diff --git "a/trainer_state.json" "b/trainer_state.json" new file mode 100644--- /dev/null +++ "b/trainer_state.json" @@ -0,0 +1,30016 @@ +{ + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 0.12281994595922378, + "global_step": 5000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.6366612111292965e-08, + "loss": 2.6258, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 3.273322422258593e-08, + "loss": 2.7051, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 4.909983633387889e-08, + "loss": 2.6564, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 6.546644844517186e-08, + "loss": 2.631, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 8.183306055646483e-08, + "loss": 2.7386, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 9.819967266775778e-08, + "loss": 2.6464, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 1.1456628477905074e-07, + "loss": 2.6867, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 1.3093289689034372e-07, + "loss": 2.6739, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 1.4729950900163667e-07, + "loss": 2.6253, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1.6366612111292966e-07, + "loss": 2.6906, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.800327332242226e-07, + "loss": 2.5914, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 1.9639934533551557e-07, + "loss": 2.6922, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 2.1276595744680852e-07, + "loss": 2.6864, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 2.2913256955810148e-07, + "loss": 2.635, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 2.4549918166939443e-07, + "loss": 2.6713, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 2.6186579378068744e-07, + "loss": 2.6551, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 2.782324058919804e-07, + "loss": 2.6033, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 2.9459901800327335e-07, + "loss": 2.6177, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 3.109656301145663e-07, + "loss": 2.589, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 3.273322422258593e-07, + "loss": 2.6455, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 3.436988543371522e-07, + "loss": 2.6395, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 3.600654664484452e-07, + "loss": 2.6208, + "step": 22 + }, + { + "epoch": 0.0, + "learning_rate": 3.764320785597381e-07, + "loss": 2.5779, + "step": 23 + }, + { + "epoch": 0.0, + "learning_rate": 3.9279869067103113e-07, + "loss": 2.5547, + "step": 24 + }, + { + "epoch": 0.0, + "learning_rate": 4.0916530278232414e-07, + "loss": 2.425, + "step": 25 + }, + { + "epoch": 0.0, + "learning_rate": 4.2553191489361704e-07, + "loss": 2.4681, + "step": 26 + }, + { + "epoch": 0.0, + "learning_rate": 4.4189852700491005e-07, + "loss": 2.5412, + "step": 27 + }, + { + "epoch": 0.0, + "learning_rate": 4.5826513911620295e-07, + "loss": 2.454, + "step": 28 + }, + { + "epoch": 0.0, + "learning_rate": 4.7463175122749596e-07, + "loss": 2.4572, + "step": 29 + }, + { + "epoch": 0.0, + "learning_rate": 4.909983633387889e-07, + "loss": 2.4587, + "step": 30 + }, + { + "epoch": 0.0, + "learning_rate": 5.073649754500819e-07, + "loss": 2.4182, + "step": 31 + }, + { + "epoch": 0.0, + "learning_rate": 5.237315875613749e-07, + "loss": 2.3269, + "step": 32 + }, + { + "epoch": 0.0, + "learning_rate": 5.400981996726678e-07, + "loss": 2.225, + "step": 33 + }, + { + "epoch": 0.0, + "learning_rate": 5.564648117839608e-07, + "loss": 2.2196, + "step": 34 + }, + { + "epoch": 0.0, + "learning_rate": 5.728314238952537e-07, + "loss": 2.2227, + "step": 35 + }, + { + "epoch": 0.0, + "learning_rate": 5.891980360065467e-07, + "loss": 2.1143, + "step": 36 + }, + { + "epoch": 0.0, + "learning_rate": 6.055646481178397e-07, + "loss": 2.1236, + "step": 37 + }, + { + "epoch": 0.0, + "learning_rate": 6.219312602291326e-07, + "loss": 2.1368, + "step": 38 + }, + { + "epoch": 0.0, + "learning_rate": 6.382978723404255e-07, + "loss": 2.1354, + "step": 39 + }, + { + "epoch": 0.0, + "learning_rate": 6.546644844517186e-07, + "loss": 2.169, + "step": 40 + }, + { + "epoch": 0.0, + "learning_rate": 6.710310965630115e-07, + "loss": 2.097, + "step": 41 + }, + { + "epoch": 0.0, + "learning_rate": 6.873977086743044e-07, + "loss": 2.0275, + "step": 42 + }, + { + "epoch": 0.0, + "learning_rate": 7.037643207855975e-07, + "loss": 2.0292, + "step": 43 + }, + { + "epoch": 0.0, + "learning_rate": 7.201309328968904e-07, + "loss": 1.9631, + "step": 44 + }, + { + "epoch": 0.0, + "learning_rate": 7.364975450081833e-07, + "loss": 1.8798, + "step": 45 + }, + { + "epoch": 0.0, + "learning_rate": 7.528641571194762e-07, + "loss": 1.8223, + "step": 46 + }, + { + "epoch": 0.0, + "learning_rate": 7.692307692307694e-07, + "loss": 1.7422, + "step": 47 + }, + { + "epoch": 0.0, + "learning_rate": 7.855973813420623e-07, + "loss": 1.7074, + "step": 48 + }, + { + "epoch": 0.0, + "learning_rate": 8.019639934533552e-07, + "loss": 1.6586, + "step": 49 + }, + { + "epoch": 0.0, + "learning_rate": 8.183306055646483e-07, + "loss": 1.6396, + "step": 50 + }, + { + "epoch": 0.0, + "learning_rate": 8.346972176759412e-07, + "loss": 1.5771, + "step": 51 + }, + { + "epoch": 0.0, + "learning_rate": 8.510638297872341e-07, + "loss": 1.614, + "step": 52 + }, + { + "epoch": 0.0, + "learning_rate": 8.67430441898527e-07, + "loss": 1.6324, + "step": 53 + }, + { + "epoch": 0.0, + "learning_rate": 8.837970540098201e-07, + "loss": 1.5733, + "step": 54 + }, + { + "epoch": 0.0, + "learning_rate": 9.00163666121113e-07, + "loss": 1.5885, + "step": 55 + }, + { + "epoch": 0.0, + "learning_rate": 9.165302782324059e-07, + "loss": 1.6219, + "step": 56 + }, + { + "epoch": 0.0, + "learning_rate": 9.32896890343699e-07, + "loss": 1.6256, + "step": 57 + }, + { + "epoch": 0.0, + "learning_rate": 9.492635024549919e-07, + "loss": 1.5878, + "step": 58 + }, + { + "epoch": 0.0, + "learning_rate": 9.656301145662848e-07, + "loss": 1.5361, + "step": 59 + }, + { + "epoch": 0.0, + "learning_rate": 9.819967266775777e-07, + "loss": 1.5139, + "step": 60 + }, + { + "epoch": 0.0, + "learning_rate": 9.983633387888708e-07, + "loss": 1.5575, + "step": 61 + }, + { + "epoch": 0.0, + "learning_rate": 1.0147299509001637e-06, + "loss": 1.5249, + "step": 62 + }, + { + "epoch": 0.0, + "learning_rate": 1.0310965630114566e-06, + "loss": 1.4732, + "step": 63 + }, + { + "epoch": 0.0, + "learning_rate": 1.0474631751227498e-06, + "loss": 1.4938, + "step": 64 + }, + { + "epoch": 0.0, + "learning_rate": 1.0638297872340427e-06, + "loss": 1.4396, + "step": 65 + }, + { + "epoch": 0.0, + "learning_rate": 1.0801963993453356e-06, + "loss": 1.5008, + "step": 66 + }, + { + "epoch": 0.0, + "learning_rate": 1.0965630114566287e-06, + "loss": 1.4616, + "step": 67 + }, + { + "epoch": 0.0, + "learning_rate": 1.1129296235679216e-06, + "loss": 1.488, + "step": 68 + }, + { + "epoch": 0.0, + "learning_rate": 1.1292962356792145e-06, + "loss": 1.4738, + "step": 69 + }, + { + "epoch": 0.0, + "learning_rate": 1.1456628477905074e-06, + "loss": 1.4244, + "step": 70 + }, + { + "epoch": 0.0, + "learning_rate": 1.1620294599018005e-06, + "loss": 1.4317, + "step": 71 + }, + { + "epoch": 0.0, + "learning_rate": 1.1783960720130934e-06, + "loss": 1.4194, + "step": 72 + }, + { + "epoch": 0.0, + "learning_rate": 1.1947626841243863e-06, + "loss": 1.4616, + "step": 73 + }, + { + "epoch": 0.0, + "learning_rate": 1.2111292962356794e-06, + "loss": 1.4109, + "step": 74 + }, + { + "epoch": 0.0, + "learning_rate": 1.2274959083469723e-06, + "loss": 1.4496, + "step": 75 + }, + { + "epoch": 0.0, + "learning_rate": 1.2438625204582652e-06, + "loss": 1.4184, + "step": 76 + }, + { + "epoch": 0.0, + "learning_rate": 1.2602291325695581e-06, + "loss": 1.4258, + "step": 77 + }, + { + "epoch": 0.0, + "learning_rate": 1.276595744680851e-06, + "loss": 1.3879, + "step": 78 + }, + { + "epoch": 0.0, + "learning_rate": 1.2929623567921443e-06, + "loss": 1.4428, + "step": 79 + }, + { + "epoch": 0.0, + "learning_rate": 1.3093289689034372e-06, + "loss": 1.4358, + "step": 80 + }, + { + "epoch": 0.0, + "learning_rate": 1.3256955810147301e-06, + "loss": 1.4048, + "step": 81 + }, + { + "epoch": 0.0, + "learning_rate": 1.342062193126023e-06, + "loss": 1.3767, + "step": 82 + }, + { + "epoch": 0.0, + "learning_rate": 1.358428805237316e-06, + "loss": 1.4008, + "step": 83 + }, + { + "epoch": 0.0, + "learning_rate": 1.3747954173486089e-06, + "loss": 1.3627, + "step": 84 + }, + { + "epoch": 0.0, + "learning_rate": 1.3911620294599018e-06, + "loss": 1.4065, + "step": 85 + }, + { + "epoch": 0.0, + "learning_rate": 1.407528641571195e-06, + "loss": 1.384, + "step": 86 + }, + { + "epoch": 0.0, + "learning_rate": 1.423895253682488e-06, + "loss": 1.4147, + "step": 87 + }, + { + "epoch": 0.0, + "learning_rate": 1.4402618657937809e-06, + "loss": 1.374, + "step": 88 + }, + { + "epoch": 0.0, + "learning_rate": 1.4566284779050738e-06, + "loss": 1.3901, + "step": 89 + }, + { + "epoch": 0.0, + "learning_rate": 1.4729950900163667e-06, + "loss": 1.356, + "step": 90 + }, + { + "epoch": 0.0, + "learning_rate": 1.4893617021276596e-06, + "loss": 1.4547, + "step": 91 + }, + { + "epoch": 0.0, + "learning_rate": 1.5057283142389525e-06, + "loss": 1.3666, + "step": 92 + }, + { + "epoch": 0.0, + "learning_rate": 1.5220949263502458e-06, + "loss": 1.3959, + "step": 93 + }, + { + "epoch": 0.0, + "learning_rate": 1.5384615384615387e-06, + "loss": 1.3565, + "step": 94 + }, + { + "epoch": 0.0, + "learning_rate": 1.5548281505728316e-06, + "loss": 1.4213, + "step": 95 + }, + { + "epoch": 0.0, + "learning_rate": 1.5711947626841245e-06, + "loss": 1.4207, + "step": 96 + }, + { + "epoch": 0.0, + "learning_rate": 1.5875613747954174e-06, + "loss": 1.3497, + "step": 97 + }, + { + "epoch": 0.0, + "learning_rate": 1.6039279869067103e-06, + "loss": 1.3487, + "step": 98 + }, + { + "epoch": 0.0, + "learning_rate": 1.6202945990180032e-06, + "loss": 1.3555, + "step": 99 + }, + { + "epoch": 0.0, + "learning_rate": 1.6366612111292966e-06, + "loss": 1.3834, + "step": 100 + }, + { + "epoch": 0.0, + "learning_rate": 1.6530278232405895e-06, + "loss": 1.3758, + "step": 101 + }, + { + "epoch": 0.0, + "learning_rate": 1.6693944353518824e-06, + "loss": 1.3809, + "step": 102 + }, + { + "epoch": 0.0, + "learning_rate": 1.6857610474631753e-06, + "loss": 1.3684, + "step": 103 + }, + { + "epoch": 0.0, + "learning_rate": 1.7021276595744682e-06, + "loss": 1.3458, + "step": 104 + }, + { + "epoch": 0.0, + "learning_rate": 1.718494271685761e-06, + "loss": 1.3872, + "step": 105 + }, + { + "epoch": 0.0, + "learning_rate": 1.734860883797054e-06, + "loss": 1.3875, + "step": 106 + }, + { + "epoch": 0.0, + "learning_rate": 1.7512274959083473e-06, + "loss": 1.3345, + "step": 107 + }, + { + "epoch": 0.0, + "learning_rate": 1.7675941080196402e-06, + "loss": 1.3437, + "step": 108 + }, + { + "epoch": 0.0, + "learning_rate": 1.783960720130933e-06, + "loss": 1.3825, + "step": 109 + }, + { + "epoch": 0.0, + "learning_rate": 1.800327332242226e-06, + "loss": 1.3513, + "step": 110 + }, + { + "epoch": 0.0, + "learning_rate": 1.816693944353519e-06, + "loss": 1.3321, + "step": 111 + }, + { + "epoch": 0.0, + "learning_rate": 1.8330605564648118e-06, + "loss": 1.3042, + "step": 112 + }, + { + "epoch": 0.0, + "learning_rate": 1.8494271685761047e-06, + "loss": 1.3707, + "step": 113 + }, + { + "epoch": 0.0, + "learning_rate": 1.865793780687398e-06, + "loss": 1.309, + "step": 114 + }, + { + "epoch": 0.0, + "learning_rate": 1.882160392798691e-06, + "loss": 1.3349, + "step": 115 + }, + { + "epoch": 0.0, + "learning_rate": 1.8985270049099838e-06, + "loss": 1.3043, + "step": 116 + }, + { + "epoch": 0.0, + "learning_rate": 1.9148936170212767e-06, + "loss": 1.318, + "step": 117 + }, + { + "epoch": 0.0, + "learning_rate": 1.9312602291325696e-06, + "loss": 1.304, + "step": 118 + }, + { + "epoch": 0.0, + "learning_rate": 1.9476268412438625e-06, + "loss": 1.3417, + "step": 119 + }, + { + "epoch": 0.0, + "learning_rate": 1.9639934533551554e-06, + "loss": 1.3389, + "step": 120 + }, + { + "epoch": 0.0, + "learning_rate": 1.9803600654664488e-06, + "loss": 1.3566, + "step": 121 + }, + { + "epoch": 0.0, + "learning_rate": 1.9967266775777417e-06, + "loss": 1.3266, + "step": 122 + }, + { + "epoch": 0.0, + "learning_rate": 2.0130932896890346e-06, + "loss": 1.3127, + "step": 123 + }, + { + "epoch": 0.0, + "learning_rate": 2.0294599018003275e-06, + "loss": 1.3109, + "step": 124 + }, + { + "epoch": 0.0, + "learning_rate": 2.0458265139116204e-06, + "loss": 1.3149, + "step": 125 + }, + { + "epoch": 0.0, + "learning_rate": 2.0621931260229133e-06, + "loss": 1.2847, + "step": 126 + }, + { + "epoch": 0.0, + "learning_rate": 2.078559738134206e-06, + "loss": 1.3338, + "step": 127 + }, + { + "epoch": 0.0, + "learning_rate": 2.0949263502454995e-06, + "loss": 1.295, + "step": 128 + }, + { + "epoch": 0.0, + "learning_rate": 2.1112929623567924e-06, + "loss": 1.3132, + "step": 129 + }, + { + "epoch": 0.0, + "learning_rate": 2.1276595744680853e-06, + "loss": 1.3144, + "step": 130 + }, + { + "epoch": 0.0, + "learning_rate": 2.144026186579378e-06, + "loss": 1.3353, + "step": 131 + }, + { + "epoch": 0.0, + "learning_rate": 2.160392798690671e-06, + "loss": 1.3534, + "step": 132 + }, + { + "epoch": 0.0, + "learning_rate": 2.176759410801964e-06, + "loss": 1.3348, + "step": 133 + }, + { + "epoch": 0.0, + "learning_rate": 2.1931260229132573e-06, + "loss": 1.3273, + "step": 134 + }, + { + "epoch": 0.0, + "learning_rate": 2.2094926350245502e-06, + "loss": 1.3079, + "step": 135 + }, + { + "epoch": 0.0, + "learning_rate": 2.225859247135843e-06, + "loss": 1.3527, + "step": 136 + }, + { + "epoch": 0.0, + "learning_rate": 2.242225859247136e-06, + "loss": 1.2992, + "step": 137 + }, + { + "epoch": 0.0, + "learning_rate": 2.258592471358429e-06, + "loss": 1.3007, + "step": 138 + }, + { + "epoch": 0.0, + "learning_rate": 2.274959083469722e-06, + "loss": 1.2791, + "step": 139 + }, + { + "epoch": 0.0, + "learning_rate": 2.2913256955810148e-06, + "loss": 1.3023, + "step": 140 + }, + { + "epoch": 0.0, + "learning_rate": 2.307692307692308e-06, + "loss": 1.3147, + "step": 141 + }, + { + "epoch": 0.0, + "learning_rate": 2.324058919803601e-06, + "loss": 1.29, + "step": 142 + }, + { + "epoch": 0.0, + "learning_rate": 2.340425531914894e-06, + "loss": 1.3149, + "step": 143 + }, + { + "epoch": 0.0, + "learning_rate": 2.3567921440261868e-06, + "loss": 1.3127, + "step": 144 + }, + { + "epoch": 0.0, + "learning_rate": 2.3731587561374797e-06, + "loss": 1.3232, + "step": 145 + }, + { + "epoch": 0.0, + "learning_rate": 2.3895253682487726e-06, + "loss": 1.3403, + "step": 146 + }, + { + "epoch": 0.0, + "learning_rate": 2.4058919803600655e-06, + "loss": 1.3003, + "step": 147 + }, + { + "epoch": 0.0, + "learning_rate": 2.422258592471359e-06, + "loss": 1.2872, + "step": 148 + }, + { + "epoch": 0.0, + "learning_rate": 2.4386252045826517e-06, + "loss": 1.3348, + "step": 149 + }, + { + "epoch": 0.0, + "learning_rate": 2.4549918166939446e-06, + "loss": 1.3698, + "step": 150 + }, + { + "epoch": 0.0, + "learning_rate": 2.4713584288052375e-06, + "loss": 1.2556, + "step": 151 + }, + { + "epoch": 0.0, + "learning_rate": 2.4877250409165304e-06, + "loss": 1.304, + "step": 152 + }, + { + "epoch": 0.0, + "learning_rate": 2.5040916530278237e-06, + "loss": 1.309, + "step": 153 + }, + { + "epoch": 0.0, + "learning_rate": 2.5204582651391162e-06, + "loss": 1.2841, + "step": 154 + }, + { + "epoch": 0.0, + "learning_rate": 2.5368248772504096e-06, + "loss": 1.287, + "step": 155 + }, + { + "epoch": 0.0, + "learning_rate": 2.553191489361702e-06, + "loss": 1.3111, + "step": 156 + }, + { + "epoch": 0.0, + "learning_rate": 2.5695581014729954e-06, + "loss": 1.2641, + "step": 157 + }, + { + "epoch": 0.0, + "learning_rate": 2.5859247135842887e-06, + "loss": 1.2689, + "step": 158 + }, + { + "epoch": 0.0, + "learning_rate": 2.602291325695581e-06, + "loss": 1.2817, + "step": 159 + }, + { + "epoch": 0.0, + "learning_rate": 2.6186579378068745e-06, + "loss": 1.2981, + "step": 160 + }, + { + "epoch": 0.0, + "learning_rate": 2.635024549918167e-06, + "loss": 1.294, + "step": 161 + }, + { + "epoch": 0.0, + "learning_rate": 2.6513911620294603e-06, + "loss": 1.2907, + "step": 162 + }, + { + "epoch": 0.0, + "learning_rate": 2.6677577741407528e-06, + "loss": 1.3094, + "step": 163 + }, + { + "epoch": 0.0, + "learning_rate": 2.684124386252046e-06, + "loss": 1.2588, + "step": 164 + }, + { + "epoch": 0.0, + "learning_rate": 2.7004909983633394e-06, + "loss": 1.3106, + "step": 165 + }, + { + "epoch": 0.0, + "learning_rate": 2.716857610474632e-06, + "loss": 1.2916, + "step": 166 + }, + { + "epoch": 0.0, + "learning_rate": 2.7332242225859252e-06, + "loss": 1.2959, + "step": 167 + }, + { + "epoch": 0.0, + "learning_rate": 2.7495908346972177e-06, + "loss": 1.2823, + "step": 168 + }, + { + "epoch": 0.0, + "learning_rate": 2.765957446808511e-06, + "loss": 1.2732, + "step": 169 + }, + { + "epoch": 0.0, + "learning_rate": 2.7823240589198035e-06, + "loss": 1.2743, + "step": 170 + }, + { + "epoch": 0.0, + "learning_rate": 2.798690671031097e-06, + "loss": 1.2633, + "step": 171 + }, + { + "epoch": 0.0, + "learning_rate": 2.81505728314239e-06, + "loss": 1.3051, + "step": 172 + }, + { + "epoch": 0.0, + "learning_rate": 2.8314238952536826e-06, + "loss": 1.255, + "step": 173 + }, + { + "epoch": 0.0, + "learning_rate": 2.847790507364976e-06, + "loss": 1.297, + "step": 174 + }, + { + "epoch": 0.0, + "learning_rate": 2.8641571194762684e-06, + "loss": 1.2937, + "step": 175 + }, + { + "epoch": 0.0, + "learning_rate": 2.8805237315875618e-06, + "loss": 1.2591, + "step": 176 + }, + { + "epoch": 0.0, + "learning_rate": 2.8968903436988542e-06, + "loss": 1.2888, + "step": 177 + }, + { + "epoch": 0.0, + "learning_rate": 2.9132569558101476e-06, + "loss": 1.284, + "step": 178 + }, + { + "epoch": 0.0, + "learning_rate": 2.929623567921441e-06, + "loss": 1.2999, + "step": 179 + }, + { + "epoch": 0.0, + "learning_rate": 2.9459901800327334e-06, + "loss": 1.2546, + "step": 180 + }, + { + "epoch": 0.0, + "learning_rate": 2.9623567921440267e-06, + "loss": 1.279, + "step": 181 + }, + { + "epoch": 0.0, + "learning_rate": 2.978723404255319e-06, + "loss": 1.3058, + "step": 182 + }, + { + "epoch": 0.0, + "learning_rate": 2.9950900163666125e-06, + "loss": 1.2931, + "step": 183 + }, + { + "epoch": 0.0, + "learning_rate": 3.011456628477905e-06, + "loss": 1.2775, + "step": 184 + }, + { + "epoch": 0.0, + "learning_rate": 3.0278232405891983e-06, + "loss": 1.3034, + "step": 185 + }, + { + "epoch": 0.0, + "learning_rate": 3.0441898527004916e-06, + "loss": 1.2647, + "step": 186 + }, + { + "epoch": 0.0, + "learning_rate": 3.060556464811784e-06, + "loss": 1.2459, + "step": 187 + }, + { + "epoch": 0.0, + "learning_rate": 3.0769230769230774e-06, + "loss": 1.2472, + "step": 188 + }, + { + "epoch": 0.0, + "learning_rate": 3.09328968903437e-06, + "loss": 1.2773, + "step": 189 + }, + { + "epoch": 0.0, + "learning_rate": 3.1096563011456632e-06, + "loss": 1.2874, + "step": 190 + }, + { + "epoch": 0.0, + "learning_rate": 3.1260229132569557e-06, + "loss": 1.2869, + "step": 191 + }, + { + "epoch": 0.0, + "learning_rate": 3.142389525368249e-06, + "loss": 1.2595, + "step": 192 + }, + { + "epoch": 0.0, + "learning_rate": 3.1587561374795424e-06, + "loss": 1.2695, + "step": 193 + }, + { + "epoch": 0.0, + "learning_rate": 3.175122749590835e-06, + "loss": 1.2376, + "step": 194 + }, + { + "epoch": 0.0, + "learning_rate": 3.191489361702128e-06, + "loss": 1.3041, + "step": 195 + }, + { + "epoch": 0.0, + "learning_rate": 3.2078559738134207e-06, + "loss": 1.2485, + "step": 196 + }, + { + "epoch": 0.0, + "learning_rate": 3.224222585924714e-06, + "loss": 1.2842, + "step": 197 + }, + { + "epoch": 0.0, + "learning_rate": 3.2405891980360065e-06, + "loss": 1.264, + "step": 198 + }, + { + "epoch": 0.0, + "learning_rate": 3.2569558101472998e-06, + "loss": 1.2616, + "step": 199 + }, + { + "epoch": 0.0, + "learning_rate": 3.273322422258593e-06, + "loss": 1.2976, + "step": 200 + }, + { + "epoch": 0.0, + "learning_rate": 3.2896890343698856e-06, + "loss": 1.2634, + "step": 201 + }, + { + "epoch": 0.0, + "learning_rate": 3.306055646481179e-06, + "loss": 1.2711, + "step": 202 + }, + { + "epoch": 0.0, + "learning_rate": 3.3224222585924714e-06, + "loss": 1.2558, + "step": 203 + }, + { + "epoch": 0.01, + "learning_rate": 3.3387888707037647e-06, + "loss": 1.2592, + "step": 204 + }, + { + "epoch": 0.01, + "learning_rate": 3.355155482815057e-06, + "loss": 1.2716, + "step": 205 + }, + { + "epoch": 0.01, + "learning_rate": 3.3715220949263505e-06, + "loss": 1.2707, + "step": 206 + }, + { + "epoch": 0.01, + "learning_rate": 3.387888707037644e-06, + "loss": 1.2925, + "step": 207 + }, + { + "epoch": 0.01, + "learning_rate": 3.4042553191489363e-06, + "loss": 1.1993, + "step": 208 + }, + { + "epoch": 0.01, + "learning_rate": 3.4206219312602296e-06, + "loss": 1.242, + "step": 209 + }, + { + "epoch": 0.01, + "learning_rate": 3.436988543371522e-06, + "loss": 1.2974, + "step": 210 + }, + { + "epoch": 0.01, + "learning_rate": 3.4533551554828155e-06, + "loss": 1.227, + "step": 211 + }, + { + "epoch": 0.01, + "learning_rate": 3.469721767594108e-06, + "loss": 1.2617, + "step": 212 + }, + { + "epoch": 0.01, + "learning_rate": 3.4860883797054013e-06, + "loss": 1.246, + "step": 213 + }, + { + "epoch": 0.01, + "learning_rate": 3.5024549918166946e-06, + "loss": 1.2216, + "step": 214 + }, + { + "epoch": 0.01, + "learning_rate": 3.518821603927987e-06, + "loss": 1.232, + "step": 215 + }, + { + "epoch": 0.01, + "learning_rate": 3.5351882160392804e-06, + "loss": 1.23, + "step": 216 + }, + { + "epoch": 0.01, + "learning_rate": 3.551554828150573e-06, + "loss": 1.2291, + "step": 217 + }, + { + "epoch": 0.01, + "learning_rate": 3.567921440261866e-06, + "loss": 1.2591, + "step": 218 + }, + { + "epoch": 0.01, + "learning_rate": 3.5842880523731587e-06, + "loss": 1.2586, + "step": 219 + }, + { + "epoch": 0.01, + "learning_rate": 3.600654664484452e-06, + "loss": 1.2393, + "step": 220 + }, + { + "epoch": 0.01, + "learning_rate": 3.6170212765957453e-06, + "loss": 1.2312, + "step": 221 + }, + { + "epoch": 0.01, + "learning_rate": 3.633387888707038e-06, + "loss": 1.2169, + "step": 222 + }, + { + "epoch": 0.01, + "learning_rate": 3.649754500818331e-06, + "loss": 1.2506, + "step": 223 + }, + { + "epoch": 0.01, + "learning_rate": 3.6661211129296236e-06, + "loss": 1.2425, + "step": 224 + }, + { + "epoch": 0.01, + "learning_rate": 3.682487725040917e-06, + "loss": 1.2605, + "step": 225 + }, + { + "epoch": 0.01, + "learning_rate": 3.6988543371522094e-06, + "loss": 1.2744, + "step": 226 + }, + { + "epoch": 0.01, + "learning_rate": 3.7152209492635027e-06, + "loss": 1.2406, + "step": 227 + }, + { + "epoch": 0.01, + "learning_rate": 3.731587561374796e-06, + "loss": 1.2363, + "step": 228 + }, + { + "epoch": 0.01, + "learning_rate": 3.7479541734860885e-06, + "loss": 1.292, + "step": 229 + }, + { + "epoch": 0.01, + "learning_rate": 3.764320785597382e-06, + "loss": 1.2337, + "step": 230 + }, + { + "epoch": 0.01, + "learning_rate": 3.7806873977086743e-06, + "loss": 1.2517, + "step": 231 + }, + { + "epoch": 0.01, + "learning_rate": 3.7970540098199677e-06, + "loss": 1.1914, + "step": 232 + }, + { + "epoch": 0.01, + "learning_rate": 3.81342062193126e-06, + "loss": 1.2467, + "step": 233 + }, + { + "epoch": 0.01, + "learning_rate": 3.8297872340425535e-06, + "loss": 1.2876, + "step": 234 + }, + { + "epoch": 0.01, + "learning_rate": 3.846153846153847e-06, + "loss": 1.23, + "step": 235 + }, + { + "epoch": 0.01, + "learning_rate": 3.862520458265139e-06, + "loss": 1.253, + "step": 236 + }, + { + "epoch": 0.01, + "learning_rate": 3.878887070376433e-06, + "loss": 1.1973, + "step": 237 + }, + { + "epoch": 0.01, + "learning_rate": 3.895253682487725e-06, + "loss": 1.2723, + "step": 238 + }, + { + "epoch": 0.01, + "learning_rate": 3.911620294599018e-06, + "loss": 1.2428, + "step": 239 + }, + { + "epoch": 0.01, + "learning_rate": 3.927986906710311e-06, + "loss": 1.2424, + "step": 240 + }, + { + "epoch": 0.01, + "learning_rate": 3.944353518821604e-06, + "loss": 1.2424, + "step": 241 + }, + { + "epoch": 0.01, + "learning_rate": 3.9607201309328975e-06, + "loss": 1.2307, + "step": 242 + }, + { + "epoch": 0.01, + "learning_rate": 3.97708674304419e-06, + "loss": 1.2321, + "step": 243 + }, + { + "epoch": 0.01, + "learning_rate": 3.993453355155483e-06, + "loss": 1.2375, + "step": 244 + }, + { + "epoch": 0.01, + "learning_rate": 4.009819967266776e-06, + "loss": 1.2338, + "step": 245 + }, + { + "epoch": 0.01, + "learning_rate": 4.026186579378069e-06, + "loss": 1.2324, + "step": 246 + }, + { + "epoch": 0.01, + "learning_rate": 4.042553191489362e-06, + "loss": 1.2577, + "step": 247 + }, + { + "epoch": 0.01, + "learning_rate": 4.058919803600655e-06, + "loss": 1.2342, + "step": 248 + }, + { + "epoch": 0.01, + "learning_rate": 4.075286415711948e-06, + "loss": 1.2174, + "step": 249 + }, + { + "epoch": 0.01, + "learning_rate": 4.091653027823241e-06, + "loss": 1.2401, + "step": 250 + }, + { + "epoch": 0.01, + "learning_rate": 4.108019639934534e-06, + "loss": 1.2025, + "step": 251 + }, + { + "epoch": 0.01, + "learning_rate": 4.1243862520458266e-06, + "loss": 1.2475, + "step": 252 + }, + { + "epoch": 0.01, + "learning_rate": 4.14075286415712e-06, + "loss": 1.2286, + "step": 253 + }, + { + "epoch": 0.01, + "learning_rate": 4.157119476268412e-06, + "loss": 1.246, + "step": 254 + }, + { + "epoch": 0.01, + "learning_rate": 4.173486088379706e-06, + "loss": 1.2834, + "step": 255 + }, + { + "epoch": 0.01, + "learning_rate": 4.189852700490999e-06, + "loss": 1.2459, + "step": 256 + }, + { + "epoch": 0.01, + "learning_rate": 4.2062193126022915e-06, + "loss": 1.2616, + "step": 257 + }, + { + "epoch": 0.01, + "learning_rate": 4.222585924713585e-06, + "loss": 1.2536, + "step": 258 + }, + { + "epoch": 0.01, + "learning_rate": 4.238952536824877e-06, + "loss": 1.2949, + "step": 259 + }, + { + "epoch": 0.01, + "learning_rate": 4.255319148936171e-06, + "loss": 1.2785, + "step": 260 + }, + { + "epoch": 0.01, + "learning_rate": 4.271685761047464e-06, + "loss": 1.2198, + "step": 261 + }, + { + "epoch": 0.01, + "learning_rate": 4.288052373158756e-06, + "loss": 1.2338, + "step": 262 + }, + { + "epoch": 0.01, + "learning_rate": 4.30441898527005e-06, + "loss": 1.2261, + "step": 263 + }, + { + "epoch": 0.01, + "learning_rate": 4.320785597381342e-06, + "loss": 1.2371, + "step": 264 + }, + { + "epoch": 0.01, + "learning_rate": 4.3371522094926355e-06, + "loss": 1.2248, + "step": 265 + }, + { + "epoch": 0.01, + "learning_rate": 4.353518821603928e-06, + "loss": 1.255, + "step": 266 + }, + { + "epoch": 0.01, + "learning_rate": 4.369885433715221e-06, + "loss": 1.2119, + "step": 267 + }, + { + "epoch": 0.01, + "learning_rate": 4.386252045826515e-06, + "loss": 1.2301, + "step": 268 + }, + { + "epoch": 0.01, + "learning_rate": 4.402618657937807e-06, + "loss": 1.233, + "step": 269 + }, + { + "epoch": 0.01, + "learning_rate": 4.4189852700491005e-06, + "loss": 1.2097, + "step": 270 + }, + { + "epoch": 0.01, + "learning_rate": 4.435351882160393e-06, + "loss": 1.2752, + "step": 271 + }, + { + "epoch": 0.01, + "learning_rate": 4.451718494271686e-06, + "loss": 1.2535, + "step": 272 + }, + { + "epoch": 0.01, + "learning_rate": 4.468085106382979e-06, + "loss": 1.246, + "step": 273 + }, + { + "epoch": 0.01, + "learning_rate": 4.484451718494272e-06, + "loss": 1.2435, + "step": 274 + }, + { + "epoch": 0.01, + "learning_rate": 4.500818330605565e-06, + "loss": 1.2314, + "step": 275 + }, + { + "epoch": 0.01, + "learning_rate": 4.517184942716858e-06, + "loss": 1.2442, + "step": 276 + }, + { + "epoch": 0.01, + "learning_rate": 4.533551554828151e-06, + "loss": 1.2294, + "step": 277 + }, + { + "epoch": 0.01, + "learning_rate": 4.549918166939444e-06, + "loss": 1.2169, + "step": 278 + }, + { + "epoch": 0.01, + "learning_rate": 4.566284779050737e-06, + "loss": 1.2348, + "step": 279 + }, + { + "epoch": 0.01, + "learning_rate": 4.5826513911620295e-06, + "loss": 1.2205, + "step": 280 + }, + { + "epoch": 0.01, + "learning_rate": 4.599018003273323e-06, + "loss": 1.2323, + "step": 281 + }, + { + "epoch": 0.01, + "learning_rate": 4.615384615384616e-06, + "loss": 1.2045, + "step": 282 + }, + { + "epoch": 0.01, + "learning_rate": 4.631751227495909e-06, + "loss": 1.2161, + "step": 283 + }, + { + "epoch": 0.01, + "learning_rate": 4.648117839607202e-06, + "loss": 1.2314, + "step": 284 + }, + { + "epoch": 0.01, + "learning_rate": 4.6644844517184944e-06, + "loss": 1.2541, + "step": 285 + }, + { + "epoch": 0.01, + "learning_rate": 4.680851063829788e-06, + "loss": 1.237, + "step": 286 + }, + { + "epoch": 0.01, + "learning_rate": 4.69721767594108e-06, + "loss": 1.2194, + "step": 287 + }, + { + "epoch": 0.01, + "learning_rate": 4.7135842880523736e-06, + "loss": 1.2099, + "step": 288 + }, + { + "epoch": 0.01, + "learning_rate": 4.729950900163667e-06, + "loss": 1.2125, + "step": 289 + }, + { + "epoch": 0.01, + "learning_rate": 4.746317512274959e-06, + "loss": 1.2515, + "step": 290 + }, + { + "epoch": 0.01, + "learning_rate": 4.762684124386253e-06, + "loss": 1.2516, + "step": 291 + }, + { + "epoch": 0.01, + "learning_rate": 4.779050736497545e-06, + "loss": 1.1901, + "step": 292 + }, + { + "epoch": 0.01, + "learning_rate": 4.7954173486088385e-06, + "loss": 1.2221, + "step": 293 + }, + { + "epoch": 0.01, + "learning_rate": 4.811783960720131e-06, + "loss": 1.1883, + "step": 294 + }, + { + "epoch": 0.01, + "learning_rate": 4.828150572831424e-06, + "loss": 1.2221, + "step": 295 + }, + { + "epoch": 0.01, + "learning_rate": 4.844517184942718e-06, + "loss": 1.2157, + "step": 296 + }, + { + "epoch": 0.01, + "learning_rate": 4.86088379705401e-06, + "loss": 1.2155, + "step": 297 + }, + { + "epoch": 0.01, + "learning_rate": 4.8772504091653034e-06, + "loss": 1.2307, + "step": 298 + }, + { + "epoch": 0.01, + "learning_rate": 4.893617021276596e-06, + "loss": 1.2359, + "step": 299 + }, + { + "epoch": 0.01, + "learning_rate": 4.909983633387889e-06, + "loss": 1.2527, + "step": 300 + }, + { + "epoch": 0.01, + "learning_rate": 4.926350245499182e-06, + "loss": 1.2312, + "step": 301 + }, + { + "epoch": 0.01, + "learning_rate": 4.942716857610475e-06, + "loss": 1.205, + "step": 302 + }, + { + "epoch": 0.01, + "learning_rate": 4.959083469721768e-06, + "loss": 1.2077, + "step": 303 + }, + { + "epoch": 0.01, + "learning_rate": 4.975450081833061e-06, + "loss": 1.2846, + "step": 304 + }, + { + "epoch": 0.01, + "learning_rate": 4.991816693944354e-06, + "loss": 1.2409, + "step": 305 + }, + { + "epoch": 0.01, + "learning_rate": 5.0081833060556475e-06, + "loss": 1.1738, + "step": 306 + }, + { + "epoch": 0.01, + "learning_rate": 5.02454991816694e-06, + "loss": 1.2259, + "step": 307 + }, + { + "epoch": 0.01, + "learning_rate": 5.0409165302782325e-06, + "loss": 1.242, + "step": 308 + }, + { + "epoch": 0.01, + "learning_rate": 5.057283142389527e-06, + "loss": 1.2095, + "step": 309 + }, + { + "epoch": 0.01, + "learning_rate": 5.073649754500819e-06, + "loss": 1.1938, + "step": 310 + }, + { + "epoch": 0.01, + "learning_rate": 5.090016366612112e-06, + "loss": 1.2169, + "step": 311 + }, + { + "epoch": 0.01, + "learning_rate": 5.106382978723404e-06, + "loss": 1.2303, + "step": 312 + }, + { + "epoch": 0.01, + "learning_rate": 5.122749590834698e-06, + "loss": 1.2252, + "step": 313 + }, + { + "epoch": 0.01, + "learning_rate": 5.139116202945991e-06, + "loss": 1.1996, + "step": 314 + }, + { + "epoch": 0.01, + "learning_rate": 5.155482815057283e-06, + "loss": 1.2389, + "step": 315 + }, + { + "epoch": 0.01, + "learning_rate": 5.171849427168577e-06, + "loss": 1.2179, + "step": 316 + }, + { + "epoch": 0.01, + "learning_rate": 5.18821603927987e-06, + "loss": 1.2367, + "step": 317 + }, + { + "epoch": 0.01, + "learning_rate": 5.204582651391162e-06, + "loss": 1.2132, + "step": 318 + }, + { + "epoch": 0.01, + "learning_rate": 5.220949263502455e-06, + "loss": 1.2148, + "step": 319 + }, + { + "epoch": 0.01, + "learning_rate": 5.237315875613749e-06, + "loss": 1.1872, + "step": 320 + }, + { + "epoch": 0.01, + "learning_rate": 5.2536824877250414e-06, + "loss": 1.2392, + "step": 321 + }, + { + "epoch": 0.01, + "learning_rate": 5.270049099836334e-06, + "loss": 1.2366, + "step": 322 + }, + { + "epoch": 0.01, + "learning_rate": 5.286415711947628e-06, + "loss": 1.2032, + "step": 323 + }, + { + "epoch": 0.01, + "learning_rate": 5.302782324058921e-06, + "loss": 1.2245, + "step": 324 + }, + { + "epoch": 0.01, + "learning_rate": 5.319148936170213e-06, + "loss": 1.2119, + "step": 325 + }, + { + "epoch": 0.01, + "learning_rate": 5.3355155482815055e-06, + "loss": 1.2248, + "step": 326 + }, + { + "epoch": 0.01, + "learning_rate": 5.3518821603928e-06, + "loss": 1.2274, + "step": 327 + }, + { + "epoch": 0.01, + "learning_rate": 5.368248772504092e-06, + "loss": 1.2467, + "step": 328 + }, + { + "epoch": 0.01, + "learning_rate": 5.384615384615385e-06, + "loss": 1.2057, + "step": 329 + }, + { + "epoch": 0.01, + "learning_rate": 5.400981996726679e-06, + "loss": 1.2367, + "step": 330 + }, + { + "epoch": 0.01, + "learning_rate": 5.417348608837971e-06, + "loss": 1.1927, + "step": 331 + }, + { + "epoch": 0.01, + "learning_rate": 5.433715220949264e-06, + "loss": 1.2023, + "step": 332 + }, + { + "epoch": 0.01, + "learning_rate": 5.450081833060556e-06, + "loss": 1.1975, + "step": 333 + }, + { + "epoch": 0.01, + "learning_rate": 5.4664484451718504e-06, + "loss": 1.2095, + "step": 334 + }, + { + "epoch": 0.01, + "learning_rate": 5.482815057283143e-06, + "loss": 1.1916, + "step": 335 + }, + { + "epoch": 0.01, + "learning_rate": 5.499181669394435e-06, + "loss": 1.2377, + "step": 336 + }, + { + "epoch": 0.01, + "learning_rate": 5.5155482815057296e-06, + "loss": 1.2127, + "step": 337 + }, + { + "epoch": 0.01, + "learning_rate": 5.531914893617022e-06, + "loss": 1.2321, + "step": 338 + }, + { + "epoch": 0.01, + "learning_rate": 5.5482815057283145e-06, + "loss": 1.1843, + "step": 339 + }, + { + "epoch": 0.01, + "learning_rate": 5.564648117839607e-06, + "loss": 1.2284, + "step": 340 + }, + { + "epoch": 0.01, + "learning_rate": 5.581014729950901e-06, + "loss": 1.2063, + "step": 341 + }, + { + "epoch": 0.01, + "learning_rate": 5.597381342062194e-06, + "loss": 1.2036, + "step": 342 + }, + { + "epoch": 0.01, + "learning_rate": 5.613747954173486e-06, + "loss": 1.2221, + "step": 343 + }, + { + "epoch": 0.01, + "learning_rate": 5.63011456628478e-06, + "loss": 1.1813, + "step": 344 + }, + { + "epoch": 0.01, + "learning_rate": 5.646481178396073e-06, + "loss": 1.1945, + "step": 345 + }, + { + "epoch": 0.01, + "learning_rate": 5.662847790507365e-06, + "loss": 1.2142, + "step": 346 + }, + { + "epoch": 0.01, + "learning_rate": 5.679214402618658e-06, + "loss": 1.1752, + "step": 347 + }, + { + "epoch": 0.01, + "learning_rate": 5.695581014729952e-06, + "loss": 1.2136, + "step": 348 + }, + { + "epoch": 0.01, + "learning_rate": 5.711947626841244e-06, + "loss": 1.2035, + "step": 349 + }, + { + "epoch": 0.01, + "learning_rate": 5.728314238952537e-06, + "loss": 1.2009, + "step": 350 + }, + { + "epoch": 0.01, + "learning_rate": 5.744680851063831e-06, + "loss": 1.2246, + "step": 351 + }, + { + "epoch": 0.01, + "learning_rate": 5.7610474631751235e-06, + "loss": 1.2131, + "step": 352 + }, + { + "epoch": 0.01, + "learning_rate": 5.777414075286416e-06, + "loss": 1.2085, + "step": 353 + }, + { + "epoch": 0.01, + "learning_rate": 5.7937806873977085e-06, + "loss": 1.2238, + "step": 354 + }, + { + "epoch": 0.01, + "learning_rate": 5.810147299509003e-06, + "loss": 1.2347, + "step": 355 + }, + { + "epoch": 0.01, + "learning_rate": 5.826513911620295e-06, + "loss": 1.184, + "step": 356 + }, + { + "epoch": 0.01, + "learning_rate": 5.842880523731588e-06, + "loss": 1.19, + "step": 357 + }, + { + "epoch": 0.01, + "learning_rate": 5.859247135842882e-06, + "loss": 1.2311, + "step": 358 + }, + { + "epoch": 0.01, + "learning_rate": 5.875613747954174e-06, + "loss": 1.1794, + "step": 359 + }, + { + "epoch": 0.01, + "learning_rate": 5.891980360065467e-06, + "loss": 1.1761, + "step": 360 + }, + { + "epoch": 0.01, + "learning_rate": 5.908346972176759e-06, + "loss": 1.1363, + "step": 361 + }, + { + "epoch": 0.01, + "learning_rate": 5.924713584288053e-06, + "loss": 1.1803, + "step": 362 + }, + { + "epoch": 0.01, + "learning_rate": 5.941080196399346e-06, + "loss": 1.2271, + "step": 363 + }, + { + "epoch": 0.01, + "learning_rate": 5.957446808510638e-06, + "loss": 1.2356, + "step": 364 + }, + { + "epoch": 0.01, + "learning_rate": 5.9738134206219325e-06, + "loss": 1.1826, + "step": 365 + }, + { + "epoch": 0.01, + "learning_rate": 5.990180032733225e-06, + "loss": 1.1911, + "step": 366 + }, + { + "epoch": 0.01, + "learning_rate": 6.0065466448445175e-06, + "loss": 1.2142, + "step": 367 + }, + { + "epoch": 0.01, + "learning_rate": 6.02291325695581e-06, + "loss": 1.2149, + "step": 368 + }, + { + "epoch": 0.01, + "learning_rate": 6.039279869067104e-06, + "loss": 1.1963, + "step": 369 + }, + { + "epoch": 0.01, + "learning_rate": 6.055646481178397e-06, + "loss": 1.1825, + "step": 370 + }, + { + "epoch": 0.01, + "learning_rate": 6.072013093289689e-06, + "loss": 1.2125, + "step": 371 + }, + { + "epoch": 0.01, + "learning_rate": 6.088379705400983e-06, + "loss": 1.2588, + "step": 372 + }, + { + "epoch": 0.01, + "learning_rate": 6.104746317512276e-06, + "loss": 1.2344, + "step": 373 + }, + { + "epoch": 0.01, + "learning_rate": 6.121112929623568e-06, + "loss": 1.2028, + "step": 374 + }, + { + "epoch": 0.01, + "learning_rate": 6.137479541734861e-06, + "loss": 1.2126, + "step": 375 + }, + { + "epoch": 0.01, + "learning_rate": 6.153846153846155e-06, + "loss": 1.2017, + "step": 376 + }, + { + "epoch": 0.01, + "learning_rate": 6.170212765957447e-06, + "loss": 1.1886, + "step": 377 + }, + { + "epoch": 0.01, + "learning_rate": 6.18657937806874e-06, + "loss": 1.2051, + "step": 378 + }, + { + "epoch": 0.01, + "learning_rate": 6.202945990180034e-06, + "loss": 1.1944, + "step": 379 + }, + { + "epoch": 0.01, + "learning_rate": 6.2193126022913265e-06, + "loss": 1.1663, + "step": 380 + }, + { + "epoch": 0.01, + "learning_rate": 6.235679214402619e-06, + "loss": 1.2284, + "step": 381 + }, + { + "epoch": 0.01, + "learning_rate": 6.2520458265139114e-06, + "loss": 1.2043, + "step": 382 + }, + { + "epoch": 0.01, + "learning_rate": 6.268412438625206e-06, + "loss": 1.2502, + "step": 383 + }, + { + "epoch": 0.01, + "learning_rate": 6.284779050736498e-06, + "loss": 1.2448, + "step": 384 + }, + { + "epoch": 0.01, + "learning_rate": 6.3011456628477906e-06, + "loss": 1.1841, + "step": 385 + }, + { + "epoch": 0.01, + "learning_rate": 6.317512274959085e-06, + "loss": 1.2179, + "step": 386 + }, + { + "epoch": 0.01, + "learning_rate": 6.333878887070377e-06, + "loss": 1.1882, + "step": 387 + }, + { + "epoch": 0.01, + "learning_rate": 6.35024549918167e-06, + "loss": 1.2136, + "step": 388 + }, + { + "epoch": 0.01, + "learning_rate": 6.366612111292962e-06, + "loss": 1.2073, + "step": 389 + }, + { + "epoch": 0.01, + "learning_rate": 6.382978723404256e-06, + "loss": 1.1697, + "step": 390 + }, + { + "epoch": 0.01, + "learning_rate": 6.399345335515549e-06, + "loss": 1.2311, + "step": 391 + }, + { + "epoch": 0.01, + "learning_rate": 6.415711947626841e-06, + "loss": 1.1959, + "step": 392 + }, + { + "epoch": 0.01, + "learning_rate": 6.4320785597381355e-06, + "loss": 1.2096, + "step": 393 + }, + { + "epoch": 0.01, + "learning_rate": 6.448445171849428e-06, + "loss": 1.2198, + "step": 394 + }, + { + "epoch": 0.01, + "learning_rate": 6.4648117839607204e-06, + "loss": 1.1922, + "step": 395 + }, + { + "epoch": 0.01, + "learning_rate": 6.481178396072013e-06, + "loss": 1.173, + "step": 396 + }, + { + "epoch": 0.01, + "learning_rate": 6.497545008183307e-06, + "loss": 1.2025, + "step": 397 + }, + { + "epoch": 0.01, + "learning_rate": 6.5139116202945996e-06, + "loss": 1.1927, + "step": 398 + }, + { + "epoch": 0.01, + "learning_rate": 6.530278232405892e-06, + "loss": 1.1836, + "step": 399 + }, + { + "epoch": 0.01, + "learning_rate": 6.546644844517186e-06, + "loss": 1.2066, + "step": 400 + }, + { + "epoch": 0.01, + "learning_rate": 6.563011456628479e-06, + "loss": 1.1885, + "step": 401 + }, + { + "epoch": 0.01, + "learning_rate": 6.579378068739771e-06, + "loss": 1.2319, + "step": 402 + }, + { + "epoch": 0.01, + "learning_rate": 6.595744680851064e-06, + "loss": 1.1678, + "step": 403 + }, + { + "epoch": 0.01, + "learning_rate": 6.612111292962358e-06, + "loss": 1.187, + "step": 404 + }, + { + "epoch": 0.01, + "learning_rate": 6.62847790507365e-06, + "loss": 1.1907, + "step": 405 + }, + { + "epoch": 0.01, + "learning_rate": 6.644844517184943e-06, + "loss": 1.2024, + "step": 406 + }, + { + "epoch": 0.01, + "learning_rate": 6.661211129296237e-06, + "loss": 1.1929, + "step": 407 + }, + { + "epoch": 0.01, + "learning_rate": 6.6775777414075294e-06, + "loss": 1.1957, + "step": 408 + }, + { + "epoch": 0.01, + "learning_rate": 6.693944353518822e-06, + "loss": 1.1766, + "step": 409 + }, + { + "epoch": 0.01, + "learning_rate": 6.710310965630114e-06, + "loss": 1.1757, + "step": 410 + }, + { + "epoch": 0.01, + "learning_rate": 6.7266775777414086e-06, + "loss": 1.1902, + "step": 411 + }, + { + "epoch": 0.01, + "learning_rate": 6.743044189852701e-06, + "loss": 1.1843, + "step": 412 + }, + { + "epoch": 0.01, + "learning_rate": 6.7594108019639935e-06, + "loss": 1.1898, + "step": 413 + }, + { + "epoch": 0.01, + "learning_rate": 6.775777414075288e-06, + "loss": 1.1808, + "step": 414 + }, + { + "epoch": 0.01, + "learning_rate": 6.79214402618658e-06, + "loss": 1.1744, + "step": 415 + }, + { + "epoch": 0.01, + "learning_rate": 6.808510638297873e-06, + "loss": 1.1823, + "step": 416 + }, + { + "epoch": 0.01, + "learning_rate": 6.824877250409165e-06, + "loss": 1.202, + "step": 417 + }, + { + "epoch": 0.01, + "learning_rate": 6.841243862520459e-06, + "loss": 1.2027, + "step": 418 + }, + { + "epoch": 0.01, + "learning_rate": 6.857610474631752e-06, + "loss": 1.249, + "step": 419 + }, + { + "epoch": 0.01, + "learning_rate": 6.873977086743044e-06, + "loss": 1.2052, + "step": 420 + }, + { + "epoch": 0.01, + "learning_rate": 6.8903436988543384e-06, + "loss": 1.2075, + "step": 421 + }, + { + "epoch": 0.01, + "learning_rate": 6.906710310965631e-06, + "loss": 1.1914, + "step": 422 + }, + { + "epoch": 0.01, + "learning_rate": 6.923076923076923e-06, + "loss": 1.1941, + "step": 423 + }, + { + "epoch": 0.01, + "learning_rate": 6.939443535188216e-06, + "loss": 1.2071, + "step": 424 + }, + { + "epoch": 0.01, + "learning_rate": 6.95581014729951e-06, + "loss": 1.2127, + "step": 425 + }, + { + "epoch": 0.01, + "learning_rate": 6.9721767594108025e-06, + "loss": 1.1951, + "step": 426 + }, + { + "epoch": 0.01, + "learning_rate": 6.988543371522095e-06, + "loss": 1.1818, + "step": 427 + }, + { + "epoch": 0.01, + "learning_rate": 7.004909983633389e-06, + "loss": 1.1865, + "step": 428 + }, + { + "epoch": 0.01, + "learning_rate": 7.021276595744682e-06, + "loss": 1.214, + "step": 429 + }, + { + "epoch": 0.01, + "learning_rate": 7.037643207855974e-06, + "loss": 1.2052, + "step": 430 + }, + { + "epoch": 0.01, + "learning_rate": 7.054009819967267e-06, + "loss": 1.1594, + "step": 431 + }, + { + "epoch": 0.01, + "learning_rate": 7.070376432078561e-06, + "loss": 1.1944, + "step": 432 + }, + { + "epoch": 0.01, + "learning_rate": 7.086743044189853e-06, + "loss": 1.1743, + "step": 433 + }, + { + "epoch": 0.01, + "learning_rate": 7.103109656301146e-06, + "loss": 1.1488, + "step": 434 + }, + { + "epoch": 0.01, + "learning_rate": 7.11947626841244e-06, + "loss": 1.2229, + "step": 435 + }, + { + "epoch": 0.01, + "learning_rate": 7.135842880523732e-06, + "loss": 1.1924, + "step": 436 + }, + { + "epoch": 0.01, + "learning_rate": 7.152209492635025e-06, + "loss": 1.2102, + "step": 437 + }, + { + "epoch": 0.01, + "learning_rate": 7.168576104746317e-06, + "loss": 1.187, + "step": 438 + }, + { + "epoch": 0.01, + "learning_rate": 7.1849427168576115e-06, + "loss": 1.2083, + "step": 439 + }, + { + "epoch": 0.01, + "learning_rate": 7.201309328968904e-06, + "loss": 1.2014, + "step": 440 + }, + { + "epoch": 0.01, + "learning_rate": 7.2176759410801965e-06, + "loss": 1.1844, + "step": 441 + }, + { + "epoch": 0.01, + "learning_rate": 7.234042553191491e-06, + "loss": 1.1985, + "step": 442 + }, + { + "epoch": 0.01, + "learning_rate": 7.250409165302783e-06, + "loss": 1.182, + "step": 443 + }, + { + "epoch": 0.01, + "learning_rate": 7.266775777414076e-06, + "loss": 1.1914, + "step": 444 + }, + { + "epoch": 0.01, + "learning_rate": 7.283142389525368e-06, + "loss": 1.2142, + "step": 445 + }, + { + "epoch": 0.01, + "learning_rate": 7.299509001636662e-06, + "loss": 1.1762, + "step": 446 + }, + { + "epoch": 0.01, + "learning_rate": 7.315875613747955e-06, + "loss": 1.2061, + "step": 447 + }, + { + "epoch": 0.01, + "learning_rate": 7.332242225859247e-06, + "loss": 1.1436, + "step": 448 + }, + { + "epoch": 0.01, + "learning_rate": 7.348608837970541e-06, + "loss": 1.1569, + "step": 449 + }, + { + "epoch": 0.01, + "learning_rate": 7.364975450081834e-06, + "loss": 1.1668, + "step": 450 + }, + { + "epoch": 0.01, + "learning_rate": 7.381342062193126e-06, + "loss": 1.223, + "step": 451 + }, + { + "epoch": 0.01, + "learning_rate": 7.397708674304419e-06, + "loss": 1.1881, + "step": 452 + }, + { + "epoch": 0.01, + "learning_rate": 7.414075286415713e-06, + "loss": 1.2325, + "step": 453 + }, + { + "epoch": 0.01, + "learning_rate": 7.4304418985270055e-06, + "loss": 1.2095, + "step": 454 + }, + { + "epoch": 0.01, + "learning_rate": 7.446808510638298e-06, + "loss": 1.1738, + "step": 455 + }, + { + "epoch": 0.01, + "learning_rate": 7.463175122749592e-06, + "loss": 1.1627, + "step": 456 + }, + { + "epoch": 0.01, + "learning_rate": 7.479541734860885e-06, + "loss": 1.2124, + "step": 457 + }, + { + "epoch": 0.01, + "learning_rate": 7.495908346972177e-06, + "loss": 1.219, + "step": 458 + }, + { + "epoch": 0.01, + "learning_rate": 7.5122749590834695e-06, + "loss": 1.1965, + "step": 459 + }, + { + "epoch": 0.01, + "learning_rate": 7.528641571194764e-06, + "loss": 1.1511, + "step": 460 + }, + { + "epoch": 0.01, + "learning_rate": 7.545008183306056e-06, + "loss": 1.2258, + "step": 461 + }, + { + "epoch": 0.01, + "learning_rate": 7.561374795417349e-06, + "loss": 1.1791, + "step": 462 + }, + { + "epoch": 0.01, + "learning_rate": 7.577741407528643e-06, + "loss": 1.1737, + "step": 463 + }, + { + "epoch": 0.01, + "learning_rate": 7.594108019639935e-06, + "loss": 1.2009, + "step": 464 + }, + { + "epoch": 0.01, + "learning_rate": 7.610474631751228e-06, + "loss": 1.1572, + "step": 465 + }, + { + "epoch": 0.01, + "learning_rate": 7.62684124386252e-06, + "loss": 1.1832, + "step": 466 + }, + { + "epoch": 0.01, + "learning_rate": 7.643207855973814e-06, + "loss": 1.2242, + "step": 467 + }, + { + "epoch": 0.01, + "learning_rate": 7.659574468085107e-06, + "loss": 1.1889, + "step": 468 + }, + { + "epoch": 0.01, + "learning_rate": 7.6759410801964e-06, + "loss": 1.1873, + "step": 469 + }, + { + "epoch": 0.01, + "learning_rate": 7.692307692307694e-06, + "loss": 1.2134, + "step": 470 + }, + { + "epoch": 0.01, + "learning_rate": 7.708674304418986e-06, + "loss": 1.1708, + "step": 471 + }, + { + "epoch": 0.01, + "learning_rate": 7.725040916530279e-06, + "loss": 1.184, + "step": 472 + }, + { + "epoch": 0.01, + "learning_rate": 7.741407528641571e-06, + "loss": 1.1998, + "step": 473 + }, + { + "epoch": 0.01, + "learning_rate": 7.757774140752865e-06, + "loss": 1.2299, + "step": 474 + }, + { + "epoch": 0.01, + "learning_rate": 7.774140752864158e-06, + "loss": 1.18, + "step": 475 + }, + { + "epoch": 0.01, + "learning_rate": 7.79050736497545e-06, + "loss": 1.1932, + "step": 476 + }, + { + "epoch": 0.01, + "learning_rate": 7.806873977086744e-06, + "loss": 1.1607, + "step": 477 + }, + { + "epoch": 0.01, + "learning_rate": 7.823240589198037e-06, + "loss": 1.1674, + "step": 478 + }, + { + "epoch": 0.01, + "learning_rate": 7.83960720130933e-06, + "loss": 1.1827, + "step": 479 + }, + { + "epoch": 0.01, + "learning_rate": 7.855973813420622e-06, + "loss": 1.172, + "step": 480 + }, + { + "epoch": 0.01, + "learning_rate": 7.872340425531916e-06, + "loss": 1.1841, + "step": 481 + }, + { + "epoch": 0.01, + "learning_rate": 7.888707037643208e-06, + "loss": 1.1702, + "step": 482 + }, + { + "epoch": 0.01, + "learning_rate": 7.905073649754501e-06, + "loss": 1.1751, + "step": 483 + }, + { + "epoch": 0.01, + "learning_rate": 7.921440261865795e-06, + "loss": 1.2247, + "step": 484 + }, + { + "epoch": 0.01, + "learning_rate": 7.937806873977088e-06, + "loss": 1.1868, + "step": 485 + }, + { + "epoch": 0.01, + "learning_rate": 7.95417348608838e-06, + "loss": 1.2012, + "step": 486 + }, + { + "epoch": 0.01, + "learning_rate": 7.970540098199672e-06, + "loss": 1.17, + "step": 487 + }, + { + "epoch": 0.01, + "learning_rate": 7.986906710310967e-06, + "loss": 1.1988, + "step": 488 + }, + { + "epoch": 0.01, + "learning_rate": 8.003273322422259e-06, + "loss": 1.1928, + "step": 489 + }, + { + "epoch": 0.01, + "learning_rate": 8.019639934533552e-06, + "loss": 1.1569, + "step": 490 + }, + { + "epoch": 0.01, + "learning_rate": 8.036006546644846e-06, + "loss": 1.1898, + "step": 491 + }, + { + "epoch": 0.01, + "learning_rate": 8.052373158756138e-06, + "loss": 1.1729, + "step": 492 + }, + { + "epoch": 0.01, + "learning_rate": 8.06873977086743e-06, + "loss": 1.1781, + "step": 493 + }, + { + "epoch": 0.01, + "learning_rate": 8.085106382978723e-06, + "loss": 1.1804, + "step": 494 + }, + { + "epoch": 0.01, + "learning_rate": 8.101472995090017e-06, + "loss": 1.1642, + "step": 495 + }, + { + "epoch": 0.01, + "learning_rate": 8.11783960720131e-06, + "loss": 1.141, + "step": 496 + }, + { + "epoch": 0.01, + "learning_rate": 8.134206219312602e-06, + "loss": 1.1767, + "step": 497 + }, + { + "epoch": 0.01, + "learning_rate": 8.150572831423897e-06, + "loss": 1.1701, + "step": 498 + }, + { + "epoch": 0.01, + "learning_rate": 8.166939443535189e-06, + "loss": 1.1499, + "step": 499 + }, + { + "epoch": 0.01, + "learning_rate": 8.183306055646481e-06, + "loss": 1.1968, + "step": 500 + }, + { + "epoch": 0.01, + "learning_rate": 8.199672667757774e-06, + "loss": 1.1625, + "step": 501 + }, + { + "epoch": 0.01, + "learning_rate": 8.216039279869068e-06, + "loss": 1.1983, + "step": 502 + }, + { + "epoch": 0.01, + "learning_rate": 8.23240589198036e-06, + "loss": 1.2316, + "step": 503 + }, + { + "epoch": 0.01, + "learning_rate": 8.248772504091653e-06, + "loss": 1.2063, + "step": 504 + }, + { + "epoch": 0.01, + "learning_rate": 8.265139116202947e-06, + "loss": 1.2117, + "step": 505 + }, + { + "epoch": 0.01, + "learning_rate": 8.28150572831424e-06, + "loss": 1.2381, + "step": 506 + }, + { + "epoch": 0.01, + "learning_rate": 8.297872340425532e-06, + "loss": 1.176, + "step": 507 + }, + { + "epoch": 0.01, + "learning_rate": 8.314238952536825e-06, + "loss": 1.1878, + "step": 508 + }, + { + "epoch": 0.01, + "learning_rate": 8.330605564648119e-06, + "loss": 1.1927, + "step": 509 + }, + { + "epoch": 0.01, + "learning_rate": 8.346972176759411e-06, + "loss": 1.2092, + "step": 510 + }, + { + "epoch": 0.01, + "learning_rate": 8.363338788870704e-06, + "loss": 1.1965, + "step": 511 + }, + { + "epoch": 0.01, + "learning_rate": 8.379705400981998e-06, + "loss": 1.194, + "step": 512 + }, + { + "epoch": 0.01, + "learning_rate": 8.39607201309329e-06, + "loss": 1.1871, + "step": 513 + }, + { + "epoch": 0.01, + "learning_rate": 8.412438625204583e-06, + "loss": 1.1848, + "step": 514 + }, + { + "epoch": 0.01, + "learning_rate": 8.428805237315877e-06, + "loss": 1.1537, + "step": 515 + }, + { + "epoch": 0.01, + "learning_rate": 8.44517184942717e-06, + "loss": 1.1882, + "step": 516 + }, + { + "epoch": 0.01, + "learning_rate": 8.461538461538462e-06, + "loss": 1.1676, + "step": 517 + }, + { + "epoch": 0.01, + "learning_rate": 8.477905073649755e-06, + "loss": 1.1868, + "step": 518 + }, + { + "epoch": 0.01, + "learning_rate": 8.494271685761049e-06, + "loss": 1.1738, + "step": 519 + }, + { + "epoch": 0.01, + "learning_rate": 8.510638297872341e-06, + "loss": 1.1531, + "step": 520 + }, + { + "epoch": 0.01, + "learning_rate": 8.527004909983634e-06, + "loss": 1.1891, + "step": 521 + }, + { + "epoch": 0.01, + "learning_rate": 8.543371522094928e-06, + "loss": 1.1632, + "step": 522 + }, + { + "epoch": 0.01, + "learning_rate": 8.55973813420622e-06, + "loss": 1.198, + "step": 523 + }, + { + "epoch": 0.01, + "learning_rate": 8.576104746317513e-06, + "loss": 1.1946, + "step": 524 + }, + { + "epoch": 0.01, + "learning_rate": 8.592471358428805e-06, + "loss": 1.1609, + "step": 525 + }, + { + "epoch": 0.01, + "learning_rate": 8.6088379705401e-06, + "loss": 1.2022, + "step": 526 + }, + { + "epoch": 0.01, + "learning_rate": 8.625204582651392e-06, + "loss": 1.1848, + "step": 527 + }, + { + "epoch": 0.01, + "learning_rate": 8.641571194762684e-06, + "loss": 1.1681, + "step": 528 + }, + { + "epoch": 0.01, + "learning_rate": 8.657937806873979e-06, + "loss": 1.1609, + "step": 529 + }, + { + "epoch": 0.01, + "learning_rate": 8.674304418985271e-06, + "loss": 1.1862, + "step": 530 + }, + { + "epoch": 0.01, + "learning_rate": 8.690671031096564e-06, + "loss": 1.1723, + "step": 531 + }, + { + "epoch": 0.01, + "learning_rate": 8.707037643207856e-06, + "loss": 1.2129, + "step": 532 + }, + { + "epoch": 0.01, + "learning_rate": 8.72340425531915e-06, + "loss": 1.148, + "step": 533 + }, + { + "epoch": 0.01, + "learning_rate": 8.739770867430443e-06, + "loss": 1.1964, + "step": 534 + }, + { + "epoch": 0.01, + "learning_rate": 8.756137479541735e-06, + "loss": 1.1712, + "step": 535 + }, + { + "epoch": 0.01, + "learning_rate": 8.77250409165303e-06, + "loss": 1.1686, + "step": 536 + }, + { + "epoch": 0.01, + "learning_rate": 8.788870703764322e-06, + "loss": 1.1925, + "step": 537 + }, + { + "epoch": 0.01, + "learning_rate": 8.805237315875614e-06, + "loss": 1.1698, + "step": 538 + }, + { + "epoch": 0.01, + "learning_rate": 8.821603927986907e-06, + "loss": 1.2068, + "step": 539 + }, + { + "epoch": 0.01, + "learning_rate": 8.837970540098201e-06, + "loss": 1.2105, + "step": 540 + }, + { + "epoch": 0.01, + "learning_rate": 8.854337152209493e-06, + "loss": 1.1782, + "step": 541 + }, + { + "epoch": 0.01, + "learning_rate": 8.870703764320786e-06, + "loss": 1.2036, + "step": 542 + }, + { + "epoch": 0.01, + "learning_rate": 8.88707037643208e-06, + "loss": 1.1671, + "step": 543 + }, + { + "epoch": 0.01, + "learning_rate": 8.903436988543373e-06, + "loss": 1.1821, + "step": 544 + }, + { + "epoch": 0.01, + "learning_rate": 8.919803600654665e-06, + "loss": 1.224, + "step": 545 + }, + { + "epoch": 0.01, + "learning_rate": 8.936170212765958e-06, + "loss": 1.1423, + "step": 546 + }, + { + "epoch": 0.01, + "learning_rate": 8.952536824877252e-06, + "loss": 1.1921, + "step": 547 + }, + { + "epoch": 0.01, + "learning_rate": 8.968903436988544e-06, + "loss": 1.2114, + "step": 548 + }, + { + "epoch": 0.01, + "learning_rate": 8.985270049099837e-06, + "loss": 1.1749, + "step": 549 + }, + { + "epoch": 0.01, + "learning_rate": 9.00163666121113e-06, + "loss": 1.1764, + "step": 550 + }, + { + "epoch": 0.01, + "learning_rate": 9.018003273322423e-06, + "loss": 1.1845, + "step": 551 + }, + { + "epoch": 0.01, + "learning_rate": 9.034369885433716e-06, + "loss": 1.1665, + "step": 552 + }, + { + "epoch": 0.01, + "learning_rate": 9.050736497545008e-06, + "loss": 1.1846, + "step": 553 + }, + { + "epoch": 0.01, + "learning_rate": 9.067103109656302e-06, + "loss": 1.1827, + "step": 554 + }, + { + "epoch": 0.01, + "learning_rate": 9.083469721767595e-06, + "loss": 1.172, + "step": 555 + }, + { + "epoch": 0.01, + "learning_rate": 9.099836333878887e-06, + "loss": 1.175, + "step": 556 + }, + { + "epoch": 0.01, + "learning_rate": 9.116202945990182e-06, + "loss": 1.1575, + "step": 557 + }, + { + "epoch": 0.01, + "learning_rate": 9.132569558101474e-06, + "loss": 1.1282, + "step": 558 + }, + { + "epoch": 0.01, + "learning_rate": 9.148936170212767e-06, + "loss": 1.1823, + "step": 559 + }, + { + "epoch": 0.01, + "learning_rate": 9.165302782324059e-06, + "loss": 1.1462, + "step": 560 + }, + { + "epoch": 0.01, + "learning_rate": 9.181669394435353e-06, + "loss": 1.2033, + "step": 561 + }, + { + "epoch": 0.01, + "learning_rate": 9.198036006546646e-06, + "loss": 1.1797, + "step": 562 + }, + { + "epoch": 0.01, + "learning_rate": 9.214402618657938e-06, + "loss": 1.1582, + "step": 563 + }, + { + "epoch": 0.01, + "learning_rate": 9.230769230769232e-06, + "loss": 1.2073, + "step": 564 + }, + { + "epoch": 0.01, + "learning_rate": 9.247135842880525e-06, + "loss": 1.2082, + "step": 565 + }, + { + "epoch": 0.01, + "learning_rate": 9.263502454991817e-06, + "loss": 1.1468, + "step": 566 + }, + { + "epoch": 0.01, + "learning_rate": 9.27986906710311e-06, + "loss": 1.1612, + "step": 567 + }, + { + "epoch": 0.01, + "learning_rate": 9.296235679214404e-06, + "loss": 1.182, + "step": 568 + }, + { + "epoch": 0.01, + "learning_rate": 9.312602291325696e-06, + "loss": 1.1489, + "step": 569 + }, + { + "epoch": 0.01, + "learning_rate": 9.328968903436989e-06, + "loss": 1.1578, + "step": 570 + }, + { + "epoch": 0.01, + "learning_rate": 9.345335515548283e-06, + "loss": 1.1556, + "step": 571 + }, + { + "epoch": 0.01, + "learning_rate": 9.361702127659576e-06, + "loss": 1.1988, + "step": 572 + }, + { + "epoch": 0.01, + "learning_rate": 9.378068739770868e-06, + "loss": 1.1651, + "step": 573 + }, + { + "epoch": 0.01, + "learning_rate": 9.39443535188216e-06, + "loss": 1.1504, + "step": 574 + }, + { + "epoch": 0.01, + "learning_rate": 9.410801963993455e-06, + "loss": 1.1495, + "step": 575 + }, + { + "epoch": 0.01, + "learning_rate": 9.427168576104747e-06, + "loss": 1.1631, + "step": 576 + }, + { + "epoch": 0.01, + "learning_rate": 9.44353518821604e-06, + "loss": 1.1697, + "step": 577 + }, + { + "epoch": 0.01, + "learning_rate": 9.459901800327334e-06, + "loss": 1.1829, + "step": 578 + }, + { + "epoch": 0.01, + "learning_rate": 9.476268412438626e-06, + "loss": 1.1738, + "step": 579 + }, + { + "epoch": 0.01, + "learning_rate": 9.492635024549919e-06, + "loss": 1.1631, + "step": 580 + }, + { + "epoch": 0.01, + "learning_rate": 9.509001636661211e-06, + "loss": 1.1909, + "step": 581 + }, + { + "epoch": 0.01, + "learning_rate": 9.525368248772505e-06, + "loss": 1.1937, + "step": 582 + }, + { + "epoch": 0.01, + "learning_rate": 9.541734860883798e-06, + "loss": 1.1929, + "step": 583 + }, + { + "epoch": 0.01, + "learning_rate": 9.55810147299509e-06, + "loss": 1.1794, + "step": 584 + }, + { + "epoch": 0.01, + "learning_rate": 9.574468085106385e-06, + "loss": 1.1976, + "step": 585 + }, + { + "epoch": 0.01, + "learning_rate": 9.590834697217677e-06, + "loss": 1.1762, + "step": 586 + }, + { + "epoch": 0.01, + "learning_rate": 9.60720130932897e-06, + "loss": 1.1391, + "step": 587 + }, + { + "epoch": 0.01, + "learning_rate": 9.623567921440262e-06, + "loss": 1.1398, + "step": 588 + }, + { + "epoch": 0.01, + "learning_rate": 9.639934533551556e-06, + "loss": 1.163, + "step": 589 + }, + { + "epoch": 0.01, + "learning_rate": 9.656301145662849e-06, + "loss": 1.2079, + "step": 590 + }, + { + "epoch": 0.01, + "learning_rate": 9.672667757774141e-06, + "loss": 1.1557, + "step": 591 + }, + { + "epoch": 0.01, + "learning_rate": 9.689034369885435e-06, + "loss": 1.1421, + "step": 592 + }, + { + "epoch": 0.01, + "learning_rate": 9.705400981996728e-06, + "loss": 1.1919, + "step": 593 + }, + { + "epoch": 0.01, + "learning_rate": 9.72176759410802e-06, + "loss": 1.173, + "step": 594 + }, + { + "epoch": 0.01, + "learning_rate": 9.738134206219313e-06, + "loss": 1.1534, + "step": 595 + }, + { + "epoch": 0.01, + "learning_rate": 9.754500818330607e-06, + "loss": 1.1691, + "step": 596 + }, + { + "epoch": 0.01, + "learning_rate": 9.7708674304419e-06, + "loss": 1.1789, + "step": 597 + }, + { + "epoch": 0.01, + "learning_rate": 9.787234042553192e-06, + "loss": 1.1489, + "step": 598 + }, + { + "epoch": 0.01, + "learning_rate": 9.803600654664486e-06, + "loss": 1.1801, + "step": 599 + }, + { + "epoch": 0.01, + "learning_rate": 9.819967266775778e-06, + "loss": 1.1662, + "step": 600 + }, + { + "epoch": 0.01, + "learning_rate": 9.836333878887071e-06, + "loss": 1.2165, + "step": 601 + }, + { + "epoch": 0.01, + "learning_rate": 9.852700490998363e-06, + "loss": 1.1645, + "step": 602 + }, + { + "epoch": 0.01, + "learning_rate": 9.869067103109658e-06, + "loss": 1.1785, + "step": 603 + }, + { + "epoch": 0.01, + "learning_rate": 9.88543371522095e-06, + "loss": 1.1719, + "step": 604 + }, + { + "epoch": 0.01, + "learning_rate": 9.901800327332243e-06, + "loss": 1.1807, + "step": 605 + }, + { + "epoch": 0.01, + "learning_rate": 9.918166939443537e-06, + "loss": 1.1602, + "step": 606 + }, + { + "epoch": 0.01, + "learning_rate": 9.93453355155483e-06, + "loss": 1.174, + "step": 607 + }, + { + "epoch": 0.01, + "learning_rate": 9.950900163666122e-06, + "loss": 1.1739, + "step": 608 + }, + { + "epoch": 0.01, + "learning_rate": 9.967266775777414e-06, + "loss": 1.1606, + "step": 609 + }, + { + "epoch": 0.01, + "learning_rate": 9.983633387888708e-06, + "loss": 1.1917, + "step": 610 + }, + { + "epoch": 0.02, + "learning_rate": 1e-05, + "loss": 1.1481, + "step": 611 + }, + { + "epoch": 0.02, + "learning_rate": 1.0016366612111295e-05, + "loss": 1.1691, + "step": 612 + }, + { + "epoch": 0.02, + "learning_rate": 1.0032733224222586e-05, + "loss": 1.1714, + "step": 613 + }, + { + "epoch": 0.02, + "learning_rate": 1.004909983633388e-05, + "loss": 1.1796, + "step": 614 + }, + { + "epoch": 0.02, + "learning_rate": 1.0065466448445174e-05, + "loss": 1.1804, + "step": 615 + }, + { + "epoch": 0.02, + "learning_rate": 1.0081833060556465e-05, + "loss": 1.1609, + "step": 616 + }, + { + "epoch": 0.02, + "learning_rate": 1.0098199672667759e-05, + "loss": 1.1612, + "step": 617 + }, + { + "epoch": 0.02, + "learning_rate": 1.0114566284779053e-05, + "loss": 1.1592, + "step": 618 + }, + { + "epoch": 0.02, + "learning_rate": 1.0130932896890344e-05, + "loss": 1.1416, + "step": 619 + }, + { + "epoch": 0.02, + "learning_rate": 1.0147299509001638e-05, + "loss": 1.1878, + "step": 620 + }, + { + "epoch": 0.02, + "learning_rate": 1.0163666121112929e-05, + "loss": 1.1896, + "step": 621 + }, + { + "epoch": 0.02, + "learning_rate": 1.0180032733224223e-05, + "loss": 1.1706, + "step": 622 + }, + { + "epoch": 0.02, + "learning_rate": 1.0196399345335517e-05, + "loss": 1.163, + "step": 623 + }, + { + "epoch": 0.02, + "learning_rate": 1.0212765957446808e-05, + "loss": 1.1792, + "step": 624 + }, + { + "epoch": 0.02, + "learning_rate": 1.0229132569558102e-05, + "loss": 1.1569, + "step": 625 + }, + { + "epoch": 0.02, + "learning_rate": 1.0245499181669396e-05, + "loss": 1.1584, + "step": 626 + }, + { + "epoch": 0.02, + "learning_rate": 1.0261865793780687e-05, + "loss": 1.159, + "step": 627 + }, + { + "epoch": 0.02, + "learning_rate": 1.0278232405891981e-05, + "loss": 1.1725, + "step": 628 + }, + { + "epoch": 0.02, + "learning_rate": 1.0294599018003276e-05, + "loss": 1.1939, + "step": 629 + }, + { + "epoch": 0.02, + "learning_rate": 1.0310965630114566e-05, + "loss": 1.1487, + "step": 630 + }, + { + "epoch": 0.02, + "learning_rate": 1.032733224222586e-05, + "loss": 1.1194, + "step": 631 + }, + { + "epoch": 0.02, + "learning_rate": 1.0343698854337155e-05, + "loss": 1.1585, + "step": 632 + }, + { + "epoch": 0.02, + "learning_rate": 1.0360065466448446e-05, + "loss": 1.1356, + "step": 633 + }, + { + "epoch": 0.02, + "learning_rate": 1.037643207855974e-05, + "loss": 1.1838, + "step": 634 + }, + { + "epoch": 0.02, + "learning_rate": 1.039279869067103e-05, + "loss": 1.1606, + "step": 635 + }, + { + "epoch": 0.02, + "learning_rate": 1.0409165302782325e-05, + "loss": 1.1865, + "step": 636 + }, + { + "epoch": 0.02, + "learning_rate": 1.0425531914893619e-05, + "loss": 1.1641, + "step": 637 + }, + { + "epoch": 0.02, + "learning_rate": 1.044189852700491e-05, + "loss": 1.1709, + "step": 638 + }, + { + "epoch": 0.02, + "learning_rate": 1.0458265139116204e-05, + "loss": 1.1359, + "step": 639 + }, + { + "epoch": 0.02, + "learning_rate": 1.0474631751227498e-05, + "loss": 1.1538, + "step": 640 + }, + { + "epoch": 0.02, + "learning_rate": 1.0490998363338789e-05, + "loss": 1.1878, + "step": 641 + }, + { + "epoch": 0.02, + "learning_rate": 1.0507364975450083e-05, + "loss": 1.184, + "step": 642 + }, + { + "epoch": 0.02, + "learning_rate": 1.0523731587561377e-05, + "loss": 1.1873, + "step": 643 + }, + { + "epoch": 0.02, + "learning_rate": 1.0540098199672668e-05, + "loss": 1.1503, + "step": 644 + }, + { + "epoch": 0.02, + "learning_rate": 1.0556464811783962e-05, + "loss": 1.1657, + "step": 645 + }, + { + "epoch": 0.02, + "learning_rate": 1.0572831423895256e-05, + "loss": 1.2033, + "step": 646 + }, + { + "epoch": 0.02, + "learning_rate": 