{ "best_metric": null, "best_model_checkpoint": null, "epoch": 2.8715003589375447, "global_step": 8000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0, "loss": 1.1808, "step": 1 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.0975, "step": 2 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.0421, "step": 3 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.1004, "step": 4 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.1075, "step": 5 }, { "epoch": 0.0, "learning_rate": 0, "loss": 0.9785, "step": 6 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.0688, "step": 7 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.1002, "step": 8 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.0137, "step": 9 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.0715, "step": 10 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.0862, "step": 11 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.1345, "step": 12 }, { "epoch": 0.0, "learning_rate": 0, "loss": 1.1899, "step": 13 }, { "epoch": 0.01, "learning_rate": 0, "loss": 1.0586, "step": 14 }, { "epoch": 0.01, "learning_rate": 0, "loss": 1.0067, "step": 15 }, { "epoch": 0.01, "learning_rate": 0, "loss": 1.0933, "step": 16 }, { "epoch": 0.01, "learning_rate": 0, "loss": 1.0936, "step": 17 }, { "epoch": 0.01, "learning_rate": 0, "loss": 1.1113, "step": 18 }, { "epoch": 0.01, "learning_rate": 0.0, "loss": 1.2256, "step": 19 }, { "epoch": 0.01, "learning_rate": 3.3333333333333335e-07, "loss": 1.0188, "step": 20 }, { "epoch": 0.01, "learning_rate": 6.666666666666667e-07, "loss": 1.0319, "step": 21 }, { "epoch": 0.01, "learning_rate": 1.0000000000000002e-06, "loss": 0.9406, "step": 22 }, { "epoch": 0.01, "learning_rate": 1.3333333333333334e-06, "loss": 0.9155, "step": 23 }, { "epoch": 0.01, "learning_rate": 1.6666666666666667e-06, "loss": 1.1723, "step": 24 }, { "epoch": 0.01, "learning_rate": 2.0000000000000003e-06, "loss": 0.9988, "step": 25 }, { "epoch": 0.01, "learning_rate": 2.3333333333333336e-06, "loss": 1.0594, "step": 26 }, { "epoch": 0.01, "learning_rate": 2.666666666666667e-06, "loss": 0.8641, "step": 27 }, { "epoch": 0.01, "learning_rate": 3e-06, "loss": 0.9298, "step": 28 }, { "epoch": 0.01, "learning_rate": 3e-06, "loss": 0.9047, "step": 29 }, { "epoch": 0.01, "learning_rate": 3.3333333333333333e-06, "loss": 1.0014, "step": 30 }, { "epoch": 0.01, "learning_rate": 3.6666666666666666e-06, "loss": 0.9678, "step": 31 }, { "epoch": 0.01, "learning_rate": 4.000000000000001e-06, "loss": 1.0063, "step": 32 }, { "epoch": 0.01, "learning_rate": 4.333333333333334e-06, "loss": 0.9735, "step": 33 }, { "epoch": 0.01, "learning_rate": 4.666666666666667e-06, "loss": 0.989, "step": 34 }, { "epoch": 0.01, "learning_rate": 5e-06, "loss": 0.9786, "step": 35 }, { "epoch": 0.01, "learning_rate": 5.333333333333334e-06, "loss": 0.9329, "step": 36 }, { "epoch": 0.01, "learning_rate": 5.666666666666667e-06, "loss": 0.9945, "step": 37 }, { "epoch": 0.01, "learning_rate": 6e-06, "loss": 0.8924, "step": 38 }, { "epoch": 0.01, "learning_rate": 6.333333333333333e-06, "loss": 1.041, "step": 39 }, { "epoch": 0.01, "learning_rate": 6.666666666666667e-06, "loss": 0.9236, "step": 40 }, { "epoch": 0.01, "learning_rate": 7e-06, "loss": 0.8724, "step": 41 }, { "epoch": 0.02, "learning_rate": 7.333333333333333e-06, "loss": 0.9631, "step": 42 }, { "epoch": 0.02, "learning_rate": 7.666666666666667e-06, "loss": 0.8526, "step": 43 }, { "epoch": 0.02, "learning_rate": 8.000000000000001e-06, "loss": 0.9662, "step": 44 }, { "epoch": 0.02, "learning_rate": 8.333333333333334e-06, "loss": 0.9569, "step": 45 }, { "epoch": 0.02, "learning_rate": 8.666666666666668e-06, "loss": 0.9834, "step": 46 }, { "epoch": 0.02, "learning_rate": 9e-06, "loss": 1.057, "step": 47 }, { "epoch": 0.02, "learning_rate": 9.333333333333334e-06, "loss": 0.8681, "step": 48 }, { "epoch": 0.02, "learning_rate": 9.666666666666667e-06, "loss": 0.9708, "step": 49 }, { "epoch": 0.02, "learning_rate": 1e-05, "loss": 0.861, "step": 50 }, { "epoch": 0.02, "learning_rate": 1e-05, "loss": 0.9933, "step": 51 }, { "epoch": 0.02, "learning_rate": 9.999100233939176e-06, "loss": 0.9161, "step": 52 }, { "epoch": 0.02, "learning_rate": 9.998200467878352e-06, "loss": 0.8866, "step": 53 }, { "epoch": 0.02, "learning_rate": 9.997300701817529e-06, "loss": 0.9641, "step": 54 }, { "epoch": 0.02, "learning_rate": 9.996400935756704e-06, "loss": 1.0002, "step": 55 }, { "epoch": 0.02, "learning_rate": 9.99550116969588e-06, "loss": 0.9254, "step": 56 }, { "epoch": 0.02, "learning_rate": 9.994601403635055e-06, "loss": 0.9598, "step": 57 }, { "epoch": 0.02, "learning_rate": 9.993701637574232e-06, "loss": 0.9077, "step": 58 }, { "epoch": 0.02, "learning_rate": 9.992801871513409e-06, "loss": 1.0076, "step": 59 }, { "epoch": 0.02, "learning_rate": 9.991902105452584e-06, "loss": 0.9566, "step": 60 }, { "epoch": 0.02, "learning_rate": 9.991002339391759e-06, "loss": 0.9241, "step": 61 }, { "epoch": 0.02, "learning_rate": 9.990102573330933e-06, "loss": 0.9878, "step": 62 }, { "epoch": 0.02, "learning_rate": 9.98920280727011e-06, "loss": 1.0074, "step": 63 }, { "epoch": 0.02, "learning_rate": 9.988303041209287e-06, "loss": 0.9124, "step": 64 }, { "epoch": 0.02, "learning_rate": 9.987403275148462e-06, "loss": 0.8968, "step": 65 }, { "epoch": 0.02, "learning_rate": 9.986503509087638e-06, "loss": 0.9743, "step": 66 }, { "epoch": 0.02, "learning_rate": 9.985603743026813e-06, "loss": 0.9691, "step": 67 }, { "epoch": 0.02, "learning_rate": 9.98470397696599e-06, "loss": 0.9524, "step": 68 }, { "epoch": 0.02, "learning_rate": 9.983804210905166e-06, "loss": 1.0561, "step": 69 }, { "epoch": 0.03, "learning_rate": 9.982904444844341e-06, "loss": 0.9845, "step": 70 }, { "epoch": 0.03, "learning_rate": 9.982004678783516e-06, "loss": 0.9321, "step": 71 }, { "epoch": 0.03, "learning_rate": 9.981104912722693e-06, "loss": 0.9795, "step": 72 }, { "epoch": 0.03, "learning_rate": 9.98020514666187e-06, "loss": 0.9491, "step": 73 }, { "epoch": 0.03, "learning_rate": 9.979305380601044e-06, "loss": 0.9665, "step": 74 }, { "epoch": 0.03, "learning_rate": 9.978405614540221e-06, "loss": 0.9965, "step": 75 }, { "epoch": 0.03, "learning_rate": 9.977505848479396e-06, "loss": 0.8404, "step": 76 }, { "epoch": 0.03, "learning_rate": 9.976606082418573e-06, "loss": 0.9704, "step": 77 }, { "epoch": 0.03, "learning_rate": 9.975706316357749e-06, "loss": 0.8572, "step": 78 }, { "epoch": 0.03, "learning_rate": 9.974806550296924e-06, "loss": 1.1222, "step": 79 }, { "epoch": 0.03, "learning_rate": 9.973906784236099e-06, "loss": 0.9126, "step": 80 }, { "epoch": 0.03, "learning_rate": 9.973007018175274e-06, "loss": 1.0166, "step": 81 }, { "epoch": 0.03, "learning_rate": 9.97210725211445e-06, "loss": 0.8678, "step": 82 }, { "epoch": 0.03, "learning_rate": 9.971207486053627e-06, "loss": 0.844, "step": 83 }, { "epoch": 0.03, "learning_rate": 9.970307719992802e-06, "loss": 0.8966, "step": 84 }, { "epoch": 0.03, "learning_rate": 9.969407953931979e-06, "loss": 0.9052, "step": 85 }, { "epoch": 0.03, "learning_rate": 9.968508187871154e-06, "loss": 0.8722, "step": 86 }, { "epoch": 0.03, "learning_rate": 9.96760842181033e-06, "loss": 0.9735, "step": 87 }, { "epoch": 0.03, "learning_rate": 9.966708655749507e-06, "loss": 0.9215, "step": 88 }, { "epoch": 0.03, "learning_rate": 9.965808889688682e-06, "loss": 0.8831, "step": 89 }, { "epoch": 0.03, "learning_rate": 9.964909123627857e-06, "loss": 1.0686, "step": 90 }, { "epoch": 0.03, "learning_rate": 9.964009357567033e-06, "loss": 0.8869, "step": 91 }, { "epoch": 0.03, "learning_rate": 9.96310959150621e-06, "loss": 0.8772, "step": 92 }, { "epoch": 0.03, "learning_rate": 9.962209825445385e-06, "loss": 1.0249, "step": 93 }, { "epoch": 0.03, "learning_rate": 9.961310059384561e-06, "loss": 0.8962, "step": 94 }, { "epoch": 0.03, "learning_rate": 9.960410293323736e-06, "loss": 0.9598, "step": 95 }, { "epoch": 0.03, "learning_rate": 9.959510527262913e-06, "loss": 1.0345, "step": 96 }, { "epoch": 0.03, "learning_rate": 9.95861076120209e-06, "loss": 0.9049, "step": 97 }, { "epoch": 0.04, "learning_rate": 9.957710995141264e-06, "loss": 0.9306, "step": 98 }, { "epoch": 0.04, "learning_rate": 9.95681122908044e-06, "loss": 0.8264, "step": 99 }, { "epoch": 0.04, "learning_rate": 9.955911463019616e-06, "loss": 0.9358, "step": 100 }, { "epoch": 0.04, "learning_rate": 9.955011696958791e-06, "loss": 0.9209, "step": 101 }, { "epoch": 0.04, "learning_rate": 9.954111930897968e-06, "loss": 0.9096, "step": 102 }, { "epoch": 0.04, "learning_rate": 9.953212164837142e-06, "loss": 0.9026, "step": 103 }, { "epoch": 0.04, "learning_rate": 9.952312398776319e-06, "loss": 1.0474, "step": 104 }, { "epoch": 0.04, "learning_rate": 9.951412632715494e-06, "loss": 0.8598, "step": 105 }, { "epoch": 0.04, "learning_rate": 9.95051286665467e-06, "loss": 1.0015, "step": 106 }, { "epoch": 0.04, "learning_rate": 9.949613100593847e-06, "loss": 0.877, "step": 107 }, { "epoch": 0.04, "learning_rate": 9.948713334533022e-06, "loss": 0.9943, "step": 108 }, { "epoch": 0.04, "learning_rate": 9.947813568472197e-06, "loss": 0.9769, "step": 109 }, { "epoch": 0.04, "learning_rate": 9.946913802411374e-06, "loss": 0.8456, "step": 110 }, { "epoch": 0.04, "learning_rate": 9.94601403635055e-06, "loss": 0.8459, "step": 111 }, { "epoch": 0.04, "learning_rate": 9.945114270289725e-06, "loss": 0.8737, "step": 112 }, { "epoch": 0.04, "learning_rate": 9.944214504228902e-06, "loss": 0.9534, "step": 113 }, { "epoch": 0.04, "learning_rate": 9.943314738168077e-06, "loss": 1.0665, "step": 114 }, { "epoch": 0.04, "learning_rate": 9.942414972107253e-06, "loss": 0.8839, "step": 115 }, { "epoch": 0.04, "learning_rate": 9.94151520604643e-06, "loss": 1.0453, "step": 116 }, { "epoch": 0.04, "learning_rate": 9.940615439985605e-06, "loss": 1.0384, "step": 117 }, { "epoch": 0.04, "learning_rate": 9.93971567392478e-06, "loss": 0.8788, "step": 118 }, { "epoch": 0.04, "learning_rate": 9.938815907863956e-06, "loss": 0.9417, "step": 119 }, { "epoch": 0.04, "learning_rate": 9.937916141803131e-06, "loss": 0.8661, "step": 120 }, { "epoch": 0.04, "learning_rate": 9.937016375742308e-06, "loss": 0.89, "step": 121 }, { "epoch": 0.04, "learning_rate": 9.936116609681483e-06, "loss": 0.9592, "step": 122 }, { "epoch": 0.04, "learning_rate": 9.93521684362066e-06, "loss": 0.9254, "step": 123 }, { "epoch": 0.04, "learning_rate": 9.934317077559834e-06, "loss": 0.9434, "step": 124 }, { "epoch": 0.04, "learning_rate": 9.933417311499011e-06, "loss": 0.9794, "step": 125 }, { "epoch": 0.05, "learning_rate": 9.932517545438188e-06, "loss": 0.814, "step": 126 }, { "epoch": 0.05, "learning_rate": 9.931617779377363e-06, "loss": 0.9172, "step": 127 }, { "epoch": 0.05, "learning_rate": 9.930718013316538e-06, "loss": 1.0689, "step": 128 }, { "epoch": 0.05, "learning_rate": 9.929818247255714e-06, "loss": 0.8845, "step": 129 }, { "epoch": 0.05, "learning_rate": 9.92891848119489e-06, "loss": 0.837, "step": 130 }, { "epoch": 0.05, "learning_rate": 9.928018715134066e-06, "loss": 0.8144, "step": 131 }, { "epoch": 0.05, "learning_rate": 9.927118949073242e-06, "loss": 0.8934, "step": 132 }, { "epoch": 0.05, "learning_rate": 9.926219183012417e-06, "loss": 0.8767, "step": 133 }, { "epoch": 0.05, "learning_rate": 9.925319416951594e-06, "loss": 0.9946, "step": 134 }, { "epoch": 0.05, "learning_rate": 9.92441965089077e-06, "loss": 0.9454, "step": 135 }, { "epoch": 0.05, "learning_rate": 9.923519884829945e-06, "loss": 0.9944, "step": 136 }, { "epoch": 0.05, "learning_rate": 9.92262011876912e-06, "loss": 0.9428, "step": 137 }, { "epoch": 0.05, "learning_rate": 9.921720352708297e-06, "loss": 0.866, "step": 138 }, { "epoch": 0.05, "learning_rate": 9.920820586647474e-06, "loss": 0.9645, "step": 139 }, { "epoch": 0.05, "learning_rate": 9.919920820586648e-06, "loss": 0.9882, "step": 140 }, { "epoch": 0.05, "learning_rate": 9.919021054525823e-06, "loss": 1.0016, "step": 141 }, { "epoch": 0.05, "learning_rate": 9.918121288465e-06, "loss": 0.9208, "step": 142 }, { "epoch": 0.05, "learning_rate": 9.917221522404175e-06, "loss": 0.8817, "step": 143 }, { "epoch": 0.05, "learning_rate": 9.916321756343352e-06, "loss": 0.9716, "step": 144 }, { "epoch": 0.05, "learning_rate": 9.915421990282528e-06, "loss": 0.9992, "step": 145 }, { "epoch": 0.05, "learning_rate": 9.914522224221703e-06, "loss": 0.8935, "step": 146 }, { "epoch": 0.05, "learning_rate": 9.913622458160878e-06, "loss": 0.8735, "step": 147 }, { "epoch": 0.05, "learning_rate": 9.912722692100055e-06, "loss": 0.8662, "step": 148 }, { "epoch": 0.05, "learning_rate": 9.911822926039231e-06, "loss": 0.9648, "step": 149 }, { "epoch": 0.05, "learning_rate": 9.910923159978406e-06, "loss": 0.922, "step": 150 }, { "epoch": 0.05, "learning_rate": 9.910023393917583e-06, "loss": 1.0019, "step": 151 }, { "epoch": 0.05, "learning_rate": 9.909123627856758e-06, "loss": 0.9107, "step": 152 }, { "epoch": 0.05, "learning_rate": 9.908223861795934e-06, "loss": 0.8982, "step": 153 }, { "epoch": 0.06, "learning_rate": 9.907324095735111e-06, "loss": 0.9812, "step": 154 }, { "epoch": 0.06, "learning_rate": 9.906424329674286e-06, "loss": 0.7969, "step": 