0688, "step": 4774 }, { "epoch": 1.78, "learning_rate": 2e-05, "loss": 0.9778, "step": 4775 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.7278, "step": 4776 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.6886, "step": 4777 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.7351, "step": 4778 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.6913, "step": 4779 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.8724, "step": 4780 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.6031, "step": 4781 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.977, "step": 4782 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.8334, "step": 4783 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.7585, "step": 4784 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.0959, "step": 4785 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.8024, "step": 4786 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.8672, "step": 4787 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.185, "step": 4788 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.0715, "step": 4789 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.2304, "step": 4790 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.888, "step": 4791 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.1106, "step": 4792 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.012, "step": 4793 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.9385, "step": 4794 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.8826, "step": 4795 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.5556, "step": 4796 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 0.8944, "step": 4797 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.2119, "step": 4798 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.4505, "step": 4799 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.1004, "step": 4800 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.1854, "step": 4801 }, { "epoch": 1.79, "learning_rate": 2e-05, "loss": 1.0051, "step": 4802 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.8968, "step": 4803 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.8318, "step": 4804 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.9948, "step": 4805 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.1347, "step": 4806 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.9831, "step": 4807 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.7925, "step": 4808 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.2606, "step": 4809 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.8668, "step": 4810 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.5222, "step": 4811 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.2226, "step": 4812 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.8845, "step": 4813 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.0471, "step": 4814 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.0348, "step": 4815 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.8304, "step": 4816 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.0391, "step": 4817 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.1988, "step": 4818 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.9336, "step": 4819 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.9512, "step": 4820 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.917, "step": 4821 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.9191, "step": 4822 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.2507, "step": 4823 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.808, "step": 4824 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.0315, "step": 4825 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.8416, "step": 4826 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.2746, "step": 4827 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 1.1598, "step": 4828 }, { "epoch": 1.8, "learning_rate": 2e-05, "loss": 0.8283, "step": 4829 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.7292, "step": 4830 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.1159, "step": 4831 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.2741, "step": 4832 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.9097, "step": 4833 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.7356, "step": 4834 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.7158, "step": 4835 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.9756, "step": 4836 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.0063, "step": 4837 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.0956, "step": 4838 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.1673, "step": 4839 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.8115, "step": 4840 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.0672, "step": 4841 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.3568, "step": 4842 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.6798, "step": 4843 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.0126, "step": 4844 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.7922, "step": 4845 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.8298, "step": 4846 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.7283, "step": 4847 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.8474, "step": 4848 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.919, "step": 4849 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.9765, "step": 4850 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.9543, "step": 4851 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.8901, "step": 4852 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.3746, "step": 4853 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.8643, "step": 4854 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 1.0186, "step": 4855 }, { "epoch": 1.81, "learning_rate": 2e-05, "loss": 0.8618, "step": 4856 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.2531, "step": 4857 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.8111, "step": 4858 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.9406, "step": 4859 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.8458, "step": 4860 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.3784, "step": 4861 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.8829, "step": 4862 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.7854, "step": 4863 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.89, "step": 4864 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.9881, "step": 4865 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.9203, "step": 4866 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.9998, "step": 4867 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.0048, "step": 4868 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.835, "step": 4869 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.8704, "step": 4870 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.8853, "step": 4871 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.0702, "step": 4872 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.2195, "step": 4873 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.8381, "step": 4874 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.9249, "step": 4875 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.9171, "step": 4876 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.2742, "step": 4877 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.0037, "step": 4878 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.0559, "step": 4879 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 1.1217, "step": 4880 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.8229, "step": 4881 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.7518, "step": 4882 }, { "epoch": 1.82, "learning_rate": 2e-05, "loss": 0.969, "step": 4883 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9583, "step": 4884 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9605, "step": 4885 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 1.1589, "step": 4886 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9331, "step": 4887 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 1.122, "step": 4888 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 1.2569, "step": 4889 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9254, "step": 4890 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.8546, "step": 4891 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 1.0102, "step": 4892 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 1.2621, "step": 4893 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9085, "step": 4894 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9065, "step": 4895 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.6954, "step": 4896 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9822, "step": 4897 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.9464, "step": 4898 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.8043, "step": 4899 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.6718, "step": 4900 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.8269, "step": 4901 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.8764, "step": 4902 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 1.1425, "step": 4903 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.7397, "step": 4904 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.8985, "step": 4905 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.8051, "step": 4906 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.8032, "step": 4907 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 1.0242, "step": 4908 }, { "epoch": 1.83, "learning_rate": 2e-05, "loss": 0.819, "step": 4909 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.9619, "step": 4910 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.0313, "step": 4911 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.2969, "step": 4912 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.0199, "step": 4913 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.2043, "step": 4914 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.0659, "step": 4915 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.8599, "step": 4916 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.9248, "step": 4917 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.9197, "step": 4918 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.905, "step": 4919 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.7473, "step": 4920 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.9166, "step": 4921 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.0524, "step": 4922 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.7416, "step": 4923 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.2554, "step": 4924 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.0128, "step": 4925 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.9715, "step": 4926 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.9507, "step": 4927 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.8075, "step": 4928 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.7113, "step": 4929 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.0502, "step": 4930 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.8366, "step": 4931 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.1162, "step": 4932 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.3543, "step": 4933 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.9703, "step": 4934 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 1.3707, "step": 4935 }, { "epoch": 1.84, "learning_rate": 2e-05, "loss": 0.7345, "step": 4936 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.943, "step": 4937 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.6859, "step": 4938 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.9429, "step": 4939 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.8911, "step": 4940 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.9314, "step": 4941 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.0929, "step": 4942 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.9626, "step": 4943 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.9627, "step": 4944 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.2002, "step": 4945 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.1817, "step": 4946 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.106, "step": 4947 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.9464, "step": 4948 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.1333, "step": 4949 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.7977, "step": 4950 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.8656, "step": 4951 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.734, "step": 4952 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.9992, "step": 4953 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.8511, "step": 4954 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.9295, "step": 4955 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.121, "step": 4956 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.8531, "step": 4957 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.4566, "step": 4958 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.8352, "step": 4959 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.0637, "step": 4960 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.9104, "step": 4961 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 0.7381, "step": 4962 }, { "epoch": 1.85, "learning_rate": 2e-05, "loss": 1.0807, "step": 4963 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.1256, "step": 4964 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.0985, "step": 4965 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.8271, "step": 4966 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.499, "step": 4967 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.8565, "step": 4968 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.6695, "step": 4969 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.627, "step": 4970 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.7684, "step": 4971 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.8196, "step": 4972 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.7345, "step": 4973 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.826, "step": 4974 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.9035, "step": 4975 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.8964, "step": 4976 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.7602, "step": 4977 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.9989, "step": 4978 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.1979, "step": 4979 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.8469, "step": 4980 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.7051, "step": 4981 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.7474, "step": 4982 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.569, "step": 4983 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.7793, "step": 4984 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.0814, "step": 4985 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.0216, "step": 4986 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.7795, "step": 4987 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.9762, "step": 4988 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 0.9484, "step": 4989 }, { "epoch": 1.86, "learning_rate": 2e-05, "loss": 1.1481, "step": 4990 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.8487, "step": 4991 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.9899, "step": 4992 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.9663, "step": 4993 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.0126, "step": 4994 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.7889, "step": 4995 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.2781, "step": 4996 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.449, "step": 4997 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.3379, "step": 4998 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.2231, "step": 4999 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.8402, "step": 5000 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.8752, "step": 5001 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.92, "step": 5002 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.0644, "step": 5003 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.0304, "step": 5004 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.6556, "step": 5005 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.9678, "step": 5006 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.9601, "step": 5007 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.7674, "step": 5008 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.2619, "step": 5009 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.9664, "step": 5010 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.8841, "step": 5011 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.1869, "step": 5012 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.1175, "step": 5013 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 0.7999, "step": 5014 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.1939, "step": 5015 }, { "epoch": 1.87, "learning_rate": 2e-05, "loss": 1.3072, "step": 5016 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.0219, "step": 5017 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.6872, "step": 5018 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.0884, "step": 5019 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.2231, "step": 5020 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.7677, "step": 5021 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.0749, "step": 5022 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.3485, "step": 5023 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.9673, "step": 5024 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.9385, "step": 5025 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.8651, "step": 5026 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.8072, "step": 5027 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.7324, "step": 5028 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.7862, "step": 5029 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.6989, "step": 5030 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.2305, "step": 5031 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.0929, "step": 5032 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.8854, "step": 5033 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.9105, "step": 5034 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.7472, "step": 5035 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.9285, "step": 5036 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.8749, "step": 5037 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.1415, "step": 5038 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.7164, "step": 5039 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 0.9054, "step": 5040 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.1026, "step": 5041 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.1249, "step": 5042 }, { "epoch": 1.88, "learning_rate": 2e-05, "loss": 1.2954, "step": 5043 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.7677, "step": 5044 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.8587, "step": 5045 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.7428, "step": 5046 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.7695, "step": 5047 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.8846, "step": 5048 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.1158, "step": 5049 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.0776, "step": 5050 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.7838, "step": 5051 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.2534, "step": 5052 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.8695, "step": 5053 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.2315, "step": 5054 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.8898, "step": 5055 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.0242, "step": 5056 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.9462, "step": 5057 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.0024, "step": 5058 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.7668, "step": 5059 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.8204, "step": 5060 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.964, "step": 5061 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.9497, "step": 5062 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.867, "step": 5063 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.002, "step": 5064 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.8073, "step": 5065 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.0764, "step": 5066 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.2704, "step": 5067 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.9798, "step": 5068 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 1.3811, "step": 5069 }, { "epoch": 1.89, "learning_rate": 2e-05, "loss": 0.8301, "step": 5070 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.9252, "step": 5071 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.735, "step": 5072 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8756, "step": 5073 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.144, "step": 5074 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8744, "step": 5075 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.0673, "step": 5076 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.032, "step": 5077 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.9537, "step": 5078 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.1543, "step": 5079 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8115, "step": 5080 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8922, "step": 5081 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.7886, "step": 5082 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.1476, "step": 5083 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.0482, "step": 5084 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8038, "step": 5085 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.8006, "step": 5086 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.9737, "step": 5087 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.0755, "step": 5088 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.7006, "step": 5089 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.1021, "step": 5090 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.9553, "step": 5091 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8481, "step": 5092 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.9622, "step": 5093 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8624, "step": 5094 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.2156, "step": 5095 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 0.8511, "step": 5096 }, { "epoch": 1.9, "learning_rate": 2e-05, "loss": 1.1094, "step": 5097 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.0439, "step": 5098 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8689, "step": 5099 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.9474, "step": 5100 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.9992, "step": 5101 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8137, "step": 5102 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.0587, "step": 5103 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8165, "step": 5104 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8002, "step": 5105 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.3166, "step": 5106 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.7554, "step": 5107 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8165, "step": 5108 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8131, "step": 5109 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.7227, "step": 5110 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.9416, "step": 5111 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8335, "step": 5112 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.2456, "step": 5113 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.0146, "step": 5114 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.9261, "step": 5115 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.4915, "step": 5116 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.9827, "step": 5117 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.7438, "step": 5118 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8852, "step": 5119 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.9066, "step": 5120 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.0196, "step": 5121 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 0.8169, "step": 5122 }, { "epoch": 1.91, "learning_rate": 2e-05, "loss": 1.0913, "step": 5123 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.0954, "step": 5124 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.1898, "step": 5125 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.9066, "step": 5126 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.9945, "step": 5127 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.0023, "step": 5128 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.8828, "step": 5129 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.8556, "step": 5130 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.0453, "step": 5131 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.7638, "step": 5132 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.1503, "step": 5133 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.7079, "step": 5134 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.5473, "step": 5135 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.1293, "step": 5136 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.6935, "step": 5137 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.9762, "step": 5138 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.1237, "step": 5139 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.8228, "step": 5140 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.7999, "step": 5141 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.9746, "step": 5142 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.9709, "step": 5143 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.897, "step": 5144 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.8226, "step": 5145 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.7692, "step": 5146 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.9072, "step": 5147 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 0.729, "step": 5148 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.1427, "step": 5149 }, { "epoch": 1.92, "learning_rate": 2e-05, "loss": 1.0832, "step": 5150 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.7139, "step": 5151 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.9305, "step": 5152 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.7667, "step": 5153 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.8545, "step": 5154 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.2093, "step": 5155 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.0111, "step": 5156 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.1103, "step": 5157 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.9337, "step": 5158 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.0832, "step": 5159 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.7399, "step": 5160 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.6538, "step": 5161 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.7211, "step": 5162 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.7066, "step": 5163 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.9064, "step": 5164 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.1068, "step": 5165 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.6219, "step": 5166 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.8143, "step": 5167 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.1445, "step": 5168 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.9289, "step": 5169 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.0821, "step": 5170 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.0428, "step": 5171 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.8927, "step": 5172 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.7777, "step": 5173 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.8367, "step": 5174 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 