1.0589198036006547e-05, + "loss": 1.1679, + "step": 647 + }, + { + "epoch": 0.02, + "learning_rate": 1.0605564648117841e-05, + "loss": 1.1424, + "step": 648 + }, + { + "epoch": 0.02, + "learning_rate": 1.0621931260229132e-05, + "loss": 1.1506, + "step": 649 + }, + { + "epoch": 0.02, + "learning_rate": 1.0638297872340426e-05, + "loss": 1.2015, + "step": 650 + }, + { + "epoch": 0.02, + "learning_rate": 1.065466448445172e-05, + "loss": 1.1562, + "step": 651 + }, + { + "epoch": 0.02, + "learning_rate": 1.0671031096563011e-05, + "loss": 1.1314, + "step": 652 + }, + { + "epoch": 0.02, + "learning_rate": 1.0687397708674305e-05, + "loss": 1.1444, + "step": 653 + }, + { + "epoch": 0.02, + "learning_rate": 1.07037643207856e-05, + "loss": 1.1769, + "step": 654 + }, + { + "epoch": 0.02, + "learning_rate": 1.072013093289689e-05, + "loss": 1.1877, + "step": 655 + }, + { + "epoch": 0.02, + "learning_rate": 1.0736497545008184e-05, + "loss": 1.1736, + "step": 656 + }, + { + "epoch": 0.02, + "learning_rate": 1.0752864157119479e-05, + "loss": 1.194, + "step": 657 + }, + { + "epoch": 0.02, + "learning_rate": 1.076923076923077e-05, + "loss": 1.1773, + "step": 658 + }, + { + "epoch": 0.02, + "learning_rate": 1.0785597381342064e-05, + "loss": 1.1417, + "step": 659 + }, + { + "epoch": 0.02, + "learning_rate": 1.0801963993453358e-05, + "loss": 1.1707, + "step": 660 + }, + { + "epoch": 0.02, + "learning_rate": 1.0818330605564648e-05, + "loss": 1.1388, + "step": 661 + }, + { + "epoch": 0.02, + "learning_rate": 1.0834697217675943e-05, + "loss": 1.1499, + "step": 662 + }, + { + "epoch": 0.02, + "learning_rate": 1.0851063829787233e-05, + "loss": 1.1465, + "step": 663 + }, + { + "epoch": 0.02, + "learning_rate": 1.0867430441898528e-05, + "loss": 1.1546, + "step": 664 + }, + { + "epoch": 0.02, + "learning_rate": 1.0883797054009822e-05, + "loss": 1.2042, + "step": 665 + }, + { + "epoch": 0.02, + "learning_rate": 1.0900163666121113e-05, + "loss": 1.1801, + "step": 666 + }, + { + "epoch": 0.02, + "learning_rate": 1.0916530278232407e-05, + "loss": 1.1675, + "step": 667 + }, + { + "epoch": 0.02, + "learning_rate": 1.0932896890343701e-05, + "loss": 1.1836, + "step": 668 + }, + { + "epoch": 0.02, + "learning_rate": 1.0949263502454992e-05, + "loss": 1.1703, + "step": 669 + }, + { + "epoch": 0.02, + "learning_rate": 1.0965630114566286e-05, + "loss": 1.1779, + "step": 670 + }, + { + "epoch": 0.02, + "learning_rate": 1.098199672667758e-05, + "loss": 1.139, + "step": 671 + }, + { + "epoch": 0.02, + "learning_rate": 1.099836333878887e-05, + "loss": 1.1468, + "step": 672 + }, + { + "epoch": 0.02, + "learning_rate": 1.1014729950900165e-05, + "loss": 1.1639, + "step": 673 + }, + { + "epoch": 0.02, + "learning_rate": 1.1031096563011459e-05, + "loss": 1.1816, + "step": 674 + }, + { + "epoch": 0.02, + "learning_rate": 1.104746317512275e-05, + "loss": 1.2107, + "step": 675 + }, + { + "epoch": 0.02, + "learning_rate": 1.1063829787234044e-05, + "loss": 1.1604, + "step": 676 + }, + { + "epoch": 0.02, + "learning_rate": 1.1080196399345335e-05, + "loss": 1.1719, + "step": 677 + }, + { + "epoch": 0.02, + "learning_rate": 1.1096563011456629e-05, + "loss": 1.1661, + "step": 678 + }, + { + "epoch": 0.02, + "learning_rate": 1.1112929623567923e-05, + "loss": 1.1559, + "step": 679 + }, + { + "epoch": 0.02, + "learning_rate": 1.1129296235679214e-05, + "loss": 1.1283, + "step": 680 + }, + { + "epoch": 0.02, + "learning_rate": 1.1145662847790508e-05, + "loss": 1.1165, + "step": 681 + }, + { + "epoch": 0.02, + "learning_rate": 1.1162029459901802e-05, + "loss": 1.1607, + "step": 682 + }, + { + "epoch": 0.02, + "learning_rate": 1.1178396072013093e-05, + "loss": 1.1298, + "step": 683 + }, + { + "epoch": 0.02, + "learning_rate": 1.1194762684124387e-05, + "loss": 1.1275, + "step": 684 + }, + { + "epoch": 0.02, + "learning_rate": 1.1211129296235681e-05, + "loss": 1.1729, + "step": 685 + }, + { + "epoch": 0.02, + "learning_rate": 1.1227495908346972e-05, + "loss": 1.1349, + "step": 686 + }, + { + "epoch": 0.02, + "learning_rate": 1.1243862520458266e-05, + "loss": 1.2012, + "step": 687 + }, + { + "epoch": 0.02, + "learning_rate": 1.126022913256956e-05, + "loss": 1.1385, + "step": 688 + }, + { + "epoch": 0.02, + "learning_rate": 1.1276595744680851e-05, + "loss": 1.1476, + "step": 689 + }, + { + "epoch": 0.02, + "learning_rate": 1.1292962356792146e-05, + "loss": 1.1696, + "step": 690 + }, + { + "epoch": 0.02, + "learning_rate": 1.1309328968903436e-05, + "loss": 1.1257, + "step": 691 + }, + { + "epoch": 0.02, + "learning_rate": 1.132569558101473e-05, + "loss": 1.1393, + "step": 692 + }, + { + "epoch": 0.02, + "learning_rate": 1.1342062193126025e-05, + "loss": 1.2084, + "step": 693 + }, + { + "epoch": 0.02, + "learning_rate": 1.1358428805237315e-05, + "loss": 1.1769, + "step": 694 + }, + { + "epoch": 0.02, + "learning_rate": 1.137479541734861e-05, + "loss": 1.1807, + "step": 695 + }, + { + "epoch": 0.02, + "learning_rate": 1.1391162029459904e-05, + "loss": 1.1907, + "step": 696 + }, + { + "epoch": 0.02, + "learning_rate": 1.1407528641571195e-05, + "loss": 1.1758, + "step": 697 + }, + { + "epoch": 0.02, + "learning_rate": 1.1423895253682489e-05, + "loss": 1.125, + "step": 698 + }, + { + "epoch": 0.02, + "learning_rate": 1.1440261865793783e-05, + "loss": 1.1492, + "step": 699 + }, + { + "epoch": 0.02, + "learning_rate": 1.1456628477905074e-05, + "loss": 1.1436, + "step": 700 + }, + { + "epoch": 0.02, + "learning_rate": 1.1472995090016368e-05, + "loss": 1.1769, + "step": 701 + }, + { + "epoch": 0.02, + "learning_rate": 1.1489361702127662e-05, + "loss": 1.124, + "step": 702 + }, + { + "epoch": 0.02, + "learning_rate": 1.1505728314238953e-05, + "loss": 1.154, + "step": 703 + }, + { + "epoch": 0.02, + "learning_rate": 1.1522094926350247e-05, + "loss": 1.1555, + "step": 704 + }, + { + "epoch": 0.02, + "learning_rate": 1.1538461538461538e-05, + "loss": 1.1609, + "step": 705 + }, + { + "epoch": 0.02, + "learning_rate": 1.1554828150572832e-05, + "loss": 1.1589, + "step": 706 + }, + { + "epoch": 0.02, + "learning_rate": 1.1571194762684126e-05, + "loss": 1.1432, + "step": 707 + }, + { + "epoch": 0.02, + "learning_rate": 1.1587561374795417e-05, + "loss": 1.1408, + "step": 708 + }, + { + "epoch": 0.02, + "learning_rate": 1.1603927986906711e-05, + "loss": 1.1804, + "step": 709 + }, + { + "epoch": 0.02, + "learning_rate": 1.1620294599018005e-05, + "loss": 1.1294, + "step": 710 + }, + { + "epoch": 0.02, + "learning_rate": 1.1636661211129296e-05, + "loss": 1.1816, + "step": 711 + }, + { + "epoch": 0.02, + "learning_rate": 1.165302782324059e-05, + "loss": 1.1473, + "step": 712 + }, + { + "epoch": 0.02, + "learning_rate": 1.1669394435351884e-05, + "loss": 1.1599, + "step": 713 + }, + { + "epoch": 0.02, + "learning_rate": 1.1685761047463175e-05, + "loss": 1.1369, + "step": 714 + }, + { + "epoch": 0.02, + "learning_rate": 1.170212765957447e-05, + "loss": 1.1785, + "step": 715 + }, + { + "epoch": 0.02, + "learning_rate": 1.1718494271685764e-05, + "loss": 1.1669, + "step": 716 + }, + { + "epoch": 0.02, + "learning_rate": 1.1734860883797054e-05, + "loss": 1.1315, + "step": 717 + }, + { + "epoch": 0.02, + "learning_rate": 1.1751227495908349e-05, + "loss": 1.126, + "step": 718 + }, + { + "epoch": 0.02, + "learning_rate": 1.176759410801964e-05, + "loss": 1.1413, + "step": 719 + }, + { + "epoch": 0.02, + "learning_rate": 1.1783960720130933e-05, + "loss": 1.155, + "step": 720 + }, + { + "epoch": 0.02, + "learning_rate": 1.1800327332242228e-05, + "loss": 1.1644, + "step": 721 + }, + { + "epoch": 0.02, + "learning_rate": 1.1816693944353518e-05, + "loss": 1.1453, + "step": 722 + }, + { + "epoch": 0.02, + "learning_rate": 1.1833060556464813e-05, + "loss": 1.1967, + "step": 723 + }, + { + "epoch": 0.02, + "learning_rate": 1.1849427168576107e-05, + "loss": 1.1672, + "step": 724 + }, + { + "epoch": 0.02, + "learning_rate": 1.1865793780687398e-05, + "loss": 1.1627, + "step": 725 + }, + { + "epoch": 0.02, + "learning_rate": 1.1882160392798692e-05, + "loss": 1.137, + "step": 726 + }, + { + "epoch": 0.02, + "learning_rate": 1.1898527004909986e-05, + "loss": 1.1428, + "step": 727 + }, + { + "epoch": 0.02, + "learning_rate": 1.1914893617021277e-05, + "loss": 1.1718, + "step": 728 + }, + { + "epoch": 0.02, + "learning_rate": 1.1931260229132571e-05, + "loss": 1.1656, + "step": 729 + }, + { + "epoch": 0.02, + "learning_rate": 1.1947626841243865e-05, + "loss": 1.1519, + "step": 730 + }, + { + "epoch": 0.02, + "learning_rate": 1.1963993453355156e-05, + "loss": 1.1567, + "step": 731 + }, + { + "epoch": 0.02, + "learning_rate": 1.198036006546645e-05, + "loss": 1.1462, + "step": 732 + }, + { + "epoch": 0.02, + "learning_rate": 1.199672667757774e-05, + "loss": 1.1548, + "step": 733 + }, + { + "epoch": 0.02, + "learning_rate": 1.2013093289689035e-05, + "loss": 1.153, + "step": 734 + }, + { + "epoch": 0.02, + "learning_rate": 1.2029459901800329e-05, + "loss": 1.1869, + "step": 735 + }, + { + "epoch": 0.02, + "learning_rate": 1.204582651391162e-05, + "loss": 1.1724, + "step": 736 + }, + { + "epoch": 0.02, + "learning_rate": 1.2062193126022914e-05, + "loss": 1.0906, + "step": 737 + }, + { + "epoch": 0.02, + "learning_rate": 1.2078559738134208e-05, + "loss": 1.1655, + "step": 738 + }, + { + "epoch": 0.02, + "learning_rate": 1.2094926350245499e-05, + "loss": 1.1351, + "step": 739 + }, + { + "epoch": 0.02, + "learning_rate": 1.2111292962356793e-05, + "loss": 1.1573, + "step": 740 + }, + { + "epoch": 0.02, + "learning_rate": 1.2127659574468087e-05, + "loss": 1.1452, + "step": 741 + }, + { + "epoch": 0.02, + "learning_rate": 1.2144026186579378e-05, + "loss": 1.1979, + "step": 742 + }, + { + "epoch": 0.02, + "learning_rate": 1.2160392798690672e-05, + "loss": 1.1808, + "step": 743 + }, + { + "epoch": 0.02, + "learning_rate": 1.2176759410801967e-05, + "loss": 1.1619, + "step": 744 + }, + { + "epoch": 0.02, + "learning_rate": 1.2193126022913257e-05, + "loss": 1.142, + "step": 745 + }, + { + "epoch": 0.02, + "learning_rate": 1.2209492635024551e-05, + "loss": 1.1314, + "step": 746 + }, + { + "epoch": 0.02, + "learning_rate": 1.2225859247135842e-05, + "loss": 1.1572, + "step": 747 + }, + { + "epoch": 0.02, + "learning_rate": 1.2242225859247136e-05, + "loss": 1.1626, + "step": 748 + }, + { + "epoch": 0.02, + "learning_rate": 1.225859247135843e-05, + "loss": 1.1643, + "step": 749 + }, + { + "epoch": 0.02, + "learning_rate": 1.2274959083469721e-05, + "loss": 1.1729, + "step": 750 + }, + { + "epoch": 0.02, + "learning_rate": 1.2291325695581016e-05, + "loss": 1.1429, + "step": 751 + }, + { + "epoch": 0.02, + "learning_rate": 1.230769230769231e-05, + "loss": 1.1733, + "step": 752 + }, + { + "epoch": 0.02, + "learning_rate": 1.23240589198036e-05, + "loss": 1.1393, + "step": 753 + }, + { + "epoch": 0.02, + "learning_rate": 1.2340425531914895e-05, + "loss": 1.1628, + "step": 754 + }, + { + "epoch": 0.02, + "learning_rate": 1.2356792144026189e-05, + "loss": 1.1669, + "step": 755 + }, + { + "epoch": 0.02, + "learning_rate": 1.237315875613748e-05, + "loss": 1.154, + "step": 756 + }, + { + "epoch": 0.02, + "learning_rate": 1.2389525368248774e-05, + "loss": 1.1458, + "step": 757 + }, + { + "epoch": 0.02, + "learning_rate": 1.2405891980360068e-05, + "loss": 1.1529, + "step": 758 + }, + { + "epoch": 0.02, + "learning_rate": 1.2422258592471359e-05, + "loss": 1.136, + "step": 759 + }, + { + "epoch": 0.02, + "learning_rate": 1.2438625204582653e-05, + "loss": 1.1907, + "step": 760 + }, + { + "epoch": 0.02, + "learning_rate": 1.2454991816693944e-05, + "loss": 1.1308, + "step": 761 + }, + { + "epoch": 0.02, + "learning_rate": 1.2471358428805238e-05, + "loss": 1.1634, + "step": 762 + }, + { + "epoch": 0.02, + "learning_rate": 1.2487725040916532e-05, + "loss": 1.1646, + "step": 763 + }, + { + "epoch": 0.02, + "learning_rate": 1.2504091653027823e-05, + "loss": 1.1501, + "step": 764 + }, + { + "epoch": 0.02, + "learning_rate": 1.2520458265139117e-05, + "loss": 1.1297, + "step": 765 + }, + { + "epoch": 0.02, + "learning_rate": 1.2536824877250411e-05, + "loss": 1.1445, + "step": 766 + }, + { + "epoch": 0.02, + "learning_rate": 1.2553191489361702e-05, + "loss": 1.1514, + "step": 767 + }, + { + "epoch": 0.02, + "learning_rate": 1.2569558101472996e-05, + "loss": 1.1332, + "step": 768 + }, + { + "epoch": 0.02, + "learning_rate": 1.258592471358429e-05, + "loss": 1.1658, + "step": 769 + }, + { + "epoch": 0.02, + "learning_rate": 1.2602291325695581e-05, + "loss": 1.1882, + "step": 770 + }, + { + "epoch": 0.02, + "learning_rate": 1.2618657937806875e-05, + "loss": 1.1902, + "step": 771 + }, + { + "epoch": 0.02, + "learning_rate": 1.263502454991817e-05, + "loss": 1.1232, + "step": 772 + }, + { + "epoch": 0.02, + "learning_rate": 1.265139116202946e-05, + "loss": 1.1869, + "step": 773 + }, + { + "epoch": 0.02, + "learning_rate": 1.2667757774140754e-05, + "loss": 1.1844, + "step": 774 + }, + { + "epoch": 0.02, + "learning_rate": 1.2684124386252045e-05, + "loss": 1.152, + "step": 775 + }, + { + "epoch": 0.02, + "learning_rate": 1.270049099836334e-05, + "loss": 1.1916, + "step": 776 + }, + { + "epoch": 0.02, + "learning_rate": 1.2716857610474634e-05, + "loss": 1.2124, + "step": 777 + }, + { + "epoch": 0.02, + "learning_rate": 1.2733224222585924e-05, + "loss": 1.1242, + "step": 778 + }, + { + "epoch": 0.02, + "learning_rate": 1.2749590834697219e-05, + "loss": 1.1873, + "step": 779 + }, + { + "epoch": 0.02, + "learning_rate": 1.2765957446808513e-05, + "loss": 1.1645, + "step": 780 + }, + { + "epoch": 0.02, + "learning_rate": 1.2782324058919803e-05, + "loss": 1.155, + "step": 781 + }, + { + "epoch": 0.02, + "learning_rate": 1.2798690671031098e-05, + "loss": 1.1988, + "step": 782 + }, + { + "epoch": 0.02, + "learning_rate": 1.2815057283142392e-05, + "loss": 1.1418, + "step": 783 + }, + { + "epoch": 0.02, + "learning_rate": 1.2831423895253683e-05, + "loss": 1.1492, + "step": 784 + }, + { + "epoch": 0.02, + "learning_rate": 1.2847790507364977e-05, + "loss": 1.1043, + "step": 785 + }, + { + "epoch": 0.02, + "learning_rate": 1.2864157119476271e-05, + "loss": 1.1594, + "step": 786 + }, + { + "epoch": 0.02, + "learning_rate": 1.2880523731587562e-05, + "loss": 1.1469, + "step": 787 + }, + { + "epoch": 0.02, + "learning_rate": 1.2896890343698856e-05, + "loss": 1.1784, + "step": 788 + }, + { + "epoch": 0.02, + "learning_rate": 1.2913256955810147e-05, + "loss": 1.1481, + "step": 789 + }, + { + "epoch": 0.02, + "learning_rate": 1.2929623567921441e-05, + "loss": 1.1501, + "step": 790 + }, + { + "epoch": 0.02, + "learning_rate": 1.2945990180032735e-05, + "loss": 1.1475, + "step": 791 + }, + { + "epoch": 0.02, + "learning_rate": 1.2962356792144026e-05, + "loss": 1.1186, + "step": 792 + }, + { + "epoch": 0.02, + "learning_rate": 1.297872340425532e-05, + "loss": 1.1935, + "step": 793 + }, + { + "epoch": 0.02, + "learning_rate": 1.2995090016366614e-05, + "loss": 1.1662, + "step": 794 + }, + { + "epoch": 0.02, + "learning_rate": 1.3011456628477905e-05, + "loss": 1.1709, + "step": 795 + }, + { + "epoch": 0.02, + "learning_rate": 1.3027823240589199e-05, + "loss": 1.1977, + "step": 796 + }, + { + "epoch": 0.02, + "learning_rate": 1.3044189852700493e-05, + "loss": 1.1391, + "step": 797 + }, + { + "epoch": 0.02, + "learning_rate": 1.3060556464811784e-05, + "loss": 1.1674, + "step": 798 + }, + { + "epoch": 0.02, + "learning_rate": 1.3076923076923078e-05, + "loss": 1.1395, + "step": 799 + }, + { + "epoch": 0.02, + "learning_rate": 1.3093289689034372e-05, + "loss": 1.1735, + "step": 800 + }, + { + "epoch": 0.02, + "learning_rate": 1.3109656301145663e-05, + "loss": 1.1462, + "step": 801 + }, + { + "epoch": 0.02, + "learning_rate": 1.3126022913256957e-05, + "loss": 1.1167, + "step": 802 + }, + { + "epoch": 0.02, + "learning_rate": 1.3142389525368248e-05, + "loss": 1.1692, + "step": 803 + }, + { + "epoch": 0.02, + "learning_rate": 1.3158756137479542e-05, + "loss": 1.1644, + "step": 804 + }, + { + "epoch": 0.02, + "learning_rate": 1.3175122749590837e-05, + "loss": 1.1561, + "step": 805 + }, + { + "epoch": 0.02, + "learning_rate": 1.3191489361702127e-05, + "loss": 1.1535, + "step": 806 + }, + { + "epoch": 0.02, + "learning_rate": 1.3207855973813421e-05, + "loss": 1.1671, + "step": 807 + }, + { + "epoch": 0.02, + "learning_rate": 1.3224222585924716e-05, + "loss": 1.1561, + "step": 808 + }, + { + "epoch": 0.02, + "learning_rate": 1.3240589198036006e-05, + "loss": 1.1559, + "step": 809 + }, + { + "epoch": 0.02, + "learning_rate": 1.32569558101473e-05, + "loss": 1.1701, + "step": 810 + }, + { + "epoch": 0.02, + "learning_rate": 1.3273322422258595e-05, + "loss": 1.1954, + "step": 811 + }, + { + "epoch": 0.02, + "learning_rate": 1.3289689034369886e-05, + "loss": 1.1304, + "step": 812 + }, + { + "epoch": 0.02, + "learning_rate": 1.330605564648118e-05, + "loss": 1.1597, + "step": 813 + }, + { + "epoch": 0.02, + "learning_rate": 1.3322422258592474e-05, + "loss": 1.1273, + "step": 814 + }, + { + "epoch": 0.02, + "learning_rate": 1.3338788870703765e-05, + "loss": 1.1439, + "step": 815 + }, + { + "epoch": 0.02, + "learning_rate": 1.3355155482815059e-05, + "loss": 1.1731, + "step": 816 + }, + { + "epoch": 0.02, + "learning_rate": 1.3371522094926353e-05, + "loss": 1.1684, + "step": 817 + }, + { + "epoch": 0.02, + "learning_rate": 1.3387888707037644e-05, + "loss": 1.1367, + "step": 818 + }, + { + "epoch": 0.02, + "learning_rate": 1.3404255319148938e-05, + "loss": 1.1769, + "step": 819 + }, + { + "epoch": 0.02, + "learning_rate": 1.3420621931260229e-05, + "loss": 1.1501, + "step": 820 + }, + { + "epoch": 0.02, + "learning_rate": 1.3436988543371523e-05, + "loss": 1.1878, + "step": 821 + }, + { + "epoch": 0.02, + "learning_rate": 1.3453355155482817e-05, + "loss": 1.1458, + "step": 822 + }, + { + "epoch": 0.02, + "learning_rate": 1.3469721767594108e-05, + "loss": 1.1183, + "step": 823 + }, + { + "epoch": 0.02, + "learning_rate": 1.3486088379705402e-05, + "loss": 1.1611, + "step": 824 + }, + { + "epoch": 0.02, + "learning_rate": 1.3502454991816696e-05, + "loss": 1.1621, + "step": 825 + }, + { + "epoch": 0.02, + "learning_rate": 1.3518821603927987e-05, + "loss": 1.172, + "step": 826 + }, + { + "epoch": 0.02, + "learning_rate": 1.3535188216039281e-05, + "loss": 1.1272, + "step": 827 + }, + { + "epoch": 0.02, + "learning_rate": 1.3551554828150575e-05, + "loss": 1.135, + "step": 828 + }, + { + "epoch": 0.02, + "learning_rate": 1.3567921440261866e-05, + "loss": 1.1544, + "step": 829 + }, + { + "epoch": 0.02, + "learning_rate": 1.358428805237316e-05, + "loss": 1.1945, + "step": 830 + }, + { + "epoch": 0.02, + "learning_rate": 1.3600654664484455e-05, + "loss": 1.1394, + "step": 831 + }, + { + "epoch": 0.02, + "learning_rate": 1.3617021276595745e-05, + "loss": 1.1508, + "step": 832 + }, + { + "epoch": 0.02, + "learning_rate": 1.363338788870704e-05, + "loss": 1.1803, + "step": 833 + }, + { + "epoch": 0.02, + "learning_rate": 1.364975450081833e-05, + "loss": 1.1642, + "step": 834 + }, + { + "epoch": 0.02, + "learning_rate": 1.3666121112929624e-05, + "loss": 1.1863, + "step": 835 + }, + { + "epoch": 0.02, + "learning_rate": 1.3682487725040919e-05, + "loss": 1.1405, + "step": 836 + }, + { + "epoch": 0.02, + "learning_rate": 1.369885433715221e-05, + "loss": 1.173, + "step": 837 + }, + { + "epoch": 0.02, + "learning_rate": 1.3715220949263504e-05, + "loss": 1.1451, + "step": 838 + }, + { + "epoch": 0.02, + "learning_rate": 1.3731587561374798e-05, + "loss": 1.1513, + "step": 839 + }, + { + "epoch": 0.02, + "learning_rate": 1.3747954173486089e-05, + "loss": 1.1695, + "step": 840 + }, + { + "epoch": 0.02, + "learning_rate": 1.3764320785597383e-05, + "loss": 1.1682, + "step": 841 + }, + { + "epoch": 0.02, + "learning_rate": 1.3780687397708677e-05, + "loss": 1.1553, + "step": 842 + }, + { + "epoch": 0.02, + "learning_rate": 1.3797054009819968e-05, + "loss": 1.1731, + "step": 843 + }, + { + "epoch": 0.02, + "learning_rate": 1.3813420621931262e-05, + "loss": 1.1483, + "step": 844 + }, + { + "epoch": 0.02, + "learning_rate": 1.3829787234042556e-05, + "loss": 1.1236, + "step": 845 + }, + { + "epoch": 0.02, + "learning_rate": 1.3846153846153847e-05, + "loss": 1.1372, + "step": 846 + }, + { + "epoch": 0.02, + "learning_rate": 1.3862520458265141e-05, + "loss": 1.1196, + "step": 847 + }, + { + "epoch": 0.02, + "learning_rate": 1.3878887070376432e-05, + "loss": 1.1453, + "step": 848 + }, + { + "epoch": 0.02, + "learning_rate": 1.3895253682487726e-05, + "loss": 1.166, + "step": 849 + }, + { + "epoch": 0.02, + "learning_rate": 1.391162029459902e-05, + "loss": 1.1348, + "step": 850 + }, + { + "epoch": 0.02, + "learning_rate": 1.3927986906710311e-05, + "loss": 1.1428, + "step": 851 + }, + { + "epoch": 0.02, + "learning_rate": 1.3944353518821605e-05, + "loss": 1.1626, + "step": 852 + }, + { + "epoch": 0.02, + "learning_rate": 1.39607201309329e-05, + "loss": 1.1327, + "step": 853 + }, + { + "epoch": 0.02, + "learning_rate": 1.397708674304419e-05, + "loss": 1.1306, + "step": 854 + }, + { + "epoch": 0.02, + "learning_rate": 1.3993453355155484e-05, + "loss": 1.1907, + "step": 855 + }, + { + "epoch": 0.02, + "learning_rate": 1.4009819967266778e-05, + "loss": 1.1777, + "step": 856 + }, + { + "epoch": 0.02, + "learning_rate": 1.4026186579378069e-05, + "loss": 1.1659, + "step": 857 + }, + { + "epoch": 0.02, + "learning_rate": 1.4042553191489363e-05, + "loss": 1.1413, + "step": 858 + }, + { + "epoch": 0.02, + "learning_rate": 1.4058919803600657e-05, + "loss": 1.1522, + "step": 859 + }, + { + "epoch": 0.02, + "learning_rate": 1.4075286415711948e-05, + "loss": 1.1112, + "step": 860 + }, + { + "epoch": 0.02, + "learning_rate": 1.4091653027823242e-05, + "loss": 1.1693, + "step": 861 + }, + { + "epoch": 0.02, + "learning_rate": 1.4108019639934533e-05, + "loss": 1.1713, + "step": 862 + }, + { + "epoch": 0.02, + "learning_rate": 1.4124386252045827e-05, + "loss": 1.1431, + "step": 863 + }, + { + "epoch": 0.02, + "learning_rate": 1.4140752864157122e-05, + "loss": 1.1616, + "step": 864 + }, + { + "epoch": 0.02, + "learning_rate": 1.4157119476268412e-05, + "loss": 1.1688, + "step": 865 + }, + { + "epoch": 0.02, + "learning_rate": 1.4173486088379706e-05, + "loss": 1.1635, + "step": 866 + }, + { + "epoch": 0.02, + "learning_rate": 1.4189852700491e-05, + "loss": 1.1132, + "step": 867 + }, + { + "epoch": 0.02, + "learning_rate": 1.4206219312602291e-05, + "loss": 1.1259, + "step": 868 + }, + { + "epoch": 0.02, + "learning_rate": 1.4222585924713586e-05, + "loss": 1.1462, + "step": 869 + }, + { + "epoch": 0.02, + "learning_rate": 1.423895253682488e-05, + "loss": 1.1741, + "step": 870 + }, + { + "epoch": 0.02, + "learning_rate": 1.425531914893617e-05, + "loss": 1.1542, + "step": 871 + }, + { + "epoch": 0.02, + "learning_rate": 1.4271685761047465e-05, + "loss": 1.1652, + "step": 872 + }, + { + "epoch": 0.02, + "learning_rate": 1.4288052373158759e-05, + "loss": 1.1701, + "step": 873 + }, + { + "epoch": 0.02, + "learning_rate": 1.430441898527005e-05, + "loss": 1.138, + "step": 874 + }, + { + "epoch": 0.02, + "learning_rate": 1.4320785597381344e-05, + "loss": 1.1641, + "step": 875 + }, + { + "epoch": 0.02, + "learning_rate": 1.4337152209492635e-05, + "loss": 1.1404, + "step": 876 + }, + { + "epoch": 0.02, + "learning_rate": 1.4353518821603929e-05, + "loss": 1.1657, + "step": 877 + }, + { + "epoch": 0.02, + "learning_rate": 1.4369885433715223e-05, + "loss": 1.0998, + "step": 878 + }, + { + "epoch": 0.02, + "learning_rate": 1.4386252045826514e-05, + "loss": 1.1116, + "step": 879 + }, + { + "epoch": 0.02, + "learning_rate": 1.4402618657937808e-05, + "loss": 1.1119, + "step": 880 + }, + { + "epoch": 0.02, + "learning_rate": 1.4418985270049102e-05, + "loss": 1.145, + "step": 881 + }, + { + "epoch": 0.02, + "learning_rate": 1.4435351882160393e-05, + "loss": 1.1658, + "step": 882 + }, + { + "epoch": 0.02, + "learning_rate": 1.4451718494271687e-05, + "loss": 1.1501, + "step": 883 + }, + { + "epoch": 0.02, + "learning_rate": 1.4468085106382981e-05, + "loss": 1.1381, + "step": 884 + }, + { + "epoch": 0.02, + "learning_rate": 1.4484451718494272e-05, + "loss": 1.1378, + "step": 885 + }, + { + "epoch": 0.02, + "learning_rate": 1.4500818330605566e-05, + "loss": 1.1326, + "step": 886 + }, + { + "epoch": 0.02, + "learning_rate": 1.451718494271686e-05, + "loss": 1.1217, + "step": 887 + }, + { + "epoch": 0.02, + "learning_rate": 1.4533551554828151e-05, + "loss": 1.1213, + "step": 888 + }, + { + "epoch": 0.02, + "learning_rate": 1.4549918166939445e-05, + "loss": 1.1523, + "step": 889 + }, + { + "epoch": 0.02, + "learning_rate": 1.4566284779050736e-05, + "loss": 1.1563, + "step": 890 + }, + { + "epoch": 0.02, + "learning_rate": 1.458265139116203e-05, + "loss": 1.1558, + "step": 891 + }, + { + "epoch": 0.02, + "learning_rate": 1.4599018003273324e-05, + "loss": 1.1378, + "step": 892 + }, + { + "epoch": 0.02, + "learning_rate": 1.4615384615384615e-05, + "loss": 1.1446, + "step": 893 + }, + { + "epoch": 0.02, + "learning_rate": 1.463175122749591e-05, + "loss": 1.1324, + "step": 894 + }, + { + "epoch": 0.02, + "learning_rate": 1.4648117839607204e-05, + "loss": 1.1411, + "step": 895 + }, + { + "epoch": 0.02, + "learning_rate": 1.4664484451718494e-05, + "loss": 1.1787, + "step": 896 + }, + { + "epoch": 0.02, + "learning_rate": 1.4680851063829789e-05, + "loss": 1.1351, + "step": 897 + }, + { + "epoch": 0.02, + "learning_rate": 1.4697217675941083e-05, + "loss": 1.193, + "step": 898 + }, + { + "epoch": 0.02, + "learning_rate": 1.4713584288052374e-05, + "loss": 1.137, + "step": 899 + }, + { + "epoch": 0.02, + "learning_rate": 1.4729950900163668e-05, + "loss": 1.195, + "step": 900 + }, + { + "epoch": 0.02, + "learning_rate": 1.4746317512274962e-05, + "loss": 1.1517, + "step": 901 + }, + { + "epoch": 0.02, + "learning_rate": 1.4762684124386253e-05, + "loss": 1.1468, + "step": 902 + }, + { + "epoch": 0.02, + "learning_rate": 1.4779050736497547e-05, + "loss": 1.1475, + "step": 903 + }, + { + "epoch": 0.02, + "learning_rate": 1.4795417348608838e-05, + "loss": 1.1399, + "step": 904 + }, + { + "epoch": 0.02, + "learning_rate": 1.4811783960720132e-05, + "loss": 1.1481, + "step": 905 + }, + { + "epoch": 0.02, + "learning_rate": 1.4828150572831426e-05, + "loss": 1.1646, + "step": 906 + }, + { + "epoch": 0.02, + "learning_rate": 1.4844517184942717e-05, + "loss": 1.1174, + "step": 907 + }, + { + "epoch": 0.02, + "learning_rate": 1.4860883797054011e-05, + "loss": 1.127, + "step": 908 + }, + { + "epoch": 0.02, + "learning_rate": 1.4877250409165305e-05, + "loss": 1.1391, + "step": 909 + }, + { + "epoch": 0.02, + "learning_rate": 1.4893617021276596e-05, + "loss": 1.1529, + "step": 910 + }, + { + "epoch": 0.02, + "learning_rate": 1.490998363338789e-05, + "loss": 1.1366, + "step": 911 + }, + { + "epoch": 0.02, + "learning_rate": 1.4926350245499184e-05, + "loss": 1.1405, + "step": 912 + }, + { + "epoch": 0.02, + "learning_rate": 1.4942716857610475e-05, + "loss": 1.179, + "step": 913 + }, + { + "epoch": 0.02, + "learning_rate": 1.495908346972177e-05, + "loss": 1.1345, + "step": 914 + }, + { + "epoch": 0.02, + "learning_rate": 1.4975450081833063e-05, + "loss": 1.1186, + "step": 915 + }, + { + "epoch": 0.02, + "learning_rate": 1.4991816693944354e-05, + "loss": 1.148, + "step": 916 + }, + { + "epoch": 0.02, + "learning_rate": 1.5008183306055648e-05, + "loss": 1.1687, + "step": 917 + }, + { + "epoch": 0.02, + "learning_rate": 1.5024549918166939e-05, + "loss": 1.176, + "step": 918 + }, + { + "epoch": 0.02, + "learning_rate": 1.5040916530278233e-05, + "loss": 1.1419, + "step": 919 + }, + { + "epoch": 0.02, + "learning_rate": 1.5057283142389527e-05, + "loss": 1.1478, + "step": 920 + }, + { + "epoch": 0.02, + "learning_rate": 1.5073649754500818e-05, + "loss": 1.1563, + "step": 921 + }, + { + "epoch": 0.02, + "learning_rate": 1.5090016366612112e-05, + "loss": 1.1298, + "step": 922 + }, + { + "epoch": 0.02, + "learning_rate": 1.5106382978723407e-05, + "loss": 1.1782, + "step": 923 + }, + { + "epoch": 0.02, + "learning_rate": 1.5122749590834697e-05, + "loss": 1.1556, + "step": 924 + }, + { + "epoch": 0.02, + "learning_rate": 1.5139116202945992e-05, + "loss": 1.1448, + "step": 925 + }, + { + "epoch": 0.02, + "learning_rate": 1.5155482815057286e-05, + "loss": 1.1474, + "step": 926 + }, + { + "epoch": 0.02, + "learning_rate": 1.5171849427168576e-05, + "loss": 1.0807, + "step": 927 + }, + { + "epoch": 0.02, + "learning_rate": 1.518821603927987e-05, + "loss": 1.1703, + "step": 928 + }, + { + "epoch": 0.02, + "learning_rate": 1.5204582651391165e-05, + "loss": 1.1698, + "step": 929 + }, + { + "epoch": 0.02, + "learning_rate": 1.5220949263502456e-05, + "loss": 1.1395, + "step": 930 + }, + { + "epoch": 0.02, + "learning_rate": 1.523731587561375e-05, + "loss": 1.1424, + "step": 931 + }, + { + "epoch": 0.02, + "learning_rate": 1.525368248772504e-05, + "loss": 1.1494, + "step": 932 + }, + { + "epoch": 0.02, + "learning_rate": 1.5270049099836336e-05, + "loss": 1.1406, + "step": 933 + }, + { + "epoch": 0.02, + "learning_rate": 1.528641571194763e-05, + "loss": 1.1535, + "step": 934 + }, + { + "epoch": 0.02, + "learning_rate": 1.530278232405892e-05, + "loss": 1.1466, + "step": 935 + }, + { + "epoch": 0.02, + "learning_rate": 1.5319148936170214e-05, + "loss": 1.136, + "step": 936 + }, + { + "epoch": 0.02, + "learning_rate": 1.5335515548281506e-05, + "loss": 1.1333, + "step": 937 + }, + { + "epoch": 0.02, + "learning_rate": 1.53518821603928e-05, + "loss": 1.1315, + "step": 938 + }, + { + "epoch": 0.02, + "learning_rate": 1.536824877250409e-05, + "loss": 1.1223, + "step": 939 + }, + { + "epoch": 0.02, + "learning_rate": 1.5384615384615387e-05, + "loss": 1.116, + "step": 940 + }, + { + "epoch": 0.02, + "learning_rate": 1.540098199672668e-05, + "loss": 1.154, + "step": 941 + }, + { + "epoch": 0.02, + "learning_rate": 1.5417348608837972e-05, + "loss": 1.1438, + "step": 942 + }, + { + "epoch": 0.02, + "learning_rate": 1.5433715220949265e-05, + "loss": 1.1466, + "step": 943 + }, + { + "epoch": 0.02, + "learning_rate": 1.5450081833060557e-05, + "loss": 1.1655, + "step": 944 + }, + { + "epoch": 0.02, + "learning_rate": 1.546644844517185e-05, + "loss": 1.1138, + "step": 945 + }, + { + "epoch": 0.02, + "learning_rate": 1.5482815057283142e-05, + "loss": 1.1762, + "step": 946 + }, + { + "epoch": 0.02, + "learning_rate": 1.5499181669394438e-05, + "loss": 1.1753, + "step": 947 + }, + { + "epoch": 0.02, + "learning_rate": 1.551554828150573e-05, + "loss": 1.1669, + "step": 948 + }, + { + "epoch": 0.02, + "learning_rate": 1.5531914893617023e-05, + "loss": 1.1458, + "step": 949 + }, + { + "epoch": 0.02, + "learning_rate": 1.5548281505728315e-05, + "loss": 1.1304, + "step": 950 + }, + { + "epoch": 0.02, + "learning_rate": 1.5564648117839608e-05, + "loss": 1.1928, + "step": 951 + }, + { + "epoch": 0.02, + "learning_rate": 1.55810147299509e-05, + "loss": 1.1737, + "step": 952 + }, + { + "epoch": 0.02, + "learning_rate": 1.5597381342062193e-05, + "loss": 1.1407, + "step": 953 + }, + { + "epoch": 0.02, + "learning_rate": 1.561374795417349e-05, + "loss": 1.1358, + "step": 954 + }, + { + "epoch": 0.02, + "learning_rate": 1.563011456628478e-05, + "loss": 1.1622, + "step": 955 + }, + { + "epoch": 0.02, + "learning_rate": 1.5646481178396074e-05, + "loss": 1.1376, + "step": 956 + }, + { + "epoch": 0.02, + "learning_rate": 1.5662847790507366e-05, + "loss": 1.1562, + "step": 957 + }, + { + "epoch": 0.02, + "learning_rate": 1.567921440261866e-05, + "loss": 1.165, + "step": 958 + }, + { + "epoch": 0.02, + "learning_rate": 1.569558101472995e-05, + "loss": 1.1293, + "step": 959 + }, + { + "epoch": 0.02, + "learning_rate": 1.5711947626841244e-05, + "loss": 1.1612, + "step": 960 + }, + { + "epoch": 0.02, + "learning_rate": 1.572831423895254e-05, + "loss": 1.1231, + "step": 961 + }, + { + "epoch": 0.02, + "learning_rate": 1.5744680851063832e-05, + "loss": 1.1293, + "step": 962 + }, + { + "epoch": 0.02, + "learning_rate": 1.5761047463175124e-05, + "loss": 1.1391, + "step": 963 + }, + { + "epoch": 0.02, + "learning_rate": 1.5777414075286417e-05, + "loss": 1.141, + "step": 964 + }, + { + "epoch": 0.02, + "learning_rate": 1.579378068739771e-05, + "loss": 1.1496, + "step": 965 + }, + { + "epoch": 0.02, + "learning_rate": 1.5810147299509002e-05, + "loss": 1.1118, + "step": 966 + }, + { + "epoch": 0.02, + "learning_rate": 1.5826513911620294e-05, + "loss": 1.1524, + "step": 967 + }, + { + "epoch": 0.02, + "learning_rate": 1.584288052373159e-05, + "loss": 1.1713, + "step": 968 + }, + { + "epoch": 0.02, + "learning_rate": 1.5859247135842883e-05, + "loss": 1.153, + "step": 969 + }, + { + "epoch": 0.02, + "learning_rate": 1.5875613747954175e-05, + "loss": 1.1207, + "step": 970 + }, + { + "epoch": 0.02, + "learning_rate": 1.5891980360065468e-05, + "loss": 1.1728, + "step": 971 + }, + { + "epoch": 0.02, + "learning_rate": 1.590834697217676e-05, + "loss": 1.1432, + "step": 972 + }, + { + "epoch": 0.02, + "learning_rate": 1.5924713584288053e-05, + "loss": 1.126, + "step": 973 + }, + { + "epoch": 0.02, + "learning_rate": 1.5941080196399345e-05, + "loss": 1.157, + "step": 974 + }, + { + "epoch": 0.02, + "learning_rate": 1.595744680851064e-05, + "loss": 1.1453, + "step": 975 + }, + { + "epoch": 0.02, + "learning_rate": 1.5973813420621933e-05, + "loss": 1.1242, + "step": 976 + }, + { + "epoch": 0.02, + "learning_rate": 1.5990180032733226e-05, + "loss": 1.1683, + "step": 977 + }, + { + "epoch": 0.02, + "learning_rate": 1.6006546644844518e-05, + "loss": 1.1416, + "step": 978 + }, + { + "epoch": 0.02, + "learning_rate": 1.602291325695581e-05, + "loss": 1.1241, + "step": 979 + }, + { + "epoch": 0.02, + "learning_rate": 1.6039279869067103e-05, + "loss": 1.1368, + "step": 980 + }, + { + "epoch": 0.02, + "learning_rate": 1.6055646481178396e-05, + "loss": 1.1327, + "step": 981 + }, + { + "epoch": 0.02, + "learning_rate": 1.607201309328969e-05, + "loss": 1.1291, + "step": 982 + }, + { + "epoch": 0.02, + "learning_rate": 1.6088379705400984e-05, + "loss": 1.1532, + "step": 983 + }, + { + "epoch": 0.02, + "learning_rate": 1.6104746317512277e-05, + "loss": 1.1354, + "step": 984 + }, + { + "epoch": 0.02, + "learning_rate": 1.612111292962357e-05, + "loss": 1.1237, + "step": 985 + }, + { + "epoch": 0.02, + "learning_rate": 1.613747954173486e-05, + "loss": 1.1797, + "step": 986 + }, + { + "epoch": 0.02, + "learning_rate": 1.6153846153846154e-05, + "loss": 1.153, + "step": 987 + }, + { + "epoch": 0.02, + "learning_rate": 1.6170212765957446e-05, + "loss": 1.1422, + "step": 988 + }, + { + "epoch": 0.02, + "learning_rate": 1.6186579378068742e-05, + "loss": 1.1401, + "step": 989 + }, + { + "epoch": 0.02, + "learning_rate": 1.6202945990180035e-05, + "loss": 1.1234, + "step": 990 + }, + { + "epoch": 0.02, + "learning_rate": 1.6219312602291327e-05, + "loss": 1.1522, + "step": 991 + }, + { + "epoch": 0.02, + "learning_rate": 1.623567921440262e-05, + "loss": 1.1644, + "step": 992 + }, + { + "epoch": 0.02, + "learning_rate": 1.6252045826513912e-05, + "loss": 1.1579, + "step": 993 + }, + { + "epoch": 0.02, + "learning_rate": 1.6268412438625205e-05, + "loss": 1.1278, + "step": 994 + }, + { + "epoch": 0.02, + "learning_rate": 1.6284779050736497e-05, + "loss": 1.0771, + "step": 995 + }, + { + "epoch": 0.02, + "learning_rate": 1.6301145662847793e-05, + "loss": 1.1293, + "step": 996 + }, + { + "epoch": 0.02, + "learning_rate": 1.6317512274959086e-05, + "loss": 1.1268, + "step": 997 + }, + { + "epoch": 0.02, + "learning_rate": 1.6333878887070378e-05, + "loss": 1.1372, + "step": 998 + }, + { + "epoch": 0.02, + "learning_rate": 1.635024549918167e-05, + "loss": 1.1649, + "step": 999 + }, + { + "epoch": 0.02, + "learning_rate": 1.6366612111292963e-05, + "loss": 1.1247, + "step": 1000 + }, + { + "epoch": 0.02, + "learning_rate": 1.6382978723404255e-05, + "loss": 1.1225, + "step": 1001 + }, + { + "epoch": 0.02, + "learning_rate": 1.6399345335515548e-05, + "loss": 1.1452, + "step": 1002 + }, + { + "epoch": 0.02, + "learning_rate": 1.6415711947626844e-05, + "loss": 1.1476, + "step": 1003 + }, + { + "epoch": 0.02, + "learning_rate": 1.6432078559738136e-05, + "loss": 1.1196, + "step": 1004 + }, + { + "epoch": 0.02, + "learning_rate": 1.644844517184943e-05, + "loss": 1.1486, + "step": 1005 + }, + { + "epoch": 0.02, + "learning_rate": 1.646481178396072e-05, + "loss": 1.1334, + "step": 1006 + }, + { + "epoch": 0.02, + "learning_rate": 1.6481178396072014e-05, + "loss": 1.166, + "step": 1007 + }, + { + "epoch": 0.02, + "learning_rate": 1.6497545008183306e-05, + "loss": 1.158, + "step": 1008 + }, + { + "epoch": 0.02, + "learning_rate": 1.65139116202946e-05, + "loss": 1.1368, + "step": 1009 + }, + { + "epoch": 0.02, + "learning_rate": 1.6530278232405895e-05, + "loss": 1.1492, + "step": 1010 + }, + { + "epoch": 0.02, + "learning_rate": 1.6546644844517187e-05, + "loss": 1.1288, + "step": 1011 + }, + { + "epoch": 0.02, + "learning_rate": 1.656301145662848e-05, + "loss": 1.1596, + "step": 1012 + }, + { + "epoch": 0.02, + "learning_rate": 1.6579378068739772e-05, + "loss": 1.1409, + "step": 1013 + }, + { + "epoch": 0.02, + "learning_rate": 1.6595744680851064e-05, + "loss": 1.1083, + "step": 1014 + }, + { + "epoch": 0.02, + "learning_rate": 1.6612111292962357e-05, + "loss": 1.1443, + "step": 1015 + }, + { + "epoch": 0.02, + "learning_rate": 1.662847790507365e-05, + "loss": 1.2075, + "step": 1016 + }, + { + "epoch": 0.02, + "learning_rate": 1.6644844517184945e-05, + "loss": 1.1078, + "step": 1017 + }, + { + "epoch": 0.03, + "learning_rate": 1.6661211129296238e-05, + "loss": 1.202, + "step": 1018 + }, + { + "epoch": 0.03, + "learning_rate": 1.667757774140753e-05, + "loss": 1.0936, + "step": 1019 + }, + { + "epoch": 0.03, + "learning_rate": 1.6693944353518823e-05, + "loss": 1.1007, + "step": 1020 + }, + { + "epoch": 0.03, + "learning_rate": 1.6710310965630115e-05, + "loss": 1.1689, + "step": 1021 + }, + { + "epoch": 0.03, + "learning_rate": 1.6726677577741408e-05, + "loss": 1.1321, + "step": 1022 + }, + { + "epoch": 0.03, + "learning_rate": 1.6743044189852704e-05, + "loss": 1.1524, + "step": 1023 + }, + { + "epoch": 0.03, + "learning_rate": 1.6759410801963996e-05, + "loss": 1.1531, + "step": 1024 + }, + { + "epoch": 0.03, + "learning_rate": 1.677577741407529e-05, + "loss": 1.1682, + "step": 1025 + }, + { + "epoch": 0.03, + "learning_rate": 1.679214402618658e-05, + "loss": 1.1276, + "step": 1026 + }, + { + "epoch": 0.03, + "learning_rate": 1.6808510638297873e-05, + "loss": 1.1649, + "step": 1027 + }, + { + "epoch": 0.03, + "learning_rate": 1.6824877250409166e-05, + "loss": 1.128, + "step": 1028 + }, + { + "epoch": 0.03, + "learning_rate": 1.684124386252046e-05, + "loss": 1.1389, + "step": 1029 + }, + { + "epoch": 0.03, + "learning_rate": 1.6857610474631754e-05, + "loss": 1.1184, + "step": 1030 + }, + { + "epoch": 0.03, + "learning_rate": 1.6873977086743047e-05, + "loss": 1.1351, + "step": 1031 + }, + { + "epoch": 0.03, + "learning_rate": 1.689034369885434e-05, + "loss": 1.1287, + "step": 1032 + }, + { + "epoch": 0.03, + "learning_rate": 1.6906710310965632e-05, + "loss": 1.1395, + "step": 1033 + }, + { + "epoch": 0.03, + "learning_rate": 1.6923076923076924e-05, + "loss": 1.1288, + "step": 1034 + }, + { + "epoch": 0.03, + "learning_rate": 1.6939443535188217e-05, + "loss": 1.1714, + "step": 1035 + }, + { + "epoch": 0.03, + "learning_rate": 1.695581014729951e-05, + "loss": 1.1758, + "step": 1036 + }, + { + "epoch": 0.03, + "learning_rate": 1.6972176759410805e-05, + "loss": 1.1282, + "step": 1037 + }, + { + "epoch": 0.03, + "learning_rate": 1.6988543371522098e-05, + "loss": 1.1675, + "step": 1038 + }, + { + "epoch": 0.03, + "learning_rate": 1.700490998363339e-05, + "loss": 1.1373, + "step": 1039 + }, + { + "epoch": 0.03, + "learning_rate": 1.7021276595744682e-05, + "loss": 1.118, + "step": 1040 + }, + { + "epoch": 0.03, + "learning_rate": 1.7037643207855975e-05, + "loss": 1.146, + "step": 1041 + }, + { + "epoch": 0.03, + "learning_rate": 1.7054009819967267e-05, + "loss": 1.1742, + "step": 1042 + }, + { + "epoch": 0.03, + "learning_rate": 1.707037643207856e-05, + "loss": 1.1238, + "step": 1043 + }, + { + "epoch": 0.03, + "learning_rate": 1.7086743044189856e-05, + "loss": 1.1394, + "step": 1044 + }, + { + "epoch": 0.03, + "learning_rate": 1.7103109656301148e-05, + "loss": 1.1544, + "step": 1045 + }, + { + "epoch": 0.03, + "learning_rate": 1.711947626841244e-05, + "loss": 1.1061, + "step": 1046 + }, + { + "epoch": 0.03, + "learning_rate": 1.7135842880523733e-05, + "loss": 1.2001, + "step": 1047 + }, + { + "epoch": 0.03, + "learning_rate": 1.7152209492635026e-05, + "loss": 1.1843, + "step": 1048 + }, + { + "epoch": 0.03, + "learning_rate": 1.7168576104746318e-05, + "loss": 1.1441, + "step": 1049 + }, + { + "epoch": 0.03, + "learning_rate": 1.718494271685761e-05, + "loss": 1.1625, + "step": 1050 + }, + { + "epoch": 0.03, + "learning_rate": 1.7201309328968906e-05, + "loss": 1.1789, + "step": 1051 + }, + { + "epoch": 0.03, + "learning_rate": 1.72176759410802e-05, + "loss": 1.1023, + "step": 1052 + }, + { + "epoch": 0.03, + "learning_rate": 1.723404255319149e-05, + "loss": 1.1682, + "step": 1053 + }, + { + "epoch": 0.03, + "learning_rate": 1.7250409165302784e-05, + "loss": 1.1321, + "step": 1054 + }, + { + "epoch": 0.03, + "learning_rate": 1.7266775777414076e-05, + "loss": 1.1254, + "step": 1055 + }, + { + "epoch": 0.03, + "learning_rate": 1.728314238952537e-05, + "loss": 1.1437, + "step": 1056 + }, + { + "epoch": 0.03, + "learning_rate": 1.729950900163666e-05, + "loss": 1.1368, + "step": 1057 + }, + { + "epoch": 0.03, + "learning_rate": 1.7315875613747957e-05, + "loss": 1.1393, + "step": 1058 + }, + { + "epoch": 0.03, + "learning_rate": 1.733224222585925e-05, + "loss": 1.1121, + "step": 1059 + }, + { + "epoch": 0.03, + "learning_rate": 1.7348608837970542e-05, + "loss": 1.1547, + "step": 1060 + }, + { + "epoch": 0.03, + "learning_rate": 1.7364975450081835e-05, + "loss": 1.1396, + "step": 1061 + }, + { + "epoch": 0.03, + "learning_rate": 1.7381342062193127e-05, + "loss": 1.1416, + "step": 1062 + }, + { + "epoch": 0.03, + "learning_rate": 1.739770867430442e-05, + "loss": 1.1426, + "step": 1063 + }, + { + "epoch": 0.03, + "learning_rate": 1.7414075286415712e-05, + "loss": 1.1498, + "step": 1064 + }, + { + "epoch": 0.03, + "learning_rate": 1.7430441898527008e-05, + "loss": 1.1099, + "step": 1065 + }, + { + "epoch": 0.03, + "learning_rate": 1.74468085106383e-05, + "loss": 1.132, + "step": 1066 + }, + { + "epoch": 0.03, + "learning_rate": 1.7463175122749593e-05, + "loss": 1.1274, + "step": 1067 + }, + { + "epoch": 0.03, + "learning_rate": 1.7479541734860885e-05, + "loss": 1.1506, + "step": 1068 + }, + { + "epoch": 0.03, + "learning_rate": 1.7495908346972178e-05, + "loss": 1.1687, + "step": 1069 + }, + { + "epoch": 0.03, + "learning_rate": 1.751227495908347e-05, + "loss": 1.1105, + "step": 1070 + }, + { + "epoch": 0.03, + "learning_rate": 1.7528641571194763e-05, + "loss": 1.1278, + "step": 1071 + }, + { + "epoch": 0.03, + "learning_rate": 1.754500818330606e-05, + "loss": 1.1386, + "step": 1072 + }, + { + "epoch": 0.03, + "learning_rate": 1.756137479541735e-05, + "loss": 1.1241, + "step": 1073 + }, + { + "epoch": 0.03, + "learning_rate": 1.7577741407528644e-05, + "loss": 1.1546, + "step": 1074 + }, + { + "epoch": 0.03, + "learning_rate": 1.7594108019639936e-05, + "loss": 1.1192, + "step": 1075 + }, + { + "epoch": 0.03, + "learning_rate": 1.761047463175123e-05, + "loss": 1.1886, + "step": 1076 + }, + { + "epoch": 0.03, + "learning_rate": 1.762684124386252e-05, + "loss": 1.1237, + "step": 1077 + }, + { + "epoch": 0.03, + "learning_rate": 1.7643207855973814e-05, + "loss": 1.1591, + "step": 1078 + }, + { + "epoch": 0.03, + "learning_rate": 1.765957446808511e-05, + "loss": 1.1278, + "step": 1079 + }, + { + "epoch": 0.03, + "learning_rate": 1.7675941080196402e-05, + "loss": 1.1381, + "step": 1080 + }, + { + "epoch": 0.03, + "learning_rate": 1.7692307692307694e-05, + "loss": 1.112, + "step": 1081 + }, + { + "epoch": 0.03, + "learning_rate": 1.7708674304418987e-05, + "loss": 1.1318, + "step": 1082 + }, + { + "epoch": 0.03, + "learning_rate": 1.772504091653028e-05, + "loss": 1.1131, + "step": 1083 + }, + { + "epoch": 0.03, + "learning_rate": 1.7741407528641572e-05, + "loss": 1.1097, + "step": 1084 + }, + { + "epoch": 0.03, + "learning_rate": 1.7757774140752864e-05, + "loss": 1.1236, + "step": 1085 + }, + { + "epoch": 0.03, + "learning_rate": 1.777414075286416e-05, + "loss": 1.1702, + "step": 1086 + }, + { + "epoch": 0.03, + "learning_rate": 1.7790507364975453e-05, + "loss": 1.1333, + "step": 1087 + }, + { + "epoch": 0.03, + "learning_rate": 1.7806873977086745e-05, + "loss": 1.1157, + "step": 1088 + }, + { + "epoch": 0.03, + "learning_rate": 1.7823240589198038e-05, + "loss": 1.1243, + "step": 1089 + }, + { + "epoch": 0.03, + "learning_rate": 1.783960720130933e-05, + "loss": 1.1738, + "step": 1090 + }, + { + "epoch": 0.03, + "learning_rate": 1.7855973813420623e-05, + "loss": 1.1346, + "step": 1091 + }, + { + "epoch": 0.03, + "learning_rate": 1.7872340425531915e-05, + "loss": 1.1263, + "step": 1092 + }, + { + "epoch": 0.03, + "learning_rate": 1.788870703764321e-05, + "loss": 1.1244, + "step": 1093 + }, + { + "epoch": 0.03, + "learning_rate": 1.7905073649754503e-05, + "loss": 1.1582, + "step": 1094 + }, + { + "epoch": 0.03, + "learning_rate": 1.7921440261865796e-05, + "loss": 1.1232, + "step": 1095 + }, + { + "epoch": 0.03, + "learning_rate": 1.793780687397709e-05, + "loss": 1.1633, + "step": 1096 + }, + { + "epoch": 0.03, + "learning_rate": 1.795417348608838e-05, + "loss": 1.1305, + "step": 1097 + }, + { + "epoch": 0.03, + "learning_rate": 1.7970540098199673e-05, + "loss": 1.1299, + "step": 1098 + }, + { + "epoch": 0.03, + "learning_rate": 1.7986906710310966e-05, + "loss": 1.1252, + "step": 1099 + }, + { + "epoch": 0.03, + "learning_rate": 1.800327332242226e-05, + "loss": 1.1337, + "step": 1100 + }, + { + "epoch": 0.03, + "learning_rate": 1.8019639934533554e-05, + "loss": 1.1508, + "step": 1101 + }, + { + "epoch": 0.03, + "learning_rate": 1.8036006546644847e-05, + "loss": 1.1018, + "step": 1102 + }, + { + "epoch": 0.03, + "learning_rate": 1.805237315875614e-05, + "loss": 1.1555, + "step": 1103 + }, + { + "epoch": 0.03, + "learning_rate": 1.806873977086743e-05, + "loss": 1.1238, + "step": 1104 + }, + { + "epoch": 0.03, + "learning_rate": 1.8085106382978724e-05, + "loss": 1.1013, + "step": 1105 + }, + { + "epoch": 0.03, + "learning_rate": 1.8101472995090017e-05, + "loss": 1.1465, + "step": 1106 + }, + { + "epoch": 0.03, + "learning_rate": 1.8117839607201312e-05, + "loss": 1.1529, + "step": 1107 + }, + { + "epoch": 0.03, + "learning_rate": 1.8134206219312605e-05, + "loss": 1.1404, + "step": 1108 + }, + { + "epoch": 0.03, + "learning_rate": 1.8150572831423897e-05, + "loss": 1.1719, + "step": 1109 + }, + { + "epoch": 0.03, + "learning_rate": 1.816693944353519e-05, + "loss": 1.143, + "step": 1110 + }, + { + "epoch": 0.03, + "learning_rate": 1.8183306055646482e-05, + "loss": 1.1206, + "step": 1111 + }, + { + "epoch": 0.03, + "learning_rate": 1.8199672667757775e-05, + "loss": 1.1498, + "step": 1112 + }, + { + "epoch": 0.03, + "learning_rate": 1.8216039279869067e-05, + "loss": 1.1446, + "step": 1113 + }, + { + "epoch": 0.03, + "learning_rate": 1.8232405891980363e-05, + "loss": 1.1309, + "step": 1114 + }, + { + "epoch": 0.03, + "learning_rate": 1.8248772504091656e-05, + "loss": 1.1558, + "step": 1115 + }, + { + "epoch": 0.03, + "learning_rate": 1.8265139116202948e-05, + "loss": 1.1411, + "step": 1116 + }, + { + "epoch": 0.03, + "learning_rate": 1.828150572831424e-05, + "loss": 1.1155, + "step": 1117 + }, + { + "epoch": 0.03, + "learning_rate": 1.8297872340425533e-05, + "loss": 1.139, + "step": 1118 + }, + { + "epoch": 0.03, + "learning_rate": 1.8314238952536826e-05, + "loss": 1.1806, + "step": 1119 + }, + { + "epoch": 0.03, + "learning_rate": 1.8330605564648118e-05, + "loss": 1.1655, + "step": 1120 + }, + { + "epoch": 0.03, + "learning_rate": 1.8346972176759414e-05, + "loss": 1.1523, + "step": 1121 + }, + { + "epoch": 0.03, + "learning_rate": 1.8363338788870706e-05, + "loss": 1.1111, + "step": 1122 + }, + { + "epoch": 0.03, + "learning_rate": 1.8379705400982e-05, + "loss": 1.1148, + "step": 1123 + }, + { + "epoch": 0.03, + "learning_rate": 1.839607201309329e-05, + "loss": 1.1403, + "step": 1124 + }, + { + "epoch": 0.03, + "learning_rate": 1.8412438625204584e-05, + "loss": 1.1606, + "step": 1125 + }, + { + "epoch": 0.03, + "learning_rate": 1.8428805237315876e-05, + "loss": 1.1269, + "step": 1126 + }, + { + "epoch": 0.03, + "learning_rate": 1.844517184942717e-05, + "loss": 1.1143, + "step": 1127 + }, + { + "epoch": 0.03, + "learning_rate": 1.8461538461538465e-05, + "loss": 1.1192, + "step": 1128 + }, + { + "epoch": 0.03, + "learning_rate": 1.8477905073649757e-05, + "loss": 1.1635, + "step": 1129 + }, + { + "epoch": 0.03, + "learning_rate": 1.849427168576105e-05, + "loss": 1.1456, + "step": 1130 + }, + { + "epoch": 0.03, + "learning_rate": 1.8510638297872342e-05, + "loss": 1.1331, + "step": 1131 + }, + { + "epoch": 0.03, + "learning_rate": 1.8527004909983635e-05, + "loss": 1.1309, + "step": 1132 + }, + { + "epoch": 0.03, + "learning_rate": 1.8543371522094927e-05, + "loss": 1.1071, + "step": 1133 + }, + { + "epoch": 0.03, + "learning_rate": 1.855973813420622e-05, + "loss": 1.1425, + "step": 1134 + }, + { + "epoch": 0.03, + "learning_rate": 1.8576104746317515e-05, + "loss": 1.1484, + "step": 1135 + }, + { + "epoch": 0.03, + "learning_rate": 1.8592471358428808e-05, + "loss": 1.1341, + "step": 1136 + }, + { + "epoch": 0.03, + "learning_rate": 1.86088379705401e-05, + "loss": 1.1189, + "step": 1137 + }, + { + "epoch": 0.03, + "learning_rate": 1.8625204582651393e-05, + "loss": 1.1682, + "step": 1138 + }, + { + "epoch": 0.03, + "learning_rate": 1.8641571194762685e-05, + "loss": 1.1305, + "step": 1139 + }, + { + "epoch": 0.03, + "learning_rate": 1.8657937806873978e-05, + "loss": 1.1055, + "step": 1140 + }, + { + "epoch": 0.03, + "learning_rate": 1.867430441898527e-05, + "loss": 1.138, + "step": 1141 + }, + { + "epoch": 0.03, + "learning_rate": 1.8690671031096566e-05, + "loss": 1.1087, + "step": 1142 + }, + { + "epoch": 0.03, + "learning_rate": 1.870703764320786e-05, + "loss": 1.1689, + "step": 1143 + }, + { + "epoch": 0.03, + "learning_rate": 1.872340425531915e-05, + "loss": 1.1688, + "step": 1144 + }, + { + "epoch": 0.03, + "learning_rate": 1.8739770867430444e-05, + "loss": 1.1921, + "step": 1145 + }, + { + "epoch": 0.03, + "learning_rate": 1.8756137479541736e-05, + "loss": 1.1401, + "step": 1146 + }, + { + "epoch": 0.03, + "learning_rate": 1.877250409165303e-05, + "loss": 1.1245, + "step": 1147 + }, + { + "epoch": 0.03, + "learning_rate": 1.878887070376432e-05, + "loss": 1.157, + "step": 1148 + }, + { + "epoch": 0.03, + "learning_rate": 1.8805237315875617e-05, + "loss": 1.1249, + "step": 1149 + }, + { + "epoch": 0.03, + "learning_rate": 1.882160392798691e-05, + "loss": 1.1598, + "step": 1150 + }, + { + "epoch": 0.03, + "learning_rate": 1.8837970540098202e-05, + "loss": 1.1544, + "step": 1151 + }, + { + "epoch": 0.03, + "learning_rate": 1.8854337152209494e-05, + "loss": 1.1084, + "step": 1152 + }, + { + "epoch": 0.03, + "learning_rate": 1.8870703764320787e-05, + "loss": 1.1715, + "step": 1153 + }, + { + "epoch": 0.03, + "learning_rate": 1.888707037643208e-05, + "loss": 1.1362, + "step": 1154 + }, + { + "epoch": 0.03, + "learning_rate": 1.8903436988543372e-05, + "loss": 1.1327, + "step": 1155 + }, + { + "epoch": 0.03, + "learning_rate": 1.8919803600654668e-05, + "loss": 1.1318, + "step": 1156 + }, + { + "epoch": 0.03, + "learning_rate": 1.893617021276596e-05, + "loss": 1.124, + "step": 1157 + }, + { + "epoch": 0.03, + "learning_rate": 1.8952536824877253e-05, + "loss": 1.1706, + "step": 1158 + }, + { + "epoch": 0.03, + "learning_rate": 1.8968903436988545e-05, + "loss": 1.0823, + "step": 1159 + }, + { + "epoch": 0.03, + "learning_rate": 1.8985270049099837e-05, + "loss": 1.1298, + "step": 1160 + }, + { + "epoch": 0.03, + "learning_rate": 1.900163666121113e-05, + "loss": 1.1193, + "step": 1161 + }, + { + "epoch": 0.03, + "learning_rate": 1.9018003273322422e-05, + "loss": 1.113, + "step": 1162 + }, + { + "epoch": 0.03, + "learning_rate": 1.9034369885433718e-05, + "loss": 1.1542, + "step": 1163 + }, + { + "epoch": 0.03, + "learning_rate": 1.905073649754501e-05, + "loss": 1.1321, + "step": 1164 + }, + { + "epoch": 0.03, + "learning_rate": 1.9067103109656303e-05, + "loss": 1.1243, + "step": 1165 + }, + { + "epoch": 0.03, + "learning_rate": 1.9083469721767596e-05, + "loss": 1.1025, + "step": 1166 + }, + { + "epoch": 0.03, + "learning_rate": 1.9099836333878888e-05, + "loss": 1.1258, + "step": 1167 + }, + { + "epoch": 0.03, + "learning_rate": 1.911620294599018e-05, + "loss": 1.1216, + "step": 1168 + }, + { + "epoch": 0.03, + "learning_rate": 1.9132569558101473e-05, + "loss": 1.1588, + "step": 1169 + }, + { + "epoch": 0.03, + "learning_rate": 1.914893617021277e-05, + "loss": 1.1469, + "step": 1170 + }, + { + "epoch": 0.03, + "learning_rate": 1.916530278232406e-05, + "loss": 1.1064, + "step": 1171 + }, + { + "epoch": 0.03, + "learning_rate": 1.9181669394435354e-05, + "loss": 1.1564, + "step": 1172 + }, + { + "epoch": 0.03, + "learning_rate": 1.9198036006546646e-05, + "loss": 1.1836, + "step": 1173 + }, + { + "epoch": 0.03, + "learning_rate": 1.921440261865794e-05, + "loss": 1.1063, + "step": 1174 + }, + { + "epoch": 0.03, + "learning_rate": 1.923076923076923e-05, + "loss": 1.156, + "step": 1175 + }, + { + "epoch": 0.03, + "learning_rate": 1.9247135842880524e-05, + "loss": 1.1124, + "step": 1176 + }, + { + "epoch": 0.03, + "learning_rate": 1.926350245499182e-05, + "loss": 1.1452, + "step": 1177 + }, + { + "epoch": 0.03, + "learning_rate": 1.9279869067103112e-05, + "loss": 1.0885, + "step": 1178 + }, + { + "epoch": 0.03, + "learning_rate": 1.9296235679214405e-05, + "loss": 1.1892, + "step": 1179 + }, + { + "epoch": 0.03, + "learning_rate": 1.9312602291325697e-05, + "loss": 1.1397, + "step": 1180 + }, + { + "epoch": 0.03, + "learning_rate": 1.932896890343699e-05, + "loss": 1.1543, + "step": 1181 + }, + { + "epoch": 0.03, + "learning_rate": 1.9345335515548282e-05, + "loss": 1.127, + "step": 1182 + }, + { + "epoch": 0.03, + "learning_rate": 1.9361702127659575e-05, + "loss": 1.1199, + "step": 1183 + }, + { + "epoch": 0.03, + "learning_rate": 1.937806873977087e-05, + "loss": 1.1395, + "step": 1184 + }, + { + "epoch": 0.03, + "learning_rate": 1.9394435351882163e-05, + "loss": 1.1455, + "step": 1185 + }, + { + "epoch": 0.03, + "learning_rate": 1.9410801963993455e-05, + "loss": 1.1183, + "step": 1186 + }, + { + "epoch": 0.03, + "learning_rate": 1.9427168576104748e-05, + "loss": 1.1689, + "step": 1187 + }, + { + "epoch": 0.03, + "learning_rate": 1.944353518821604e-05, + "loss": 1.1692, + "step": 1188 + }, + { + "epoch": 0.03, + "learning_rate": 1.9459901800327333e-05, + "loss": 1.1241, + "step": 1189 + }, + { + "epoch": 0.03, + "learning_rate": 1.9476268412438625e-05, + "loss": 1.1103, + "step": 1190 + }, + { + "epoch": 0.03, + "learning_rate": 1.949263502454992e-05, + "loss": 1.0914, + "step": 1191 + }, + { + "epoch": 0.03, + "learning_rate": 1.9509001636661214e-05, + "loss": 1.1551, + "step": 1192 + }, + { + "epoch": 0.03, + "learning_rate": 1.9525368248772506e-05, + "loss": 1.1203, + "step": 1193 + }, + { + "epoch": 0.03, + "learning_rate": 1.95417348608838e-05, + "loss": 1.1561, + "step": 1194 + }, + { + "epoch": 0.03, + "learning_rate": 1.955810147299509e-05, + "loss": 1.1523, + "step": 1195 + }, + { + "epoch": 0.03, + "learning_rate": 1.9574468085106384e-05, + "loss": 1.1084, + "step": 1196 + }, + { + "epoch": 0.03, + "learning_rate": 1.9590834697217676e-05, + "loss": 1.1456, + "step": 1197 + }, + { + "epoch": 0.03, + "learning_rate": 1.9607201309328972e-05, + "loss": 1.1232, + "step": 1198 + }, + { + "epoch": 0.03, + "learning_rate": 1.9623567921440264e-05, + "loss": 1.1272, + "step": 1199 + }, + { + "epoch": 0.03, + "learning_rate": 1.9639934533551557e-05, + "loss": 1.1239, + "step": 1200 + }, + { + "epoch": 0.03, + "learning_rate": 1.965630114566285e-05, + "loss": 1.1434, + "step": 1201 + }, + { + "epoch": 0.03, + "learning_rate": 1.9672667757774142e-05, + "loss": 1.0807, + "step": 1202 + }, + { + "epoch": 0.03, + "learning_rate": 1.9689034369885434e-05, + "loss": 1.1733, + "step": 1203 + }, + { + "epoch": 0.03, + "learning_rate": 1.9705400981996727e-05, + "loss": 1.1343, + "step": 1204 + }, + { + "epoch": 0.03, + "learning_rate": 1.9721767594108023e-05, + "loss": 1.1269, + "step": 1205 + }, + { + "epoch": 0.03, + "learning_rate": 1.9738134206219315e-05, + "loss": 1.1277, + "step": 1206 + }, + { + "epoch": 0.03, + "learning_rate": 1.9754500818330608e-05, + "loss": 1.1423, + "step": 1207 + }, + { + "epoch": 0.03, + "learning_rate": 1.97708674304419e-05, + "loss": 1.1647, + "step": 1208 + }, + { + "epoch": 0.03, + "learning_rate": 1.9787234042553193e-05, + "loss": 1.131, + "step": 1209 + }, + { + "epoch": 0.03, + "learning_rate": 1.9803600654664485e-05, + "loss": 1.1578, + "step": 1210 + }, + { + "epoch": 0.03, + "learning_rate": 1.9819967266775778e-05, + "loss": 1.1508, + "step": 1211 + }, + { + "epoch": 0.03, + "learning_rate": 1.9836333878887073e-05, + "loss": 1.1278, + "step": 1212 + }, + { + "epoch": 0.03, + "learning_rate": 1.9852700490998366e-05, + "loss": 1.0853, + "step": 1213 + }, + { + "epoch": 0.03, + "learning_rate": 1.986906710310966e-05, + "loss": 1.1534, + "step": 1214 + }, + { + "epoch": 0.03, + "learning_rate": 1.988543371522095e-05, + "loss": 1.1576, + "step": 1215 + }, + { + "epoch": 0.03, + "learning_rate": 1.9901800327332243e-05, + "loss": 1.1374, + "step": 1216 + }, + { + "epoch": 0.03, + "learning_rate": 1.9918166939443536e-05, + "loss": 1.1498, + "step": 1217 + }, + { + "epoch": 0.03, + "learning_rate": 1.993453355155483e-05, + "loss": 1.1098, + "step": 1218 + }, + { + "epoch": 0.03, + "learning_rate": 1.9950900163666124e-05, + "loss": 1.0934, + "step": 1219 + }, + { + "epoch": 0.03, + "learning_rate": 1.9967266775777417e-05, + "loss": 1.1339, + "step": 1220 + }, + { + "epoch": 0.03, + "learning_rate": 1.998363338788871e-05, + "loss": 1.128, + "step": 1221 + }, + { + "epoch": 0.03, + "learning_rate": 2e-05, + "loss": 1.1735, + "step": 1222 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999968352496e-05, + "loss": 1.1113, + "step": 1223 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999873409982e-05, + "loss": 1.1388, + "step": 1224 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999971517246e-05, + "loss": 1.1658, + "step": 1225 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999949363993e-05, + "loss": 1.1411, + "step": 1226 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999999208812394e-05, + "loss": 1.1158, + "step": 1227 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999998860689852e-05, + "loss": 1.1576, + "step": 1228 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999844927231e-05, + "loss": 1.1595, + "step": 1229 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999997974559765e-05, + "loss": 1.1268, + "step": 1230 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999997436552226e-05, + "loss": 1.1386, + "step": 1231 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999996835249696e-05, + "loss": 1.1452, + "step": 1232 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999996170652172e-05, + "loss": 1.1391, + "step": 1233 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999995442759664e-05, + "loss": 1.1556, + "step": 1234 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999994651572175e-05, + "loss": 1.0971, + "step": 1235 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999993797089715e-05, + "loss": 1.1172, + "step": 1236 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999992879312278e-05, + "loss": 1.1342, + "step": 1237 + }, + { + "epoch": 0.03, + "learning_rate": 1.999999189823988e-05, + "loss": 1.0942, + "step": 1238 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999990853872522e-05, + "loss": 1.1435, + "step": 1239 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999989746210214e-05, + "loss": 1.1212, + "step": 1240 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999988575252965e-05, + "loss": 1.1208, + "step": 1241 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999987341000776e-05, + "loss": 1.1651, + "step": 1242 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999986043453654e-05, + "loss": 1.0838, + "step": 1243 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999984682611612e-05, + "loss": 1.1639, + "step": 1244 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999983258474664e-05, + "loss": 1.1422, + "step": 1245 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999981771042806e-05, + "loss": 1.1232, + "step": 1246 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999980220316055e-05, + "loss": 1.156, + "step": 1247 + }, + { + "epoch": 0.03, + "learning_rate": 1.999997860629442e-05, + "loss": 1.0926, + "step": 1248 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999976928977912e-05, + "loss": 1.1381, + "step": 1249 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999975188366538e-05, + "loss": 1.1284, + "step": 1250 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999973384460314e-05, + "loss": 1.1273, + "step": 1251 + }, + { + "epoch": 0.03, + "learning_rate": 1.999997151725925e-05, + "loss": 1.1582, + "step": 1252 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999969586763355e-05, + "loss": 1.143, + "step": 1253 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999967592972643e-05, + "loss": 1.1284, + "step": 1254 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999965535887126e-05, + "loss": 1.1574, + "step": 1255 + }, + { + "epoch": 0.03, + "learning_rate": 1.999996341550682e-05, + "loss": 1.1753, + "step": 1256 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999961231831733e-05, + "loss": 1.1138, + "step": 1257 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999958984861884e-05, + "loss": 1.1509, + "step": 1258 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999956674597288e-05, + "loss": 1.0941, + "step": 1259 + }, + { + "epoch": 0.03, + "learning_rate": 1.999995430103795e-05, + "loss": 1.141, + "step": 1260 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999951864183898e-05, + "loss": 1.1086, + "step": 1261 + }, + { + "epoch": 0.03, + "learning_rate": 1.999994936403514e-05, + "loss": 1.1224, + "step": 1262 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999946800591693e-05, + "loss": 1.1535, + "step": 1263 + }, + { + "epoch": 0.03, + "learning_rate": 1.999994417385357e-05, + "loss": 1.1277, + "step": 1264 + }, + { + "epoch": 0.03, + "learning_rate": 1.999994148382079e-05, + "loss": 1.1379, + "step": 1265 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999938730493376e-05, + "loss": 1.1293, + "step": 1266 + }, + { + "epoch": 0.03, + "learning_rate": 1.999993591387134e-05, + "loss": 1.1222, + "step": 1267 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999933033954696e-05, + "loss": 1.1449, + "step": 1268 + }, + { + "epoch": 0.03, + "learning_rate": 1.999993009074347e-05, + "loss": 1.1458, + "step": 1269 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999927084237676e-05, + "loss": 1.1325, + "step": 1270 + }, + { + "epoch": 0.03, + "learning_rate": 1.999992401443733e-05, + "loss": 1.1513, + "step": 1271 + }, + { + "epoch": 0.03, + "learning_rate": 1.999992088134246e-05, + "loss": 1.1349, + "step": 1272 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999917684953082e-05, + "loss": 1.1069, + "step": 1273 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999914425269215e-05, + "loss": 1.1114, + "step": 1274 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999911102290878e-05, + "loss": 1.1574, + "step": 1275 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999907716018096e-05, + "loss": 1.1082, + "step": 1276 + }, + { + "epoch": 0.03, + "learning_rate": 1.999990426645089e-05, + "loss": 1.1069, + "step": 1277 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999900753589277e-05, + "loss": 1.1216, + "step": 1278 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999897177433284e-05, + "loss": 1.1229, + "step": 1279 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999893537982933e-05, + "loss": 1.1299, + "step": 1280 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999889835238245e-05, + "loss": 1.0989, + "step": 1281 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999886069199247e-05, + "loss": 1.1569, + "step": 1282 + }, + { + "epoch": 0.03, + "learning_rate": 1.999988223986596e-05, + "loss": 