155 }, { "epoch": 0.06, "learning_rate": 9.90552456361346e-06, "loss": 0.9178, "step": 156 }, { "epoch": 0.06, "learning_rate": 9.904624797552637e-06, "loss": 0.8655, "step": 157 }, { "epoch": 0.06, "learning_rate": 9.903725031491814e-06, "loss": 0.9484, "step": 158 }, { "epoch": 0.06, "learning_rate": 9.902825265430989e-06, "loss": 0.8646, "step": 159 }, { "epoch": 0.06, "learning_rate": 9.901925499370164e-06, "loss": 1.0642, "step": 160 }, { "epoch": 0.06, "learning_rate": 9.90102573330934e-06, "loss": 1.0673, "step": 161 }, { "epoch": 0.06, "learning_rate": 9.900125967248515e-06, "loss": 0.9983, "step": 162 }, { "epoch": 0.06, "learning_rate": 9.899226201187692e-06, "loss": 0.8179, "step": 163 }, { "epoch": 0.06, "learning_rate": 9.898326435126869e-06, "loss": 1.0134, "step": 164 }, { "epoch": 0.06, "learning_rate": 9.897426669066044e-06, "loss": 0.9437, "step": 165 }, { "epoch": 0.06, "learning_rate": 9.896526903005218e-06, "loss": 0.9421, "step": 166 }, { "epoch": 0.06, "learning_rate": 9.895627136944395e-06, "loss": 1.0129, "step": 167 }, { "epoch": 0.06, "learning_rate": 9.894727370883572e-06, "loss": 0.8048, "step": 168 }, { "epoch": 0.06, "learning_rate": 9.893827604822747e-06, "loss": 0.8791, "step": 169 }, { "epoch": 0.06, "learning_rate": 9.892927838761923e-06, "loss": 0.9347, "step": 170 }, { "epoch": 0.06, "learning_rate": 9.892028072701098e-06, "loss": 0.8988, "step": 171 }, { "epoch": 0.06, "learning_rate": 9.891128306640275e-06, "loss": 0.8523, "step": 172 }, { "epoch": 0.06, "learning_rate": 9.890228540579451e-06, "loss": 0.9922, "step": 173 }, { "epoch": 0.06, "learning_rate": 9.889328774518626e-06, "loss": 0.9518, "step": 174 }, { "epoch": 0.06, "learning_rate": 9.888429008457801e-06, "loss": 0.9386, "step": 175 }, { "epoch": 0.06, "learning_rate": 9.887529242396978e-06, "loss": 0.8858, "step": 176 }, { "epoch": 0.06, "learning_rate": 9.886629476336154e-06, "loss": 0.9528, "step": 177 }, { "epoch": 0.06, "learning_rate": 9.88572971027533e-06, "loss": 1.0115, "step": 178 }, { "epoch": 0.06, "learning_rate": 9.884829944214504e-06, "loss": 0.9216, "step": 179 }, { "epoch": 0.06, "learning_rate": 9.883930178153681e-06, "loss": 0.9166, "step": 180 }, { "epoch": 0.06, "learning_rate": 9.883030412092856e-06, "loss": 0.9155, "step": 181 }, { "epoch": 0.07, "learning_rate": 9.882130646032032e-06, "loss": 0.9347, "step": 182 }, { "epoch": 0.07, "learning_rate": 9.881230879971209e-06, "loss": 0.9241, "step": 183 }, { "epoch": 0.07, "learning_rate": 9.880331113910384e-06, "loss": 0.7903, "step": 184 }, { "epoch": 0.07, "learning_rate": 9.879431347849559e-06, "loss": 1.0577, "step": 185 }, { "epoch": 0.07, "learning_rate": 9.878531581788736e-06, "loss": 0.8351, "step": 186 }, { "epoch": 0.07, "learning_rate": 9.877631815727912e-06, "loss": 0.8954, "step": 187 }, { "epoch": 0.07, "learning_rate": 9.876732049667087e-06, "loss": 0.9258, "step": 188 }, { "epoch": 0.07, "learning_rate": 9.875832283606264e-06, "loss": 0.8713, "step": 189 }, { "epoch": 0.07, "learning_rate": 9.874932517545439e-06, "loss": 0.9698, "step": 190 }, { "epoch": 0.07, "learning_rate": 9.874032751484615e-06, "loss": 0.8559, "step": 191 }, { "epoch": 0.07, "learning_rate": 9.873132985423792e-06, "loss": 0.9477, "step": 192 }, { "epoch": 0.07, "learning_rate": 9.872233219362967e-06, "loss": 0.8652, "step": 193 }, { "epoch": 0.07, "learning_rate": 9.871333453302142e-06, "loss": 0.8587, "step": 194 }, { "epoch": 0.07, "learning_rate": 9.870433687241318e-06, "loss": 0.9339, "step": 195 }, { "epoch": 0.07, "learning_rate": 9.869533921180495e-06, "loss": 0.9443, "step": 196 }, { "epoch": 0.07, "learning_rate": 9.86863415511967e-06, "loss": 0.8807, "step": 197 }, { "epoch": 0.07, "learning_rate": 9.867734389058845e-06, "loss": 0.928, "step": 198 }, { "epoch": 0.07, "learning_rate": 9.866834622998021e-06, "loss": 0.9455, "step": 199 }, { "epoch": 0.07, "learning_rate": 9.865934856937196e-06, "loss": 0.9789, "step": 200 }, { "epoch": 0.07, "learning_rate": 9.865035090876373e-06, "loss": 0.927, "step": 201 }, { "epoch": 0.07, "learning_rate": 9.86413532481555e-06, "loss": 0.8587, "step": 202 }, { "epoch": 0.07, "learning_rate": 9.863235558754724e-06, "loss": 0.9724, "step": 203 }, { "epoch": 0.07, "learning_rate": 9.8623357926939e-06, "loss": 0.8146, "step": 204 }, { "epoch": 0.07, "learning_rate": 9.861436026633076e-06, "loss": 0.8024, "step": 205 }, { "epoch": 0.07, "learning_rate": 9.860536260572253e-06, "loss": 0.9887, "step": 206 }, { "epoch": 0.07, "learning_rate": 9.859636494511428e-06, "loss": 0.8701, "step": 207 }, { "epoch": 0.07, "learning_rate": 9.858736728450604e-06, "loss": 0.8591, "step": 208 }, { "epoch": 0.08, "learning_rate": 9.857836962389779e-06, "loss": 0.9454, "step": 209 }, { "epoch": 0.08, "learning_rate": 9.856937196328956e-06, "loss": 0.9379, "step": 210 }, { "epoch": 0.08, "learning_rate": 9.856037430268132e-06, "loss": 0.9173, "step": 211 }, { "epoch": 0.08, "learning_rate": 9.855137664207307e-06, "loss": 0.8715, "step": 212 }, { "epoch": 0.08, "learning_rate": 9.854237898146482e-06, "loss": 0.9577, "step": 213 }, { "epoch": 0.08, "learning_rate": 9.853338132085659e-06, "loss": 0.8376, "step": 214 }, { "epoch": 0.08, "learning_rate": 9.852438366024835e-06, "loss": 0.9789, "step": 215 }, { "epoch": 0.08, "learning_rate": 9.85153859996401e-06, "loss": 0.7204, "step": 216 }, { "epoch": 0.08, "learning_rate": 9.850638833903185e-06, "loss": 0.9708, "step": 217 }, { "epoch": 0.08, "learning_rate": 9.849739067842362e-06, "loss": 0.8805, "step": 218 }, { "epoch": 0.08, "learning_rate": 9.848839301781537e-06, "loss": 0.903, "step": 219 }, { "epoch": 0.08, "learning_rate": 9.847939535720713e-06, "loss": 0.846, "step": 220 }, { "epoch": 0.08, "learning_rate": 9.84703976965989e-06, "loss": 0.9666, "step": 221 }, { "epoch": 0.08, "learning_rate": 9.846140003599065e-06, "loss": 0.8346, "step": 222 }, { "epoch": 0.08, "learning_rate": 9.84524023753824e-06, "loss": 0.7454, "step": 223 }, { "epoch": 0.08, "learning_rate": 9.844340471477416e-06, "loss": 0.7904, "step": 224 }, { "epoch": 0.08, "learning_rate": 9.843440705416593e-06, "loss": 0.8578, "step": 225 }, { "epoch": 0.08, "learning_rate": 9.842540939355768e-06, "loss": 0.9147, "step": 226 }, { "epoch": 0.08, "learning_rate": 9.841641173294945e-06, "loss": 0.913, "step": 227 }, { "epoch": 0.08, "learning_rate": 9.84074140723412e-06, "loss": 0.9032, "step": 228 }, { "epoch": 0.08, "learning_rate": 9.839841641173296e-06, "loss": 0.9186, "step": 229 }, { "epoch": 0.08, "learning_rate": 9.838941875112471e-06, "loss": 0.9279, "step": 230 }, { "epoch": 0.08, "learning_rate": 9.838042109051648e-06, "loss": 0.8786, "step": 231 }, { "epoch": 0.08, "learning_rate": 9.837142342990823e-06, "loss": 0.9432, "step": 232 }, { "epoch": 0.08, "learning_rate": 9.83624257693e-06, "loss": 0.9695, "step": 233 }, { "epoch": 0.08, "learning_rate": 9.835342810869176e-06, "loss": 0.915, "step": 234 }, { "epoch": 0.08, "learning_rate": 9.83444304480835e-06, "loss": 0.9561, "step": 235 }, { "epoch": 0.08, "learning_rate": 9.833543278747526e-06, "loss": 0.9211, "step": 236 }, { "epoch": 0.09, "learning_rate": 9.832643512686702e-06, "loss": 0.8611, "step": 237 }, { "epoch": 0.09, "learning_rate": 9.831743746625877e-06, "loss": 0.9464, "step": 238 }, { "epoch": 0.09, "learning_rate": 9.830843980565054e-06, "loss": 0.8101, "step": 239 }, { "epoch": 0.09, "learning_rate": 9.82994421450423e-06, "loss": 0.8788, "step": 240 }, { "epoch": 0.09, "learning_rate": 9.829044448443405e-06, "loss": 0.7038, "step": 241 }, { "epoch": 0.09, "learning_rate": 9.82814468238258e-06, "loss": 0.8038, "step": 242 }, { "epoch": 0.09, "learning_rate": 9.827244916321757e-06, "loss": 0.8928, "step": 243 }, { "epoch": 0.09, "learning_rate": 9.826345150260934e-06, "loss": 0.8892, "step": 244 }, { "epoch": 0.09, "learning_rate": 9.825445384200108e-06, "loss": 0.7961, "step": 245 }, { "epoch": 0.09, "learning_rate": 9.824545618139285e-06, "loss": 0.9315, "step": 246 }, { "epoch": 0.09, "learning_rate": 9.82364585207846e-06, "loss": 0.9186, "step": 247 }, { "epoch": 0.09, "learning_rate": 9.822746086017637e-06, "loss": 0.8554, "step": 248 }, { "epoch": 0.09, "learning_rate": 9.821846319956812e-06, "loss": 0.9657, "step": 249 }, { "epoch": 0.09, "learning_rate": 9.820946553895988e-06, "loss": 1.0353, "step": 250 }, { "epoch": 0.09, "learning_rate": 9.820046787835163e-06, "loss": 0.8966, "step": 251 }, { "epoch": 0.09, "learning_rate": 9.81914702177434e-06, "loss": 0.9434, "step": 252 }, { "epoch": 0.09, "learning_rate": 9.818247255713516e-06, "loss": 0.9313, "step": 253 }, { "epoch": 0.09, "learning_rate": 9.817347489652691e-06, "loss": 0.9076, "step": 254 }, { "epoch": 0.09, "learning_rate": 9.816447723591866e-06, "loss": 0.8613, "step": 255 }, { "epoch": 0.09, "learning_rate": 9.815547957531043e-06, "loss": 1.0325, "step": 256 }, { "epoch": 0.09, "learning_rate": 9.81464819147022e-06, "loss": 0.8823, "step": 257 }, { "epoch": 0.09, "learning_rate": 9.813748425409394e-06, "loss": 0.9215, "step": 258 }, { "epoch": 0.09, "learning_rate": 9.812848659348571e-06, "loss": 0.7986, "step": 259 }, { "epoch": 0.09, "learning_rate": 9.811948893287746e-06, "loss": 0.8571, "step": 260 }, { "epoch": 0.09, "learning_rate": 9.81104912722692e-06, "loss": 0.8555, "step": 261 }, { "epoch": 0.09, "learning_rate": 9.810149361166097e-06, "loss": 0.8803, "step": 262 }, { "epoch": 0.09, "learning_rate": 9.809249595105274e-06, "loss": 0.7793, "step": 263 }, { "epoch": 0.09, "learning_rate": 9.808349829044449e-06, "loss": 0.9324, "step": 264 }, { "epoch": 0.1, "learning_rate": 9.807450062983626e-06, "loss": 0.9411, "step": 265 }, { "epoch": 0.1, "learning_rate": 9.8065502969228e-06, "loss": 0.8483, "step": 266 }, { "epoch": 0.1, "learning_rate": 9.805650530861977e-06, "loss": 0.8264, "step": 267 }, { "epoch": 0.1, "learning_rate": 9.804750764801152e-06, "loss": 1.0402, "step": 268 }, { "epoch": 0.1, "learning_rate": 9.803850998740329e-06, "loss": 0.9411, "step": 269 }, { "epoch": 0.1, "learning_rate": 9.802951232679504e-06, "loss": 1.0282, "step": 270 }, { "epoch": 0.1, "learning_rate": 9.80205146661868e-06, "loss": 0.8995, "step": 271 }, { "epoch": 0.1, "learning_rate": 9.801151700557857e-06, "loss": 0.9482, "step": 272 }, { "epoch": 0.1, "learning_rate": 9.800251934497032e-06, "loss": 0.8936, "step": 273 }, { "epoch": 0.1, "learning_rate": 9.799352168436207e-06, "loss": 0.866, "step": 274 }, { "epoch": 0.1, "learning_rate": 9.798452402375383e-06, "loss": 0.8328, "step": 275 }, { "epoch": 0.1, "learning_rate": 9.79755263631456e-06, "loss": 0.8124, "step": 276 }, { "epoch": 0.1, "learning_rate": 9.796652870253735e-06, "loss": 0.8745, "step": 277 }, { "epoch": 0.1, "learning_rate": 9.795753104192911e-06, "loss": 0.8807, "step": 278 }, { "epoch": 0.1, "learning_rate": 9.794853338132086e-06, "loss": 0.8928, "step": 279 }, { "epoch": 0.1, "learning_rate": 9.793953572071261e-06, "loss": 0.9627, "step": 280 }, { "epoch": 0.1, "learning_rate": 9.793053806010438e-06, "loss": 0.9351, "step": 281 }, { "epoch": 0.1, "learning_rate": 9.792154039949614e-06, "loss": 0.802, "step": 282 }, { "epoch": 0.1, "learning_rate": 9.79125427388879e-06, "loss": 0.8741, "step": 283 }, { "epoch": 0.1, "learning_rate": 9.790354507827966e-06, "loss": 0.8944, "step": 284 }, { "epoch": 0.1, "learning_rate": 9.789454741767141e-06, "loss": 0.9611, "step": 285 }, { "epoch": 0.1, "learning_rate": 9.788554975706317e-06, "loss": 0.8238, "step": 286 }, { "epoch": 0.1, "learning_rate": 9.787655209645492e-06, "loss": 0.778, "step": 287 }, { "epoch": 0.1, "learning_rate": 9.786755443584669e-06, "loss": 0.9175, "step": 288 }, { "epoch": 0.1, "learning_rate": 9.785855677523844e-06, "loss": 0.8534, "step": 289 }, { "epoch": 0.1, "learning_rate": 9.78495591146302e-06, "loss": 0.8472, "step": 290 }, { "epoch": 0.1, "learning_rate": 9.784056145402197e-06, "loss": 1.0503, "step": 291 }, { "epoch": 0.1, "learning_rate": 9.783156379341372e-06, "loss": 1.0033, "step": 292 }, { "epoch": 0.11, "learning_rate": 9.782256613280547e-06, "loss": 0.8719, "step": 293 }, { "epoch": 0.11, "learning_rate": 9.781356847219724e-06, "loss": 0.9096, "step": 294 }, { "epoch": 0.11, "learning_rate": 9.7804570811589e-06, "loss": 0.9566, "step": 295 }, { "epoch": 0.11, "learning_rate": 9.779557315098075e-06, "loss": 0.9584, "step": 296 }, { "epoch": 0.11, "learning_rate": 9.778657549037252e-06, "loss": 0.8764, "step": 297 }, { "epoch": 0.11, "learning_rate": 9.777757782976427e-06, "loss": 0.8452, "step": 298 }, { "epoch": 0.11, "learning_rate": 9.776858016915602e-06, "loss": 0.9519, "step": 299 }, { "epoch": 0.11, "learning_rate": 9.775958250854778e-06, "loss": 0.7644, "step": 300 }, { "epoch": 0.11, "learning_rate": 9.775058484793955e-06, "loss": 0.8021, "step": 301 }, { "epoch": 0.11, "learning_rate": 9.77415871873313e-06, "loss": 0.8153, "step": 302 }, { "epoch": 0.11, "learning_rate": 9.773258952672306e-06, "loss": 0.7978, "step": 303 }, { "epoch": 0.11, "learning_rate": 9.772359186611481e-06, "loss": 0.806, "step": 304 }, { "epoch": 0.11, "learning_rate": 9.771459420550658e-06, "loss": 0.8473, "step": 305 }, { "epoch": 