1.1468, "step": 5175 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.9743, "step": 5176 }, { "epoch": 1.93, "learning_rate": 2e-05, "loss": 0.9089, "step": 5177 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.952, "step": 5178 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.8673, "step": 5179 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 1.0143, "step": 5180 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 1.1375, "step": 5181 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.7935, "step": 5182 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 1.0082, "step": 5183 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.8699, "step": 5184 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 1.0255, "step": 5185 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.776, "step": 5186 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.6951, "step": 5187 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.8608, "step": 5188 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.8538, "step": 5189 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.9994, "step": 5190 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.8176, "step": 5191 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.8431, "step": 5192 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.7033, "step": 5193 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.5451, "step": 5194 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 1.0179, "step": 5195 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.9439, "step": 5196 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.7998, "step": 5197 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.9245, "step": 5198 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.6413, "step": 5199 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.9374, "step": 5200 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 1.5779, "step": 5201 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.6983, "step": 5202 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 0.7555, "step": 5203 }, { "epoch": 1.94, "learning_rate": 2e-05, "loss": 1.1123, "step": 5204 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.9834, "step": 5205 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.885, "step": 5206 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.1294, "step": 5207 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8886, "step": 5208 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.7478, "step": 5209 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.729, "step": 5210 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.1576, "step": 5211 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.7791, "step": 5212 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.1334, "step": 5213 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8095, "step": 5214 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.276, "step": 5215 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8416, "step": 5216 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.9603, "step": 5217 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8219, "step": 5218 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.748, "step": 5219 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8552, "step": 5220 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.9958, "step": 5221 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.1024, "step": 5222 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.6988, "step": 5223 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.76, "step": 5224 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8544, "step": 5225 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8689, "step": 5226 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.077, "step": 5227 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.7497, "step": 5228 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 1.1148, "step": 5229 }, { "epoch": 1.95, "learning_rate": 2e-05, "loss": 0.8559, "step": 5230 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.1151, "step": 5231 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.8656, "step": 5232 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.7967, "step": 5233 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.8923, "step": 5234 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.9194, "step": 5235 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.6845, "step": 5236 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.967, "step": 5237 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.8523, "step": 5238 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.9987, "step": 5239 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.9216, "step": 5240 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.0749, "step": 5241 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.8529, "step": 5242 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.7863, "step": 5243 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.479, "step": 5244 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.1069, "step": 5245 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.858, "step": 5246 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.1211, "step": 5247 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.8237, "step": 5248 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.9499, "step": 5249 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.0358, "step": 5250 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.849, "step": 5251 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.9503, "step": 5252 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.0802, "step": 5253 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 1.3892, "step": 5254 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.8891, "step": 5255 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.7835, "step": 5256 }, { "epoch": 1.96, "learning_rate": 2e-05, "loss": 0.9531, "step": 5257 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 2.0087, "step": 5258 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.7373, "step": 5259 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.8869, "step": 5260 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.8401, "step": 5261 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.0144, "step": 5262 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.1446, "step": 5263 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.145, "step": 5264 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.7448, "step": 5265 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.9374, "step": 5266 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.0827, "step": 5267 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.4747, "step": 5268 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.7185, "step": 5269 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.7525, "step": 5270 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.8231, "step": 5271 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.8227, "step": 5272 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.906, "step": 5273 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.8453, "step": 5274 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.2691, "step": 5275 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.2441, "step": 5276 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.336, "step": 5277 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.9117, "step": 5278 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.2799, "step": 5279 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.9105, "step": 5280 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.9779, "step": 5281 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 0.9055, "step": 5282 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.2032, "step": 5283 }, { "epoch": 1.97, "learning_rate": 2e-05, "loss": 1.1389, "step": 5284 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.9269, "step": 5285 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.9142, "step": 5286 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.7694, "step": 5287 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.9967, "step": 5288 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.1854, "step": 5289 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.7682, "step": 5290 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.7928, "step": 5291 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.8295, "step": 5292 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.7625, "step": 5293 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.0268, "step": 5294 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.805, "step": 5295 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.1444, "step": 5296 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.85, "step": 5297 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.8479, "step": 5298 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.8576, "step": 5299 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.8537, "step": 5300 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.2307, "step": 5301 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.0778, "step": 5302 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.7931, "step": 5303 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.6997, "step": 5304 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.007, "step": 5305 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.9507, "step": 5306 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.9365, "step": 5307 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.9769, "step": 5308 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 0.9302, "step": 5309 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.1294, "step": 5310 }, { "epoch": 1.98, "learning_rate": 2e-05, "loss": 1.2054, "step": 5311 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.1561, "step": 5312 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.0813, "step": 5313 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.1111, "step": 5314 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.1122, "step": 5315 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.6162, "step": 5316 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.2146, "step": 5317 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.6261, "step": 5318 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.207, "step": 5319 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.2127, "step": 5320 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.151, "step": 5321 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.1198, "step": 5322 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.319, "step": 5323 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.1465, "step": 5324 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.7386, "step": 5325 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.2747, "step": 5326 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.7515, "step": 5327 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.9556, "step": 5328 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.8143, "step": 5329 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.9416, "step": 5330 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.6854, "step": 5331 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.8822, "step": 5332 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.0001, "step": 5333 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.8052, "step": 5334 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.8604, "step": 5335 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 1.1073, "step": 5336 }, { "epoch": 1.99, "learning_rate": 2e-05, "loss": 0.7181, "step": 5337 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.797, "step": 5338 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.8973, "step": 5339 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.8752, "step": 5340 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.9296, "step": 5341 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.8938, "step": 5342 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 1.2295, "step": 5343 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.8478, "step": 5344 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.7115, "step": 5345 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.8236, "step": 5346 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.784, "step": 5347 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.7165, "step": 5348 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.94, "step": 5349 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.9741, "step": 5350 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.8971, "step": 5351 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.607, "step": 5352 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.5201, "step": 5353 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.4569, "step": 5354 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.4574, "step": 5355 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.3953, "step": 5356 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.3353, "step": 5357 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.3928, "step": 5358 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.3563, "step": 5359 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.3955, "step": 5360 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.4281, "step": 5361 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.7514, "step": 5362 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.3879, "step": 5363 }, { "epoch": 2.0, "learning_rate": 2e-05, "loss": 0.3814, "step": 5364 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3003, "step": 5365 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3625, "step": 5366 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3517, "step": 5367 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3618, "step": 5368 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.4393, "step": 5369 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.5737, "step": 5370 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3833, "step": 5371 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 1.0597, "step": 5372 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.4631, "step": 5373 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.2962, "step": 5374 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.6289, "step": 5375 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.4264, "step": 5376 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.5169, "step": 5377 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.324, "step": 5378 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3598, "step": 5379 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3859, "step": 5380 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.4227, "step": 5381 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.5555, "step": 5382 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.5031, "step": 5383 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.6033, "step": 5384 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3191, "step": 5385 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.4919, "step": 5386 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3445, "step": 5387 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.4267, "step": 5388 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.4171, "step": 5389 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.3896, "step": 5390 }, { "epoch": 2.01, "learning_rate": 2e-05, "loss": 0.5127, "step": 5391 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3662, "step": 5392 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4176, "step": 5393 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3841, "step": 5394 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4657, "step": 5395 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3531, "step": 5396 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4566, "step": 5397 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4781, "step": 5398 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3634, "step": 5399 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.6265, "step": 5400 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.5153, "step": 5401 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.349, "step": 5402 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3615, "step": 5403 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3466, "step": 5404 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4203, "step": 5405 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4312, "step": 5406 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.494, "step": 5407 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4523, "step": 5408 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3336, "step": 5409 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4044, "step": 5410 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3882, "step": 5411 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4525, "step": 5412 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3449, "step": 5413 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3244, "step": 5414 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4672, "step": 5415 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.4189, "step": 5416 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.5739, "step": 5417 }, { "epoch": 2.02, "learning_rate": 2e-05, "loss": 0.3971, "step": 5418 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3565, "step": 5419 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4047, "step": 5420 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.402, "step": 5421 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.2818, "step": 5422 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3692, "step": 5423 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4611, "step": 5424 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4024, "step": 5425 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4483, "step": 5426 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.5557, "step": 5427 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3504, "step": 5428 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4133, "step": 5429 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.5164, "step": 5430 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4114, "step": 5431 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4632, "step": 5432 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.313, "step": 5433 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3523, "step": 5434 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4219, "step": 5435 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4117, "step": 5436 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3027, "step": 5437 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3243, "step": 5438 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4803, "step": 5439 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3018, "step": 5440 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4942, "step": 5441 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4499, "step": 5442 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.4066, "step": 5443 }, { "epoch": 2.03, "learning_rate": 2e-05, "loss": 0.3681, "step": 5444 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.517, "step": 5445 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.5139, "step": 5446 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.4315, "step": 5447 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.436, "step": 5448 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.2883, "step": 5449 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.4818, "step": 5450 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.325, "step": 5451 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.377, "step": 5452 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3461, "step": 5453 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3887, "step": 5454 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3512, "step": 5455 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.4112, "step": 5456 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3923, "step": 5457 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.5529, "step": 5458 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.359, "step": 5459 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3872, "step": 5460 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.357, "step": 5461 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.4022, "step": 5462 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.391, "step": 5463 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.4503, "step": 5464 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3463, "step": 5465 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3866, "step": 5466 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3496, "step": 5467 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3137, "step": 5468 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.4946, "step": 5469 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.4859, "step": 5470 }, { "epoch": 2.04, "learning_rate": 2e-05, "loss": 0.3341, "step": 5471 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3809, "step": 5472 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.391, "step": 5473 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3929, "step": 5474 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3902, "step": 5475 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4534, "step": 5476 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4033, "step": 5477 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4393, "step": 5478 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3741, "step": 5479 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4407, "step": 5480 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.2308, "step": 5481 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.6162, "step": 5482 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4816, "step": 5483 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.402, "step": 5484 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.5061, "step": 5485 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3469, "step": 5486 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3181, "step": 5487 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.6531, "step": 5488 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3445, "step": 5489 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.6637, "step": 5490 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3411, "step": 5491 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4102, "step": 5492 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4796, "step": 5493 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3894, "step": 5494 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3905, "step": 5495 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.5091, "step": 5496 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.4123, "step": 5497 }, { "epoch": 2.05, "learning_rate": 2e-05, "loss": 0.3348, "step": 5498 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3737, "step": 5499 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.519, "step": 5500 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.348, "step": 5501 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3917, "step": 5502 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3575, "step": 5503 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.318, "step": 5504 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3348, "step": 5505 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3797, "step": 5506 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.2853, "step": 5507 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3291, "step": 5508 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3739, "step": 5509 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4317, "step": 5510 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4783, "step": 5511 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3997, "step": 5512 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4511, "step": 5513 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.5427, "step": 5514 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.2516, "step": 5515 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4832, "step": 5516 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.7015, "step": 5517 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3568, "step": 5518 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4062, "step": 5519 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.3057, "step": 5520 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4188, "step": 5521 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4154, "step": 5522 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.5926, "step": 5523 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4506, "step": 5524 }, { "epoch": 2.06, "learning_rate": 2e-05, "loss": 0.4358, "step": 5525 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4731, "step": 5526 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3905, "step": 5527 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3956, "step": 5528 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4689, "step": 5529 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.5135, "step": 5530 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3576, "step": 5531 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.337, "step": 5532 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4485, "step": 5533 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3992, "step": 5534 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.5908, "step": 5535 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3574, "step": 5536 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3169, "step": 5537 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4525, "step": 5538 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3909, "step": 5539 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3846, "step": 5540 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4309, "step": 5541 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3955, "step": 5542 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4287, "step": 5543 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.399, "step": 5544 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4266, "step": 5545 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3242, "step": 5546 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3997, "step": 5547 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.2881, "step": 5548 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.4388, "step": 5549 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3794, "step": 5550 }, { "epoch": 2.07, "learning_rate": 2e-05, "loss": 0.3237, "step": 5551 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3113, "step": 5552 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3095, "step": 5553 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4361, "step": 5554 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3818, "step": 5555 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4711, "step": 5556 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3915, "step": 5557 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3017, "step": 5558 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4562, "step": 5559 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4401, "step": 5560 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4363, "step": 5561 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4863, "step": 5562 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4776, "step": 5563 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4282, "step": 5564 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3991, "step": 5565 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4031, "step": 5566 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3211, "step": 5567 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4823, "step": 5568 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.5292, "step": 5569 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.382, "step": 5570 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.7387, "step": 5571 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4338, "step": 5572 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3886, "step": 5573 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.388, "step": 5574 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.2484, "step": 5575 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.3779, "step": 5576 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.5112, "step": 5577 }, { "epoch": 2.08, "learning_rate": 