1.1351, + "step": 1283 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999878347238407e-05, + "loss": 1.1273, + "step": 1284 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999874391316618e-05, + "loss": 1.1143, + "step": 1285 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999870372100614e-05, + "loss": 1.1223, + "step": 1286 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999866289590416e-05, + "loss": 1.1389, + "step": 1287 + }, + { + "epoch": 0.03, + "learning_rate": 1.999986214378606e-05, + "loss": 1.1411, + "step": 1288 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999857934687566e-05, + "loss": 1.117, + "step": 1289 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999853662294963e-05, + "loss": 1.15, + "step": 1290 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999849326608273e-05, + "loss": 1.0741, + "step": 1291 + }, + { + "epoch": 0.03, + "learning_rate": 1.999984492762753e-05, + "loss": 1.1364, + "step": 1292 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999840465352756e-05, + "loss": 1.1351, + "step": 1293 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999835939783987e-05, + "loss": 1.1254, + "step": 1294 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999831350921247e-05, + "loss": 1.1061, + "step": 1295 + }, + { + "epoch": 0.03, + "learning_rate": 1.999982669876456e-05, + "loss": 1.113, + "step": 1296 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999821983313964e-05, + "loss": 1.1177, + "step": 1297 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999817204569486e-05, + "loss": 1.1138, + "step": 1298 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999812362531152e-05, + "loss": 1.127, + "step": 1299 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999807457199e-05, + "loss": 1.1517, + "step": 1300 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999802488573056e-05, + "loss": 1.1249, + "step": 1301 + }, + { + "epoch": 0.03, + "learning_rate": 1.999979745665335e-05, + "loss": 1.1261, + "step": 1302 + }, + { + "epoch": 0.03, + "learning_rate": 1.999979236143992e-05, + "loss": 1.1359, + "step": 1303 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999787202932793e-05, + "loss": 1.1613, + "step": 1304 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999781981132004e-05, + "loss": 1.1379, + "step": 1305 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999776696037583e-05, + "loss": 1.1244, + "step": 1306 + }, + { + "epoch": 0.03, + "learning_rate": 1.999977134764957e-05, + "loss": 1.1256, + "step": 1307 + }, + { + "epoch": 0.03, + "learning_rate": 1.999976593596799e-05, + "loss": 1.1196, + "step": 1308 + }, + { + "epoch": 0.03, + "learning_rate": 1.999976046099289e-05, + "loss": 1.1597, + "step": 1309 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999754922724288e-05, + "loss": 1.105, + "step": 1310 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999749321162234e-05, + "loss": 1.0866, + "step": 1311 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999743656306754e-05, + "loss": 1.1316, + "step": 1312 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999737928157886e-05, + "loss": 1.1061, + "step": 1313 + }, + { + "epoch": 0.03, + "learning_rate": 1.999973213671567e-05, + "loss": 1.1456, + "step": 1314 + }, + { + "epoch": 0.03, + "learning_rate": 1.999972628198014e-05, + "loss": 1.1506, + "step": 1315 + }, + { + "epoch": 0.03, + "learning_rate": 1.999972036395133e-05, + "loss": 1.1153, + "step": 1316 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999714382629284e-05, + "loss": 1.1444, + "step": 1317 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999708338014033e-05, + "loss": 1.1124, + "step": 1318 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999702230105624e-05, + "loss": 1.1337, + "step": 1319 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999696058904087e-05, + "loss": 1.1491, + "step": 1320 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999689824409464e-05, + "loss": 1.1643, + "step": 1321 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999683526621797e-05, + "loss": 1.109, + "step": 1322 + }, + { + "epoch": 0.03, + "learning_rate": 1.999967716554112e-05, + "loss": 1.1238, + "step": 1323 + }, + { + "epoch": 0.03, + "learning_rate": 1.999967074116748e-05, + "loss": 1.1123, + "step": 1324 + }, + { + "epoch": 0.03, + "learning_rate": 1.999966425350091e-05, + "loss": 1.1429, + "step": 1325 + }, + { + "epoch": 0.03, + "learning_rate": 1.999965770254146e-05, + "loss": 1.1202, + "step": 1326 + }, + { + "epoch": 0.03, + "learning_rate": 1.999965108828917e-05, + "loss": 1.1481, + "step": 1327 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999644410744072e-05, + "loss": 1.1175, + "step": 1328 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999637669906217e-05, + "loss": 1.1444, + "step": 1329 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999630865775648e-05, + "loss": 1.1018, + "step": 1330 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999623998352405e-05, + "loss": 1.1103, + "step": 1331 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999617067636533e-05, + "loss": 1.1271, + "step": 1332 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999610073628073e-05, + "loss": 1.1313, + "step": 1333 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999603016327074e-05, + "loss": 1.1418, + "step": 1334 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999595895733578e-05, + "loss": 1.1004, + "step": 1335 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999588711847632e-05, + "loss": 1.1243, + "step": 1336 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999581464669277e-05, + "loss": 1.1152, + "step": 1337 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999574154198563e-05, + "loss": 1.0986, + "step": 1338 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999566780435532e-05, + "loss": 1.1722, + "step": 1339 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999559343380233e-05, + "loss": 1.1247, + "step": 1340 + }, + { + "epoch": 0.03, + "learning_rate": 1.999955184303272e-05, + "loss": 1.1553, + "step": 1341 + }, + { + "epoch": 0.03, + "learning_rate": 1.999954427939303e-05, + "loss": 1.1663, + "step": 1342 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999536652461215e-05, + "loss": 1.0996, + "step": 1343 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999528962237326e-05, + "loss": 1.1325, + "step": 1344 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999521208721405e-05, + "loss": 1.1378, + "step": 1345 + }, + { + "epoch": 0.03, + "learning_rate": 1.999951339191351e-05, + "loss": 1.1254, + "step": 1346 + }, + { + "epoch": 0.03, + "learning_rate": 1.999950551181368e-05, + "loss": 1.136, + "step": 1347 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999497568421973e-05, + "loss": 1.1771, + "step": 1348 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999489561738438e-05, + "loss": 1.156, + "step": 1349 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999481491763123e-05, + "loss": 1.1331, + "step": 1350 + }, + { + "epoch": 0.03, + "learning_rate": 1.999947335849608e-05, + "loss": 1.1261, + "step": 1351 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999465161937363e-05, + "loss": 1.1018, + "step": 1352 + }, + { + "epoch": 0.03, + "learning_rate": 1.999945690208702e-05, + "loss": 1.1072, + "step": 1353 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999448578945103e-05, + "loss": 1.1538, + "step": 1354 + }, + { + "epoch": 0.03, + "learning_rate": 1.999944019251167e-05, + "loss": 1.1448, + "step": 1355 + }, + { + "epoch": 0.03, + "learning_rate": 1.999943174278677e-05, + "loss": 1.1523, + "step": 1356 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999423229770457e-05, + "loss": 1.152, + "step": 1357 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999414653462787e-05, + "loss": 1.1409, + "step": 1358 + }, + { + "epoch": 0.03, + "learning_rate": 1.999940601386381e-05, + "loss": 1.1042, + "step": 1359 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999397310973583e-05, + "loss": 1.0896, + "step": 1360 + }, + { + "epoch": 0.03, + "learning_rate": 1.999938854479216e-05, + "loss": 1.1153, + "step": 1361 + }, + { + "epoch": 0.03, + "learning_rate": 1.99993797153196e-05, + "loss": 1.1464, + "step": 1362 + }, + { + "epoch": 0.03, + "learning_rate": 1.999937082255596e-05, + "loss": 1.1263, + "step": 1363 + }, + { + "epoch": 0.03, + "learning_rate": 1.999936186650129e-05, + "loss": 1.1093, + "step": 1364 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999352847155647e-05, + "loss": 1.1056, + "step": 1365 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999343764519095e-05, + "loss": 1.1183, + "step": 1366 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999334618591685e-05, + "loss": 1.159, + "step": 1367 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999325409373478e-05, + "loss": 1.129, + "step": 1368 + }, + { + "epoch": 0.03, + "learning_rate": 1.999931613686453e-05, + "loss": 1.1044, + "step": 1369 + }, + { + "epoch": 0.03, + "learning_rate": 1.99993068010649e-05, + "loss": 1.125, + "step": 1370 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999297401974652e-05, + "loss": 1.1283, + "step": 1371 + }, + { + "epoch": 0.03, + "learning_rate": 1.999928793959384e-05, + "loss": 1.119, + "step": 1372 + }, + { + "epoch": 0.03, + "learning_rate": 1.999927841392252e-05, + "loss": 1.169, + "step": 1373 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999268824960765e-05, + "loss": 1.122, + "step": 1374 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999259172708627e-05, + "loss": 1.0982, + "step": 1375 + }, + { + "epoch": 0.03, + "learning_rate": 1.999924945716617e-05, + "loss": 1.1725, + "step": 1376 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999239678333447e-05, + "loss": 1.1028, + "step": 1377 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999229836210535e-05, + "loss": 1.0839, + "step": 1378 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999219930797484e-05, + "loss": 1.1016, + "step": 1379 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999209962094365e-05, + "loss": 1.1124, + "step": 1380 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999199930101232e-05, + "loss": 1.1093, + "step": 1381 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999189834818156e-05, + "loss": 1.1208, + "step": 1382 + }, + { + "epoch": 0.03, + "learning_rate": 1.99991796762452e-05, + "loss": 1.1438, + "step": 1383 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999169454382424e-05, + "loss": 1.1339, + "step": 1384 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999159169229896e-05, + "loss": 1.0989, + "step": 1385 + }, + { + "epoch": 0.03, + "learning_rate": 1.999914882078768e-05, + "loss": 1.0875, + "step": 1386 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999138409055843e-05, + "loss": 1.1317, + "step": 1387 + }, + { + "epoch": 0.03, + "learning_rate": 1.999912793403445e-05, + "loss": 1.1619, + "step": 1388 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999117395723567e-05, + "loss": 1.1423, + "step": 1389 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999106794123264e-05, + "loss": 1.0812, + "step": 1390 + }, + { + "epoch": 0.03, + "learning_rate": 1.99990961292336e-05, + "loss": 1.1133, + "step": 1391 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999085401054653e-05, + "loss": 1.0931, + "step": 1392 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999074609586483e-05, + "loss": 1.1135, + "step": 1393 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999063754829157e-05, + "loss": 1.0826, + "step": 1394 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999052836782753e-05, + "loss": 1.1613, + "step": 1395 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999041855447334e-05, + "loss": 1.1809, + "step": 1396 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999030810822968e-05, + "loss": 1.1288, + "step": 1397 + }, + { + "epoch": 0.03, + "learning_rate": 1.9999019702909727e-05, + "loss": 1.1504, + "step": 1398 + }, + { + "epoch": 0.03, + "learning_rate": 1.999900853170768e-05, + "loss": 1.1457, + "step": 1399 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998997297216903e-05, + "loss": 1.1272, + "step": 1400 + }, + { + "epoch": 0.03, + "learning_rate": 1.999898599943746e-05, + "loss": 1.1156, + "step": 1401 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998974638369425e-05, + "loss": 1.1159, + "step": 1402 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998963214012877e-05, + "loss": 1.1392, + "step": 1403 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998951726367878e-05, + "loss": 1.1403, + "step": 1404 + }, + { + "epoch": 0.03, + "learning_rate": 1.99989401754345e-05, + "loss": 1.1443, + "step": 1405 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998928561212825e-05, + "loss": 1.1007, + "step": 1406 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998916883702924e-05, + "loss": 1.1527, + "step": 1407 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998905142904866e-05, + "loss": 1.102, + "step": 1408 + }, + { + "epoch": 0.03, + "learning_rate": 1.999889333881873e-05, + "loss": 1.1049, + "step": 1409 + }, + { + "epoch": 0.03, + "learning_rate": 1.999888147144459e-05, + "loss": 1.1416, + "step": 1410 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998869540782518e-05, + "loss": 1.1496, + "step": 1411 + }, + { + "epoch": 0.03, + "learning_rate": 1.999885754683259e-05, + "loss": 1.1238, + "step": 1412 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998845489594888e-05, + "loss": 1.1408, + "step": 1413 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998833369069483e-05, + "loss": 1.0954, + "step": 1414 + }, + { + "epoch": 0.03, + "learning_rate": 1.999882118525645e-05, + "loss": 1.1691, + "step": 1415 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998808938155872e-05, + "loss": 1.1034, + "step": 1416 + }, + { + "epoch": 0.03, + "learning_rate": 1.999879662776782e-05, + "loss": 1.0919, + "step": 1417 + }, + { + "epoch": 0.03, + "learning_rate": 1.999878425409238e-05, + "loss": 1.1154, + "step": 1418 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998771817129624e-05, + "loss": 1.1352, + "step": 1419 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998759316879627e-05, + "loss": 1.1256, + "step": 1420 + }, + { + "epoch": 0.03, + "learning_rate": 1.999874675334248e-05, + "loss": 1.1055, + "step": 1421 + }, + { + "epoch": 0.03, + "learning_rate": 1.9998734126518254e-05, + "loss": 1.1485, + "step": 1422 + }, + { + "epoch": 0.03, + "learning_rate": 1.999872143640703e-05, + "loss": 1.1108, + "step": 1423 + }, + { + "epoch": 0.03, + "learning_rate": 1.999870868300889e-05, + "loss": 1.0965, + "step": 1424 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998695866323914e-05, + "loss": 1.1071, + "step": 1425 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998682986352184e-05, + "loss": 1.1281, + "step": 1426 + }, + { + "epoch": 0.04, + "learning_rate": 1.999867004309378e-05, + "loss": 1.1183, + "step": 1427 + }, + { + "epoch": 0.04, + "learning_rate": 1.999865703654878e-05, + "loss": 1.1304, + "step": 1428 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998643966717276e-05, + "loss": 1.1687, + "step": 1429 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998630833599348e-05, + "loss": 1.0959, + "step": 1430 + }, + { + "epoch": 0.04, + "learning_rate": 1.999861763719507e-05, + "loss": 1.0939, + "step": 1431 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998604377504536e-05, + "loss": 1.0955, + "step": 1432 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998591054527825e-05, + "loss": 1.1033, + "step": 1433 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998577668265023e-05, + "loss": 1.1064, + "step": 1434 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998564218716213e-05, + "loss": 1.0844, + "step": 1435 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998550705881485e-05, + "loss": 1.1125, + "step": 1436 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998537129760914e-05, + "loss": 1.1109, + "step": 1437 + }, + { + "epoch": 0.04, + "learning_rate": 1.99985234903546e-05, + "loss": 1.1426, + "step": 1438 + }, + { + "epoch": 0.04, + "learning_rate": 1.999850978766262e-05, + "loss": 1.1231, + "step": 1439 + }, + { + "epoch": 0.04, + "learning_rate": 1.999849602168506e-05, + "loss": 1.1683, + "step": 1440 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998482192422012e-05, + "loss": 1.1314, + "step": 1441 + }, + { + "epoch": 0.04, + "learning_rate": 1.999846829987356e-05, + "loss": 1.1156, + "step": 1442 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998454344039798e-05, + "loss": 1.1231, + "step": 1443 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998440324920806e-05, + "loss": 1.1213, + "step": 1444 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998426242516676e-05, + "loss": 1.1111, + "step": 1445 + }, + { + "epoch": 0.04, + "learning_rate": 1.99984120968275e-05, + "loss": 1.119, + "step": 1446 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998397887853362e-05, + "loss": 1.0866, + "step": 1447 + }, + { + "epoch": 0.04, + "learning_rate": 1.999838361559436e-05, + "loss": 1.1259, + "step": 1448 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998369280050573e-05, + "loss": 1.1648, + "step": 1449 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998354881222103e-05, + "loss": 1.1395, + "step": 1450 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998340419109033e-05, + "loss": 1.1157, + "step": 1451 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998325893711462e-05, + "loss": 1.0919, + "step": 1452 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998311305029478e-05, + "loss": 1.1226, + "step": 1453 + }, + { + "epoch": 0.04, + "learning_rate": 1.999829665306317e-05, + "loss": 1.1244, + "step": 1454 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998281937812632e-05, + "loss": 1.1491, + "step": 1455 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998267159277964e-05, + "loss": 1.1393, + "step": 1456 + }, + { + "epoch": 0.04, + "learning_rate": 1.999825231745925e-05, + "loss": 1.0963, + "step": 1457 + }, + { + "epoch": 0.04, + "learning_rate": 1.999823741235659e-05, + "loss": 1.1053, + "step": 1458 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998222443970078e-05, + "loss": 1.0855, + "step": 1459 + }, + { + "epoch": 0.04, + "learning_rate": 1.999820741229981e-05, + "loss": 1.1462, + "step": 1460 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998192317345873e-05, + "loss": 1.127, + "step": 1461 + }, + { + "epoch": 0.04, + "learning_rate": 1.999817715910837e-05, + "loss": 1.1244, + "step": 1462 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998161937587398e-05, + "loss": 1.1059, + "step": 1463 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998146652783047e-05, + "loss": 1.132, + "step": 1464 + }, + { + "epoch": 0.04, + "learning_rate": 1.999813130469542e-05, + "loss": 1.0796, + "step": 1465 + }, + { + "epoch": 0.04, + "learning_rate": 1.999811589332461e-05, + "loss": 1.123, + "step": 1466 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998100418670718e-05, + "loss": 1.0832, + "step": 1467 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998084880733837e-05, + "loss": 1.1175, + "step": 1468 + }, + { + "epoch": 0.04, + "learning_rate": 1.999806927951407e-05, + "loss": 1.112, + "step": 1469 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998053615011513e-05, + "loss": 1.102, + "step": 1470 + }, + { + "epoch": 0.04, + "learning_rate": 1.999803788722627e-05, + "loss": 1.1323, + "step": 1471 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998022096158434e-05, + "loss": 1.125, + "step": 1472 + }, + { + "epoch": 0.04, + "learning_rate": 1.9998006241808107e-05, + "loss": 1.1336, + "step": 1473 + }, + { + "epoch": 0.04, + "learning_rate": 1.999799032417539e-05, + "loss": 1.1151, + "step": 1474 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997974343260386e-05, + "loss": 1.137, + "step": 1475 + }, + { + "epoch": 0.04, + "learning_rate": 1.999795829906319e-05, + "loss": 1.1164, + "step": 1476 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997942191583906e-05, + "loss": 1.1351, + "step": 1477 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997926020822643e-05, + "loss": 1.107, + "step": 1478 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997909786779497e-05, + "loss": 1.1229, + "step": 1479 + }, + { + "epoch": 0.04, + "learning_rate": 1.999789348945457e-05, + "loss": 1.1424, + "step": 1480 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997877128847966e-05, + "loss": 1.1298, + "step": 1481 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997860704959794e-05, + "loss": 1.0916, + "step": 1482 + }, + { + "epoch": 0.04, + "learning_rate": 1.999784421779015e-05, + "loss": 1.1303, + "step": 1483 + }, + { + "epoch": 0.04, + "learning_rate": 1.999782766733914e-05, + "loss": 1.1268, + "step": 1484 + }, + { + "epoch": 0.04, + "learning_rate": 1.999781105360687e-05, + "loss": 1.1114, + "step": 1485 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997794376593448e-05, + "loss": 1.1367, + "step": 1486 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997777636298977e-05, + "loss": 1.1414, + "step": 1487 + }, + { + "epoch": 0.04, + "learning_rate": 1.999776083272356e-05, + "loss": 1.1258, + "step": 1488 + }, + { + "epoch": 0.04, + "learning_rate": 1.999774396586731e-05, + "loss": 1.1515, + "step": 1489 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997727035730327e-05, + "loss": 1.098, + "step": 1490 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997710042312725e-05, + "loss": 1.1375, + "step": 1491 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997692985614603e-05, + "loss": 1.131, + "step": 1492 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997675865636076e-05, + "loss": 1.1218, + "step": 1493 + }, + { + "epoch": 0.04, + "learning_rate": 1.999765868237725e-05, + "loss": 1.1148, + "step": 1494 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997641435838237e-05, + "loss": 1.1222, + "step": 1495 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997624126019138e-05, + "loss": 1.1085, + "step": 1496 + }, + { + "epoch": 0.04, + "learning_rate": 1.999760675292007e-05, + "loss": 1.0829, + "step": 1497 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997589316541142e-05, + "loss": 1.1117, + "step": 1498 + }, + { + "epoch": 0.04, + "learning_rate": 1.999757181688246e-05, + "loss": 1.1351, + "step": 1499 + }, + { + "epoch": 0.04, + "learning_rate": 1.999755425394414e-05, + "loss": 1.0898, + "step": 1500 + }, + { + "epoch": 0.04, + "learning_rate": 1.999753662772629e-05, + "loss": 1.1351, + "step": 1501 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997518938229023e-05, + "loss": 1.106, + "step": 1502 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997501185452452e-05, + "loss": 1.1144, + "step": 1503 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997483369396686e-05, + "loss": 1.1094, + "step": 1504 + }, + { + "epoch": 0.04, + "learning_rate": 1.999746549006184e-05, + "loss": 1.0991, + "step": 1505 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997447547448027e-05, + "loss": 1.1293, + "step": 1506 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997429541555355e-05, + "loss": 1.103, + "step": 1507 + }, + { + "epoch": 0.04, + "learning_rate": 1.999741147238395e-05, + "loss": 1.1067, + "step": 1508 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997393339933915e-05, + "loss": 1.1307, + "step": 1509 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997375144205375e-05, + "loss": 1.1313, + "step": 1510 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997356885198435e-05, + "loss": 1.1506, + "step": 1511 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997338562913214e-05, + "loss": 1.1243, + "step": 1512 + }, + { + "epoch": 0.04, + "learning_rate": 1.999732017734983e-05, + "loss": 1.1363, + "step": 1513 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997301728508405e-05, + "loss": 1.1148, + "step": 1514 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997283216389043e-05, + "loss": 1.1194, + "step": 1515 + }, + { + "epoch": 0.04, + "learning_rate": 1.999726464099187e-05, + "loss": 1.1147, + "step": 1516 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997246002317e-05, + "loss": 1.1744, + "step": 1517 + }, + { + "epoch": 0.04, + "learning_rate": 1.999722730036455e-05, + "loss": 1.1215, + "step": 1518 + }, + { + "epoch": 0.04, + "learning_rate": 1.999720853513464e-05, + "loss": 1.1233, + "step": 1519 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997189706627393e-05, + "loss": 1.1017, + "step": 1520 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997170814842924e-05, + "loss": 1.1131, + "step": 1521 + }, + { + "epoch": 0.04, + "learning_rate": 1.999715185978135e-05, + "loss": 1.1301, + "step": 1522 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997132841442797e-05, + "loss": 1.1146, + "step": 1523 + }, + { + "epoch": 0.04, + "learning_rate": 1.999711375982738e-05, + "loss": 1.1135, + "step": 1524 + }, + { + "epoch": 0.04, + "learning_rate": 1.999709461493522e-05, + "loss": 1.1481, + "step": 1525 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997075406766445e-05, + "loss": 1.1212, + "step": 1526 + }, + { + "epoch": 0.04, + "learning_rate": 1.999705613532117e-05, + "loss": 1.153, + "step": 1527 + }, + { + "epoch": 0.04, + "learning_rate": 1.9997036800599516e-05, + "loss": 1.0977, + "step": 1528 + }, + { + "epoch": 0.04, + "learning_rate": 1.999701740260161e-05, + "loss": 1.0977, + "step": 1529 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996997941327572e-05, + "loss": 1.1076, + "step": 1530 + }, + { + "epoch": 0.04, + "learning_rate": 1.999697841677753e-05, + "loss": 1.1386, + "step": 1531 + }, + { + "epoch": 0.04, + "learning_rate": 1.99969588289516e-05, + "loss": 1.1112, + "step": 1532 + }, + { + "epoch": 0.04, + "learning_rate": 1.999693917784991e-05, + "loss": 1.1191, + "step": 1533 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996919463472583e-05, + "loss": 1.1179, + "step": 1534 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996899685819747e-05, + "loss": 1.0879, + "step": 1535 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996879844891523e-05, + "loss": 1.0946, + "step": 1536 + }, + { + "epoch": 0.04, + "learning_rate": 1.999685994068804e-05, + "loss": 1.1198, + "step": 1537 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996839973209423e-05, + "loss": 1.1479, + "step": 1538 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996819942455796e-05, + "loss": 1.1105, + "step": 1539 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996799848427287e-05, + "loss": 1.1211, + "step": 1540 + }, + { + "epoch": 0.04, + "learning_rate": 1.999677969112403e-05, + "loss": 1.1041, + "step": 1541 + }, + { + "epoch": 0.04, + "learning_rate": 1.999675947054614e-05, + "loss": 1.1136, + "step": 1542 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996739186693754e-05, + "loss": 1.1497, + "step": 1543 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996718839566996e-05, + "loss": 1.1278, + "step": 1544 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996698429166003e-05, + "loss": 1.1002, + "step": 1545 + }, + { + "epoch": 0.04, + "learning_rate": 1.999667795549089e-05, + "loss": 1.1388, + "step": 1546 + }, + { + "epoch": 0.04, + "learning_rate": 1.99966574185418e-05, + "loss": 1.0842, + "step": 1547 + }, + { + "epoch": 0.04, + "learning_rate": 1.999663681831885e-05, + "loss": 1.1011, + "step": 1548 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996616154822184e-05, + "loss": 1.1476, + "step": 1549 + }, + { + "epoch": 0.04, + "learning_rate": 1.999659542805193e-05, + "loss": 1.1558, + "step": 1550 + }, + { + "epoch": 0.04, + "learning_rate": 1.999657463800821e-05, + "loss": 1.1497, + "step": 1551 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996553784691162e-05, + "loss": 1.1228, + "step": 1552 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996532868100918e-05, + "loss": 1.1041, + "step": 1553 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996511888237608e-05, + "loss": 1.1518, + "step": 1554 + }, + { + "epoch": 0.04, + "learning_rate": 1.999649084510137e-05, + "loss": 1.1272, + "step": 1555 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996469738692333e-05, + "loss": 1.1224, + "step": 1556 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996448569010627e-05, + "loss": 1.1011, + "step": 1557 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996427336056393e-05, + "loss": 1.1182, + "step": 1558 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996406039829766e-05, + "loss": 1.1163, + "step": 1559 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996384680330873e-05, + "loss": 1.0853, + "step": 1560 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996363257559855e-05, + "loss": 1.1551, + "step": 1561 + }, + { + "epoch": 0.04, + "learning_rate": 1.999634177151685e-05, + "loss": 1.0988, + "step": 1562 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996320222201985e-05, + "loss": 1.137, + "step": 1563 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996298609615404e-05, + "loss": 1.0593, + "step": 1564 + }, + { + "epoch": 0.04, + "learning_rate": 1.999627693375724e-05, + "loss": 1.1162, + "step": 1565 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996255194627633e-05, + "loss": 1.1104, + "step": 1566 + }, + { + "epoch": 0.04, + "learning_rate": 1.999623339222672e-05, + "loss": 1.0996, + "step": 1567 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996211526554637e-05, + "loss": 1.1161, + "step": 1568 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996189597611523e-05, + "loss": 1.1193, + "step": 1569 + }, + { + "epoch": 0.04, + "learning_rate": 1.999616760539752e-05, + "loss": 1.1003, + "step": 1570 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996145549912766e-05, + "loss": 1.1237, + "step": 1571 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996123431157395e-05, + "loss": 1.103, + "step": 1572 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996101249131552e-05, + "loss": 1.1167, + "step": 1573 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996079003835378e-05, + "loss": 1.1021, + "step": 1574 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996056695269013e-05, + "loss": 1.0914, + "step": 1575 + }, + { + "epoch": 0.04, + "learning_rate": 1.9996034323432595e-05, + "loss": 1.0888, + "step": 1576 + }, + { + "epoch": 0.04, + "learning_rate": 1.999601188832627e-05, + "loss": 1.1032, + "step": 1577 + }, + { + "epoch": 0.04, + "learning_rate": 1.999598938995018e-05, + "loss": 1.0864, + "step": 1578 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995966828304466e-05, + "loss": 1.1259, + "step": 1579 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995944203389266e-05, + "loss": 1.1047, + "step": 1580 + }, + { + "epoch": 0.04, + "learning_rate": 1.999592151520473e-05, + "loss": 1.1341, + "step": 1581 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995898763751e-05, + "loss": 1.1209, + "step": 1582 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995875949028217e-05, + "loss": 1.1297, + "step": 1583 + }, + { + "epoch": 0.04, + "learning_rate": 1.999585307103653e-05, + "loss": 1.1231, + "step": 1584 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995830129776084e-05, + "loss": 1.1258, + "step": 1585 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995807125247016e-05, + "loss": 1.1104, + "step": 1586 + }, + { + "epoch": 0.04, + "learning_rate": 1.999578405744948e-05, + "loss": 1.1396, + "step": 1587 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995760926383624e-05, + "loss": 1.164, + "step": 1588 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995737732049584e-05, + "loss": 1.1278, + "step": 1589 + }, + { + "epoch": 0.04, + "learning_rate": 1.999571447444751e-05, + "loss": 1.1174, + "step": 1590 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995691153577557e-05, + "loss": 1.1135, + "step": 1591 + }, + { + "epoch": 0.04, + "learning_rate": 1.999566776943987e-05, + "loss": 1.1262, + "step": 1592 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995644322034587e-05, + "loss": 1.1407, + "step": 1593 + }, + { + "epoch": 0.04, + "learning_rate": 1.999562081136187e-05, + "loss": 1.101, + "step": 1594 + }, + { + "epoch": 0.04, + "learning_rate": 1.999559723742186e-05, + "loss": 1.1272, + "step": 1595 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995573600214704e-05, + "loss": 1.1448, + "step": 1596 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995549899740558e-05, + "loss": 1.1206, + "step": 1597 + }, + { + "epoch": 0.04, + "learning_rate": 1.999552613599957e-05, + "loss": 1.1115, + "step": 1598 + }, + { + "epoch": 0.04, + "learning_rate": 1.999550230899189e-05, + "loss": 1.1271, + "step": 1599 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995478418717668e-05, + "loss": 1.116, + "step": 1600 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995454465177057e-05, + "loss": 1.148, + "step": 1601 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995430448370206e-05, + "loss": 1.1235, + "step": 1602 + }, + { + "epoch": 0.04, + "learning_rate": 1.999540636829727e-05, + "loss": 1.1008, + "step": 1603 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995382224958397e-05, + "loss": 1.1426, + "step": 1604 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995358018353744e-05, + "loss": 1.1208, + "step": 1605 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995333748483464e-05, + "loss": 1.1435, + "step": 1606 + }, + { + "epoch": 0.04, + "learning_rate": 1.999530941534771e-05, + "loss": 1.1095, + "step": 1607 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995285018946637e-05, + "loss": 1.1204, + "step": 1608 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995260559280394e-05, + "loss": 1.1163, + "step": 1609 + }, + { + "epoch": 0.04, + "learning_rate": 1.999523603634914e-05, + "loss": 1.137, + "step": 1610 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995211450153032e-05, + "loss": 1.1073, + "step": 1611 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995186800692225e-05, + "loss": 1.1026, + "step": 1612 + }, + { + "epoch": 0.04, + "learning_rate": 1.999516208796687e-05, + "loss": 1.0976, + "step": 1613 + }, + { + "epoch": 0.04, + "learning_rate": 1.999513731197713e-05, + "loss": 1.1138, + "step": 1614 + }, + { + "epoch": 0.04, + "learning_rate": 1.999511247272316e-05, + "loss": 1.0957, + "step": 1615 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995087570205113e-05, + "loss": 1.1153, + "step": 1616 + }, + { + "epoch": 0.04, + "learning_rate": 1.9995062604423152e-05, + "loss": 1.1061, + "step": 1617 + }, + { + "epoch": 0.04, + "learning_rate": 1.999503757537743e-05, + "loss": 1.1458, + "step": 1618 + }, + { + "epoch": 0.04, + "learning_rate": 1.999501248306811e-05, + "loss": 1.122, + "step": 1619 + }, + { + "epoch": 0.04, + "learning_rate": 1.999498732749535e-05, + "loss": 1.0967, + "step": 1620 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994962108659308e-05, + "loss": 1.131, + "step": 1621 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994936826560143e-05, + "loss": 1.1507, + "step": 1622 + }, + { + "epoch": 0.04, + "learning_rate": 1.999491148119802e-05, + "loss": 1.1019, + "step": 1623 + }, + { + "epoch": 0.04, + "learning_rate": 1.999488607257309e-05, + "loss": 1.111, + "step": 1624 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994860600685518e-05, + "loss": 1.0897, + "step": 1625 + }, + { + "epoch": 0.04, + "learning_rate": 1.999483506553547e-05, + "loss": 1.0977, + "step": 1626 + }, + { + "epoch": 0.04, + "learning_rate": 1.99948094671231e-05, + "loss": 1.1149, + "step": 1627 + }, + { + "epoch": 0.04, + "learning_rate": 1.999478380544858e-05, + "loss": 1.0853, + "step": 1628 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994758080512066e-05, + "loss": 1.0808, + "step": 1629 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994732292313717e-05, + "loss": 1.1085, + "step": 1630 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994706440853703e-05, + "loss": 1.1047, + "step": 1631 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994680526132183e-05, + "loss": 1.1268, + "step": 1632 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994654548149323e-05, + "loss": 1.1136, + "step": 1633 + }, + { + "epoch": 0.04, + "learning_rate": 1.999462850690529e-05, + "loss": 1.1197, + "step": 1634 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994602402400246e-05, + "loss": 1.115, + "step": 1635 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994576234634358e-05, + "loss": 1.119, + "step": 1636 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994550003607788e-05, + "loss": 1.1462, + "step": 1637 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994523709320704e-05, + "loss": 1.1155, + "step": 1638 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994497351773276e-05, + "loss": 1.1353, + "step": 1639 + }, + { + "epoch": 0.04, + "learning_rate": 1.999447093096566e-05, + "loss": 1.1008, + "step": 1640 + }, + { + "epoch": 0.04, + "learning_rate": 1.999444444689804e-05, + "loss": 1.0825, + "step": 1641 + }, + { + "epoch": 0.04, + "learning_rate": 1.999441789957057e-05, + "loss": 1.1201, + "step": 1642 + }, + { + "epoch": 0.04, + "learning_rate": 1.999439128898342e-05, + "loss": 1.0956, + "step": 1643 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994364615136764e-05, + "loss": 1.1115, + "step": 1644 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994337878030766e-05, + "loss": 1.0962, + "step": 1645 + }, + { + "epoch": 0.04, + "learning_rate": 1.99943110776656e-05, + "loss": 1.1239, + "step": 1646 + }, + { + "epoch": 0.04, + "learning_rate": 1.999428421404143e-05, + "loss": 1.0891, + "step": 1647 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994257287158428e-05, + "loss": 1.1505, + "step": 1648 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994230297016766e-05, + "loss": 1.1098, + "step": 1649 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994203243616617e-05, + "loss": 1.1332, + "step": 1650 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994176126958143e-05, + "loss": 1.1238, + "step": 1651 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994148947041528e-05, + "loss": 1.0964, + "step": 1652 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994121703866933e-05, + "loss": 1.1261, + "step": 1653 + }, + { + "epoch": 0.04, + "learning_rate": 1.999409439743454e-05, + "loss": 1.1425, + "step": 1654 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994067027744512e-05, + "loss": 1.0956, + "step": 1655 + }, + { + "epoch": 0.04, + "learning_rate": 1.999403959479703e-05, + "loss": 1.0634, + "step": 1656 + }, + { + "epoch": 0.04, + "learning_rate": 1.9994012098592265e-05, + "loss": 1.1158, + "step": 1657 + }, + { + "epoch": 0.04, + "learning_rate": 1.999398453913039e-05, + "loss": 1.1226, + "step": 1658 + }, + { + "epoch": 0.04, + "learning_rate": 1.999395691641158e-05, + "loss": 1.1298, + "step": 1659 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993929230436012e-05, + "loss": 1.17, + "step": 1660 + }, + { + "epoch": 0.04, + "learning_rate": 1.999390148120386e-05, + "loss": 1.1017, + "step": 1661 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993873668715298e-05, + "loss": 1.1055, + "step": 1662 + }, + { + "epoch": 0.04, + "learning_rate": 1.99938457929705e-05, + "loss": 1.0933, + "step": 1663 + }, + { + "epoch": 0.04, + "learning_rate": 1.999381785396965e-05, + "loss": 1.1049, + "step": 1664 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993789851712918e-05, + "loss": 1.1106, + "step": 1665 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993761786200486e-05, + "loss": 1.1253, + "step": 1666 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993733657432525e-05, + "loss": 1.1068, + "step": 1667 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993705465409224e-05, + "loss": 1.0943, + "step": 1668 + }, + { + "epoch": 0.04, + "learning_rate": 1.999367721013075e-05, + "loss": 1.1126, + "step": 1669 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993648891597284e-05, + "loss": 1.1158, + "step": 1670 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993620509809015e-05, + "loss": 1.1403, + "step": 1671 + }, + { + "epoch": 0.04, + "learning_rate": 1.999359206476611e-05, + "loss": 1.1258, + "step": 1672 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993563556468757e-05, + "loss": 1.1212, + "step": 1673 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993534984917135e-05, + "loss": 1.1004, + "step": 1674 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993506350111423e-05, + "loss": 1.1163, + "step": 1675 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993477652051802e-05, + "loss": 1.095, + "step": 1676 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993448890738453e-05, + "loss": 1.1579, + "step": 1677 + }, + { + "epoch": 0.04, + "learning_rate": 1.999342006617156e-05, + "loss": 1.136, + "step": 1678 + }, + { + "epoch": 0.04, + "learning_rate": 1.999339117835131e-05, + "loss": 1.1375, + "step": 1679 + }, + { + "epoch": 0.04, + "learning_rate": 1.999336222727788e-05, + "loss": 1.1179, + "step": 1680 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993333212951453e-05, + "loss": 1.1131, + "step": 1681 + }, + { + "epoch": 0.04, + "learning_rate": 1.999330413537221e-05, + "loss": 1.0877, + "step": 1682 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993274994540346e-05, + "loss": 1.1307, + "step": 1683 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993245790456035e-05, + "loss": 1.1234, + "step": 1684 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993216523119465e-05, + "loss": 1.1456, + "step": 1685 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993187192530824e-05, + "loss": 1.1201, + "step": 1686 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993157798690293e-05, + "loss": 1.0998, + "step": 1687 + }, + { + "epoch": 0.04, + "learning_rate": 1.999312834159806e-05, + "loss": 1.1268, + "step": 1688 + }, + { + "epoch": 0.04, + "learning_rate": 1.999309882125431e-05, + "loss": 1.1488, + "step": 1689 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993069237659236e-05, + "loss": 1.095, + "step": 1690 + }, + { + "epoch": 0.04, + "learning_rate": 1.999303959081302e-05, + "loss": 1.1183, + "step": 1691 + }, + { + "epoch": 0.04, + "learning_rate": 1.9993009880715847e-05, + "loss": 1.0845, + "step": 1692 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992980107367913e-05, + "loss": 1.1212, + "step": 1693 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992950270769396e-05, + "loss": 1.1294, + "step": 1694 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992920370920496e-05, + "loss": 1.1302, + "step": 1695 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992890407821396e-05, + "loss": 1.083, + "step": 1696 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992860381472282e-05, + "loss": 1.1398, + "step": 1697 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992830291873355e-05, + "loss": 1.1205, + "step": 1698 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992800139024798e-05, + "loss": 1.1158, + "step": 1699 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992769922926802e-05, + "loss": 1.1123, + "step": 1700 + }, + { + "epoch": 0.04, + "learning_rate": 1.999273964357956e-05, + "loss": 1.1267, + "step": 1701 + }, + { + "epoch": 0.04, + "learning_rate": 1.999270930098326e-05, + "loss": 1.0859, + "step": 1702 + }, + { + "epoch": 0.04, + "learning_rate": 1.99926788951381e-05, + "loss": 1.1038, + "step": 1703 + }, + { + "epoch": 0.04, + "learning_rate": 1.999264842604427e-05, + "loss": 1.1206, + "step": 1704 + }, + { + "epoch": 0.04, + "learning_rate": 1.999261789370196e-05, + "loss": 1.0794, + "step": 1705 + }, + { + "epoch": 0.04, + "learning_rate": 1.999258729811137e-05, + "loss": 1.1221, + "step": 1706 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992556639272684e-05, + "loss": 1.0941, + "step": 1707 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992525917186104e-05, + "loss": 1.1029, + "step": 1708 + }, + { + "epoch": 0.04, + "learning_rate": 1.999249513185182e-05, + "loss": 1.1281, + "step": 1709 + }, + { + "epoch": 0.04, + "learning_rate": 1.999246428327003e-05, + "loss": 1.0922, + "step": 1710 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992433371440927e-05, + "loss": 1.0949, + "step": 1711 + }, + { + "epoch": 0.04, + "learning_rate": 1.999240239636471e-05, + "loss": 1.1298, + "step": 1712 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992371358041568e-05, + "loss": 1.116, + "step": 1713 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992340256471708e-05, + "loss": 1.0821, + "step": 1714 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992309091655317e-05, + "loss": 1.0697, + "step": 1715 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992277863592596e-05, + "loss": 1.1276, + "step": 1716 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992246572283744e-05, + "loss": 1.1004, + "step": 1717 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992215217728958e-05, + "loss": 1.0966, + "step": 1718 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992183799928434e-05, + "loss": 1.1251, + "step": 1719 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992152318882375e-05, + "loss": 1.084, + "step": 1720 + }, + { + "epoch": 0.04, + "learning_rate": 1.999212077459098e-05, + "loss": 1.1216, + "step": 1721 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992089167054444e-05, + "loss": 1.0979, + "step": 1722 + }, + { + "epoch": 0.04, + "learning_rate": 1.9992057496272972e-05, + "loss": 1.1128, + "step": 1723 + }, + { + "epoch": 0.04, + "learning_rate": 1.999202576224676e-05, + "loss": 1.0901, + "step": 1724 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991993964976016e-05, + "loss": 1.1206, + "step": 1725 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991962104460935e-05, + "loss": 1.0966, + "step": 1726 + }, + { + "epoch": 0.04, + "learning_rate": 1.999193018070172e-05, + "loss": 1.0737, + "step": 1727 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991898193698575e-05, + "loss": 1.086, + "step": 1728 + }, + { + "epoch": 0.04, + "learning_rate": 1.99918661434517e-05, + "loss": 1.1398, + "step": 1729 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991834029961294e-05, + "loss": 1.0704, + "step": 1730 + }, + { + "epoch": 0.04, + "learning_rate": 1.999180185322757e-05, + "loss": 1.1071, + "step": 1731 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991769613250728e-05, + "loss": 1.1163, + "step": 1732 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991737310030967e-05, + "loss": 1.081, + "step": 1733 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991704943568497e-05, + "loss": 1.1187, + "step": 1734 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991672513863522e-05, + "loss": 1.0977, + "step": 1735 + }, + { + "epoch": 0.04, + "learning_rate": 1.999164002091624e-05, + "loss": 1.1055, + "step": 1736 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991607464726872e-05, + "loss": 1.1181, + "step": 1737 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991574845295614e-05, + "loss": 1.0948, + "step": 1738 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991542162622667e-05, + "loss": 1.102, + "step": 1739 + }, + { + "epoch": 0.04, + "learning_rate": 1.999150941670825e-05, + "loss": 1.1348, + "step": 1740 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991476607552563e-05, + "loss": 1.1089, + "step": 1741 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991443735155816e-05, + "loss": 1.088, + "step": 1742 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991410799518214e-05, + "loss": 1.1065, + "step": 1743 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991377800639974e-05, + "loss": 1.0847, + "step": 1744 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991344738521294e-05, + "loss": 1.1055, + "step": 1745 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991311613162386e-05, + "loss": 1.1342, + "step": 1746 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991278424563467e-05, + "loss": 1.1069, + "step": 1747 + }, + { + "epoch": 0.04, + "learning_rate": 1.999124517272474e-05, + "loss": 1.0808, + "step": 1748 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991211857646412e-05, + "loss": 1.1169, + "step": 1749 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991178479328707e-05, + "loss": 1.1381, + "step": 1750 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991145037771824e-05, + "loss": 1.136, + "step": 1751 + }, + { + "epoch": 0.04, + "learning_rate": 1.999111153297598e-05, + "loss": 1.125, + "step": 1752 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991077964941387e-05, + "loss": 1.1061, + "step": 1753 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991044333668253e-05, + "loss": 1.0963, + "step": 1754 + }, + { + "epoch": 0.04, + "learning_rate": 1.9991010639156798e-05, + "loss": 1.1175, + "step": 1755 + }, + { + "epoch": 0.04, + "learning_rate": 1.999097688140723e-05, + "loss": 1.0838, + "step": 1756 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990943060419764e-05, + "loss": 1.1306, + "step": 1757 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990909176194617e-05, + "loss": 1.1132, + "step": 1758 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990875228731998e-05, + "loss": 1.0972, + "step": 1759 + }, + { + "epoch": 0.04, + "learning_rate": 1.999084121803213e-05, + "loss": 1.1208, + "step": 1760 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990807144095216e-05, + "loss": 1.1476, + "step": 1761 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990773006921484e-05, + "loss": 1.0666, + "step": 1762 + }, + { + "epoch": 0.04, + "learning_rate": 1.999073880651114e-05, + "loss": 1.1039, + "step": 1763 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990704542864408e-05, + "loss": 1.1072, + "step": 1764 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990670215981502e-05, + "loss": 1.1345, + "step": 1765 + }, + { + "epoch": 0.04, + "learning_rate": 1.999063582586264e-05, + "loss": 1.1454, + "step": 1766 + }, + { + "epoch": 0.04, + "learning_rate": 1.999060137250804e-05, + "loss": 1.1226, + "step": 1767 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990566855917918e-05, + "loss": 1.0954, + "step": 1768 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990532276092495e-05, + "loss": 1.0751, + "step": 1769 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990497633031984e-05, + "loss": 1.1064, + "step": 1770 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990462926736614e-05, + "loss": 1.0868, + "step": 1771 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990428157206595e-05, + "loss": 1.1089, + "step": 1772 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990393324442155e-05, + "loss": 1.1098, + "step": 1773 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990358428443508e-05, + "loss": 1.1341, + "step": 1774 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990323469210883e-05, + "loss": 1.1206, + "step": 1775 + }, + { + "epoch": 0.04, + "learning_rate": 1.999028844674449e-05, + "loss": 1.0997, + "step": 1776 + }, + { + "epoch": 0.04, + "learning_rate": 1.999025336104456e-05, + "loss": 1.1211, + "step": 1777 + }, + { + "epoch": 0.04, + "learning_rate": 1.999021821211131e-05, + "loss": 1.1181, + "step": 1778 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990182999944967e-05, + "loss": 1.1055, + "step": 1779 + }, + { + "epoch": 0.04, + "learning_rate": 1.999014772454575e-05, + "loss": 1.101, + "step": 1780 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990112385913886e-05, + "loss": 1.142, + "step": 1781 + }, + { + "epoch": 0.04, + "learning_rate": 1.999007698404959e-05, + "loss": 1.1135, + "step": 1782 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990041518953098e-05, + "loss": 1.1046, + "step": 1783 + }, + { + "epoch": 0.04, + "learning_rate": 1.9990005990624626e-05, + "loss": 1.1176, + "step": 1784 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989970399064402e-05, + "loss": 1.0963, + "step": 1785 + }, + { + "epoch": 0.04, + "learning_rate": 1.998993474427265e-05, + "loss": 1.1122, + "step": 1786 + }, + { + "epoch": 0.04, + "learning_rate": 1.99898990262496e-05, + "loss": 1.138, + "step": 1787 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989863244995473e-05, + "loss": 1.1124, + "step": 1788 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989827400510496e-05, + "loss": 1.1315, + "step": 1789 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989791492794897e-05, + "loss": 1.1306, + "step": 1790 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989755521848904e-05, + "loss": 1.1381, + "step": 1791 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989719487672743e-05, + "loss": 1.1186, + "step": 1792 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989683390266645e-05, + "loss": 1.0694, + "step": 1793 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989647229630836e-05, + "loss": 1.1237, + "step": 1794 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989611005765546e-05, + "loss": 1.1005, + "step": 1795 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989574718671e-05, + "loss": 1.1068, + "step": 1796 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989538368347435e-05, + "loss": 1.0933, + "step": 1797 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989501954795076e-05, + "loss": 1.1165, + "step": 1798 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989465478014157e-05, + "loss": 1.1228, + "step": 1799 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989428938004908e-05, + "loss": 1.1314, + "step": 1800 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989392334767553e-05, + "loss": 1.1163, + "step": 1801 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989355668302333e-05, + "loss": 1.13, + "step": 1802 + }, + { + "epoch": 0.04, + "learning_rate": 1.998931893860948e-05, + "loss": 1.0835, + "step": 1803 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989282145689216e-05, + "loss": 1.0985, + "step": 1804 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989245289541785e-05, + "loss": 1.0832, + "step": 1805 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989208370167418e-05, + "loss": 1.0941, + "step": 1806 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989171387566348e-05, + "loss": 1.0808, + "step": 1807 + }, + { + "epoch": 0.04, + "learning_rate": 1.9989134341738805e-05, + "loss": 1.0926, + "step": 1808 + }, + { + "epoch": 0.04, + "learning_rate": 1.998909723268503e-05, + "loss": 1.1209, + "step": 1809 + }, + { + "epoch": 0.04, + "learning_rate": 1.998906006040525e-05, + "loss": 1.1274, + "step": 1810 + }, + { + "epoch": 0.04, + "learning_rate": 1.998902282489971e-05, + "loss": 1.1204, + "step": 1811 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988985526168637e-05, + "loss": 1.0778, + "step": 1812 + }, + { + "epoch": 0.04, + "learning_rate": 1.998894816421227e-05, + "loss": 1.1177, + "step": 1813 + }, + { + "epoch": 0.04, + "learning_rate": 1.998891073903085e-05, + "loss": 1.1167, + "step": 1814 + }, + { + "epoch": 0.04, + "learning_rate": 1.998887325062461e-05, + "loss": 1.0964, + "step": 1815 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988835698993784e-05, + "loss": 1.1613, + "step": 1816 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988798084138614e-05, + "loss": 1.1183, + "step": 1817 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988760406059336e-05, + "loss": 1.1151, + "step": 1818 + }, + { + "epoch": 0.04, + "learning_rate": 1.998872266475619e-05, + "loss": 1.0992, + "step": 1819 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988684860229416e-05, + "loss": 1.0899, + "step": 1820 + }, + { + "epoch": 0.04, + "learning_rate": 1.998864699247925e-05, + "loss": 1.1449, + "step": 1821 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988609061505937e-05, + "loss": 1.0658, + "step": 1822 + }, + { + "epoch": 0.04, + "learning_rate": 1.998857106730971e-05, + "loss": 1.0655, + "step": 1823 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988533009890814e-05, + "loss": 1.1183, + "step": 1824 + }, + { + "epoch": 0.04, + "learning_rate": 1.998849488924949e-05, + "loss": 1.1303, + "step": 1825 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988456705385975e-05, + "loss": 1.0831, + "step": 1826 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988418458300514e-05, + "loss": 1.0883, + "step": 1827 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988380147993355e-05, + "loss": 1.1421, + "step": 1828 + }, + { + "epoch": 0.04, + "learning_rate": 1.998834177446473e-05, + "loss": 1.0798, + "step": 1829 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988303337714887e-05, + "loss": 1.0983, + "step": 1830 + }, + { + "epoch": 0.04, + "learning_rate": 1.9988264837744067e-05, + "loss": 1.0886, + "step": 1831 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988226274552517e-05, + "loss": 1.1131, + "step": 1832 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988187648140477e-05, + "loss": 1.0559, + "step": 1833 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988148958508196e-05, + "loss": 1.0904, + "step": 1834 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988110205655917e-05, + "loss": 1.1161, + "step": 1835 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988071389583883e-05, + "loss": 1.1037, + "step": 1836 + }, + { + "epoch": 0.05, + "learning_rate": 1.9988032510292343e-05, + "loss": 1.1191, + "step": 1837 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987993567781543e-05, + "loss": 1.0885, + "step": 1838 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987954562051724e-05, + "loss": 1.1268, + "step": 1839 + }, + { + "epoch": 0.05, + "learning_rate": 1.998791549310314e-05, + "loss": 1.0932, + "step": 1840 + }, + { + "epoch": 0.05, + "learning_rate": 1.998787636093604e-05, + "loss": 1.0806, + "step": 1841 + }, + { + "epoch": 0.05, + "learning_rate": 1.998783716555066e-05, + "loss": 1.0691, + "step": 1842 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987797906947257e-05, + "loss": 1.0368, + "step": 1843 + }, + { + "epoch": 0.05, + "learning_rate": 1.998775858512608e-05, + "loss": 1.1022, + "step": 1844 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987719200087373e-05, + "loss": 1.1244, + "step": 1845 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987679751831388e-05, + "loss": 1.1167, + "step": 1846 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987640240358376e-05, + "loss": 1.1012, + "step": 1847 + }, + { + "epoch": 0.05, + "learning_rate": 1.998760066566859e-05, + "loss": 1.1123, + "step": 1848 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987561027762265e-05, + "loss": 1.0833, + "step": 1849 + }, + { + "epoch": 0.05, + "learning_rate": 1.998752132663967e-05, + "loss": 1.1101, + "step": 1850 + }, + { + "epoch": 0.05, + "learning_rate": 1.998748156230105e-05, + "loss": 1.1208, + "step": 1851 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987441734746655e-05, + "loss": 1.0724, + "step": 1852 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987401843976738e-05, + "loss": 1.1323, + "step": 1853 + }, + { + "epoch": 0.05, + "learning_rate": 1.998736188999155e-05, + "loss": 1.0819, + "step": 1854 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987321872791346e-05, + "loss": 1.1358, + "step": 1855 + }, + { + "epoch": 0.05, + "learning_rate": 1.998728179237638e-05, + "loss": 1.0816, + "step": 1856 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987241648746904e-05, + "loss": 1.0966, + "step": 1857 + }, + { + "epoch": 0.05, + "learning_rate": 1.9987201441903174e-05, + "loss": 1.102, + "step": 1858 + }, + { + "epoch": 0.05, + "learning_rate": 1.998716117184544e-05, + "loss": 1.0781, + "step": 1859 + }, + { + "epoch": 0.05, + "learning_rate": 1.998712083857396e-05, + "loss": 1.0952, + "step": 1860 + }, + { + "epoch": 0.05, + "learning_rate": 1.998708044208899e-05, + "loss": 1.1216, + "step": 1861 + }, + { + "epoch": 0.05, + "learning_rate": 1.998703998239079e-05, + "loss": 1.0923, + "step": 1862 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986999459479603e-05, + "loss": 1.1399, + "step": 1863 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986958873355698e-05, + "loss": 1.1121, + "step": 1864 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986918224019326e-05, + "loss": 1.0732, + "step": 1865 + }, + { + "epoch": 0.05, + "learning_rate": 1.998687751147075e-05, + "loss": 1.0991, + "step": 1866 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986836735710214e-05, + "loss": 1.1092, + "step": 1867 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986795896737995e-05, + "loss": 1.0838, + "step": 1868 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986754994554337e-05, + "loss": 1.0764, + "step": 1869 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986714029159503e-05, + "loss": 1.0616, + "step": 1870 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986673000553755e-05, + "loss": 1.0838, + "step": 1871 + }, + { + "epoch": 0.05, + "learning_rate": 1.998663190873735e-05, + "loss": 1.1086, + "step": 1872 + }, + { + "epoch": 0.05, + "learning_rate": 1.998659075371055e-05, + "loss": 1.108, + "step": 1873 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986549535473612e-05, + "loss": 1.1215, + "step": 1874 + }, + { + "epoch": 0.05, + "learning_rate": 1.99865082540268e-05, + "loss": 1.1279, + "step": 1875 + }, + { + "epoch": 0.05, + "learning_rate": 1.998646690937038e-05, + "loss": 1.1336, + "step": 1876 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986425501504604e-05, + "loss": 1.0749, + "step": 1877 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986384030429735e-05, + "loss": 1.1295, + "step": 1878 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986342496146042e-05, + "loss": 1.1225, + "step": 1879 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986300898653787e-05, + "loss": 1.1269, + "step": 1880 + }, + { + "epoch": 0.05, + "learning_rate": 1.998625923795323e-05, + "loss": 1.1373, + "step": 1881 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986217514044633e-05, + "loss": 1.1251, + "step": 1882 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986175726928263e-05, + "loss": 1.1273, + "step": 1883 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986133876604384e-05, + "loss": 1.0903, + "step": 1884 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986091963073263e-05, + "loss": 1.1212, + "step": 1885 + }, + { + "epoch": 0.05, + "learning_rate": 1.9986049986335165e-05, + "loss": 1.0914, + "step": 1886 + }, + { + "epoch": 0.05, + "learning_rate": 1.998600794639035e-05, + "loss": 1.1046, + "step": 1887 + }, + { + "epoch": 0.05, + "learning_rate": 1.998596584323909e-05, + "loss": 1.1038, + "step": 1888 + }, + { + "epoch": 0.05, + "learning_rate": 1.998592367688165e-05, + "loss": 1.092, + "step": 1889 + }, + { + "epoch": 0.05, + "learning_rate": 1.99858814473183e-05, + "loss": 1.0859, + "step": 1890 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985839154549297e-05, + "loss": 1.088, + "step": 1891 + }, + { + "epoch": 0.05, + "learning_rate": 1.998579679857492e-05, + "loss": 1.081, + "step": 1892 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985754379395434e-05, + "loss": 1.1172, + "step": 1893 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985711897011104e-05, + "loss": 1.1029, + "step": 1894 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985669351422202e-05, + "loss": 1.0825, + "step": 1895 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985626742628996e-05, + "loss": 1.1024, + "step": 1896 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985584070631753e-05, + "loss": 1.0944, + "step": 1897 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985541335430752e-05, + "loss": 1.0911, + "step": 1898 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985498537026253e-05, + "loss": 1.0835, + "step": 1899 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985455675418533e-05, + "loss": 1.1263, + "step": 1900 