0.11, "learning_rate": 9.770559654489833e-06, "loss": 0.8817, "step": 306 }, { "epoch": 0.11, "learning_rate": 9.76965988842901e-06, "loss": 0.8614, "step": 307 }, { "epoch": 0.11, "learning_rate": 9.768760122368184e-06, "loss": 0.974, "step": 308 }, { "epoch": 0.11, "learning_rate": 9.767860356307361e-06, "loss": 0.7739, "step": 309 }, { "epoch": 0.11, "learning_rate": 9.766960590246538e-06, "loss": 0.7661, "step": 310 }, { "epoch": 0.11, "learning_rate": 9.766060824185713e-06, "loss": 0.9215, "step": 311 }, { "epoch": 0.11, "learning_rate": 9.765161058124887e-06, "loss": 0.9185, "step": 312 }, { "epoch": 0.11, "learning_rate": 9.764261292064064e-06, "loss": 1.0879, "step": 313 }, { "epoch": 0.11, "learning_rate": 9.76336152600324e-06, "loss": 0.9046, "step": 314 }, { "epoch": 0.11, "learning_rate": 9.762461759942416e-06, "loss": 0.9784, "step": 315 }, { "epoch": 0.11, "learning_rate": 9.761561993881592e-06, "loss": 0.8655, "step": 316 }, { "epoch": 0.11, "learning_rate": 9.760662227820767e-06, "loss": 0.819, "step": 317 }, { "epoch": 0.11, "learning_rate": 9.759762461759942e-06, "loss": 0.9341, "step": 318 }, { "epoch": 0.11, "learning_rate": 9.758862695699119e-06, "loss": 0.9188, "step": 319 }, { "epoch": 0.11, "learning_rate": 9.757962929638295e-06, "loss": 0.9513, "step": 320 }, { "epoch": 0.12, "learning_rate": 9.75706316357747e-06, "loss": 0.9451, "step": 321 }, { "epoch": 0.12, "learning_rate": 9.756163397516647e-06, "loss": 1.0171, "step": 322 }, { "epoch": 0.12, "learning_rate": 9.755263631455822e-06, "loss": 0.8934, "step": 323 }, { "epoch": 0.12, "learning_rate": 9.754363865394998e-06, "loss": 0.8811, "step": 324 }, { "epoch": 0.12, "learning_rate": 9.753464099334173e-06, "loss": 0.9598, "step": 325 }, { "epoch": 0.12, "learning_rate": 9.75256433327335e-06, "loss": 0.7458, "step": 326 }, { "epoch": 0.12, "learning_rate": 9.751664567212525e-06, "loss": 0.837, "step": 327 }, { "epoch": 0.12, "learning_rate": 9.750764801151701e-06, "loss": 0.8957, "step": 328 }, { "epoch": 0.12, "learning_rate": 9.749865035090878e-06, "loss": 0.9619, "step": 329 }, { "epoch": 0.12, "learning_rate": 9.748965269030053e-06, "loss": 0.9955, "step": 330 }, { "epoch": 0.12, "learning_rate": 9.748065502969228e-06, "loss": 0.8404, "step": 331 }, { "epoch": 0.12, "learning_rate": 9.747165736908405e-06, "loss": 0.9288, "step": 332 }, { "epoch": 0.12, "learning_rate": 9.746265970847581e-06, "loss": 0.8237, "step": 333 }, { "epoch": 0.12, "learning_rate": 9.745366204786756e-06, "loss": 0.8819, "step": 334 }, { "epoch": 0.12, "learning_rate": 9.744466438725933e-06, "loss": 0.9771, "step": 335 }, { "epoch": 0.12, "learning_rate": 9.743566672665108e-06, "loss": 0.8298, "step": 336 }, { "epoch": 0.12, "learning_rate": 9.742666906604283e-06, "loss": 0.8761, "step": 337 }, { "epoch": 0.12, "learning_rate": 9.74176714054346e-06, "loss": 1.007, "step": 338 }, { "epoch": 0.12, "learning_rate": 9.740867374482636e-06, "loss": 0.9351, "step": 339 }, { "epoch": 0.12, "learning_rate": 9.73996760842181e-06, "loss": 0.7861, "step": 340 }, { "epoch": 0.12, "learning_rate": 9.739067842360987e-06, "loss": 0.9545, "step": 341 }, { "epoch": 0.12, "learning_rate": 9.738168076300162e-06, "loss": 1.002, "step": 342 }, { "epoch": 0.12, "learning_rate": 9.737268310239339e-06, "loss": 1.0108, "step": 343 }, { "epoch": 0.12, "learning_rate": 9.736368544178514e-06, "loss": 0.9753, "step": 344 }, { "epoch": 0.12, "learning_rate": 9.73546877811769e-06, "loss": 0.9516, "step": 345 }, { "epoch": 0.12, "learning_rate": 9.734569012056865e-06, "loss": 0.9165, "step": 346 }, { "epoch": 0.12, "learning_rate": 9.733669245996042e-06, "loss": 0.9975, "step": 347 }, { "epoch": 0.12, "learning_rate": 9.732769479935219e-06, "loss": 0.8383, "step": 348 }, { "epoch": 0.13, "learning_rate": 9.731869713874393e-06, "loss": 0.9784, "step": 349 }, { "epoch": 0.13, "learning_rate": 9.730969947813568e-06, "loss": 0.9126, "step": 350 }, { "epoch": 0.13, "learning_rate": 9.730070181752745e-06, "loss": 0.8348, "step": 351 }, { "epoch": 0.13, "learning_rate": 9.729170415691922e-06, "loss": 0.8428, "step": 352 }, { "epoch": 0.13, "learning_rate": 9.728270649631097e-06, "loss": 0.9504, "step": 353 }, { "epoch": 0.13, "learning_rate": 9.727370883570273e-06, "loss": 0.7588, "step": 354 }, { "epoch": 0.13, "learning_rate": 9.726471117509448e-06, "loss": 0.8522, "step": 355 }, { "epoch": 0.13, "learning_rate": 9.725571351448625e-06, "loss": 0.8527, "step": 356 }, { "epoch": 0.13, "learning_rate": 9.7246715853878e-06, "loss": 0.9724, "step": 357 }, { "epoch": 0.13, "learning_rate": 9.723771819326976e-06, "loss": 0.8749, "step": 358 }, { "epoch": 0.13, "learning_rate": 9.722872053266151e-06, "loss": 0.9618, "step": 359 }, { "epoch": 0.13, "learning_rate": 9.721972287205328e-06, "loss": 0.8861, "step": 360 }, { "epoch": 0.13, "learning_rate": 9.721072521144503e-06, "loss": 0.811, "step": 361 }, { "epoch": 0.13, "learning_rate": 9.72017275508368e-06, "loss": 0.9483, "step": 362 }, { "epoch": 0.13, "learning_rate": 9.719272989022854e-06, "loss": 0.8663, "step": 363 }, { "epoch": 0.13, "learning_rate": 9.718373222962031e-06, "loss": 0.9028, "step": 364 }, { "epoch": 0.13, "learning_rate": 9.717473456901206e-06, "loss": 0.8867, "step": 365 }, { "epoch": 0.13, "learning_rate": 9.716573690840382e-06, "loss": 1.0768, "step": 366 }, { "epoch": 0.13, "learning_rate": 9.715673924779559e-06, "loss": 0.7092, "step": 367 }, { "epoch": 0.13, "learning_rate": 9.714774158718734e-06, "loss": 0.8399, "step": 368 }, { "epoch": 0.13, "learning_rate": 9.713874392657909e-06, "loss": 0.8894, "step": 369 }, { "epoch": 0.13, "learning_rate": 9.712974626597085e-06, "loss": 0.9957, "step": 370 }, { "epoch": 0.13, "learning_rate": 9.712074860536262e-06, "loss": 0.9631, "step": 371 }, { "epoch": 0.13, "learning_rate": 9.711175094475437e-06, "loss": 0.8457, "step": 372 }, { "epoch": 0.13, "learning_rate": 9.710275328414614e-06, "loss": 0.975, "step": 373 }, { "epoch": 0.13, "learning_rate": 9.709375562353789e-06, "loss": 0.9597, "step": 374 }, { "epoch": 0.13, "learning_rate": 9.708475796292965e-06, "loss": 0.8591, "step": 375 }, { "epoch": 0.13, "learning_rate": 9.70757603023214e-06, "loss": 0.9541, "step": 376 }, { "epoch": 0.14, "learning_rate": 9.706676264171317e-06, "loss": 0.8366, "step": 377 }, { "epoch": 0.14, "learning_rate": 9.705776498110492e-06, "loss": 0.9975, "step": 378 }, { "epoch": 0.14, "learning_rate": 9.704876732049668e-06, "loss": 0.9332, "step": 379 }, { "epoch": 0.14, "learning_rate": 9.703976965988843e-06, "loss": 0.896, "step": 380 }, { "epoch": 0.14, "learning_rate": 9.70307719992802e-06, "loss": 0.9166, "step": 381 }, { "epoch": 0.14, "learning_rate": 9.702177433867195e-06, "loss": 0.892, "step": 382 }, { "epoch": 0.14, "learning_rate": 9.701277667806371e-06, "loss": 0.9952, "step": 383 }, { "epoch": 0.14, "learning_rate": 9.700377901745546e-06, "loss": 0.8472, "step": 384 }, { "epoch": 0.14, "learning_rate": 9.699478135684723e-06, "loss": 1.054, "step": 385 }, { "epoch": 0.14, "learning_rate": 9.6985783696239e-06, "loss": 0.8834, "step": 386 }, { "epoch": 0.14, "learning_rate": 9.697678603563074e-06, "loss": 0.8845, "step": 387 }, { "epoch": 0.14, "learning_rate": 9.69677883750225e-06, "loss": 0.8992, "step": 388 }, { "epoch": 0.14, "learning_rate": 9.695879071441426e-06, "loss": 0.9172, "step": 389 }, { "epoch": 0.14, "learning_rate": 9.694979305380603e-06, "loss": 1.0601, "step": 390 }, { "epoch": 0.14, "learning_rate": 9.694079539319777e-06, "loss": 0.9695, "step": 391 }, { "epoch": 0.14, "learning_rate": 9.693179773258954e-06, "loss": 1.0169, "step": 392 }, { "epoch": 0.14, "learning_rate": 9.692280007198129e-06, "loss": 0.8305, "step": 393 }, { "epoch": 0.14, "learning_rate": 9.691380241137306e-06, "loss": 0.8298, "step": 394 }, { "epoch": 0.14, "learning_rate": 9.69048047507648e-06, "loss": 0.8805, "step": 395 }, { "epoch": 0.14, "learning_rate": 9.689580709015657e-06, "loss": 0.9155, "step": 396 }, { "epoch": 0.14, "learning_rate": 9.688680942954832e-06, "loss": 1.0281, "step": 397 }, { "epoch": 0.14, "learning_rate": 9.687781176894009e-06, "loss": 0.9925, "step": 398 }, { "epoch": 0.14, "learning_rate": 9.686881410833184e-06, "loss": 0.8893, "step": 399 }, { "epoch": 0.14, "learning_rate": 9.68598164477236e-06, "loss": 0.8898, "step": 400 }, { "epoch": 0.14, "learning_rate": 9.685081878711535e-06, "loss": 0.8527, "step": 401 }, { "epoch": 0.14, "learning_rate": 9.684182112650712e-06, "loss": 0.9386, "step": 402 }, { "epoch": 0.14, "learning_rate": 9.683282346589887e-06, "loss": 0.8276, "step": 403 }, { "epoch": 0.15, "learning_rate": 9.682382580529063e-06, "loss": 0.9066, "step": 404 }, { "epoch": 0.15, "learning_rate": 9.68148281446824e-06, "loss": 0.9036, "step": 405 }, { "epoch": 0.15, "learning_rate": 9.680583048407415e-06, "loss": 0.8525, "step": 406 }, { "epoch": 0.15, "learning_rate": 9.67968328234659e-06, "loss": 0.807, "step": 407 }, { "epoch": 0.15, "learning_rate": 9.678783516285766e-06, "loss": 0.956, "step": 408 }, { "epoch": 0.15, "learning_rate": 9.677883750224943e-06, "loss": 0.9574, "step": 409 }, { "epoch": 0.15, "learning_rate": 9.676983984164118e-06, "loss": 0.9453, "step": 410 }, { "epoch": 0.15, "learning_rate": 9.676084218103295e-06, "loss": 0.8109, "step": 411 }, { "epoch": 0.15, "learning_rate": 9.67518445204247e-06, "loss": 0.9014, "step": 412 }, { "epoch": 0.15, "learning_rate": 9.674284685981646e-06, "loss": 0.9307, "step": 413 }, { "epoch": 0.15, "learning_rate": 9.673384919920823e-06, "loss": 1.0123, "step": 414 }, { "epoch": 0.15, "learning_rate": 9.672485153859998e-06, "loss": 0.9389, "step": 415 }, { "epoch": 0.15, "learning_rate": 9.671585387799173e-06, "loss": 0.9559, "step": 416 }, { "epoch": 0.15, "learning_rate": 9.670685621738349e-06, "loss": 0.968, "step": 417 }, { "epoch": 0.15, "learning_rate": 9.669785855677524e-06, "loss": 0.9739, "step": 418 }, { "epoch": 0.15, "learning_rate": 9.6688860896167e-06, "loss": 0.9433, "step": 419 }, { "epoch": 0.15, "learning_rate": 9.667986323555876e-06, "loss": 1.0463, "step": 420 }, { "epoch": 0.15, "learning_rate": 9.667086557495052e-06, "loss": 0.7731, "step": 421 }, { "epoch": 0.15, "learning_rate": 9.666186791434227e-06, "loss": 0.7974, "step": 422 }, { "epoch": 0.15, "learning_rate": 9.665287025373404e-06, "loss": 0.8897, "step": 423 }, { "epoch": 0.15, "learning_rate": 9.66438725931258e-06, "loss": 0.9365, "step": 424 }, { "epoch": 0.15, "learning_rate": 9.663487493251755e-06, "loss": 0.856, "step": 425 }, { "epoch": 0.15, "learning_rate": 9.66258772719093e-06, "loss": 0.9545, "step": 426 }, { "epoch": 0.15, "learning_rate": 9.661687961130107e-06, "loss": 0.8327, "step": 427 }, { "epoch": 0.15, "learning_rate": 9.660788195069283e-06, "loss": 0.9594, "step": 428 }, { "epoch": 0.15, "learning_rate": 9.659888429008458e-06, "loss": 0.8262, "step": 429 }, { "epoch": 0.15, "learning_rate": 9.658988662947635e-06, "loss": 0.936, "step": 430 }, { "epoch": 0.15, "learning_rate": 9.65808889688681e-06, "loss": 0.8522, "step": 431 }, { "epoch": 0.16, "learning_rate": 9.657189130825987e-06, "loss": 1.0147, "step": 432 }, { "epoch": 0.16, "learning_rate": 9.656289364765163e-06, "loss": 0.8063, "step": 433 }, { "epoch": 0.16, "learning_rate": 9.655389598704338e-06, "loss": 0.9163, "step": 434 }, { "epoch": 0.16, "learning_rate": 9.654489832643513e-06, "loss": 0.8547, "step": 435 }, { "epoch": 0.16, "learning_rate": 9.65359006658269e-06, "loss": 1.0187, "step": 436 }, { "epoch": 0.16, "learning_rate": 9.652690300521865e-06, "loss": 0.8631, "step": 437 }, { "epoch": 0.16, "learning_rate": 9.651790534461041e-06, "loss": 0.894, "step": 438 }, { "epoch": 0.16, "learning_rate": 9.650890768400216e-06, "loss": 1.0042, "step": 439 }, { "epoch": 0.16, "learning_rate": 9.649991002339393e-06, "loss": 0.9059, "step": 440 }, { "epoch": 0.16, "learning_rate": 9.649091236278568e-06, "loss": 0.7164, "step": 441 }, { "epoch": 0.16, "learning_rate": 9.648191470217744e-06, "loss": 0.9543, "step": 442 }, { "epoch": 0.16, "learning_rate": 9.64729170415692e-06, "loss": 0.7908, "step": 443 }, { "epoch": 0.16, "learning_rate": 9.646391938096096e-06, "loss": 0.9138, "step": 444 }, { "epoch": 0.16, "learning_rate": 9.64549217203527e-06, "loss": 0.8206, "step": 445 }, { "epoch": 0.16, "learning_rate": 9.644592405974447e-06, "loss": 0.9728, "step": 446 }, { "epoch": 0.16, "learning_rate": 9.643692639913624e-06, "loss": 0.9074, "step": 447 }, { "epoch": 0.16, "learning_rate": 9.642792873852799e-06, "loss": 0.947, "step": 448 }, { "epoch": 0.16, "learning_rate": 9.641893107791975e-06, "loss": 0.8909, "step": 449 }, { "epoch": 0.16, "learning_rate": 9.64099334173115e-06, "loss": 0.7314, "step": 450 }, { "epoch": 0.16, "learning_rate": 9.640093575670327e-06, "loss": 1.0409, "step": 451 }, { "epoch": 0.16, "learning_rate": 9.639193809609504e-06, "loss": 0.9954, "step": 452 }, { "epoch": 0.16, "learning_rate": 9.638294043548679e-06, "loss": 0.8145, "step": 453 }, { "epoch": 0.16, "learning_rate": 9.637394277487853e-06, "loss": 0.8812, "step": 454 }, { "epoch": 0.16, "learning_rate": 9.63649451142703e-06, "loss": 0.8953, "step": 455 }, { "epoch": 0.16, "learning_rate": 9.635594745366205e-06, "loss": 0.8838, "step": 456 }, { "epoch": 0.16, "learning_rate": 9.634694979305382e-06, "loss": 0.8676, "step": 457 }, { "epoch": 0.16, "learning_rate": 9.633795213244557e-06, "loss": 0.9584, "step": 458 }, { "epoch": 0.16, "learning_rate": 9.632895447183733e-06, "loss": 0.9111, "step": 459 }, { "epoch": 0.17, "learning_rate": 9.631995681122908e-06, "loss": 0.9394, "step": 460 }, { "epoch": 0.17, "learning_rate": 9.631095915062085e-06, "loss": 1.0217, "step": 461 }, { "epoch": 0.17, "learning_rate": 9.630196149001261e-06, "loss": 0.8804, "step": 462 }, { "epoch": 0.17, "learning_rate": 9.629296382940436e-06, "loss": 0.8964, "step": 463 }, { "epoch": 0.17, "learning_rate": 9.628396616879611e-06, "loss": 0.9059, "step": 464 }, { "epoch": 0.17, "learning_rate": 9.627496850818788e-06, "loss": 0.7636, "step": 465 }, { "epoch": 0.17, "learning_rate": 9.626597084757964e-06, "loss": 0.971, "step": 466 }, { "epoch": 0.17, "learning_rate": 9.62569731869714e-06, "loss": 0.8102, "step": 467 }, { "epoch": 0.17, "learning_rate": 9.624797552636316e-06, "loss": 0.9978, "step": 468 }, { "epoch": 0.17, "learning_rate": 9.62389778657549e-06, "loss": 0.8738, "step": 469 }, { "epoch": 0.17, "learning_rate": 9.622998020514667e-06, "loss": 0.9076, "step": 470 }, { "epoch": 0.17, "learning_rate": 9.622098254453844e-06, "loss": 1.0225, "step": 471 }, { "epoch": 0.17, "learning_rate": 9.621198488393019e-06, "loss": 0.8338, "step": 472 }, { "epoch": 0.17, "learning_rate": 9.620298722332194e-06, "loss": 0.8879, "step": 473 }, { "epoch": 0.17, "learning_rate": 9.61939895627137e-06, "loss": 0.8512, "step": 474 }, { "epoch": 0.17, "learning_rate": 9.618499190210545e-06, "loss": 0.8969, "step": 475 }, { "epoch": 0.17, "learning_rate": 9.617599424149722e-06, "loss": 0.8804, "step": 476 }, { "epoch": 0.17, "learning_rate": 9.616699658088897e-06, "loss": 0.9076, "step": 477 }, { "epoch": 0.17, "learning_rate": 9.615799892028074e-06, "loss": 0.8989, "step": 478 }, { "epoch": 0.17, "learning_rate": 9.614900125967248e-06, "loss": 0.9342, "step": 479 }, { "epoch": 0.17, "learning_rate": 9.614000359906425e-06, "loss": 0.9495, "step": 480 }, { "epoch": 0.17, "learning_rate": 9.613100593845602e-06, "loss": 0.8741, "step": 481 }, { "epoch": 0.17, "learning_rate": 9.612200827784777e-06, "loss": 0.8178, "step": 482 }, { "epoch": 0.17, "learning_rate": 9.611301061723952e-06, "loss": 1.0048, "step": 483 }, { "epoch": 0.17, "learning_rate": 9.610401295663128e-06, "loss": 0.883, "step": 484 }, { "epoch": 0.17, "learning_rate": 9.609501529602305e-06, "loss": 0.7528, "step": 485 }, { "epoch": 0.17, "learning_rate": 9.60860176354148e-06, "loss": 0.8496, "step": 486 }, { "epoch": 0.17, "learning_rate": 9.607701997480656e-06, "loss": 0.8533, "step": 487 }, { "epoch": 0.18, "learning_rate": 9.606802231419831e-06, "loss": 0.9628, "step": 488 }, { "epoch": 0.18, "learning_rate": 9.605902465359008e-06, "loss": 0.8786, "step": 489 }, { "epoch": 0.18, "learning_rate": 9.605002699298184e-06, "loss": 0.7681, "step": 490 }, { "epoch": 0.18, "learning_rate": 9.60410293323736e-06, "loss": 0.9972, "step": 491 }, { "epoch": 0.18, "learning_rate": 9.603203167176534e-06, "loss": 0.7387, "step": 492 }, { "epoch": 0.18, "learning_rate": 9.602303401115711e-06, "loss": 0.8644, "step": 493 }, { "epoch": 0.18, "learning_rate": 9.601403635054886e-06, "loss": 0.9138, "step": 494 }, { "epoch": 0.18, "learning_rate": 9.600503868994062e-06, "loss": 0.8121, "step": 495 }, { "epoch": 0.18, "learning_rate": 9.599604102933237e-06, "loss": 0.8467, "step": 496 }, { "epoch": 0.18, "learning_rate": 9.598704336872414e-06, "loss": 0.9136, "step": 497 }, { "epoch": 0.18, "learning_rate": 9.597804570811589e-06, "loss": 0.8354, "step": 498 }, { "epoch": 0.18, "learning_rate": 9.596904804750766e-06, "loss": 0.8907, "step": 499 }, { "epoch": 0.18, "learning_rate": 9.596005038689942e-06, "loss": 0.7903, "step": 500 }, { "epoch": 0.18, "learning_rate": 9.595105272629117e-06, "loss": 0.9093, "step": 501 }, { "epoch": 0.18, "learning_rate": 9.594205506568292e-06, "loss": 0.9033, "step": 502 }, { "epoch": 0.18, "learning_rate": 9.593305740507469e-06, "loss": 0.9458, "step": 503 }, { "epoch": 0.18, "learning_rate": 9.592405974446645e-06, "loss": 0.8594, "step": 504 }, { "epoch": 0.18, "learning_rate": 9.59150620838582e-06, "loss": 0.9443, "step": 505 }, { "epoch": 0.18, "learning_rate": 9.590606442324997e-06, "loss": 0.767, "step": 506 }, { "epoch": 0.18, "learning_rate": 9.589706676264172e-06, "loss": 0.9596, "step": 507 }, { "epoch": 0.18, "learning_rate": 9.588806910203348e-06, "loss": 0.8615, "step": 508 }, { "epoch": 0.18, "learning_rate": 9.587907144142525e-06, "loss": 0.9273, "step": 509 }, { "epoch": 0.18, "learning_rate": 9.5870073780817e-06, "loss": 0.9231, "step": 510 }, { "epoch": 0.18, "learning_rate": 9.586107612020875e-06, "loss": 1.0446, "step": 511 }, { "epoch": 0.18, "learning_rate": 9.585207845960051e-06, "loss": 0.8992, "step": 512 }, { "epoch": 0.18, "learning_rate": 9.584308079899228e-06, "loss": 0.8417, "step": 513 }, { "epoch": 0.18, "learning_rate": 9.583408313838403e-06, "loss": 0.8967, "step": 514 }, { "epoch": 0.18, "learning_rate": 9.582508547777578e-06, "loss": 0.9174, "step": 515 }, { "epoch": 0.19, "learning_rate": 9.581608781716754e-06, "loss": 0.8203, "step": 516 }, { "epoch": 0.19, "learning_rate": 9.58070901565593e-06, "loss": 0.9999, "step": 517 }, { "epoch": 0.19, "learning_rate": 9.579809249595106e-06, "loss": 0.9436, "step": 518 }, { "epoch": 0.19, "learning_rate": 9.578909483534283e-06, "loss": 0.8768, "step": 519 }, { "epoch": 0.19, "learning_rate": 9.578009717473458e-06, "loss": 0.9074, "step": 520 }, { "epoch": 0.19, "learning_rate": 9.577109951412632e-06, "loss": 0.8962, "step": 521 }, { "epoch": 0.19, "learning_rate": 9.576210185351809e-06, "loss": 0.9208, "step": 522 }, { "epoch": 0.19, "learning_rate": 9.575310419290986e-06, "loss": 0.878, "step": 523 }, { "epoch": 0.19, "learning_rate": 9.57441065323016e-06, "loss": 0.8965, "step": 524 }, { "epoch": 0.19, "learning_rate": 9.573510887169337e-06, "loss": 0.8584, "step": 525 }, { "epoch": 0.19, "learning_rate": 9.572611121108512e-06, "loss": 0.8561, "step": 526 }, { "epoch": 0.19, "learning_rate": 9.571711355047689e-06, "loss": 0.8266, "step": 527 }, { "epoch": 0.19, "learning_rate": 9.570811588986865e-06, "loss": 0.8745, "step": 528 }, { "epoch": 0.19, "learning_rate": 9.56991182292604e-06, "loss": 0.9711, "step": 529 }, { "epoch": 0.19, "learning_rate": 9.569012056865215e-06, "loss": 0.8589, "step": 530 }, { "epoch": 0.19, "learning_rate": 9.568112290804392e-06, "loss": 0.812, "step": 531 }, { "epoch": 0.19, "learning_rate": 9.567212524743568e-06, "loss": 0.8115, "step": 532 }, { "epoch": 0.19, "learning_rate": 9.566312758682743e-06, "loss": 0.9389, "step": 533 }, { "epoch": 0.19, "learning_rate": 9.565412992621918e-06, "loss": 1.038, "step": 534 }, { "epoch": 0.19, "learning_rate": 9.564513226561095e-06, "loss": 0.8683, "step": 535 }, { "epoch": 0.19, "learning_rate": 9.56361346050027e-06, "loss": 0.9311, "step": 536 }, { "epoch": 0.19, "learning_rate": 9.562713694439446e-06, "loss": 0.8371, "step": 537 }, { "epoch": 0.19, "learning_rate": 9.561813928378623e-06, "loss": 0.9625, "step": 538 }, { "epoch": 0.19, "learning_rate": 9.560914162317798e-06, "loss": 0.8475, "step": 539 }, { "epoch": 0.19, "learning_rate": 9.560014396256973e-06, "loss": 0.9831, "step": 540 }, { "epoch": 0.19, "learning_rate": 9.55911463019615e-06, "loss": 0.9152, "step": 541 }, { "epoch": 0.19, "learning_rate": 9.558214864135326e-06, "loss": 0.97, "step": 542 }, { "epoch": 0.19, "learning_rate": 9.557315098074501e-06, "loss": 0.8396, "step": 543 }, { "epoch": 0.2, "learning_rate": 9.556415332013678e-06, "loss": 0.9302, "step": 544 }, { "epoch": 0.2, "learning_rate": 9.555515565952853e-06, "loss": 0.9003, "step": 545 }, { "epoch": 0.2, "learning_rate": 9.55461579989203e-06, "loss": 0.9765, "step": 546 }, { "epoch": 0.2, "learning_rate": 9.553716033831206e-06, "loss": 0.7834, "step": 547 }, { "epoch": 0.2, "learning_rate": 9.55281626777038e-06, "loss": 0.9618, "step": 548 }, { "epoch": 0.2, "learning_rate": 9.551916501709556e-06, "loss": 0.8588, "step": 549 }, { "epoch": 0.2, "learning_rate": 9.551016735648732e-06, "loss": 0.8226, "step": 550 }, { "epoch": 0.2, "learning_rate": 9.550116969587909e-06, "loss": 0.9321, "step": 551 }, { "epoch": 0.2, "learning_rate": 9.549217203527084e-06, "loss": 0.804, "step": 552 }, { "epoch": 0.2, "learning_rate": 9.548317437466259e-06, "loss": 0.8472, "step": 553 }, { "epoch": 0.2, "learning_rate": 9.547417671405435e-06, "loss": 0.8383, "step": 554 }, { "epoch": 0.2, "learning_rate": 9.54651790534461e-06, "loss": 0.9001, "step": 555 }, { "epoch": 0.2, "learning_rate": 9.545618139283787e-06, "loss": 0.8829, "step": 556 }, { "epoch": 0.2, "learning_rate": 9.544718373222964e-06, "loss": 0.9914, "step": 557 }, { "epoch": 0.2, "learning_rate": 9.543818607162138e-06, "loss": 1.013, "step": 558 }, { "epoch": 0.2, "learning_rate": 9.542918841101313e-06, "loss": 0.8709, "step": 559 }, { "epoch": 0.2, "learning_rate": 9.54201907504049e-06, "loss": 0.8773, "step": 560 }, { "epoch": 0.2, "learning_rate": 9.541119308979667e-06, "loss": 0.8578, "step": 561 }, { "epoch": 0.2, "learning_rate": 9.540219542918842e-06, "loss": 0.8965, "step": 562 }, { "epoch": 0.2, "learning_rate": 9.539319776858018e-06, "loss": 0.8906, "step": 563 }, { "epoch": 0.2, "learning_rate": 9.538420010797193e-06, "loss": 1.0673, "step": 564 }, { "epoch": 0.2, "learning_rate": 9.53752024473637e-06, "loss": 0.868, "step": 565 }, { "epoch": 0.2, "learning_rate": 9.536620478675546e-06, "loss": 1.0298, "step": 566 }, { "epoch": 0.2, "learning_rate": 9.535720712614721e-06, "loss": 0.8982, "step": 567 }, { "epoch": 0.2, "learning_rate": 9.534820946553896e-06, "loss": 0.8499, "step": 568 }, { "epoch": 0.2, "learning_rate": 9.533921180493073e-06, "loss": 0.7938, "step": 569 }, { "epoch": 0.2, "learning_rate": 9.53302141443225e-06, "loss": 0.8709, "step": 570 }, { "epoch": 0.2, "learning_rate": 9.532121648371424e-06, "loss": 0.8669, "step": 571 }, { "epoch": 0.21, "learning_rate": 9.5312218823106e-06, "loss": 0.8065, "step": 572 }, { "epoch": 0.21, "learning_rate": 9.530322116249776e-06, "loss": 1.0002, "step": 573 }, { "epoch": 0.21, "learning_rate": 9.52942235018895e-06, "loss": 0.8445, "step": 574 }, { "epoch": 0.21, "learning_rate": 9.528522584128127e-06, "loss": 0.8646, "step": 575 }, { "epoch": 0.21, "learning_rate": 9.527622818067304e-06, "loss": 0.8756, "step": 576 }, { "epoch": 0.21, "learning_rate": 9.526723052006479e-06, "loss": 0.8671, "step": 577 }, { "epoch": 0.21, "learning_rate": 9.525823285945654e-06, "loss": 0.8313, "step": 578 }, { "epoch": 0.21, "learning_rate": 9.52492351988483e-06, "loss": 0.8274, "step": 579 }, { "epoch": 0.21, "learning_rate": 9.524023753824007e-06, "loss": 0.8751, "step": 580 }, { "epoch": 0.21, "learning_rate": 9.523123987763182e-06, "loss": 0.9716, "step": 581 }, { "epoch": 0.21, "learning_rate": 9.522224221702359e-06, "loss": 0.8402, "step": 582 }, { "epoch": 0.21, "learning_rate": 9.521324455641534e-06, "loss": 0.8939, "step": 583 }, { "epoch": 0.21, "learning_rate": 9.52042468958071e-06, "loss": 1.0017, "step": 584 }, { "epoch": 0.21, "learning_rate": 9.519524923519887e-06, "loss": 0.8231, "step": 585 }, { "epoch": 0.21, "learning_rate": 9.518625157459062e-06, "loss": 0.7424, "step": 586 }, { "epoch": 0.21, "learning_rate": 9.517725391398237e-06, "loss": 0.8941, "step": 587 }, { "epoch": 0.21, "learning_rate": 9.516825625337413e-06, "loss": 0.8212, "step": 588 }, { "epoch": 0.21, "learning_rate": 9.51592585927659e-06, "loss": 0.7851, "step": 589 }, { "epoch": 0.21, "learning_rate": 9.515026093215765e-06, "loss": 0.992, "step": 590 }, { "epoch": 0.21, "learning_rate": 9.51412632715494e-06, "loss": 1.0166, "step": 591 }, { "epoch": 0.21, "learning_rate": 9.513226561094116e-06, "loss": 0.7765, "step": 592 }, { "epoch": 0.21, "learning_rate": 9.512326795033291e-06, "loss": 0.8851, "step": 593 }, { "epoch": 0.21, "learning_rate": 9.511427028972468e-06, "loss": 0.8803, "step": 594 }, { "epoch": 0.21, "learning_rate": 9.510527262911644e-06, "loss": 0.9177, "step": 595 }, { "epoch": 0.21, "learning_rate": 9.50962749685082e-06, "loss": 0.9284, "step": 596 }, { "epoch": 0.21, "learning_rate": 9.508727730789994e-06, "loss": 0.8524, "step": 597 }, { "epoch": 0.21, "learning_rate": 9.507827964729171e-06, "loss": 0.8164, "step": 598 }, { "epoch": 0.22, "learning_rate": 9.506928198668348e-06, "loss": 0.8096, "step": 599 }, { "epoch": 0.22, "learning_rate": 9.506028432607522e-06, "loss": 0.9689, "step": 600 }, { "epoch": 0.22, "learning_rate": 9.505128666546699e-06, "loss": 0.9344, "step": 601 }, { "epoch": 0.22, "learning_rate": 9.504228900485874e-06, "loss": 0.876, "step": 602 }, { "epoch": 0.22, "learning_rate": 9.50332913442505e-06, "loss": 0.844, "step": 603 }, { "epoch": 0.22, "learning_rate": 9.502429368364227e-06, "loss": 0.8466, "step": 604 }, { "epoch": 0.22, "learning_rate": 9.501529602303402e-06, "loss": 0.8977, "step": 605 }, { "epoch": 