2e-05, "loss": 0.4705, "step": 5578 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.6162, "step": 5579 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.2799, "step": 5580 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3765, "step": 5581 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4892, "step": 5582 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4042, "step": 5583 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4977, "step": 5584 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3046, "step": 5585 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4361, "step": 5586 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3747, "step": 5587 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4457, "step": 5588 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.5613, "step": 5589 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3748, "step": 5590 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3664, "step": 5591 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3537, "step": 5592 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.2765, "step": 5593 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3444, "step": 5594 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3637, "step": 5595 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3565, "step": 5596 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.2786, "step": 5597 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3656, "step": 5598 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4018, "step": 5599 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4426, "step": 5600 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4226, "step": 5601 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3245, "step": 5602 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4258, "step": 5603 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.4309, "step": 5604 }, { "epoch": 2.09, "learning_rate": 2e-05, "loss": 0.3798, "step": 5605 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3266, "step": 5606 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3118, "step": 5607 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3676, "step": 5608 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.2886, "step": 5609 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.6568, "step": 5610 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.5218, "step": 5611 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.356, "step": 5612 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3768, "step": 5613 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3629, "step": 5614 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.2904, "step": 5615 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3841, "step": 5616 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3383, "step": 5617 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3207, "step": 5618 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.2814, "step": 5619 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.5152, "step": 5620 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3155, "step": 5621 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.5357, "step": 5622 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4033, "step": 5623 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.5628, "step": 5624 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.3797, "step": 5625 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4373, "step": 5626 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4006, "step": 5627 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4181, "step": 5628 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4602, "step": 5629 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4675, "step": 5630 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4158, "step": 5631 }, { "epoch": 2.1, "learning_rate": 2e-05, "loss": 0.4199, "step": 5632 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.2765, "step": 5633 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4041, "step": 5634 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3409, "step": 5635 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4282, "step": 5636 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4018, "step": 5637 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3636, "step": 5638 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.509, "step": 5639 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3878, "step": 5640 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4188, "step": 5641 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.439, "step": 5642 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3176, "step": 5643 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4171, "step": 5644 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4108, "step": 5645 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.5665, "step": 5646 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.2958, "step": 5647 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3235, "step": 5648 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4813, "step": 5649 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.8018, "step": 5650 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4513, "step": 5651 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.352, "step": 5652 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4775, "step": 5653 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4121, "step": 5654 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3567, "step": 5655 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3936, "step": 5656 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.3524, "step": 5657 }, { "epoch": 2.11, "learning_rate": 2e-05, "loss": 0.4655, "step": 5658 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3365, "step": 5659 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3036, "step": 5660 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.4446, "step": 5661 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.4503, "step": 5662 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.5943, "step": 5663 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.5181, "step": 5664 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.4121, "step": 5665 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.2612, "step": 5666 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3579, "step": 5667 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3842, "step": 5668 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3603, "step": 5669 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.499, "step": 5670 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.6992, "step": 5671 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.4744, "step": 5672 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.5294, "step": 5673 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.5166, "step": 5674 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.6653, "step": 5675 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3966, "step": 5676 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.4223, "step": 5677 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3713, "step": 5678 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3882, "step": 5679 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3825, "step": 5680 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3161, "step": 5681 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.3348, "step": 5682 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.4003, "step": 5683 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.5872, "step": 5684 }, { "epoch": 2.12, "learning_rate": 2e-05, "loss": 0.475, "step": 5685 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4927, "step": 5686 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3029, "step": 5687 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4497, "step": 5688 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3441, "step": 5689 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3432, "step": 5690 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.2936, "step": 5691 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.5242, "step": 5692 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4048, "step": 5693 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4664, "step": 5694 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3955, "step": 5695 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.5305, "step": 5696 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4026, "step": 5697 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.376, "step": 5698 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3596, "step": 5699 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3582, "step": 5700 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.5742, "step": 5701 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4512, "step": 5702 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4288, "step": 5703 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.5308, "step": 5704 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.2936, "step": 5705 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4664, "step": 5706 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4707, "step": 5707 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.4395, "step": 5708 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3697, "step": 5709 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.3447, "step": 5710 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.5035, "step": 5711 }, { "epoch": 2.13, "learning_rate": 2e-05, "loss": 0.5342, "step": 5712 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.328, "step": 5713 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.2773, "step": 5714 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3877, "step": 5715 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.5298, "step": 5716 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.4503, "step": 5717 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.4205, "step": 5718 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3276, "step": 5719 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.5432, "step": 5720 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3456, "step": 5721 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.485, "step": 5722 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3759, "step": 5723 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.4837, "step": 5724 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3541, "step": 5725 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.5535, "step": 5726 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.4471, "step": 5727 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3756, "step": 5728 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.4189, "step": 5729 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3431, "step": 5730 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3235, "step": 5731 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.4471, "step": 5732 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.4411, "step": 5733 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.279, "step": 5734 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.5239, "step": 5735 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.5503, "step": 5736 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3265, "step": 5737 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3586, "step": 5738 }, { "epoch": 2.14, "learning_rate": 2e-05, "loss": 0.3896, "step": 5739 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3598, "step": 5740 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.4783, "step": 5741 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.488, "step": 5742 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.5346, "step": 5743 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.327, "step": 5744 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3675, "step": 5745 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3779, "step": 5746 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3879, "step": 5747 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.4257, "step": 5748 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.405, "step": 5749 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.5161, "step": 5750 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3498, "step": 5751 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.5229, "step": 5752 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.4405, "step": 5753 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3347, "step": 5754 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.355, "step": 5755 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.4565, "step": 5756 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3419, "step": 5757 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.4126, "step": 5758 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3908, "step": 5759 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3867, "step": 5760 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.5084, "step": 5761 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3762, "step": 5762 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.2508, "step": 5763 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.4178, "step": 5764 }, { "epoch": 2.15, "learning_rate": 2e-05, "loss": 0.3551, "step": 5765 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.5471, "step": 5766 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4147, "step": 5767 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4635, "step": 5768 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3681, "step": 5769 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3908, "step": 5770 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.7367, "step": 5771 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3287, "step": 5772 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4969, "step": 5773 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3488, "step": 5774 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3411, "step": 5775 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4276, "step": 5776 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.2943, "step": 5777 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3075, "step": 5778 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4288, "step": 5779 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4949, "step": 5780 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3148, "step": 5781 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4476, "step": 5782 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4124, "step": 5783 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.5604, "step": 5784 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4012, "step": 5785 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.2965, "step": 5786 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3419, "step": 5787 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4272, "step": 5788 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.3612, "step": 5789 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.4193, "step": 5790 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.5206, "step": 5791 }, { "epoch": 2.16, "learning_rate": 2e-05, "loss": 0.361, "step": 5792 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.8675, "step": 5793 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3618, "step": 5794 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.388, "step": 5795 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3388, "step": 5796 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3685, "step": 5797 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.351, "step": 5798 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4321, "step": 5799 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.337, "step": 5800 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4186, "step": 5801 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.5082, "step": 5802 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.2658, "step": 5803 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.509, "step": 5804 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3985, "step": 5805 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3586, "step": 5806 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3143, "step": 5807 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3785, "step": 5808 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3188, "step": 5809 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4061, "step": 5810 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.405, "step": 5811 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4172, "step": 5812 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.201, "step": 5813 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3639, "step": 5814 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.3581, "step": 5815 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4629, "step": 5816 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4998, "step": 5817 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4385, "step": 5818 }, { "epoch": 2.17, "learning_rate": 2e-05, "loss": 0.4868, "step": 5819 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.2909, "step": 5820 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3911, "step": 5821 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3499, "step": 5822 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.5997, "step": 5823 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.2918, "step": 5824 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3346, "step": 5825 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.4636, "step": 5826 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.2722, "step": 5827 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.4653, "step": 5828 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3768, "step": 5829 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.5411, "step": 5830 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3674, "step": 5831 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3759, "step": 5832 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3372, "step": 5833 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.4194, "step": 5834 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.392, "step": 5835 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.264, "step": 5836 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3067, "step": 5837 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3709, "step": 5838 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3752, "step": 5839 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.543, "step": 5840 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3328, "step": 5841 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3576, "step": 5842 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.4008, "step": 5843 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3223, "step": 5844 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3493, "step": 5845 }, { "epoch": 2.18, "learning_rate": 2e-05, "loss": 0.3414, "step": 5846 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3541, "step": 5847 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.4237, "step": 5848 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3834, "step": 5849 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.4329, "step": 5850 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.7559, "step": 5851 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3789, "step": 5852 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3646, "step": 5853 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.4401, "step": 5854 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.277, "step": 5855 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3622, "step": 5856 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.4177, "step": 5857 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3478, "step": 5858 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.5905, "step": 5859 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.5187, "step": 5860 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3202, "step": 5861 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.6154, "step": 5862 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.68, "step": 5863 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.4943, "step": 5864 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3894, "step": 5865 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.4529, "step": 5866 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.5827, "step": 5867 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3702, "step": 5868 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.5899, "step": 5869 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3139, "step": 5870 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.5794, "step": 5871 }, { "epoch": 2.19, "learning_rate": 2e-05, "loss": 0.3136, "step": 5872 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.419, "step": 5873 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3684, "step": 5874 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4138, "step": 5875 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.346, "step": 5876 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3216, "step": 5877 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3393, "step": 5878 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.6012, "step": 5879 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4398, "step": 5880 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.6475, "step": 5881 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4087, "step": 5882 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3717, "step": 5883 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.5127, "step": 5884 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4416, "step": 5885 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3478, "step": 5886 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4396, "step": 5887 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.6163, "step": 5888 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4002, "step": 5889 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4767, "step": 5890 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.266, "step": 5891 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.333, "step": 5892 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.2774, "step": 5893 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.6057, "step": 5894 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3284, "step": 5895 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3452, "step": 5896 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.4256, "step": 5897 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.3457, "step": 5898 }, { "epoch": 2.2, "learning_rate": 2e-05, "loss": 0.634, "step": 5899 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4006, "step": 5900 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4494, "step": 5901 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.3326, "step": 5902 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4232, "step": 5903 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4868, "step": 5904 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4074, "step": 5905 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.6458, "step": 5906 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.2944, "step": 5907 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.3646, "step": 5908 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.2948, "step": 5909 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.3241, "step": 5910 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4149, "step": 5911 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.5227, "step": 5912 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.3717, "step": 5913 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4345, "step": 5914 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4945, "step": 5915 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4812, "step": 5916 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.3321, "step": 5917 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4776, "step": 5918 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4574, "step": 5919 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4615, "step": 5920 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.2693, "step": 5921 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.3971, "step": 5922 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4127, "step": 5923 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.4894, "step": 5924 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.3962, "step": 5925 }, { "epoch": 2.21, "learning_rate": 2e-05, "loss": 0.5185, "step": 5926 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3836, "step": 5927 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4948, "step": 5928 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4072, "step": 5929 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3431, "step": 5930 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4725, "step": 5931 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3281, "step": 5932 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.5855, "step": 5933 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3668, "step": 5934 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3232, "step": 5935 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.5371, "step": 5936 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4788, "step": 5937 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.6984, "step": 5938 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4545, "step": 5939 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3634, "step": 5940 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.46, "step": 5941 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3904, "step": 5942 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.7199, "step": 5943 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4412, "step": 5944 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.6072, "step": 5945 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4231, "step": 5946 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4087, "step": 5947 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.6001, "step": 5948 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.333, "step": 5949 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4096, "step": 5950 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.4302, "step": 5951 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3438, "step": 5952 }, { "epoch": 2.22, "learning_rate": 2e-05, "loss": 0.3895, "step": 5953 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.4504, "step": 5954 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.386, "step": 5955 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3471, "step": 5956 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3024, "step": 5957 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3956, "step": 5958 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.7906, "step": 5959 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3353, "step": 5960 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.4379, "step": 5961 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3957, "step": 5962 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3565, "step": 5963 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.5185, "step": 5964 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3498, "step": 5965 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.4267, "step": 5966 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.5999, "step": 5967 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.5293, "step": 5968 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.585, "step": 5969 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.4527, "step": 5970 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3298, "step": 5971 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.37, "step": 5972 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3397, "step": 5973 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.4151, "step": 5974 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3829, "step": 5975 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3528, "step": 5976 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.4777, "step": 5977 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.7187, "step": 5978 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.3405, "step": 5979 }, { "epoch": 2.23, "learning_rate": 2e-05, "loss": 0.2688, "step": 5980 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.4804, "step": 5981 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.6969, "step": 