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985412750607865e-05, + "loss": 1.1027, + "step": 1901 + }, + { + "epoch": 0.05, + "learning_rate": 1.998536976259451e-05, + "loss": 1.0839, + "step": 1902 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985326711378755e-05, + "loss": 1.0613, + "step": 1903 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985283596960862e-05, + "loss": 1.0593, + "step": 1904 + }, + { + "epoch": 0.05, + "learning_rate": 1.998524041934111e-05, + "loss": 1.083, + "step": 1905 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985197178519768e-05, + "loss": 1.0893, + "step": 1906 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985153874497107e-05, + "loss": 1.1113, + "step": 1907 + }, + { + "epoch": 0.05, + "learning_rate": 1.998511050727341e-05, + "loss": 1.0971, + "step": 1908 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985067076848944e-05, + "loss": 1.1087, + "step": 1909 + }, + { + "epoch": 0.05, + "learning_rate": 1.9985023583223986e-05, + "loss": 1.0764, + "step": 1910 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984980026398813e-05, + "loss": 1.1066, + "step": 1911 + }, + { + "epoch": 0.05, + "learning_rate": 1.99849364063737e-05, + "loss": 1.0856, + "step": 1912 + }, + { + "epoch": 0.05, + "learning_rate": 1.998489272314892e-05, + "loss": 1.1101, + "step": 1913 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984848976724754e-05, + "loss": 1.1171, + "step": 1914 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984805167101477e-05, + "loss": 1.0903, + "step": 1915 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984761294279367e-05, + "loss": 1.1151, + "step": 1916 + }, + { + "epoch": 0.05, + "learning_rate": 1.99847173582587e-05, + "loss": 1.1258, + "step": 1917 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984673359039758e-05, + "loss": 1.1255, + "step": 1918 + }, + { + "epoch": 0.05, + "learning_rate": 1.998462929662281e-05, + "loss": 1.0642, + "step": 1919 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984585171008146e-05, + "loss": 1.1118, + "step": 1920 + }, + { + "epoch": 0.05, + "learning_rate": 1.998454098219604e-05, + "loss": 1.1195, + "step": 1921 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984496730186772e-05, + "loss": 1.1049, + "step": 1922 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984452414980625e-05, + "loss": 1.1007, + "step": 1923 + }, + { + "epoch": 0.05, + "learning_rate": 1.998440803657787e-05, + "loss": 1.1125, + "step": 1924 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984363594978802e-05, + "loss": 1.1039, + "step": 1925 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984319090183692e-05, + "loss": 1.0899, + "step": 1926 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984274522192827e-05, + "loss": 1.1182, + "step": 1927 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984229891006483e-05, + "loss": 1.1113, + "step": 1928 + }, + { + "epoch": 0.05, + "learning_rate": 1.998418519662495e-05, + "loss": 1.1028, + "step": 1929 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984140439048507e-05, + "loss": 1.1127, + "step": 1930 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984095618277436e-05, + "loss": 1.0529, + "step": 1931 + }, + { + "epoch": 0.05, + "learning_rate": 1.9984050734312026e-05, + "loss": 1.1075, + "step": 1932 + }, + { + "epoch": 0.05, + "learning_rate": 1.998400578715255e-05, + "loss": 1.0787, + "step": 1933 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983960776799307e-05, + "loss": 1.0744, + "step": 1934 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983915703252572e-05, + "loss": 1.1085, + "step": 1935 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983870566512635e-05, + "loss": 1.1254, + "step": 1936 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983825366579778e-05, + "loss": 1.0891, + "step": 1937 + }, + { + "epoch": 0.05, + "learning_rate": 1.998378010345429e-05, + "loss": 1.1504, + "step": 1938 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983734777136455e-05, + "loss": 1.0759, + "step": 1939 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983689387626562e-05, + "loss": 1.1184, + "step": 1940 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983643934924898e-05, + "loss": 1.0935, + "step": 1941 + }, + { + "epoch": 0.05, + "learning_rate": 1.998359841903175e-05, + "loss": 1.1147, + "step": 1942 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983552839947406e-05, + "loss": 1.1281, + "step": 1943 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983507197672155e-05, + "loss": 1.0684, + "step": 1944 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983461492206288e-05, + "loss": 1.0968, + "step": 1945 + }, + { + "epoch": 0.05, + "learning_rate": 1.998341572355009e-05, + "loss": 1.1287, + "step": 1946 + }, + { + "epoch": 0.05, + "learning_rate": 1.998336989170385e-05, + "loss": 1.1584, + "step": 1947 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983323996667865e-05, + "loss": 1.0939, + "step": 1948 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983278038442416e-05, + "loss": 1.0993, + "step": 1949 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983232017027805e-05, + "loss": 1.0824, + "step": 1950 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983185932424315e-05, + "loss": 1.1225, + "step": 1951 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983139784632242e-05, + "loss": 1.1077, + "step": 1952 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983093573651874e-05, + "loss": 1.1155, + "step": 1953 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983047299483508e-05, + "loss": 1.1074, + "step": 1954 + }, + { + "epoch": 0.05, + "learning_rate": 1.9983000962127433e-05, + "loss": 1.1229, + "step": 1955 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982954561583944e-05, + "loss": 1.0741, + "step": 1956 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982908097853335e-05, + "loss": 1.0825, + "step": 1957 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982861570935902e-05, + "loss": 1.1216, + "step": 1958 + }, + { + "epoch": 0.05, + "learning_rate": 1.998281498083193e-05, + "loss": 1.106, + "step": 1959 + }, + { + "epoch": 0.05, + "learning_rate": 1.998276832754173e-05, + "loss": 1.0842, + "step": 1960 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982721611065582e-05, + "loss": 1.067, + "step": 1961 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982674831403793e-05, + "loss": 1.0694, + "step": 1962 + }, + { + "epoch": 0.05, + "learning_rate": 1.998262798855665e-05, + "loss": 1.0698, + "step": 1963 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982581082524457e-05, + "loss": 1.0816, + "step": 1964 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982534113307506e-05, + "loss": 1.1063, + "step": 1965 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982487080906095e-05, + "loss": 1.0793, + "step": 1966 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982439985320522e-05, + "loss": 1.0868, + "step": 1967 + }, + { + "epoch": 0.05, + "learning_rate": 1.998239282655109e-05, + "loss": 1.1446, + "step": 1968 + }, + { + "epoch": 0.05, + "learning_rate": 1.998234560459809e-05, + "loss": 1.0692, + "step": 1969 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982298319461825e-05, + "loss": 1.1145, + "step": 1970 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982250971142593e-05, + "loss": 1.0985, + "step": 1971 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982203559640694e-05, + "loss": 1.1085, + "step": 1972 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982156084956425e-05, + "loss": 1.0955, + "step": 1973 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982108547090095e-05, + "loss": 1.0966, + "step": 1974 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982060946041995e-05, + "loss": 1.1294, + "step": 1975 + }, + { + "epoch": 0.05, + "learning_rate": 1.9982013281812437e-05, + "loss": 1.1039, + "step": 1976 + }, + { + "epoch": 0.05, + "learning_rate": 1.998196555440171e-05, + "loss": 1.0904, + "step": 1977 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981917763810123e-05, + "loss": 1.0697, + "step": 1978 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981869910037983e-05, + "loss": 1.1154, + "step": 1979 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981821993085585e-05, + "loss": 1.0924, + "step": 1980 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981774012953236e-05, + "loss": 1.0738, + "step": 1981 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981725969641235e-05, + "loss": 1.0608, + "step": 1982 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981677863149893e-05, + "loss": 1.0701, + "step": 1983 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981629693479512e-05, + "loss": 1.1078, + "step": 1984 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981581460630398e-05, + "loss": 1.0984, + "step": 1985 + }, + { + "epoch": 0.05, + "learning_rate": 1.998153316460285e-05, + "loss": 1.0772, + "step": 1986 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981484805397183e-05, + "loss": 1.0799, + "step": 1987 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981436383013696e-05, + "loss": 1.0697, + "step": 1988 + }, + { + "epoch": 0.05, + "learning_rate": 1.99813878974527e-05, + "loss": 1.1474, + "step": 1989 + }, + { + "epoch": 0.05, + "learning_rate": 1.99813393487145e-05, + "loss": 1.095, + "step": 1990 + }, + { + "epoch": 0.05, + "learning_rate": 1.99812907367994e-05, + "loss": 1.103, + "step": 1991 + }, + { + "epoch": 0.05, + "learning_rate": 1.998124206170771e-05, + "loss": 1.0765, + "step": 1992 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981193323439743e-05, + "loss": 1.1003, + "step": 1993 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981144521995805e-05, + "loss": 1.132, + "step": 1994 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981095657376196e-05, + "loss": 1.15, + "step": 1995 + }, + { + "epoch": 0.05, + "learning_rate": 1.9981046729581238e-05, + "loss": 1.1286, + "step": 1996 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980997738611235e-05, + "loss": 1.1079, + "step": 1997 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980948684466496e-05, + "loss": 1.1101, + "step": 1998 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980899567147333e-05, + "loss": 1.1435, + "step": 1999 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980850386654057e-05, + "loss": 1.1138, + "step": 2000 + }, + { + "epoch": 0.05, + "learning_rate": 1.998080114298698e-05, + "loss": 1.1136, + "step": 2001 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980751836146412e-05, + "loss": 1.1489, + "step": 2002 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980702466132665e-05, + "loss": 1.1317, + "step": 2003 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980653032946053e-05, + "loss": 1.1187, + "step": 2004 + }, + { + "epoch": 0.05, + "learning_rate": 1.998060353658689e-05, + "loss": 1.0916, + "step": 2005 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980553977055484e-05, + "loss": 1.0934, + "step": 2006 + }, + { + "epoch": 0.05, + "learning_rate": 1.998050435435215e-05, + "loss": 1.1071, + "step": 2007 + }, + { + "epoch": 0.05, + "learning_rate": 1.998045466847721e-05, + "loss": 1.1195, + "step": 2008 + }, + { + "epoch": 0.05, + "learning_rate": 1.998040491943097e-05, + "loss": 1.1173, + "step": 2009 + }, + { + "epoch": 0.05, + "learning_rate": 1.998035510721375e-05, + "loss": 1.0859, + "step": 2010 + }, + { + "epoch": 0.05, + "learning_rate": 1.998030523182586e-05, + "loss": 1.0969, + "step": 2011 + }, + { + "epoch": 0.05, + "learning_rate": 1.998025529326762e-05, + "loss": 1.0926, + "step": 2012 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980205291539344e-05, + "loss": 1.1211, + "step": 2013 + }, + { + "epoch": 0.05, + "learning_rate": 1.998015522664135e-05, + "loss": 1.1464, + "step": 2014 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980105098573953e-05, + "loss": 1.0896, + "step": 2015 + }, + { + "epoch": 0.05, + "learning_rate": 1.998005490733747e-05, + "loss": 1.1101, + "step": 2016 + }, + { + "epoch": 0.05, + "learning_rate": 1.9980004652932226e-05, + "loss": 1.1032, + "step": 2017 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979954335358527e-05, + "loss": 1.1048, + "step": 2018 + }, + { + "epoch": 0.05, + "learning_rate": 1.99799039546167e-05, + "loss": 1.1009, + "step": 2019 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979853510707063e-05, + "loss": 1.079, + "step": 2020 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979803003629932e-05, + "loss": 1.0847, + "step": 2021 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979752433385627e-05, + "loss": 1.1087, + "step": 2022 + }, + { + "epoch": 0.05, + "learning_rate": 1.997970179997447e-05, + "loss": 1.1205, + "step": 2023 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979651103396784e-05, + "loss": 1.0614, + "step": 2024 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979600343652886e-05, + "loss": 1.1221, + "step": 2025 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979549520743094e-05, + "loss": 1.0685, + "step": 2026 + }, + { + "epoch": 0.05, + "learning_rate": 1.997949863466774e-05, + "loss": 1.0679, + "step": 2027 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979447685427137e-05, + "loss": 1.0827, + "step": 2028 + }, + { + "epoch": 0.05, + "learning_rate": 1.997939667302161e-05, + "loss": 1.1061, + "step": 2029 + }, + { + "epoch": 0.05, + "learning_rate": 1.997934559745148e-05, + "loss": 1.12, + "step": 2030 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979294458717076e-05, + "loss": 1.1042, + "step": 2031 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979243256818717e-05, + "loss": 1.1043, + "step": 2032 + }, + { + "epoch": 0.05, + "learning_rate": 1.997919199175673e-05, + "loss": 1.0798, + "step": 2033 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979140663531437e-05, + "loss": 1.1133, + "step": 2034 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979089272143165e-05, + "loss": 1.1117, + "step": 2035 + }, + { + "epoch": 0.05, + "learning_rate": 1.9979037817592234e-05, + "loss": 1.0955, + "step": 2036 + }, + { + "epoch": 0.05, + "learning_rate": 1.997898629987898e-05, + "loss": 1.0614, + "step": 2037 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978934719003716e-05, + "loss": 1.0929, + "step": 2038 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978883074966776e-05, + "loss": 1.0916, + "step": 2039 + }, + { + "epoch": 0.05, + "learning_rate": 1.997883136776849e-05, + "loss": 1.1108, + "step": 2040 + }, + { + "epoch": 0.05, + "learning_rate": 1.997877959740918e-05, + "loss": 1.0847, + "step": 2041 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978727763889175e-05, + "loss": 1.1283, + "step": 2042 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978675867208802e-05, + "loss": 1.0921, + "step": 2043 + }, + { + "epoch": 0.05, + "learning_rate": 1.997862390736839e-05, + "loss": 1.0864, + "step": 2044 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978571884368272e-05, + "loss": 1.0779, + "step": 2045 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978519798208773e-05, + "loss": 1.0922, + "step": 2046 + }, + { + "epoch": 0.05, + "learning_rate": 1.997846764889022e-05, + "loss": 1.0795, + "step": 2047 + }, + { + "epoch": 0.05, + "learning_rate": 1.997841543641295e-05, + "loss": 1.0528, + "step": 2048 + }, + { + "epoch": 0.05, + "learning_rate": 1.997836316077729e-05, + "loss": 1.0962, + "step": 2049 + }, + { + "epoch": 0.05, + "learning_rate": 1.997831082198357e-05, + "loss": 1.1088, + "step": 2050 + }, + { + "epoch": 0.05, + "learning_rate": 1.9978258420032122e-05, + "loss": 1.0667, + "step": 2051 + }, + { + "epoch": 0.05, + "learning_rate": 1.997820595492328e-05, + "loss": 1.0779, + "step": 2052 + }, + { + "epoch": 0.05, + "learning_rate": 1.997815342665737e-05, + "loss": 1.1002, + "step": 2053 + }, + { + "epoch": 0.05, + "learning_rate": 1.997810083523473e-05, + "loss": 1.0837, + "step": 2054 + }, + { + "epoch": 0.05, + "learning_rate": 1.997804818065569e-05, + "loss": 1.087, + "step": 2055 + }, + { + "epoch": 0.05, + "learning_rate": 1.997799546292059e-05, + "loss": 1.0792, + "step": 2056 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977942682029756e-05, + "loss": 1.1524, + "step": 2057 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977889837983522e-05, + "loss": 1.1311, + "step": 2058 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977836930782224e-05, + "loss": 1.0919, + "step": 2059 + }, + { + "epoch": 0.05, + "learning_rate": 1.99777839604262e-05, + "loss": 1.0659, + "step": 2060 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977730926915783e-05, + "loss": 1.0978, + "step": 2061 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977677830251307e-05, + "loss": 1.0559, + "step": 2062 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977624670433112e-05, + "loss": 1.1339, + "step": 2063 + }, + { + "epoch": 0.05, + "learning_rate": 1.997757144746153e-05, + "loss": 1.1089, + "step": 2064 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977518161336902e-05, + "loss": 1.0697, + "step": 2065 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977464812059565e-05, + "loss": 1.1104, + "step": 2066 + }, + { + "epoch": 0.05, + "learning_rate": 1.997741139962985e-05, + "loss": 1.0844, + "step": 2067 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977357924048105e-05, + "loss": 1.077, + "step": 2068 + }, + { + "epoch": 0.05, + "learning_rate": 1.997730438531466e-05, + "loss": 1.1038, + "step": 2069 + }, + { + "epoch": 0.05, + "learning_rate": 1.997725078342986e-05, + "loss": 1.0866, + "step": 2070 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977197118394038e-05, + "loss": 1.1262, + "step": 2071 + }, + { + "epoch": 0.05, + "learning_rate": 1.997714339020754e-05, + "loss": 1.0851, + "step": 2072 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977089598870704e-05, + "loss": 1.0941, + "step": 2073 + }, + { + "epoch": 0.05, + "learning_rate": 1.9977035744383868e-05, + "loss": 1.0625, + "step": 2074 + }, + { + "epoch": 0.05, + "learning_rate": 1.997698182674738e-05, + "loss": 1.1146, + "step": 2075 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976927845961568e-05, + "loss": 1.1331, + "step": 2076 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976873802026788e-05, + "loss": 1.0734, + "step": 2077 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976819694943374e-05, + "loss": 1.1141, + "step": 2078 + }, + { + "epoch": 0.05, + "learning_rate": 1.997676552471167e-05, + "loss": 1.1037, + "step": 2079 + }, + { + "epoch": 0.05, + "learning_rate": 1.997671129133202e-05, + "loss": 1.0689, + "step": 2080 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976656994804764e-05, + "loss": 1.0963, + "step": 2081 + }, + { + "epoch": 0.05, + "learning_rate": 1.997660263513025e-05, + "loss": 1.0658, + "step": 2082 + }, + { + "epoch": 0.05, + "learning_rate": 1.997654821230882e-05, + "loss": 1.0424, + "step": 2083 + }, + { + "epoch": 0.05, + "learning_rate": 1.997649372634082e-05, + "loss": 1.0838, + "step": 2084 + }, + { + "epoch": 0.05, + "learning_rate": 1.997643917722659e-05, + "loss": 1.1012, + "step": 2085 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976384564966483e-05, + "loss": 1.0611, + "step": 2086 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976329889560838e-05, + "loss": 1.1296, + "step": 2087 + }, + { + "epoch": 0.05, + "learning_rate": 1.997627515101e-05, + "loss": 1.1482, + "step": 2088 + }, + { + "epoch": 0.05, + "learning_rate": 1.997622034931433e-05, + "loss": 1.0881, + "step": 2089 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976165484474156e-05, + "loss": 1.122, + "step": 2090 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976110556489832e-05, + "loss": 1.0929, + "step": 2091 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976055565361706e-05, + "loss": 1.0643, + "step": 2092 + }, + { + "epoch": 0.05, + "learning_rate": 1.9976000511090132e-05, + "loss": 1.0697, + "step": 2093 + }, + { + "epoch": 0.05, + "learning_rate": 1.997594539367545e-05, + "loss": 1.0899, + "step": 2094 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975890213118015e-05, + "loss": 1.0992, + "step": 2095 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975834969418172e-05, + "loss": 1.1174, + "step": 2096 + }, + { + "epoch": 0.05, + "learning_rate": 1.997577966257627e-05, + "loss": 1.1, + "step": 2097 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975724292592663e-05, + "loss": 1.1095, + "step": 2098 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975668859467698e-05, + "loss": 1.0806, + "step": 2099 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975613363201732e-05, + "loss": 1.079, + "step": 2100 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975557803795108e-05, + "loss": 1.1069, + "step": 2101 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975502181248182e-05, + "loss": 1.1126, + "step": 2102 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975446495561306e-05, + "loss": 1.0918, + "step": 2103 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975390746734832e-05, + "loss": 1.0692, + "step": 2104 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975334934769113e-05, + "loss": 1.1079, + "step": 2105 + }, + { + "epoch": 0.05, + "learning_rate": 1.99752790596645e-05, + "loss": 1.0845, + "step": 2106 + }, + { + "epoch": 0.05, + "learning_rate": 1.997522312142135e-05, + "loss": 1.0713, + "step": 2107 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975167120040016e-05, + "loss": 1.1246, + "step": 2108 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975111055520853e-05, + "loss": 1.0883, + "step": 2109 + }, + { + "epoch": 0.05, + "learning_rate": 1.9975054927864216e-05, + "loss": 1.1248, + "step": 2110 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974998737070458e-05, + "loss": 1.1162, + "step": 2111 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974942483139935e-05, + "loss": 1.0978, + "step": 2112 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974886166073006e-05, + "loss": 1.0954, + "step": 2113 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974829785870026e-05, + "loss": 1.0837, + "step": 2114 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974773342531352e-05, + "loss": 1.0748, + "step": 2115 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974716836057336e-05, + "loss": 1.0623, + "step": 2116 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974660266448343e-05, + "loss": 1.098, + "step": 2117 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974603633704726e-05, + "loss": 1.1007, + "step": 2118 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974546937826852e-05, + "loss": 1.0841, + "step": 2119 + }, + { + "epoch": 0.05, + "learning_rate": 1.997449017881507e-05, + "loss": 1.0784, + "step": 2120 + }, + { + "epoch": 0.05, + "learning_rate": 1.997443335666974e-05, + "loss": 1.1254, + "step": 2121 + }, + { + "epoch": 0.05, + "learning_rate": 1.997437647139123e-05, + "loss": 1.0813, + "step": 2122 + }, + { + "epoch": 0.05, + "learning_rate": 1.997431952297989e-05, + "loss": 1.0883, + "step": 2123 + }, + { + "epoch": 0.05, + "learning_rate": 1.997426251143609e-05, + "loss": 1.1123, + "step": 2124 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974205436760186e-05, + "loss": 1.1273, + "step": 2125 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974148298952537e-05, + "loss": 1.1131, + "step": 2126 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974091098013507e-05, + "loss": 1.0859, + "step": 2127 + }, + { + "epoch": 0.05, + "learning_rate": 1.9974033833943455e-05, + "loss": 1.0964, + "step": 2128 + }, + { + "epoch": 0.05, + "learning_rate": 1.997397650674275e-05, + "loss": 1.0874, + "step": 2129 + }, + { + "epoch": 0.05, + "learning_rate": 1.997391911641175e-05, + "loss": 1.1252, + "step": 2130 + }, + { + "epoch": 0.05, + "learning_rate": 1.997386166295082e-05, + "loss": 1.0808, + "step": 2131 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973804146360323e-05, + "loss": 1.0997, + "step": 2132 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973746566640624e-05, + "loss": 1.1025, + "step": 2133 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973688923792088e-05, + "loss": 1.131, + "step": 2134 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973631217815074e-05, + "loss": 1.0523, + "step": 2135 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973573448709955e-05, + "loss": 1.0753, + "step": 2136 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973515616477093e-05, + "loss": 1.0866, + "step": 2137 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973457721116855e-05, + "loss": 1.0664, + "step": 2138 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973399762629605e-05, + "loss": 1.0923, + "step": 2139 + }, + { + "epoch": 0.05, + "learning_rate": 1.997334174101571e-05, + "loss": 1.1297, + "step": 2140 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973283656275545e-05, + "loss": 1.0998, + "step": 2141 + }, + { + "epoch": 0.05, + "learning_rate": 1.997322550840947e-05, + "loss": 1.1286, + "step": 2142 + }, + { + "epoch": 0.05, + "learning_rate": 1.997316729741785e-05, + "loss": 1.0906, + "step": 2143 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973109023301065e-05, + "loss": 1.0993, + "step": 2144 + }, + { + "epoch": 0.05, + "learning_rate": 1.9973050686059472e-05, + "loss": 1.0759, + "step": 2145 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972992285693447e-05, + "loss": 1.1062, + "step": 2146 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972933822203356e-05, + "loss": 1.1044, + "step": 2147 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972875295589574e-05, + "loss": 1.122, + "step": 2148 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972816705852467e-05, + "loss": 1.0941, + "step": 2149 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972758052992405e-05, + "loss": 1.1146, + "step": 2150 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972699337009764e-05, + "loss": 1.0946, + "step": 2151 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972640557904914e-05, + "loss": 1.0994, + "step": 2152 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972581715678223e-05, + "loss": 1.094, + "step": 2153 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972522810330067e-05, + "loss": 1.1081, + "step": 2154 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972463841860823e-05, + "loss": 1.0896, + "step": 2155 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972404810270856e-05, + "loss": 1.0877, + "step": 2156 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972345715560544e-05, + "loss": 1.0888, + "step": 2157 + }, + { + "epoch": 0.05, + "learning_rate": 1.997228655773026e-05, + "loss": 1.111, + "step": 2158 + }, + { + "epoch": 0.05, + "learning_rate": 1.997222733678038e-05, + "loss": 1.1037, + "step": 2159 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972168052711274e-05, + "loss": 1.0908, + "step": 2160 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972108705523327e-05, + "loss": 1.0991, + "step": 2161 + }, + { + "epoch": 0.05, + "learning_rate": 1.9972049295216903e-05, + "loss": 1.1099, + "step": 2162 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971989821792386e-05, + "loss": 1.0798, + "step": 2163 + }, + { + "epoch": 0.05, + "learning_rate": 1.997193028525015e-05, + "loss": 1.0695, + "step": 2164 + }, + { + "epoch": 0.05, + "learning_rate": 1.997187068559057e-05, + "loss": 1.0967, + "step": 2165 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971811022814027e-05, + "loss": 1.0904, + "step": 2166 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971751296920895e-05, + "loss": 1.1027, + "step": 2167 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971691507911556e-05, + "loss": 1.1003, + "step": 2168 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971631655786384e-05, + "loss": 1.0887, + "step": 2169 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971571740545762e-05, + "loss": 1.0639, + "step": 2170 + }, + { + "epoch": 0.05, + "learning_rate": 1.997151176219007e-05, + "loss": 1.098, + "step": 2171 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971451720719677e-05, + "loss": 1.0752, + "step": 2172 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971391616134976e-05, + "loss": 1.0956, + "step": 2173 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971331448436343e-05, + "loss": 1.0775, + "step": 2174 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971271217624154e-05, + "loss": 1.1037, + "step": 2175 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971210923698797e-05, + "loss": 1.1029, + "step": 2176 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971150566660653e-05, + "loss": 1.0872, + "step": 2177 + }, + { + "epoch": 0.05, + "learning_rate": 1.9971090146510098e-05, + "loss": 1.1173, + "step": 2178 + }, + { + "epoch": 0.05, + "learning_rate": 1.997102966324752e-05, + "loss": 1.0697, + "step": 2179 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970969116873302e-05, + "loss": 1.0905, + "step": 2180 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970908507387823e-05, + "loss": 1.0911, + "step": 2181 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970847834791472e-05, + "loss": 1.0981, + "step": 2182 + }, + { + "epoch": 0.05, + "learning_rate": 1.997078709908463e-05, + "loss": 1.1175, + "step": 2183 + }, + { + "epoch": 0.05, + "learning_rate": 1.997072630026768e-05, + "loss": 1.0746, + "step": 2184 + }, + { + "epoch": 0.05, + "learning_rate": 1.997066543834101e-05, + "loss": 1.121, + "step": 2185 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970604513305e-05, + "loss": 1.0855, + "step": 2186 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970543525160043e-05, + "loss": 1.1167, + "step": 2187 + }, + { + "epoch": 0.05, + "learning_rate": 1.997048247390652e-05, + "loss": 1.1284, + "step": 2188 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970421359544817e-05, + "loss": 1.0811, + "step": 2189 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970360182075324e-05, + "loss": 1.1033, + "step": 2190 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970298941498428e-05, + "loss": 1.0764, + "step": 2191 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970237637814517e-05, + "loss": 1.1146, + "step": 2192 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970176271023972e-05, + "loss": 1.0947, + "step": 2193 + }, + { + "epoch": 0.05, + "learning_rate": 1.9970114841127192e-05, + "loss": 1.1048, + "step": 2194 + }, + { + "epoch": 0.05, + "learning_rate": 1.997005334812456e-05, + "loss": 1.0653, + "step": 2195 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969991792016462e-05, + "loss": 1.1031, + "step": 2196 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969930172803292e-05, + "loss": 1.0723, + "step": 2197 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969868490485442e-05, + "loss": 1.0872, + "step": 2198 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969806745063302e-05, + "loss": 1.1218, + "step": 2199 + }, + { + "epoch": 0.05, + "learning_rate": 1.996974493653726e-05, + "loss": 1.0767, + "step": 2200 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969683064907704e-05, + "loss": 1.1056, + "step": 2201 + }, + { + "epoch": 0.05, + "learning_rate": 1.996962113017503e-05, + "loss": 1.0637, + "step": 2202 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969559132339632e-05, + "loss": 1.0812, + "step": 2203 + }, + { + "epoch": 0.05, + "learning_rate": 1.99694970714019e-05, + "loss": 1.096, + "step": 2204 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969434947362226e-05, + "loss": 1.0726, + "step": 2205 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969372760221004e-05, + "loss": 1.109, + "step": 2206 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969310509978628e-05, + "loss": 1.0961, + "step": 2207 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969248196635493e-05, + "loss": 1.0607, + "step": 2208 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969185820191992e-05, + "loss": 1.0851, + "step": 2209 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969123380648517e-05, + "loss": 1.0624, + "step": 2210 + }, + { + "epoch": 0.05, + "learning_rate": 1.9969060878005468e-05, + "loss": 1.1041, + "step": 2211 + }, + { + "epoch": 0.05, + "learning_rate": 1.996899831226324e-05, + "loss": 1.0932, + "step": 2212 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968935683422225e-05, + "loss": 1.1072, + "step": 2213 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968872991482824e-05, + "loss": 1.12, + "step": 2214 + }, + { + "epoch": 0.05, + "learning_rate": 1.996881023644543e-05, + "loss": 1.1185, + "step": 2215 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968747418310446e-05, + "loss": 1.0753, + "step": 2216 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968684537078263e-05, + "loss": 1.0604, + "step": 2217 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968621592749284e-05, + "loss": 1.0992, + "step": 2218 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968558585323903e-05, + "loss": 1.0524, + "step": 2219 + }, + { + "epoch": 0.05, + "learning_rate": 1.996849551480252e-05, + "loss": 1.1026, + "step": 2220 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968432381185536e-05, + "loss": 1.0809, + "step": 2221 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968369184473353e-05, + "loss": 1.1337, + "step": 2222 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968305924666363e-05, + "loss": 1.1127, + "step": 2223 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968242601764977e-05, + "loss": 1.0695, + "step": 2224 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968179215769584e-05, + "loss": 1.1284, + "step": 2225 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968115766680594e-05, + "loss": 1.0577, + "step": 2226 + }, + { + "epoch": 0.05, + "learning_rate": 1.9968052254498406e-05, + "loss": 1.0921, + "step": 2227 + }, + { + "epoch": 0.05, + "learning_rate": 1.996798867922342e-05, + "loss": 1.1026, + "step": 2228 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967925040856042e-05, + "loss": 1.0839, + "step": 2229 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967861339396672e-05, + "loss": 1.1107, + "step": 2230 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967797574845717e-05, + "loss": 1.1417, + "step": 2231 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967733747203578e-05, + "loss": 1.0737, + "step": 2232 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967669856470657e-05, + "loss": 1.0906, + "step": 2233 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967605902647357e-05, + "loss": 1.0462, + "step": 2234 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967541885734086e-05, + "loss": 1.087, + "step": 2235 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967477805731253e-05, + "loss": 1.058, + "step": 2236 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967413662639262e-05, + "loss": 1.0681, + "step": 2237 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967349456458512e-05, + "loss": 1.0988, + "step": 2238 + }, + { + "epoch": 0.05, + "learning_rate": 1.9967285187189416e-05, + "loss": 1.0564, + "step": 2239 + }, + { + "epoch": 0.06, + "learning_rate": 1.9967220854832378e-05, + "loss": 1.092, + "step": 2240 + }, + { + "epoch": 0.06, + "learning_rate": 1.9967156459387808e-05, + "loss": 1.0743, + "step": 2241 + }, + { + "epoch": 0.06, + "learning_rate": 1.996709200085611e-05, + "loss": 1.1079, + "step": 2242 + }, + { + "epoch": 0.06, + "learning_rate": 1.9967027479237693e-05, + "loss": 1.0879, + "step": 2243 + }, + { + "epoch": 0.06, + "learning_rate": 1.996696289453297e-05, + "loss": 1.1037, + "step": 2244 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966898246742342e-05, + "loss": 1.1235, + "step": 2245 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966833535866223e-05, + "loss": 1.0803, + "step": 2246 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966768761905024e-05, + "loss": 1.0861, + "step": 2247 + }, + { + "epoch": 0.06, + "learning_rate": 1.996670392485915e-05, + "loss": 1.0824, + "step": 2248 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966639024729015e-05, + "loss": 1.1147, + "step": 2249 + }, + { + "epoch": 0.06, + "learning_rate": 1.996657406151503e-05, + "loss": 1.1388, + "step": 2250 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966509035217604e-05, + "loss": 1.0524, + "step": 2251 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966443945837152e-05, + "loss": 1.1239, + "step": 2252 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966378793374076e-05, + "loss": 1.1076, + "step": 2253 + }, + { + "epoch": 0.06, + "learning_rate": 1.99663135778288e-05, + "loss": 1.0864, + "step": 2254 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966248299201737e-05, + "loss": 1.1196, + "step": 2255 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966182957493294e-05, + "loss": 1.047, + "step": 2256 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966117552703884e-05, + "loss": 1.1385, + "step": 2257 + }, + { + "epoch": 0.06, + "learning_rate": 1.9966052084833928e-05, + "loss": 1.0516, + "step": 2258 + }, + { + "epoch": 0.06, + "learning_rate": 1.996598655388383e-05, + "loss": 1.0981, + "step": 2259 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965920959854015e-05, + "loss": 1.1037, + "step": 2260 + }, + { + "epoch": 0.06, + "learning_rate": 1.996585530274489e-05, + "loss": 1.0987, + "step": 2261 + }, + { + "epoch": 0.06, + "learning_rate": 1.996578958255688e-05, + "loss": 1.0756, + "step": 2262 + }, + { + "epoch": 0.06, + "learning_rate": 1.996572379929039e-05, + "loss": 1.0702, + "step": 2263 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965657952945845e-05, + "loss": 1.0728, + "step": 2264 + }, + { + "epoch": 0.06, + "learning_rate": 1.996559204352366e-05, + "loss": 1.1, + "step": 2265 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965526071024246e-05, + "loss": 1.0645, + "step": 2266 + }, + { + "epoch": 0.06, + "learning_rate": 1.996546003544803e-05, + "loss": 1.0775, + "step": 2267 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965393936795422e-05, + "loss": 1.1549, + "step": 2268 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965327775066843e-05, + "loss": 1.0843, + "step": 2269 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965261550262713e-05, + "loss": 1.071, + "step": 2270 + }, + { + "epoch": 0.06, + "learning_rate": 1.996519526238345e-05, + "loss": 1.0735, + "step": 2271 + }, + { + "epoch": 0.06, + "learning_rate": 1.996512891142948e-05, + "loss": 1.0801, + "step": 2272 + }, + { + "epoch": 0.06, + "learning_rate": 1.9965062497401212e-05, + "loss": 1.097, + "step": 2273 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964996020299072e-05, + "loss": 1.1206, + "step": 2274 + }, + { + "epoch": 0.06, + "learning_rate": 1.996492948012348e-05, + "loss": 1.0668, + "step": 2275 + }, + { + "epoch": 0.06, + "learning_rate": 1.996486287687486e-05, + "loss": 1.0735, + "step": 2276 + }, + { + "epoch": 0.06, + "learning_rate": 1.996479621055363e-05, + "loss": 1.092, + "step": 2277 + }, + { + "epoch": 0.06, + "learning_rate": 1.996472948116021e-05, + "loss": 1.0962, + "step": 2278 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964662688695027e-05, + "loss": 1.1001, + "step": 2279 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964595833158502e-05, + "loss": 1.0649, + "step": 2280 + }, + { + "epoch": 0.06, + "learning_rate": 1.996452891455106e-05, + "loss": 1.0832, + "step": 2281 + }, + { + "epoch": 0.06, + "learning_rate": 1.996446193287312e-05, + "loss": 1.0766, + "step": 2282 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964394888125113e-05, + "loss": 1.1149, + "step": 2283 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964327780307454e-05, + "loss": 1.0603, + "step": 2284 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964260609420573e-05, + "loss": 1.1114, + "step": 2285 + }, + { + "epoch": 0.06, + "learning_rate": 1.99641933754649e-05, + "loss": 1.0705, + "step": 2286 + }, + { + "epoch": 0.06, + "learning_rate": 1.9964126078440853e-05, + "loss": 1.0707, + "step": 2287 + }, + { + "epoch": 0.06, + "learning_rate": 1.996405871834886e-05, + "loss": 1.0804, + "step": 2288 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963991295189345e-05, + "loss": 1.1067, + "step": 2289 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963923808962743e-05, + "loss": 1.0811, + "step": 2290 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963856259669473e-05, + "loss": 1.0884, + "step": 2291 + }, + { + "epoch": 0.06, + "learning_rate": 1.996378864730997e-05, + "loss": 1.0826, + "step": 2292 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963720971884653e-05, + "loss": 1.1054, + "step": 2293 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963653233393953e-05, + "loss": 1.0844, + "step": 2294 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963585431838303e-05, + "loss": 1.0809, + "step": 2295 + }, + { + "epoch": 0.06, + "learning_rate": 1.996351756721813e-05, + "loss": 1.1102, + "step": 2296 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963449639533863e-05, + "loss": 1.0921, + "step": 2297 + }, + { + "epoch": 0.06, + "learning_rate": 1.996338164878593e-05, + "loss": 1.0917, + "step": 2298 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963313594974765e-05, + "loss": 1.1068, + "step": 2299 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963245478100797e-05, + "loss": 1.0783, + "step": 2300 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963177298164457e-05, + "loss": 1.0839, + "step": 2301 + }, + { + "epoch": 0.06, + "learning_rate": 1.996310905516618e-05, + "loss": 1.1032, + "step": 2302 + }, + { + "epoch": 0.06, + "learning_rate": 1.9963040749106392e-05, + "loss": 1.1134, + "step": 2303 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962972379985528e-05, + "loss": 1.0806, + "step": 2304 + }, + { + "epoch": 0.06, + "learning_rate": 1.996290394780402e-05, + "loss": 1.101, + "step": 2305 + }, + { + "epoch": 0.06, + "learning_rate": 1.99628354525623e-05, + "loss": 1.0812, + "step": 2306 + }, + { + "epoch": 0.06, + "learning_rate": 1.996276689426081e-05, + "loss": 1.1118, + "step": 2307 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962698272899974e-05, + "loss": 1.0923, + "step": 2308 + }, + { + "epoch": 0.06, + "learning_rate": 1.996262958848023e-05, + "loss": 1.0842, + "step": 2309 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962560841002013e-05, + "loss": 1.0965, + "step": 2310 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962492030465758e-05, + "loss": 1.0795, + "step": 2311 + }, + { + "epoch": 0.06, + "learning_rate": 1.99624231568719e-05, + "loss": 1.1195, + "step": 2312 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962354220220873e-05, + "loss": 1.0437, + "step": 2313 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962285220513118e-05, + "loss": 1.1419, + "step": 2314 + }, + { + "epoch": 0.06, + "learning_rate": 1.996221615774907e-05, + "loss": 1.0893, + "step": 2315 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962147031929163e-05, + "loss": 1.0596, + "step": 2316 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962077843053837e-05, + "loss": 1.1221, + "step": 2317 + }, + { + "epoch": 0.06, + "learning_rate": 1.9962008591123533e-05, + "loss": 1.0772, + "step": 2318 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961939276138683e-05, + "loss": 1.0635, + "step": 2319 + }, + { + "epoch": 0.06, + "learning_rate": 1.996186989809973e-05, + "loss": 1.0622, + "step": 2320 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961800457007112e-05, + "loss": 1.1381, + "step": 2321 + }, + { + "epoch": 0.06, + "learning_rate": 1.996173095286127e-05, + "loss": 1.0857, + "step": 2322 + }, + { + "epoch": 0.06, + "learning_rate": 1.996166138566264e-05, + "loss": 1.1106, + "step": 2323 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961591755411668e-05, + "loss": 1.1114, + "step": 2324 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961522062108788e-05, + "loss": 1.0799, + "step": 2325 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961452305754446e-05, + "loss": 1.11, + "step": 2326 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961382486349082e-05, + "loss": 1.1489, + "step": 2327 + }, + { + "epoch": 0.06, + "learning_rate": 1.996131260389314e-05, + "loss": 1.067, + "step": 2328 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961242658387058e-05, + "loss": 1.0922, + "step": 2329 + }, + { + "epoch": 0.06, + "learning_rate": 1.996117264983128e-05, + "loss": 1.0865, + "step": 2330 + }, + { + "epoch": 0.06, + "learning_rate": 1.9961102578226255e-05, + "loss": 1.1262, + "step": 2331 + }, + { + "epoch": 0.06, + "learning_rate": 1.996103244357242e-05, + "loss": 1.097, + "step": 2332 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960962245870216e-05, + "loss": 1.1012, + "step": 2333 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960891985120096e-05, + "loss": 1.0662, + "step": 2334 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960821661322503e-05, + "loss": 1.0705, + "step": 2335 + }, + { + "epoch": 0.06, + "learning_rate": 1.996075127447788e-05, + "loss": 1.1061, + "step": 2336 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960680824586668e-05, + "loss": 1.0786, + "step": 2337 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960610311649317e-05, + "loss": 1.1157, + "step": 2338 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960539735666278e-05, + "loss": 1.0822, + "step": 2339 + }, + { + "epoch": 0.06, + "learning_rate": 1.996046909663799e-05, + "loss": 1.112, + "step": 2340 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960398394564905e-05, + "loss": 1.0858, + "step": 2341 + }, + { + "epoch": 0.06, + "learning_rate": 1.996032762944747e-05, + "loss": 1.089, + "step": 2342 + }, + { + "epoch": 0.06, + "learning_rate": 1.996025680128613e-05, + "loss": 1.0594, + "step": 2343 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960185910081334e-05, + "loss": 1.088, + "step": 2344 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960114955833534e-05, + "loss": 1.1037, + "step": 2345 + }, + { + "epoch": 0.06, + "learning_rate": 1.9960043938543178e-05, + "loss": 1.0428, + "step": 2346 + }, + { + "epoch": 0.06, + "learning_rate": 1.995997285821071e-05, + "loss": 1.0771, + "step": 2347 + }, + { + "epoch": 0.06, + "learning_rate": 1.995990171483659e-05, + "loss": 1.0783, + "step": 2348 + }, + { + "epoch": 0.06, + "learning_rate": 1.995983050842126e-05, + "loss": 1.0853, + "step": 2349 + }, + { + "epoch": 0.06, + "learning_rate": 1.995975923896517e-05, + "loss": 1.0942, + "step": 2350 + }, + { + "epoch": 0.06, + "learning_rate": 1.995968790646878e-05, + "loss": 1.0934, + "step": 2351 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959616510932535e-05, + "loss": 1.0804, + "step": 2352 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959545052356886e-05, + "loss": 1.0686, + "step": 2353 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959473530742287e-05, + "loss": 1.0674, + "step": 2354 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959401946089195e-05, + "loss": 1.0946, + "step": 2355 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959330298398055e-05, + "loss": 1.0715, + "step": 2356 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959258587669323e-05, + "loss": 1.1037, + "step": 2357 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959186813903458e-05, + "loss": 1.0627, + "step": 2358 + }, + { + "epoch": 0.06, + "learning_rate": 1.9959114977100906e-05, + "loss": 1.1088, + "step": 2359 + }, + { + "epoch": 0.06, + "learning_rate": 1.995904307726213e-05, + "loss": 1.0957, + "step": 2360 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958971114387582e-05, + "loss": 1.1014, + "step": 2361 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958899088477714e-05, + "loss": 1.1021, + "step": 2362 + }, + { + "epoch": 0.06, + "learning_rate": 1.995882699953299e-05, + "loss": 1.126, + "step": 2363 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958754847553856e-05, + "loss": 1.0702, + "step": 2364 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958682632540776e-05, + "loss": 1.0761, + "step": 2365 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958610354494205e-05, + "loss": 1.1032, + "step": 2366 + }, + { + "epoch": 0.06, + "learning_rate": 1.99585380134146e-05, + "loss": 1.0591, + "step": 2367 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958465609302417e-05, + "loss": 1.0993, + "step": 2368 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958393142158118e-05, + "loss": 1.1028, + "step": 2369 + }, + { + "epoch": 0.06, + "learning_rate": 1.995832061198216e-05, + "loss": 1.0708, + "step": 2370 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958248018775003e-05, + "loss": 1.1092, + "step": 2371 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958175362537108e-05, + "loss": 1.0763, + "step": 2372 + }, + { + "epoch": 0.06, + "learning_rate": 1.9958102643268926e-05, + "loss": 1.091, + "step": 2373 + }, + { + "epoch": 0.06, + "learning_rate": 1.995802986097093e-05, + "loss": 1.0742, + "step": 2374 + }, + { + "epoch": 0.06, + "learning_rate": 1.995795701564357e-05, + "loss": 1.0832, + "step": 2375 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957884107287318e-05, + "loss": 1.0686, + "step": 2376 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957811135902622e-05, + "loss": 1.0812, + "step": 2377 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957738101489957e-05, + "loss": 1.0818, + "step": 2378 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957665004049774e-05, + "loss": 1.0844, + "step": 2379 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957591843582547e-05, + "loss": 1.1203, + "step": 2380 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957518620088727e-05, + "loss": 1.0769, + "step": 2381 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957445333568784e-05, + "loss": 1.0771, + "step": 2382 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957371984023186e-05, + "loss": 1.1024, + "step": 2383 + }, + { + "epoch": 0.06, + "learning_rate": 1.995729857145239e-05, + "loss": 1.085, + "step": 2384 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957225095856866e-05, + "loss": 1.0791, + "step": 2385 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957151557237072e-05, + "loss": 1.122, + "step": 2386 + }, + { + "epoch": 0.06, + "learning_rate": 1.9957077955593477e-05, + "loss": 1.0675, + "step": 2387 + }, + { + "epoch": 0.06, + "learning_rate": 1.995700429092655e-05, + "loss": 1.0663, + "step": 2388 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956930563236754e-05, + "loss": 1.0983, + "step": 2389 + }, + { + "epoch": 0.06, + "learning_rate": 1.995685677252456e-05, + "loss": 1.1081, + "step": 2390 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956782918790427e-05, + "loss": 1.0814, + "step": 2391 + }, + { + "epoch": 0.06, + "learning_rate": 1.995670900203483e-05, + "loss": 1.1056, + "step": 2392 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956635022258234e-05, + "loss": 1.0804, + "step": 2393 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956560979461106e-05, + "loss": 1.0617, + "step": 2394 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956486873643917e-05, + "loss": 1.076, + "step": 2395 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956412704807134e-05, + "loss": 1.1209, + "step": 2396 + }, + { + "epoch": 0.06, + "learning_rate": 1.995633847295123e-05, + "loss": 1.1047, + "step": 2397 + }, + { + "epoch": 0.06, + "learning_rate": 1.995626417807667e-05, + "loss": 1.0553, + "step": 2398 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956189820183924e-05, + "loss": 1.0632, + "step": 2399 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956115399273467e-05, + "loss": 1.1386, + "step": 2400 + }, + { + "epoch": 0.06, + "learning_rate": 1.9956040915345768e-05, + "loss": 1.0777, + "step": 2401 + }, + { + "epoch": 0.06, + "learning_rate": 1.99559663684013e-05, + "loss": 1.1022, + "step": 2402 + }, + { + "epoch": 0.06, + "learning_rate": 1.995589175844053e-05, + "loss": 1.09, + "step": 2403 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955817085463937e-05, + "loss": 1.0865, + "step": 2404 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955742349471992e-05, + "loss": 1.0941, + "step": 2405 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955667550465163e-05, + "loss": 1.0826, + "step": 2406 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955592688443928e-05, + "loss": 1.0624, + "step": 2407 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955517763408758e-05, + "loss": 1.0822, + "step": 2408 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955442775360133e-05, + "loss": 1.1071, + "step": 2409 + }, + { + "epoch": 0.06, + "learning_rate": 1.995536772429852e-05, + "loss": 1.0876, + "step": 2410 + }, + { + "epoch": 0.06, + "learning_rate": 1.99552926102244e-05, + "loss": 1.0854, + "step": 2411 + }, + { + "epoch": 0.06, + "learning_rate": 1.995521743313824e-05, + "loss": 1.0651, + "step": 2412 + }, + { + "epoch": 0.06, + "learning_rate": 1.995514219304053e-05, + "loss": 1.0938, + "step": 2413 + }, + { + "epoch": 0.06, + "learning_rate": 1.9955066889931732e-05, + "loss": 1.0904, + "step": 2414 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954991523812334e-05, + "loss": 1.0938, + "step": 2415 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954916094682805e-05, + "loss": 1.1049, + "step": 2416 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954840602543623e-05, + "loss": 1.128, + "step": 2417 + }, + { + "epoch": 0.06, + "learning_rate": 1.995476504739527e-05, + "loss": 1.0542, + "step": 2418 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954689429238222e-05, + "loss": 1.0805, + "step": 2419 + }, + { + "epoch": 0.06, + "learning_rate": 1.995461374807296e-05, + "loss": 1.1058, + "step": 2420 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954538003899957e-05, + "loss": 1.0767, + "step": 2421 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954462196719698e-05, + "loss": 1.1129, + "step": 2422 + }, + { + "epoch": 0.06, + "learning_rate": 1.995438632653266e-05, + "loss": 1.1157, + "step": 2423 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954310393339326e-05, + "loss": 1.0933, + "step": 2424 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954234397140175e-05, + "loss": 1.0587, + "step": 2425 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954158337935685e-05, + "loss": 1.1018, + "step": 2426 + }, + { + "epoch": 0.06, + "learning_rate": 1.9954082215726346e-05, + "loss": 1.0711, + "step": 2427 + }, + { + "epoch": 0.06, + "learning_rate": 1.995400603051263e-05, + "loss": 1.0832, + "step": 2428 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953929782295024e-05, + "loss": 1.0699, + "step": 2429 + }, + { + "epoch": 0.06, + "learning_rate": 1.995385347107401e-05, + "loss": 1.1009, + "step": 2430 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953777096850075e-05, + "loss": 1.1011, + "step": 2431 + }, + { + "epoch": 0.06, + "learning_rate": 1.995370065962369e-05, + "loss": 1.0634, + "step": 2432 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953624159395355e-05, + "loss": 1.092, + "step": 2433 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953547596165542e-05, + "loss": 1.0836, + "step": 2434 + }, + { + "epoch": 0.06, + "learning_rate": 1.995347096993474e-05, + "loss": 1.0763, + "step": 2435 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953394280703437e-05, + "loss": 1.1076, + "step": 2436 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953317528472117e-05, + "loss": 1.1111, + "step": 2437 + }, + { + "epoch": 0.06, + "learning_rate": 1.995324071324126e-05, + "loss": 1.0812, + "step": 2438 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953163835011357e-05, + "loss": 1.0733, + "step": 2439 + }, + { + "epoch": 0.06, + "learning_rate": 1.9953086893782897e-05, + "loss": 1.0871, + "step": 2440 + }, + { + "epoch": 0.06, + "learning_rate": 1.995300988955636e-05, + "loss": 1.0968, + "step": 2441 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952932822332244e-05, + "loss": 1.0659, + "step": 2442 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952855692111025e-05, + "loss": 1.1185, + "step": 2443 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952778498893197e-05, + "loss": 1.0741, + "step": 2444 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952701242679248e-05, + "loss": 1.0505, + "step": 2445 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952623923469667e-05, + "loss": 1.1278, + "step": 2446 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952546541264944e-05, + "loss": 1.0561, + "step": 2447 + }, + { + "epoch": 0.06, + "learning_rate": 1.995246909606557e-05, + "loss": 1.0755, + "step": 2448 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952391587872027e-05, + "loss": 1.0859, + "step": 2449 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952314016684818e-05, + "loss": 1.0934, + "step": 2450 + }, + { + "epoch": 0.06, + "learning_rate": 1.995223638250443e-05, + "loss": 1.0772, + "step": 2451 + }, + { + "epoch": 0.06, + "learning_rate": 1.995215868533135e-05, + "loss": 1.1024, + "step": 2452 + }, + { + "epoch": 0.06, + "learning_rate": 1.995208092516607e-05, + "loss": 1.0797, + "step": 2453 + }, + { + "epoch": 0.06, + "learning_rate": 1.9952003102009085e-05, + "loss": 1.0884, + "step": 2454 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951925215860888e-05, + "loss": 1.0718, + "step": 2455 + }, + { + "epoch": 0.06, + "learning_rate": 1.995184726672197e-05, + "loss": 1.0994, + "step": 2456 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951769254592825e-05, + "loss": 1.1271, + "step": 2457 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951691179473948e-05, + "loss": 1.0918, + "step": 2458 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951613041365833e-05, + "loss": 1.1132, + "step": 2459 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951534840268972e-05, + "loss": 1.1088, + "step": 2460 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951456576183866e-05, + "loss": 1.082, + "step": 2461 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951378249111e-05, + "loss": 1.0659, + "step": 2462 + }, + { + "epoch": 0.06, + "learning_rate": 1.995129985905088e-05, + "loss": 1.1099, + "step": 2463 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951221406003997e-05, + "loss": 1.0884, + "step": 2464 + }, + { + "epoch": 0.06, + "learning_rate": 1.995114288997085e-05, + "loss": 1.0822, + "step": 2465 + }, + { + "epoch": 0.06, + "learning_rate": 1.9951064310951932e-05, + "loss": 1.0964, + "step": 2466 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950985668947748e-05, + "loss": 1.0774, + "step": 2467 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950906963958787e-05, + "loss": 1.0618, + "step": 2468 + }, + { + "epoch": 0.06, + "learning_rate": 1.995082819598555e-05, + "loss": 1.079, + "step": 2469 + }, + { + "epoch": 0.06, + "learning_rate": 1.995074936502854e-05, + "loss": 1.0909, + "step": 2470 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950670471088246e-05, + "loss": 1.0866, + "step": 2471 + }, + { + "epoch": 0.06, + "learning_rate": 1.995059151416518e-05, + "loss": 1.0752, + "step": 2472 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950512494259836e-05, + "loss": 1.0555, + "step": 2473 + }, + { + "epoch": 0.06, + "learning_rate": 1.995043341137271e-05, + "loss": 1.0856, + "step": 2474 + }, + { + "epoch": 0.06, + "learning_rate": 1.995035426550431e-05, + "loss": 1.1277, + "step": 2475 + }, + { + "epoch": 0.06, + "learning_rate": 1.995027505665513e-05, + "loss": 1.079, + "step": 2476 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950195784825676e-05, + "loss": 1.0998, + "step": 2477 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950116450016446e-05, + "loss": 1.1145, + "step": 2478 + }, + { + "epoch": 0.06, + "learning_rate": 1.9950037052227945e-05, + "loss": 1.0633, + "step": 2479 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949957591460677e-05, + "loss": 1.1105, + "step": 2480 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949878067715144e-05, + "loss": 1.0938, + "step": 2481 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949798480991848e-05, + "loss": 1.1131, + "step": 2482 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949718831291293e-05, + "loss": 1.0802, + "step": 2483 + }, + { + "epoch": 0.06, + "learning_rate": 1.994963911861398e-05, + "loss": 1.099, + "step": 2484 + }, + { + "epoch": 0.06, + "learning_rate": 1.994955934296042e-05, + "loss": 1.1166, + "step": 2485 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949479504331113e-05, + "loss": 1.0881, + "step": 2486 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949399602726568e-05, + "loss": 1.0936, + "step": 2487 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949319638147288e-05, + "loss": 1.0622, + "step": 2488 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949239610593782e-05, + "loss": 1.0705, + "step": 2489 + }, + { + "epoch": 0.06, + "learning_rate": 1.994915952006655e-05, + "loss": 1.0855, + "step": 2490 + }, + { + "epoch": 0.06, + "learning_rate": 1.9949079366566106e-05, + "loss": 1.0469, + "step": 2491 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948999150092953e-05, + "loss": 1.0906, + "step": 2492 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948918870647603e-05, + "loss": 1.0914, + "step": 2493 + }, + { + "epoch": 0.06, + "learning_rate": 1.994883852823056e-05, + "loss": 1.1104, + "step": 2494 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948758122842332e-05, + "loss": 1.1124, + "step": 2495 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948677654483434e-05, + "loss": 1.1119, + "step": 2496 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948597123154368e-05, + "loss": 1.0932, + "step": 2497 + }, + { + "epoch": 0.06, + "learning_rate": 1.994851652885565e-05, + "loss": 1.0778, + "step": 2498 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948435871587788e-05, + "loss": 1.0816, + "step": 2499 + }, + { + "epoch": 0.06, + "learning_rate": 1.994835515135129e-05, + "loss": 1.0992, + "step": 2500 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948274368146667e-05, + "loss": 1.0914, + "step": 2501 + }, + { + "epoch": 0.06, + "learning_rate": 1.9948193521974436e-05, + "loss": 1.1011, + "step": 2502 + }, + { + "epoch": 0.06, + "learning_rate": 1.99481126128351e-05, + "loss": 1.1012, + "step": 2503 + }, + { + "epoch": 0.06, + "learning_rate": 1.994803164072918e-05, + "loss": 1.0596, + "step": 2504 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947950605657183e-05, + "loss": 1.1062, + "step": 2505 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947869507619623e-05, + "loss": 1.1052, + "step": 2506 + }, + { + "epoch": 0.06, + "learning_rate": 1.994778834661701e-05, + "loss": 1.1297, + "step": 2507 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947707122649867e-05, + "loss": 1.0949, + "step": 2508 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947625835718698e-05, + "loss": 1.1138, + "step": 2509 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947544485824023e-05, + "loss": 1.0538, + "step": 2510 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947463072966356e-05, + "loss": 1.0631, + "step": 2511 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947381597146216e-05, + "loss": 1.0898, + "step": 2512 + }, + { + "epoch": 0.06, + "learning_rate": 1.994730005836411e-05, + "loss": 1.1156, + "step": 2513 + }, + { + "epoch": 0.06, + "learning_rate": 1.9947218456620558e-05, + "loss": 1.1108, + "step": 2514 + }, + { + "epoch": 0.06, + "learning_rate": 1.994713679191608e-05, + "loss": 1.1149, + "step": 2515 + }, + { + "epoch": 0.06, + "learning_rate": 1.994705506425119e-05, + "loss": 1.092, + "step": 2516 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946973273626408e-05, + "loss": 1.0942, + "step": 2517 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946891420042245e-05, + "loss": 