0.22, "learning_rate": 9.500629836242577e-06, "loss": 0.8746, "step": 606 }, { "epoch": 0.22, "learning_rate": 9.499730070181754e-06, "loss": 0.993, "step": 607 }, { "epoch": 0.22, "learning_rate": 9.49883030412093e-06, "loss": 0.8704, "step": 608 }, { "epoch": 0.22, "learning_rate": 9.497930538060105e-06, "loss": 0.9734, "step": 609 }, { "epoch": 0.22, "learning_rate": 9.49703077199928e-06, "loss": 0.96, "step": 610 }, { "epoch": 0.22, "learning_rate": 9.496131005938457e-06, "loss": 0.862, "step": 611 }, { "epoch": 0.22, "learning_rate": 9.495231239877632e-06, "loss": 0.8751, "step": 612 }, { "epoch": 0.22, "learning_rate": 9.494331473816808e-06, "loss": 0.9226, "step": 613 }, { "epoch": 0.22, "learning_rate": 9.493431707755985e-06, "loss": 0.9507, "step": 614 }, { "epoch": 0.22, "learning_rate": 9.49253194169516e-06, "loss": 0.8565, "step": 615 }, { "epoch": 0.22, "learning_rate": 9.491632175634335e-06, "loss": 0.8928, "step": 616 }, { "epoch": 0.22, "learning_rate": 9.490732409573511e-06, "loss": 0.8646, "step": 617 }, { "epoch": 0.22, "learning_rate": 9.489832643512688e-06, "loss": 0.8232, "step": 618 }, { "epoch": 0.22, "learning_rate": 9.488932877451863e-06, "loss": 0.9304, "step": 619 }, { "epoch": 0.22, "learning_rate": 9.48803311139104e-06, "loss": 1.1176, "step": 620 }, { "epoch": 0.22, "learning_rate": 9.487133345330214e-06, "loss": 0.8811, "step": 621 }, { "epoch": 0.22, "learning_rate": 9.486233579269391e-06, "loss": 0.9005, "step": 622 }, { "epoch": 0.22, "learning_rate": 9.485333813208568e-06, "loss": 0.9069, "step": 623 }, { "epoch": 0.22, "learning_rate": 9.484434047147743e-06, "loss": 0.87, "step": 624 }, { "epoch": 0.22, "learning_rate": 9.483534281086918e-06, "loss": 0.8352, "step": 625 }, { "epoch": 0.22, "learning_rate": 9.482634515026094e-06, "loss": 0.861, "step": 626 }, { "epoch": 0.23, "learning_rate": 9.48173474896527e-06, "loss": 1.0083, "step": 627 }, { "epoch": 0.23, "learning_rate": 9.480834982904446e-06, "loss": 0.9525, "step": 628 }, { "epoch": 0.23, "learning_rate": 9.47993521684362e-06, "loss": 0.8825, "step": 629 }, { "epoch": 0.23, "learning_rate": 9.479035450782797e-06, "loss": 1.0026, "step": 630 }, { "epoch": 0.23, "learning_rate": 9.478135684721974e-06, "loss": 0.8977, "step": 631 }, { "epoch": 0.23, "learning_rate": 9.477235918661149e-06, "loss": 0.8544, "step": 632 }, { "epoch": 0.23, "learning_rate": 9.476336152600325e-06, "loss": 0.7888, "step": 633 }, { "epoch": 0.23, "learning_rate": 9.4754363865395e-06, "loss": 0.9793, "step": 634 }, { "epoch": 0.23, "learning_rate": 9.474536620478675e-06, "loss": 0.9994, "step": 635 }, { "epoch": 0.23, "learning_rate": 9.473636854417852e-06, "loss": 0.7589, "step": 636 }, { "epoch": 0.23, "learning_rate": 9.472737088357028e-06, "loss": 0.9013, "step": 637 }, { "epoch": 0.23, "learning_rate": 9.471837322296203e-06, "loss": 0.9175, "step": 638 }, { "epoch": 0.23, "learning_rate": 9.47093755623538e-06, "loss": 0.8714, "step": 639 }, { "epoch": 0.23, "learning_rate": 9.470037790174555e-06, "loss": 0.9299, "step": 640 }, { "epoch": 0.23, "learning_rate": 9.469138024113732e-06, "loss": 0.8808, "step": 641 }, { "epoch": 0.23, "learning_rate": 9.468238258052908e-06, "loss": 0.8799, "step": 642 }, { "epoch": 0.23, "learning_rate": 9.467338491992083e-06, "loss": 0.8842, "step": 643 }, { "epoch": 0.23, "learning_rate": 9.466438725931258e-06, "loss": 0.8937, "step": 644 }, { "epoch": 0.23, "learning_rate": 9.465538959870435e-06, "loss": 0.8253, "step": 645 }, { "epoch": 0.23, "learning_rate": 9.464639193809611e-06, "loss": 0.8446, "step": 646 }, { "epoch": 0.23, "learning_rate": 9.463739427748786e-06, "loss": 0.7923, "step": 647 }, { "epoch": 0.23, "learning_rate": 9.462839661687961e-06, "loss": 0.9168, "step": 648 }, { "epoch": 0.23, "learning_rate": 9.461939895627138e-06, "loss": 0.9707, "step": 649 }, { "epoch": 0.23, "learning_rate": 9.461040129566314e-06, "loss": 0.8569, "step": 650 }, { "epoch": 0.23, "learning_rate": 9.46014036350549e-06, "loss": 0.8873, "step": 651 }, { "epoch": 0.23, "learning_rate": 9.459240597444666e-06, "loss": 0.9129, "step": 652 }, { "epoch": 0.23, "learning_rate": 9.45834083138384e-06, "loss": 0.8955, "step": 653 }, { "epoch": 0.23, "learning_rate": 9.457441065323016e-06, "loss": 0.7968, "step": 654 }, { "epoch": 0.24, "learning_rate": 9.456541299262192e-06, "loss": 0.8642, "step": 655 }, { "epoch": 0.24, "learning_rate": 9.455641533201369e-06, "loss": 0.9634, "step": 656 }, { "epoch": 0.24, "learning_rate": 9.454741767140544e-06, "loss": 0.8605, "step": 657 }, { "epoch": 0.24, "learning_rate": 9.45384200107972e-06, "loss": 0.887, "step": 658 }, { "epoch": 0.24, "learning_rate": 9.452942235018895e-06, "loss": 0.9597, "step": 659 }, { "epoch": 0.24, "learning_rate": 9.452042468958072e-06, "loss": 0.8449, "step": 660 }, { "epoch": 0.24, "learning_rate": 9.451142702897249e-06, "loss": 0.978, "step": 661 }, { "epoch": 0.24, "learning_rate": 9.450242936836424e-06, "loss": 0.8293, "step": 662 }, { "epoch": 0.24, "learning_rate": 9.449343170775598e-06, "loss": 0.975, "step": 663 }, { "epoch": 0.24, "learning_rate": 9.448443404714775e-06, "loss": 0.9557, "step": 664 }, { "epoch": 0.24, "learning_rate": 9.447543638653952e-06, "loss": 0.8788, "step": 665 }, { "epoch": 0.24, "learning_rate": 9.446643872593127e-06, "loss": 0.9037, "step": 666 }, { "epoch": 0.24, "learning_rate": 9.445744106532301e-06, "loss": 0.8504, "step": 667 }, { "epoch": 0.24, "learning_rate": 9.444844340471478e-06, "loss": 0.7426, "step": 668 }, { "epoch": 0.24, "learning_rate": 9.443944574410655e-06, "loss": 0.7847, "step": 669 }, { "epoch": 0.24, "learning_rate": 9.44304480834983e-06, "loss": 0.904, "step": 670 }, { "epoch": 0.24, "learning_rate": 9.442145042289006e-06, "loss": 0.8531, "step": 671 }, { "epoch": 0.24, "learning_rate": 9.441245276228181e-06, "loss": 0.8996, "step": 672 }, { "epoch": 0.24, "learning_rate": 9.440345510167356e-06, "loss": 0.8615, "step": 673 }, { "epoch": 0.24, "learning_rate": 9.439445744106533e-06, "loss": 0.7879, "step": 674 }, { "epoch": 0.24, "learning_rate": 9.43854597804571e-06, "loss": 0.8235, "step": 675 }, { "epoch": 0.24, "learning_rate": 9.437646211984884e-06, "loss": 1.0531, "step": 676 }, { "epoch": 0.24, "learning_rate": 9.436746445924061e-06, "loss": 0.9119, "step": 677 }, { "epoch": 0.24, "learning_rate": 9.435846679863236e-06, "loss": 0.7906, "step": 678 }, { "epoch": 0.24, "learning_rate": 9.434946913802412e-06, "loss": 0.837, "step": 679 }, { "epoch": 0.24, "learning_rate": 9.434047147741589e-06, "loss": 0.8543, "step": 680 }, { "epoch": 0.24, "learning_rate": 9.433147381680764e-06, "loss": 0.8399, "step": 681 }, { "epoch": 0.24, "learning_rate": 9.432247615619939e-06, "loss": 1.1176, "step": 682 }, { "epoch": 0.25, "learning_rate": 9.431347849559115e-06, "loss": 0.8394, "step": 683 }, { "epoch": 0.25, "learning_rate": 9.430448083498292e-06, "loss": 0.9621, "step": 684 }, { "epoch": 0.25, "learning_rate": 9.429548317437467e-06, "loss": 0.9575, "step": 685 }, { "epoch": 0.25, "learning_rate": 9.428648551376642e-06, "loss": 0.9146, "step": 686 }, { "epoch": 0.25, "learning_rate": 9.427748785315819e-06, "loss": 0.8646, "step": 687 }, { "epoch": 0.25, "learning_rate": 9.426849019254995e-06, "loss": 0.99, "step": 688 }, { "epoch": 0.25, "learning_rate": 9.42594925319417e-06, "loss": 0.9794, "step": 689 }, { "epoch": 0.25, "learning_rate": 9.425049487133347e-06, "loss": 0.7298, "step": 690 }, { "epoch": 0.25, "learning_rate": 9.424149721072522e-06, "loss": 0.9136, "step": 691 }, { "epoch": 0.25, "learning_rate": 9.423249955011697e-06, "loss": 0.8824, "step": 692 }, { "epoch": 0.25, "learning_rate": 9.422350188950873e-06, "loss": 0.8805, "step": 693 }, { "epoch": 0.25, "learning_rate": 9.42145042289005e-06, "loss": 0.8297, "step": 694 }, { "epoch": 0.25, "learning_rate": 9.420550656829225e-06, "loss": 0.8008, "step": 695 }, { "epoch": 0.25, "learning_rate": 9.419650890768401e-06, "loss": 0.9434, "step": 696 }, { "epoch": 0.25, "learning_rate": 9.418751124707576e-06, "loss": 0.9827, "step": 697 }, { "epoch": 0.25, "learning_rate": 9.417851358646753e-06, "loss": 0.8447, "step": 698 }, { "epoch": 0.25, "learning_rate": 9.41695159258593e-06, "loss": 0.8443, "step": 699 }, { "epoch": 0.25, "learning_rate": 9.416051826525104e-06, "loss": 0.9086, "step": 700 }, { "epoch": 0.25, "learning_rate": 9.41515206046428e-06, "loss": 0.8064, "step": 701 }, { "epoch": 0.25, "learning_rate": 9.414252294403456e-06, "loss": 0.9501, "step": 702 }, { "epoch": 0.25, "learning_rate": 9.413352528342633e-06, "loss": 0.8164, "step": 703 }, { "epoch": 0.25, "learning_rate": 9.412452762281807e-06, "loss": 0.8572, "step": 704 }, { "epoch": 0.25, "learning_rate": 9.411552996220982e-06, "loss": 0.9424, "step": 705 }, { "epoch": 0.25, "learning_rate": 9.410653230160159e-06, "loss": 0.9589, "step": 706 }, { "epoch": 0.25, "learning_rate": 9.409753464099336e-06, "loss": 0.9032, "step": 707 }, { "epoch": 0.25, "learning_rate": 9.40885369803851e-06, "loss": 0.8058, "step": 708 }, { "epoch": 0.25, "learning_rate": 9.407953931977687e-06, "loss": 0.8086, "step": 709 }, { "epoch": 0.25, "learning_rate": 9.407054165916862e-06, "loss": 0.9284, "step": 710 }, { "epoch": 0.26, "learning_rate": 9.406154399856037e-06, "loss": 0.8547, "step": 711 }, { "epoch": 0.26, "learning_rate": 9.405254633795214e-06, "loss": 0.9481, "step": 712 }, { "epoch": 0.26, "learning_rate": 9.40435486773439e-06, "loss": 0.9495, "step": 713 }, { "epoch": 0.26, "learning_rate": 9.403455101673565e-06, "loss": 0.8416, "step": 714 }, { "epoch": 0.26, "learning_rate": 9.402555335612742e-06, "loss": 0.8595, "step": 715 }, { "epoch": 0.26, "learning_rate": 9.401655569551917e-06, "loss": 0.87, "step": 716 }, { "epoch": 0.26, "learning_rate": 9.400755803491093e-06, "loss": 0.9736, "step": 717 }, { "epoch": 0.26, "learning_rate": 9.39985603743027e-06, "loss": 0.9437, "step": 718 }, { "epoch": 0.26, "learning_rate": 9.398956271369445e-06, "loss": 0.7999, "step": 719 }, { "epoch": 0.26, "learning_rate": 9.39805650530862e-06, "loss": 0.8703, "step": 720 }, { "epoch": 0.26, "learning_rate": 9.397156739247796e-06, "loss": 0.8758, "step": 721 }, { "epoch": 0.26, "learning_rate": 9.396256973186973e-06, "loss": 0.8098, "step": 722 }, { "epoch": 0.26, "learning_rate": 9.395357207126148e-06, "loss": 0.8259, "step": 723 }, { "epoch": 0.26, "learning_rate": 9.394457441065323e-06, "loss": 0.9589, "step": 724 }, { "epoch": 0.26, "learning_rate": 9.3935576750045e-06, "loss": 0.8495, "step": 725 }, { "epoch": 0.26, "learning_rate": 9.392657908943676e-06, "loss": 0.8866, "step": 726 }, { "epoch": 0.26, "learning_rate": 9.391758142882851e-06, "loss": 0.8896, "step": 727 }, { "epoch": 0.26, "learning_rate": 9.390858376822028e-06, "loss": 0.8878, "step": 728 }, { "epoch": 0.26, "learning_rate": 9.389958610761203e-06, "loss": 0.934, "step": 729 }, { "epoch": 0.26, "learning_rate": 9.389058844700377e-06, "loss": 0.9386, "step": 730 }, { "epoch": 0.26, "learning_rate": 9.388159078639554e-06, "loss": 0.9224, "step": 731 }, { "epoch": 0.26, "learning_rate": 9.38725931257873e-06, "loss": 0.9447, "step": 732 }, { "epoch": 0.26, "learning_rate": 9.386359546517906e-06, "loss": 0.8337, "step": 733 }, { "epoch": 0.26, "learning_rate": 9.385459780457082e-06, "loss": 0.8292, "step": 734 }, { "epoch": 0.26, "learning_rate": 9.384560014396257e-06, "loss": 0.94, "step": 735 }, { "epoch": 0.26, "learning_rate": 9.383660248335434e-06, "loss": 0.8949, "step": 736 }, { "epoch": 0.26, "learning_rate": 9.38276048227461e-06, "loss": 0.8673, "step": 737 }, { "epoch": 0.26, "learning_rate": 9.381860716213785e-06, "loss": 0.8096, "step": 738 }, { "epoch": 0.27, "learning_rate": 9.38096095015296e-06, "loss": 0.8924, "step": 739 }, { "epoch": 0.27, "learning_rate": 9.380061184092137e-06, "loss": 0.8892, "step": 740 }, { "epoch": 0.27, "learning_rate": 9.379161418031313e-06, "loss": 0.9018, "step": 741 }, { "epoch": 0.27, "learning_rate": 9.378261651970488e-06, "loss": 0.9394, "step": 742 }, { "epoch": 0.27, "learning_rate": 9.377361885909663e-06, "loss": 0.873, "step": 743 }, { "epoch": 0.27, "learning_rate": 9.37646211984884e-06, "loss": 0.8328, "step": 744 }, { "epoch": 0.27, "learning_rate": 9.375562353788017e-06, "loss": 0.8594, "step": 745 }, { "epoch": 0.27, "learning_rate": 9.374662587727191e-06, "loss": 1.0163, "step": 746 }, { "epoch": 0.27, "learning_rate": 9.373762821666368e-06, "loss": 0.9264, "step": 747 }, { "epoch": 0.27, "learning_rate": 9.372863055605543e-06, "loss": 0.9081, "step": 748 }, { "epoch": 0.27, "learning_rate": 9.37196328954472e-06, "loss": 0.9195, "step": 749 }, { "epoch": 0.27, "learning_rate": 9.371063523483895e-06, "loss": 0.9409, "step": 750 }, { "epoch": 0.27, "learning_rate": 9.370163757423071e-06, "loss": 0.8397, "step": 751 }, { "epoch": 0.27, "learning_rate": 9.369263991362246e-06, "loss": 0.9001, "step": 752 }, { "epoch": 0.27, "learning_rate": 9.368364225301423e-06, "loss": 0.8563, "step": 753 }, { "epoch": 0.27, "learning_rate": 9.367464459240598e-06, "loss": 0.8371, "step": 754 }, { "epoch": 0.27, "learning_rate": 9.366564693179774e-06, "loss": 0.9211, "step": 755 }, { "epoch": 