5982 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.499, "step": 5983 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3251, "step": 5984 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3832, "step": 5985 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3173, "step": 5986 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.4099, "step": 5987 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.551, "step": 5988 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.6248, "step": 5989 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.4011, "step": 5990 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3521, "step": 5991 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.4714, "step": 5992 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3482, "step": 5993 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.431, "step": 5994 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.476, "step": 5995 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3875, "step": 5996 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.5383, "step": 5997 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.5111, "step": 5998 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.57, "step": 5999 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3499, "step": 6000 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3724, "step": 6001 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.4888, "step": 6002 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.4243, "step": 6003 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3563, "step": 6004 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.4186, "step": 6005 }, { "epoch": 2.24, "learning_rate": 2e-05, "loss": 0.3809, "step": 6006 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3111, "step": 6007 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3859, "step": 6008 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.5621, "step": 6009 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3644, "step": 6010 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4601, "step": 6011 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3591, "step": 6012 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3805, "step": 6013 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4456, "step": 6014 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4191, "step": 6015 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4554, "step": 6016 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.5572, "step": 6017 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3223, "step": 6018 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4721, "step": 6019 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.445, "step": 6020 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4249, "step": 6021 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.2879, "step": 6022 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.5554, "step": 6023 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3634, "step": 6024 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3875, "step": 6025 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3524, "step": 6026 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4017, "step": 6027 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.4799, "step": 6028 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3386, "step": 6029 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3913, "step": 6030 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.2458, "step": 6031 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.3723, "step": 6032 }, { "epoch": 2.25, "learning_rate": 2e-05, "loss": 0.5039, "step": 6033 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.5017, "step": 6034 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3378, "step": 6035 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.5597, "step": 6036 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3405, "step": 6037 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.2778, "step": 6038 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3503, "step": 6039 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.5651, "step": 6040 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3488, "step": 6041 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.432, "step": 6042 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.9566, "step": 6043 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3613, "step": 6044 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3321, "step": 6045 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.4659, "step": 6046 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3783, "step": 6047 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.4084, "step": 6048 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3638, "step": 6049 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.4307, "step": 6050 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3748, "step": 6051 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3851, "step": 6052 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.4205, "step": 6053 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.5116, "step": 6054 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.5067, "step": 6055 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.4785, "step": 6056 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.5315, "step": 6057 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3119, "step": 6058 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.3989, "step": 6059 }, { "epoch": 2.26, "learning_rate": 2e-05, "loss": 0.7819, "step": 6060 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3695, "step": 6061 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.2976, "step": 6062 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3128, "step": 6063 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4402, "step": 6064 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.507, "step": 6065 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.477, "step": 6066 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.6519, "step": 6067 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4611, "step": 6068 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3723, "step": 6069 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4091, "step": 6070 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3109, "step": 6071 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.5249, "step": 6072 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.47, "step": 6073 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4093, "step": 6074 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.439, "step": 6075 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3002, "step": 6076 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4673, "step": 6077 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4971, "step": 6078 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3411, "step": 6079 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3778, "step": 6080 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.484, "step": 6081 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.342, "step": 6082 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4497, "step": 6083 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4481, "step": 6084 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.353, "step": 6085 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.3289, "step": 6086 }, { "epoch": 2.27, "learning_rate": 2e-05, "loss": 0.4626, "step": 6087 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.5869, "step": 6088 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3917, "step": 6089 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.5356, "step": 6090 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3615, "step": 6091 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3597, "step": 6092 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3527, "step": 6093 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.4718, "step": 6094 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3654, "step": 6095 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.4787, "step": 6096 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.5235, "step": 6097 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.54, "step": 6098 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3968, "step": 6099 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.329, "step": 6100 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3769, "step": 6101 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.5563, "step": 6102 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3758, "step": 6103 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.5065, "step": 6104 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3596, "step": 6105 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.5057, "step": 6106 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.437, "step": 6107 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.2809, "step": 6108 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.4212, "step": 6109 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3866, "step": 6110 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.5516, "step": 6111 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.2868, "step": 6112 }, { "epoch": 2.28, "learning_rate": 2e-05, "loss": 0.3443, "step": 6113 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.4172, "step": 6114 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3763, "step": 6115 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.411, "step": 6116 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.4549, "step": 6117 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3392, "step": 6118 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3526, "step": 6119 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.423, "step": 6120 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3501, "step": 6121 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3931, "step": 6122 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3453, "step": 6123 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.5167, "step": 6124 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3575, "step": 6125 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3832, "step": 6126 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.5313, "step": 6127 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3528, "step": 6128 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.5289, "step": 6129 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.4187, "step": 6130 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3878, "step": 6131 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.6347, "step": 6132 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.4092, "step": 6133 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.4362, "step": 6134 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.4736, "step": 6135 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3619, "step": 6136 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3885, "step": 6137 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.3937, "step": 6138 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.5017, "step": 6139 }, { "epoch": 2.29, "learning_rate": 2e-05, "loss": 0.4343, "step": 6140 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4811, "step": 6141 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.6961, "step": 6142 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3634, "step": 6143 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3723, "step": 6144 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4629, "step": 6145 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.2816, "step": 6146 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3034, "step": 6147 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.5842, "step": 6148 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4552, "step": 6149 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.5278, "step": 6150 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3944, "step": 6151 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4128, "step": 6152 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3943, "step": 6153 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.361, "step": 6154 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4541, "step": 6155 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.5646, "step": 6156 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.6584, "step": 6157 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3324, "step": 6158 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.5008, "step": 6159 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.38, "step": 6160 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3919, "step": 6161 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4042, "step": 6162 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3126, "step": 6163 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.3162, "step": 6164 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4017, "step": 6165 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.6287, "step": 6166 }, { "epoch": 2.3, "learning_rate": 2e-05, "loss": 0.4782, "step": 6167 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4077, "step": 6168 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 1.0238, "step": 6169 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.581, "step": 6170 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.5439, "step": 6171 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.536, "step": 6172 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3343, "step": 6173 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4015, "step": 6174 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3589, "step": 6175 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4203, "step": 6176 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.478, "step": 6177 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4493, "step": 6178 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3342, "step": 6179 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.2917, "step": 6180 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4157, "step": 6181 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4824, "step": 6182 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.5329, "step": 6183 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3285, "step": 6184 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4619, "step": 6185 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3238, "step": 6186 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3594, "step": 6187 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3072, "step": 6188 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3245, "step": 6189 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.446, "step": 6190 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.3555, "step": 6191 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.5145, "step": 6192 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.4157, "step": 6193 }, { "epoch": 2.31, "learning_rate": 2e-05, "loss": 0.398, "step": 6194 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3308, "step": 6195 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.2582, "step": 6196 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.5309, "step": 6197 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.2911, "step": 6198 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.8668, "step": 6199 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.5452, "step": 6200 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.5971, "step": 6201 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3424, "step": 6202 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3311, "step": 6203 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3359, "step": 6204 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.4069, "step": 6205 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3269, "step": 6206 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.52, "step": 6207 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.4405, "step": 6208 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3767, "step": 6209 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.4216, "step": 6210 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3746, "step": 6211 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3597, "step": 6212 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3705, "step": 6213 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3824, "step": 6214 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.4491, "step": 6215 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.4841, "step": 6216 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.5114, "step": 6217 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.4953, "step": 6218 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.4227, "step": 6219 }, { "epoch": 2.32, "learning_rate": 2e-05, "loss": 0.3876, "step": 6220 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.4458, "step": 6221 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.4392, "step": 6222 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3631, "step": 6223 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.2793, "step": 6224 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3157, "step": 6225 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.418, "step": 6226 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3512, "step": 6227 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.4023, "step": 6228 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3984, "step": 6229 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.5135, "step": 6230 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.5186, "step": 6231 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.478, "step": 6232 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.5445, "step": 6233 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.4571, "step": 6234 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3132, "step": 6235 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.7902, "step": 6236 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.4362, "step": 6237 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.2952, "step": 6238 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3953, "step": 6239 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3578, "step": 6240 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.4813, "step": 6241 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.425, "step": 6242 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3503, "step": 6243 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.4992, "step": 6244 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.7399, "step": 6245 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.3388, "step": 6246 }, { "epoch": 2.33, "learning_rate": 2e-05, "loss": 0.448, "step": 6247 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.7735, "step": 6248 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.5673, "step": 6249 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4687, "step": 6250 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.5006, "step": 6251 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4361, "step": 6252 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3975, "step": 6253 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4051, "step": 6254 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4709, "step": 6255 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.6237, "step": 6256 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.316, "step": 6257 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3754, "step": 6258 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.5781, "step": 6259 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4424, "step": 6260 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3553, "step": 6261 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3424, "step": 6262 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3305, "step": 6263 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4714, "step": 6264 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3817, "step": 6265 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3845, "step": 6266 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3292, "step": 6267 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4969, "step": 6268 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4046, "step": 6269 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4329, "step": 6270 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.3647, "step": 6271 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4384, "step": 6272 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.5758, "step": 6273 }, { "epoch": 2.34, "learning_rate": 2e-05, "loss": 0.4654, "step": 6274 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.361, "step": 6275 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.5013, "step": 6276 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.482, "step": 6277 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.446, "step": 6278 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3779, "step": 6279 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.4856, "step": 6280 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.5571, "step": 6281 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.4319, "step": 6282 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3201, "step": 6283 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.5294, "step": 6284 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3958, "step": 6285 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.4051, "step": 6286 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.2869, "step": 6287 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3937, "step": 6288 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.7303, "step": 6289 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.5592, "step": 6290 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3386, "step": 6291 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3332, "step": 6292 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.5577, "step": 6293 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.4932, "step": 6294 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3306, "step": 6295 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.4284, "step": 6296 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3604, "step": 6297 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.5949, "step": 6298 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3086, "step": 6299 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.3562, "step": 6300 }, { "epoch": 2.35, "learning_rate": 2e-05, "loss": 0.5815, "step": 6301 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4468, "step": 6302 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4498, "step": 6303 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3874, "step": 6304 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4197, "step": 6305 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.6465, "step": 6306 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4794, "step": 6307 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4289, "step": 6308 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3923, "step": 6309 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3581, "step": 6310 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.6665, "step": 6311 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3626, "step": 6312 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.482, "step": 6313 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4377, "step": 6314 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3537, "step": 6315 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3619, "step": 6316 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.5576, "step": 6317 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.6182, "step": 6318 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4596, "step": 6319 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4012, "step": 6320 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3554, "step": 6321 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4332, "step": 6322 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3499, "step": 6323 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.3295, "step": 6324 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.5245, "step": 6325 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.5801, "step": 6326 }, { "epoch": 2.36, "learning_rate": 2e-05, "loss": 0.4174, "step": 6327 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.6016, "step": 6328 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3687, "step": 6329 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3996, "step": 6330 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4349, "step": 6331 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4048, "step": 6332 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3552, "step": 6333 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.2946, "step": 6334 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4593, "step": 6335 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3711, "step": 6336 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4109, "step": 6337 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4131, "step": 6338 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4, "step": 6339 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4798, "step": 6340 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.2637, "step": 6341 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4521, "step": 6342 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3328, "step": 6343 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4252, "step": 6344 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.5421, "step": 6345 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3433, "step": 6346 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4304, "step": 6347 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4933, "step": 6348 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.4292, "step": 6349 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3637, "step": 6350 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.5482, "step": 6351 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3906, "step": 6352 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.6778, "step": 6353 }, { "epoch": 2.37, "learning_rate": 2e-05, "loss": 0.3736, "step": 6354 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3824, "step": 6355 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4624, "step": 6356 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.365, "step": 6357 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4673, "step": 6358 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3444, "step": 6359 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.387, "step": 6360 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4253, "step": 6361 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.5251, "step": 6362 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3578, "step": 6363 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4709, "step": 6364 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3858, "step": 6365 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4669, "step": 6366 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4567, "step": 6367 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3927, "step": 6368 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.6317, "step": 6369 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3926, "step": 6370 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3455, "step": 6371 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3629, "step": 6372 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.5698, "step": 6373 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3493, "step": 6374 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.5384, "step": 6375 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.3707, "step": 6376 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4488, "step": 6377 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.5948, "step": 6378 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.2926, "step": 6379 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.4321, "step": 6380 }, { "epoch": 2.38, "learning_rate": 2e-05, "loss": 0.5098, "step": 6381 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3604, "step": 6382 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4382, "step": 6383 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.5707, "step": 6384 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4267, "step": 6385 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.394, "step": 6386 