1.1125, + "step": 2518 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946809503499225e-05, + "loss": 1.0701, + "step": 2519 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946727523997865e-05, + "loss": 1.0823, + "step": 2520 + }, + { + "epoch": 0.06, + "learning_rate": 1.994664548153868e-05, + "loss": 1.0932, + "step": 2521 + }, + { + "epoch": 0.06, + "learning_rate": 1.99465633761222e-05, + "loss": 1.1163, + "step": 2522 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946481207748933e-05, + "loss": 1.065, + "step": 2523 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946398976419406e-05, + "loss": 1.0608, + "step": 2524 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946316682134138e-05, + "loss": 1.1039, + "step": 2525 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946234324893644e-05, + "loss": 1.0935, + "step": 2526 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946151904698454e-05, + "loss": 1.0963, + "step": 2527 + }, + { + "epoch": 0.06, + "learning_rate": 1.9946069421549084e-05, + "loss": 1.0794, + "step": 2528 + }, + { + "epoch": 0.06, + "learning_rate": 1.994598687544606e-05, + "loss": 1.0926, + "step": 2529 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945904266389906e-05, + "loss": 1.1109, + "step": 2530 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945821594381138e-05, + "loss": 1.0326, + "step": 2531 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945738859420284e-05, + "loss": 1.1066, + "step": 2532 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945656061507864e-05, + "loss": 1.0683, + "step": 2533 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945573200644406e-05, + "loss": 1.0639, + "step": 2534 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945490276830432e-05, + "loss": 1.0843, + "step": 2535 + }, + { + "epoch": 0.06, + "learning_rate": 1.994540729006647e-05, + "loss": 1.1305, + "step": 2536 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945324240353042e-05, + "loss": 1.0601, + "step": 2537 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945241127690676e-05, + "loss": 1.088, + "step": 2538 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945157952079896e-05, + "loss": 1.1149, + "step": 2539 + }, + { + "epoch": 0.06, + "learning_rate": 1.9945074713521228e-05, + "loss": 1.1038, + "step": 2540 + }, + { + "epoch": 0.06, + "learning_rate": 1.99449914120152e-05, + "loss": 1.1137, + "step": 2541 + }, + { + "epoch": 0.06, + "learning_rate": 1.994490804756234e-05, + "loss": 1.0801, + "step": 2542 + }, + { + "epoch": 0.06, + "learning_rate": 1.994482462016318e-05, + "loss": 1.0679, + "step": 2543 + }, + { + "epoch": 0.06, + "learning_rate": 1.994474112981824e-05, + "loss": 1.0619, + "step": 2544 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944657576528047e-05, + "loss": 1.0834, + "step": 2545 + }, + { + "epoch": 0.06, + "learning_rate": 1.994457396029314e-05, + "loss": 1.1393, + "step": 2546 + }, + { + "epoch": 0.06, + "learning_rate": 1.994449028111404e-05, + "loss": 1.0672, + "step": 2547 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944406538991277e-05, + "loss": 1.104, + "step": 2548 + }, + { + "epoch": 0.06, + "learning_rate": 1.994432273392539e-05, + "loss": 1.0408, + "step": 2549 + }, + { + "epoch": 0.06, + "learning_rate": 1.9944238865916898e-05, + "loss": 1.1172, + "step": 2550 + }, + { + "epoch": 0.06, + "learning_rate": 1.994415493496634e-05, + "loss": 1.0737, + "step": 2551 + }, + { + "epoch": 0.06, + "learning_rate": 1.994407094107424e-05, + "loss": 1.0934, + "step": 2552 + }, + { + "epoch": 0.06, + "learning_rate": 1.994398688424114e-05, + "loss": 1.0508, + "step": 2553 + }, + { + "epoch": 0.06, + "learning_rate": 1.994390276446756e-05, + "loss": 1.1051, + "step": 2554 + }, + { + "epoch": 0.06, + "learning_rate": 1.994381858175404e-05, + "loss": 1.0697, + "step": 2555 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943734336101108e-05, + "loss": 1.1282, + "step": 2556 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943650027509303e-05, + "loss": 1.0735, + "step": 2557 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943565655979154e-05, + "loss": 1.076, + "step": 2558 + }, + { + "epoch": 0.06, + "learning_rate": 1.99434812215112e-05, + "loss": 1.121, + "step": 2559 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943396724105968e-05, + "loss": 1.0762, + "step": 2560 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943312163764e-05, + "loss": 1.0782, + "step": 2561 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943227540485828e-05, + "loss": 1.0558, + "step": 2562 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943142854271986e-05, + "loss": 1.0894, + "step": 2563 + }, + { + "epoch": 0.06, + "learning_rate": 1.9943058105123013e-05, + "loss": 1.0659, + "step": 2564 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942973293039442e-05, + "loss": 1.0751, + "step": 2565 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942888418021814e-05, + "loss": 1.0648, + "step": 2566 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942803480070668e-05, + "loss": 1.1067, + "step": 2567 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942718479186532e-05, + "loss": 1.1139, + "step": 2568 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942633415369953e-05, + "loss": 1.101, + "step": 2569 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942548288621468e-05, + "loss": 1.0684, + "step": 2570 + }, + { + "epoch": 0.06, + "learning_rate": 1.994246309894161e-05, + "loss": 1.0772, + "step": 2571 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942377846330927e-05, + "loss": 1.1143, + "step": 2572 + }, + { + "epoch": 0.06, + "learning_rate": 1.994229253078995e-05, + "loss": 1.097, + "step": 2573 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942207152319227e-05, + "loss": 1.0672, + "step": 2574 + }, + { + "epoch": 0.06, + "learning_rate": 1.9942121710919294e-05, + "loss": 1.1156, + "step": 2575 + }, + { + "epoch": 0.06, + "learning_rate": 1.994203620659069e-05, + "loss": 1.0841, + "step": 2576 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941950639333956e-05, + "loss": 1.0852, + "step": 2577 + }, + { + "epoch": 0.06, + "learning_rate": 1.994186500914964e-05, + "loss": 1.0907, + "step": 2578 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941779316038278e-05, + "loss": 1.0856, + "step": 2579 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941693560000414e-05, + "loss": 1.0811, + "step": 2580 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941607741036594e-05, + "loss": 1.0468, + "step": 2581 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941521859147358e-05, + "loss": 1.0968, + "step": 2582 + }, + { + "epoch": 0.06, + "learning_rate": 1.994143591433325e-05, + "loss": 1.0949, + "step": 2583 + }, + { + "epoch": 0.06, + "learning_rate": 1.9941349906594812e-05, + "loss": 1.1236, + "step": 2584 + }, + { + "epoch": 0.06, + "learning_rate": 1.994126383593259e-05, + "loss": 1.0622, + "step": 2585 + }, + { + "epoch": 0.06, + "learning_rate": 1.994117770234713e-05, + "loss": 1.1041, + "step": 2586 + }, + { + "epoch": 0.06, + "learning_rate": 1.994109150583898e-05, + "loss": 1.081, + "step": 2587 + }, + { + "epoch": 0.06, + "learning_rate": 1.994100524640868e-05, + "loss": 1.0837, + "step": 2588 + }, + { + "epoch": 0.06, + "learning_rate": 1.994091892405677e-05, + "loss": 1.0853, + "step": 2589 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940832538783815e-05, + "loss": 1.0886, + "step": 2590 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940746090590343e-05, + "loss": 1.1011, + "step": 2591 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940659579476915e-05, + "loss": 1.0914, + "step": 2592 + }, + { + "epoch": 0.06, + "learning_rate": 1.994057300544407e-05, + "loss": 1.0579, + "step": 2593 + }, + { + "epoch": 0.06, + "learning_rate": 1.994048636849236e-05, + "loss": 1.0557, + "step": 2594 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940399668622334e-05, + "loss": 1.0623, + "step": 2595 + }, + { + "epoch": 0.06, + "learning_rate": 1.9940312905834536e-05, + "loss": 1.085, + "step": 2596 + }, + { + "epoch": 0.06, + "learning_rate": 1.994022608012952e-05, + "loss": 1.0801, + "step": 2597 + }, + { + "epoch": 0.06, + "learning_rate": 1.994013919150783e-05, + "loss": 1.0665, + "step": 2598 + }, + { + "epoch": 0.06, + "learning_rate": 1.994005223997002e-05, + "loss": 1.0735, + "step": 2599 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939965225516644e-05, + "loss": 1.0404, + "step": 2600 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939878148148246e-05, + "loss": 1.075, + "step": 2601 + }, + { + "epoch": 0.06, + "learning_rate": 1.993979100786538e-05, + "loss": 1.0955, + "step": 2602 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939703804668598e-05, + "loss": 1.0856, + "step": 2603 + }, + { + "epoch": 0.06, + "learning_rate": 1.993961653855845e-05, + "loss": 1.1056, + "step": 2604 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939529209535494e-05, + "loss": 1.0931, + "step": 2605 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939441817600273e-05, + "loss": 1.0969, + "step": 2606 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939354362753347e-05, + "loss": 1.0848, + "step": 2607 + }, + { + "epoch": 0.06, + "learning_rate": 1.993926684499527e-05, + "loss": 1.0728, + "step": 2608 + }, + { + "epoch": 0.06, + "learning_rate": 1.9939179264326592e-05, + "loss": 1.0889, + "step": 2609 + }, + { + "epoch": 0.06, + "learning_rate": 1.993909162074787e-05, + "loss": 1.0691, + "step": 2610 + }, + { + "epoch": 0.06, + "learning_rate": 1.993900391425966e-05, + "loss": 1.0495, + "step": 2611 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938916144862516e-05, + "loss": 1.0749, + "step": 2612 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938828312556987e-05, + "loss": 1.0961, + "step": 2613 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938740417343642e-05, + "loss": 1.0463, + "step": 2614 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938652459223027e-05, + "loss": 1.0633, + "step": 2615 + }, + { + "epoch": 0.06, + "learning_rate": 1.99385644381957e-05, + "loss": 1.1367, + "step": 2616 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938476354262223e-05, + "loss": 1.0893, + "step": 2617 + }, + { + "epoch": 0.06, + "learning_rate": 1.993838820742315e-05, + "loss": 1.0467, + "step": 2618 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938299997679037e-05, + "loss": 1.1167, + "step": 2619 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938211725030447e-05, + "loss": 1.1155, + "step": 2620 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938123389477937e-05, + "loss": 1.1049, + "step": 2621 + }, + { + "epoch": 0.06, + "learning_rate": 1.9938034991022064e-05, + "loss": 1.0828, + "step": 2622 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937946529663387e-05, + "loss": 1.0912, + "step": 2623 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937858005402474e-05, + "loss": 1.0481, + "step": 2624 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937769418239874e-05, + "loss": 1.0687, + "step": 2625 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937680768176154e-05, + "loss": 1.0654, + "step": 2626 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937592055211876e-05, + "loss": 1.0782, + "step": 2627 + }, + { + "epoch": 0.06, + "learning_rate": 1.99375032793476e-05, + "loss": 1.0719, + "step": 2628 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937414440583888e-05, + "loss": 1.0513, + "step": 2629 + }, + { + "epoch": 0.06, + "learning_rate": 1.99373255389213e-05, + "loss": 1.0622, + "step": 2630 + }, + { + "epoch": 0.06, + "learning_rate": 1.9937236574360402e-05, + "loss": 1.0798, + "step": 2631 + }, + { + "epoch": 0.06, + "learning_rate": 1.993714754690175e-05, + "loss": 1.0745, + "step": 2632 + }, + { + "epoch": 0.06, + "learning_rate": 1.993705845654592e-05, + "loss": 1.0824, + "step": 2633 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936969303293466e-05, + "loss": 1.0496, + "step": 2634 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936880087144956e-05, + "loss": 1.0713, + "step": 2635 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936790808100956e-05, + "loss": 1.0936, + "step": 2636 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936701466162026e-05, + "loss": 1.0983, + "step": 2637 + }, + { + "epoch": 0.06, + "learning_rate": 1.993661206132874e-05, + "loss": 1.0655, + "step": 2638 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936522593601657e-05, + "loss": 1.0582, + "step": 2639 + }, + { + "epoch": 0.06, + "learning_rate": 1.993643306298134e-05, + "loss": 1.0689, + "step": 2640 + }, + { + "epoch": 0.06, + "learning_rate": 1.9936343469468362e-05, + "loss": 1.1215, + "step": 2641 + }, + { + "epoch": 0.06, + "learning_rate": 1.993625381306329e-05, + "loss": 1.0728, + "step": 2642 + }, + { + "epoch": 0.06, + "learning_rate": 1.993616409376669e-05, + "loss": 1.0887, + "step": 2643 + }, + { + "epoch": 0.06, + "learning_rate": 1.993607431157913e-05, + "loss": 1.0504, + "step": 2644 + }, + { + "epoch": 0.06, + "learning_rate": 1.993598446650118e-05, + "loss": 1.091, + "step": 2645 + }, + { + "epoch": 0.06, + "learning_rate": 1.99358945585334e-05, + "loss": 1.0819, + "step": 2646 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935804587676376e-05, + "loss": 1.1128, + "step": 2647 + }, + { + "epoch": 0.07, + "learning_rate": 1.993571455393066e-05, + "loss": 1.1045, + "step": 2648 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935624457296834e-05, + "loss": 1.0638, + "step": 2649 + }, + { + "epoch": 0.07, + "learning_rate": 1.993553429777546e-05, + "loss": 1.0612, + "step": 2650 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935444075367114e-05, + "loss": 1.1092, + "step": 2651 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935353790072364e-05, + "loss": 1.0843, + "step": 2652 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935263441891784e-05, + "loss": 1.091, + "step": 2653 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935173030825946e-05, + "loss": 1.0685, + "step": 2654 + }, + { + "epoch": 0.07, + "learning_rate": 1.9935082556875417e-05, + "loss": 1.1127, + "step": 2655 + }, + { + "epoch": 0.07, + "learning_rate": 1.993499202004078e-05, + "loss": 1.0923, + "step": 2656 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934901420322594e-05, + "loss": 1.0816, + "step": 2657 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934810757721443e-05, + "loss": 1.085, + "step": 2658 + }, + { + "epoch": 0.07, + "learning_rate": 1.99347200322379e-05, + "loss": 1.0563, + "step": 2659 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934629243872535e-05, + "loss": 1.069, + "step": 2660 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934538392625925e-05, + "loss": 1.0988, + "step": 2661 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934447478498648e-05, + "loss": 1.1078, + "step": 2662 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934356501491274e-05, + "loss": 1.0661, + "step": 2663 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934265461604378e-05, + "loss": 1.0846, + "step": 2664 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934174358838544e-05, + "loss": 1.0826, + "step": 2665 + }, + { + "epoch": 0.07, + "learning_rate": 1.9934083193194344e-05, + "loss": 1.0731, + "step": 2666 + }, + { + "epoch": 0.07, + "learning_rate": 1.993399196467235e-05, + "loss": 1.0796, + "step": 2667 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933900673273146e-05, + "loss": 1.0465, + "step": 2668 + }, + { + "epoch": 0.07, + "learning_rate": 1.993380931899731e-05, + "loss": 1.0978, + "step": 2669 + }, + { + "epoch": 0.07, + "learning_rate": 1.993371790184542e-05, + "loss": 1.021, + "step": 2670 + }, + { + "epoch": 0.07, + "learning_rate": 1.993362642181805e-05, + "loss": 1.0508, + "step": 2671 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933534878915787e-05, + "loss": 1.0709, + "step": 2672 + }, + { + "epoch": 0.07, + "learning_rate": 1.99334432731392e-05, + "loss": 1.0966, + "step": 2673 + }, + { + "epoch": 0.07, + "learning_rate": 1.993335160448888e-05, + "loss": 1.0813, + "step": 2674 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933259872965395e-05, + "loss": 1.0941, + "step": 2675 + }, + { + "epoch": 0.07, + "learning_rate": 1.9933168078569336e-05, + "loss": 1.0816, + "step": 2676 + }, + { + "epoch": 0.07, + "learning_rate": 1.993307622130128e-05, + "loss": 1.061, + "step": 2677 + }, + { + "epoch": 0.07, + "learning_rate": 1.993298430116181e-05, + "loss": 1.0614, + "step": 2678 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932892318151502e-05, + "loss": 1.0925, + "step": 2679 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932800272270948e-05, + "loss": 1.068, + "step": 2680 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932708163520722e-05, + "loss": 1.0956, + "step": 2681 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932615991901413e-05, + "loss": 1.1017, + "step": 2682 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932523757413598e-05, + "loss": 1.0607, + "step": 2683 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932431460057865e-05, + "loss": 1.109, + "step": 2684 + }, + { + "epoch": 0.07, + "learning_rate": 1.99323390998348e-05, + "loss": 1.0605, + "step": 2685 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932246676744984e-05, + "loss": 1.1203, + "step": 2686 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932154190789004e-05, + "loss": 1.0923, + "step": 2687 + }, + { + "epoch": 0.07, + "learning_rate": 1.9932061641967445e-05, + "loss": 1.0773, + "step": 2688 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931969030280893e-05, + "loss": 1.0979, + "step": 2689 + }, + { + "epoch": 0.07, + "learning_rate": 1.993187635572993e-05, + "loss": 1.0552, + "step": 2690 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931783618315147e-05, + "loss": 1.0924, + "step": 2691 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931690818037133e-05, + "loss": 1.0907, + "step": 2692 + }, + { + "epoch": 0.07, + "learning_rate": 1.993159795489647e-05, + "loss": 1.0394, + "step": 2693 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931505028893748e-05, + "loss": 1.0917, + "step": 2694 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931412040029557e-05, + "loss": 1.1193, + "step": 2695 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931318988304483e-05, + "loss": 1.0758, + "step": 2696 + }, + { + "epoch": 0.07, + "learning_rate": 1.993122587371912e-05, + "loss": 1.0585, + "step": 2697 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931132696274053e-05, + "loss": 1.0685, + "step": 2698 + }, + { + "epoch": 0.07, + "learning_rate": 1.9931039455969866e-05, + "loss": 1.0597, + "step": 2699 + }, + { + "epoch": 0.07, + "learning_rate": 1.993094615280716e-05, + "loss": 1.0667, + "step": 2700 + }, + { + "epoch": 0.07, + "learning_rate": 1.993085278678652e-05, + "loss": 1.039, + "step": 2701 + }, + { + "epoch": 0.07, + "learning_rate": 1.993075935790854e-05, + "loss": 1.0981, + "step": 2702 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930665866173804e-05, + "loss": 1.0758, + "step": 2703 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930572311582915e-05, + "loss": 1.0439, + "step": 2704 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930478694136454e-05, + "loss": 1.0982, + "step": 2705 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930385013835023e-05, + "loss": 1.0463, + "step": 2706 + }, + { + "epoch": 0.07, + "learning_rate": 1.993029127067921e-05, + "loss": 1.0478, + "step": 2707 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930197464669606e-05, + "loss": 1.0361, + "step": 2708 + }, + { + "epoch": 0.07, + "learning_rate": 1.993010359580681e-05, + "loss": 1.0629, + "step": 2709 + }, + { + "epoch": 0.07, + "learning_rate": 1.9930009664091414e-05, + "loss": 1.0548, + "step": 2710 + }, + { + "epoch": 0.07, + "learning_rate": 1.992991566952401e-05, + "loss": 1.078, + "step": 2711 + }, + { + "epoch": 0.07, + "learning_rate": 1.99298216121052e-05, + "loss": 1.0983, + "step": 2712 + }, + { + "epoch": 0.07, + "learning_rate": 1.992972749183557e-05, + "loss": 1.0858, + "step": 2713 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929633308715723e-05, + "loss": 1.0931, + "step": 2714 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929539062746252e-05, + "loss": 1.0571, + "step": 2715 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929444753927756e-05, + "loss": 1.0713, + "step": 2716 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929350382260825e-05, + "loss": 1.0717, + "step": 2717 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929255947746065e-05, + "loss": 1.1394, + "step": 2718 + }, + { + "epoch": 0.07, + "learning_rate": 1.992916145038407e-05, + "loss": 1.0745, + "step": 2719 + }, + { + "epoch": 0.07, + "learning_rate": 1.9929066890175438e-05, + "loss": 1.0733, + "step": 2720 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928972267120766e-05, + "loss": 1.0982, + "step": 2721 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928877581220658e-05, + "loss": 1.0461, + "step": 2722 + }, + { + "epoch": 0.07, + "learning_rate": 1.992878283247571e-05, + "loss": 1.0602, + "step": 2723 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928688020886516e-05, + "loss": 1.0944, + "step": 2724 + }, + { + "epoch": 0.07, + "learning_rate": 1.992859314645369e-05, + "loss": 1.0828, + "step": 2725 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928498209177817e-05, + "loss": 1.0841, + "step": 2726 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928403209059507e-05, + "loss": 1.0775, + "step": 2727 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928308146099362e-05, + "loss": 1.0843, + "step": 2728 + }, + { + "epoch": 0.07, + "learning_rate": 1.9928213020297978e-05, + "loss": 1.0742, + "step": 2729 + }, + { + "epoch": 0.07, + "learning_rate": 1.992811783165596e-05, + "loss": 1.0934, + "step": 2730 + }, + { + "epoch": 0.07, + "learning_rate": 1.992802258017392e-05, + "loss": 1.0395, + "step": 2731 + }, + { + "epoch": 0.07, + "learning_rate": 1.992792726585244e-05, + "loss": 1.0971, + "step": 2732 + }, + { + "epoch": 0.07, + "learning_rate": 1.992783188869214e-05, + "loss": 1.0924, + "step": 2733 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927736448693617e-05, + "loss": 1.112, + "step": 2734 + }, + { + "epoch": 0.07, + "learning_rate": 1.992764094585748e-05, + "loss": 1.0708, + "step": 2735 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927545380184326e-05, + "loss": 1.0801, + "step": 2736 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927449751674767e-05, + "loss": 1.0808, + "step": 2737 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927354060329406e-05, + "loss": 1.0664, + "step": 2738 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927258306148847e-05, + "loss": 1.0597, + "step": 2739 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927162489133694e-05, + "loss": 1.0837, + "step": 2740 + }, + { + "epoch": 0.07, + "learning_rate": 1.9927066609284563e-05, + "loss": 1.0846, + "step": 2741 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926970666602047e-05, + "loss": 1.0732, + "step": 2742 + }, + { + "epoch": 0.07, + "learning_rate": 1.992687466108677e-05, + "loss": 1.0844, + "step": 2743 + }, + { + "epoch": 0.07, + "learning_rate": 1.992677859273932e-05, + "loss": 1.0576, + "step": 2744 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926682461560322e-05, + "loss": 1.0993, + "step": 2745 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926586267550378e-05, + "loss": 1.0954, + "step": 2746 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926490010710097e-05, + "loss": 1.0592, + "step": 2747 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926393691040084e-05, + "loss": 1.0891, + "step": 2748 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926297308540958e-05, + "loss": 1.0698, + "step": 2749 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926200863213322e-05, + "loss": 1.0594, + "step": 2750 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926104355057783e-05, + "loss": 1.0696, + "step": 2751 + }, + { + "epoch": 0.07, + "learning_rate": 1.9926007784074965e-05, + "loss": 1.0579, + "step": 2752 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925911150265466e-05, + "loss": 1.0909, + "step": 2753 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925814453629904e-05, + "loss": 1.1051, + "step": 2754 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925717694168892e-05, + "loss": 1.0984, + "step": 2755 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925620871883036e-05, + "loss": 1.0813, + "step": 2756 + }, + { + "epoch": 0.07, + "learning_rate": 1.992552398677296e-05, + "loss": 1.0711, + "step": 2757 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925427038839267e-05, + "loss": 1.089, + "step": 2758 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925330028082574e-05, + "loss": 1.0672, + "step": 2759 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925232954503496e-05, + "loss": 1.109, + "step": 2760 + }, + { + "epoch": 0.07, + "learning_rate": 1.9925135818102644e-05, + "loss": 1.1002, + "step": 2761 + }, + { + "epoch": 0.07, + "learning_rate": 1.992503861888064e-05, + "loss": 1.0934, + "step": 2762 + }, + { + "epoch": 0.07, + "learning_rate": 1.992494135683809e-05, + "loss": 1.0392, + "step": 2763 + }, + { + "epoch": 0.07, + "learning_rate": 1.992484403197562e-05, + "loss": 1.0691, + "step": 2764 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924746644293834e-05, + "loss": 1.0706, + "step": 2765 + }, + { + "epoch": 0.07, + "learning_rate": 1.992464919379336e-05, + "loss": 1.0635, + "step": 2766 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924551680474806e-05, + "loss": 1.0707, + "step": 2767 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924454104338792e-05, + "loss": 1.0591, + "step": 2768 + }, + { + "epoch": 0.07, + "learning_rate": 1.992435646538594e-05, + "loss": 1.0738, + "step": 2769 + }, + { + "epoch": 0.07, + "learning_rate": 1.992425876361686e-05, + "loss": 1.0707, + "step": 2770 + }, + { + "epoch": 0.07, + "learning_rate": 1.9924160999032178e-05, + "loss": 1.1009, + "step": 2771 + }, + { + "epoch": 0.07, + "learning_rate": 1.992406317163251e-05, + "loss": 1.0964, + "step": 2772 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923965281418472e-05, + "loss": 1.0601, + "step": 2773 + }, + { + "epoch": 0.07, + "learning_rate": 1.992386732839069e-05, + "loss": 1.1001, + "step": 2774 + }, + { + "epoch": 0.07, + "learning_rate": 1.992376931254978e-05, + "loss": 1.0747, + "step": 2775 + }, + { + "epoch": 0.07, + "learning_rate": 1.992367123389636e-05, + "loss": 1.0462, + "step": 2776 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923573092431054e-05, + "loss": 1.1033, + "step": 2777 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923474888154485e-05, + "loss": 1.0751, + "step": 2778 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923376621067268e-05, + "loss": 1.0628, + "step": 2779 + }, + { + "epoch": 0.07, + "learning_rate": 1.992327829117003e-05, + "loss": 1.1244, + "step": 2780 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923179898463397e-05, + "loss": 1.1173, + "step": 2781 + }, + { + "epoch": 0.07, + "learning_rate": 1.9923081442947986e-05, + "loss": 1.0651, + "step": 2782 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922982924624417e-05, + "loss": 1.0481, + "step": 2783 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922884343493325e-05, + "loss": 1.0792, + "step": 2784 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922785699555323e-05, + "loss": 1.1012, + "step": 2785 + }, + { + "epoch": 0.07, + "learning_rate": 1.992268699281104e-05, + "loss": 1.1056, + "step": 2786 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922588223261102e-05, + "loss": 1.0773, + "step": 2787 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922489390906133e-05, + "loss": 1.0943, + "step": 2788 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922390495746754e-05, + "loss": 1.0794, + "step": 2789 + }, + { + "epoch": 0.07, + "learning_rate": 1.99222915377836e-05, + "loss": 1.1, + "step": 2790 + }, + { + "epoch": 0.07, + "learning_rate": 1.9922192517017288e-05, + "loss": 1.1071, + "step": 2791 + }, + { + "epoch": 0.07, + "learning_rate": 1.992209343344845e-05, + "loss": 1.074, + "step": 2792 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921994287077715e-05, + "loss": 1.0719, + "step": 2793 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921895077905706e-05, + "loss": 1.1107, + "step": 2794 + }, + { + "epoch": 0.07, + "learning_rate": 1.992179580593305e-05, + "loss": 1.0919, + "step": 2795 + }, + { + "epoch": 0.07, + "learning_rate": 1.992169647116038e-05, + "loss": 1.0595, + "step": 2796 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921597073588326e-05, + "loss": 1.0767, + "step": 2797 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921497613217512e-05, + "loss": 1.0715, + "step": 2798 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921398090048564e-05, + "loss": 1.0808, + "step": 2799 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921298504082124e-05, + "loss": 1.0785, + "step": 2800 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921198855318816e-05, + "loss": 1.0652, + "step": 2801 + }, + { + "epoch": 0.07, + "learning_rate": 1.9921099143759267e-05, + "loss": 1.0977, + "step": 2802 + }, + { + "epoch": 0.07, + "learning_rate": 1.992099936940411e-05, + "loss": 1.0706, + "step": 2803 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920899532253984e-05, + "loss": 1.0588, + "step": 2804 + }, + { + "epoch": 0.07, + "learning_rate": 1.992079963230951e-05, + "loss": 1.0587, + "step": 2805 + }, + { + "epoch": 0.07, + "learning_rate": 1.992069966957133e-05, + "loss": 1.0697, + "step": 2806 + }, + { + "epoch": 0.07, + "learning_rate": 1.992059964404007e-05, + "loss": 1.075, + "step": 2807 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920499555716362e-05, + "loss": 1.1066, + "step": 2808 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920399404600846e-05, + "loss": 1.0578, + "step": 2809 + }, + { + "epoch": 0.07, + "learning_rate": 1.992029919069415e-05, + "loss": 1.0784, + "step": 2810 + }, + { + "epoch": 0.07, + "learning_rate": 1.9920198913996916e-05, + "loss": 1.0688, + "step": 2811 + }, + { + "epoch": 0.07, + "learning_rate": 1.992009857450977e-05, + "loss": 1.1241, + "step": 2812 + }, + { + "epoch": 0.07, + "learning_rate": 1.991999817223335e-05, + "loss": 1.0869, + "step": 2813 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919897707168293e-05, + "loss": 1.0931, + "step": 2814 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919797179315235e-05, + "loss": 1.0983, + "step": 2815 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919696588674807e-05, + "loss": 1.097, + "step": 2816 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919595935247656e-05, + "loss": 1.0714, + "step": 2817 + }, + { + "epoch": 0.07, + "learning_rate": 1.991949521903441e-05, + "loss": 1.0758, + "step": 2818 + }, + { + "epoch": 0.07, + "learning_rate": 1.991939444003571e-05, + "loss": 1.0535, + "step": 2819 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919293598252193e-05, + "loss": 1.0981, + "step": 2820 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919192693684502e-05, + "loss": 1.1086, + "step": 2821 + }, + { + "epoch": 0.07, + "learning_rate": 1.9919091726333265e-05, + "loss": 1.0709, + "step": 2822 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918990696199134e-05, + "loss": 1.0459, + "step": 2823 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918889603282737e-05, + "loss": 1.0808, + "step": 2824 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918788447584722e-05, + "loss": 1.0538, + "step": 2825 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918687229105723e-05, + "loss": 1.089, + "step": 2826 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918585947846383e-05, + "loss": 1.0204, + "step": 2827 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918484603807347e-05, + "loss": 1.0732, + "step": 2828 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918383196989254e-05, + "loss": 1.0542, + "step": 2829 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918281727392745e-05, + "loss": 1.0801, + "step": 2830 + }, + { + "epoch": 0.07, + "learning_rate": 1.991818019501846e-05, + "loss": 1.0664, + "step": 2831 + }, + { + "epoch": 0.07, + "learning_rate": 1.9918078599867044e-05, + "loss": 1.1199, + "step": 2832 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917976941939143e-05, + "loss": 1.0741, + "step": 2833 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917875221235397e-05, + "loss": 1.089, + "step": 2834 + }, + { + "epoch": 0.07, + "learning_rate": 1.991777343775645e-05, + "loss": 1.0984, + "step": 2835 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917671591502943e-05, + "loss": 1.0728, + "step": 2836 + }, + { + "epoch": 0.07, + "learning_rate": 1.991756968247553e-05, + "loss": 1.0605, + "step": 2837 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917467710674846e-05, + "loss": 1.0889, + "step": 2838 + }, + { + "epoch": 0.07, + "learning_rate": 1.991736567610154e-05, + "loss": 1.0995, + "step": 2839 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917263578756266e-05, + "loss": 1.0761, + "step": 2840 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917161418639654e-05, + "loss": 1.0868, + "step": 2841 + }, + { + "epoch": 0.07, + "learning_rate": 1.9917059195752364e-05, + "loss": 1.074, + "step": 2842 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916956910095037e-05, + "loss": 1.0881, + "step": 2843 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916854561668323e-05, + "loss": 1.0574, + "step": 2844 + }, + { + "epoch": 0.07, + "learning_rate": 1.991675215047287e-05, + "loss": 1.0163, + "step": 2845 + }, + { + "epoch": 0.07, + "learning_rate": 1.991664967650932e-05, + "loss": 1.1204, + "step": 2846 + }, + { + "epoch": 0.07, + "learning_rate": 1.991654713977833e-05, + "loss": 1.0361, + "step": 2847 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916444540280546e-05, + "loss": 1.079, + "step": 2848 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916341878016616e-05, + "loss": 1.0635, + "step": 2849 + }, + { + "epoch": 0.07, + "learning_rate": 1.991623915298719e-05, + "loss": 1.0203, + "step": 2850 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916136365192918e-05, + "loss": 1.1002, + "step": 2851 + }, + { + "epoch": 0.07, + "learning_rate": 1.9916033514634458e-05, + "loss": 1.0743, + "step": 2852 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915930601312446e-05, + "loss": 1.0765, + "step": 2853 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915827625227545e-05, + "loss": 1.0789, + "step": 2854 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915724586380407e-05, + "loss": 1.0802, + "step": 2855 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915621484771677e-05, + "loss": 1.0791, + "step": 2856 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915518320402012e-05, + "loss": 1.0982, + "step": 2857 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915415093272063e-05, + "loss": 1.084, + "step": 2858 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915311803382486e-05, + "loss": 1.0652, + "step": 2859 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915208450733937e-05, + "loss": 1.0883, + "step": 2860 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915105035327063e-05, + "loss": 1.0986, + "step": 2861 + }, + { + "epoch": 0.07, + "learning_rate": 1.9915001557162522e-05, + "loss": 1.1047, + "step": 2862 + }, + { + "epoch": 0.07, + "learning_rate": 1.991489801624097e-05, + "loss": 1.0662, + "step": 2863 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914794412563062e-05, + "loss": 1.1393, + "step": 2864 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914690746129453e-05, + "loss": 1.0923, + "step": 2865 + }, + { + "epoch": 0.07, + "learning_rate": 1.99145870169408e-05, + "loss": 1.0961, + "step": 2866 + }, + { + "epoch": 0.07, + "learning_rate": 1.991448322499776e-05, + "loss": 1.1072, + "step": 2867 + }, + { + "epoch": 0.07, + "learning_rate": 1.9914379370300988e-05, + "loss": 1.0501, + "step": 2868 + }, + { + "epoch": 0.07, + "learning_rate": 1.991427545285114e-05, + "loss": 1.0919, + "step": 2869 + }, + { + "epoch": 0.07, + "learning_rate": 1.991417147264888e-05, + "loss": 1.1093, + "step": 2870 + }, + { + "epoch": 0.07, + "learning_rate": 1.991406742969486e-05, + "loss": 1.0557, + "step": 2871 + }, + { + "epoch": 0.07, + "learning_rate": 1.991396332398974e-05, + "loss": 1.0682, + "step": 2872 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913859155534182e-05, + "loss": 1.0384, + "step": 2873 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913754924328844e-05, + "loss": 1.0939, + "step": 2874 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913650630374383e-05, + "loss": 1.0557, + "step": 2875 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913546273671464e-05, + "loss": 1.088, + "step": 2876 + }, + { + "epoch": 0.07, + "learning_rate": 1.991344185422074e-05, + "loss": 1.1026, + "step": 2877 + }, + { + "epoch": 0.07, + "learning_rate": 1.991333737202288e-05, + "loss": 1.0546, + "step": 2878 + }, + { + "epoch": 0.07, + "learning_rate": 1.9913232827078537e-05, + "loss": 1.0828, + "step": 2879 + }, + { + "epoch": 0.07, + "learning_rate": 1.991312821938838e-05, + "loss": 1.0595, + "step": 2880 + }, + { + "epoch": 0.07, + "learning_rate": 1.991302354895307e-05, + "loss": 1.0797, + "step": 2881 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912918815773267e-05, + "loss": 1.0967, + "step": 2882 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912814019849636e-05, + "loss": 1.1037, + "step": 2883 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912709161182836e-05, + "loss": 1.0953, + "step": 2884 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912604239773537e-05, + "loss": 1.0631, + "step": 2885 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912499255622397e-05, + "loss": 1.0712, + "step": 2886 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912394208730087e-05, + "loss": 1.0763, + "step": 2887 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912289099097267e-05, + "loss": 1.0902, + "step": 2888 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912183926724605e-05, + "loss": 1.0616, + "step": 2889 + }, + { + "epoch": 0.07, + "learning_rate": 1.9912078691612768e-05, + "loss": 1.0804, + "step": 2890 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911973393762413e-05, + "loss": 1.073, + "step": 2891 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911868033174215e-05, + "loss": 1.0839, + "step": 2892 + }, + { + "epoch": 0.07, + "learning_rate": 1.991176260984884e-05, + "loss": 1.0703, + "step": 2893 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911657123786954e-05, + "loss": 1.0627, + "step": 2894 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911551574989224e-05, + "loss": 1.0815, + "step": 2895 + }, + { + "epoch": 0.07, + "learning_rate": 1.991144596345632e-05, + "loss": 1.0962, + "step": 2896 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911340289188913e-05, + "loss": 1.0726, + "step": 2897 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911234552187663e-05, + "loss": 1.0641, + "step": 2898 + }, + { + "epoch": 0.07, + "learning_rate": 1.9911128752453245e-05, + "loss": 1.0555, + "step": 2899 + }, + { + "epoch": 0.07, + "learning_rate": 1.991102288998633e-05, + "loss": 1.0702, + "step": 2900 + }, + { + "epoch": 0.07, + "learning_rate": 1.991091696478759e-05, + "loss": 1.0762, + "step": 2901 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910810976857684e-05, + "loss": 1.0906, + "step": 2902 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910704926197294e-05, + "loss": 1.047, + "step": 2903 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910598812807087e-05, + "loss": 1.072, + "step": 2904 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910492636687734e-05, + "loss": 1.0604, + "step": 2905 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910386397839913e-05, + "loss": 1.1155, + "step": 2906 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910280096264288e-05, + "loss": 1.0778, + "step": 2907 + }, + { + "epoch": 0.07, + "learning_rate": 1.991017373196154e-05, + "loss": 1.1062, + "step": 2908 + }, + { + "epoch": 0.07, + "learning_rate": 1.9910067304932333e-05, + "loss": 1.0664, + "step": 2909 + }, + { + "epoch": 0.07, + "learning_rate": 1.990996081517735e-05, + "loss": 1.0591, + "step": 2910 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909854262697258e-05, + "loss": 1.0751, + "step": 2911 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909747647492736e-05, + "loss": 1.0953, + "step": 2912 + }, + { + "epoch": 0.07, + "learning_rate": 1.990964096956445e-05, + "loss": 1.1075, + "step": 2913 + }, + { + "epoch": 0.07, + "learning_rate": 1.990953422891309e-05, + "loss": 1.098, + "step": 2914 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909427425539324e-05, + "loss": 1.092, + "step": 2915 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909320559443822e-05, + "loss": 1.0579, + "step": 2916 + }, + { + "epoch": 0.07, + "learning_rate": 1.990921363062727e-05, + "loss": 1.0559, + "step": 2917 + }, + { + "epoch": 0.07, + "learning_rate": 1.9909106639090342e-05, + "loss": 1.078, + "step": 2918 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908999584833708e-05, + "loss": 1.085, + "step": 2919 + }, + { + "epoch": 0.07, + "learning_rate": 1.990889246785806e-05, + "loss": 1.111, + "step": 2920 + }, + { + "epoch": 0.07, + "learning_rate": 1.990878528816406e-05, + "loss": 1.097, + "step": 2921 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908678045752398e-05, + "loss": 1.0573, + "step": 2922 + }, + { + "epoch": 0.07, + "learning_rate": 1.990857074062375e-05, + "loss": 1.0798, + "step": 2923 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908463372778793e-05, + "loss": 1.0863, + "step": 2924 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908355942218208e-05, + "loss": 1.0824, + "step": 2925 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908248448942676e-05, + "loss": 1.0387, + "step": 2926 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908140892952873e-05, + "loss": 1.1029, + "step": 2927 + }, + { + "epoch": 0.07, + "learning_rate": 1.9908033274249482e-05, + "loss": 1.0475, + "step": 2928 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907925592833187e-05, + "loss": 1.0756, + "step": 2929 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907817848704668e-05, + "loss": 1.0453, + "step": 2930 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907710041864607e-05, + "loss": 1.0716, + "step": 2931 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907602172313686e-05, + "loss": 1.1291, + "step": 2932 + }, + { + "epoch": 0.07, + "learning_rate": 1.990749424005259e-05, + "loss": 1.0645, + "step": 2933 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907386245081994e-05, + "loss": 1.1075, + "step": 2934 + }, + { + "epoch": 0.07, + "learning_rate": 1.990727818740259e-05, + "loss": 1.1035, + "step": 2935 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907170067015057e-05, + "loss": 1.0944, + "step": 2936 + }, + { + "epoch": 0.07, + "learning_rate": 1.9907061883920088e-05, + "loss": 1.1066, + "step": 2937 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906953638118355e-05, + "loss": 1.088, + "step": 2938 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906845329610552e-05, + "loss": 1.0534, + "step": 2939 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906736958397365e-05, + "loss": 1.0877, + "step": 2940 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906628524479474e-05, + "loss": 1.0539, + "step": 2941 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906520027857566e-05, + "loss": 1.0483, + "step": 2942 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906411468532334e-05, + "loss": 1.0511, + "step": 2943 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906302846504458e-05, + "loss": 1.0646, + "step": 2944 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906194161774636e-05, + "loss": 1.0842, + "step": 2945 + }, + { + "epoch": 0.07, + "learning_rate": 1.9906085414343542e-05, + "loss": 1.1105, + "step": 2946 + }, + { + "epoch": 0.07, + "learning_rate": 1.990597660421187e-05, + "loss": 1.0882, + "step": 2947 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905867731380314e-05, + "loss": 1.0688, + "step": 2948 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905758795849555e-05, + "loss": 1.1004, + "step": 2949 + }, + { + "epoch": 0.07, + "learning_rate": 1.990564979762029e-05, + "loss": 1.1234, + "step": 2950 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905540736693205e-05, + "loss": 1.0888, + "step": 2951 + }, + { + "epoch": 0.07, + "learning_rate": 1.990543161306899e-05, + "loss": 1.0498, + "step": 2952 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905322426748332e-05, + "loss": 1.055, + "step": 2953 + }, + { + "epoch": 0.07, + "learning_rate": 1.990521317773193e-05, + "loss": 1.1395, + "step": 2954 + }, + { + "epoch": 0.07, + "learning_rate": 1.9905103866020473e-05, + "loss": 1.1137, + "step": 2955 + }, + { + "epoch": 0.07, + "learning_rate": 1.990499449161465e-05, + "loss": 1.0825, + "step": 2956 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904885054515154e-05, + "loss": 1.0238, + "step": 2957 + }, + { + "epoch": 0.07, + "learning_rate": 1.990477555472268e-05, + "loss": 1.0744, + "step": 2958 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904665992237922e-05, + "loss": 1.0584, + "step": 2959 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904556367061567e-05, + "loss": 1.0727, + "step": 2960 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904446679194315e-05, + "loss": 1.0852, + "step": 2961 + }, + { + "epoch": 0.07, + "learning_rate": 1.990433692863686e-05, + "loss": 1.076, + "step": 2962 + }, + { + "epoch": 0.07, + "learning_rate": 1.9904227115389895e-05, + "loss": 1.0889, + "step": 2963 + }, + { + "epoch": 0.07, + "learning_rate": 1.990411723945411e-05, + "loss": 1.0671, + "step": 2964 + }, + { + "epoch": 0.07, + "learning_rate": 1.990400730083021e-05, + "loss": 1.131, + "step": 2965 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903897299518886e-05, + "loss": 1.048, + "step": 2966 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903787235520837e-05, + "loss": 1.0753, + "step": 2967 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903677108836752e-05, + "loss": 1.0409, + "step": 2968 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903566919467336e-05, + "loss": 1.0583, + "step": 2969 + }, + { + "epoch": 0.07, + "learning_rate": 1.990345666741328e-05, + "loss": 1.0966, + "step": 2970 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903346352675292e-05, + "loss": 1.0702, + "step": 2971 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903235975254062e-05, + "loss": 1.0697, + "step": 2972 + }, + { + "epoch": 0.07, + "learning_rate": 1.990312553515029e-05, + "loss": 1.0698, + "step": 2973 + }, + { + "epoch": 0.07, + "learning_rate": 1.9903015032364675e-05, + "loss": 1.0724, + "step": 2974 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902904466897916e-05, + "loss": 1.1059, + "step": 2975 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902793838750716e-05, + "loss": 1.0842, + "step": 2976 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902683147923772e-05, + "loss": 1.0868, + "step": 2977 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902572394417787e-05, + "loss": 1.074, + "step": 2978 + }, + { + "epoch": 0.07, + "learning_rate": 1.990246157823346e-05, + "loss": 1.0952, + "step": 2979 + }, + { + "epoch": 0.07, + "learning_rate": 1.990235069937149e-05, + "loss": 1.0824, + "step": 2980 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902239757832586e-05, + "loss": 1.0513, + "step": 2981 + }, + { + "epoch": 0.07, + "learning_rate": 1.990212875361744e-05, + "loss": 1.0698, + "step": 2982 + }, + { + "epoch": 0.07, + "learning_rate": 1.9902017686726768e-05, + "loss": 1.0335, + "step": 2983 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901906557161264e-05, + "loss": 1.0946, + "step": 2984 + }, + { + "epoch": 0.07, + "learning_rate": 1.990179536492163e-05, + "loss": 1.0571, + "step": 2985 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901684110008576e-05, + "loss": 1.0454, + "step": 2986 + }, + { + "epoch": 0.07, + "learning_rate": 1.99015727924228e-05, + "loss": 1.0893, + "step": 2987 + }, + { + "epoch": 0.07, + "learning_rate": 1.990146141216501e-05, + "loss": 1.0514, + "step": 2988 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901349969235915e-05, + "loss": 1.0395, + "step": 2989 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901238463636212e-05, + "loss": 1.0644, + "step": 2990 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901126895366612e-05, + "loss": 1.0267, + "step": 2991 + }, + { + "epoch": 0.07, + "learning_rate": 1.9901015264427822e-05, + "loss": 1.0863, + "step": 2992 + }, + { + "epoch": 0.07, + "learning_rate": 1.990090357082054e-05, + "loss": 1.0635, + "step": 2993 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900791814545486e-05, + "loss": 1.0437, + "step": 2994 + }, + { + "epoch": 0.07, + "learning_rate": 1.990067999560336e-05, + "loss": 1.0728, + "step": 2995 + }, + { + "epoch": 0.07, + "learning_rate": 1.990056811399487e-05, + "loss": 1.077, + "step": 2996 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900456169720723e-05, + "loss": 1.0962, + "step": 2997 + }, + { + "epoch": 0.07, + "learning_rate": 1.990034416278163e-05, + "loss": 1.0714, + "step": 2998 + }, + { + "epoch": 0.07, + "learning_rate": 1.99002320931783e-05, + "loss": 1.0621, + "step": 2999 + }, + { + "epoch": 0.07, + "learning_rate": 1.990011996091144e-05, + "loss": 1.0959, + "step": 3000 + }, + { + "epoch": 0.07, + "learning_rate": 1.9900007765981762e-05, + "loss": 1.0698, + "step": 3001 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899895508389977e-05, + "loss": 1.0746, + "step": 3002 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899783188136794e-05, + "loss": 1.0676, + "step": 3003 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899670805222923e-05, + "loss": 1.1068, + "step": 3004 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899558359649075e-05, + "loss": 1.0999, + "step": 3005 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899445851415968e-05, + "loss": 1.0842, + "step": 3006 + }, + { + "epoch": 0.07, + "learning_rate": 1.9899333280524303e-05, + "loss": 1.0665, + "step": 3007 + }, + { + "epoch": 0.07, + "learning_rate": 1.98992206469748e-05, + "loss": 1.0648, + "step": 3008 + }, + { + "epoch": 0.07, + "learning_rate": 1.989910795076817e-05, + "loss": 1.067, + "step": 3009 + }, + { + "epoch": 0.07, + "learning_rate": 1.989899519190513e-05, + "loss": 1.0517, + "step": 3010 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898882370386387e-05, + "loss": 1.0531, + "step": 3011 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898769486212665e-05, + "loss": 1.0679, + "step": 3012 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898656539384665e-05, + "loss": 1.0733, + "step": 3013 + }, + { + "epoch": 0.07, + "learning_rate": 1.989854352990311e-05, + "loss": 1.1153, + "step": 3014 + }, + { + "epoch": 0.07, + "learning_rate": 1.989843045776872e-05, + "loss": 1.0645, + "step": 3015 + }, + { + "epoch": 0.07, + "learning_rate": 1.98983173229822e-05, + "loss": 1.0728, + "step": 3016 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898204125544272e-05, + "loss": 1.0866, + "step": 3017 + }, + { + "epoch": 0.07, + "learning_rate": 1.9898090865455654e-05, + "loss": 1.0903, + "step": 3018 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897977542717054e-05, + "loss": 1.0713, + "step": 3019 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897864157329204e-05, + "loss": 1.0801, + "step": 3020 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897750709292808e-05, + "loss": 1.1368, + "step": 3021 + }, + { + "epoch": 0.07, + "learning_rate": 1.989763719860859e-05, + "loss": 1.1199, + "step": 3022 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897523625277272e-05, + "loss": 1.0386, + "step": 3023 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897409989299566e-05, + "loss": 1.0855, + "step": 3024 + }, + { + "epoch": 0.07, + "learning_rate": 1.989729629067619e-05, + "loss": 1.0876, + "step": 3025 + }, + { + "epoch": 0.07, + "learning_rate": 1.9897182529407875e-05, + "loss": 1.0818, + "step": 3026 + }, + { + "epoch": 0.07, + "learning_rate": 1.989706870549533e-05, + "loss": 1.0962, + "step": 3027 + }, + { + "epoch": 0.07, + "learning_rate": 1.989695481893928e-05, + "loss": 1.0506, + "step": 3028 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896840869740445e-05, + "loss": 1.109, + "step": 3029 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896726857899546e-05, + "loss": 1.0331, + "step": 3030 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896612783417305e-05, + "loss": 1.0901, + "step": 3031 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896498646294446e-05, + "loss": 1.1012, + "step": 3032 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896384446531688e-05, + "loss": 1.0788, + "step": 3033 + }, + { + "epoch": 0.07, + "learning_rate": 1.9896270184129753e-05, + "loss": 1.1046, + "step": 3034 + }, + { + "epoch": 0.07, + "learning_rate": 1.989615585908937e-05, + "loss": 1.0322, + "step": 3035 + }, + { + "epoch": 0.07, + "learning_rate": 1.989604147141126e-05, + "loss": 1.0635, + "step": 3036 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895927021096143e-05, + "loss": 1.0593, + "step": 3037 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895812508144748e-05, + "loss": 1.0663, + "step": 3038 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895697932557798e-05, + "loss": 1.0334, + "step": 3039 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895583294336017e-05, + "loss": 1.079, + "step": 3040 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895468593480135e-05, + "loss": 1.0587, + "step": 3041 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895353829990872e-05, + "loss": 1.1097, + "step": 3042 + }, + { + "epoch": 0.07, + "learning_rate": 1.989523900386896e-05, + "loss": 1.0754, + "step": 3043 + }, + { + "epoch": 0.07, + "learning_rate": 1.9895124115115123e-05, + "loss": 1.0689, + "step": 3044 + }, + { + "epoch": 0.07, + "learning_rate": 1.989500916373009e-05, + "loss": 1.058, + "step": 3045 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894894149714586e-05, + "loss": 1.0914, + "step": 3046 + }, + { + "epoch": 0.07, + "learning_rate": 1.989477907306934e-05, + "loss": 1.0397, + "step": 3047 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894663933795077e-05, + "loss": 1.0697, + "step": 3048 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894548731892532e-05, + "loss": 1.0713, + "step": 3049 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894433467362434e-05, + "loss": 1.0606, + "step": 3050 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894318140205508e-05, + "loss": 1.0965, + "step": 3051 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894202750422486e-05, + "loss": 1.0858, + "step": 3052 + }, + { + "epoch": 0.07, + "learning_rate": 1.9894087298014096e-05, + "loss": 1.0995, + "step": 3053 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893971782981075e-05, + "loss": 1.0701, + "step": 3054 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893856205324146e-05, + "loss": 1.066, + "step": 3055 + }, + { + "epoch": 0.08, + "learning_rate": 1.989374056504405e-05, + "loss": 1.0711, + "step": 3056 + }, + { + "epoch": 0.08, + "learning_rate": 1.989362486214151e-05, + "loss": 1.0993, + "step": 3057 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893509096617264e-05, + "loss": 1.0829, + "step": 3058 + }, + { + "epoch": 0.08, + "learning_rate": 1.989339326847204e-05, + "loss": 1.0558, + "step": 3059 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893277377706576e-05, + "loss": 1.09, + "step": 3060 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893161424321603e-05, + "loss": 1.0617, + "step": 3061 + }, + { + "epoch": 0.08, + "learning_rate": 1.9893045408317854e-05, + "loss": 1.0938, + "step": 3062 + }, + { + "epoch": 0.08, + "learning_rate": 1.989292932969607e-05, + "loss": 1.0776, + "step": 3063 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892813188456976e-05, + "loss": 1.0626, + "step": 3064 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892696984601312e-05, + "loss": 1.016, + "step": 3065 + }, + { + "epoch": 0.08, + "learning_rate": 1.989258071812981e-05, + "loss": 1.0857, + "step": 3066 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892464389043216e-05, + "loss": 1.077, + "step": 3067 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892347997342255e-05, + "loss": 1.077, + "step": 3068 + }, + { + "epoch": 0.08, + "learning_rate": 1.989223154302767e-05, + "loss": 1.0516, + "step": 3069 + }, + { + "epoch": 0.08, + "learning_rate": 1.9892115026100195e-05, + "loss": 1.034, + "step": 3070 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891998446560564e-05, + "loss": 1.0816, + "step": 3071 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891881804409524e-05, + "loss": 1.1, + "step": 3072 + }, + { + "epoch": 0.08, + "learning_rate": 1.989176509964781e-05, + "loss": 1.0885, + "step": 3073 + }, + { + "epoch": 0.08, + "learning_rate": 1.989164833227616e-05, + "loss": 1.0498, + "step": 3074 + }, + { + "epoch": 0.08, + "learning_rate": 1.989153150229531e-05, + "loss": 1.0992, + "step": 3075 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891414609706e-05, + "loss": 1.0777, + "step": 3076 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891297654508978e-05, + "loss": 1.0622, + "step": 3077 + }, + { + "epoch": 0.08, + "learning_rate": 1.9891180636704975e-05, + "loss": 1.1101, + "step": 3078 + }, + { + "epoch": 0.08, + "learning_rate": 1.989106355629473e-05, + "loss": 1.0727, + "step": 3079 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890946413278995e-05, + "loss": 1.0765, + "step": 3080 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890829207658508e-05, + "loss": 1.0509, + "step": 3081 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890711939434002e-05, + "loss": 1.0768, + "step": 3082 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890594608606226e-05, + "loss": 1.0914, + "step": 3083 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890477215175922e-05, + "loss": 1.0475, + "step": 3084 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890359759143836e-05, + "loss": 1.0522, + "step": 3085 + }, + { + "epoch": 0.08, + "learning_rate": 1.9890242240510705e-05, + "loss": 1.0723, + "step": 3086 + }, + { + "epoch": 0.08, + "learning_rate": 1.989012465927728e-05, + "loss": 1.1278, + "step": 3087 + }, + { + "epoch": 0.08, + "learning_rate": 1.98900070154443e-05, + "loss": 1.0578, + "step": 3088 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889889309012507e-05, + "loss": 1.0873, + "step": 3089 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889771539982655e-05, + "loss": 1.0352, + "step": 3090 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889653708355485e-05, + "loss": 1.0658, + "step": 3091 + }, + { + "epoch": 0.08, + "learning_rate": 1.988953581413174e-05, + "loss": 1.0954, + "step": 3092 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889417857312173e-05, + "loss": 1.0617, + "step": 3093 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889299837897524e-05, + "loss": 1.0688, + "step": 3094 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889181755888544e-05, + "loss": 1.0623, + "step": 3095 + }, + { + "epoch": 0.08, + "learning_rate": 1.9889063611285976e-05, + "loss": 1.059, + "step": 3096 + }, + { + "epoch": 0.08, + "learning_rate": 1.988894540409057e-05, + "loss": 1.0423, + "step": 3097 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888827134303076e-05, + "loss": 1.1036, + "step": 3098 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888708801924238e-05, + "loss": 1.0618, + "step": 3099 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888590406954815e-05, + "loss": 1.1106, + "step": 3100 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888471949395546e-05, + "loss": 1.0696, + "step": 3101 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888353429247186e-05, + "loss": 1.0722, + "step": 3102 + }, + { + "epoch": 0.08, + "learning_rate": 1.988823484651048e-05, + "loss": 1.0797, + "step": 3103 + }, + { + "epoch": 0.08, + "learning_rate": 1.9888116201186185e-05, + "loss": 1.0535, + "step": 3104 + }, + { + "epoch": 0.08, + "learning_rate": 1.988799749327505e-05, + "loss": 1.0576, + "step": 3105 + }, + { + "epoch": 0.08, + "learning_rate": 1.988787872277782e-05, + "loss": 1.1174, + "step": 3106 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887759889695257e-05, + "loss": 1.0864, + "step": 3107 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887640994028107e-05, + "loss": 1.0735, + "step": 3108 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887522035777126e-05, + "loss": 1.0506, + "step": 3109 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887403014943064e-05, + "loss": 1.0701, + "step": 3110 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887283931526674e-05, + "loss": 1.0799, + "step": 3111 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887164785528712e-05, + "loss": 1.0517, + "step": 3112 + }, + { + "epoch": 0.08, + "learning_rate": 1.9887045576949934e-05, + "loss": 1.047, + "step": 3113 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886926305791087e-05, + "loss": 1.058, + "step": 3114 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886806972052934e-05, + "loss": 1.0305, + "step": 3115 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886687575736224e-05, + "loss": 1.0629, + "step": 3116 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886568116841713e-05, + "loss": 1.0059, + "step": 3117 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886448595370166e-05, + "loss": 1.0946, + "step": 3118 + }, + { + "epoch": 0.08, + "learning_rate": 1.988632901132233e-05, + "loss": 1.0746, + "step": 3119 + }, + { + "epoch": 0.08, + "learning_rate": 1.9886209364698964e-05, + "loss": 1.0589, + "step": 3120 + }, + { + "epoch": 0.08, + "learning_rate": 1.988608965550083e-05, + "loss": 1.0644, + "step": 3121 + }, + { + "epoch": 0.08, + "learning_rate": 1.988596988372868e-05, + "loss": 1.0695, + "step": 3122 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885850049383276e-05, + "loss": 1.0318, + "step": 3123 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885730152465372e-05, + "loss": 1.0364, + "step": 3124 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885610192975734e-05, + "loss": 1.0945, + "step": 3125 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885490170915115e-05, + "loss": 1.0717, + "step": 3126 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885370086284274e-05, + "loss": 1.0865, + "step": 3127 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885249939083976e-05, + "loss": 1.0973, + "step": 3128 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885129729314978e-05, + "loss": 1.0929, + "step": 3129 + }, + { + "epoch": 0.08, + "learning_rate": 1.9885009456978044e-05, + "loss": 1.0816, + "step": 3130 + }, + { + "epoch": 0.08, + "learning_rate": 1.988488912207393e-05, + "loss": 1.0984, + "step": 3131 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884768724603404e-05, + "loss": 1.0592, + "step": 3132 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884648264567225e-05, + "loss": 1.0533, + "step": 3133 + }, + { + "epoch": 0.08, + "learning_rate": 1.988452774196615e-05, + "loss": 1.0592, + "step": 3134 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884407156800955e-05, + "loss": 1.0535, + "step": 3135 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884286509072392e-05, + "loss": 