0.27, "learning_rate": 9.365664927118951e-06, "loss": 1.025, "step": 756 }, { "epoch": 0.27, "learning_rate": 9.364765161058126e-06, "loss": 0.8751, "step": 757 }, { "epoch": 0.27, "learning_rate": 9.3638653949973e-06, "loss": 0.8094, "step": 758 }, { "epoch": 0.27, "learning_rate": 9.362965628936477e-06, "loss": 0.7525, "step": 759 }, { "epoch": 0.27, "learning_rate": 9.362065862875654e-06, "loss": 0.8836, "step": 760 }, { "epoch": 0.27, "learning_rate": 9.361166096814829e-06, "loss": 0.8583, "step": 761 }, { "epoch": 0.27, "learning_rate": 9.360266330754004e-06, "loss": 0.7841, "step": 762 }, { "epoch": 0.27, "learning_rate": 9.35936656469318e-06, "loss": 0.8924, "step": 763 }, { "epoch": 0.27, "learning_rate": 9.358466798632357e-06, "loss": 0.8721, "step": 764 }, { "epoch": 0.27, "learning_rate": 9.357567032571532e-06, "loss": 0.7824, "step": 765 }, { "epoch": 0.27, "learning_rate": 9.356667266510709e-06, "loss": 0.8786, "step": 766 }, { "epoch": 0.28, "learning_rate": 9.355767500449883e-06, "loss": 0.867, "step": 767 }, { "epoch": 0.28, "learning_rate": 9.35486773438906e-06, "loss": 0.8712, "step": 768 }, { "epoch": 0.28, "learning_rate": 9.353967968328235e-06, "loss": 0.756, "step": 769 }, { "epoch": 0.28, "learning_rate": 9.353068202267412e-06, "loss": 0.9202, "step": 770 }, { "epoch": 0.28, "learning_rate": 9.352168436206587e-06, "loss": 1.0414, "step": 771 }, { "epoch": 0.28, "learning_rate": 9.351268670145763e-06, "loss": 0.8797, "step": 772 }, { "epoch": 0.28, "learning_rate": 9.350368904084938e-06, "loss": 0.9018, "step": 773 }, { "epoch": 0.28, "learning_rate": 9.349469138024115e-06, "loss": 0.8899, "step": 774 }, { "epoch": 0.28, "learning_rate": 9.348569371963291e-06, "loss": 0.9084, "step": 775 }, { "epoch": 0.28, "learning_rate": 9.347669605902466e-06, "loss": 0.7766, "step": 776 }, { "epoch": 0.28, "learning_rate": 9.346769839841641e-06, "loss": 0.8855, "step": 777 }, { "epoch": 0.28, "learning_rate": 9.345870073780818e-06, "loss": 0.9406, "step": 778 }, { "epoch": 0.28, "learning_rate": 9.344970307719994e-06, "loss": 0.9028, "step": 779 }, { "epoch": 0.28, "learning_rate": 9.34407054165917e-06, "loss": 0.8585, "step": 780 }, { "epoch": 0.28, "learning_rate": 9.343170775598344e-06, "loss": 0.9153, "step": 781 }, { "epoch": 0.28, "learning_rate": 9.34227100953752e-06, "loss": 0.9577, "step": 782 }, { "epoch": 0.28, "learning_rate": 9.341371243476697e-06, "loss": 0.8926, "step": 783 }, { "epoch": 0.28, "learning_rate": 9.340471477415872e-06, "loss": 0.8354, "step": 784 }, { "epoch": 0.28, "learning_rate": 9.339571711355049e-06, "loss": 0.8774, "step": 785 }, { "epoch": 0.28, "learning_rate": 9.338671945294224e-06, "loss": 0.8995, "step": 786 }, { "epoch": 0.28, "learning_rate": 9.3377721792334e-06, "loss": 0.9256, "step": 787 }, { "epoch": 0.28, "learning_rate": 9.336872413172577e-06, "loss": 0.8668, "step": 788 }, { "epoch": 0.28, "learning_rate": 9.335972647111752e-06, "loss": 0.8613, "step": 789 }, { "epoch": 0.28, "learning_rate": 9.335072881050927e-06, "loss": 0.9657, "step": 790 }, { "epoch": 0.28, "learning_rate": 9.334173114990104e-06, "loss": 0.8793, "step": 791 }, { "epoch": 0.28, "learning_rate": 9.333273348929279e-06, "loss": 0.8634, "step": 792 }, { "epoch": 0.28, "learning_rate": 9.332373582868455e-06, "loss": 0.9284, "step": 793 }, { "epoch": 0.28, "learning_rate": 9.331473816807632e-06, "loss": 0.9001, "step": 794 }, { "epoch": 0.29, "learning_rate": 9.330574050746807e-06, "loss": 0.7616, "step": 795 }, { "epoch": 0.29, "learning_rate": 9.329674284685982e-06, "loss": 0.8407, "step": 796 }, { "epoch": 0.29, "learning_rate": 9.328774518625158e-06, "loss": 0.8926, "step": 797 }, { "epoch": 0.29, "learning_rate": 9.327874752564335e-06, "loss": 0.9383, "step": 798 }, { "epoch": 0.29, "learning_rate": 9.32697498650351e-06, "loss": 0.837, "step": 799 }, { "epoch": 0.29, "learning_rate": 9.326075220442685e-06, "loss": 0.8876, "step": 800 }, { "epoch": 0.29, "learning_rate": 9.325175454381861e-06, "loss": 0.9506, "step": 801 }, { "epoch": 0.29, "learning_rate": 9.324275688321038e-06, "loss": 0.9768, "step": 802 }, { "epoch": 0.29, "learning_rate": 9.323375922260213e-06, "loss": 0.7889, "step": 803 }, { "epoch": 0.29, "learning_rate": 9.32247615619939e-06, "loss": 0.8398, "step": 804 }, { "epoch": 0.29, "learning_rate": 9.321576390138564e-06, "loss": 0.8868, "step": 805 }, { "epoch": 0.29, "learning_rate": 9.320676624077741e-06, "loss": 0.8164, "step": 806 }, { "epoch": 0.29, "learning_rate": 9.319776858016918e-06, "loss": 0.8737, "step": 807 }, { "epoch": 0.29, "learning_rate": 9.318877091956093e-06, "loss": 0.9286, "step": 808 }, { "epoch": 0.29, "learning_rate": 9.317977325895267e-06, "loss": 0.7801, "step": 809 }, { "epoch": 0.29, "learning_rate": 9.317077559834444e-06, "loss": 0.8546, "step": 810 }, { "epoch": 0.29, "learning_rate": 9.316177793773619e-06, "loss": 0.8172, "step": 811 }, { "epoch": 0.29, "learning_rate": 9.315278027712796e-06, "loss": 0.955, "step": 812 }, { "epoch": 0.29, "learning_rate": 9.314378261651972e-06, "loss": 0.9191, "step": 813 }, { "epoch": 0.29, "learning_rate": 9.313478495591147e-06, "loss": 1.0594, "step": 814 }, { "epoch": 0.29, "learning_rate": 9.312578729530322e-06, "loss": 0.8924, "step": 815 }, { "epoch": 0.29, "learning_rate": 9.311678963469499e-06, "loss": 0.8153, "step": 816 }, { "epoch": 0.29, "learning_rate": 9.310779197408675e-06, "loss": 0.9376, "step": 817 }, { "epoch": 0.29, "learning_rate": 9.30987943134785e-06, "loss": 0.861, "step": 818 }, { "epoch": 0.29, "learning_rate": 9.308979665287025e-06, "loss": 0.961, "step": 819 }, { "epoch": 0.29, "learning_rate": 9.308079899226202e-06, "loss": 0.8507, "step": 820 }, { "epoch": 0.29, "learning_rate": 9.307180133165378e-06, "loss": 0.8485, "step": 821 }, { "epoch": 0.3, "learning_rate": 9.306280367104553e-06, "loss": 0.8007, "step": 822 }, { "epoch": 0.3, "learning_rate": 9.30538060104373e-06, "loss": 0.907, "step": 823 }, { "epoch": 0.3, "learning_rate": 9.304480834982905e-06, "loss": 0.8955, "step": 824 }, { "epoch": 0.3, "learning_rate": 9.303581068922081e-06, "loss": 0.765, "step": 825 }, { "epoch": 0.3, "learning_rate": 9.302681302861258e-06, "loss": 0.7643, "step": 826 }, { "epoch": 0.3, "learning_rate": 9.301781536800433e-06, "loss": 0.9545, "step": 827 }, { "epoch": 0.3, "learning_rate": 9.300881770739608e-06, "loss": 0.9608, "step": 828 }, { "epoch": 0.3, "learning_rate": 9.299982004678785e-06, "loss": 0.8182, "step": 829 }, { "epoch": 0.3, "learning_rate": 9.29908223861796e-06, "loss": 0.9194, "step": 830 }, { "epoch": 0.3, "learning_rate": 9.298182472557136e-06, "loss": 0.9672, "step": 831 }, { "epoch": 0.3, "learning_rate": 9.297282706496313e-06, "loss": 0.8798, "step": 832 }, { "epoch": 0.3, "learning_rate": 9.296382940435488e-06, "loss": 0.874, "step": 833 }, { "epoch": 0.3, "learning_rate": 9.295483174374663e-06, "loss": 0.8816, "step": 834 }, { "epoch": 0.3, "learning_rate": 9.294583408313839e-06, "loss": 0.8067, "step": 835 }, { "epoch": 0.3, "learning_rate": 9.293683642253016e-06, "loss": 0.8799, "step": 836 }, { "epoch": 0.3, "learning_rate": 9.29278387619219e-06, "loss": 1.0053, "step": 837 }, { "epoch": 0.3, "learning_rate": 9.291884110131366e-06, "loss": 0.9308, "step": 838 }, { "epoch": 0.3, "learning_rate": 9.290984344070542e-06, "loss": 0.8546, "step": 839 }, { "epoch": 0.3, "learning_rate": 9.290084578009719e-06, "loss": 0.9351, "step": 840 }, { "epoch": 0.3, "learning_rate": 9.289184811948894e-06, "loss": 0.7221, "step": 841 }, { "epoch": 0.3, "learning_rate": 9.28828504588807e-06, "loss": 0.8767, "step": 842 }, { "epoch": 0.3, "learning_rate": 9.287385279827245e-06, "loss": 0.802, "step": 843 }, { "epoch": 0.3, "learning_rate": 9.286485513766422e-06, "loss": 0.9537, "step": 844 }, { "epoch": 0.3, "learning_rate": 9.285585747705599e-06, "loss": 0.8237, "step": 845 }, { "epoch": 0.3, "learning_rate": 9.284685981644773e-06, "loss": 0.8051, "step": 846 }, { "epoch": 0.3, "learning_rate": 9.283786215583948e-06, "loss": 0.8692, "step": 847 }, { "epoch": 0.3, "learning_rate": 9.282886449523125e-06, "loss": 0.884, "step": 848 }, { "epoch": 0.3, "learning_rate": 9.2819866834623e-06, "loss": 0.9786, "step": 849 }, { "epoch": 0.31, "learning_rate": 9.281086917401477e-06, "loss": 0.8434, "step": 850 }, { "epoch": 0.31, "learning_rate": 9.280187151340653e-06, "loss": 0.9774, "step": 851 }, { "epoch": 0.31, "learning_rate": 9.279287385279828e-06, "loss": 0.9603, "step": 852 }, { "epoch": 0.31, "learning_rate": 9.278387619219003e-06, "loss": 0.9013, "step": 853 }, { "epoch": 0.31, "learning_rate": 9.27748785315818e-06, "loss": 0.9989, "step": 854 }, { "epoch": 0.31, "learning_rate": 9.276588087097356e-06, "loss": 0.6652, "step": 855 }, { "epoch": 0.31, "learning_rate": 9.275688321036531e-06, "loss": 0.9182, "step": 856 }, { "epoch": 0.31, "learning_rate": 9.274788554975706e-06, "loss": 0.8572, "step": 857 }, { "epoch": 0.31, "learning_rate": 9.273888788914883e-06, "loss": 0.9003, "step": 858 }, { "epoch": 0.31, "learning_rate": 9.27298902285406e-06, "loss": 0.8811, "step": 859 }, { "epoch": 0.31, "learning_rate": 9.272089256793234e-06, "loss": 0.8285, "step": 860 }, { "epoch": 0.31, "learning_rate": 9.27118949073241e-06, "loss": 0.9128, "step": 861 }, { "epoch": 0.31, "learning_rate": 9.270289724671586e-06, "loss": 0.9187, "step": 862 }, { "epoch": 0.31, "learning_rate": 9.269389958610762e-06, "loss": 0.8239, "step": 863 }, { "epoch": 0.31, "learning_rate": 9.268490192549939e-06, "loss": 0.9186, "step": 864 }, { "epoch": 0.31, "learning_rate": 9.267590426489114e-06, "loss": 0.8424, "step": 865 }, { "epoch": 0.31, "learning_rate": 9.266690660428289e-06, "loss": 0.8512, "step": 866 }, { "epoch": 0.31, "learning_rate": 9.265790894367465e-06, "loss": 0.9512, "step": 867 }, { "epoch": 0.31, "learning_rate": 9.26489112830664e-06, "loss": 0.8748, "step": 868 }, { "epoch": 0.31, "learning_rate": 9.263991362245817e-06, "loss": 0.8395, "step": 869 }, { "epoch": 0.31, "learning_rate": 9.263091596184994e-06, "loss": 0.7709, "step": 870 }, { "epoch": 0.31, "learning_rate": 9.262191830124168e-06, "loss": 0.8644, "step": 871 }, { "epoch": 0.31, "learning_rate": 9.261292064063343e-06, "loss": 0.9015, "step": 872 }, { "epoch": 0.31, "learning_rate": 9.26039229800252e-06, "loss": 0.829, "step": 873 }, { "epoch": 0.31, "learning_rate": 9.259492531941697e-06, "loss": 0.7871, "step": 874 }, { "epoch": 0.31, "learning_rate": 9.258592765880872e-06, "loss": 0.8723, "step": 875 }, { "epoch": 0.31, "learning_rate": 9.257692999820046e-06, "loss": 0.7756, "step": 876 }, { "epoch": 0.31, "learning_rate": 9.256793233759223e-06, "loss": 0.8609, "step": 877 }, { "epoch": 0.32, "learning_rate": 9.2558934676984e-06, "loss": 0.8213, "step": 878 }, { "epoch": 0.32, "learning_rate": 9.254993701637575e-06, "loss": 0.8943, "step": 879 }, { "epoch": 0.32, "learning_rate": 9.254093935576751e-06, "loss": 0.8077, "step": 880 }, { "epoch": 0.32, "learning_rate": 9.253194169515926e-06, "loss": 0.7804, "step": 881 }, { "epoch": 0.32, "learning_rate": 9.252294403455103e-06, "loss": 0.946, "step": 882 }, { "epoch": 0.32, "learning_rate": 9.25139463739428e-06, "loss": 0.9336, "step": 883 }, { "epoch": 0.32, "learning_rate": 9.250494871333454e-06, "loss": 0.8854, "step": 884 }, { "epoch": 0.32, "learning_rate": 9.24959510527263e-06, "loss": 0.9531, "step": 885 }, { "epoch": 0.32, "learning_rate": 9.248695339211806e-06, "loss": 0.7965, "step": 886 }, { "epoch": 0.32, "learning_rate": 9.24779557315098e-06, "loss": 0.6525, "step": 887 }, { "epoch": 0.32, "learning_rate": 9.246895807090157e-06, "loss": 0.8558, "step": 888 }, { "epoch": 0.32, "learning_rate": 9.245996041029334e-06, "loss": 0.7872, "step": 889 }, { "epoch": 0.32, "learning_rate": 9.245096274968509e-06, "loss": 0.7164, "step": 890 }, { "epoch": 0.32, "learning_rate": 9.244196508907684e-06, "loss": 0.8386, "step": 891 }, { "epoch": 0.32, "learning_rate": 9.24329674284686e-06, "loss": 0.8768, "step": 892 }, { "epoch": 0.32, "learning_rate": 9.242396976786037e-06, "loss": 0.9062, "step": 893 }, { "epoch": 0.32, "learning_rate": 9.241497210725212e-06, "loss": 0.9277, "step": 894 }, { "epoch": 0.32, "learning_rate": 9.240597444664387e-06, "loss": 0.746, "step": 895 }, { "epoch": 0.32, "learning_rate": 9.239697678603564e-06, "loss": 0.8367, "step": 896 }, { "epoch": 0.32, "learning_rate": 9.23879791254274e-06, "loss": 0.9632, "step": 897 }, { "epoch": 0.32, "learning_rate": 9.237898146481915e-06, "loss": 0.8204, "step": 898 }, { "epoch": 0.32, "learning_rate": 9.236998380421092e-06, "loss": 0.8705, "step": 899 }, { "epoch": 0.32, "learning_rate": 9.236098614360267e-06, "loss": 0.9951, "step": 900 }, { "epoch": 0.32, "learning_rate": 9.235198848299443e-06, "loss": 0.9927, "step": 901 }, { "epoch": 0.32, "learning_rate": 9.23429908223862e-06, "loss": 1.0241, "step": 902 }, { "epoch": 0.32, "learning_rate": 9.233399316177795e-06, "loss": 0.9341, "step": 903 }, { "epoch": 0.32, "learning_rate": 9.23249955011697e-06, "loss": 0.9192, "step": 904 }, { "epoch": 0.32, "learning_rate": 9.231599784056146e-06, "loss": 0.8454, "step": 905 }, { "epoch": 0.33, "learning_rate": 