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4064, "step": 6387 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.5456, "step": 6388 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.6776, "step": 6389 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3337, "step": 6390 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.5767, "step": 6391 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3087, "step": 6392 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.5875, "step": 6393 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3294, "step": 6394 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3493, "step": 6395 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3627, "step": 6396 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4522, "step": 6397 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3787, "step": 6398 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3761, "step": 6399 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.6613, "step": 6400 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.6643, "step": 6401 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3495, "step": 6402 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4852, "step": 6403 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4793, "step": 6404 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4639, "step": 6405 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.4388, "step": 6406 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3262, "step": 6407 }, { "epoch": 2.39, "learning_rate": 2e-05, "loss": 0.3853, "step": 6408 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.4338, "step": 6409 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3011, "step": 6410 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3265, "step": 6411 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3087, "step": 6412 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.4242, "step": 6413 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3564, "step": 6414 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3584, "step": 6415 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.4316, "step": 6416 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3904, "step": 6417 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.4558, "step": 6418 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.5535, "step": 6419 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.364, "step": 6420 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3865, "step": 6421 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3325, "step": 6422 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.4773, "step": 6423 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3947, "step": 6424 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3423, "step": 6425 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.5087, "step": 6426 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.6917, "step": 6427 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.503, "step": 6428 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.3964, "step": 6429 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.4485, "step": 6430 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.552, "step": 6431 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.5716, "step": 6432 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.5071, "step": 6433 }, { "epoch": 2.4, "learning_rate": 2e-05, "loss": 0.5357, "step": 6434 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.6766, "step": 6435 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.4309, "step": 6436 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.6352, "step": 6437 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.346, "step": 6438 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.5562, "step": 6439 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.5196, "step": 6440 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.2927, "step": 6441 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.5742, "step": 6442 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3212, "step": 6443 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3026, "step": 6444 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3695, "step": 6445 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.4411, "step": 6446 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.4434, "step": 6447 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.4515, "step": 6448 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3171, "step": 6449 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.5661, "step": 6450 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3595, "step": 6451 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.405, "step": 6452 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.4943, "step": 6453 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.5703, "step": 6454 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.5955, "step": 6455 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3001, "step": 6456 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3475, "step": 6457 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3353, "step": 6458 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3819, "step": 6459 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.3288, "step": 6460 }, { "epoch": 2.41, "learning_rate": 2e-05, "loss": 0.4276, "step": 6461 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3062, "step": 6462 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.287, "step": 6463 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4765, "step": 6464 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.449, "step": 6465 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4428, "step": 6466 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3713, "step": 6467 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.5514, "step": 6468 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3753, "step": 6469 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3338, "step": 6470 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4689, "step": 6471 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4175, "step": 6472 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3802, "step": 6473 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3748, "step": 6474 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3455, "step": 6475 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.527, "step": 6476 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4178, "step": 6477 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3703, "step": 6478 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3959, "step": 6479 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4445, "step": 6480 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.7599, "step": 6481 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4407, "step": 6482 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4134, "step": 6483 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3931, "step": 6484 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.454, "step": 6485 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.4551, "step": 6486 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.3416, "step": 6487 }, { "epoch": 2.42, "learning_rate": 2e-05, "loss": 0.5183, "step": 6488 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3769, "step": 6489 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3188, "step": 6490 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3534, "step": 6491 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3519, "step": 6492 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3538, "step": 6493 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4757, "step": 6494 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3148, "step": 6495 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4875, "step": 6496 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3287, "step": 6497 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4601, "step": 6498 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4031, "step": 6499 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4545, "step": 6500 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3348, "step": 6501 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.2989, "step": 6502 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3385, "step": 6503 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4967, "step": 6504 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4366, "step": 6505 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.339, "step": 6506 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.339, "step": 6507 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4612, "step": 6508 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.3648, "step": 6509 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.488, "step": 6510 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.4285, "step": 6511 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.5128, "step": 6512 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.336, "step": 6513 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.448, "step": 6514 }, { "epoch": 2.43, "learning_rate": 2e-05, "loss": 0.5128, "step": 6515 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.3487, "step": 6516 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4423, "step": 6517 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.48, "step": 6518 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4688, "step": 6519 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.5271, "step": 6520 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4116, "step": 6521 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.3114, "step": 6522 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4357, "step": 6523 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.3649, "step": 6524 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.56, "step": 6525 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4679, "step": 6526 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.441, "step": 6527 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4131, "step": 6528 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.3323, "step": 6529 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4081, "step": 6530 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4054, "step": 6531 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4549, "step": 6532 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.5269, "step": 6533 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.5502, "step": 6534 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4736, "step": 6535 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.3634, "step": 6536 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.5208, "step": 6537 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.4516, "step": 6538 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.3685, "step": 6539 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.3285, "step": 6540 }, { "epoch": 2.44, "learning_rate": 2e-05, "loss": 0.6155, "step": 6541 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4324, "step": 6542 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3042, "step": 6543 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.367, "step": 6544 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3086, "step": 6545 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3276, "step": 6546 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4123, "step": 6547 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4312, "step": 6548 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3939, "step": 6549 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3907, "step": 6550 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4007, "step": 6551 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.535, "step": 6552 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3143, "step": 6553 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3261, "step": 6554 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.7697, "step": 6555 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3851, "step": 6556 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4274, "step": 6557 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4497, "step": 6558 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.5096, "step": 6559 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3262, "step": 6560 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3728, "step": 6561 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3908, "step": 6562 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.5317, "step": 6563 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3497, "step": 6564 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.2892, "step": 6565 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.3494, "step": 6566 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4021, "step": 6567 }, { "epoch": 2.45, "learning_rate": 2e-05, "loss": 0.4609, "step": 6568 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4894, "step": 6569 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.3611, "step": 6570 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4574, "step": 6571 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.3951, "step": 6572 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4695, "step": 6573 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.415, "step": 6574 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4169, "step": 6575 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4116, "step": 6576 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.375, "step": 6577 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4434, "step": 6578 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.3713, "step": 6579 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.3494, "step": 6580 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4521, "step": 6581 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4308, "step": 6582 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.3726, "step": 6583 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.6207, "step": 6584 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.3747, "step": 6585 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.513, "step": 6586 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.5101, "step": 6587 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.3664, "step": 6588 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4233, "step": 6589 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4176, "step": 6590 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.5673, "step": 6591 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.2876, "step": 6592 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.4538, "step": 6593 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.5557, "step": 6594 }, { "epoch": 2.46, "learning_rate": 2e-05, "loss": 0.5117, "step": 6595 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.4261, "step": 6596 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.461, "step": 6597 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.347, "step": 6598 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.4287, "step": 6599 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.5157, "step": 6600 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3467, "step": 6601 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3472, "step": 6602 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.322, "step": 6603 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.506, "step": 6604 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.4433, "step": 6605 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3371, "step": 6606 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3527, "step": 6607 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.5957, "step": 6608 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3427, "step": 6609 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.4809, "step": 6610 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3735, "step": 6611 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.6221, "step": 6612 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3608, "step": 6613 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.2979, "step": 6614 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.256, "step": 6615 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3851, "step": 6616 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3704, "step": 6617 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3538, "step": 6618 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3497, "step": 6619 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.3749, "step": 6620 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.4012, "step": 6621 }, { "epoch": 2.47, "learning_rate": 2e-05, "loss": 0.639, "step": 6622 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.5124, "step": 6623 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.5861, "step": 6624 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.3853, "step": 6625 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.449, "step": 6626 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.3224, "step": 6627 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.3376, "step": 6628 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.5652, "step": 6629 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.323, "step": 6630 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.6524, "step": 6631 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.3112, "step": 6632 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.4278, "step": 6633 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.3482, "step": 6634 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.6082, "step": 6635 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.6829, "step": 6636 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.322, "step": 6637 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.4538, "step": 6638 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.3693, "step": 6639 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.5125, "step": 6640 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.4339, "step": 6641 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.5092, "step": 6642 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.4185, "step": 6643 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.334, "step": 6644 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.6564, "step": 6645 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.3524, "step": 6646 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.4831, "step": 6647 }, { "epoch": 2.48, "learning_rate": 2e-05, "loss": 0.4349, "step": 6648 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4938, "step": 6649 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3429, "step": 6650 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3139, "step": 6651 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.6625, "step": 6652 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3917, "step": 6653 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3958, "step": 6654 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.6268, "step": 6655 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3782, "step": 6656 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3593, "step": 6657 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3184, "step": 6658 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.5734, "step": 6659 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3544, "step": 6660 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3626, "step": 6661 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4623, "step": 6662 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4857, "step": 6663 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.5092, "step": 6664 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3405, "step": 6665 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4654, "step": 6666 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4707, "step": 6667 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.5633, "step": 6668 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4385, "step": 6669 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4822, "step": 6670 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4658, "step": 6671 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.3646, "step": 6672 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4617, "step": 6673 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.4693, "step": 6674 }, { "epoch": 2.49, "learning_rate": 2e-05, "loss": 0.5497, "step": 6675 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.5533, "step": 6676 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.3372, "step": 6677 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.458, "step": 6678 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4118, "step": 6679 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.3697, "step": 6680 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.3284, "step": 6681 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4853, "step": 6682 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.6436, "step": 6683 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4838, "step": 6684 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.5271, "step": 6685 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.3391, "step": 6686 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.356, "step": 6687 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4624, "step": 6688 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.329, "step": 6689 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.2533, "step": 6690 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4959, "step": 6691 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.6199, "step": 6692 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4072, "step": 6693 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4668, "step": 6694 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.5519, "step": 6695 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.435, "step": 6696 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4666, "step": 6697 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.3931, "step": 6698 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4694, "step": 6699 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.3144, "step": 6700 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.4407, "step": 6701 }, { "epoch": 2.5, "learning_rate": 2e-05, "loss": 0.3805, "step": 6702 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.6863, "step": 6703 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3707, "step": 6704 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.4905, "step": 6705 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.7644, "step": 6706 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3817, "step": 6707 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3128, "step": 6708 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3354, "step": 6709 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3127, "step": 6710 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.4085, "step": 6711 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3976, "step": 6712 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3347, "step": 6713 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3217, "step": 6714 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.338, "step": 6715 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.4937, "step": 6716 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.4357, "step": 6717 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.351, "step": 6718 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.335, "step": 6719 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.4952, "step": 6720 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3196, "step": 6721 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.5382, "step": 6722 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3768, "step": 6723 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.4029, "step": 6724 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3613, "step": 6725 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.5151, "step": 6726 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3403, "step": 6727 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3459, "step": 6728 }, { "epoch": 2.51, "learning_rate": 2e-05, "loss": 0.3909, "step": 6729 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.5146, "step": 6730 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4587, "step": 6731 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.5537, "step": 6732 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.529, "step": 6733 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.508, "step": 6734 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.7344, "step": 6735 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.3573, "step": 6736 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4246, "step": 6737 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.357, "step": 6738 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4562, "step": 6739 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.3045, "step": 6740 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4438, "step": 6741 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.3687, "step": 6742 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4723, "step": 6743 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.5023, "step": 6744 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4731, "step": 6745 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4327, "step": 6746 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.5415, "step": 6747 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4455, "step": 6748 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.3192, "step": 6749 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.435, "step": 6750 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4184, "step": 6751 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4087, "step": 6752 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.3434, "step": 6753 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.3941, "step": 6754 }, { "epoch": 2.52, "learning_rate": 2e-05, "loss": 0.4938, "step": 6755 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.3926, "step": 6756 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4217, "step": 6757 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4082, "step": 6758 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.5599, "step": 6759 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4275, "step": 6760 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.2847, "step": 6761 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4978, "step": 6762 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.3909, "step": 6763 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.3475, "step": 6764 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.7522, "step": 6765 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.6185, "step": 6766 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.5181, "step": 6767 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.5255, "step": 6768 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.7064, "step": 6769 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4781, "step": 6770 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4307, "step": 6771 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.2718, "step": 6772 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4569, "step": 6773 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.6426, "step": 6774 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4733, "step": 6775 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.6193, "step": 6776 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4364, "step": 6777 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.4135, "step": 6778 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.3604, "step": 6779 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.5327, "step": 6780 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.3561, "step": 6781 }, { "epoch": 2.53, "learning_rate": 2e-05, "loss": 0.3185, "step": 6782 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3649, "step": 6783 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5914, "step": 6784 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3643, "step": 6785 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.625, "step": 6786 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3703, "step": 6787 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.443, "step": 6788 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3461, "step": 6789 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3884, "step": 6790 