1.0859, + "step": 3136 + }, + { + "epoch": 0.08, + "learning_rate": 1.9884165798781227e-05, + "loss": 1.0747, + "step": 3137 + }, + { + "epoch": 0.08, + "learning_rate": 1.988404502592823e-05, + "loss": 1.0493, + "step": 3138 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883924190514156e-05, + "loss": 1.0948, + "step": 3139 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883803292539775e-05, + "loss": 1.0795, + "step": 3140 + }, + { + "epoch": 0.08, + "learning_rate": 1.988368233200585e-05, + "loss": 1.0663, + "step": 3141 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883561308913154e-05, + "loss": 1.0938, + "step": 3142 + }, + { + "epoch": 0.08, + "learning_rate": 1.988344022326244e-05, + "loss": 1.1008, + "step": 3143 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883319075054487e-05, + "loss": 1.072, + "step": 3144 + }, + { + "epoch": 0.08, + "learning_rate": 1.988319786429006e-05, + "loss": 1.0705, + "step": 3145 + }, + { + "epoch": 0.08, + "learning_rate": 1.9883076590969917e-05, + "loss": 1.0813, + "step": 3146 + }, + { + "epoch": 0.08, + "learning_rate": 1.988295525509483e-05, + "loss": 1.0794, + "step": 3147 + }, + { + "epoch": 0.08, + "learning_rate": 1.988283385666557e-05, + "loss": 1.0652, + "step": 3148 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882712395682905e-05, + "loss": 1.1157, + "step": 3149 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882590872147604e-05, + "loss": 1.0457, + "step": 3150 + }, + { + "epoch": 0.08, + "learning_rate": 1.988246928606043e-05, + "loss": 1.0517, + "step": 3151 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882347637422162e-05, + "loss": 1.0822, + "step": 3152 + }, + { + "epoch": 0.08, + "learning_rate": 1.9882225926233563e-05, + "loss": 1.0893, + "step": 3153 + }, + { + "epoch": 0.08, + "learning_rate": 1.988210415249541e-05, + "loss": 1.0518, + "step": 3154 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881982316208463e-05, + "loss": 1.0928, + "step": 3155 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881860417373503e-05, + "loss": 1.0894, + "step": 3156 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881738455991297e-05, + "loss": 1.0677, + "step": 3157 + }, + { + "epoch": 0.08, + "learning_rate": 1.988161643206262e-05, + "loss": 1.0829, + "step": 3158 + }, + { + "epoch": 0.08, + "learning_rate": 1.988149434558824e-05, + "loss": 1.0431, + "step": 3159 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881372196568933e-05, + "loss": 1.0691, + "step": 3160 + }, + { + "epoch": 0.08, + "learning_rate": 1.988124998500547e-05, + "loss": 1.1013, + "step": 3161 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881127710898627e-05, + "loss": 1.0456, + "step": 3162 + }, + { + "epoch": 0.08, + "learning_rate": 1.9881005374249177e-05, + "loss": 1.0402, + "step": 3163 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880882975057893e-05, + "loss": 1.057, + "step": 3164 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880760513325553e-05, + "loss": 1.1017, + "step": 3165 + }, + { + "epoch": 0.08, + "learning_rate": 1.988063798905293e-05, + "loss": 1.0815, + "step": 3166 + }, + { + "epoch": 0.08, + "learning_rate": 1.98805154022408e-05, + "loss": 1.075, + "step": 3167 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880392752889936e-05, + "loss": 1.0809, + "step": 3168 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880270041001116e-05, + "loss": 1.082, + "step": 3169 + }, + { + "epoch": 0.08, + "learning_rate": 1.9880147266575117e-05, + "loss": 1.0803, + "step": 3170 + }, + { + "epoch": 0.08, + "learning_rate": 1.988002442961272e-05, + "loss": 1.0818, + "step": 3171 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879901530114698e-05, + "loss": 1.0787, + "step": 3172 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879778568081832e-05, + "loss": 1.0859, + "step": 3173 + }, + { + "epoch": 0.08, + "learning_rate": 1.987965554351489e-05, + "loss": 1.1008, + "step": 3174 + }, + { + "epoch": 0.08, + "learning_rate": 1.987953245641467e-05, + "loss": 1.1134, + "step": 3175 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879409306781933e-05, + "loss": 1.0777, + "step": 3176 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879286094617466e-05, + "loss": 1.0981, + "step": 3177 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879162819922052e-05, + "loss": 1.0788, + "step": 3178 + }, + { + "epoch": 0.08, + "learning_rate": 1.9879039482696464e-05, + "loss": 1.0892, + "step": 3179 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878916082941487e-05, + "loss": 1.0971, + "step": 3180 + }, + { + "epoch": 0.08, + "learning_rate": 1.98787926206579e-05, + "loss": 1.062, + "step": 3181 + }, + { + "epoch": 0.08, + "learning_rate": 1.987866909584649e-05, + "loss": 1.0961, + "step": 3182 + }, + { + "epoch": 0.08, + "learning_rate": 1.987854550850803e-05, + "loss": 1.0892, + "step": 3183 + }, + { + "epoch": 0.08, + "learning_rate": 1.987842185864331e-05, + "loss": 1.0377, + "step": 3184 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878298146253107e-05, + "loss": 1.0518, + "step": 3185 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878174371338212e-05, + "loss": 1.0697, + "step": 3186 + }, + { + "epoch": 0.08, + "learning_rate": 1.9878050533899397e-05, + "loss": 1.0553, + "step": 3187 + }, + { + "epoch": 0.08, + "learning_rate": 1.987792663393745e-05, + "loss": 1.0518, + "step": 3188 + }, + { + "epoch": 0.08, + "learning_rate": 1.987780267145316e-05, + "loss": 1.0974, + "step": 3189 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877678646447313e-05, + "loss": 1.0568, + "step": 3190 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877554558920686e-05, + "loss": 1.0275, + "step": 3191 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877430408874065e-05, + "loss": 1.0762, + "step": 3192 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877306196308244e-05, + "loss": 1.0787, + "step": 3193 + }, + { + "epoch": 0.08, + "learning_rate": 1.9877181921224e-05, + "loss": 1.0444, + "step": 3194 + }, + { + "epoch": 0.08, + "learning_rate": 1.987705758362213e-05, + "loss": 1.0881, + "step": 3195 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876933183503406e-05, + "loss": 1.073, + "step": 3196 + }, + { + "epoch": 0.08, + "learning_rate": 1.987680872086863e-05, + "loss": 1.0699, + "step": 3197 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876684195718582e-05, + "loss": 1.05, + "step": 3198 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876559608054053e-05, + "loss": 1.0411, + "step": 3199 + }, + { + "epoch": 0.08, + "learning_rate": 1.987643495787583e-05, + "loss": 1.1073, + "step": 3200 + }, + { + "epoch": 0.08, + "learning_rate": 1.98763102451847e-05, + "loss": 1.0428, + "step": 3201 + }, + { + "epoch": 0.08, + "learning_rate": 1.987618546998146e-05, + "loss": 1.1001, + "step": 3202 + }, + { + "epoch": 0.08, + "learning_rate": 1.9876060632266888e-05, + "loss": 1.104, + "step": 3203 + }, + { + "epoch": 0.08, + "learning_rate": 1.987593573204179e-05, + "loss": 1.1017, + "step": 3204 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875810769306942e-05, + "loss": 1.037, + "step": 3205 + }, + { + "epoch": 0.08, + "learning_rate": 1.987568574406314e-05, + "loss": 1.0691, + "step": 3206 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875560656311177e-05, + "loss": 1.1058, + "step": 3207 + }, + { + "epoch": 0.08, + "learning_rate": 1.987543550605184e-05, + "loss": 1.0851, + "step": 3208 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875310293285927e-05, + "loss": 1.0552, + "step": 3209 + }, + { + "epoch": 0.08, + "learning_rate": 1.987518501801423e-05, + "loss": 1.0939, + "step": 3210 + }, + { + "epoch": 0.08, + "learning_rate": 1.9875059680237537e-05, + "loss": 1.0596, + "step": 3211 + }, + { + "epoch": 0.08, + "learning_rate": 1.987493427995665e-05, + "loss": 1.0732, + "step": 3212 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874808817172353e-05, + "loss": 1.0763, + "step": 3213 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874683291885444e-05, + "loss": 1.0918, + "step": 3214 + }, + { + "epoch": 0.08, + "learning_rate": 1.987455770409672e-05, + "loss": 1.0787, + "step": 3215 + }, + { + "epoch": 0.08, + "learning_rate": 1.987443205380697e-05, + "loss": 1.05, + "step": 3216 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874306341016996e-05, + "loss": 1.0326, + "step": 3217 + }, + { + "epoch": 0.08, + "learning_rate": 1.987418056572759e-05, + "loss": 1.0696, + "step": 3218 + }, + { + "epoch": 0.08, + "learning_rate": 1.9874054727939547e-05, + "loss": 1.0643, + "step": 3219 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873928827653666e-05, + "loss": 1.0637, + "step": 3220 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873802864870744e-05, + "loss": 1.0662, + "step": 3221 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873676839591577e-05, + "loss": 1.0803, + "step": 3222 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873550751816966e-05, + "loss": 1.1088, + "step": 3223 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873424601547702e-05, + "loss": 1.0664, + "step": 3224 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873298388784592e-05, + "loss": 1.039, + "step": 3225 + }, + { + "epoch": 0.08, + "learning_rate": 1.9873172113528427e-05, + "loss": 1.0586, + "step": 3226 + }, + { + "epoch": 0.08, + "learning_rate": 1.987304577578001e-05, + "loss": 1.0605, + "step": 3227 + }, + { + "epoch": 0.08, + "learning_rate": 1.987291937554014e-05, + "loss": 1.0769, + "step": 3228 + }, + { + "epoch": 0.08, + "learning_rate": 1.987279291280962e-05, + "loss": 1.0474, + "step": 3229 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872666387589247e-05, + "loss": 1.0626, + "step": 3230 + }, + { + "epoch": 0.08, + "learning_rate": 1.987253979987982e-05, + "loss": 1.0439, + "step": 3231 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872413149682145e-05, + "loss": 1.0608, + "step": 3232 + }, + { + "epoch": 0.08, + "learning_rate": 1.987228643699702e-05, + "loss": 1.0621, + "step": 3233 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872159661825253e-05, + "loss": 1.1057, + "step": 3234 + }, + { + "epoch": 0.08, + "learning_rate": 1.9872032824167638e-05, + "loss": 1.046, + "step": 3235 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871905924024985e-05, + "loss": 1.0761, + "step": 3236 + }, + { + "epoch": 0.08, + "learning_rate": 1.987177896139809e-05, + "loss": 1.0776, + "step": 3237 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871651936287764e-05, + "loss": 1.0625, + "step": 3238 + }, + { + "epoch": 0.08, + "learning_rate": 1.987152484869481e-05, + "loss": 1.0465, + "step": 3239 + }, + { + "epoch": 0.08, + "learning_rate": 1.9871397698620028e-05, + "loss": 1.0809, + "step": 3240 + }, + { + "epoch": 0.08, + "learning_rate": 1.987127048606422e-05, + "loss": 1.0718, + "step": 3241 + }, + { + "epoch": 0.08, + "learning_rate": 1.98711432110282e-05, + "loss": 1.1037, + "step": 3242 + }, + { + "epoch": 0.08, + "learning_rate": 1.987101587351277e-05, + "loss": 1.08, + "step": 3243 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870888473518736e-05, + "loss": 1.0824, + "step": 3244 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870761011046905e-05, + "loss": 1.0853, + "step": 3245 + }, + { + "epoch": 0.08, + "learning_rate": 1.987063348609808e-05, + "loss": 1.056, + "step": 3246 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870505898673072e-05, + "loss": 1.0619, + "step": 3247 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870378248772687e-05, + "loss": 1.0732, + "step": 3248 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870250536397737e-05, + "loss": 1.0627, + "step": 3249 + }, + { + "epoch": 0.08, + "learning_rate": 1.9870122761549025e-05, + "loss": 1.0705, + "step": 3250 + }, + { + "epoch": 0.08, + "learning_rate": 1.986999492422736e-05, + "loss": 1.0994, + "step": 3251 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869867024433557e-05, + "loss": 1.0705, + "step": 3252 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869739062168417e-05, + "loss": 1.0812, + "step": 3253 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869611037432755e-05, + "loss": 1.0726, + "step": 3254 + }, + { + "epoch": 0.08, + "learning_rate": 1.986948295022738e-05, + "loss": 1.0584, + "step": 3255 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869354800553106e-05, + "loss": 1.0681, + "step": 3256 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869226588410736e-05, + "loss": 1.0531, + "step": 3257 + }, + { + "epoch": 0.08, + "learning_rate": 1.9869098313801093e-05, + "loss": 1.1225, + "step": 3258 + }, + { + "epoch": 0.08, + "learning_rate": 1.986896997672498e-05, + "loss": 1.0511, + "step": 3259 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868841577183216e-05, + "loss": 1.0715, + "step": 3260 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868713115176607e-05, + "loss": 1.0808, + "step": 3261 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868584590705966e-05, + "loss": 1.0606, + "step": 3262 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868456003772113e-05, + "loss": 1.0837, + "step": 3263 + }, + { + "epoch": 0.08, + "learning_rate": 1.986832735437586e-05, + "loss": 1.0854, + "step": 3264 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868198642518016e-05, + "loss": 1.0947, + "step": 3265 + }, + { + "epoch": 0.08, + "learning_rate": 1.9868069868199398e-05, + "loss": 1.0654, + "step": 3266 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867941031420828e-05, + "loss": 1.0651, + "step": 3267 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867812132183113e-05, + "loss": 1.0946, + "step": 3268 + }, + { + "epoch": 0.08, + "learning_rate": 1.986768317048707e-05, + "loss": 1.0534, + "step": 3269 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867554146333517e-05, + "loss": 1.0798, + "step": 3270 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867425059723274e-05, + "loss": 1.0863, + "step": 3271 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867295910657155e-05, + "loss": 1.073, + "step": 3272 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867166699135976e-05, + "loss": 1.0621, + "step": 3273 + }, + { + "epoch": 0.08, + "learning_rate": 1.9867037425160554e-05, + "loss": 1.0722, + "step": 3274 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866908088731712e-05, + "loss": 1.0774, + "step": 3275 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866778689850264e-05, + "loss": 1.0794, + "step": 3276 + }, + { + "epoch": 0.08, + "learning_rate": 1.986664922851703e-05, + "loss": 1.0698, + "step": 3277 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866519704732834e-05, + "loss": 1.0579, + "step": 3278 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866390118498492e-05, + "loss": 1.0462, + "step": 3279 + }, + { + "epoch": 0.08, + "learning_rate": 1.986626046981482e-05, + "loss": 1.0689, + "step": 3280 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866130758682645e-05, + "loss": 1.0783, + "step": 3281 + }, + { + "epoch": 0.08, + "learning_rate": 1.9866000985102786e-05, + "loss": 1.0555, + "step": 3282 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865871149076067e-05, + "loss": 1.0956, + "step": 3283 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865741250603304e-05, + "loss": 1.0849, + "step": 3284 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865611289685324e-05, + "loss": 1.0684, + "step": 3285 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865481266322947e-05, + "loss": 1.0717, + "step": 3286 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865351180517e-05, + "loss": 1.0812, + "step": 3287 + }, + { + "epoch": 0.08, + "learning_rate": 1.98652210322683e-05, + "loss": 1.0445, + "step": 3288 + }, + { + "epoch": 0.08, + "learning_rate": 1.9865090821577676e-05, + "loss": 1.0678, + "step": 3289 + }, + { + "epoch": 0.08, + "learning_rate": 1.986496054844595e-05, + "loss": 1.0393, + "step": 3290 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864830212873944e-05, + "loss": 1.0685, + "step": 3291 + }, + { + "epoch": 0.08, + "learning_rate": 1.986469981486249e-05, + "loss": 1.054, + "step": 3292 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864569354412407e-05, + "loss": 1.1022, + "step": 3293 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864438831524522e-05, + "loss": 1.0839, + "step": 3294 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864308246199664e-05, + "loss": 1.0885, + "step": 3295 + }, + { + "epoch": 0.08, + "learning_rate": 1.9864177598438656e-05, + "loss": 1.0567, + "step": 3296 + }, + { + "epoch": 0.08, + "learning_rate": 1.986404688824233e-05, + "loss": 1.0634, + "step": 3297 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863916115611505e-05, + "loss": 1.0758, + "step": 3298 + }, + { + "epoch": 0.08, + "learning_rate": 1.986378528054702e-05, + "loss": 1.0891, + "step": 3299 + }, + { + "epoch": 0.08, + "learning_rate": 1.986365438304969e-05, + "loss": 1.091, + "step": 3300 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863523423120354e-05, + "loss": 1.0731, + "step": 3301 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863392400759837e-05, + "loss": 1.0607, + "step": 3302 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863261315968968e-05, + "loss": 1.1134, + "step": 3303 + }, + { + "epoch": 0.08, + "learning_rate": 1.9863130168748575e-05, + "loss": 1.044, + "step": 3304 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862998959099493e-05, + "loss": 1.0878, + "step": 3305 + }, + { + "epoch": 0.08, + "learning_rate": 1.986286768702255e-05, + "loss": 1.0623, + "step": 3306 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862736352518576e-05, + "loss": 1.071, + "step": 3307 + }, + { + "epoch": 0.08, + "learning_rate": 1.98626049555884e-05, + "loss": 1.0865, + "step": 3308 + }, + { + "epoch": 0.08, + "learning_rate": 1.986247349623286e-05, + "loss": 1.0612, + "step": 3309 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862341974452784e-05, + "loss": 1.0312, + "step": 3310 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862210390249006e-05, + "loss": 1.0589, + "step": 3311 + }, + { + "epoch": 0.08, + "learning_rate": 1.9862078743622355e-05, + "loss": 1.0409, + "step": 3312 + }, + { + "epoch": 0.08, + "learning_rate": 1.986194703457367e-05, + "loss": 1.1203, + "step": 3313 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861815263103783e-05, + "loss": 1.0721, + "step": 3314 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861683429213526e-05, + "loss": 1.0817, + "step": 3315 + }, + { + "epoch": 0.08, + "learning_rate": 1.986155153290373e-05, + "loss": 1.0442, + "step": 3316 + }, + { + "epoch": 0.08, + "learning_rate": 1.986141957417524e-05, + "loss": 1.0661, + "step": 3317 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861287553028882e-05, + "loss": 1.0642, + "step": 3318 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861155469465496e-05, + "loss": 1.0758, + "step": 3319 + }, + { + "epoch": 0.08, + "learning_rate": 1.9861023323485916e-05, + "loss": 1.0768, + "step": 3320 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860891115090985e-05, + "loss": 1.0738, + "step": 3321 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860758844281527e-05, + "loss": 1.0907, + "step": 3322 + }, + { + "epoch": 0.08, + "learning_rate": 1.986062651105839e-05, + "loss": 1.0676, + "step": 3323 + }, + { + "epoch": 0.08, + "learning_rate": 1.986049411542241e-05, + "loss": 1.0786, + "step": 3324 + }, + { + "epoch": 0.08, + "learning_rate": 1.986036165737442e-05, + "loss": 1.053, + "step": 3325 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860229136915262e-05, + "loss": 1.0353, + "step": 3326 + }, + { + "epoch": 0.08, + "learning_rate": 1.9860096554045774e-05, + "loss": 1.0776, + "step": 3327 + }, + { + "epoch": 0.08, + "learning_rate": 1.98599639087668e-05, + "loss": 1.0349, + "step": 3328 + }, + { + "epoch": 0.08, + "learning_rate": 1.985983120107917e-05, + "loss": 1.0904, + "step": 3329 + }, + { + "epoch": 0.08, + "learning_rate": 1.985969843098373e-05, + "loss": 1.0438, + "step": 3330 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859565598481323e-05, + "loss": 1.067, + "step": 3331 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859432703572782e-05, + "loss": 1.0753, + "step": 3332 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859299746258953e-05, + "loss": 1.0515, + "step": 3333 + }, + { + "epoch": 0.08, + "learning_rate": 1.985916672654068e-05, + "loss": 1.1082, + "step": 3334 + }, + { + "epoch": 0.08, + "learning_rate": 1.9859033644418798e-05, + "loss": 1.0609, + "step": 3335 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858900499894157e-05, + "loss": 1.0838, + "step": 3336 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858767292967592e-05, + "loss": 1.0736, + "step": 3337 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858634023639954e-05, + "loss": 1.052, + "step": 3338 + }, + { + "epoch": 0.08, + "learning_rate": 1.985850069191208e-05, + "loss": 1.0621, + "step": 3339 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858367297784816e-05, + "loss": 1.1135, + "step": 3340 + }, + { + "epoch": 0.08, + "learning_rate": 1.9858233841259008e-05, + "loss": 1.0557, + "step": 3341 + }, + { + "epoch": 0.08, + "learning_rate": 1.98581003223355e-05, + "loss": 1.0968, + "step": 3342 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857966741015135e-05, + "loss": 1.0817, + "step": 3343 + }, + { + "epoch": 0.08, + "learning_rate": 1.985783309729876e-05, + "loss": 1.0667, + "step": 3344 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857699391187224e-05, + "loss": 1.0708, + "step": 3345 + }, + { + "epoch": 0.08, + "learning_rate": 1.985756562268137e-05, + "loss": 1.0609, + "step": 3346 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857431791782044e-05, + "loss": 1.0919, + "step": 3347 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857297898490096e-05, + "loss": 1.0209, + "step": 3348 + }, + { + "epoch": 0.08, + "learning_rate": 1.985716394280637e-05, + "loss": 1.0629, + "step": 3349 + }, + { + "epoch": 0.08, + "learning_rate": 1.9857029924731712e-05, + "loss": 1.0976, + "step": 3350 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856895844266977e-05, + "loss": 1.0467, + "step": 3351 + }, + { + "epoch": 0.08, + "learning_rate": 1.985676170141301e-05, + "loss": 1.0798, + "step": 3352 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856627496170663e-05, + "loss": 1.0577, + "step": 3353 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856493228540778e-05, + "loss": 1.0731, + "step": 3354 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856358898524216e-05, + "loss": 1.0467, + "step": 3355 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856224506121817e-05, + "loss": 1.0621, + "step": 3356 + }, + { + "epoch": 0.08, + "learning_rate": 1.9856090051334436e-05, + "loss": 1.0881, + "step": 3357 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855955534162925e-05, + "loss": 1.0788, + "step": 3358 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855820954608132e-05, + "loss": 1.1159, + "step": 3359 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855686312670912e-05, + "loss": 1.0732, + "step": 3360 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855551608352117e-05, + "loss": 1.0628, + "step": 3361 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855416841652597e-05, + "loss": 1.085, + "step": 3362 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855282012573208e-05, + "loss": 1.068, + "step": 3363 + }, + { + "epoch": 0.08, + "learning_rate": 1.9855147121114803e-05, + "loss": 1.0578, + "step": 3364 + }, + { + "epoch": 0.08, + "learning_rate": 1.985501216727823e-05, + "loss": 1.0467, + "step": 3365 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854877151064354e-05, + "loss": 1.0583, + "step": 3366 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854742072474022e-05, + "loss": 1.0633, + "step": 3367 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854606931508093e-05, + "loss": 1.0327, + "step": 3368 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854471728167416e-05, + "loss": 1.0898, + "step": 3369 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854336462452853e-05, + "loss": 1.0182, + "step": 3370 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854201134365256e-05, + "loss": 1.0853, + "step": 3371 + }, + { + "epoch": 0.08, + "learning_rate": 1.9854065743905488e-05, + "loss": 1.0356, + "step": 3372 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853930291074397e-05, + "loss": 1.1384, + "step": 3373 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853794775872845e-05, + "loss": 1.0508, + "step": 3374 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853659198301693e-05, + "loss": 1.0597, + "step": 3375 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853523558361793e-05, + "loss": 1.0731, + "step": 3376 + }, + { + "epoch": 0.08, + "learning_rate": 1.985338785605401e-05, + "loss": 1.0583, + "step": 3377 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853252091379196e-05, + "loss": 1.0794, + "step": 3378 + }, + { + "epoch": 0.08, + "learning_rate": 1.9853116264338212e-05, + "loss": 1.0592, + "step": 3379 + }, + { + "epoch": 0.08, + "learning_rate": 1.985298037493192e-05, + "loss": 1.0811, + "step": 3380 + }, + { + "epoch": 0.08, + "learning_rate": 1.985284442316118e-05, + "loss": 1.0565, + "step": 3381 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852708409026856e-05, + "loss": 1.0345, + "step": 3382 + }, + { + "epoch": 0.08, + "learning_rate": 1.98525723325298e-05, + "loss": 1.0678, + "step": 3383 + }, + { + "epoch": 0.08, + "learning_rate": 1.985243619367088e-05, + "loss": 1.0609, + "step": 3384 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852299992450957e-05, + "loss": 1.0388, + "step": 3385 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852163728870887e-05, + "loss": 1.0337, + "step": 3386 + }, + { + "epoch": 0.08, + "learning_rate": 1.9852027402931543e-05, + "loss": 1.082, + "step": 3387 + }, + { + "epoch": 0.08, + "learning_rate": 1.985189101463378e-05, + "loss": 1.0591, + "step": 3388 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851754563978463e-05, + "loss": 1.0529, + "step": 3389 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851618050966458e-05, + "loss": 1.0833, + "step": 3390 + }, + { + "epoch": 0.08, + "learning_rate": 1.9851481475598625e-05, + "loss": 1.0397, + "step": 3391 + }, + { + "epoch": 0.08, + "learning_rate": 1.985134483787583e-05, + "loss": 1.0502, + "step": 3392 + }, + { + "epoch": 0.08, + "learning_rate": 1.985120813779894e-05, + "loss": 1.0146, + "step": 3393 + }, + { + "epoch": 0.08, + "learning_rate": 1.985107137536882e-05, + "loss": 1.0716, + "step": 3394 + }, + { + "epoch": 0.08, + "learning_rate": 1.985093455058633e-05, + "loss": 1.0404, + "step": 3395 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850797663452347e-05, + "loss": 1.0636, + "step": 3396 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850660713967727e-05, + "loss": 1.0553, + "step": 3397 + }, + { + "epoch": 0.08, + "learning_rate": 1.985052370213334e-05, + "loss": 1.0465, + "step": 3398 + }, + { + "epoch": 0.08, + "learning_rate": 1.9850386627950055e-05, + "loss": 1.0725, + "step": 3399 + }, + { + "epoch": 0.08, + "learning_rate": 1.985024949141874e-05, + "loss": 1.0832, + "step": 3400 + }, + { + "epoch": 0.08, + "learning_rate": 1.985011229254026e-05, + "loss": 1.0672, + "step": 3401 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849975031315485e-05, + "loss": 1.0634, + "step": 3402 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849837707745286e-05, + "loss": 1.0789, + "step": 3403 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849700321830525e-05, + "loss": 1.0406, + "step": 3404 + }, + { + "epoch": 0.08, + "learning_rate": 1.984956287357208e-05, + "loss": 1.074, + "step": 3405 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849425362970818e-05, + "loss": 1.036, + "step": 3406 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849287790027607e-05, + "loss": 1.0525, + "step": 3407 + }, + { + "epoch": 0.08, + "learning_rate": 1.9849150154743317e-05, + "loss": 1.1081, + "step": 3408 + }, + { + "epoch": 0.08, + "learning_rate": 1.984901245711883e-05, + "loss": 1.0698, + "step": 3409 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848874697155002e-05, + "loss": 1.0884, + "step": 3410 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848736874852715e-05, + "loss": 1.0459, + "step": 3411 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848598990212835e-05, + "loss": 1.0737, + "step": 3412 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848461043236243e-05, + "loss": 1.0941, + "step": 3413 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848323033923806e-05, + "loss": 1.0677, + "step": 3414 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848184962276398e-05, + "loss": 1.0373, + "step": 3415 + }, + { + "epoch": 0.08, + "learning_rate": 1.9848046828294894e-05, + "loss": 1.0797, + "step": 3416 + }, + { + "epoch": 0.08, + "learning_rate": 1.984790863198017e-05, + "loss": 1.0195, + "step": 3417 + }, + { + "epoch": 0.08, + "learning_rate": 1.98477703733331e-05, + "loss": 1.0827, + "step": 3418 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847632052354555e-05, + "loss": 1.0631, + "step": 3419 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847493669045416e-05, + "loss": 1.0556, + "step": 3420 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847355223406555e-05, + "loss": 1.0236, + "step": 3421 + }, + { + "epoch": 0.08, + "learning_rate": 1.984721671543885e-05, + "loss": 1.074, + "step": 3422 + }, + { + "epoch": 0.08, + "learning_rate": 1.9847078145143176e-05, + "loss": 1.0569, + "step": 3423 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846939512520417e-05, + "loss": 1.0525, + "step": 3424 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846800817571438e-05, + "loss": 1.0829, + "step": 3425 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846662060297128e-05, + "loss": 1.0655, + "step": 3426 + }, + { + "epoch": 0.08, + "learning_rate": 1.984652324069836e-05, + "loss": 1.0602, + "step": 3427 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846384358776017e-05, + "loss": 1.0681, + "step": 3428 + }, + { + "epoch": 0.08, + "learning_rate": 1.984624541453097e-05, + "loss": 1.0431, + "step": 3429 + }, + { + "epoch": 0.08, + "learning_rate": 1.9846106407964108e-05, + "loss": 1.0494, + "step": 3430 + }, + { + "epoch": 0.08, + "learning_rate": 1.98459673390763e-05, + "loss": 1.0572, + "step": 3431 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845828207868437e-05, + "loss": 1.1006, + "step": 3432 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845689014341395e-05, + "loss": 1.0618, + "step": 3433 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845549758496055e-05, + "loss": 1.0905, + "step": 3434 + }, + { + "epoch": 0.08, + "learning_rate": 1.98454104403333e-05, + "loss": 1.0496, + "step": 3435 + }, + { + "epoch": 0.08, + "learning_rate": 1.9845271059854006e-05, + "loss": 1.0582, + "step": 3436 + }, + { + "epoch": 0.08, + "learning_rate": 1.984513161705906e-05, + "loss": 1.0705, + "step": 3437 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844992111949345e-05, + "loss": 1.0951, + "step": 3438 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844852544525744e-05, + "loss": 1.0509, + "step": 3439 + }, + { + "epoch": 0.08, + "learning_rate": 1.984471291478914e-05, + "loss": 1.0463, + "step": 3440 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844573222740415e-05, + "loss": 1.111, + "step": 3441 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844433468380454e-05, + "loss": 1.0947, + "step": 3442 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844293651710145e-05, + "loss": 1.0823, + "step": 3443 + }, + { + "epoch": 0.08, + "learning_rate": 1.9844153772730367e-05, + "loss": 1.0194, + "step": 3444 + }, + { + "epoch": 0.08, + "learning_rate": 1.984401383144201e-05, + "loss": 1.046, + "step": 3445 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843873827845958e-05, + "loss": 1.0854, + "step": 3446 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843733761943095e-05, + "loss": 1.0473, + "step": 3447 + }, + { + "epoch": 0.08, + "learning_rate": 1.984359363373431e-05, + "loss": 1.0624, + "step": 3448 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843453443220495e-05, + "loss": 1.0843, + "step": 3449 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843313190402526e-05, + "loss": 1.0706, + "step": 3450 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843172875281303e-05, + "loss": 1.0571, + "step": 3451 + }, + { + "epoch": 0.08, + "learning_rate": 1.9843032497857703e-05, + "loss": 1.0612, + "step": 3452 + }, + { + "epoch": 0.08, + "learning_rate": 1.984289205813262e-05, + "loss": 1.0784, + "step": 3453 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842751556106944e-05, + "loss": 1.0782, + "step": 3454 + }, + { + "epoch": 0.08, + "learning_rate": 1.984261099178156e-05, + "loss": 1.0721, + "step": 3455 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842470365157364e-05, + "loss": 1.0761, + "step": 3456 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842329676235242e-05, + "loss": 1.0759, + "step": 3457 + }, + { + "epoch": 0.08, + "learning_rate": 1.9842188925016084e-05, + "loss": 1.0833, + "step": 3458 + }, + { + "epoch": 0.08, + "learning_rate": 1.984204811150078e-05, + "loss": 1.0722, + "step": 3459 + }, + { + "epoch": 0.08, + "learning_rate": 1.9841907235690224e-05, + "loss": 1.0701, + "step": 3460 + }, + { + "epoch": 0.09, + "learning_rate": 1.984176629758531e-05, + "loss": 1.1098, + "step": 3461 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841625297186925e-05, + "loss": 1.0636, + "step": 3462 + }, + { + "epoch": 0.09, + "learning_rate": 1.984148423449596e-05, + "loss": 1.0833, + "step": 3463 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841343109513317e-05, + "loss": 1.0421, + "step": 3464 + }, + { + "epoch": 0.09, + "learning_rate": 1.984120192223988e-05, + "loss": 1.064, + "step": 3465 + }, + { + "epoch": 0.09, + "learning_rate": 1.9841060672676547e-05, + "loss": 1.0492, + "step": 3466 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840919360824215e-05, + "loss": 1.0661, + "step": 3467 + }, + { + "epoch": 0.09, + "learning_rate": 1.984077798668377e-05, + "loss": 1.047, + "step": 3468 + }, + { + "epoch": 0.09, + "learning_rate": 1.984063655025611e-05, + "loss": 1.0944, + "step": 3469 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840495051542135e-05, + "loss": 1.0734, + "step": 3470 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840353490542735e-05, + "loss": 1.0994, + "step": 3471 + }, + { + "epoch": 0.09, + "learning_rate": 1.984021186725881e-05, + "loss": 1.0465, + "step": 3472 + }, + { + "epoch": 0.09, + "learning_rate": 1.9840070181691252e-05, + "loss": 1.0609, + "step": 3473 + }, + { + "epoch": 0.09, + "learning_rate": 1.983992843384096e-05, + "loss": 1.0379, + "step": 3474 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839786623708836e-05, + "loss": 1.074, + "step": 3475 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839644751295774e-05, + "loss": 1.034, + "step": 3476 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839502816602666e-05, + "loss": 1.0707, + "step": 3477 + }, + { + "epoch": 0.09, + "learning_rate": 1.983936081963042e-05, + "loss": 1.0784, + "step": 3478 + }, + { + "epoch": 0.09, + "learning_rate": 1.983921876037993e-05, + "loss": 1.0204, + "step": 3479 + }, + { + "epoch": 0.09, + "learning_rate": 1.9839076638852092e-05, + "loss": 1.0847, + "step": 3480 + }, + { + "epoch": 0.09, + "learning_rate": 1.983893445504781e-05, + "loss": 1.0757, + "step": 3481 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838792208967984e-05, + "loss": 1.0906, + "step": 3482 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838649900613515e-05, + "loss": 1.0614, + "step": 3483 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838507529985302e-05, + "loss": 1.0733, + "step": 3484 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838365097084244e-05, + "loss": 1.0797, + "step": 3485 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838222601911248e-05, + "loss": 1.069, + "step": 3486 + }, + { + "epoch": 0.09, + "learning_rate": 1.9838080044467208e-05, + "loss": 1.0756, + "step": 3487 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837937424753036e-05, + "loss": 1.0971, + "step": 3488 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837794742769624e-05, + "loss": 1.075, + "step": 3489 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837651998517886e-05, + "loss": 1.0791, + "step": 3490 + }, + { + "epoch": 0.09, + "learning_rate": 1.983750919199872e-05, + "loss": 1.1068, + "step": 3491 + }, + { + "epoch": 0.09, + "learning_rate": 1.983736632321303e-05, + "loss": 1.0768, + "step": 3492 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837223392161716e-05, + "loss": 1.0793, + "step": 3493 + }, + { + "epoch": 0.09, + "learning_rate": 1.9837080398845693e-05, + "loss": 1.0303, + "step": 3494 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836937343265856e-05, + "loss": 1.0725, + "step": 3495 + }, + { + "epoch": 0.09, + "learning_rate": 1.983679422542312e-05, + "loss": 1.0475, + "step": 3496 + }, + { + "epoch": 0.09, + "learning_rate": 1.983665104531838e-05, + "loss": 1.0406, + "step": 3497 + }, + { + "epoch": 0.09, + "learning_rate": 1.983650780295255e-05, + "loss": 1.0386, + "step": 3498 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836364498326536e-05, + "loss": 1.0195, + "step": 3499 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836221131441244e-05, + "loss": 1.0322, + "step": 3500 + }, + { + "epoch": 0.09, + "learning_rate": 1.9836077702297577e-05, + "loss": 1.0673, + "step": 3501 + }, + { + "epoch": 0.09, + "learning_rate": 1.983593421089645e-05, + "loss": 1.0564, + "step": 3502 + }, + { + "epoch": 0.09, + "learning_rate": 1.983579065723877e-05, + "loss": 1.0864, + "step": 3503 + }, + { + "epoch": 0.09, + "learning_rate": 1.983564704132544e-05, + "loss": 1.0673, + "step": 3504 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835503363157375e-05, + "loss": 1.0457, + "step": 3505 + }, + { + "epoch": 0.09, + "learning_rate": 1.983535962273548e-05, + "loss": 1.0457, + "step": 3506 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835215820060668e-05, + "loss": 1.0571, + "step": 3507 + }, + { + "epoch": 0.09, + "learning_rate": 1.9835071955133847e-05, + "loss": 1.0451, + "step": 3508 + }, + { + "epoch": 0.09, + "learning_rate": 1.983492802795593e-05, + "loss": 1.0476, + "step": 3509 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834784038527827e-05, + "loss": 1.0843, + "step": 3510 + }, + { + "epoch": 0.09, + "learning_rate": 1.983463998685045e-05, + "loss": 1.0917, + "step": 3511 + }, + { + "epoch": 0.09, + "learning_rate": 1.983449587292471e-05, + "loss": 1.0767, + "step": 3512 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834351696751515e-05, + "loss": 1.0534, + "step": 3513 + }, + { + "epoch": 0.09, + "learning_rate": 1.9834207458331787e-05, + "loss": 1.0594, + "step": 3514 + }, + { + "epoch": 0.09, + "learning_rate": 1.983406315766643e-05, + "loss": 1.0391, + "step": 3515 + }, + { + "epoch": 0.09, + "learning_rate": 1.983391879475636e-05, + "loss": 1.0662, + "step": 3516 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833774369602493e-05, + "loss": 1.0879, + "step": 3517 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833629882205745e-05, + "loss": 1.0664, + "step": 3518 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833485332567024e-05, + "loss": 1.0933, + "step": 3519 + }, + { + "epoch": 0.09, + "learning_rate": 1.983334072068725e-05, + "loss": 1.1013, + "step": 3520 + }, + { + "epoch": 0.09, + "learning_rate": 1.9833196046567335e-05, + "loss": 1.0816, + "step": 3521 + }, + { + "epoch": 0.09, + "learning_rate": 1.98330513102082e-05, + "loss": 1.0618, + "step": 3522 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832906511610754e-05, + "loss": 1.0549, + "step": 3523 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832761650775916e-05, + "loss": 1.0469, + "step": 3524 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832616727704607e-05, + "loss": 1.0598, + "step": 3525 + }, + { + "epoch": 0.09, + "learning_rate": 1.983247174239774e-05, + "loss": 1.0793, + "step": 3526 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832326694856232e-05, + "loss": 1.0973, + "step": 3527 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832181585081003e-05, + "loss": 1.0372, + "step": 3528 + }, + { + "epoch": 0.09, + "learning_rate": 1.9832036413072972e-05, + "loss": 1.0838, + "step": 3529 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831891178833058e-05, + "loss": 1.052, + "step": 3530 + }, + { + "epoch": 0.09, + "learning_rate": 1.983174588236218e-05, + "loss": 1.0866, + "step": 3531 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831600523661255e-05, + "loss": 1.0869, + "step": 3532 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831455102731207e-05, + "loss": 1.0578, + "step": 3533 + }, + { + "epoch": 0.09, + "learning_rate": 1.9831309619572955e-05, + "loss": 1.0372, + "step": 3534 + }, + { + "epoch": 0.09, + "learning_rate": 1.983116407418742e-05, + "loss": 1.1122, + "step": 3535 + }, + { + "epoch": 0.09, + "learning_rate": 1.983101846657552e-05, + "loss": 1.0952, + "step": 3536 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830872796738178e-05, + "loss": 1.0983, + "step": 3537 + }, + { + "epoch": 0.09, + "learning_rate": 1.983072706467632e-05, + "loss": 1.0783, + "step": 3538 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830581270390868e-05, + "loss": 1.0771, + "step": 3539 + }, + { + "epoch": 0.09, + "learning_rate": 1.983043541388274e-05, + "loss": 1.0944, + "step": 3540 + }, + { + "epoch": 0.09, + "learning_rate": 1.983028949515286e-05, + "loss": 1.114, + "step": 3541 + }, + { + "epoch": 0.09, + "learning_rate": 1.9830143514202155e-05, + "loss": 1.0696, + "step": 3542 + }, + { + "epoch": 0.09, + "learning_rate": 1.982999747103155e-05, + "loss": 1.0619, + "step": 3543 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829851365641962e-05, + "loss": 1.0503, + "step": 3544 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829705198034322e-05, + "loss": 1.063, + "step": 3545 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829558968209558e-05, + "loss": 1.0731, + "step": 3546 + }, + { + "epoch": 0.09, + "learning_rate": 1.9829412676168587e-05, + "loss": 1.0905, + "step": 3547 + }, + { + "epoch": 0.09, + "learning_rate": 1.982926632191234e-05, + "loss": 1.0479, + "step": 3548 + }, + { + "epoch": 0.09, + "learning_rate": 1.982911990544174e-05, + "loss": 1.076, + "step": 3549 + }, + { + "epoch": 0.09, + "learning_rate": 1.982897342675772e-05, + "loss": 1.0851, + "step": 3550 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828826885861202e-05, + "loss": 1.0812, + "step": 3551 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828680282753114e-05, + "loss": 1.0782, + "step": 3552 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828533617434386e-05, + "loss": 1.0472, + "step": 3553 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828386889905947e-05, + "loss": 1.0788, + "step": 3554 + }, + { + "epoch": 0.09, + "learning_rate": 1.9828240100168717e-05, + "loss": 1.0584, + "step": 3555 + }, + { + "epoch": 0.09, + "learning_rate": 1.982809324822364e-05, + "loss": 1.0326, + "step": 3556 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827946334071634e-05, + "loss": 1.0541, + "step": 3557 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827799357713632e-05, + "loss": 1.0463, + "step": 3558 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827652319150568e-05, + "loss": 1.076, + "step": 3559 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827505218383366e-05, + "loss": 1.0452, + "step": 3560 + }, + { + "epoch": 0.09, + "learning_rate": 1.982735805541296e-05, + "loss": 1.0768, + "step": 3561 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827210830240285e-05, + "loss": 1.0712, + "step": 3562 + }, + { + "epoch": 0.09, + "learning_rate": 1.9827063542866266e-05, + "loss": 1.0768, + "step": 3563 + }, + { + "epoch": 0.09, + "learning_rate": 1.982691619329184e-05, + "loss": 1.0829, + "step": 3564 + }, + { + "epoch": 0.09, + "learning_rate": 1.982676878151794e-05, + "loss": 1.0844, + "step": 3565 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826621307545498e-05, + "loss": 1.0782, + "step": 3566 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826473771375446e-05, + "loss": 1.0803, + "step": 3567 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826326173008717e-05, + "loss": 1.0813, + "step": 3568 + }, + { + "epoch": 0.09, + "learning_rate": 1.982617851244625e-05, + "loss": 1.0489, + "step": 3569 + }, + { + "epoch": 0.09, + "learning_rate": 1.9826030789688978e-05, + "loss": 1.0707, + "step": 3570 + }, + { + "epoch": 0.09, + "learning_rate": 1.982588300473783e-05, + "loss": 1.0955, + "step": 3571 + }, + { + "epoch": 0.09, + "learning_rate": 1.982573515759375e-05, + "loss": 1.0592, + "step": 3572 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825587248257672e-05, + "loss": 1.0707, + "step": 3573 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825439276730524e-05, + "loss": 1.0863, + "step": 3574 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825291243013255e-05, + "loss": 1.0924, + "step": 3575 + }, + { + "epoch": 0.09, + "learning_rate": 1.9825143147106792e-05, + "loss": 1.0966, + "step": 3576 + }, + { + "epoch": 0.09, + "learning_rate": 1.982499498901208e-05, + "loss": 1.0895, + "step": 3577 + }, + { + "epoch": 0.09, + "learning_rate": 1.982484676873005e-05, + "loss": 1.0646, + "step": 3578 + }, + { + "epoch": 0.09, + "learning_rate": 1.982469848626164e-05, + "loss": 1.0653, + "step": 3579 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824550141607798e-05, + "loss": 1.0569, + "step": 3580 + }, + { + "epoch": 0.09, + "learning_rate": 1.982440173476945e-05, + "loss": 1.1098, + "step": 3581 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824253265747545e-05, + "loss": 1.0648, + "step": 3582 + }, + { + "epoch": 0.09, + "learning_rate": 1.9824104734543018e-05, + "loss": 1.0925, + "step": 3583 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823956141156816e-05, + "loss": 1.0149, + "step": 3584 + }, + { + "epoch": 0.09, + "learning_rate": 1.982380748558987e-05, + "loss": 1.058, + "step": 3585 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823658767843125e-05, + "loss": 1.0988, + "step": 3586 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823509987917527e-05, + "loss": 1.0197, + "step": 3587 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823361145814012e-05, + "loss": 1.0636, + "step": 3588 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823212241533518e-05, + "loss": 1.0469, + "step": 3589 + }, + { + "epoch": 0.09, + "learning_rate": 1.9823063275076998e-05, + "loss": 1.0953, + "step": 3590 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822914246445386e-05, + "loss": 1.0652, + "step": 3591 + }, + { + "epoch": 0.09, + "learning_rate": 1.982276515563963e-05, + "loss": 1.0653, + "step": 3592 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822616002660676e-05, + "loss": 1.0662, + "step": 3593 + }, + { + "epoch": 0.09, + "learning_rate": 1.982246678750946e-05, + "loss": 1.0479, + "step": 3594 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822317510186936e-05, + "loss": 1.0523, + "step": 3595 + }, + { + "epoch": 0.09, + "learning_rate": 1.982216817069404e-05, + "loss": 1.064, + "step": 3596 + }, + { + "epoch": 0.09, + "learning_rate": 1.9822018769031723e-05, + "loss": 1.0387, + "step": 3597 + }, + { + "epoch": 0.09, + "learning_rate": 1.982186930520093e-05, + "loss": 1.0451, + "step": 3598 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821719779202605e-05, + "loss": 1.0437, + "step": 3599 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821570191037695e-05, + "loss": 1.0905, + "step": 3600 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821420540707146e-05, + "loss": 1.0613, + "step": 3601 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821270828211906e-05, + "loss": 1.0004, + "step": 3602 + }, + { + "epoch": 0.09, + "learning_rate": 1.9821121053552926e-05, + "loss": 1.088, + "step": 3603 + }, + { + "epoch": 0.09, + "learning_rate": 1.982097121673115e-05, + "loss": 1.024, + "step": 3604 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820821317747526e-05, + "loss": 1.0586, + "step": 3605 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820671356603004e-05, + "loss": 1.0675, + "step": 3606 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820521333298535e-05, + "loss": 1.0538, + "step": 3607 + }, + { + "epoch": 0.09, + "learning_rate": 1.982037124783507e-05, + "loss": 1.0705, + "step": 3608 + }, + { + "epoch": 0.09, + "learning_rate": 1.982022110021355e-05, + "loss": 1.0338, + "step": 3609 + }, + { + "epoch": 0.09, + "learning_rate": 1.9820070890434934e-05, + "loss": 1.0492, + "step": 3610 + }, + { + "epoch": 0.09, + "learning_rate": 1.981992061850017e-05, + "loss": 1.0471, + "step": 3611 + }, + { + "epoch": 0.09, + "learning_rate": 1.981977028441021e-05, + "loss": 1.057, + "step": 3612 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819619888166e-05, + "loss": 1.0986, + "step": 3613 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819469429768503e-05, + "loss": 1.0479, + "step": 3614 + }, + { + "epoch": 0.09, + "learning_rate": 1.981931890921866e-05, + "loss": 1.0536, + "step": 3615 + }, + { + "epoch": 0.09, + "learning_rate": 1.981916832651743e-05, + "loss": 1.0566, + "step": 3616 + }, + { + "epoch": 0.09, + "learning_rate": 1.9819017681665764e-05, + "loss": 1.1094, + "step": 3617 + }, + { + "epoch": 0.09, + "learning_rate": 1.981886697466462e-05, + "loss": 1.0994, + "step": 3618 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818716205514944e-05, + "loss": 1.0425, + "step": 3619 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818565374217697e-05, + "loss": 1.072, + "step": 3620 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818414480773832e-05, + "loss": 1.1166, + "step": 3621 + }, + { + "epoch": 0.09, + "learning_rate": 1.9818263525184302e-05, + "loss": 1.094, + "step": 3622 + }, + { + "epoch": 0.09, + "learning_rate": 1.981811250745006e-05, + "loss": 1.0685, + "step": 3623 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817961427572073e-05, + "loss": 1.046, + "step": 3624 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817810285551286e-05, + "loss": 1.0774, + "step": 3625 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817659081388658e-05, + "loss": 1.0403, + "step": 3626 + }, + { + "epoch": 0.09, + "learning_rate": 1.981750781508515e-05, + "loss": 1.0816, + "step": 3627 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817356486641713e-05, + "loss": 1.0455, + "step": 3628 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817205096059314e-05, + "loss": 1.0458, + "step": 3629 + }, + { + "epoch": 0.09, + "learning_rate": 1.9817053643338906e-05, + "loss": 1.0563, + "step": 3630 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816902128481446e-05, + "loss": 1.0626, + "step": 3631 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816750551487895e-05, + "loss": 1.059, + "step": 3632 + }, + { + "epoch": 0.09, + "learning_rate": 1.981659891235921e-05, + "loss": 1.0532, + "step": 3633 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816447211096356e-05, + "loss": 1.0566, + "step": 3634 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816295447700286e-05, + "loss": 1.1029, + "step": 3635 + }, + { + "epoch": 0.09, + "learning_rate": 1.9816143622171967e-05, + "loss": 1.0731, + "step": 3636 + }, + { + "epoch": 0.09, + "learning_rate": 1.981599173451236e-05, + "loss": 1.0839, + "step": 3637 + }, + { + "epoch": 0.09, + "learning_rate": 1.981583978472242e-05, + "loss": 1.0678, + "step": 3638 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815687772803115e-05, + "loss": 1.1024, + "step": 3639 + }, + { + "epoch": 0.09, + "learning_rate": 1.9815535698755404e-05, + "loss": 1.0125, + "step": 3640 + }, + { + "epoch": 0.09, + "learning_rate": 1.981538356258025e-05, + "loss": 1.0666, + "step": 3641 + }, + { + "epoch": 0.09, + "learning_rate": 1.981523136427862e-05, + "loss": 1.0877, + "step": 3642 + }, + { + "epoch": 0.09, + "learning_rate": 1.981507910385147e-05, + "loss": 1.073, + "step": 3643 + }, + { + "epoch": 0.09, + "learning_rate": 1.981492678129977e-05, + "loss": 1.0697, + "step": 3644 + }, + { + "epoch": 0.09, + "learning_rate": 1.981477439662448e-05, + "loss": 1.0601, + "step": 3645 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814621949826564e-05, + "loss": 1.0643, + "step": 3646 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814469440906992e-05, + "loss": 1.0939, + "step": 3647 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814316869866728e-05, + "loss": 1.0813, + "step": 3648 + }, + { + "epoch": 0.09, + "learning_rate": 1.981416423670673e-05, + "loss": 1.0859, + "step": 3649 + }, + { + "epoch": 0.09, + "learning_rate": 1.9814011541427977e-05, + "loss": 1.0626, + "step": 3650 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813858784031428e-05, + "loss": 1.0819, + "step": 3651 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813705964518047e-05, + "loss": 1.0657, + "step": 3652 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813553082888805e-05, + "loss": 1.0691, + "step": 3653 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813400139144673e-05, + "loss": 1.1093, + "step": 3654 + }, + { + "epoch": 0.09, + "learning_rate": 1.981324713328661e-05, + "loss": 1.0522, + "step": 3655 + }, + { + "epoch": 0.09, + "learning_rate": 1.9813094065315597e-05, + "loss": 1.068, + "step": 3656 + }, + { + "epoch": 0.09, + "learning_rate": 1.981294093523259e-05, + "loss": 1.0778, + "step": 3657 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812787743038567e-05, + "loss": 1.0881, + "step": 3658 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812634488734494e-05, + "loss": 1.089, + "step": 3659 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812481172321343e-05, + "loss": 1.0501, + "step": 3660 + }, + { + "epoch": 0.09, + "learning_rate": 1.981232779380008e-05, + "loss": 1.0455, + "step": 3661 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812174353171684e-05, + "loss": 1.0753, + "step": 3662 + }, + { + "epoch": 0.09, + "learning_rate": 1.9812020850437116e-05, + "loss": 1.0468, + "step": 3663 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811867285597354e-05, + "loss": 1.0624, + "step": 3664 + }, + { + "epoch": 0.09, + "learning_rate": 1.981171365865337e-05, + "loss": 1.0619, + "step": 3665 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811559969606133e-05, + "loss": 1.0187, + "step": 3666 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811406218456622e-05, + "loss": 1.0776, + "step": 3667 + }, + { + "epoch": 0.09, + "learning_rate": 1.98112524052058e-05, + "loss": 1.022, + "step": 3668 + }, + { + "epoch": 0.09, + "learning_rate": 1.9811098529854652e-05, + "loss": 1.0554, + "step": 3669 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810944592404144e-05, + "loss": 1.05, + "step": 3670 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810790592855252e-05, + "loss": 1.0629, + "step": 3671 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810636531208956e-05, + "loss": 1.0429, + "step": 3672 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810482407466223e-05, + "loss": 1.0699, + "step": 3673 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810328221628033e-05, + "loss": 1.0966, + "step": 3674 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810173973695362e-05, + "loss": 1.1009, + "step": 3675 + }, + { + "epoch": 0.09, + "learning_rate": 1.9810019663669183e-05, + "loss": 1.0158, + "step": 3676 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809865291550478e-05, + "loss": 1.072, + "step": 3677 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809710857340222e-05, + "loss": 1.0642, + "step": 3678 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809556361039392e-05, + "loss": 1.0518, + "step": 3679 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809401802648963e-05, + "loss": 1.0466, + "step": 3680 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809247182169918e-05, + "loss": 1.0647, + "step": 3681 + }, + { + "epoch": 0.09, + "learning_rate": 1.9809092499603232e-05, + "loss": 1.0609, + "step": 3682 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808937754949885e-05, + "loss": 1.0703, + "step": 3683 + }, + { + "epoch": 0.09, + "learning_rate": 1.980878294821086e-05, + "loss": 1.0673, + "step": 3684 + }, + { + "epoch": 0.09, + "learning_rate": 1.980862807938713e-05, + "loss": 1.0685, + "step": 3685 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808473148479686e-05, + "loss": 1.0797, + "step": 3686 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808318155489497e-05, + "loss": 1.0749, + "step": 3687 + }, + { + "epoch": 0.09, + "learning_rate": 1.980816310041755e-05, + "loss": 1.0921, + "step": 3688 + }, + { + "epoch": 0.09, + "learning_rate": 1.9808007983264827e-05, + "loss": 1.0668, + "step": 3689 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807852804032306e-05, + "loss": 1.0576, + "step": 3690 + }, + { + "epoch": 0.09, + "learning_rate": 1.980769756272097e-05, + "loss": 1.0874, + "step": 3691 + }, + { + "epoch": 0.09, + "learning_rate": 1.980754225933181e-05, + "loss": 1.0517, + "step": 3692 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807386893865793e-05, + "loss": 1.0721, + "step": 3693 + }, + { + "epoch": 0.09, + "learning_rate": 1.9807231466323917e-05, + "loss": 1.0323, + "step": 3694 + }, + { + "epoch": 0.09, + "learning_rate": 1.980707597670716e-05, + "loss": 1.0475, + "step": 3695 + }, + { + "epoch": 0.09, + "learning_rate": 1.98069204250165e-05, + "loss": 1.0706, + "step": 3696 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806764811252937e-05, + "loss": 1.0562, + "step": 3697 + }, + { + "epoch": 0.09, + "learning_rate": 1.980660913541744e-05, + "loss": 1.0512, + "step": 3698 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806453397511006e-05, + "loss": 1.0341, + "step": 3699 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806297597534614e-05, + "loss": 1.1105, + "step": 3700 + }, + { + "epoch": 0.09, + "learning_rate": 1.9806141735489254e-05, + "loss": 1.0374, + "step": 3701 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805985811375908e-05, + "loss": 1.0688, + "step": 3702 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805829825195562e-05, + "loss": 1.064, + "step": 3703 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805673776949213e-05, + "loss": 1.0107, + "step": 3704 + }, + { + "epoch": 0.09, + "learning_rate": 1.980551766663784e-05, + "loss": 1.0457, + "step": 3705 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805361494262432e-05, + "loss": 1.0687, + "step": 3706 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805205259823984e-05, + "loss": 1.0448, + "step": 3707 + }, + { + "epoch": 0.09, + "learning_rate": 1.9805048963323475e-05, + "loss": 1.0657, + "step": 3708 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804892604761902e-05, + "loss": 1.0483, + "step": 3709 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804736184140255e-05, + "loss": 1.0328, + "step": 3710 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804579701459516e-05, + "loss": 1.0243, + "step": 3711 + }, + { + "epoch": 0.09, + "learning_rate": 1.980442315672068e-05, + "loss": 1.0685, + "step": 3712 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804266549924743e-05, + "loss": 1.0495, + "step": 3713 + }, + { + "epoch": 0.09, + "learning_rate": 1.9804109881072688e-05, + "loss": 1.0377, + "step": 3714 + }, + { + "epoch": 0.09, + "learning_rate": 1.980395315016551e-05, + "loss": 1.0758, + "step": 3715 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803796357204203e-05, + "loss": 1.0559, + "step": 3716 + }, + { + "epoch": 0.09, + "learning_rate": 1.980363950218976e-05, + "loss": 1.0644, + "step": 3717 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803482585123165e-05, + "loss": 1.0434, + "step": 3718 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803325606005425e-05, + "loss": 1.0785, + "step": 3719 + }, + { + "epoch": 0.09, + "learning_rate": 1.980316856483752e-05, + "loss": 1.0575, + "step": 3720 + }, + { + "epoch": 0.09, + "learning_rate": 1.9803011461620454e-05, + "loss": 1.0629, + "step": 3721 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802854296355216e-05, + "loss": 1.0384, + "step": 3722 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802697069042805e-05, + "loss": 1.0775, + "step": 3723 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802539779684208e-05, + "loss": 1.0842, + "step": 3724 + }, + { + "epoch": 0.09, + "learning_rate": 1.980238242828043e-05, + "loss": 1.0548, + "step": 3725 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802225014832465e-05, + "loss": 1.0776, + "step": 3726 + }, + { + "epoch": 0.09, + "learning_rate": 1.9802067539341305e-05, + "loss": 1.0206, + "step": 3727 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801910001807953e-05, + "loss": 1.0551, + "step": 3728 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801752402233395e-05, + "loss": 1.0653, + "step": 3729 + }, + { + "epoch": 0.09, + "learning_rate": 1.980159474061864e-05, + "loss": 1.0662, + "step": 3730 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801437016964682e-05, + "loss": 1.05, + "step": 3731 + }, + { + "epoch": 0.09, + "learning_rate": 1.980127923127252e-05, + "loss": 1.0935, + "step": 3732 + }, + { + "epoch": 0.09, + "learning_rate": 1.9801121383543152e-05, + "loss": 1.09, + "step": 3733 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800963473777573e-05, + "loss": 1.0866, + "step": 3734 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800805501976792e-05, + "loss": 1.0549, + "step": 3735 + }, + { + "epoch": 0.09, + "learning_rate": 1.98006474681418e-05, + "loss": 1.0731, + "step": 3736 + }, + { + "epoch": 0.09, + "learning_rate": 1.98004893722736e-05, + "loss": 1.0997, + "step": 3737 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800331214373196e-05, + "loss": 1.0912, + "step": 3738 + }, + { + "epoch": 0.09, + "learning_rate": 1.9800172994441584e-05, + "loss": 1.0578, + "step": 3739 + }, + { + "epoch": 0.09, + "learning_rate": 1.980001471247977e-05, + "loss": 1.0468, + "step": 3740 + }, + { + "epoch": 0.09, + "learning_rate": 1.979985636848875e-05, + "loss": 1.0127, + "step": 3741 + }, + { + "epoch": 0.09, + "learning_rate": 1.979969796246953e-05, + "loss": 1.0304, + "step": 3742 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799539494423115e-05, + "loss": 1.0122, + "step": 3743 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799380964350504e-05, + "loss": 1.0871, + "step": 3744 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799222372252703e-05, + "loss": 1.0338, + "step": 3745 + }, + { + "epoch": 0.09, + "learning_rate": 1.9799063718130718e-05, + "loss": 1.0644, + "step": 3746 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798905001985545e-05, + "loss": 1.0184, + "step": 3747 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798746223818194e-05, + "loss": 1.0357, + "step": 3748 + }, + { + "epoch": 0.09, + "learning_rate": 1.979858738362967e-05, + "loss": 1.0421, + "step": 3749 + }, + { + "epoch": 0.09, + "learning_rate": 1.979842848142098e-05, + "loss": 1.045, + "step": 3750 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798269517193125e-05, + "loss": 1.0615, + "step": 3751 + }, + { + "epoch": 0.09, + "learning_rate": 1.9798110490947113e-05, + "loss": 1.0676, + "step": 3752 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797951402683954e-05, + "loss": 1.0672, + "step": 3753 + }, + { + "epoch": 0.09, + "learning_rate": 1.979779225240465e-05, + "loss": 1.062, + "step": 3754 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797633040110217e-05, + "loss": 1.0523, + "step": 3755 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797473765801648e-05, + "loss": 