9.230700017995323e-06, "loss": 1.0115, "step": 906 }, { "epoch": 0.33, "learning_rate": 9.229800251934498e-06, "loss": 0.8403, "step": 907 }, { "epoch": 0.33, "learning_rate": 9.228900485873674e-06, "loss": 0.8213, "step": 908 }, { "epoch": 0.33, "learning_rate": 9.22800071981285e-06, "loss": 0.902, "step": 909 }, { "epoch": 0.33, "learning_rate": 9.227100953752024e-06, "loss": 0.8366, "step": 910 }, { "epoch": 0.33, "learning_rate": 9.226201187691201e-06, "loss": 0.8716, "step": 911 }, { "epoch": 0.33, "learning_rate": 9.225301421630378e-06, "loss": 0.9049, "step": 912 }, { "epoch": 0.33, "learning_rate": 9.224401655569552e-06, "loss": 0.7897, "step": 913 }, { "epoch": 0.33, "learning_rate": 9.223501889508727e-06, "loss": 0.8991, "step": 914 }, { "epoch": 0.33, "learning_rate": 9.222602123447904e-06, "loss": 0.9183, "step": 915 }, { "epoch": 0.33, "learning_rate": 9.22170235738708e-06, "loss": 0.9732, "step": 916 }, { "epoch": 0.33, "learning_rate": 9.220802591326256e-06, "loss": 0.9443, "step": 917 }, { "epoch": 0.33, "learning_rate": 9.219902825265432e-06, "loss": 0.9491, "step": 918 }, { "epoch": 0.33, "learning_rate": 9.219003059204607e-06, "loss": 0.8402, "step": 919 }, { "epoch": 0.33, "learning_rate": 9.218103293143784e-06, "loss": 0.8487, "step": 920 }, { "epoch": 0.33, "learning_rate": 9.21720352708296e-06, "loss": 0.7278, "step": 921 }, { "epoch": 0.33, "learning_rate": 9.216303761022135e-06, "loss": 0.7418, "step": 922 }, { "epoch": 0.33, "learning_rate": 9.21540399496131e-06, "loss": 0.7844, "step": 923 }, { "epoch": 0.33, "learning_rate": 9.214504228900487e-06, "loss": 0.8984, "step": 924 }, { "epoch": 0.33, "learning_rate": 9.213604462839663e-06, "loss": 0.8663, "step": 925 }, { "epoch": 0.33, "learning_rate": 9.212704696778838e-06, "loss": 0.8844, "step": 926 }, { "epoch": 0.33, "learning_rate": 9.211804930718013e-06, "loss": 0.7878, "step": 927 }, { "epoch": 0.33, "learning_rate": 9.21090516465719e-06, "loss": 0.7964, "step": 928 }, { "epoch": 0.33, "learning_rate": 9.210005398596365e-06, "loss": 0.942, "step": 929 }, { "epoch": 0.33, "learning_rate": 9.209105632535541e-06, "loss": 0.8735, "step": 930 }, { "epoch": 0.33, "learning_rate": 9.208205866474718e-06, "loss": 0.9036, "step": 931 }, { "epoch": 0.33, "learning_rate": 9.207306100413893e-06, "loss": 0.9002, "step": 932 }, { "epoch": 0.33, "learning_rate": 9.206406334353068e-06, "loss": 0.925, "step": 933 }, { "epoch": 0.34, "learning_rate": 9.205506568292244e-06, "loss": 0.9267, "step": 934 }, { "epoch": 0.34, "learning_rate": 9.204606802231421e-06, "loss": 0.9258, "step": 935 }, { "epoch": 0.34, "learning_rate": 9.203707036170596e-06, "loss": 0.7187, "step": 936 }, { "epoch": 0.34, "learning_rate": 9.202807270109773e-06, "loss": 0.8093, "step": 937 }, { "epoch": 0.34, "learning_rate": 9.201907504048948e-06, "loss": 0.7875, "step": 938 }, { "epoch": 0.34, "learning_rate": 9.201007737988124e-06, "loss": 0.8786, "step": 939 }, { "epoch": 0.34, "learning_rate": 9.2001079719273e-06, "loss": 0.738, "step": 940 }, { "epoch": 0.34, "learning_rate": 9.199208205866476e-06, "loss": 0.8706, "step": 941 }, { "epoch": 0.34, "learning_rate": 9.19830843980565e-06, "loss": 0.868, "step": 942 }, { "epoch": 0.34, "learning_rate": 9.197408673744827e-06, "loss": 0.8836, "step": 943 }, { "epoch": 0.34, "learning_rate": 9.196508907684004e-06, "loss": 0.8539, "step": 944 }, { "epoch": 0.34, "learning_rate": 9.195609141623179e-06, "loss": 0.8834, "step": 945 }, { "epoch": 0.34, "learning_rate": 9.194709375562354e-06, "loss": 0.9577, "step": 946 }, { "epoch": 0.34, "learning_rate": 9.19380960950153e-06, "loss": 0.7947, "step": 947 }, { "epoch": 0.34, "learning_rate": 9.192909843440705e-06, "loss": 0.7773, "step": 948 }, { "epoch": 0.34, "learning_rate": 9.192010077379882e-06, "loss": 1.0034, "step": 949 }, { "epoch": 0.34, "learning_rate": 9.191110311319058e-06, "loss": 0.8686, "step": 950 }, { "epoch": 0.34, "learning_rate": 9.190210545258233e-06, "loss": 0.8431, "step": 951 }, { "epoch": 0.34, "learning_rate": 9.189310779197408e-06, "loss": 0.9694, "step": 952 }, { "epoch": 0.34, "learning_rate": 9.188411013136585e-06, "loss": 0.8496, "step": 953 }, { "epoch": 0.34, "learning_rate": 9.187511247075762e-06, "loss": 0.7615, "step": 954 }, { "epoch": 0.34, "learning_rate": 9.186611481014936e-06, "loss": 0.8057, "step": 955 }, { "epoch": 0.34, "learning_rate": 9.185711714954113e-06, "loss": 0.8765, "step": 956 }, { "epoch": 0.34, "learning_rate": 9.184811948893288e-06, "loss": 0.908, "step": 957 }, { "epoch": 0.34, "learning_rate": 9.183912182832465e-06, "loss": 0.9319, "step": 958 }, { "epoch": 0.34, "learning_rate": 9.183012416771641e-06, "loss": 0.995, "step": 959 }, { "epoch": 0.34, "learning_rate": 9.182112650710816e-06, "loss": 0.879, "step": 960 }, { "epoch": 0.34, "learning_rate": 9.181212884649991e-06, "loss": 0.9644, "step": 961 }, { "epoch": 0.35, "learning_rate": 9.180313118589168e-06, "loss": 0.81, "step": 962 }, { "epoch": 0.35, "learning_rate": 9.179413352528344e-06, "loss": 0.8442, "step": 963 }, { "epoch": 0.35, "learning_rate": 9.17851358646752e-06, "loss": 0.8079, "step": 964 }, { "epoch": 0.35, "learning_rate": 9.177613820406694e-06, "loss": 0.8559, "step": 965 }, { "epoch": 0.35, "learning_rate": 9.17671405434587e-06, "loss": 0.8594, "step": 966 }, { "epoch": 0.35, "learning_rate": 9.175814288285046e-06, "loss": 0.8269, "step": 967 }, { "epoch": 0.35, "learning_rate": 9.174914522224222e-06, "loss": 0.9453, "step": 968 }, { "epoch": 0.35, "learning_rate": 9.174014756163399e-06, "loss": 0.9154, "step": 969 }, { "epoch": 0.35, "learning_rate": 9.173114990102574e-06, "loss": 0.9562, "step": 970 }, { "epoch": 0.35, "learning_rate": 9.172215224041749e-06, "loss": 0.8424, "step": 971 }, { "epoch": 0.35, "learning_rate": 9.171315457980925e-06, "loss": 0.9195, "step": 972 }, { "epoch": 0.35, "learning_rate": 9.170415691920102e-06, "loss": 0.9962, "step": 973 }, { "epoch": 0.35, "learning_rate": 9.169515925859277e-06, "loss": 0.8302, "step": 974 }, { "epoch": 0.35, "learning_rate": 9.168616159798454e-06, "loss": 0.9252, "step": 975 }, { "epoch": 0.35, "learning_rate": 9.167716393737628e-06, "loss": 0.8615, "step": 976 }, { "epoch": 0.35, "learning_rate": 9.166816627676805e-06, "loss": 0.8175, "step": 977 }, { "epoch": 0.35, "learning_rate": 9.165916861615982e-06, "loss": 0.9403, "step": 978 }, { "epoch": 0.35, "learning_rate": 9.165017095555157e-06, "loss": 0.7958, "step": 979 }, { "epoch": 0.35, "learning_rate": 9.164117329494332e-06, "loss": 0.8983, "step": 980 }, { "epoch": 0.35, "learning_rate": 9.163217563433508e-06, "loss": 0.8556, "step": 981 }, { "epoch": 0.35, "learning_rate": 9.162317797372685e-06, "loss": 0.9081, "step": 982 }, { "epoch": 0.35, "learning_rate": 9.16141803131186e-06, "loss": 0.9326, "step": 983 }, { "epoch": 0.35, "learning_rate": 9.160518265251035e-06, "loss": 0.7999, "step": 984 }, { "epoch": 0.35, "learning_rate": 9.159618499190211e-06, "loss": 0.7849, "step": 985 }, { "epoch": 0.35, "learning_rate": 9.158718733129386e-06, "loss": 0.9192, "step": 986 }, { "epoch": 0.35, "learning_rate": 9.157818967068563e-06, "loss": 0.865, "step": 987 }, { "epoch": 0.35, "learning_rate": 9.15691920100774e-06, "loss": 0.8032, "step": 988 }, { "epoch": 0.35, "learning_rate": 9.156019434946914e-06, "loss": 0.852, "step": 989 }, { "epoch": 0.36, "learning_rate": 9.15511966888609e-06, "loss": 0.8455, "step": 990 }, { "epoch": 0.36, "learning_rate": 9.154219902825266e-06, "loss": 0.9233, "step": 991 }, { "epoch": 0.36, "learning_rate": 9.153320136764442e-06, "loss": 0.8658, "step": 992 }, { "epoch": 0.36, "learning_rate": 9.152420370703617e-06, "loss": 0.9847, "step": 993 }, { "epoch": 0.36, "learning_rate": 9.151520604642794e-06, "loss": 0.7753, "step": 994 }, { "epoch": 0.36, "learning_rate": 9.150620838581969e-06, "loss": 0.9086, "step": 995 }, { "epoch": 0.36, "learning_rate": 9.149721072521146e-06, "loss": 0.9302, "step": 996 }, { "epoch": 0.36, "learning_rate": 9.148821306460322e-06, "loss": 0.8195, "step": 997 }, { "epoch": 0.36, "learning_rate": 9.147921540399497e-06, "loss": 0.9282, "step": 998 }, { "epoch": 0.36, "learning_rate": 9.147021774338672e-06, "loss": 0.9205, "step": 999 }, { "epoch": 0.36, "learning_rate": 9.146122008277849e-06, "loss": 0.8296, "step": 1000 }, { "epoch": 0.36, "learning_rate": 9.145222242217025e-06, "loss": 0.8174, "step": 1001 }, { "epoch": 0.36, "learning_rate": 9.1443224761562e-06, "loss": 0.909, "step": 1002 }, { "epoch": 0.36, "learning_rate": 9.143422710095375e-06, "loss": 0.9243, "step": 1003 }, { "epoch": 0.36, "learning_rate": 9.142522944034552e-06, "loss": 0.8328, "step": 1004 }, { "epoch": 0.36, "learning_rate": 9.141623177973728e-06, "loss": 1.0019, "step": 1005 }, { "epoch": 0.36, "learning_rate": 9.140723411912903e-06, "loss": 0.8714, "step": 1006 }, { "epoch": 0.36, "learning_rate": 9.13982364585208e-06, "loss": 0.9782, "step": 1007 }, { "epoch": 0.36, "learning_rate": 9.138923879791255e-06, "loss": 0.8398, "step": 1008 }, { "epoch": 0.36, "learning_rate": 9.13802411373043e-06, "loss": 0.8996, "step": 1009 }, { "epoch": 0.36, "learning_rate": 9.137124347669606e-06, "loss": 0.9346, "step": 1010 }, { "epoch": 0.36, "learning_rate": 9.136224581608783e-06, "loss": 0.8173, "step": 1011 }, { "epoch": 0.36, "learning_rate": 9.135324815547958e-06, "loss": 0.9647, "step": 1012 }, { "epoch": 0.36, "learning_rate": 9.134425049487134e-06, "loss": 0.8593, "step": 1013 }, { "epoch": 0.36, "learning_rate": 9.13352528342631e-06, "loss": 0.8614, "step": 1014 }, { "epoch": 0.36, "learning_rate": 9.132625517365486e-06, "loss": 0.8722, "step": 1015 }, { "epoch": 0.36, "learning_rate": 9.131725751304663e-06, "loss": 0.836, "step": 1016 }, { "epoch": 0.37, "learning_rate": 9.130825985243838e-06, "loss": 0.7893, "step": 1017 }, { "epoch": 0.37, "learning_rate": 9.129926219183012e-06, "loss": 0.8181, "step": 1018 }, { "epoch": 0.37, "learning_rate": 9.129026453122189e-06, "loss": 0.8235, "step": 1019 }, { "epoch": 0.37, "learning_rate": 9.128126687061366e-06, "loss": 0.8551, "step": 1020 }, { "epoch": 0.37, "learning_rate": 9.12722692100054e-06, "loss": 0.7658, "step": 1021 }, { "epoch": 0.37, "learning_rate": 9.126327154939716e-06, "loss": 0.8273, "step": 1022 }, { "epoch": 0.37, "learning_rate": 9.125427388878892e-06, "loss": 0.8786, "step": 1023 }, { "epoch": 0.37, "learning_rate": 9.124527622818069e-06, "loss": 0.7893, "step": 1024 }, { "epoch": 0.37, "learning_rate": 9.123627856757244e-06, "loss": 0.816, "step": 1025 }, { "epoch": 0.37, "learning_rate": 9.12272809069642e-06, "loss": 0.8623, "step": 1026 }, { "epoch": 0.37, "learning_rate": 9.121828324635595e-06, "loss": 0.7635, "step": 1027 }, { "epoch": 0.37, "learning_rate": 9.12092855857477e-06, "loss": 0.9245, "step": 1028 }, { "epoch": 0.37, "learning_rate": 9.120028792513947e-06, "loss": 0.7848, "step": 1029 }, { "epoch": 0.37, "learning_rate": 9.119129026453123e-06, "loss": 0.7805, "step": 1030 }, { "epoch": 0.37, "learning_rate": 9.118229260392298e-06, "loss": 0.8868, "step": 1031 }, { "epoch": 0.37, "learning_rate": 9.117329494331475e-06, "loss": 0.9378, "step": 1032 }, { "epoch": 0.37, "learning_rate": 9.11642972827065e-06, "loss": 0.7751, "step": 1033 }, { "epoch": 0.37, "learning_rate": 9.115529962209826e-06, "loss": 1.0112, "step": 1034 }, { "epoch": 0.37, "learning_rate": 9.114630196149003e-06, "loss": 0.764, "step": 1035 }, { "epoch": 0.37, "learning_rate": 9.113730430088178e-06, "loss": 0.8423, "step": 1036 }, { "epoch": 0.37, "learning_rate": 9.112830664027353e-06, "loss": 0.8973, "step": 1037 }, { "epoch": 0.37, "learning_rate": 9.11193089796653e-06, "loss": 0.8448, "step": 1038 }, { "epoch": 0.37, "learning_rate": 9.111031131905706e-06, "loss": 0.858, "step": 1039 }, { "epoch": 0.37, "learning_rate": 9.110131365844881e-06, "loss": 0.7963, "step": 1040 }, { "epoch": 0.37, "learning_rate": 9.109231599784056e-06, "loss": 0.7535, "step": 1041 }, { "epoch": 0.37, "learning_rate": 9.108331833723233e-06, "loss": 0.7531, "step": 1042 }, { "epoch": 0.37, "learning_rate": 9.10743206766241e-06, "loss": 0.8474, "step": 1043 }, { "epoch": 0.37, "learning_rate": 9.106532301601584e-06, "loss": 0.8648, "step": 1044 }, { "epoch": 0.38, "learning_rate": 9.10563253554076e-06, "loss": 0.8875, "step": 1045 }, { "epoch": 0.38, "learning_rate": 9.104732769479936e-06, "loss": 0.8813, "step": 1046 }, { "epoch": 0.38, "learning_rate": 9.10383300341911e-06, "loss": 0.9325, "step": 1047 }, { "epoch": 0.38, "learning_rate": 9.102933237358287e-06, "loss": 0.8131, "step": 1048 }, { "epoch": 0.38, "learning_rate": 9.102033471297464e-06, "loss": 0.8015, "step": 1049 }, { "epoch": 0.38, "learning_rate": 9.101133705236639e-06, "loss": 1.0626, "step": 1050 }, { "epoch": 0.38, "learning_rate": 9.100233939175815e-06, "loss": 0.8417, "step": 1051 }, { "epoch": 0.38, "learning_rate": 9.09933417311499e-06, "loss": 0.8926, "step": 1052 }, { "epoch": 0.38, "learning_rate": 9.098434407054167e-06, "loss": 0.9191, "step": 1053 }, { "epoch": 0.38, "learning_rate": 9.097534640993344e-06, "loss": 0.8912, "step": 1054 }, { "epoch": 0.38, "learning_rate": 9.096634874932518e-06, "loss": 