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.7359, "step": 6791 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5401, "step": 6792 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.743, "step": 6793 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5242, "step": 6794 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3464, "step": 6795 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5893, "step": 6796 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.2749, "step": 6797 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5277, "step": 6798 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3255, "step": 6799 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3037, "step": 6800 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5554, "step": 6801 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5051, "step": 6802 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.404, "step": 6803 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.333, "step": 6804 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.4277, "step": 6805 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.4383, "step": 6806 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.4777, "step": 6807 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.3926, "step": 6808 }, { "epoch": 2.54, "learning_rate": 2e-05, "loss": 0.5204, "step": 6809 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4173, "step": 6810 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.5743, "step": 6811 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4598, "step": 6812 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.3925, "step": 6813 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4106, "step": 6814 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4468, "step": 6815 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.5173, "step": 6816 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.3788, "step": 6817 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4436, "step": 6818 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4781, "step": 6819 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.458, "step": 6820 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4059, "step": 6821 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.3388, "step": 6822 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4135, "step": 6823 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.8076, "step": 6824 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.5882, "step": 6825 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.5048, "step": 6826 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4742, "step": 6827 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4359, "step": 6828 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.329, "step": 6829 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.6179, "step": 6830 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.3147, "step": 6831 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.3604, "step": 6832 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4144, "step": 6833 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.4463, "step": 6834 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.3083, "step": 6835 }, { "epoch": 2.55, "learning_rate": 2e-05, "loss": 0.3664, "step": 6836 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4282, "step": 6837 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.5715, "step": 6838 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3949, "step": 6839 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4984, "step": 6840 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.2771, "step": 6841 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.358, "step": 6842 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3663, "step": 6843 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.6707, "step": 6844 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4797, "step": 6845 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.2774, "step": 6846 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.2908, "step": 6847 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3949, "step": 6848 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3542, "step": 6849 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3348, "step": 6850 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4186, "step": 6851 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3719, "step": 6852 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3301, "step": 6853 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.392, "step": 6854 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3752, "step": 6855 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.3686, "step": 6856 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4748, "step": 6857 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4717, "step": 6858 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4327, "step": 6859 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4667, "step": 6860 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.2421, "step": 6861 }, { "epoch": 2.56, "learning_rate": 2e-05, "loss": 0.4406, "step": 6862 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.636, "step": 6863 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.5602, "step": 6864 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.5258, "step": 6865 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3311, "step": 6866 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3628, "step": 6867 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.5213, "step": 6868 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4533, "step": 6869 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3884, "step": 6870 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4005, "step": 6871 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4213, "step": 6872 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3225, "step": 6873 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.301, "step": 6874 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4404, "step": 6875 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4204, "step": 6876 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3225, "step": 6877 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.5034, "step": 6878 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4447, "step": 6879 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3708, "step": 6880 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4793, "step": 6881 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.5539, "step": 6882 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.367, "step": 6883 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3592, "step": 6884 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4449, "step": 6885 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3678, "step": 6886 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.4702, "step": 6887 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.5065, "step": 6888 }, { "epoch": 2.57, "learning_rate": 2e-05, "loss": 0.3619, "step": 6889 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4737, "step": 6890 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4133, "step": 6891 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3725, "step": 6892 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.2984, "step": 6893 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4898, "step": 6894 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3421, "step": 6895 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3435, "step": 6896 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3227, "step": 6897 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4756, "step": 6898 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4809, "step": 6899 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3321, "step": 6900 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3131, "step": 6901 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3885, "step": 6902 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3776, "step": 6903 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4302, "step": 6904 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4094, "step": 6905 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.38, "step": 6906 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3642, "step": 6907 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3245, "step": 6908 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.58, "step": 6909 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4368, "step": 6910 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4237, "step": 6911 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4402, "step": 6912 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.2356, "step": 6913 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3342, "step": 6914 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.4021, "step": 6915 }, { "epoch": 2.58, "learning_rate": 2e-05, "loss": 0.3501, "step": 6916 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.3292, "step": 6917 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.922, "step": 6918 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.5929, "step": 6919 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.6193, "step": 6920 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4368, "step": 6921 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4843, "step": 6922 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.3281, "step": 6923 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4611, "step": 6924 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.3999, "step": 6925 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.3207, "step": 6926 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.7388, "step": 6927 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.5564, "step": 6928 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.5643, "step": 6929 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.5501, "step": 6930 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4282, "step": 6931 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.5914, "step": 6932 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4303, "step": 6933 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.31, "step": 6934 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.6651, "step": 6935 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4293, "step": 6936 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.3285, "step": 6937 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4663, "step": 6938 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.7031, "step": 6939 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.4548, "step": 6940 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.3209, "step": 6941 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.3487, "step": 6942 }, { "epoch": 2.59, "learning_rate": 2e-05, "loss": 0.2432, "step": 6943 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3228, "step": 6944 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4021, "step": 6945 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3499, "step": 6946 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3921, "step": 6947 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3424, "step": 6948 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4125, "step": 6949 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4304, "step": 6950 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3329, "step": 6951 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3943, "step": 6952 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.7524, "step": 6953 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4481, "step": 6954 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3379, "step": 6955 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3471, "step": 6956 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.5434, "step": 6957 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.2988, "step": 6958 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4069, "step": 6959 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 1.1781, "step": 6960 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4181, "step": 6961 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.2972, "step": 6962 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.2707, "step": 6963 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4964, "step": 6964 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3914, "step": 6965 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.4339, "step": 6966 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.5881, "step": 6967 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.6032, "step": 6968 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.5614, "step": 6969 }, { "epoch": 2.6, "learning_rate": 2e-05, "loss": 0.3781, "step": 6970 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3626, "step": 6971 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3501, "step": 6972 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.463, "step": 6973 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3774, "step": 6974 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3615, "step": 6975 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4565, "step": 6976 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4052, "step": 6977 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4357, "step": 6978 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4118, "step": 6979 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4074, "step": 6980 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4381, "step": 6981 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3308, "step": 6982 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.672, "step": 6983 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3114, "step": 6984 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3982, "step": 6985 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.3404, "step": 6986 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.6679, "step": 6987 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4813, "step": 6988 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.5102, "step": 6989 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.5354, "step": 6990 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.6319, "step": 6991 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.5134, "step": 6992 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.4257, "step": 6993 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.6187, "step": 6994 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.5315, "step": 6995 }, { "epoch": 2.61, "learning_rate": 2e-05, "loss": 0.6461, "step": 6996 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.3602, "step": 6997 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.6361, "step": 6998 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.3975, "step": 6999 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4581, "step": 7000 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.5315, "step": 7001 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4623, "step": 7002 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.448, "step": 7003 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.5087, "step": 7004 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4824, "step": 7005 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4573, "step": 7006 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4637, "step": 7007 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.5959, "step": 7008 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.2797, "step": 7009 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4581, "step": 7010 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4117, "step": 7011 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.6058, "step": 7012 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.3927, "step": 7013 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.6203, "step": 7014 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4818, "step": 7015 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.5377, "step": 7016 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.3743, "step": 7017 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4622, "step": 7018 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.3535, "step": 7019 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.6148, "step": 7020 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.3297, "step": 7021 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.4927, "step": 7022 }, { "epoch": 2.62, "learning_rate": 2e-05, "loss": 0.3578, "step": 7023 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4444, "step": 7024 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4139, "step": 7025 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.5073, "step": 7026 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4007, "step": 7027 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4869, "step": 7028 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.374, "step": 7029 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.5347, "step": 7030 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4442, "step": 7031 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.5817, "step": 7032 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.3811, "step": 7033 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4962, "step": 7034 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4299, "step": 7035 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4563, "step": 7036 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4017, "step": 7037 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.535, "step": 7038 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.5518, "step": 7039 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.3635, "step": 7040 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4602, "step": 7041 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4167, "step": 7042 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4107, "step": 7043 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.4231, "step": 7044 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.3813, "step": 7045 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.3915, "step": 7046 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.3671, "step": 7047 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.3808, "step": 7048 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.5418, "step": 7049 }, { "epoch": 2.63, "learning_rate": 2e-05, "loss": 0.5388, "step": 7050 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3933, "step": 7051 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3707, "step": 7052 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.464, "step": 7053 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3805, "step": 7054 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3147, "step": 7055 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.2327, "step": 7056 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.4355, "step": 7057 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3645, "step": 7058 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.4873, "step": 7059 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.489, "step": 7060 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3955, "step": 7061 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.9338, "step": 7062 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.4532, "step": 7063 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3348, "step": 7064 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.4616, "step": 7065 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3472, "step": 7066 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3906, "step": 7067 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.4337, "step": 7068 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.4955, "step": 7069 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3467, "step": 7070 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3536, "step": 7071 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.585, "step": 7072 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3742, "step": 7073 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.3941, "step": 7074 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.4654, "step": 7075 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.5543, "step": 7076 }, { "epoch": 2.64, "learning_rate": 2e-05, "loss": 0.5305, "step": 7077 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.4637, "step": 7078 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5033, "step": 7079 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.436, "step": 7080 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.3626, "step": 7081 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5704, "step": 7082 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.4254, "step": 7083 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.3402, "step": 7084 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5178, "step": 7085 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.425, "step": 7086 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.4036, "step": 7087 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5027, "step": 7088 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5032, "step": 7089 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5624, "step": 7090 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.424, "step": 7091 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.3809, "step": 7092 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.3968, "step": 7093 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.3551, "step": 7094 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.4187, "step": 7095 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5083, "step": 7096 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.3343, "step": 7097 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.4, "step": 7098 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.4945, "step": 7099 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.6099, "step": 7100 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.3456, "step": 7101 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5099, "step": 7102 }, { "epoch": 2.65, "learning_rate": 2e-05, "loss": 0.5117, "step": 7103 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.5547, "step": 7104 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3751, "step": 7105 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3994, "step": 7106 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3781, "step": 7107 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3463, "step": 7108 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4508, "step": 7109 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4851, "step": 7110 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4198, "step": 7111 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.6619, "step": 7112 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3511, "step": 7113 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.5032, "step": 7114 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4608, "step": 7115 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4125, "step": 7116 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3143, "step": 7117 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4936, "step": 7118 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.457, "step": 7119 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.312, "step": 7120 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.6144, "step": 7121 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4153, "step": 7122 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.5457, "step": 7123 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.6179, "step": 7124 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3476, "step": 7125 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4575, "step": 7126 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.4361, "step": 7127 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3674, "step": 7128 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.3544, "step": 7129 }, { "epoch": 2.66, "learning_rate": 2e-05, "loss": 0.5831, "step": 7130 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4463, "step": 7131 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4983, "step": 7132 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4263, "step": 7133 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3451, "step": 7134 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3678, "step": 7135 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4416, "step": 7136 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4809, "step": 7137 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3679, "step": 7138 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.6346, "step": 7139 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4589, "step": 7140 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3663, "step": 7141 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4386, "step": 7142 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3721, "step": 7143 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.7147, "step": 7144 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3431, "step": 7145 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.2563, "step": 7146 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3975, "step": 7147 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.574, "step": 7148 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4155, "step": 7149 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4751, "step": 7150 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4109, "step": 7151 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4012, "step": 7152 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4392, "step": 7153 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.354, "step": 7154 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.6888, "step": 7155 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.4827, "step": 7156 }, { "epoch": 2.67, "learning_rate": 2e-05, "loss": 0.3691, "step": 7157 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3487, "step": 7158 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3412, "step": 7159 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.537, "step": 7160 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.7671, "step": 7161 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3759, "step": 7162 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3881, "step": 7163 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4844, "step": 7164 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.5493, "step": 7165 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4873, "step": 7166 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.432, "step": 7167 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4426, "step": 7168 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.5142, "step": 7169 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4487, "step": 7170 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3737, "step": 7171 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.39, "step": 7172 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4504, "step": 7173 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3772, "step": 7174 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.5348, "step": 7175 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4958, "step": 7176 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4467, "step": 7177 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3943, "step": 7178 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.5478, "step": 7179 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.3992, "step": 7180 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4604, "step": 7181 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.382, "step": 7182 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.5728, "step": 7183 }, { "epoch": 2.68, "learning_rate": 2e-05, "loss": 0.4944, "step": 7184 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.373, "step": 7185 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4708, "step": 7186 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3923, "step": 7187 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3648, "step": 7188 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.5204, "step": 7189 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4043, "step": 7190 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3469, "step": 7191 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3819, "step": 7192 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4326, "step": 7193 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.5104, "step": 7194 