1.0968, + "step": 3756 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797314429479964e-05, + "loss": 1.0511, + "step": 3757 + }, + { + "epoch": 0.09, + "learning_rate": 1.9797155031146168e-05, + "loss": 1.09, + "step": 3758 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796995570801273e-05, + "loss": 1.0807, + "step": 3759 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796836048446278e-05, + "loss": 1.0813, + "step": 3760 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796676464082204e-05, + "loss": 1.0408, + "step": 3761 + }, + { + "epoch": 0.09, + "learning_rate": 1.979651681771006e-05, + "loss": 1.0601, + "step": 3762 + }, + { + "epoch": 0.09, + "learning_rate": 1.979635710933085e-05, + "loss": 1.0439, + "step": 3763 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796197338945588e-05, + "loss": 1.011, + "step": 3764 + }, + { + "epoch": 0.09, + "learning_rate": 1.9796037506555287e-05, + "loss": 1.1005, + "step": 3765 + }, + { + "epoch": 0.09, + "learning_rate": 1.979587761216096e-05, + "loss": 1.0407, + "step": 3766 + }, + { + "epoch": 0.09, + "learning_rate": 1.979571765576361e-05, + "loss": 1.0064, + "step": 3767 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795557637364262e-05, + "loss": 1.0659, + "step": 3768 + }, + { + "epoch": 0.09, + "learning_rate": 1.979539755696392e-05, + "loss": 1.0879, + "step": 3769 + }, + { + "epoch": 0.09, + "learning_rate": 1.97952374145636e-05, + "loss": 1.1672, + "step": 3770 + }, + { + "epoch": 0.09, + "learning_rate": 1.9795077210164314e-05, + "loss": 1.0824, + "step": 3771 + }, + { + "epoch": 0.09, + "learning_rate": 1.979491694376708e-05, + "loss": 1.0847, + "step": 3772 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794756615372908e-05, + "loss": 1.1147, + "step": 3773 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794596224982817e-05, + "loss": 1.0754, + "step": 3774 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794435772597817e-05, + "loss": 1.0905, + "step": 3775 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794275258218932e-05, + "loss": 1.0983, + "step": 3776 + }, + { + "epoch": 0.09, + "learning_rate": 1.9794114681847166e-05, + "loss": 1.0501, + "step": 3777 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793954043483546e-05, + "loss": 1.0746, + "step": 3778 + }, + { + "epoch": 0.09, + "learning_rate": 1.979379334312908e-05, + "loss": 1.0845, + "step": 3779 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793632580784794e-05, + "loss": 1.0614, + "step": 3780 + }, + { + "epoch": 0.09, + "learning_rate": 1.97934717564517e-05, + "loss": 1.0701, + "step": 3781 + }, + { + "epoch": 0.09, + "learning_rate": 1.979331087013082e-05, + "loss": 1.0641, + "step": 3782 + }, + { + "epoch": 0.09, + "learning_rate": 1.9793149921823164e-05, + "loss": 1.0691, + "step": 3783 + }, + { + "epoch": 0.09, + "learning_rate": 1.979298891152976e-05, + "loss": 1.1288, + "step": 3784 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792827839251623e-05, + "loss": 1.0556, + "step": 3785 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792666704989772e-05, + "loss": 1.0774, + "step": 3786 + }, + { + "epoch": 0.09, + "learning_rate": 1.979250550874523e-05, + "loss": 1.0538, + "step": 3787 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792344250519014e-05, + "loss": 1.0591, + "step": 3788 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792182930312143e-05, + "loss": 1.0597, + "step": 3789 + }, + { + "epoch": 0.09, + "learning_rate": 1.9792021548125646e-05, + "loss": 1.0943, + "step": 3790 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791860103960538e-05, + "loss": 1.0762, + "step": 3791 + }, + { + "epoch": 0.09, + "learning_rate": 1.979169859781784e-05, + "loss": 1.0705, + "step": 3792 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791537029698577e-05, + "loss": 1.0098, + "step": 3793 + }, + { + "epoch": 0.09, + "learning_rate": 1.979137539960377e-05, + "loss": 1.1031, + "step": 3794 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791213707534447e-05, + "loss": 1.0881, + "step": 3795 + }, + { + "epoch": 0.09, + "learning_rate": 1.9791051953491626e-05, + "loss": 1.0042, + "step": 3796 + }, + { + "epoch": 0.09, + "learning_rate": 1.979089013747633e-05, + "loss": 1.0544, + "step": 3797 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790728259489587e-05, + "loss": 1.0866, + "step": 3798 + }, + { + "epoch": 0.09, + "learning_rate": 1.979056631953242e-05, + "loss": 1.0197, + "step": 3799 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790404317605856e-05, + "loss": 1.0739, + "step": 3800 + }, + { + "epoch": 0.09, + "learning_rate": 1.9790242253710915e-05, + "loss": 1.0583, + "step": 3801 + }, + { + "epoch": 0.09, + "learning_rate": 1.979008012784863e-05, + "loss": 1.0518, + "step": 3802 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789917940020022e-05, + "loss": 1.0208, + "step": 3803 + }, + { + "epoch": 0.09, + "learning_rate": 1.978975569022612e-05, + "loss": 1.0396, + "step": 3804 + }, + { + "epoch": 0.09, + "learning_rate": 1.978959337846795e-05, + "loss": 1.0369, + "step": 3805 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789431004746538e-05, + "loss": 1.0365, + "step": 3806 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789268569062914e-05, + "loss": 1.0497, + "step": 3807 + }, + { + "epoch": 0.09, + "learning_rate": 1.9789106071418104e-05, + "loss": 1.0921, + "step": 3808 + }, + { + "epoch": 0.09, + "learning_rate": 1.978894351181314e-05, + "loss": 1.0345, + "step": 3809 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788780890249047e-05, + "loss": 1.0728, + "step": 3810 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788618206726856e-05, + "loss": 1.0861, + "step": 3811 + }, + { + "epoch": 0.09, + "learning_rate": 1.97884554612476e-05, + "loss": 1.057, + "step": 3812 + }, + { + "epoch": 0.09, + "learning_rate": 1.97882926538123e-05, + "loss": 1.0433, + "step": 3813 + }, + { + "epoch": 0.09, + "learning_rate": 1.9788129784422e-05, + "loss": 1.0751, + "step": 3814 + }, + { + "epoch": 0.09, + "learning_rate": 1.978796685307772e-05, + "loss": 1.0487, + "step": 3815 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787803859780494e-05, + "loss": 1.0483, + "step": 3816 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787640804531354e-05, + "loss": 1.0255, + "step": 3817 + }, + { + "epoch": 0.09, + "learning_rate": 1.978747768733133e-05, + "loss": 1.0835, + "step": 3818 + }, + { + "epoch": 0.09, + "learning_rate": 1.978731450818146e-05, + "loss": 1.0344, + "step": 3819 + }, + { + "epoch": 0.09, + "learning_rate": 1.9787151267082775e-05, + "loss": 1.0842, + "step": 3820 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786987964036305e-05, + "loss": 1.0566, + "step": 3821 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786824599043083e-05, + "loss": 1.0697, + "step": 3822 + }, + { + "epoch": 0.09, + "learning_rate": 1.978666117210415e-05, + "loss": 1.0584, + "step": 3823 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786497683220532e-05, + "loss": 1.0882, + "step": 3824 + }, + { + "epoch": 0.09, + "learning_rate": 1.978633413239327e-05, + "loss": 1.0638, + "step": 3825 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786170519623395e-05, + "loss": 1.0734, + "step": 3826 + }, + { + "epoch": 0.09, + "learning_rate": 1.9786006844911943e-05, + "loss": 1.1158, + "step": 3827 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785843108259956e-05, + "loss": 1.0439, + "step": 3828 + }, + { + "epoch": 0.09, + "learning_rate": 1.978567930966846e-05, + "loss": 1.0819, + "step": 3829 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785515449138502e-05, + "loss": 1.0703, + "step": 3830 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785351526671115e-05, + "loss": 1.0254, + "step": 3831 + }, + { + "epoch": 0.09, + "learning_rate": 1.9785187542267334e-05, + "loss": 1.0575, + "step": 3832 + }, + { + "epoch": 0.09, + "learning_rate": 1.97850234959282e-05, + "loss": 1.0799, + "step": 3833 + }, + { + "epoch": 0.09, + "learning_rate": 1.978485938765475e-05, + "loss": 1.0701, + "step": 3834 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784695217448025e-05, + "loss": 1.0498, + "step": 3835 + }, + { + "epoch": 0.09, + "learning_rate": 1.9784530985309062e-05, + "loss": 1.0833, + "step": 3836 + }, + { + "epoch": 0.09, + "learning_rate": 1.97843666912389e-05, + "loss": 1.0622, + "step": 3837 + }, + { + "epoch": 0.09, + "learning_rate": 1.978420233523858e-05, + "loss": 1.0714, + "step": 3838 + }, + { + "epoch": 0.09, + "learning_rate": 1.978403791730914e-05, + "loss": 1.0495, + "step": 3839 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783873437451627e-05, + "loss": 1.0925, + "step": 3840 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783708895667075e-05, + "loss": 1.0815, + "step": 3841 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783544291956527e-05, + "loss": 1.0621, + "step": 3842 + }, + { + "epoch": 0.09, + "learning_rate": 1.978337962632103e-05, + "loss": 1.0315, + "step": 3843 + }, + { + "epoch": 0.09, + "learning_rate": 1.9783214898761618e-05, + "loss": 1.0338, + "step": 3844 + }, + { + "epoch": 0.09, + "learning_rate": 1.978305010927934e-05, + "loss": 1.0414, + "step": 3845 + }, + { + "epoch": 0.09, + "learning_rate": 1.978288525787524e-05, + "loss": 1.0263, + "step": 3846 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782720344550354e-05, + "loss": 1.0796, + "step": 3847 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782555369305734e-05, + "loss": 1.0478, + "step": 3848 + }, + { + "epoch": 0.09, + "learning_rate": 1.978239033214242e-05, + "loss": 1.0607, + "step": 3849 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782225233061454e-05, + "loss": 1.0426, + "step": 3850 + }, + { + "epoch": 0.09, + "learning_rate": 1.9782060072063888e-05, + "loss": 1.0642, + "step": 3851 + }, + { + "epoch": 0.09, + "learning_rate": 1.978189484915076e-05, + "loss": 1.0757, + "step": 3852 + }, + { + "epoch": 0.09, + "learning_rate": 1.978172956432312e-05, + "loss": 1.0689, + "step": 3853 + }, + { + "epoch": 0.09, + "learning_rate": 1.9781564217582014e-05, + "loss": 1.0453, + "step": 3854 + }, + { + "epoch": 0.09, + "learning_rate": 1.978139880892849e-05, + "loss": 1.0463, + "step": 3855 + }, + { + "epoch": 0.09, + "learning_rate": 1.978123333836359e-05, + "loss": 1.0478, + "step": 3856 + }, + { + "epoch": 0.09, + "learning_rate": 1.978106780588837e-05, + "loss": 1.0786, + "step": 3857 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780902211503864e-05, + "loss": 1.1013, + "step": 3858 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780736555211134e-05, + "loss": 1.1083, + "step": 3859 + }, + { + "epoch": 0.09, + "learning_rate": 1.978057083701122e-05, + "loss": 1.0612, + "step": 3860 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780405056905173e-05, + "loss": 1.0359, + "step": 3861 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780239214894045e-05, + "loss": 1.0978, + "step": 3862 + }, + { + "epoch": 0.09, + "learning_rate": 1.9780073310978883e-05, + "loss": 1.0527, + "step": 3863 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779907345160735e-05, + "loss": 1.0455, + "step": 3864 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779741317440656e-05, + "loss": 1.0934, + "step": 3865 + }, + { + "epoch": 0.09, + "learning_rate": 1.9779575227819697e-05, + "loss": 1.0509, + "step": 3866 + }, + { + "epoch": 0.09, + "learning_rate": 1.977940907629891e-05, + "loss": 1.0244, + "step": 3867 + }, + { + "epoch": 0.1, + "learning_rate": 1.9779242862879338e-05, + "loss": 1.0329, + "step": 3868 + }, + { + "epoch": 0.1, + "learning_rate": 1.9779076587562042e-05, + "loss": 1.0634, + "step": 3869 + }, + { + "epoch": 0.1, + "learning_rate": 1.977891025034807e-05, + "loss": 1.0584, + "step": 3870 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778743851238474e-05, + "loss": 1.0714, + "step": 3871 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778577390234317e-05, + "loss": 1.052, + "step": 3872 + }, + { + "epoch": 0.1, + "learning_rate": 1.977841086733664e-05, + "loss": 1.0773, + "step": 3873 + }, + { + "epoch": 0.1, + "learning_rate": 1.97782442825465e-05, + "loss": 1.051, + "step": 3874 + }, + { + "epoch": 0.1, + "learning_rate": 1.9778077635864957e-05, + "loss": 1.0573, + "step": 3875 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777910927293066e-05, + "loss": 1.0581, + "step": 3876 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777744156831873e-05, + "loss": 1.0382, + "step": 3877 + }, + { + "epoch": 0.1, + "learning_rate": 1.977757732448244e-05, + "loss": 1.0434, + "step": 3878 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777410430245824e-05, + "loss": 1.0683, + "step": 3879 + }, + { + "epoch": 0.1, + "learning_rate": 1.9777243474123076e-05, + "loss": 1.0643, + "step": 3880 + }, + { + "epoch": 0.1, + "learning_rate": 1.977707645611526e-05, + "loss": 1.0795, + "step": 3881 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776909376223424e-05, + "loss": 1.0393, + "step": 3882 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776742234448635e-05, + "loss": 1.0484, + "step": 3883 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776575030791946e-05, + "loss": 1.0353, + "step": 3884 + }, + { + "epoch": 0.1, + "learning_rate": 1.977640776525442e-05, + "loss": 1.0532, + "step": 3885 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776240437837105e-05, + "loss": 1.0695, + "step": 3886 + }, + { + "epoch": 0.1, + "learning_rate": 1.9776073048541068e-05, + "loss": 1.0691, + "step": 3887 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775905597367372e-05, + "loss": 1.052, + "step": 3888 + }, + { + "epoch": 0.1, + "learning_rate": 1.977573808431707e-05, + "loss": 1.0502, + "step": 3889 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775570509391224e-05, + "loss": 1.0576, + "step": 3890 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775402872590895e-05, + "loss": 1.0398, + "step": 3891 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775235173917144e-05, + "loss": 1.0374, + "step": 3892 + }, + { + "epoch": 0.1, + "learning_rate": 1.9775067413371034e-05, + "loss": 1.0571, + "step": 3893 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774899590953623e-05, + "loss": 1.0564, + "step": 3894 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774731706665978e-05, + "loss": 1.0719, + "step": 3895 + }, + { + "epoch": 0.1, + "learning_rate": 1.977456376050916e-05, + "loss": 1.0645, + "step": 3896 + }, + { + "epoch": 0.1, + "learning_rate": 1.977439575248423e-05, + "loss": 1.0381, + "step": 3897 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774227682592256e-05, + "loss": 1.0669, + "step": 3898 + }, + { + "epoch": 0.1, + "learning_rate": 1.9774059550834296e-05, + "loss": 1.0927, + "step": 3899 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773891357211415e-05, + "loss": 1.0285, + "step": 3900 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773723101724682e-05, + "loss": 1.0771, + "step": 3901 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773554784375156e-05, + "loss": 1.0235, + "step": 3902 + }, + { + "epoch": 0.1, + "learning_rate": 1.977338640516391e-05, + "loss": 1.0648, + "step": 3903 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773217964092005e-05, + "loss": 1.0273, + "step": 3904 + }, + { + "epoch": 0.1, + "learning_rate": 1.9773049461160504e-05, + "loss": 1.0916, + "step": 3905 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772880896370482e-05, + "loss": 1.0602, + "step": 3906 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772712269723e-05, + "loss": 1.0731, + "step": 3907 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772543581219123e-05, + "loss": 1.0868, + "step": 3908 + }, + { + "epoch": 0.1, + "learning_rate": 1.9772374830859923e-05, + "loss": 1.0574, + "step": 3909 + }, + { + "epoch": 0.1, + "learning_rate": 1.977220601864647e-05, + "loss": 1.0951, + "step": 3910 + }, + { + "epoch": 0.1, + "learning_rate": 1.977203714457983e-05, + "loss": 1.0671, + "step": 3911 + }, + { + "epoch": 0.1, + "learning_rate": 1.977186820866107e-05, + "loss": 1.078, + "step": 3912 + }, + { + "epoch": 0.1, + "learning_rate": 1.977169921089126e-05, + "loss": 1.0739, + "step": 3913 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771530151271473e-05, + "loss": 1.0947, + "step": 3914 + }, + { + "epoch": 0.1, + "learning_rate": 1.9771361029802775e-05, + "loss": 1.0602, + "step": 3915 + }, + { + "epoch": 0.1, + "learning_rate": 1.977119184648624e-05, + "loss": 1.0578, + "step": 3916 + }, + { + "epoch": 0.1, + "learning_rate": 1.977102260132293e-05, + "loss": 1.0362, + "step": 3917 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770853294313933e-05, + "loss": 1.0484, + "step": 3918 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770683925460304e-05, + "loss": 1.0738, + "step": 3919 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770514494763123e-05, + "loss": 1.0473, + "step": 3920 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770345002223464e-05, + "loss": 1.1025, + "step": 3921 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770175447842393e-05, + "loss": 1.0616, + "step": 3922 + }, + { + "epoch": 0.1, + "learning_rate": 1.9770005831620992e-05, + "loss": 1.0402, + "step": 3923 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769836153560328e-05, + "loss": 1.0382, + "step": 3924 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769666413661474e-05, + "loss": 1.0747, + "step": 3925 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769496611925512e-05, + "loss": 1.0596, + "step": 3926 + }, + { + "epoch": 0.1, + "learning_rate": 1.976932674835351e-05, + "loss": 1.0392, + "step": 3927 + }, + { + "epoch": 0.1, + "learning_rate": 1.9769156822946542e-05, + "loss": 1.0535, + "step": 3928 + }, + { + "epoch": 0.1, + "learning_rate": 1.976898683570569e-05, + "loss": 1.0629, + "step": 3929 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768816786632027e-05, + "loss": 1.0211, + "step": 3930 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768646675726627e-05, + "loss": 1.0502, + "step": 3931 + }, + { + "epoch": 0.1, + "learning_rate": 1.976847650299057e-05, + "loss": 1.0435, + "step": 3932 + }, + { + "epoch": 0.1, + "learning_rate": 1.9768306268424932e-05, + "loss": 1.0685, + "step": 3933 + }, + { + "epoch": 0.1, + "learning_rate": 1.976813597203079e-05, + "loss": 1.0388, + "step": 3934 + }, + { + "epoch": 0.1, + "learning_rate": 1.976796561380922e-05, + "loss": 1.029, + "step": 3935 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767795193761304e-05, + "loss": 1.0345, + "step": 3936 + }, + { + "epoch": 0.1, + "learning_rate": 1.976762471188812e-05, + "loss": 1.0396, + "step": 3937 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767454168190742e-05, + "loss": 1.0285, + "step": 3938 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767283562670258e-05, + "loss": 1.1077, + "step": 3939 + }, + { + "epoch": 0.1, + "learning_rate": 1.9767112895327742e-05, + "loss": 1.0442, + "step": 3940 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766942166164276e-05, + "loss": 1.0664, + "step": 3941 + }, + { + "epoch": 0.1, + "learning_rate": 1.976677137518094e-05, + "loss": 1.048, + "step": 3942 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766600522378817e-05, + "loss": 1.0524, + "step": 3943 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766429607758984e-05, + "loss": 1.0842, + "step": 3944 + }, + { + "epoch": 0.1, + "learning_rate": 1.976625863132253e-05, + "loss": 1.0665, + "step": 3945 + }, + { + "epoch": 0.1, + "learning_rate": 1.9766087593070526e-05, + "loss": 1.0482, + "step": 3946 + }, + { + "epoch": 0.1, + "learning_rate": 1.976591649300407e-05, + "loss": 1.0838, + "step": 3947 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765745331124226e-05, + "loss": 1.0583, + "step": 3948 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765574107432094e-05, + "loss": 1.0461, + "step": 3949 + }, + { + "epoch": 0.1, + "learning_rate": 1.976540282192875e-05, + "loss": 1.1001, + "step": 3950 + }, + { + "epoch": 0.1, + "learning_rate": 1.9765231474615284e-05, + "loss": 1.0445, + "step": 3951 + }, + { + "epoch": 0.1, + "learning_rate": 1.976506006549277e-05, + "loss": 1.0794, + "step": 3952 + }, + { + "epoch": 0.1, + "learning_rate": 1.97648885945623e-05, + "loss": 1.0745, + "step": 3953 + }, + { + "epoch": 0.1, + "learning_rate": 1.976471706182496e-05, + "loss": 1.0945, + "step": 3954 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764545467281833e-05, + "loss": 1.0966, + "step": 3955 + }, + { + "epoch": 0.1, + "learning_rate": 1.976437381093401e-05, + "loss": 1.0752, + "step": 3956 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764202092782568e-05, + "loss": 1.0593, + "step": 3957 + }, + { + "epoch": 0.1, + "learning_rate": 1.9764030312828605e-05, + "loss": 1.0681, + "step": 3958 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763858471073197e-05, + "loss": 1.122, + "step": 3959 + }, + { + "epoch": 0.1, + "learning_rate": 1.976368656751744e-05, + "loss": 1.0564, + "step": 3960 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763514602162424e-05, + "loss": 1.0353, + "step": 3961 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763342575009227e-05, + "loss": 1.0738, + "step": 3962 + }, + { + "epoch": 0.1, + "learning_rate": 1.9763170486058946e-05, + "loss": 1.0826, + "step": 3963 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762998335312668e-05, + "loss": 1.0569, + "step": 3964 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762826122771487e-05, + "loss": 1.0365, + "step": 3965 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762653848436485e-05, + "loss": 1.0439, + "step": 3966 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762481512308756e-05, + "loss": 1.0405, + "step": 3967 + }, + { + "epoch": 0.1, + "learning_rate": 1.9762309114389387e-05, + "loss": 1.0804, + "step": 3968 + }, + { + "epoch": 0.1, + "learning_rate": 1.976213665467948e-05, + "loss": 1.1188, + "step": 3969 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761964133180118e-05, + "loss": 1.1012, + "step": 3970 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761791549892397e-05, + "loss": 1.0626, + "step": 3971 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761618904817403e-05, + "loss": 1.0722, + "step": 3972 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761446197956234e-05, + "loss": 1.0725, + "step": 3973 + }, + { + "epoch": 0.1, + "learning_rate": 1.9761273429309982e-05, + "loss": 1.1004, + "step": 3974 + }, + { + "epoch": 0.1, + "learning_rate": 1.976110059887974e-05, + "loss": 1.0828, + "step": 3975 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760927706666604e-05, + "loss": 1.106, + "step": 3976 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760754752671667e-05, + "loss": 1.0256, + "step": 3977 + }, + { + "epoch": 0.1, + "learning_rate": 1.976058173689602e-05, + "loss": 1.0814, + "step": 3978 + }, + { + "epoch": 0.1, + "learning_rate": 1.976040865934077e-05, + "loss": 1.091, + "step": 3979 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760235520006998e-05, + "loss": 1.0333, + "step": 3980 + }, + { + "epoch": 0.1, + "learning_rate": 1.9760062318895806e-05, + "loss": 1.0546, + "step": 3981 + }, + { + "epoch": 0.1, + "learning_rate": 1.975988905600829e-05, + "loss": 1.0637, + "step": 3982 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759715731345547e-05, + "loss": 1.0372, + "step": 3983 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759542344908676e-05, + "loss": 1.0259, + "step": 3984 + }, + { + "epoch": 0.1, + "learning_rate": 1.975936889669877e-05, + "loss": 1.0923, + "step": 3985 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759195386716933e-05, + "loss": 1.028, + "step": 3986 + }, + { + "epoch": 0.1, + "learning_rate": 1.9759021814964255e-05, + "loss": 1.0646, + "step": 3987 + }, + { + "epoch": 0.1, + "learning_rate": 1.9758848181441843e-05, + "loss": 1.0289, + "step": 3988 + }, + { + "epoch": 0.1, + "learning_rate": 1.975867448615079e-05, + "loss": 1.0838, + "step": 3989 + }, + { + "epoch": 0.1, + "learning_rate": 1.97585007290922e-05, + "loss": 1.0475, + "step": 3990 + }, + { + "epoch": 0.1, + "learning_rate": 1.975832691026717e-05, + "loss": 1.0851, + "step": 3991 + }, + { + "epoch": 0.1, + "learning_rate": 1.97581530296768e-05, + "loss": 1.0999, + "step": 3992 + }, + { + "epoch": 0.1, + "learning_rate": 1.975797908732219e-05, + "loss": 1.0689, + "step": 3993 + }, + { + "epoch": 0.1, + "learning_rate": 1.975780508320444e-05, + "loss": 1.0759, + "step": 3994 + }, + { + "epoch": 0.1, + "learning_rate": 1.975763101732466e-05, + "loss": 1.0618, + "step": 3995 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757456889683938e-05, + "loss": 1.0334, + "step": 3996 + }, + { + "epoch": 0.1, + "learning_rate": 1.9757282700283387e-05, + "loss": 1.0721, + "step": 3997 + }, + { + "epoch": 0.1, + "learning_rate": 1.975710844912411e-05, + "loss": 1.0827, + "step": 3998 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756934136207198e-05, + "loss": 1.0781, + "step": 3999 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756759761533765e-05, + "loss": 1.0644, + "step": 4000 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756585325104912e-05, + "loss": 1.0465, + "step": 4001 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756410826921746e-05, + "loss": 1.0893, + "step": 4002 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756236266985365e-05, + "loss": 1.0574, + "step": 4003 + }, + { + "epoch": 0.1, + "learning_rate": 1.9756061645296878e-05, + "loss": 1.1124, + "step": 4004 + }, + { + "epoch": 0.1, + "learning_rate": 1.975588696185739e-05, + "loss": 1.0632, + "step": 4005 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755712216668006e-05, + "loss": 1.0467, + "step": 4006 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755537409729832e-05, + "loss": 1.0549, + "step": 4007 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755362541043976e-05, + "loss": 1.062, + "step": 4008 + }, + { + "epoch": 0.1, + "learning_rate": 1.9755187610611545e-05, + "loss": 1.0435, + "step": 4009 + }, + { + "epoch": 0.1, + "learning_rate": 1.975501261843364e-05, + "loss": 1.0637, + "step": 4010 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754837564511376e-05, + "loss": 1.0377, + "step": 4011 + }, + { + "epoch": 0.1, + "learning_rate": 1.975466244884586e-05, + "loss": 1.0601, + "step": 4012 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754487271438196e-05, + "loss": 1.0628, + "step": 4013 + }, + { + "epoch": 0.1, + "learning_rate": 1.9754312032289497e-05, + "loss": 1.0779, + "step": 4014 + }, + { + "epoch": 0.1, + "learning_rate": 1.975413673140087e-05, + "loss": 1.0794, + "step": 4015 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753961368773426e-05, + "loss": 1.0508, + "step": 4016 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753785944408274e-05, + "loss": 1.0729, + "step": 4017 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753610458306526e-05, + "loss": 1.0522, + "step": 4018 + }, + { + "epoch": 0.1, + "learning_rate": 1.975343491046929e-05, + "loss": 1.0268, + "step": 4019 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753259300897677e-05, + "loss": 1.0574, + "step": 4020 + }, + { + "epoch": 0.1, + "learning_rate": 1.9753083629592802e-05, + "loss": 1.0475, + "step": 4021 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752907896555777e-05, + "loss": 1.0559, + "step": 4022 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752732101787708e-05, + "loss": 1.0836, + "step": 4023 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752556245289714e-05, + "loss": 1.0481, + "step": 4024 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752380327062907e-05, + "loss": 1.0789, + "step": 4025 + }, + { + "epoch": 0.1, + "learning_rate": 1.9752204347108398e-05, + "loss": 1.1227, + "step": 4026 + }, + { + "epoch": 0.1, + "learning_rate": 1.97520283054273e-05, + "loss": 1.0574, + "step": 4027 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751852202020734e-05, + "loss": 1.0462, + "step": 4028 + }, + { + "epoch": 0.1, + "learning_rate": 1.975167603688981e-05, + "loss": 1.0974, + "step": 4029 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751499810035642e-05, + "loss": 1.0735, + "step": 4030 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751323521459345e-05, + "loss": 1.0959, + "step": 4031 + }, + { + "epoch": 0.1, + "learning_rate": 1.9751147171162038e-05, + "loss": 1.0591, + "step": 4032 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750970759144835e-05, + "loss": 1.0799, + "step": 4033 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750794285408857e-05, + "loss": 1.0713, + "step": 4034 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750617749955213e-05, + "loss": 1.0266, + "step": 4035 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750441152785027e-05, + "loss": 1.0868, + "step": 4036 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750264493899414e-05, + "loss": 1.0317, + "step": 4037 + }, + { + "epoch": 0.1, + "learning_rate": 1.9750087773299492e-05, + "loss": 1.01, + "step": 4038 + }, + { + "epoch": 0.1, + "learning_rate": 1.974991099098638e-05, + "loss": 1.0247, + "step": 4039 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749734146961197e-05, + "loss": 1.0616, + "step": 4040 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749557241225067e-05, + "loss": 1.0645, + "step": 4041 + }, + { + "epoch": 0.1, + "learning_rate": 1.97493802737791e-05, + "loss": 1.081, + "step": 4042 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749203244624422e-05, + "loss": 1.063, + "step": 4043 + }, + { + "epoch": 0.1, + "learning_rate": 1.9749026153762153e-05, + "loss": 1.0922, + "step": 4044 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748849001193416e-05, + "loss": 1.0948, + "step": 4045 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748671786919327e-05, + "loss": 1.1153, + "step": 4046 + }, + { + "epoch": 0.1, + "learning_rate": 1.9748494510941013e-05, + "loss": 1.0551, + "step": 4047 + }, + { + "epoch": 0.1, + "learning_rate": 1.974831717325959e-05, + "loss": 1.0816, + "step": 4048 + }, + { + "epoch": 0.1, + "learning_rate": 1.974813977387619e-05, + "loss": 1.0469, + "step": 4049 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747962312791928e-05, + "loss": 1.0439, + "step": 4050 + }, + { + "epoch": 0.1, + "learning_rate": 1.974778479000793e-05, + "loss": 1.0546, + "step": 4051 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747607205525314e-05, + "loss": 1.0684, + "step": 4052 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747429559345216e-05, + "loss": 1.0375, + "step": 4053 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747251851468746e-05, + "loss": 1.0451, + "step": 4054 + }, + { + "epoch": 0.1, + "learning_rate": 1.9747074081897042e-05, + "loss": 1.063, + "step": 4055 + }, + { + "epoch": 0.1, + "learning_rate": 1.974689625063122e-05, + "loss": 1.0489, + "step": 4056 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746718357672413e-05, + "loss": 1.0529, + "step": 4057 + }, + { + "epoch": 0.1, + "learning_rate": 1.974654040302174e-05, + "loss": 1.0889, + "step": 4058 + }, + { + "epoch": 0.1, + "learning_rate": 1.974636238668033e-05, + "loss": 1.0688, + "step": 4059 + }, + { + "epoch": 0.1, + "learning_rate": 1.974618430864931e-05, + "loss": 1.0863, + "step": 4060 + }, + { + "epoch": 0.1, + "learning_rate": 1.9746006168929808e-05, + "loss": 1.0658, + "step": 4061 + }, + { + "epoch": 0.1, + "learning_rate": 1.974582796752295e-05, + "loss": 1.0574, + "step": 4062 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745649704429865e-05, + "loss": 1.0419, + "step": 4063 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745471379651683e-05, + "loss": 1.0388, + "step": 4064 + }, + { + "epoch": 0.1, + "learning_rate": 1.974529299318953e-05, + "loss": 1.0694, + "step": 4065 + }, + { + "epoch": 0.1, + "learning_rate": 1.9745114545044533e-05, + "loss": 1.0758, + "step": 4066 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744936035217824e-05, + "loss": 1.025, + "step": 4067 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744757463710535e-05, + "loss": 1.0669, + "step": 4068 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744578830523794e-05, + "loss": 1.0236, + "step": 4069 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744400135658734e-05, + "loss": 1.0302, + "step": 4070 + }, + { + "epoch": 0.1, + "learning_rate": 1.974422137911648e-05, + "loss": 1.0477, + "step": 4071 + }, + { + "epoch": 0.1, + "learning_rate": 1.9744042560898175e-05, + "loss": 1.0687, + "step": 4072 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743863681004936e-05, + "loss": 1.0496, + "step": 4073 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743684739437908e-05, + "loss": 1.0772, + "step": 4074 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743505736198214e-05, + "loss": 1.0686, + "step": 4075 + }, + { + "epoch": 0.1, + "learning_rate": 1.974332667128699e-05, + "loss": 1.0806, + "step": 4076 + }, + { + "epoch": 0.1, + "learning_rate": 1.9743147544705373e-05, + "loss": 1.05, + "step": 4077 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742968356454496e-05, + "loss": 1.0419, + "step": 4078 + }, + { + "epoch": 0.1, + "learning_rate": 1.974278910653549e-05, + "loss": 1.0599, + "step": 4079 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742609794949492e-05, + "loss": 1.0332, + "step": 4080 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742430421697634e-05, + "loss": 1.0967, + "step": 4081 + }, + { + "epoch": 0.1, + "learning_rate": 1.9742250986781056e-05, + "loss": 1.0773, + "step": 4082 + }, + { + "epoch": 0.1, + "learning_rate": 1.974207149020089e-05, + "loss": 1.0297, + "step": 4083 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741891931958273e-05, + "loss": 1.0698, + "step": 4084 + }, + { + "epoch": 0.1, + "learning_rate": 1.974171231205434e-05, + "loss": 1.0607, + "step": 4085 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741532630490233e-05, + "loss": 1.0784, + "step": 4086 + }, + { + "epoch": 0.1, + "learning_rate": 1.9741352887267086e-05, + "loss": 1.0781, + "step": 4087 + }, + { + "epoch": 0.1, + "learning_rate": 1.974117308238604e-05, + "loss": 1.0668, + "step": 4088 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740993215848224e-05, + "loss": 1.0509, + "step": 4089 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740813287654782e-05, + "loss": 1.0612, + "step": 4090 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740633297806853e-05, + "loss": 1.0384, + "step": 4091 + }, + { + "epoch": 0.1, + "learning_rate": 1.974045324630558e-05, + "loss": 1.0323, + "step": 4092 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740273133152096e-05, + "loss": 1.0381, + "step": 4093 + }, + { + "epoch": 0.1, + "learning_rate": 1.9740092958347543e-05, + "loss": 1.0681, + "step": 4094 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739912721893065e-05, + "loss": 1.06, + "step": 4095 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739732423789802e-05, + "loss": 1.0602, + "step": 4096 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739552064038888e-05, + "loss": 1.0739, + "step": 4097 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739371642641475e-05, + "loss": 1.087, + "step": 4098 + }, + { + "epoch": 0.1, + "learning_rate": 1.9739191159598698e-05, + "loss": 1.0498, + "step": 4099 + }, + { + "epoch": 0.1, + "learning_rate": 1.97390106149117e-05, + "loss": 1.0506, + "step": 4100 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738830008581626e-05, + "loss": 1.0601, + "step": 4101 + }, + { + "epoch": 0.1, + "learning_rate": 1.973864934060962e-05, + "loss": 1.0554, + "step": 4102 + }, + { + "epoch": 0.1, + "learning_rate": 1.973846861099682e-05, + "loss": 1.0609, + "step": 4103 + }, + { + "epoch": 0.1, + "learning_rate": 1.9738287819744376e-05, + "loss": 1.0289, + "step": 4104 + }, + { + "epoch": 0.1, + "learning_rate": 1.973810696685343e-05, + "loss": 1.0526, + "step": 4105 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737926052325126e-05, + "loss": 1.0566, + "step": 4106 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737745076160614e-05, + "loss": 1.0606, + "step": 4107 + }, + { + "epoch": 0.1, + "learning_rate": 1.973756403836103e-05, + "loss": 1.0148, + "step": 4108 + }, + { + "epoch": 0.1, + "learning_rate": 1.973738293892753e-05, + "loss": 1.0306, + "step": 4109 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737201777861253e-05, + "loss": 1.1178, + "step": 4110 + }, + { + "epoch": 0.1, + "learning_rate": 1.9737020555163346e-05, + "loss": 1.0558, + "step": 4111 + }, + { + "epoch": 0.1, + "learning_rate": 1.973683927083496e-05, + "loss": 1.0346, + "step": 4112 + }, + { + "epoch": 0.1, + "learning_rate": 1.973665792487724e-05, + "loss": 1.0148, + "step": 4113 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736476517291336e-05, + "loss": 1.0709, + "step": 4114 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736295048078392e-05, + "loss": 1.0408, + "step": 4115 + }, + { + "epoch": 0.1, + "learning_rate": 1.9736113517239563e-05, + "loss": 1.0474, + "step": 4116 + }, + { + "epoch": 0.1, + "learning_rate": 1.973593192477599e-05, + "loss": 1.0627, + "step": 4117 + }, + { + "epoch": 0.1, + "learning_rate": 1.973575027068883e-05, + "loss": 1.0854, + "step": 4118 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735568554979226e-05, + "loss": 1.0546, + "step": 4119 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735386777648337e-05, + "loss": 1.0751, + "step": 4120 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735204938697302e-05, + "loss": 1.0647, + "step": 4121 + }, + { + "epoch": 0.1, + "learning_rate": 1.9735023038127284e-05, + "loss": 1.0458, + "step": 4122 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734841075939426e-05, + "loss": 1.0348, + "step": 4123 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734659052134882e-05, + "loss": 1.0529, + "step": 4124 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734476966714803e-05, + "loss": 1.0764, + "step": 4125 + }, + { + "epoch": 0.1, + "learning_rate": 1.9734294819680345e-05, + "loss": 1.072, + "step": 4126 + }, + { + "epoch": 0.1, + "learning_rate": 1.973411261103266e-05, + "loss": 1.0566, + "step": 4127 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733930340772894e-05, + "loss": 1.0785, + "step": 4128 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733748008902213e-05, + "loss": 1.0581, + "step": 4129 + }, + { + "epoch": 0.1, + "learning_rate": 1.973356561542176e-05, + "loss": 1.0615, + "step": 4130 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733383160332698e-05, + "loss": 1.1006, + "step": 4131 + }, + { + "epoch": 0.1, + "learning_rate": 1.9733200643636175e-05, + "loss": 1.084, + "step": 4132 + }, + { + "epoch": 0.1, + "learning_rate": 1.973301806533335e-05, + "loss": 1.0337, + "step": 4133 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732835425425382e-05, + "loss": 1.0498, + "step": 4134 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732652723913422e-05, + "loss": 1.0647, + "step": 4135 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732469960798623e-05, + "loss": 1.0404, + "step": 4136 + }, + { + "epoch": 0.1, + "learning_rate": 1.973228713608215e-05, + "loss": 1.0566, + "step": 4137 + }, + { + "epoch": 0.1, + "learning_rate": 1.9732104249765155e-05, + "loss": 1.065, + "step": 4138 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731921301848794e-05, + "loss": 1.0549, + "step": 4139 + }, + { + "epoch": 0.1, + "learning_rate": 1.973173829233423e-05, + "loss": 1.017, + "step": 4140 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731555221222623e-05, + "loss": 1.0627, + "step": 4141 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731372088515125e-05, + "loss": 1.0758, + "step": 4142 + }, + { + "epoch": 0.1, + "learning_rate": 1.9731188894212895e-05, + "loss": 1.0776, + "step": 4143 + }, + { + "epoch": 0.1, + "learning_rate": 1.97310056383171e-05, + "loss": 1.0537, + "step": 4144 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730822320828895e-05, + "loss": 1.0659, + "step": 4145 + }, + { + "epoch": 0.1, + "learning_rate": 1.973063894174944e-05, + "loss": 1.0436, + "step": 4146 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730455501079898e-05, + "loss": 1.0427, + "step": 4147 + }, + { + "epoch": 0.1, + "learning_rate": 1.9730271998821428e-05, + "loss": 1.0822, + "step": 4148 + }, + { + "epoch": 0.1, + "learning_rate": 1.973008843497519e-05, + "loss": 1.0304, + "step": 4149 + }, + { + "epoch": 0.1, + "learning_rate": 1.972990480954235e-05, + "loss": 1.0642, + "step": 4150 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729721122524067e-05, + "loss": 1.1082, + "step": 4151 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729537373921505e-05, + "loss": 1.0831, + "step": 4152 + }, + { + "epoch": 0.1, + "learning_rate": 1.9729353563735828e-05, + "loss": 1.0469, + "step": 4153 + }, + { + "epoch": 0.1, + "learning_rate": 1.97291696919682e-05, + "loss": 1.0488, + "step": 4154 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728985758619783e-05, + "loss": 1.055, + "step": 4155 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728801763691742e-05, + "loss": 1.0896, + "step": 4156 + }, + { + "epoch": 0.1, + "learning_rate": 1.972861770718524e-05, + "loss": 1.1059, + "step": 4157 + }, + { + "epoch": 0.1, + "learning_rate": 1.972843358910144e-05, + "loss": 1.0446, + "step": 4158 + }, + { + "epoch": 0.1, + "learning_rate": 1.9728249409441513e-05, + "loss": 1.0699, + "step": 4159 + }, + { + "epoch": 0.1, + "learning_rate": 1.972806516820662e-05, + "loss": 1.0639, + "step": 4160 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727880865397933e-05, + "loss": 1.0287, + "step": 4161 + }, + { + "epoch": 0.1, + "learning_rate": 1.972769650101661e-05, + "loss": 1.0599, + "step": 4162 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727512075063826e-05, + "loss": 1.0348, + "step": 4163 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727327587540745e-05, + "loss": 1.0561, + "step": 4164 + }, + { + "epoch": 0.1, + "learning_rate": 1.9727143038448534e-05, + "loss": 1.0505, + "step": 4165 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726958427788367e-05, + "loss": 1.0179, + "step": 4166 + }, + { + "epoch": 0.1, + "learning_rate": 1.97267737555614e-05, + "loss": 1.0248, + "step": 4167 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726589021768813e-05, + "loss": 1.0216, + "step": 4168 + }, + { + "epoch": 0.1, + "learning_rate": 1.972640422641177e-05, + "loss": 1.0427, + "step": 4169 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726219369491445e-05, + "loss": 1.0862, + "step": 4170 + }, + { + "epoch": 0.1, + "learning_rate": 1.9726034451009e-05, + "loss": 1.085, + "step": 4171 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725849470965617e-05, + "loss": 1.1156, + "step": 4172 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725664429362456e-05, + "loss": 1.0574, + "step": 4173 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725479326200694e-05, + "loss": 1.0445, + "step": 4174 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725294161481505e-05, + "loss": 1.0357, + "step": 4175 + }, + { + "epoch": 0.1, + "learning_rate": 1.9725108935206052e-05, + "loss": 1.0491, + "step": 4176 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724923647375516e-05, + "loss": 1.0489, + "step": 4177 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724738297991065e-05, + "loss": 1.0547, + "step": 4178 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724552887053875e-05, + "loss": 1.059, + "step": 4179 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724367414565118e-05, + "loss": 1.0711, + "step": 4180 + }, + { + "epoch": 0.1, + "learning_rate": 1.9724181880525966e-05, + "loss": 1.0729, + "step": 4181 + }, + { + "epoch": 0.1, + "learning_rate": 1.97239962849376e-05, + "loss": 1.0617, + "step": 4182 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723810627801186e-05, + "loss": 1.0493, + "step": 4183 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723624909117905e-05, + "loss": 1.0525, + "step": 4184 + }, + { + "epoch": 0.1, + "learning_rate": 1.9723439128888933e-05, + "loss": 1.0468, + "step": 4185 + }, + { + "epoch": 0.1, + "learning_rate": 1.972325328711544e-05, + "loss": 1.0617, + "step": 4186 + }, + { + "epoch": 0.1, + "learning_rate": 1.972306738379861e-05, + "loss": 1.0646, + "step": 4187 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722881418939614e-05, + "loss": 1.0557, + "step": 4188 + }, + { + "epoch": 0.1, + "learning_rate": 1.972269539253963e-05, + "loss": 1.0631, + "step": 4189 + }, + { + "epoch": 0.1, + "learning_rate": 1.972250930459984e-05, + "loss": 1.0685, + "step": 4190 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722323155121415e-05, + "loss": 1.0857, + "step": 4191 + }, + { + "epoch": 0.1, + "learning_rate": 1.9722136944105536e-05, + "loss": 1.0523, + "step": 4192 + }, + { + "epoch": 0.1, + "learning_rate": 1.972195067155338e-05, + "loss": 1.0958, + "step": 4193 + }, + { + "epoch": 0.1, + "learning_rate": 1.9721764337466132e-05, + "loss": 1.0404, + "step": 4194 + }, + { + "epoch": 0.1, + "learning_rate": 1.972157794184497e-05, + "loss": 1.0494, + "step": 4195 + }, + { + "epoch": 0.1, + "learning_rate": 1.9721391484691063e-05, + "loss": 1.0555, + "step": 4196 + }, + { + "epoch": 0.1, + "learning_rate": 1.9721204966005604e-05, + "loss": 1.0562, + "step": 4197 + }, + { + "epoch": 0.1, + "learning_rate": 1.972101838578977e-05, + "loss": 1.0661, + "step": 4198 + }, + { + "epoch": 0.1, + "learning_rate": 1.972083174404474e-05, + "loss": 1.0748, + "step": 4199 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720645040771696e-05, + "loss": 1.0594, + "step": 4200 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720458275971822e-05, + "loss": 1.0274, + "step": 4201 + }, + { + "epoch": 0.1, + "learning_rate": 1.9720271449646294e-05, + "loss": 1.0953, + "step": 4202 + }, + { + "epoch": 0.1, + "learning_rate": 1.97200845617963e-05, + "loss": 1.0315, + "step": 4203 + }, + { + "epoch": 0.1, + "learning_rate": 1.971989761242302e-05, + "loss": 1.0698, + "step": 4204 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719710601527645e-05, + "loss": 1.0504, + "step": 4205 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719523529111348e-05, + "loss": 1.0271, + "step": 4206 + }, + { + "epoch": 0.1, + "learning_rate": 1.9719336395175318e-05, + "loss": 1.0343, + "step": 4207 + }, + { + "epoch": 0.1, + "learning_rate": 1.971914919972074e-05, + "loss": 1.0299, + "step": 4208 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718961942748797e-05, + "loss": 1.1032, + "step": 4209 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718774624260674e-05, + "loss": 1.062, + "step": 4210 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718587244257562e-05, + "loss": 1.0782, + "step": 4211 + }, + { + "epoch": 0.1, + "learning_rate": 1.971839980274064e-05, + "loss": 1.0364, + "step": 4212 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718212299711095e-05, + "loss": 1.0779, + "step": 4213 + }, + { + "epoch": 0.1, + "learning_rate": 1.9718024735170123e-05, + "loss": 1.0528, + "step": 4214 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717837109118898e-05, + "loss": 1.0488, + "step": 4215 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717649421558614e-05, + "loss": 1.044, + "step": 4216 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717461672490464e-05, + "loss": 1.0529, + "step": 4217 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717273861915626e-05, + "loss": 1.0516, + "step": 4218 + }, + { + "epoch": 0.1, + "learning_rate": 1.9717085989835296e-05, + "loss": 1.0551, + "step": 4219 + }, + { + "epoch": 0.1, + "learning_rate": 1.971689805625066e-05, + "loss": 1.0924, + "step": 4220 + }, + { + "epoch": 0.1, + "learning_rate": 1.971671006116291e-05, + "loss": 1.0641, + "step": 4221 + }, + { + "epoch": 0.1, + "learning_rate": 1.971652200457323e-05, + "loss": 1.0707, + "step": 4222 + }, + { + "epoch": 0.1, + "learning_rate": 1.971633388648282e-05, + "loss": 1.0578, + "step": 4223 + }, + { + "epoch": 0.1, + "learning_rate": 1.9716145706892863e-05, + "loss": 1.0072, + "step": 4224 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715957465804558e-05, + "loss": 1.0663, + "step": 4225 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715769163219082e-05, + "loss": 1.046, + "step": 4226 + }, + { + "epoch": 0.1, + "learning_rate": 1.9715580799137642e-05, + "loss": 1.0141, + "step": 4227 + }, + { + "epoch": 0.1, + "learning_rate": 1.971539237356142e-05, + "loss": 1.055, + "step": 4228 + }, + { + "epoch": 0.1, + "learning_rate": 1.971520388649162e-05, + "loss": 1.0297, + "step": 4229 + }, + { + "epoch": 0.1, + "learning_rate": 1.971501533792942e-05, + "loss": 1.0957, + "step": 4230 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714826727876026e-05, + "loss": 1.0879, + "step": 4231 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714638056332624e-05, + "loss": 1.0233, + "step": 4232 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714449323300414e-05, + "loss": 1.0139, + "step": 4233 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714260528780584e-05, + "loss": 1.0547, + "step": 4234 + }, + { + "epoch": 0.1, + "learning_rate": 1.9714071672774335e-05, + "loss": 1.0581, + "step": 4235 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713882755282862e-05, + "loss": 1.0835, + "step": 4236 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713693776307358e-05, + "loss": 1.028, + "step": 4237 + }, + { + "epoch": 0.1, + "learning_rate": 1.971350473584902e-05, + "loss": 1.0623, + "step": 4238 + }, + { + "epoch": 0.1, + "learning_rate": 1.971331563390904e-05, + "loss": 1.078, + "step": 4239 + }, + { + "epoch": 0.1, + "learning_rate": 1.9713126470488625e-05, + "loss": 1.0757, + "step": 4240 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712937245588963e-05, + "loss": 1.0282, + "step": 4241 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712747959211257e-05, + "loss": 1.0177, + "step": 4242 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712558611356706e-05, + "loss": 1.0586, + "step": 4243 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712369202026506e-05, + "loss": 1.0959, + "step": 4244 + }, + { + "epoch": 0.1, + "learning_rate": 1.9712179731221856e-05, + "loss": 1.019, + "step": 4245 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711990198943953e-05, + "loss": 1.0387, + "step": 4246 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711800605194e-05, + "loss": 1.058, + "step": 4247 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711610949973194e-05, + "loss": 1.0939, + "step": 4248 + }, + { + "epoch": 0.1, + "learning_rate": 1.9711421233282736e-05, + "loss": 1.0288, + "step": 4249 + }, + { + "epoch": 0.1, + "learning_rate": 1.971123145512383e-05, + "loss": 1.065, + "step": 4250 + }, + { + "epoch": 0.1, + "learning_rate": 1.971104161549768e-05, + "loss": 1.0616, + "step": 4251 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710851714405478e-05, + "loss": 1.0874, + "step": 4252 + }, + { + "epoch": 0.1, + "learning_rate": 1.971066175184843e-05, + "loss": 1.0636, + "step": 4253 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710471727827742e-05, + "loss": 1.0331, + "step": 4254 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710281642344615e-05, + "loss": 1.0151, + "step": 4255 + }, + { + "epoch": 0.1, + "learning_rate": 1.9710091495400245e-05, + "loss": 1.1142, + "step": 4256 + }, + { + "epoch": 0.1, + "learning_rate": 1.970990128699585e-05, + "loss": 1.0562, + "step": 4257 + }, + { + "epoch": 0.1, + "learning_rate": 1.970971101713262e-05, + "loss": 1.0638, + "step": 4258 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709520685811765e-05, + "loss": 1.057, + "step": 4259 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709330293034494e-05, + "loss": 1.0615, + "step": 4260 + }, + { + "epoch": 0.1, + "learning_rate": 1.9709139838802e-05, + "loss": 1.0786, + "step": 4261 + }, + { + "epoch": 0.1, + "learning_rate": 1.97089493231155e-05, + "loss": 1.0664, + "step": 4262 + }, + { + "epoch": 0.1, + "learning_rate": 1.97087587459762e-05, + "loss": 1.06, + "step": 4263 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708568107385298e-05, + "loss": 1.0658, + "step": 4264 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708377407344005e-05, + "loss": 1.0108, + "step": 4265 + }, + { + "epoch": 0.1, + "learning_rate": 1.9708186645853532e-05, + "loss": 1.0222, + "step": 4266 + }, + { + "epoch": 0.1, + "learning_rate": 1.970799582291508e-05, + "loss": 1.0907, + "step": 4267 + }, + { + "epoch": 0.1, + "learning_rate": 1.970780493852986e-05, + "loss": 1.0809, + "step": 4268 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707613992699077e-05, + "loss": 1.1085, + "step": 4269 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707422985423945e-05, + "loss": 1.0619, + "step": 4270 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707231916705668e-05, + "loss": 1.0285, + "step": 4271 + }, + { + "epoch": 0.1, + "learning_rate": 1.9707040786545456e-05, + "loss": 1.0521, + "step": 4272 + }, + { + "epoch": 0.1, + "learning_rate": 1.9706849594944524e-05, + "loss": 1.0615, + "step": 4273 + }, + { + "epoch": 0.1, + "learning_rate": 1.9706658341904074e-05, + "loss": 1.0713, + "step": 4274 + }, + { + "epoch": 0.11, + "learning_rate": 1.9706467027425326e-05, + "loss": 1.0606, + "step": 4275 + }, + { + "epoch": 0.11, + "learning_rate": 1.9706275651509482e-05, + "loss": 1.0714, + "step": 4276 + }, + { + "epoch": 0.11, + "learning_rate": 1.9706084214157756e-05, + "loss": 1.0612, + "step": 4277 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705892715371364e-05, + "loss": 1.0492, + "step": 4278 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705701155151512e-05, + "loss": 1.0617, + "step": 4279 + }, + { + "epoch": 0.11, + "learning_rate": 1.970550953349942e-05, + "loss": 1.0432, + "step": 4280 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705317850416294e-05, + "loss": 1.0801, + "step": 4281 + }, + { + "epoch": 0.11, + "learning_rate": 1.9705126105903347e-05, + "loss": 1.0594, + "step": 4282 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704934299961798e-05, + "loss": 1.0458, + "step": 4283 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704742432592855e-05, + "loss": 1.0605, + "step": 4284 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704550503797742e-05, + "loss": 1.0888, + "step": 4285 + }, + { + "epoch": 0.11, + "learning_rate": 1.970435851357766e-05, + "loss": 1.0205, + "step": 4286 + }, + { + "epoch": 0.11, + "learning_rate": 1.9704166461933836e-05, + "loss": 1.0803, + "step": 4287 + }, + { + "epoch": 0.11, + "learning_rate": 1.970397434886748e-05, + "loss": 1.0674, + "step": 4288 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703782174379808e-05, + "loss": 1.0482, + "step": 4289 + }, + { + "epoch": 0.11, + "learning_rate": 1.970358993847204e-05, + "loss": 1.0777, + "step": 4290 + }, + { + "epoch": 0.11, + "learning_rate": 1.970339764114539e-05, + "loss": 1.0431, + "step": 4291 + }, + { + "epoch": 0.11, + "learning_rate": 1.9703205282401074e-05, + "loss": 1.0557, + "step": 4292 + }, + { + "epoch": 0.11, + "learning_rate": 1.970301286224031e-05, + "loss": 1.0729, + "step": 4293 + }, + { + "epoch": 0.11, + "learning_rate": 1.970282038066432e-05, + "loss": 1.0355, + "step": 4294 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702627837674315e-05, + "loss": 1.0753, + "step": 4295 + }, + { + "epoch": 0.11, + "learning_rate": 1.970243523327152e-05, + "loss": 1.0368, + "step": 4296 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702242567457154e-05, + "loss": 1.0834, + "step": 4297 + }, + { + "epoch": 0.11, + "learning_rate": 1.9702049840232434e-05, + "loss": 1.0532, + "step": 4298 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701857051598578e-05, + "loss": 1.0132, + "step": 4299 + }, + { + "epoch": 0.11, + "learning_rate": 1.970166420155681e-05, + "loss": 1.0618, + "step": 4300 + }, + { + "epoch": 0.11, + "learning_rate": 1.970147129010835e-05, + "loss": 1.0204, + "step": 4301 + }, + { + "epoch": 0.11, + "learning_rate": 1.970127831725442e-05, + "loss": 1.0468, + "step": 4302 + }, + { + "epoch": 0.11, + "learning_rate": 1.9701085282996234e-05, + "loss": 1.0673, + "step": 4303 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700892187335022e-05, + "loss": 1.0461, + "step": 4304 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700699030272002e-05, + "loss": 1.0332, + "step": 4305 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700505811808404e-05, + "loss": 1.0552, + "step": 4306 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700312531945444e-05, + "loss": 1.0644, + "step": 4307 + }, + { + "epoch": 0.11, + "learning_rate": 1.9700119190684345e-05, + "loss": 1.0663, + "step": 4308 + }, + { + "epoch": 0.11, + "learning_rate": 1.969992578802633e-05, + "loss": 1.0463, + "step": 4309 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699732323972628e-05, + "loss": 1.0265, + "step": 4310 + }, + { + "epoch": 0.11, + "learning_rate": 1.969953879852446e-05, + "loss": 1.0585, + "step": 4311 + }, + { + "epoch": 0.11, + "learning_rate": 1.9699345211683054e-05, + "loss": 1.0285, + "step": 4312 + }, + { + "epoch": 0.11, + "learning_rate": 1.969915156344963e-05, + "loss": 1.0141, + "step": 4313 + }, + { + "epoch": 0.11, + "learning_rate": 1.969895785382542e-05, + "loss": 1.0518, + "step": 4314 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698764082811647e-05, + "loss": 1.0549, + "step": 4315 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698570250409535e-05, + "loss": 1.082, + "step": 4316 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698376356620314e-05, + "loss": 1.0406, + "step": 4317 + }, + { + "epoch": 0.11, + "learning_rate": 1.9698182401445215e-05, + "loss": 1.0509, + "step": 4318 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697988384885456e-05, + "loss": 1.048, + "step": 4319 + }, + { + "epoch": 0.11, + "learning_rate": 1.969779430694227e-05, + "loss": 1.0537, + "step": 4320 + }, + { + "epoch": 0.11, + "learning_rate": 1.969760016761689e-05, + "loss": 1.0306, + "step": 4321 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697405966910538e-05, + "loss": 1.0789, + "step": 4322 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697211704824445e-05, + "loss": 1.0775, + "step": 4323 + }, + { + "epoch": 0.11, + "learning_rate": 1.9697017381359845e-05, + "loss": 1.0262, + "step": 4324 + }, + { + "epoch": 0.11, + "learning_rate": 1.969682299651796e-05, + "loss": 1.0659, + "step": 4325 + }, + { + "epoch": 0.11, + "learning_rate": 1.969662855030003e-05, + "loss": 1.0735, + "step": 4326 + }, + { + "epoch": 0.11, + "learning_rate": 1.969643404270728e-05, + "loss": 1.0658, + "step": 4327 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696239473740936e-05, + "loss": 1.0667, + "step": 4328 + }, + { + "epoch": 0.11, + "learning_rate": 1.9696044843402235e-05, + "loss": 1.0564, + "step": 4329 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695850151692415e-05, + "loss": 1.0186, + "step": 4330 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695655398612697e-05, + "loss": 1.0499, + "step": 4331 + }, + { + "epoch": 0.11, + "learning_rate": 1.969546058416432e-05, + "loss": 1.0665, + "step": 4332 + }, + { + "epoch": 0.11, + "learning_rate": 1.969526570834852e-05, + "loss": 1.0468, + "step": 4333 + }, + { + "epoch": 0.11, + "learning_rate": 1.9695070771166522e-05, + "loss": 1.024, + "step": 4334 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694875772619566e-05, + "loss": 1.0285, + "step": 4335 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694680712708885e-05, + "loss": 1.0301, + "step": 4336 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694485591435712e-05, + "loss": 1.0808, + "step": 4337 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694290408801287e-05, + "loss": 1.0487, + "step": 4338 + }, + { + "epoch": 0.11, + "learning_rate": 1.9694095164806836e-05, + "loss": 1.0607, + "step": 4339 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693899859453603e-05, + "loss": 1.0535, + "step": 4340 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693704492742825e-05, + "loss": 1.0699, + "step": 4341 + }, + { + "epoch": 0.11, + "learning_rate": 1.969350906467573e-05, + "loss": 1.0404, + "step": 4342 + }, + { + "epoch": 0.11, + "learning_rate": 1.969331357525356e-05, + "loss": 1.0726, + "step": 4343 + }, + { + "epoch": 0.11, + "learning_rate": 1.9693118024477557e-05, + "loss": 1.0463, + "step": 4344 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692922412348953e-05, + "loss": 1.0597, + "step": 4345 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692726738868984e-05, + "loss": 1.0253, + "step": 4346 + }, + { + "epoch": 0.11, + "learning_rate": 1.96925310040389e-05, + "loss": 1.0637, + "step": 4347 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692335207859924e-05, + "loss": 1.0247, + "step": 4348 + }, + { + "epoch": 0.11, + "learning_rate": 1.9692139350333305e-05, + "loss": 1.0176, + "step": 4349 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691943431460282e-05, + "loss": 1.0533, + "step": 4350 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691747451242093e-05, + "loss": 1.0225, + "step": 4351 + }, + { + "epoch": 0.11, + "learning_rate": 1.969155140967998e-05, + "loss": 1.0352, + "step": 4352 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691355306775184e-05, + "loss": 1.0343, + "step": 4353 + }, + { + "epoch": 0.11, + "learning_rate": 1.9691159142528943e-05, + "loss": 1.0669, + "step": 4354 + }, + { + "epoch": 0.11, + "learning_rate": 1.96909629169425e-05, + "loss": 1.0286, + "step": 4355 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690766630017104e-05, + "loss": 1.0708, + "step": 4356 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690570281753986e-05, + "loss": 1.0493, + "step": 4357 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690373872154396e-05, + "loss": 1.0501, + "step": 4358 + }, + { + "epoch": 0.11, + "learning_rate": 1.9690177401219575e-05, + "loss": 1.0616, + "step": 4359 + }, + { + "epoch": 0.11, + "learning_rate": 1.968998086895077e-05, + "loss": 1.0226, + "step": 4360 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689784275349217e-05, + "loss": 1.0477, + "step": 4361 + }, + { + "epoch": 0.11, + "learning_rate": 1.968958762041617e-05, + "loss": 1.0403, + "step": 4362 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689390904152866e-05, + "loss": 1.0516, + "step": 4363 + }, + { + "epoch": 0.11, + "learning_rate": 1.9689194126560554e-05, + "loss": 1.0384, + "step": 4364 + }, + { + "epoch": 0.11, + "learning_rate": 1.968899728764048e-05, + "loss": 1.0409, + "step": 4365 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688800387393883e-05, + "loss": 1.054, + "step": 4366 + }, + { + "epoch": 0.11, + "learning_rate": 1.968860342582202e-05, + "loss": 1.0711, + "step": 4367 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688406402926132e-05, + "loss": 1.0504, + "step": 4368 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688209318707466e-05, + "loss": 1.0663, + "step": 4369 + }, + { + "epoch": 0.11, + "learning_rate": 1.9688012173167268e-05, + "loss": 1.0391, + "step": 4370 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687814966306787e-05, + "loss": 1.0496, + "step": 4371 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687617698127277e-05, + "loss": 1.0486, + "step": 4372 + }, + { + "epoch": 0.11, + "learning_rate": 1.968742036862998e-05, + "loss": 1.0667, + "step": 4373 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687222977816142e-05, + "loss": 1.0394, + "step": 4374 + }, + { + "epoch": 0.11, + "learning_rate": 1.9687025525687022e-05, + "loss": 1.0428, + "step": 4375 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686828012243864e-05, + "loss": 1.0887, + "step": 4376 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686630437487917e-05, + "loss": 1.0355, + "step": 4377 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686432801420436e-05, + "loss": 1.0847, + "step": 4378 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686235104042667e-05, + "loss": 1.0412, + "step": 4379 + }, + { + "epoch": 0.11, + "learning_rate": 1.9686037345355867e-05, + "loss": 1.0492, + "step": 4380 + }, + { + "epoch": 0.11, + "learning_rate": 1.968583952536128e-05, + "loss": 1.0517, + "step": 4381 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685641644060166e-05, + "loss": 1.0584, + "step": 4382 + }, + { + "epoch": 0.11, + "learning_rate": 1.968544370145377e-05, + "loss": 1.0308, + "step": 4383 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685245697543354e-05, + "loss": 1.0752, + "step": 4384 + }, + { + "epoch": 0.11, + "learning_rate": 1.9685047632330164e-05, + "loss": 1.0166, + "step": 4385 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684849505815454e-05, + "loss": 1.0581, + "step": 4386 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684651318000486e-05, + "loss": 1.0737, + "step": 4387 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684453068886502e-05, + "loss": 1.0447, + "step": 4388 + }, + { + "epoch": 0.11, + "learning_rate": 1.9684254758474766e-05, + "loss": 1.0663, + "step": 4389 + }, + { + "epoch": 0.11, + "learning_rate": 1.968405638676653e-05, + "loss": 1.0187, + "step": 4390 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683857953763048e-05, + "loss": 1.036, + "step": 4391 + }, + { + "epoch": 0.11, + "learning_rate": 1.968365945946558e-05, + "loss": 1.0565, + "step": 4392 + }, + { + "epoch": 0.11, + "learning_rate": 1.968346090387538e-05, + "loss": 1.085, + "step": 4393 + }, + { + "epoch": 0.11, + "learning_rate": 1.9683262286993707e-05, + "loss": 1.0665, + "step": 4394 + }, + { + "epoch": 0.11, + "learning_rate": 1.968306360882181e-05, + "loss": 1.0359, + "step": 4395 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682864869360955e-05, + "loss": 1.0769, + "step": 4396 + }, + { + "epoch": 0.11, + "learning_rate": 1.96826660686124e-05, + "loss": 1.0654, + "step": 4397 + }, + { + "epoch": 0.11, + "learning_rate": 1.96824672065774e-05, + "loss": 1.0674, + "step": 4398 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682268283257215e-05, + "loss": 1.0688, + "step": 4399 + }, + { + "epoch": 0.11, + "learning_rate": 1.9682069298653105e-05, + "loss": 1.0413, + "step": 4400 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681870252766327e-05, + "loss": 1.0463, + "step": 4401 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681671145598142e-05, + "loss": 1.0348, + "step": 4402 + }, + { + "epoch": 0.11, + "learning_rate": 1.968147197714981e-05, + "loss": 1.0793, + "step": 4403 + }, + { + "epoch": 0.11, + "learning_rate": 1.9681272747422592e-05, + "loss": 1.0611, + "step": 4404 + }, + { + "epoch": 0.11, + "learning_rate": 1.968107345641775e-05, + "loss": 1.0541, + "step": 4405 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680874104136543e-05, + "loss": 1.0698, + "step": 4406 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680674690580234e-05, + "loss": 1.0397, + "step": 4407 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680475215750087e-05, + "loss": 1.0713, + "step": 4408 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680275679647362e-05, + "loss": 1.0508, + "step": 4409 + }, + { + "epoch": 0.11, + "learning_rate": 1.9680076082273322e-05, + "loss": 1.0213, + "step": 4410 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679876423629233e-05, + "loss": 1.0291, + "step": 4411 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679676703716357e-05, + "loss": 1.0258, + "step": 4412 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679476922535958e-05, + "loss": 1.0154, + "step": 4413 + }, + { + "epoch": 0.11, + "learning_rate": 1.96792770800893e-05, + "loss": 1.0274, + "step": 4414 + }, + { + "epoch": 0.11, + "learning_rate": 1.9679077176377647e-05, + "loss": 1.0472, + "step": 4415 + }, + { + "epoch": 0.11, + "learning_rate": 1.967887721140227e-05, + "loss": 1.0745, + "step": 4416 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678677185164427e-05, + "loss": 1.0726, + "step": 4417 + }, + { + "epoch": 0.11, + "learning_rate": 1.967847709766539e-05, + "loss": 1.0427, + "step": 4418 + }, + { + "epoch": 0.11, + "learning_rate": 1.9678276948906422e-05, + "loss": 1.0912, + "step": 4419 + }, + { + "epoch": 0.11, + "learning_rate": 1.967807673888879e-05, + "loss": 1.0447, + "step": 4420 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677876467613762e-05, + "loss": 1.087, + "step": 4421 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677676135082606e-05, + "loss": 1.0745, + "step": 4422 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677475741296592e-05, + "loss": 1.0474, + "step": 4423 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677275286256984e-05, + "loss": 1.0919, + "step": 4424 + }, + { + "epoch": 0.11, + "learning_rate": 1.9677074769965055e-05, + "loss": 1.0706, + "step": 4425 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676874192422067e-05, + "loss": 1.0333, + "step": 4426 + }, + { + "epoch": 0.11, + "learning_rate": 1.96766735536293e-05, + "loss": 1.0186, + "step": 4427 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676472853588017e-05, + "loss": 1.025, + "step": 4428 + }, + { + "epoch": 0.11, + "learning_rate": 1.967627209229949e-05, + "loss": 1.0625, + "step": 4429 + }, + { + "epoch": 0.11, + "learning_rate": 1.9676071269764988e-05, + "loss": 1.066, + "step": 4430 + }, + { + "epoch": 0.11, + "learning_rate": 1.9675870385985786e-05, + "loss": 1.03, + "step": 4431 + }, + { + "epoch": 0.11, + "learning_rate": 1.967566944096315e-05, + "loss": 1.0009, + "step": 4432 + }, + { + "epoch": 0.11, + "learning_rate": 1.967546843469836e-05, + "loss": 1.0809, + "step": 4433 + }, + { + "epoch": 0.11, + "learning_rate": 1.967526736719268e-05, + "loss": 1.0413, + "step": 4434 + }, + { + "epoch": 0.11, + "learning_rate": 1.9675066238447385e-05, + "loss": 1.0823, + "step": 4435 + }, + { + "epoch": 0.11, + "learning_rate": 1.967486504846375e-05, + "loss": 1.0816, + "step": 4436 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674663797243048e-05, + "loss": 1.0628, + "step": 4437 + }, + { + "epoch": 0.11, + "learning_rate": 1.9674462484786553e-05, + "loss": 1.0459, + "step": 4438 + }, + { + "epoch": 0.11, + "learning_rate": 1.967426111109554e-05, + "loss": 1.0445, + "step": 4439 + }, + { + "epoch": 0.11, + "learning_rate": 1.967405967617128e-05, + "loss": 1.1243, + "step": 4440 + }, + { + "epoch": 0.11, + "learning_rate": 1.967385818001505e-05, + "loss": 1.0851, + "step": 4441 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673656622628126e-05, + "loss": 1.0627, + "step": 4442 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673455004011783e-05, + "loss": 1.0885, + "step": 4443 + }, + { + "epoch": 0.11, + "learning_rate": 1.96732533241673e-05, + "loss": 1.0607, + "step": 4444 + }, + { + "epoch": 0.11, + "learning_rate": 1.9673051583095945e-05, + "loss": 1.0647, + "step": 4445 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672849780799006e-05, + "loss": 1.0277, + "step": 4446 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672647917277753e-05, + "loss": 1.0406, + "step": 4447 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672445992533466e-05, + "loss": 1.0545, + "step": 4448 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672244006567424e-05, + "loss": 1.0689, + "step": 4449 + }, + { + "epoch": 0.11, + "learning_rate": 1.9672041959380907e-05, + "loss": 1.0713, + "step": 4450 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671839850975185e-05, + "loss": 1.0212, + "step": 4451 + }, + { + "epoch": 0.11, + "learning_rate": 1.967163768135155e-05, + "loss": 1.0746, + "step": 4452 + }, + { + "epoch": 0.11, + "learning_rate": 1.967143545051127e-05, + "loss": 1.0573, + "step": 4453 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671233158455632e-05, + "loss": 1.0532, + "step": 4454 + }, + { + "epoch": 0.11, + "learning_rate": 1.9671030805185912e-05, + "loss": 1.0338, + "step": 4455 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670828390703394e-05, + "loss": 1.0837, + "step": 4456 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670625915009363e-05, + "loss": 1.0598, + "step": 4457 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670423378105093e-05, + "loss": 1.0683, + "step": 4458 + }, + { + "epoch": 0.11, + "learning_rate": 1.967022077999187e-05, + "loss": 1.0382, + "step": 4459 + }, + { + "epoch": 0.11, + "learning_rate": 1.9670018120670976e-05, + "loss": 1.0434, + "step": 4460 + }, + { + "epoch": 0.11, + "learning_rate": 1.966981540014369e-05, + "loss": 1.0708, + "step": 4461 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669612618411303e-05, + "loss": 1.0867, + "step": 4462 + }, + { + "epoch": 0.11, + "learning_rate": 1.966940977547509e-05, + "loss": 1.0302, + "step": 4463 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669206871336343e-05, + "loss": 1.0266, + "step": 4464 + }, + { + "epoch": 0.11, + "learning_rate": 1.9669003905996336e-05, + "loss": 1.0461, + "step": 4465 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668800879456363e-05, + "loss": 1.0546, + "step": 4466 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668597791717703e-05, + "loss": 1.0404, + "step": 4467 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668394642781648e-05, + "loss": 1.0568, + "step": 4468 + }, + { + "epoch": 0.11, + "learning_rate": 1.9668191432649475e-05, + "loss": 1.0323, + "step": 4469 + }, + { + "epoch": 0.11, + "learning_rate": 1.966798816132248e-05, + "loss": 1.0345, + "step": 4470 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667784828801943e-05, + "loss": 1.0693, + "step": 4471 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667581435089155e-05, + "loss": 1.0673, + "step": 4472 + }, + { + "epoch": 0.11, + "learning_rate": 1.96673779801854e-05, + "loss": 1.0551, + "step": 4473 + }, + { + "epoch": 0.11, + "learning_rate": 1.9667174464091963e-05, + "loss": 1.0829, + "step": 4474 + }, + { + "epoch": 0.11, + "learning_rate": 1.966697088681014e-05, + "loss": 1.0347, + "step": 4475 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666767248341213e-05, + "loss": 1.0173, + "step": 4476 + }, + { + "epoch": 0.11, + "learning_rate": 1.966656354868648e-05, + "loss": 1.0445, + "step": 4477 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666359787847214e-05, + "loss": 1.0301, + "step": 4478 + }, + { + "epoch": 0.11, + "learning_rate": 1.9666155965824726e-05, + "loss": 1.0501, + "step": 4479 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665952082620285e-05, + "loss": 1.0797, + "step": 4480 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665748138235196e-05, + "loss": 1.0276, + "step": 4481 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665544132670744e-05, + "loss": 1.0388, + "step": 4482 + }, + { + "epoch": 0.11, + "learning_rate": 1.966534006592822e-05, + "loss": 1.0789, + "step": 4483 + }, + { + "epoch": 0.11, + "learning_rate": 1.9665135938008916e-05, + "loss": 1.0608, + "step": 4484 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664931748914128e-05, + "loss": 1.0561, + "step": 4485 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664727498645144e-05, + "loss": 1.0452, + "step": 4486 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664523187203257e-05, + "loss": 1.0548, + "step": 4487 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664318814589766e-05, + "loss": 1.0721, + "step": 4488 + }, + { + "epoch": 0.11, + "learning_rate": 1.9664114380805954e-05, + "loss": 1.027, + "step": 4489 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663909885853126e-05, + "loss": 1.0192, + "step": 4490 + }, + { + "epoch": 0.11, + "learning_rate": 1.966370532973257e-05, + "loss": 1.0547, + "step": 4491 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663500712445577e-05, + "loss": 1.0662, + "step": 4492 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663296033993453e-05, + "loss": 1.0881, + "step": 4493 + }, + { + "epoch": 0.11, + "learning_rate": 1.9663091294377482e-05, + "loss": 1.0849, + "step": 4494 + }, + { + "epoch": 0.11, + "learning_rate": 1.966288649359897e-05, + "loss": 1.045, + "step": 4495 + }, + { + "epoch": 0.11, + "learning_rate": 1.966268163165921e-05, + "loss": 1.0423, + "step": 4496 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662476708559494e-05, + "loss": 1.066, + "step": 4497 + }, + { + "epoch": 0.11, + "learning_rate": 1.966227172430112e-05, + "loss": 1.0533, + "step": 4498 + }, + { + "epoch": 0.11, + "learning_rate": 1.9662066678885394e-05, + "loss": 1.0655, + "step": 4499 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661861572313605e-05, + "loss": 1.0644, + "step": 4500 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661656404587054e-05, + "loss": 1.0509, + "step": 4501 + }, + { + "epoch": 0.11, + "learning_rate": 1.966145117570704e-05, + "loss": 1.0699, + "step": 4502 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661245885674862e-05, + "loss": 1.0708, + "step": 4503 + }, + { + "epoch": 0.11, + "learning_rate": 1.9661040534491816e-05, + "loss": 1.0521, + "step": 4504 + }, + { + "epoch": 0.11, + "learning_rate": 1.966083512215921e-05, + "loss": 1.0366, + "step": 4505 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660629648678338e-05, + "loss": 1.0109, + "step": 4506 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660424114050497e-05, + "loss": 1.0748, + "step": 4507 + }, + { + "epoch": 0.11, + "learning_rate": 1.9660218518276998e-05, + "loss": 1.0519, + "step": 4508 + }, + { + "epoch": 0.11, + "learning_rate": 1.966001286135913e-05, + "loss": 1.0571, + "step": 4509 + }, + { + "epoch": 0.11, + "learning_rate": 1.965980714329821e-05, + "loss": 1.1031, + "step": 4510 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659601364095527e-05, + "loss": 1.0662, + "step": 4511 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659395523752387e-05, + "loss": 1.0299, + "step": 4512 + }, + { + "epoch": 0.11, + "learning_rate": 1.9659189622270095e-05, + "loss": 1.0169, + "step": 4513 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658983659649953e-05, + "loss": 1.0415, + "step": 4514 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658777635893264e-05, + "loss": 1.0283, + "step": 4515 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658571551001335e-05, + "loss": 1.0782, + "step": 4516 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658365404975467e-05, + "loss": 1.0867, + "step": 4517 + }, + { + "epoch": 0.11, + "learning_rate": 1.9658159197816968e-05, + "loss": 1.028, + "step": 4518 + }, + { + "epoch": 0.11, + "learning_rate": 1.965795292952714e-05, + "loss": 1.0257, + "step": 4519 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657746600107287e-05, + "loss": 1.0599, + "step": 4520 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657540209558718e-05, + "loss": 1.0554, + "step": 4521 + }, + { + "epoch": 0.11, + "learning_rate": 1.9657333757882742e-05, + "loss": 1.0121, + "step": 4522 + }, + { + "epoch": 0.11, + "learning_rate": 1.965712724508066e-05, + "loss": 1.0477, + "step": 4523 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656920671153786e-05, + "loss": 1.0701, + "step": 4524 + }, + { + "epoch": 0.11, + "learning_rate": 1.965671403610342e-05, + "loss": 1.0864, + "step": 4525 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656507339930876e-05, + "loss": 1.0538, + "step": 4526 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656300582637457e-05, + "loss": 1.0267, + "step": 4527 + }, + { + "epoch": 0.11, + "learning_rate": 1.9656093764224476e-05, + "loss": 1.0791, + "step": 4528 + }, + { + "epoch": 0.11, + "learning_rate": 1.965588688469324e-05, + "loss": 1.0416, + "step": 4529 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655679944045062e-05, + "loss": 1.0523, + "step": 4530 + }, + { + "epoch": 0.11, + "learning_rate": 1.965547294228124e-05, + "loss": 1.0553, + "step": 4531 + }, + { + "epoch": 0.11, + "learning_rate": 1.96552658794031e-05, + "loss": 1.0633, + "step": 4532 + }, + { + "epoch": 0.11, + "learning_rate": 1.9655058755411944e-05, + "loss": 1.1002, + "step": 4533 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654851570309086e-05, + "loss": 1.0542, + "step": 4534 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654644324095833e-05, + "loss": 0.9987, + "step": 4535 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654437016773503e-05, + "loss": 1.0565, + "step": 4536 + }, + { + "epoch": 0.11, + "learning_rate": 1.96542296483434e-05, + "loss": 1.0515, + "step": 4537 + }, + { + "epoch": 0.11, + "learning_rate": 1.9654022218806844e-05, + "loss": 1.0576, + "step": 4538 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653814728165144e-05, + "loss": 1.0474, + "step": 4539 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653607176419616e-05, + "loss": 1.0426, + "step": 4540 + }, + { + "epoch": 0.11, + "learning_rate": 1.9653399563571574e-05, + "loss": 1.0241, + "step": 4541 + }, + { + "epoch": 0.11, + "learning_rate": 1.965319188962233e-05, + "loss": 1.0426, + "step": 4542 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652984154573195e-05, + "loss": 1.0731, + "step": 4543 + }, + { + "epoch": 0.11, + "learning_rate": 1.965277635842549e-05, + "loss": 1.0524, + "step": 4544 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652568501180526e-05, + "loss": 1.0182, + "step": 4545 + }, + { + "epoch": 0.11, + "learning_rate": 1.9652360582839623e-05, + "loss": 1.0453, + "step": 4546 + }, + { + "epoch": 0.11, + "learning_rate": 1.965215260340409e-05, + "loss": 1.0381, + "step": 4547 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651944562875255e-05, + "loss": 1.0526, + "step": 4548 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651736461254423e-05, + "loss": 1.0921, + "step": 4549 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651528298542918e-05, + "loss": 1.0666, + "step": 4550 + }, + { + "epoch": 0.11, + "learning_rate": 1.9651320074742056e-05, + "loss": 1.0337, + "step": 4551 + }, + { + "epoch": 0.11, + "learning_rate": 1.965111178985315e-05, + "loss": 0.9982, + "step": 4552 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650903443877526e-05, + "loss": 1.0168, + "step": 4553 + }, + { + "epoch": 0.11, + "learning_rate": 1.96506950368165e-05, + "loss": 1.081, + "step": 4554 + }, + { + "epoch": 0.11, + "learning_rate": 1.965048656867139e-05, + "loss": 1.0377, + "step": 4555 + }, + { + "epoch": 0.11, + "learning_rate": 1.965027803944352e-05, + "loss": 1.0541, + "step": 4556 + }, + { + "epoch": 0.11, + "learning_rate": 1.9650069449134198e-05, + "loss": 1.01, + "step": 4557 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649860797744758e-05, + "loss": 1.0722, + "step": 4558 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649652085276514e-05, + "loss": 1.0409, + "step": 4559 + }, + { + "epoch": 0.11, + "learning_rate": 1.964944331173079e-05, + "loss": 1.0479, + "step": 4560 + }, + { + "epoch": 0.11, + "learning_rate": 1.96492344771089e-05, + "loss": 1.049, + "step": 4561 + }, + { + "epoch": 0.11, + "learning_rate": 1.9649025581412173e-05, + "loss": 1.0808, + "step": 4562 + }, + { + "epoch": 0.11, + "learning_rate": 1.964881662464193e-05, + "loss": 1.0624, + "step": 4563 + }, + { + "epoch": 0.11, + "learning_rate": 1.96486076067995e-05, + "loss": 1.0431, + "step": 4564 + }, + { + "epoch": 0.11, + "learning_rate": 1.964839852788619e-05, + "loss": 1.0608, + "step": 4565 + }, + { + "epoch": 0.11, + "learning_rate": 1.9648189387903335e-05, + "loss": 1.0422, + "step": 4566 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647980186852257e-05, + "loss": 1.0471, + "step": 4567 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647770924734278e-05, + "loss": 1.0212, + "step": 4568 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647561601550725e-05, + "loss": 1.031, + "step": 4569 + }, + { + "epoch": 0.11, + "learning_rate": 1.964735221730292e-05, + "loss": 1.0877, + "step": 4570 + }, + { + "epoch": 0.11, + "learning_rate": 1.9647142771992193e-05, + "loss": 1.0249, + "step": 4571 + }, + { + "epoch": 0.11, + "learning_rate": 1.964693326561986e-05, + "loss": 1.0533, + "step": 4572 + }, + { + "epoch": 0.11, + "learning_rate": 1.964672369818726e-05, + "loss": 1.0225, + "step": 4573 + }, + { + "epoch": 0.11, + "learning_rate": 1.964651406969571e-05, + "loss": 1.0309, + "step": 4574 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646304380146546e-05, + "loss": 1.084, + "step": 4575 + }, + { + "epoch": 0.11, + "learning_rate": 1.9646094629541085e-05, + "loss": 1.0396, + "step": 4576 + }, + { + "epoch": 0.11, + "learning_rate": 1.964588481788066e-05, + "loss": 1.0276, + "step": 4577 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645674945166596e-05, + "loss": 1.0194, + "step": 4578 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645465011400224e-05, + "loss": 1.0925, + "step": 4579 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645255016582872e-05, + "loss": 1.0302, + "step": 4580 + }, + { + "epoch": 0.11, + "learning_rate": 1.9645044960715868e-05, + "loss": 1.0537, + "step": 4581 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644834843800546e-05, + "loss": 1.0531, + "step": 4582 + }, + { + "epoch": 0.11, + "learning_rate": 1.964462466583823e-05, + "loss": 1.0594, + "step": 4583 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644414426830252e-05, + "loss": 1.0656, + "step": 4584 + }, + { + "epoch": 0.11, + "learning_rate": 1.9644204126777947e-05, + "loss": 1.0576, + "step": 4585 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643993765682644e-05, + "loss": 1.0496, + "step": 4586 + }, + { + "epoch": 0.11, + "learning_rate": 1.964378334354567e-05, + "loss": 1.0323, + "step": 4587 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643572860368362e-05, + "loss": 1.0608, + "step": 4588 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643362316152047e-05, + "loss": 1.0767, + "step": 4589 + }, + { + "epoch": 0.11, + "learning_rate": 1.9643151710898064e-05, + "loss": 1.093, + "step": 4590 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642941044607747e-05, + "loss": 1.0473, + "step": 4591 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642730317282417e-05, + "loss": 1.0801, + "step": 4592 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642519528923424e-05, + "loss": 1.0411, + "step": 4593 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642308679532092e-05, + "loss": 1.0804, + "step": 4594 + }, + { + "epoch": 0.11, + "learning_rate": 1.9642097769109756e-05, + "loss": 1.0729, + "step": 4595 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641886797657754e-05, + "loss": 1.0366, + "step": 4596 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641675765177425e-05, + "loss": 1.0487, + "step": 4597 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641464671670093e-05, + "loss": 1.0267, + "step": 4598 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641253517137103e-05, + "loss": 1.0608, + "step": 4599 + }, + { + "epoch": 0.11, + "learning_rate": 1.9641042301579795e-05, + "loss": 1.0592, + "step": 4600 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640831024999495e-05, + "loss": 1.0521, + "step": 4601 + }, + { + "epoch": 0.11, + "learning_rate": 1.964061968739755e-05, + "loss": 1.0608, + "step": 4602 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640408288775288e-05, + "loss": 1.0574, + "step": 4603 + }, + { + "epoch": 0.11, + "learning_rate": 1.9640196829134055e-05, + "loss": 1.0403, + "step": 4604 + }, + { + "epoch": 0.11, + "learning_rate": 1.963998530847519e-05, + "loss": 1.0443, + "step": 4605 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639773726800023e-05, + "loss": 1.0334, + "step": 4606 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639562084109904e-05, + "loss": 1.037, + "step": 4607 + }, + { + "epoch": 0.11, + "learning_rate": 1.963935038040617e-05, + "loss": 1.0766, + "step": 4608 + }, + { + "epoch": 0.11, + "learning_rate": 1.9639138615690154e-05, + "loss": 1.084, + "step": 4609 + }, + { + "epoch": 0.11, + "learning_rate": 1.96389267899632e-05, + "loss": 1.0877, + "step": 4610 + }, + { + "epoch": 0.11, + "learning_rate": 1.963871490322665e-05, + "loss": 1.0422, + "step": 4611 + }, + { + "epoch": 0.11, + "learning_rate": 1.963850295548185e-05, + "loss": 1.0156, + "step": 4612 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638290946730132e-05, + "loss": 1.0626, + "step": 4613 + }, + { + "epoch": 0.11, + "learning_rate": 1.9638078876972842e-05, + "loss": 1.0569, + "step": 4614 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637866746211328e-05, + "loss": 1.0464, + "step": 4615 + }, + { + "epoch": 0.11, + "learning_rate": 1.963765455444692e-05, + "loss": 1.0593, + "step": 4616 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637442301680973e-05, + "loss": 1.0732, + "step": 4617 + }, + { + "epoch": 0.11, + "learning_rate": 1.9637229987914823e-05, + "loss": 1.0491, + "step": 4618 + }, + { + "epoch": 0.11, + "learning_rate": 1.963701761314982e-05, + "loss": 1.0364, + "step": 4619 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636805177387305e-05, + "loss": 1.0371, + "step": 4620 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636592680628618e-05, + "loss": 1.0399, + "step": 4621 + }, + { + "epoch": 0.11, + "learning_rate": 1.963638012287511e-05, + "loss": 1.0136, + "step": 4622 + }, + { + "epoch": 0.11, + "learning_rate": 1.9636167504128128e-05, + "loss": 1.07, + "step": 4623 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635954824389013e-05, + "loss": 1.0759, + "step": 4624 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635742083659114e-05, + "loss": 1.0777, + "step": 4625 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635529281939773e-05, + "loss": 1.0447, + "step": 4626 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635316419232346e-05, + "loss": 1.0617, + "step": 4627 + }, + { + "epoch": 0.11, + "learning_rate": 1.9635103495538167e-05, + "loss": 1.0628, + "step": 4628 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634890510858594e-05, + "loss": 1.0856, + "step": 4629 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634677465194974e-05, + "loss": 1.0342, + "step": 4630 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634464358548655e-05, + "loss": 1.0364, + "step": 4631 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634251190920984e-05, + "loss": 1.09, + "step": 4632 + }, + { + "epoch": 0.11, + "learning_rate": 1.9634037962313308e-05, + "loss": 1.0342, + "step": 4633 + }, + { + "epoch": 0.11, + "learning_rate": 1.963382467272698e-05, + "loss": 1.0547, + "step": 4634 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633611322163352e-05, + "loss": 1.0204, + "step": 4635 + }, + { + "epoch": 0.11, + "learning_rate": 1.9633397910623776e-05, + "loss": 1.061, + "step": 4636 + }, + { + "epoch": 0.11, + "learning_rate": 1.963318443810959e-05, + "loss": 1.0667, + "step": 4637 + }, + { + "epoch": 0.11, + "learning_rate": 1.963297090462216e-05, + "loss": 1.0387, + "step": 4638 + }, + { + "epoch": 0.11, + "learning_rate": 1.963275731016283e-05, + "loss": 1.041, + "step": 4639 + }, + { + "epoch": 0.11, + "learning_rate": 1.9632543654732954e-05, + "loss": 1.0575, + "step": 4640 + }, + { + "epoch": 0.11, + "learning_rate": 1.9632329938333884e-05, + "loss": 1.0455, + "step": 4641 + }, + { + "epoch": 0.11, + "learning_rate": 1.963211616096697e-05, + "loss": 1.0311, + "step": 4642 + }, + { + "epoch": 0.11, + "learning_rate": 1.963190232263357e-05, + "loss": 1.0513, + "step": 4643 + }, + { + "epoch": 0.11, + "learning_rate": 1.963168842333503e-05, + "loss": 1.035, + "step": 4644 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631474463072715e-05, + "loss": 1.0762, + "step": 4645 + }, + { + "epoch": 0.11, + "learning_rate": 1.963126044184797e-05, + "loss": 1.046, + "step": 4646 + }, + { + "epoch": 0.11, + "learning_rate": 1.9631046359662158e-05, + "loss": 1.0718, + "step": 4647 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630832216516623e-05, + "loss": 1.0399, + "step": 4648 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630618012412733e-05, + "loss": 1.0248, + "step": 4649 + }, + { + "epoch": 0.11, + "learning_rate": 1.963040374735183e-05, + "loss": 1.0742, + "step": 4650 + }, + { + "epoch": 0.11, + "learning_rate": 1.9630189421335283e-05, + "loss": 1.0534, + "step": 4651 + }, + { + "epoch": 0.11, + "learning_rate": 1.962997503436444e-05, + "loss": 1.0566, + "step": 4652 + }, + { + "epoch": 0.11, + "learning_rate": 1.962976058644066e-05, + "loss": 1.069, + "step": 4653 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629546077565302e-05, + "loss": 1.0343, + "step": 4654 + }, + { + "epoch": 0.11, + "learning_rate": 1.9629331507739724e-05, + "loss": 1.0735, + "step": 4655 + }, + { + "epoch": 0.11, + "learning_rate": 1.962911687696528e-05, + "loss": 1.0594, + "step": 4656 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628902185243337e-05, + "loss": 1.0801, + "step": 4657 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628687432575245e-05, + "loss": 1.0454, + "step": 4658 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628472618962367e-05, + "loss": 1.0179, + "step": 4659 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628257744406065e-05, + "loss": 1.0543, + "step": 4660 + }, + { + "epoch": 0.11, + "learning_rate": 1.9628042808907693e-05, + "loss": 1.0328, + "step": 4661 + }, + { + "epoch": 0.11, + "learning_rate": 1.962782781246862e-05, + "loss": 1.0647, + "step": 4662 + }, + { + "epoch": 0.11, + "learning_rate": 1.9627612755090195e-05, + "loss": 1.0495, + "step": 4663 + }, + { + "epoch": 0.11, + "learning_rate": 1.962739763677379e-05, + "loss": 1.0763, + "step": 4664 + }, + { + "epoch": 0.11, + "learning_rate": 1.962718245752076e-05, + "loss": 1.0041, + "step": 4665 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626967217332473e-05, + "loss": 1.0483, + "step": 4666 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626751916210286e-05, + "loss": 1.1217, + "step": 4667 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626536554155562e-05, + "loss": 1.067, + "step": 4668 + }, + { + "epoch": 0.11, + "learning_rate": 1.962632113116967e-05, + "loss": 1.0492, + "step": 4669 + }, + { + "epoch": 0.11, + "learning_rate": 1.9626105647253964e-05, + "loss": 1.0397, + "step": 4670 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625890102409818e-05, + "loss": 1.0679, + "step": 4671 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625674496638588e-05, + "loss": 1.0103, + "step": 4672 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625458829941644e-05, + "loss": 1.0325, + "step": 4673 + }, + { + "epoch": 0.11, + "learning_rate": 1.962524310232035e-05, + "loss": 1.0254, + "step": 4674 + }, + { + "epoch": 0.11, + "learning_rate": 1.9625027313776067e-05, + "loss": 1.0535, + "step": 4675 + }, + { + "epoch": 0.11, + "learning_rate": 1.962481146431017e-05, + "loss": 1.0363, + "step": 4676 + }, + { + "epoch": 0.11, + "learning_rate": 1.9624595553924014e-05, + "loss": 1.0362, + "step": 4677 + }, + { + "epoch": 0.11, + "learning_rate": 1.9624379582618976e-05, + "loss": 1.0514, + "step": 4678 + }, + { + "epoch": 0.11, + "learning_rate": 1.9624163550396415e-05, + "loss": 1.0602, + "step": 4679 + }, + { + "epoch": 0.11, + "learning_rate": 1.9623947457257705e-05, + "loss": 1.0593, + "step": 4680 + }, + { + "epoch": 0.11, + "learning_rate": 1.9623731303204207e-05, + "loss": 1.0691, + "step": 4681 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623515088237296e-05, + "loss": 1.093, + "step": 4682 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623298812358332e-05, + "loss": 1.0282, + "step": 4683 + }, + { + "epoch": 0.12, + "learning_rate": 1.9623082475568694e-05, + "loss": 1.0555, + "step": 4684 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622866077869743e-05, + "loss": 1.0663, + "step": 4685 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622649619262855e-05, + "loss": 1.0398, + "step": 4686 + }, + { + "epoch": 0.12, + "learning_rate": 1.9622433099749396e-05, + "loss": 1.0454, + "step": 4687 + }, + { + "epoch": 0.12, + "learning_rate": 1.962221651933074e-05, + "loss": 1.0399, + "step": 4688 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621999878008254e-05, + "loss": 1.0413, + "step": 4689 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621783175783307e-05, + "loss": 1.0601, + "step": 4690 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621566412657276e-05, + "loss": 1.0681, + "step": 4691 + }, + { + "epoch": 0.12, + "learning_rate": 1.9621349588631532e-05, + "loss": 1.0734, + "step": 4692 + }, + { + "epoch": 0.12, + "learning_rate": 1.962113270370745e-05, + "loss": 1.0513, + "step": 4693 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620915757886397e-05, + "loss": 1.0176, + "step": 4694 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620698751169747e-05, + "loss": 1.0575, + "step": 4695 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620481683558873e-05, + "loss": 1.0553, + "step": 4696 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620264555055155e-05, + "loss": 1.0353, + "step": 4697 + }, + { + "epoch": 0.12, + "learning_rate": 1.9620047365659963e-05, + "loss": 1.0104, + "step": 4698 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619830115374673e-05, + "loss": 1.0666, + "step": 4699 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619612804200657e-05, + "loss": 1.0576, + "step": 4700 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619395432139293e-05, + "loss": 1.026, + "step": 4701 + }, + { + "epoch": 0.12, + "learning_rate": 1.9619177999191958e-05, + "loss": 1.0675, + "step": 4702 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618960505360026e-05, + "loss": 1.0359, + "step": 4703 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618742950644874e-05, + "loss": 1.0571, + "step": 4704 + }, + { + "epoch": 0.12, + "learning_rate": 1.961852533504788e-05, + "loss": 1.075, + "step": 4705 + }, + { + "epoch": 0.12, + "learning_rate": 1.9618307658570418e-05, + "loss": 1.085, + "step": 4706 + }, + { + "epoch": 0.12, + "learning_rate": 1.961808992121387e-05, + "loss": 1.0148, + "step": 4707 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617872122979617e-05, + "loss": 1.0467, + "step": 4708 + }, + { + "epoch": 0.12, + "learning_rate": 1.961765426386903e-05, + "loss": 1.0646, + "step": 4709 + }, + { + "epoch": 0.12, + "learning_rate": 1.961743634388349e-05, + "loss": 1.0595, + "step": 4710 + }, + { + "epoch": 0.12, + "learning_rate": 1.961721836302438e-05, + "loss": 1.0168, + "step": 4711 + }, + { + "epoch": 0.12, + "learning_rate": 1.9617000321293074e-05, + "loss": 1.022, + "step": 4712 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616782218690957e-05, + "loss": 1.0662, + "step": 4713 + }, + { + "epoch": 0.12, + "learning_rate": 1.9616564055219412e-05, + "loss": 1.0257, + "step": 4714 + }, + { + "epoch": 0.12, + "learning_rate": 1.961634583087981e-05, + "loss": 1.0401, + "step": 4715 + }, + { + "epoch": 0.12, + "learning_rate": 1.961612754567354e-05, + "loss": 1.0798, + "step": 4716 + }, + { + "epoch": 0.12, + "learning_rate": 1.961590919960198e-05, + "loss": 1.09, + "step": 4717 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615690792666516e-05, + "loss": 1.0254, + "step": 4718 + }, + { + "epoch": 0.12, + "learning_rate": 1.961547232486853e-05, + "loss": 1.0366, + "step": 4719 + }, + { + "epoch": 0.12, + "learning_rate": 1.9615253796209398e-05, + "loss": 1.029, + "step": 4720 + }, + { + "epoch": 0.12, + "learning_rate": 1.961503520669051e-05, + "loss": 1.0157, + "step": 4721 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614816556313246e-05, + "loss": 1.0465, + "step": 4722 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614597845078993e-05, + "loss": 1.0483, + "step": 4723 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614379072989136e-05, + "loss": 1.0494, + "step": 4724 + }, + { + "epoch": 0.12, + "learning_rate": 1.9614160240045054e-05, + "loss": 1.0778, + "step": 4725 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613941346248137e-05, + "loss": 1.0713, + "step": 4726 + }, + { + "epoch": 0.12, + "learning_rate": 1.961372239159977e-05, + "loss": 1.0343, + "step": 4727 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613503376101337e-05, + "loss": 1.0574, + "step": 4728 + }, + { + "epoch": 0.12, + "learning_rate": 1.9613284299754223e-05, + "loss": 1.0469, + "step": 4729 + }, + { + "epoch": 0.12, + "learning_rate": 1.961306516255982e-05, + "loss": 1.0609, + "step": 4730 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612845964519513e-05, + "loss": 1.0631, + "step": 4731 + }, + { + "epoch": 0.12, + "learning_rate": 1.961262670563469e-05, + "loss": 1.0711, + "step": 4732 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612407385906735e-05, + "loss": 1.0844, + "step": 4733 + }, + { + "epoch": 0.12, + "learning_rate": 1.9612188005337035e-05, + "loss": 1.017, + "step": 4734 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611968563926984e-05, + "loss": 1.0365, + "step": 4735 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611749061677966e-05, + "loss": 1.0217, + "step": 4736 + }, + { + "epoch": 0.12, + "learning_rate": 1.961152949859138e-05, + "loss": 1.0762, + "step": 4737 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611309874668604e-05, + "loss": 1.036, + "step": 4738 + }, + { + "epoch": 0.12, + "learning_rate": 1.9611090189911036e-05, + "loss": 1.0605, + "step": 4739 + }, + { + "epoch": 0.12, + "learning_rate": 1.961087044432006e-05, + "loss": 1.0618, + "step": 4740 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610650637897072e-05, + "loss": 1.0633, + "step": 4741 + }, + { + "epoch": 0.12, + "learning_rate": 1.9610430770643464e-05, + "loss": 1.0799, + "step": 4742 + }, + { + "epoch": 0.12, + "learning_rate": 1.961021084256062e-05, + "loss": 1.0927, + "step": 4743 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609990853649945e-05, + "loss": 1.0598, + "step": 4744 + }, + { + "epoch": 0.12, + "learning_rate": 1.960977080391282e-05, + "loss": 1.057, + "step": 4745 + }, + { + "epoch": 0.12, + "learning_rate": 1.960955069335064e-05, + "loss": 1.0759, + "step": 4746 + }, + { + "epoch": 0.12, + "learning_rate": 1.96093305219648e-05, + "loss": 1.0503, + "step": 4747 + }, + { + "epoch": 0.12, + "learning_rate": 1.9609110289756696e-05, + "loss": 1.0368, + "step": 4748 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608889996727716e-05, + "loss": 1.0348, + "step": 4749 + }, + { + "epoch": 0.12, + "learning_rate": 1.960866964287926e-05, + "loss": 1.0372, + "step": 4750 + }, + { + "epoch": 0.12, + "learning_rate": 1.9608449228212717e-05, + "loss": 1.0326, + "step": 4751 + }, + { + "epoch": 0.12, + "learning_rate": 1.960822875272949e-05, + "loss": 1.0243, + "step": 4752 + }, + { + "epoch": 0.12, + "learning_rate": 1.960800821643097e-05, + "loss": 1.0361, + "step": 4753 + }, + { + "epoch": 0.12, + "learning_rate": 1.960778761931855e-05, + "loss": 1.0835, + "step": 4754 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607566961393632e-05, + "loss": 1.0274, + "step": 4755 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607346242657606e-05, + "loss": 1.0644, + "step": 4756 + }, + { + "epoch": 0.12, + "learning_rate": 1.9607125463111875e-05, + "loss": 1.0439, + "step": 4757 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606904622757835e-05, + "loss": 1.052, + "step": 4758 + }, + { + "epoch": 0.12, + "learning_rate": 1.960668372159688e-05, + "loss": 1.029, + "step": 4759 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606462759630415e-05, + "loss": 1.0303, + "step": 4760 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606241736859834e-05, + "loss": 1.0622, + "step": 4761 + }, + { + "epoch": 0.12, + "learning_rate": 1.9606020653286537e-05, + "loss": 1.0321, + "step": 4762 + }, + { + "epoch": 0.12, + "learning_rate": 1.960579950891192e-05, + "loss": 1.0026, + "step": 4763 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605578303737393e-05, + "loss": 1.0215, + "step": 4764 + }, + { + "epoch": 0.12, + "learning_rate": 1.9605357037764343e-05, + "loss": 1.0836, + "step": 4765 + }, + { + "epoch": 0.12, + "learning_rate": 1.960513571099418e-05, + "loss": 1.0356, + "step": 4766 + }, + { + "epoch": 0.12, + "learning_rate": 1.96049143234283e-05, + "loss": 1.0685, + "step": 4767 + }, + { + "epoch": 0.12, + "learning_rate": 1.96046928750681e-05, + "loss": 1.0887, + "step": 4768 + }, + { + "epoch": 0.12, + "learning_rate": 1.9604471365914993e-05, + "loss": 1.0835, + "step": 4769 + }, + { + "epoch": 0.12, + "learning_rate": 1.9604249795970378e-05, + "loss": 1.0467, + "step": 4770 + }, + { + "epoch": 0.12, + "learning_rate": 1.960402816523565e-05, + "loss": 1.0512, + "step": 4771 + }, + { + "epoch": 0.12, + "learning_rate": 1.960380647371222e-05, + "loss": 1.0584, + "step": 4772 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603584721401487e-05, + "loss": 1.0224, + "step": 4773 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603362908304856e-05, + "loss": 1.086, + "step": 4774 + }, + { + "epoch": 0.12, + "learning_rate": 1.9603141034423728e-05, + "loss": 1.0742, + "step": 4775 + }, + { + "epoch": 0.12, + "learning_rate": 1.960291909975951e-05, + "loss": 1.0205, + "step": 4776 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602697104313608e-05, + "loss": 1.0521, + "step": 4777 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602475048087425e-05, + "loss": 1.0713, + "step": 4778 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602252931082365e-05, + "loss": 1.0661, + "step": 4779 + }, + { + "epoch": 0.12, + "learning_rate": 1.9602030753299843e-05, + "loss": 1.0544, + "step": 4780 + }, + { + "epoch": 0.12, + "learning_rate": 1.960180851474125e-05, + "loss": 1.0139, + "step": 4781 + }, + { + "epoch": 0.12, + "learning_rate": 1.9601586215408004e-05, + "loss": 1.067, + "step": 4782 + }, + { + "epoch": 0.12, + "learning_rate": 1.960136385530151e-05, + "loss": 1.0128, + "step": 4783 + }, + { + "epoch": 0.12, + "learning_rate": 1.960114143442317e-05, + "loss": 1.0025, + "step": 4784 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600918952774402e-05, + "loss": 1.0559, + "step": 4785 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600696410356606e-05, + "loss": 1.0926, + "step": 4786 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600473807171193e-05, + "loss": 1.0312, + "step": 4787 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600251143219572e-05, + "loss": 1.0502, + "step": 4788 + }, + { + "epoch": 0.12, + "learning_rate": 1.9600028418503155e-05, + "loss": 1.1074, + "step": 4789 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599805633023346e-05, + "loss": 1.0846, + "step": 4790 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599582786781558e-05, + "loss": 1.0171, + "step": 4791 + }, + { + "epoch": 0.12, + "learning_rate": 1.9599359879779202e-05, + "loss": 1.0507, + "step": 4792 + }, + { + "epoch": 0.12, + "learning_rate": 1.959913691201769e-05, + "loss": 1.0406, + "step": 4793 + }, + { + "epoch": 0.12, + "learning_rate": 1.959891388349843e-05, + "loss": 1.05, + "step": 4794 + }, + { + "epoch": 0.12, + "learning_rate": 1.9598690794222834e-05, + "loss": 1.091, + "step": 4795 + }, + { + "epoch": 0.12, + "learning_rate": 1.959846764419232e-05, + "loss": 1.0879, + "step": 4796 + }, + { + "epoch": 0.12, + "learning_rate": 1.9598244433408292e-05, + "loss": 1.0705, + "step": 4797 + }, + { + "epoch": 0.12, + "learning_rate": 1.959802116187217e-05, + "loss": 1.0772, + "step": 4798 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597797829585362e-05, + "loss": 1.0524, + "step": 4799 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597574436549284e-05, + "loss": 1.0745, + "step": 4800 + }, + { + "epoch": 0.12, + "learning_rate": 1.959735098276535e-05, + "loss": 1.0352, + "step": 4801 + }, + { + "epoch": 0.12, + "learning_rate": 1.9597127468234975e-05, + "loss": 1.0529, + "step": 4802 + }, + { + "epoch": 0.12, + "learning_rate": 1.959690389295957e-05, + "loss": 1.0232, + "step": 4803 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596680256940553e-05, + "loss": 1.0305, + "step": 4804 + }, + { + "epoch": 0.12, + "learning_rate": 1.959645656017934e-05, + "loss": 1.0681, + "step": 4805 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596232802677347e-05, + "loss": 1.0371, + "step": 4806 + }, + { + "epoch": 0.12, + "learning_rate": 1.9596008984435988e-05, + "loss": 1.0736, + "step": 4807 + }, + { + "epoch": 0.12, + "learning_rate": 1.959578510545668e-05, + "loss": 1.0408, + "step": 4808 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595561165740843e-05, + "loss": 1.0287, + "step": 4809 + }, + { + "epoch": 0.12, + "learning_rate": 1.959533716528989e-05, + "loss": 1.0865, + "step": 4810 + }, + { + "epoch": 0.12, + "learning_rate": 1.9595113104105245e-05, + "loss": 1.0636, + "step": 4811 + }, + { + "epoch": 0.12, + "learning_rate": 1.959488898218832e-05, + "loss": 1.0685, + "step": 4812 + }, + { + "epoch": 0.12, + "learning_rate": 1.959466479954054e-05, + "loss": 1.0124, + "step": 4813 + }, + { + "epoch": 0.12, + "learning_rate": 1.9594440556163317e-05, + "loss": 1.0538, + "step": 4814 + }, + { + "epoch": 0.12, + "learning_rate": 1.9594216252058072e-05, + "loss": 1.054, + "step": 4815 + }, + { + "epoch": 0.12, + "learning_rate": 1.959399188722623e-05, + "loss": 1.0445, + "step": 4816 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593767461669204e-05, + "loss": 1.0258, + "step": 4817 + }, + { + "epoch": 0.12, + "learning_rate": 1.959354297538842e-05, + "loss": 1.0434, + "step": 4818 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593318428385294e-05, + "loss": 1.0704, + "step": 4819 + }, + { + "epoch": 0.12, + "learning_rate": 1.9593093820661252e-05, + "loss": 1.026, + "step": 4820 + }, + { + "epoch": 0.12, + "learning_rate": 1.959286915221771e-05, + "loss": 1.0597, + "step": 4821 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592644423056097e-05, + "loss": 1.0443, + "step": 4822 + }, + { + "epoch": 0.12, + "learning_rate": 1.9592419633177834e-05, + "loss": 1.0543, + "step": 4823 + }, + { + "epoch": 0.12, + "learning_rate": 1.959219478258434e-05, + "loss": 1.0498, + "step": 4824 + }, + { + "epoch": 0.12, + "learning_rate": 1.959196987127704e-05, + "loss": 1.033, + "step": 4825 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591744899257358e-05, + "loss": 1.0083, + "step": 4826 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591519866526715e-05, + "loss": 1.0111, + "step": 4827 + }, + { + "epoch": 0.12, + "learning_rate": 1.959129477308654e-05, + "loss": 1.079, + "step": 4828 + }, + { + "epoch": 0.12, + "learning_rate": 1.9591069618938255e-05, + "loss": 1.0767, + "step": 4829 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590844404083284e-05, + "loss": 1.0734, + "step": 4830 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590619128523057e-05, + "loss": 1.0608, + "step": 4831 + }, + { + "epoch": 0.12, + "learning_rate": 1.9590393792258998e-05, + "loss": 1.0857, + "step": 4832 + }, + { + "epoch": 0.12, + "learning_rate": 1.959016839529253e-05, + "loss": 1.0363, + "step": 4833 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589942937625085e-05, + "loss": 1.0229, + "step": 4834 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589717419258086e-05, + "loss": 1.0403, + "step": 4835 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589491840192958e-05, + "loss": 1.0431, + "step": 4836 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589266200431136e-05, + "loss": 1.0499, + "step": 4837 + }, + { + "epoch": 0.12, + "learning_rate": 1.9589040499974043e-05, + "loss": 1.0385, + "step": 4838 + }, + { + "epoch": 0.12, + "learning_rate": 1.958881473882311e-05, + "loss": 1.0068, + "step": 4839 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588588916979764e-05, + "loss": 1.076, + "step": 4840 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588363034445438e-05, + "loss": 1.0888, + "step": 4841 + }, + { + "epoch": 0.12, + "learning_rate": 1.9588137091221556e-05, + "loss": 1.0341, + "step": 4842 + }, + { + "epoch": 0.12, + "learning_rate": 1.9587911087309553e-05, + "loss": 1.0464, + "step": 4843 + }, + { + "epoch": 0.12, + "learning_rate": 1.9587685022710855e-05, + "loss": 1.0708, + "step": 4844 + }, + { + "epoch": 0.12, + "learning_rate": 1.9587458897426898e-05, + "loss": 1.0323, + "step": 4845 + }, + { + "epoch": 0.12, + "learning_rate": 1.958723271145911e-05, + "loss": 1.055, + "step": 4846 + }, + { + "epoch": 0.12, + "learning_rate": 1.9587006464808924e-05, + "loss": 1.032, + "step": 4847 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586780157477772e-05, + "loss": 1.0422, + "step": 4848 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586553789467085e-05, + "loss": 1.064, + "step": 4849 + }, + { + "epoch": 0.12, + "learning_rate": 1.9586327360778298e-05, + "loss": 1.0409, + "step": 4850 + }, + { + "epoch": 0.12, + "learning_rate": 1.958610087141284e-05, + "loss": 1.0912, + "step": 4851 + }, + { + "epoch": 0.12, + "learning_rate": 1.958587432137215e-05, + "loss": 1.0987, + "step": 4852 + }, + { + "epoch": 0.12, + "learning_rate": 1.958564771065766e-05, + "loss": 1.0915, + "step": 4853 + }, + { + "epoch": 0.12, + "learning_rate": 1.95854210392708e-05, + "loss": 1.1108, + "step": 4854 + }, + { + "epoch": 0.12, + "learning_rate": 1.9585194307213013e-05, + "loss": 1.0856, + "step": 4855 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584967514485728e-05, + "loss": 1.077, + "step": 4856 + }, + { + "epoch": 0.12, + "learning_rate": 1.9584740661090383e-05, + "loss": 1.0886, + "step": 4857 + }, + { + "epoch": 0.12, + "learning_rate": 1.958451374702841e-05, + "loss": 1.0917, + "step": 4858 + }, + { + "epoch": 0.12, + "learning_rate": 1.958428677230125e-05, + "loss": 1.0643, + "step": 4859 + }, + { + "epoch": 0.12, + "learning_rate": 1.958405973691034e-05, + "loss": 1.064, + "step": 4860 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583832640857106e-05, + "loss": 1.0786, + "step": 4861 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583605484143003e-05, + "loss": 1.1019, + "step": 4862 + }, + { + "epoch": 0.12, + "learning_rate": 1.958337826676946e-05, + "loss": 1.0582, + "step": 4863 + }, + { + "epoch": 0.12, + "learning_rate": 1.9583150988737913e-05, + "loss": 1.0771, + "step": 4864 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582923650049802e-05, + "loss": 1.0935, + "step": 4865 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582696250706564e-05, + "loss": 1.0609, + "step": 4866 + }, + { + "epoch": 0.12, + "learning_rate": 1.9582468790709644e-05, + "loss": 1.0618, + "step": 4867 + }, + { + "epoch": 0.12, + "learning_rate": 1.958224127006048e-05, + "loss": 1.0452, + "step": 4868 + }, + { + "epoch": 0.12, + "learning_rate": 1.958201368876051e-05, + "loss": 1.0863, + "step": 4869 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581786046811175e-05, + "loss": 1.0502, + "step": 4870 + }, + { + "epoch": 0.12, + "learning_rate": 1.958155834421392e-05, + "loss": 1.06, + "step": 4871 + }, + { + "epoch": 0.12, + "learning_rate": 1.958133058097018e-05, + "loss": 1.0706, + "step": 4872 + }, + { + "epoch": 0.12, + "learning_rate": 1.9581102757081392e-05, + "loss": 1.0869, + "step": 4873 + }, + { + "epoch": 0.12, + "learning_rate": 1.958087487254901e-05, + "loss": 1.021, + "step": 4874 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580646927374472e-05, + "loss": 1.0141, + "step": 4875 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580418921559225e-05, + "loss": 1.0099, + "step": 4876 + }, + { + "epoch": 0.12, + "learning_rate": 1.9580190855104702e-05, + "loss": 1.0494, + "step": 4877 + }, + { + "epoch": 0.12, + "learning_rate": 1.957996272801235e-05, + "loss": 1.0731, + "step": 4878 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579734540283616e-05, + "loss": 1.0583, + "step": 4879 + }, + { + "epoch": 0.12, + "learning_rate": 1.9579506291919945e-05, + "loss": 1.0617, + "step": 4880 + }, + { + "epoch": 0.12, + "learning_rate": 1.957927798292278e-05, + "loss": 1.0455, + "step": 4881 + }, + { + "epoch": 0.12, + "learning_rate": 1.957904961329357e-05, + "loss": 1.0741, + "step": 4882 + }, + { + "epoch": 0.12, + "learning_rate": 1.957882118303375e-05, + "loss": 1.0417, + "step": 4883 + }, + { + "epoch": 0.12, + "learning_rate": 1.957859269214477e-05, + "loss": 1.0518, + "step": 4884 + }, + { + "epoch": 0.12, + "learning_rate": 1.9578364140628086e-05, + "loss": 1.0657, + "step": 4885 + }, + { + "epoch": 0.12, + "learning_rate": 1.957813552848513e-05, + "loss": 1.0587, + "step": 4886 + }, + { + "epoch": 0.12, + "learning_rate": 1.957790685571736e-05, + "loss": 1.0233, + "step": 4887 + }, + { + "epoch": 0.12, + "learning_rate": 1.9577678122326222e-05, + "loss": 1.0221, + "step": 4888 + }, + { + "epoch": 0.12, + "learning_rate": 1.957744932831316e-05, + "loss": 1.0486, + "step": 4889 + }, + { + "epoch": 0.12, + "learning_rate": 1.9577220473679622e-05, + "loss": 1.077, + "step": 4890 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576991558427058e-05, + "loss": 1.0739, + "step": 4891 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576762582556913e-05, + "loss": 1.0365, + "step": 4892 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576533546070647e-05, + "loss": 1.0747, + "step": 4893 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576304448969697e-05, + "loss": 1.0886, + "step": 4894 + }, + { + "epoch": 0.12, + "learning_rate": 1.9576075291255523e-05, + "loss": 1.0483, + "step": 4895 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575846072929573e-05, + "loss": 1.0839, + "step": 4896 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575616793993292e-05, + "loss": 1.0452, + "step": 4897 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575387454448135e-05, + "loss": 1.0541, + "step": 4898 + }, + { + "epoch": 0.12, + "learning_rate": 1.9575158054295554e-05, + "loss": 1.0599, + "step": 4899 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574928593537007e-05, + "loss": 1.0976, + "step": 4900 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574699072173934e-05, + "loss": 1.0423, + "step": 4901 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574469490207798e-05, + "loss": 1.0789, + "step": 4902 + }, + { + "epoch": 0.12, + "learning_rate": 1.9574239847640046e-05, + "loss": 1.0309, + "step": 4903 + }, + { + "epoch": 0.12, + "learning_rate": 1.957401014447213e-05, + "loss": 1.0601, + "step": 4904 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573780380705515e-05, + "loss": 1.0804, + "step": 4905 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573550556341646e-05, + "loss": 1.0304, + "step": 4906 + }, + { + "epoch": 0.12, + "learning_rate": 1.957332067138198e-05, + "loss": 1.0733, + "step": 4907 + }, + { + "epoch": 0.12, + "learning_rate": 1.9573090725827974e-05, + "loss": 1.0285, + "step": 4908 + }, + { + "epoch": 0.12, + "learning_rate": 1.9572860719681074e-05, + "loss": 1.0775, + "step": 4909 + }, + { + "epoch": 0.12, + "learning_rate": 1.957263065294275e-05, + "loss": 1.0422, + "step": 4910 + }, + { + "epoch": 0.12, + "learning_rate": 1.9572400525614446e-05, + "loss": 1.0916, + "step": 4911 + }, + { + "epoch": 0.12, + "learning_rate": 1.9572170337697626e-05, + "loss": 1.0781, + "step": 4912 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571940089193746e-05, + "loss": 1.0358, + "step": 4913 + }, + { + "epoch": 0.12, + "learning_rate": 1.957170978010426e-05, + "loss": 1.0486, + "step": 4914 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571479410430628e-05, + "loss": 1.0798, + "step": 4915 + }, + { + "epoch": 0.12, + "learning_rate": 1.957124898017431e-05, + "loss": 1.0449, + "step": 4916 + }, + { + "epoch": 0.12, + "learning_rate": 1.9571018489336763e-05, + "loss": 1.0414, + "step": 4917 + }, + { + "epoch": 0.12, + "learning_rate": 1.957078793791944e-05, + "loss": 1.0476, + "step": 4918 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570557325923812e-05, + "loss": 1.061, + "step": 4919 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570326653351328e-05, + "loss": 1.0553, + "step": 4920 + }, + { + "epoch": 0.12, + "learning_rate": 1.9570095920203455e-05, + "loss": 1.0756, + "step": 4921 + }, + { + "epoch": 0.12, + "learning_rate": 1.956986512648165e-05, + "loss": 1.0441, + "step": 4922 + }, + { + "epoch": 0.12, + "learning_rate": 1.9569634272187376e-05, + "loss": 1.0628, + "step": 4923 + }, + { + "epoch": 0.12, + "learning_rate": 1.956940335732209e-05, + "loss": 1.0671, + "step": 4924 + }, + { + "epoch": 0.12, + "learning_rate": 1.956917238188726e-05, + "loss": 1.0734, + "step": 4925 + }, + { + "epoch": 0.12, + "learning_rate": 1.956894134588434e-05, + "loss": 1.0506, + "step": 4926 + }, + { + "epoch": 0.12, + "learning_rate": 1.95687102493148e-05, + "loss": 1.0468, + "step": 4927 + }, + { + "epoch": 0.12, + "learning_rate": 1.95684790921801e-05, + "loss": 1.0103, + "step": 4928 + }, + { + "epoch": 0.12, + "learning_rate": 1.95682478744817e-05, + "loss": 1.0372, + "step": 4929 + }, + { + "epoch": 0.12, + "learning_rate": 1.9568016596221064e-05, + "loss": 1.0704, + "step": 4930 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567785257399663e-05, + "loss": 1.04, + "step": 4931 + }, + { + "epoch": 0.12, + "learning_rate": 1.956755385801895e-05, + "loss": 1.0349, + "step": 4932 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567322398080405e-05, + "loss": 1.0543, + "step": 4933 + }, + { + "epoch": 0.12, + "learning_rate": 1.9567090877585477e-05, + "loss": 1.0095, + "step": 4934 + }, + { + "epoch": 0.12, + "learning_rate": 1.956685929653564e-05, + "loss": 1.0364, + "step": 4935 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566627654932357e-05, + "loss": 1.0717, + "step": 4936 + }, + { + "epoch": 0.12, + "learning_rate": 1.9566395952777095e-05, + "loss": 1.0626, + "step": 4937 + }, + { + "epoch": 0.12, + "learning_rate": 1.956616419007132e-05, + "loss": 1.0379, + "step": 4938 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565932366816504e-05, + "loss": 1.0303, + "step": 4939 + }, + { + "epoch": 0.12, + "learning_rate": 1.956570048301411e-05, + "loss": 1.0553, + "step": 4940 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565468538665602e-05, + "loss": 1.0616, + "step": 4941 + }, + { + "epoch": 0.12, + "learning_rate": 1.956523653377245e-05, + "loss": 1.0921, + "step": 4942 + }, + { + "epoch": 0.12, + "learning_rate": 1.9565004468336132e-05, + "loss": 1.0357, + "step": 4943 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564772342358103e-05, + "loss": 1.0636, + "step": 4944 + }, + { + "epoch": 0.12, + "learning_rate": 1.956454015583984e-05, + "loss": 1.0581, + "step": 4945 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564307908782812e-05, + "loss": 1.0616, + "step": 4946 + }, + { + "epoch": 0.12, + "learning_rate": 1.9564075601188485e-05, + "loss": 1.0708, + "step": 4947 + }, + { + "epoch": 0.12, + "learning_rate": 1.956384323305834e-05, + "loss": 1.0616, + "step": 4948 + }, + { + "epoch": 0.12, + "learning_rate": 1.956361080439383e-05, + "loss": 0.9885, + "step": 4949 + }, + { + "epoch": 0.12, + "learning_rate": 1.9563378315196443e-05, + "loss": 1.0078, + "step": 4950 + }, + { + "epoch": 0.12, + "learning_rate": 1.956314576546764e-05, + "loss": 1.0287, + "step": 4951 + }, + { + "epoch": 0.12, + "learning_rate": 1.95629131552089e-05, + "loss": 1.0443, + "step": 4952 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562680484421694e-05, + "loss": 1.0266, + "step": 4953 + }, + { + "epoch": 0.12, + "learning_rate": 1.9562447753107486e-05, + "loss": 1.0466, + "step": 4954 + }, + { + "epoch": 0.12, + "learning_rate": 1.956221496126776e-05, + "loss": 1.0372, + "step": 4955 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561982108903986e-05, + "loss": 1.0776, + "step": 4956 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561749196017637e-05, + "loss": 1.0512, + "step": 4957 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561516222610185e-05, + "loss": 1.0543, + "step": 4958 + }, + { + "epoch": 0.12, + "learning_rate": 1.956128318868311e-05, + "loss": 1.0505, + "step": 4959 + }, + { + "epoch": 0.12, + "learning_rate": 1.9561050094237882e-05, + "loss": 1.0418, + "step": 4960 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560816939275978e-05, + "loss": 1.0258, + "step": 4961 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560583723798876e-05, + "loss": 1.0349, + "step": 4962 + }, + { + "epoch": 0.12, + "learning_rate": 1.956035044780805e-05, + "loss": 1.0604, + "step": 4963 + }, + { + "epoch": 0.12, + "learning_rate": 1.9560117111304973e-05, + "loss": 1.0402, + "step": 4964 + }, + { + "epoch": 0.12, + "learning_rate": 1.955988371429113e-05, + "loss": 1.0192, + "step": 4965 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559650256767994e-05, + "loss": 1.0493, + "step": 4966 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559416738737042e-05, + "loss": 1.0489, + "step": 4967 + }, + { + "epoch": 0.12, + "learning_rate": 1.9559183160199752e-05, + "loss": 1.0397, + "step": 4968 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558949521157604e-05, + "loss": 1.0478, + "step": 4969 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558715821612076e-05, + "loss": 1.0745, + "step": 4970 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558482061564645e-05, + "loss": 1.0384, + "step": 4971 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558248241016798e-05, + "loss": 1.0308, + "step": 4972 + }, + { + "epoch": 0.12, + "learning_rate": 1.9558014359970005e-05, + "loss": 1.036, + "step": 4973 + }, + { + "epoch": 0.12, + "learning_rate": 1.955778041842575e-05, + "loss": 1.0049, + "step": 4974 + }, + { + "epoch": 0.12, + "learning_rate": 1.9557546416385516e-05, + "loss": 1.0428, + "step": 4975 + }, + { + "epoch": 0.12, + "learning_rate": 1.9557312353850785e-05, + "loss": 1.0469, + "step": 4976 + }, + { + "epoch": 0.12, + "learning_rate": 1.9557078230823037e-05, + "loss": 1.0164, + "step": 4977 + }, + { + "epoch": 0.12, + "learning_rate": 1.955684404730375e-05, + "loss": 1.0374, + "step": 4978 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556609803294412e-05, + "loss": 1.0284, + "step": 4979 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556375498796503e-05, + "loss": 1.0315, + "step": 4980 + }, + { + "epoch": 0.12, + "learning_rate": 1.9556141133811507e-05, + "loss": 1.0102, + "step": 4981 + }, + { + "epoch": 0.12, + "learning_rate": 1.95559067083409e-05, + "loss": 1.0693, + "step": 4982 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555672222386183e-05, + "loss": 1.0113, + "step": 4983 + }, + { + "epoch": 0.12, + "learning_rate": 1.955543767594882e-05, + "loss": 0.9881, + "step": 4984 + }, + { + "epoch": 0.12, + "learning_rate": 1.9555203069030312e-05, + "loss": 1.0427, + "step": 4985 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554968401632135e-05, + "loss": 1.0452, + "step": 4986 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554733673755774e-05, + "loss": 1.0523, + "step": 4987 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554498885402718e-05, + "loss": 1.0552, + "step": 4988 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554264036574454e-05, + "loss": 0.9931, + "step": 4989 + }, + { + "epoch": 0.12, + "learning_rate": 1.9554029127272466e-05, + "loss": 1.0349, + "step": 4990 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553794157498244e-05, + "loss": 1.0197, + "step": 4991 + }, + { + "epoch": 0.12, + "learning_rate": 1.955355912725327e-05, + "loss": 1.0836, + "step": 4992 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553324036539037e-05, + "loss": 1.0378, + "step": 4993 + }, + { + "epoch": 0.12, + "learning_rate": 1.9553088885357026e-05, + "loss": 1.0753, + "step": 4994 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552853673708734e-05, + "loss": 1.0216, + "step": 4995 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552618401595643e-05, + "loss": 1.0143, + "step": 4996 + }, + { + "epoch": 0.12, + "learning_rate": 1.9552383069019246e-05, + "loss": 1.0326, + "step": 4997 + }, + { + "epoch": 0.12, + "learning_rate": 1.955214767598103e-05, + "loss": 1.0674, + "step": 4998 + }, + { + "epoch": 0.12, + "learning_rate": 1.9551912222482488e-05, + "loss": 1.0266, + "step": 4999 + }, + { + "epoch": 0.12, + "learning_rate": 1.955167670852511e-05, + "loss": 1.0399, + "step": 5000 + } + ], + "max_steps": 40710, + "num_train_epochs": 1, + "total_flos": 4.7325016486041354e+19, + "trial_name": null, + "trial_params": null +}