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3841, "step": 7195 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3916, "step": 7196 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.286, "step": 7197 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3429, "step": 7198 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.5172, "step": 7199 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4348, "step": 7200 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.415, "step": 7201 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.6796, "step": 7202 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4377, "step": 7203 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4122, "step": 7204 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.5352, "step": 7205 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.3044, "step": 7206 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.5409, "step": 7207 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4779, "step": 7208 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.4055, "step": 7209 }, { "epoch": 2.69, "learning_rate": 2e-05, "loss": 0.5552, "step": 7210 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3312, "step": 7211 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4416, "step": 7212 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.6206, "step": 7213 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.5211, "step": 7214 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3393, "step": 7215 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.364, "step": 7216 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.5411, "step": 7217 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4117, "step": 7218 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3823, "step": 7219 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.2891, "step": 7220 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4218, "step": 7221 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3263, "step": 7222 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3859, "step": 7223 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3369, "step": 7224 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4837, "step": 7225 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4313, "step": 7226 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3335, "step": 7227 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3257, "step": 7228 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.5238, "step": 7229 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4416, "step": 7230 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3022, "step": 7231 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4081, "step": 7232 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3969, "step": 7233 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3364, "step": 7234 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.7072, "step": 7235 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.3551, "step": 7236 }, { "epoch": 2.7, "learning_rate": 2e-05, "loss": 0.4021, "step": 7237 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4022, "step": 7238 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.8505, "step": 7239 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.33, "step": 7240 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.5867, "step": 7241 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4649, "step": 7242 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.5941, "step": 7243 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.3419, "step": 7244 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4366, "step": 7245 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.5849, "step": 7246 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4688, "step": 7247 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.5404, "step": 7248 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.3591, "step": 7249 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.453, "step": 7250 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4728, "step": 7251 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4041, "step": 7252 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4652, "step": 7253 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.6815, "step": 7254 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.3434, "step": 7255 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4795, "step": 7256 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4417, "step": 7257 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.318, "step": 7258 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4598, "step": 7259 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.3713, "step": 7260 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.4347, "step": 7261 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.472, "step": 7262 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.2682, "step": 7263 }, { "epoch": 2.71, "learning_rate": 2e-05, "loss": 0.5983, "step": 7264 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4378, "step": 7265 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3985, "step": 7266 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3744, "step": 7267 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.559, "step": 7268 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.356, "step": 7269 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4234, "step": 7270 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4015, "step": 7271 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.8271, "step": 7272 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3514, "step": 7273 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4306, "step": 7274 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.42, "step": 7275 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3418, "step": 7276 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.5525, "step": 7277 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3905, "step": 7278 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4997, "step": 7279 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 1.4373, "step": 7280 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4195, "step": 7281 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.5337, "step": 7282 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.5169, "step": 7283 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.6858, "step": 7284 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4491, "step": 7285 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3625, "step": 7286 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4598, "step": 7287 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4561, "step": 7288 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3912, "step": 7289 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.3182, "step": 7290 }, { "epoch": 2.72, "learning_rate": 2e-05, "loss": 0.4524, "step": 7291 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.4824, "step": 7292 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3562, "step": 7293 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.8966, "step": 7294 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.5675, "step": 7295 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.379, "step": 7296 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.4453, "step": 7297 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3705, "step": 7298 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3907, "step": 7299 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.451, "step": 7300 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.4776, "step": 7301 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.4166, "step": 7302 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3375, "step": 7303 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3489, "step": 7304 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.8333, "step": 7305 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3593, "step": 7306 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.4787, "step": 7307 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.6197, "step": 7308 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.5549, "step": 7309 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.492, "step": 7310 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.4188, "step": 7311 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.5783, "step": 7312 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.576, "step": 7313 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3705, "step": 7314 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3564, "step": 7315 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.371, "step": 7316 }, { "epoch": 2.73, "learning_rate": 2e-05, "loss": 0.3609, "step": 7317 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3983, "step": 7318 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3596, "step": 7319 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3359, "step": 7320 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3196, "step": 7321 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.508, "step": 7322 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.4023, "step": 7323 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3695, "step": 7324 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.5741, "step": 7325 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.4386, "step": 7326 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3543, "step": 7327 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.5171, "step": 7328 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.2926, "step": 7329 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3586, "step": 7330 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3883, "step": 7331 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.4093, "step": 7332 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.5157, "step": 7333 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.5807, "step": 7334 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3687, "step": 7335 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.4667, "step": 7336 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3787, "step": 7337 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.4299, "step": 7338 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.4831, "step": 7339 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3943, "step": 7340 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3705, "step": 7341 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.4028, "step": 7342 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.438, "step": 7343 }, { "epoch": 2.74, "learning_rate": 2e-05, "loss": 0.3606, "step": 7344 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3596, "step": 7345 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4028, "step": 7346 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3943, "step": 7347 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.2591, "step": 7348 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3741, "step": 7349 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3365, "step": 7350 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4157, "step": 7351 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.2761, "step": 7352 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4052, "step": 7353 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.5233, "step": 7354 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.5975, "step": 7355 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4544, "step": 7356 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4282, "step": 7357 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4489, "step": 7358 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.5427, "step": 7359 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.5642, "step": 7360 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.352, "step": 7361 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4433, "step": 7362 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3432, "step": 7363 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4218, "step": 7364 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4247, "step": 7365 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3884, "step": 7366 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3394, "step": 7367 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.3722, "step": 7368 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4675, "step": 7369 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.494, "step": 7370 }, { "epoch": 2.75, "learning_rate": 2e-05, "loss": 0.4018, "step": 7371 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.6943, "step": 7372 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4843, "step": 7373 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.3466, "step": 7374 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.3871, "step": 7375 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.5183, "step": 7376 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4518, "step": 7377 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4719, "step": 7378 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.408, "step": 7379 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.5369, "step": 7380 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.6052, "step": 7381 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.305, "step": 7382 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4735, "step": 7383 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4826, "step": 7384 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.5527, "step": 7385 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4266, "step": 7386 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.329, "step": 7387 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.3744, "step": 7388 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4694, "step": 7389 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.3128, "step": 7390 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4808, "step": 7391 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4214, "step": 7392 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.3735, "step": 7393 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4867, "step": 7394 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.5434, "step": 7395 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4408, "step": 7396 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4521, "step": 7397 }, { "epoch": 2.76, "learning_rate": 2e-05, "loss": 0.4318, "step": 7398 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.5269, "step": 7399 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4063, "step": 7400 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.372, "step": 7401 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4948, "step": 7402 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3785, "step": 7403 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4771, "step": 7404 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3539, "step": 7405 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3771, "step": 7406 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4881, "step": 7407 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3584, "step": 7408 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3662, "step": 7409 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.302, "step": 7410 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4325, "step": 7411 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4623, "step": 7412 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4682, "step": 7413 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4742, "step": 7414 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4985, "step": 7415 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4269, "step": 7416 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3706, "step": 7417 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3816, "step": 7418 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4991, "step": 7419 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4542, "step": 7420 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3422, "step": 7421 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.5722, "step": 7422 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.3292, "step": 7423 }, { "epoch": 2.77, "learning_rate": 2e-05, "loss": 0.4061, "step": 7424 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3794, "step": 7425 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.5498, "step": 7426 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3185, "step": 7427 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3472, "step": 7428 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4941, "step": 7429 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3684, "step": 7430 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3347, "step": 7431 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4032, "step": 7432 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4826, "step": 7433 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4334, "step": 7434 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3589, "step": 7435 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.5258, "step": 7436 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3851, "step": 7437 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4688, "step": 7438 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3093, "step": 7439 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.5791, "step": 7440 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4187, "step": 7441 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3302, "step": 7442 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3676, "step": 7443 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.5736, "step": 7444 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.5343, "step": 7445 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.345, "step": 7446 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4173, "step": 7447 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.3456, "step": 7448 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4112, "step": 7449 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.2897, "step": 7450 }, { "epoch": 2.78, "learning_rate": 2e-05, "loss": 0.4178, "step": 7451 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4511, "step": 7452 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.3834, "step": 7453 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.6953, "step": 7454 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.5024, "step": 7455 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.5772, "step": 7456 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.7135, "step": 7457 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4417, "step": 7458 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4218, "step": 7459 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.6223, "step": 7460 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4562, "step": 7461 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.3898, "step": 7462 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.5998, "step": 7463 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4938, "step": 7464 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.5356, "step": 7465 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4051, "step": 7466 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.3503, "step": 7467 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4669, "step": 7468 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.2573, "step": 7469 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.6876, "step": 7470 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4125, "step": 7471 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4449, "step": 7472 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.3474, "step": 7473 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4087, "step": 7474 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.383, "step": 7475 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.4064, "step": 7476 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.6658, "step": 7477 }, { "epoch": 2.79, "learning_rate": 2e-05, "loss": 0.3783, "step": 7478 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3988, "step": 7479 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.5845, "step": 7480 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3407, "step": 7481 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.4994, "step": 7482 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3818, "step": 7483 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3673, "step": 7484 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.7203, "step": 7485 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.413, "step": 7486 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.356, "step": 7487 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.4876, "step": 7488 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3961, "step": 7489 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.4298, "step": 7490 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.5403, "step": 7491 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.5872, "step": 7492 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.4022, "step": 7493 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3849, "step": 7494 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3986, "step": 7495 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.4696, "step": 7496 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3828, "step": 7497 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.5116, "step": 7498 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.4389, "step": 7499 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.5553, "step": 7500 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3951, "step": 7501 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.5417, "step": 7502 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3699, "step": 7503 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.3881, "step": 7504 }, { "epoch": 2.8, "learning_rate": 2e-05, "loss": 0.474, "step": 7505 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.3625, "step": 7506 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.3058, "step": 7507 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.4726, "step": 7508 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.3862, "step": 7509 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.2992, "step": 7510 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.3916, "step": 7511 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.6203, "step": 7512 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.6018, "step": 7513 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.4778, "step": 7514 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.3567, "step": 7515 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.4286, "step": 7516 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.4524, "step": 7517 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.409, "step": 7518 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.5345, "step": 7519 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.422, "step": 7520 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.7123, "step": 7521 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.3993, "step": 7522 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.5926, "step": 7523 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.4973, "step": 7524 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.4249, "step": 7525 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.4202, "step": 7526 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.3708, "step": 7527 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.2858, "step": 7528 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.5781, "step": 7529 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.546, "step": 7530 }, { "epoch": 2.81, "learning_rate": 2e-05, "loss": 0.444, "step": 7531 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4709, "step": 7532 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3655, "step": 7533 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4929, "step": 7534 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3372, "step": 7535 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3758, "step": 7536 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.5514, "step": 7537 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.6695, "step": 7538 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4659, "step": 7539 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3603, "step": 7540 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4317, "step": 7541 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4981, "step": 7542 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3632, "step": 7543 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4508, "step": 7544 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4822, "step": 7545 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3673, "step": 7546 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4219, "step": 7547 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4436, "step": 7548 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3291, "step": 7549 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3916, "step": 7550 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4631, "step": 7551 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3859, "step": 7552 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4426, "step": 7553 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.5063, "step": 7554 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4009, "step": 7555 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4708, "step": 7556 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.4353, "step": 7557 }, { "epoch": 2.82, "learning_rate": 2e-05, "loss": 0.3181, "step": 7558 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.5371, "step": 7559 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.3153, "step": 7560 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4372, "step": 7561 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.3825, "step": 7562 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.5252, "step": 7563 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4647, "step": 7564 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4297, "step": 7565 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4998, "step": 7566 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.6972, "step": 7567 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.2972, "step": 7568 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.3682, "step": 7569 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.3757, "step": 7570 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4384, "step": 7571 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.3831, "step": 7572 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4789, "step": 7573 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4143, "step": 7574 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.3394, "step": 7575 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.426, "step": 7576 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.7983, "step": 7577 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4302, "step": 7578 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.2583, "step": 7579 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4379, "step": 7580 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.5649, "step": 7581 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.4321, "step": 7582 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.6755, "step": 7583 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.3547, "step": 7584 }, { "epoch": 2.83, "learning_rate": 2e-05, "loss": 0.5232, "step": 7585 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.3851, "step": 7586 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.3117, "step": 7587 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.3746, "step": 7588 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4076, "step": 7589 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.3571, "step": 7590 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4818, "step": 7591 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.3528, "step": 7592 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4929, "step": 7593 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.5155, "step": 7594 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4017, "step": 7595 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.351, "step": 7596 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4315, "step": 7597 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.401, "step": 7598 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4058, "step": 7599 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.6101, "step": 7600 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.495, "step": 7601 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.5033, "step": 7602 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4654, "step": 7603 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.3816, "step": 7604 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4373, "step": 7605 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4889, "step": 7606 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4713, "step": 7607 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.24, "step": 7608 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4286, "step": 7609 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.5204, "step": 7610 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.3846, "step": 7611 }, { "epoch": 2.84, "learning_rate": 2e-05, "loss": 0.4634, "step": 7612 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3419, "step": 7613 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.7331, "step": 7614 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.6196, "step": 7615 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3765, "step": 7616 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3442, "step": 7617 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3842, "step": 7618 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3346, "step": 7619 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4902, "step": 7620 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4695, "step": 7621 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.5435, "step": 7622 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4319, "step": 7623 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3803, "step": 7624 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4227, "step": 7625 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.376, "step": 7626 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3536, "step": 7627 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4236, "step": 7628 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.5366, "step": 7629 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.409, "step": 7630 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.3795, "step": 7631 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4569, "step": 7632 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4209, "step": 7633 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.5849, "step": 7634 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.368, "step": 7635 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.6817, "step": 7636 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4954, "step": 7637 }, { "epoch": 2.85, "learning_rate": 2e-05, "loss": 0.4833, "step": 7638 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.4225, "step": 7639 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.584, "step": 7640 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3317, "step": 7641 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.5291, "step": 7642 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.579, "step": 7643 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3705, "step": 7644 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3654, "step": 7645 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.6319, "step": 7646 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.4941, "step": 7647 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.4301, "step": 7648 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.5139, "step": 7649 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3791, "step": 7650 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3841, "step": 7651 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.4995, "step": 7652 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3169, "step": 7653 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.4167, "step": 7654 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3899, "step": 7655 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.4911, "step": 7656 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.6106, "step": 7657 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.5121, "step": 7658 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.4479, "step": 7659 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.5532, "step": 7660 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3974, "step": 7661 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3685, "step": 7662 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.5056, "step": 7663 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.3652, "step": 7664 }, { "epoch": 2.86, "learning_rate": 2e-05, "loss": 0.5142, "step": 7665 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4644, "step": 7666 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.375, "step": 7667 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3575, "step": 7668 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4081, "step": 7669 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3807, "step": 7670 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4438, "step": 7671 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3211, "step": 7672 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.385, "step": 7673 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3726, "step": 7674 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4862, "step": 7675 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.353, "step": 7676 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4049, "step": 7677 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3787, "step": 7678 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4612, "step": 7679 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3894, "step": 7680 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4404, "step": 7681 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3749, "step": 7682 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4565, "step": 7683 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4103, "step": 7684 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3781, "step": 7685 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.5279, "step": 7686 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.3292, "step": 7687 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.5836, "step": 7688 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.7941, "step": 7689 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4068, "step": 7690 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.433, "step": 7691 }, { "epoch": 2.87, "learning_rate": 2e-05, "loss": 0.4331, "step": 7692 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.519, "step": 7693 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.4828, "step": 7694 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3464, "step": 7695 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.4843, "step": 7696 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.4234, "step": 7697 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3923, "step": 7698 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.4092, "step": 7699 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.6342, "step": 7700 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3944, "step": 7701 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3281, "step": 7702 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3324, "step": 7703 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.2493, "step": 7704 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.5482, "step": 7705 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3497, "step": 7706 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.7509, "step": 7707 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.5212, "step": 7708 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.6034, "step": 7709 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.4551, "step": 7710 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3625, "step": 7711 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.4162, "step": 7712 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3342, "step": 7713 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3626, "step": 7714 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.4865, "step": 7715 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.825, "step": 7716 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3809, "step": 7717 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3869, "step": 7718 }, { "epoch": 2.88, "learning_rate": 2e-05, "loss": 0.3772, "step": 7719 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.6895, "step": 7720 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3856, "step": 7721 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3508, "step": 7722 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.4062, "step": 7723 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3952, "step": 7724 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.4104, "step": 7725 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3465, "step": 7726 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.632, "step": 7727 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.4414, "step": 7728 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.6522, "step": 7729 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.5413, "step": 7730 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.5474, "step": 7731 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3366, "step": 7732 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.7359, "step": 7733 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3146, "step": 7734 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3998, "step": 7735 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.5247, "step": 7736 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3987, "step": 7737 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.5489, "step": 7738 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.5769, "step": 7739 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.5391, "step": 7740 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3894, "step": 7741 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.4385, "step": 7742 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.3747, "step": 7743 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.4065, "step": 7744 }, { "epoch": 2.89, "learning_rate": 2e-05, "loss": 0.4691, "step": 7745 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.4466, "step": 7746 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.526, "step": 7747 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3433, "step": 7748 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.5696, "step": 7749 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3931, "step": 7750 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.4344, "step": 7751 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3605, "step": 7752 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.5265, "step": 7753 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.543, "step": 7754 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.5006, "step": 7755 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.4699, "step": 7756 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.5149, "step": 7757 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.525, "step": 7758 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.369, "step": 7759 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3552, "step": 7760 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.355, "step": 7761 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.394, "step": 7762 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3846, "step": 7763 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.5927, "step": 7764 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3486, "step": 7765 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3346, "step": 7766 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.5233, "step": 7767 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.333, "step": 7768 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.4441, "step": 7769 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.2938, "step": 7770 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.3641, "step": 7771 }, { "epoch": 2.9, "learning_rate": 2e-05, "loss": 0.4113, "step": 7772 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.6017, "step": 7773 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3636, "step": 7774 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4275, "step": 7775 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4664, "step": 7776 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4158, "step": 7777 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3309, "step": 7778 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.2894, "step": 7779 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3724, "step": 7780 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3583, "step": 7781 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4802, "step": 7782 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.5226, "step": 7783 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3878, "step": 7784 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.6209, "step": 7785 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3688, "step": 7786 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3645, "step": 7787 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3658, "step": 7788 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4262, "step": 7789 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3821, "step": 7790 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4354, "step": 7791 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4576, "step": 7792 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.437, "step": 7793 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.6238, "step": 7794 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4379, "step": 7795 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.3822, "step": 7796 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4183, "step": 7797 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.5599, "step": 7798 }, { "epoch": 2.91, "learning_rate": 2e-05, "loss": 0.4099, "step": 7799 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.7191, "step": 7800 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.347, "step": 7801 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.3108, "step": 7802 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.2467, "step": 7803 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.6082, "step": 7804 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.4128, "step": 7805 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.3705, "step": 7806 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.4435, "step": 7807 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.5541, "step": 7808 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.3593, "step": 7809 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.5305, "step": 7810 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.6749, "step": 7811 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.4384, "step": 7812 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.4739, "step": 7813 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.614, "step": 7814 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.5922, "step": 7815 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.59, "step": 7816 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.5527, "step": 7817 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.3673, "step": 7818 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.4109, "step": 7819 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.3416, "step": 7820 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.5549, "step": 7821 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.3563, "step": 7822 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.5018, "step": 7823 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.4321, "step": 7824 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.7285, "step": 7825 }, { "epoch": 2.92, "learning_rate": 2e-05, "loss": 0.415, "step": 7826 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.5412, "step": 7827 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.3342, "step": 7828 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.2768, "step": 7829 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.353, "step": 7830 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.54, "step": 7831 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4866, "step": 7832 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.3435, "step": 7833 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.3846, "step": 7834 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.3961, "step": 7835 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.5009, "step": 7836 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4932, "step": 7837 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4476, "step": 7838 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.3749, "step": 7839 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4402, "step": 7840 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.6112, "step": 7841 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4616, "step": 7842 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4344, "step": 7843 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.5974, "step": 7844 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4482, "step": 7845 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4324, "step": 7846 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.3898, "step": 7847 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4518, "step": 7848 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.3785, "step": 7849 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.7547, "step": 7850 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.51, "step": 7851 }, { "epoch": 2.93, "learning_rate": 2e-05, "loss": 0.4123, "step": 7852 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4106, "step": 7853 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4275, "step": 7854 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.6012, "step": 7855 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4634, "step": 7856 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.5177, "step": 7857 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4249, "step": 7858 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.5783, "step": 7859 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.3687, "step": 7860 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.377, "step": 7861 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.6061, "step": 7862 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4095, "step": 7863 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.3956, "step": 7864 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.3895, "step": 7865 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.407, "step": 7866 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.724, "step": 7867 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4958, "step": 7868 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4128, "step": 7869 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.3434, "step": 7870 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.6179, "step": 7871 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.375, "step": 7872 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.3872, "step": 7873 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4479, "step": 7874 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.2949, "step": 7875 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.3506, "step": 7876 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.3806, "step": 7877 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.5517, "step": 7878 }, { "epoch": 2.94, "learning_rate": 2e-05, "loss": 0.4171, "step": 7879 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4181, "step": 7880 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.3154, "step": 7881 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.5964, "step": 7882 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4589, "step": 7883 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.38, "step": 7884 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.3917, "step": 7885 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.3459, "step": 7886 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.7305, "step": 7887 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.3461, "step": 7888 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4704, "step": 7889 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.53, "step": 7890 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.9271, "step": 7891 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4946, "step": 7892 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4999, "step": 7893 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.5199, "step": 7894 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4085, "step": 7895 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.5509, "step": 7896 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4589, "step": 7897 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4613, "step": 7898 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4496, "step": 7899 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4267, "step": 7900 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.3924, "step": 7901 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.347, "step": 7902 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.4194, "step": 7903 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.3717, "step": 7904 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.6593, "step": 7905 }, { "epoch": 2.95, "learning_rate": 2e-05, "loss": 0.3886, "step": 7906 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.3169, "step": 7907 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5066, "step": 7908 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 1.0899, "step": 7909 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.495, "step": 7910 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.4028, "step": 7911 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.3514, "step": 7912 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.4872, "step": 7913 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.4192, "step": 7914 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5118, "step": 7915 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5302, "step": 7916 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.3014, "step": 7917 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 1.0035, "step": 7918 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.3915, "step": 7919 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.4622, "step": 7920 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5809, "step": 7921 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5196, "step": 7922 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5427, "step": 7923 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.4443, "step": 7924 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5438, "step": 7925 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5316, "step": 7926 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.4202, "step": 7927 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.3571, "step": 7928 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5961, "step": 7929 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.3906, "step": 7930 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5758, "step": 7931 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5416, "step": 7932 }, { "epoch": 2.96, "learning_rate": 2e-05, "loss": 0.5503, "step": 7933 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4842, "step": 7934 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3546, "step": 7935 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4874, "step": 7936 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.443, "step": 7937 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3419, "step": 7938 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.44, "step": 7939 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3025, "step": 7940 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3464, "step": 7941 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3242, "step": 7942 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4516, "step": 7943 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4905, "step": 7944 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.805, "step": 7945 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.5666, "step": 7946 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3661, "step": 7947 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3925, "step": 7948 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.2539, "step": 7949 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4954, "step": 7950 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4752, "step": 7951 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4292, "step": 7952 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.395, "step": 7953 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.5691, "step": 7954 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.3747, "step": 7955 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4343, "step": 7956 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4783, "step": 7957 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.461, "step": 7958 }, { "epoch": 2.97, "learning_rate": 2e-05, "loss": 0.4074, "step": 7959 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.556, "step": 7960 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4173, "step": 7961 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4634, "step": 7962 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4405, "step": 7963 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3999, "step": 7964 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.6092, "step": 7965 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.5503, "step": 7966 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.5096, "step": 7967 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.5832, "step": 7968 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3957, "step": 7969 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.8837, "step": 7970 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3601, "step": 7971 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.5124, "step": 7972 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3478, "step": 7973 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3379, "step": 7974 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3903, "step": 7975 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3968, "step": 7976 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4376, "step": 7977 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4122, "step": 7978 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.536, "step": 7979 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.3614, "step": 7980 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.5562, "step": 7981 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4839, "step": 7982 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.6032, "step": 7983 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4386, "step": 7984 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.4708, "step": 7985 }, { "epoch": 2.98, "learning_rate": 2e-05, "loss": 0.6846, "step": 7986 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.4764, "step": 7987 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.5133, "step": 7988 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.4448, "step": 7989 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.4427, "step": 7990 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.5345, "step": 7991 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.5987, "step": 7992 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.4198, "step": 7993 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.4156, "step": 7994 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.4164, "step": 7995 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.3839, "step": 7996 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.6234, "step": 7997 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.8303, "step": 7998 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.3319, "step": 7999 }, { "epoch": 2.99, "learning_rate": 2e-05, "loss": 0.4429, "step": 8000 } ], "max_steps": 8025, "num_train_epochs": 3, "total_flos": 1024787635126272.0, "trial_name": null, "trial_params": null }