0.9867, "step": 1055 }, { "epoch": 0.38, "learning_rate": 9.095735108871693e-06, "loss": 0.9757, "step": 1056 }, { "epoch": 0.38, "learning_rate": 9.09483534281087e-06, "loss": 0.7209, "step": 1057 }, { "epoch": 0.38, "learning_rate": 9.093935576750047e-06, "loss": 0.7514, "step": 1058 }, { "epoch": 0.38, "learning_rate": 9.093035810689221e-06, "loss": 0.8688, "step": 1059 }, { "epoch": 0.38, "learning_rate": 9.092136044628396e-06, "loss": 0.7955, "step": 1060 }, { "epoch": 0.38, "learning_rate": 9.091236278567573e-06, "loss": 0.86, "step": 1061 }, { "epoch": 0.38, "learning_rate": 9.09033651250675e-06, "loss": 0.8979, "step": 1062 }, { "epoch": 0.38, "learning_rate": 9.089436746445925e-06, "loss": 0.8694, "step": 1063 }, { "epoch": 0.38, "learning_rate": 9.088536980385101e-06, "loss": 0.8853, "step": 1064 }, { "epoch": 0.38, "learning_rate": 9.087637214324276e-06, "loss": 0.8503, "step": 1065 }, { "epoch": 0.38, "learning_rate": 9.086737448263451e-06, "loss": 0.7916, "step": 1066 }, { "epoch": 0.38, "learning_rate": 9.085837682202628e-06, "loss": 0.7943, "step": 1067 }, { "epoch": 0.38, "learning_rate": 9.084937916141804e-06, "loss": 0.8383, "step": 1068 }, { "epoch": 0.38, "learning_rate": 9.08403815008098e-06, "loss": 0.8002, "step": 1069 }, { "epoch": 0.38, "learning_rate": 9.083138384020156e-06, "loss": 0.7359, "step": 1070 }, { "epoch": 0.38, "learning_rate": 9.08223861795933e-06, "loss": 0.782, "step": 1071 }, { "epoch": 0.38, "learning_rate": 9.081338851898507e-06, "loss": 0.8878, "step": 1072 }, { "epoch": 0.39, "learning_rate": 9.080439085837684e-06, "loss": 0.8735, "step": 1073 }, { "epoch": 0.39, "learning_rate": 9.079539319776859e-06, "loss": 0.7656, "step": 1074 }, { "epoch": 0.39, "learning_rate": 9.078639553716034e-06, "loss": 0.7825, "step": 1075 }, { "epoch": 0.39, "learning_rate": 9.07773978765521e-06, "loss": 0.8999, "step": 1076 }, { "epoch": 0.39, "learning_rate": 9.076840021594387e-06, "loss": 0.7945, "step": 1077 }, { "epoch": 0.39, "learning_rate": 9.075940255533562e-06, "loss": 0.9638, "step": 1078 }, { "epoch": 0.39, "learning_rate": 9.075040489472737e-06, "loss": 0.9093, "step": 1079 }, { "epoch": 0.39, "learning_rate": 9.074140723411913e-06, "loss": 0.901, "step": 1080 }, { "epoch": 0.39, "learning_rate": 9.074140723411913e-06, "loss": 0.9001, "step": 1081 }, { "epoch": 0.39, "learning_rate": 9.07324095735109e-06, "loss": 0.7476, "step": 1082 }, { "epoch": 0.39, "learning_rate": 9.072341191290265e-06, "loss": 0.9258, "step": 1083 }, { "epoch": 0.39, "learning_rate": 9.071441425229442e-06, "loss": 0.8131, "step": 1084 }, { "epoch": 0.39, "learning_rate": 9.070541659168617e-06, "loss": 0.9515, "step": 1085 }, { "epoch": 0.39, "learning_rate": 9.069641893107791e-06, "loss": 0.9512, "step": 1086 }, { "epoch": 0.39, "learning_rate": 9.068742127046968e-06, "loss": 0.8956, "step": 1087 }, { "epoch": 0.39, "learning_rate": 9.067842360986145e-06, "loss": 1.0106, "step": 1088 }, { "epoch": 0.39, "learning_rate": 9.06694259492532e-06, "loss": 0.8698, "step": 1089 }, { "epoch": 0.39, "learning_rate": 9.066042828864496e-06, "loss": 0.7977, "step": 1090 }, { "epoch": 0.39, "learning_rate": 9.065143062803671e-06, "loss": 0.7914, "step": 1091 }, { "epoch": 0.39, "learning_rate": 9.064243296742848e-06, "loss": 0.8453, "step": 1092 }, { "epoch": 0.39, "learning_rate": 9.063343530682024e-06, "loss": 0.8006, "step": 1093 }, { "epoch": 0.39, "learning_rate": 9.0624437646212e-06, "loss": 0.8797, "step": 1094 }, { "epoch": 0.39, "learning_rate": 9.061543998560374e-06, "loss": 1.0731, "step": 1095 }, { "epoch": 0.39, "learning_rate": 9.060644232499551e-06, "loss": 0.8861, "step": 1096 }, { "epoch": 0.39, "learning_rate": 9.059744466438727e-06, "loss": 0.8763, "step": 1097 }, { "epoch": 0.39, "learning_rate": 9.058844700377902e-06, "loss": 0.7874, "step": 1098 }, { "epoch": 0.39, "learning_rate": 9.057944934317077e-06, "loss": 0.937, "step": 1099 }, { "epoch": 0.39, "learning_rate": 9.057045168256254e-06, "loss": 0.8221, "step": 1100 }, { "epoch": 0.4, "learning_rate": 9.05614540219543e-06, "loss": 0.9266, "step": 1101 }, { "epoch": 0.4, "learning_rate": 9.055245636134605e-06, "loss": 1.013, "step": 1102 }, { "epoch": 0.4, "learning_rate": 9.054345870073782e-06, "loss": 0.9172, "step": 1103 }, { "epoch": 0.4, "learning_rate": 9.053446104012957e-06, "loss": 0.8868, "step": 1104 }, { "epoch": 0.4, "learning_rate": 9.052546337952132e-06, "loss": 0.8098, "step": 1105 }, { "epoch": 0.4, "learning_rate": 9.051646571891309e-06, "loss": 0.9195, "step": 1106 }, { "epoch": 0.4, "learning_rate": 9.050746805830485e-06, "loss": 0.8094, "step": 1107 }, { "epoch": 0.4, "learning_rate": 9.04984703976966e-06, "loss": 0.8791, "step": 1108 }, { "epoch": 0.4, "learning_rate": 9.048947273708837e-06, "loss": 0.8414, "step": 1109 }, { "epoch": 0.4, "learning_rate": 9.048047507648012e-06, "loss": 0.9324, "step": 1110 }, { "epoch": 0.4, "learning_rate": 9.047147741587188e-06, "loss": 0.9061, "step": 1111 }, { "epoch": 0.4, "learning_rate": 9.046247975526365e-06, "loss": 0.8187, "step": 1112 }, { "epoch": 0.4, "learning_rate": 9.04534820946554e-06, "loss": 0.9285, "step": 1113 }, { "epoch": 0.4, "learning_rate": 9.044448443404715e-06, "loss": 0.8043, "step": 1114 }, { "epoch": 0.4, "learning_rate": 9.043548677343891e-06, "loss": 0.8752, "step": 1115 }, { "epoch": 0.4, "learning_rate": 9.043548677343891e-06, "loss": 0.8065, "step": 1116 }, { "epoch": 0.4, "learning_rate": 9.042648911283068e-06, "loss": 0.8333, "step": 1117 }, { "epoch": 0.4, "learning_rate": 9.041749145222243e-06, "loss": 0.8815, "step": 1118 }, { "epoch": 0.4, "learning_rate": 9.040849379161418e-06, "loss": 0.8275, "step": 1119 }, { "epoch": 0.4, "learning_rate": 9.039949613100594e-06, "loss": 0.9233, "step": 1120 }, { "epoch": 0.4, "learning_rate": 9.039049847039771e-06, "loss": 0.9166, "step": 1121 }, { "epoch": 0.4, "learning_rate": 9.038150080978946e-06, "loss": 0.8574, "step": 1122 }, { "epoch": 0.4, "learning_rate": 9.037250314918123e-06, "loss": 0.899, "step": 1123 }, { "epoch": 0.4, "learning_rate": 9.036350548857297e-06, "loss": 0.8772, "step": 1124 }, { "epoch": 0.4, "learning_rate": 9.035450782796474e-06, "loss": 0.8422, "step": 1125 }, { "epoch": 0.4, "learning_rate": 9.034551016735649e-06, "loss": 0.7543, "step": 1126 }, { "epoch": 0.4, "learning_rate": 9.033651250674826e-06, "loss": 0.8422, "step": 1127 }, { "epoch": 0.4, "learning_rate": 9.032751484614e-06, "loss": 0.7521, "step": 1128 }, { "epoch": 0.41, "learning_rate": 9.031851718553177e-06, "loss": 0.7911, "step": 1129 }, { "epoch": 0.41, "learning_rate": 9.030951952492352e-06, "loss": 0.8765, "step": 1130 }, { "epoch": 0.41, "learning_rate": 9.030052186431529e-06, "loss": 0.9065, "step": 1131 }, { "epoch": 0.41, "learning_rate": 9.029152420370705e-06, "loss": 0.7167, "step": 1132 }, { "epoch": 0.41, "learning_rate": 9.02825265430988e-06, "loss": 0.8854, "step": 1133 }, { "epoch": 0.41, "learning_rate": 9.027352888249055e-06, "loss": 0.8526, "step": 1134 }, { "epoch": 0.41, "learning_rate": 9.026453122188232e-06, "loss": 0.9218, "step": 1135 }, { "epoch": 0.41, "learning_rate": 9.025553356127408e-06, "loss": 0.8195, "step": 1136 }, { "epoch": 0.41, "learning_rate": 9.024653590066583e-06, "loss": 0.8914, "step": 1137 }, { "epoch": 0.41, "learning_rate": 9.023753824005758e-06, "loss": 1.0163, "step": 1138 }, { "epoch": 0.41, "learning_rate": 9.022854057944935e-06, "loss": 0.7644, "step": 1139 }, { "epoch": 0.41, "learning_rate": 9.021954291884111e-06, "loss": 1.0169, "step": 1140 }, { "epoch": 0.41, "learning_rate": 9.021954291884111e-06, "loss": 0.9409, "step": 1141 }, { "epoch": 0.41, "learning_rate": 9.021054525823286e-06, "loss": 1.0454, "step": 1142 }, { "epoch": 0.41, "learning_rate": 9.020154759762463e-06, "loss": 0.8711, "step": 1143 }, { "epoch": 0.41, "learning_rate": 9.019254993701638e-06, "loss": 0.7785, "step": 1144 }, { "epoch": 0.41, "learning_rate": 9.018355227640815e-06, "loss": 0.7983, "step": 1145 }, { "epoch": 0.41, "learning_rate": 9.01745546157999e-06, "loss": 0.9214, "step": 1146 }, { "epoch": 0.41, "learning_rate": 9.016555695519166e-06, "loss": 0.7772, "step": 1147 }, { "epoch": 0.41, "learning_rate": 9.015655929458341e-06, "loss": 0.8317, "step": 1148 }, { "epoch": 0.41, "learning_rate": 9.014756163397518e-06, "loss": 0.7665, "step": 1149 }, { "epoch": 0.41, "learning_rate": 9.013856397336693e-06, "loss": 0.8095, "step": 1150 }, { "epoch": 0.41, "learning_rate": 9.01295663127587e-06, "loss": 0.8554, "step": 1151 }, { "epoch": 0.41, "learning_rate": 9.012056865215046e-06, "loss": 0.8337, "step": 1152 }, { "epoch": 0.41, "learning_rate": 9.01115709915422e-06, "loss": 0.8288, "step": 1153 }, { "epoch": 0.41, "learning_rate": 9.010257333093396e-06, "loss": 0.8269, "step": 1154 }, { "epoch": 0.41, "learning_rate": 9.009357567032572e-06, "loss": 0.8555, "step": 1155 }, { "epoch": 0.41, "learning_rate": 9.008457800971749e-06, "loss": 0.8011, "step": 1156 }, { "epoch": 0.42, "learning_rate": 9.007558034910924e-06, "loss": 0.9373, "step": 1157 }, { "epoch": 0.42, "learning_rate": 9.006658268850099e-06, "loss": 0.8656, "step": 1158 }, { "epoch": 0.42, "learning_rate": 9.005758502789275e-06, "loss": 0.8593, "step": 1159 }, { "epoch": 0.42, "learning_rate": 9.004858736728452e-06, "loss": 0.8077, "step": 1160 }, { "epoch": 0.42, "learning_rate": 9.003958970667627e-06, "loss": 0.7889, "step": 1161 }, { "epoch": 0.42, "learning_rate": 9.003059204606803e-06, "loss": 0.9272, "step": 1162 }, { "epoch": 0.42, "learning_rate": 9.002159438545978e-06, "loss": 0.9774, "step": 1163 }, { "epoch": 0.42, "learning_rate": 9.001259672485155e-06, "loss": 0.9224, "step": 1164 }, { "epoch": 0.42, "learning_rate": 9.000359906424332e-06, "loss": 0.847, "step": 1165 }, { "epoch": 0.42, "learning_rate": 8.999460140363507e-06, "loss": 0.9381, "step": 1166 }, { "epoch": 0.42, "learning_rate": 8.998560374302681e-06, "loss": 0.8371, "step": 1167 }, { "epoch": 0.42, "learning_rate": 8.997660608241858e-06, "loss": 0.8479, "step": 1168 }, { "epoch": 0.42, "learning_rate": 8.996760842181033e-06, "loss": 0.8147, "step": 1169 }, { "epoch": 0.42, "learning_rate": 8.99586107612021e-06, "loss": 0.8741, "step": 1170 }, { "epoch": 0.42, "learning_rate": 8.994961310059386e-06, "loss": 0.9339, "step": 1171 }, { "epoch": 0.42, "learning_rate": 8.994061543998561e-06, "loss": 0.9206, "step": 1172 }, { "epoch": 0.42, "learning_rate": 8.993161777937736e-06, "loss": 0.8792, "step": 1173 }, { "epoch": 0.42, "learning_rate": 8.992262011876913e-06, "loss": 0.8023, "step": 1174 }, { "epoch": 0.42, "learning_rate": 8.99136224581609e-06, "loss": 0.9065, "step": 1175 }, { "epoch": 0.42, "learning_rate": 8.990462479755264e-06, "loss": 0.8644, "step": 1176 }, { "epoch": 0.42, "learning_rate": 8.989562713694439e-06, "loss": 0.9152, "step": 1177 }, { "epoch": 0.42, "learning_rate": 8.988662947633616e-06, "loss": 0.7606, "step": 1178 }, { "epoch": 0.42, "learning_rate": 8.987763181572792e-06, "loss": 0.9029, "step": 1179 }, { "epoch": 0.42, "learning_rate": 8.986863415511967e-06, "loss": 1.0094, "step": 1180 }, { "epoch": 0.42, "learning_rate": 8.985963649451144e-06, "loss": 0.8028, "step": 1181 }, { "epoch": 0.42, "learning_rate": 8.985063883390319e-06, "loss": 0.7278, "step": 1182 }, { "epoch": 0.42, "learning_rate": 8.984164117329495e-06, "loss": 0.903, "step": 1183 }, { "epoch": 0.42, "learning_rate": 8.983264351268672e-06, "loss": 0.8305, "step": 1184 }, { "epoch": 0.43, "learning_rate": 8.982364585207847e-06, "loss": 0.8592, "step": 1185 }, { "epoch": 0.43, "learning_rate": 8.981464819147022e-06, "loss": 0.9052, "step": 1186 }, { "epoch": 0.43, "learning_rate": 8.980565053086199e-06, "loss": 0.8416, "step": 1187 }, { "epoch": 0.43, "learning_rate": 8.979665287025373e-06, "loss": 0.8879, "step": 1188 }, { "epoch": 0.43, "learning_rate": 8.97876552096455e-06, "loss": 0.9284, "step": 1189 }, { "epoch": 0.43, "learning_rate": 8.977865754903727e-06, "loss": 0.9194, "step": 1190 }, { "epoch": 0.43, "learning_rate": 8.976965988842902e-06, "loss": 0.9031, "step": 1191 }, { "epoch": 0.43, "learning_rate": 8.976066222782077e-06, "loss": 0.9119, "step": 1192 }, { "epoch": 0.43, "learning_rate": 8.975166456721253e-06, "loss": 0.9384, "step": 1193 }, { "epoch": 0.43, "learning_rate": 8.97426669066043e-06, "loss": 0.8385, "step": 1194 }, { "epoch": 0.43, "learning_rate": 8.973366924599605e-06, "loss": 0.8628, "step": 1195 }, { "epoch": 0.43, "learning_rate": 8.97246715853878e-06, "loss": 0.8761, "step": 1196 }, { "epoch": 0.43, "learning_rate": 8.971567392477956e-06, "loss": 0.8688, "step": 1197 }, { "epoch": 0.43, "learning_rate": 8.970667626417133e-06, "loss": 0.9173, "step": 1198 }, { "epoch": 0.43, "learning_rate": 8.969767860356308e-06, "loss": 0.9741, "step": 1199 }, { "epoch": 0.43, "learning_rate": 8.968868094295484e-06, "loss": 0.8108, "step": 1200 }, { "epoch": 0.43, "learning_rate": 8.96796832823466e-06, "loss": 0.8126, "step": 1201 }, { "epoch": 0.43, "learning_rate": 8.967068562173836e-06, "loss": 0.7308, "step": 1202 }, { "epoch": 0.43, "learning_rate": 8.966168796113013e-06, "loss": 0.8353, "step": 1203 }, { "epoch": 0.43, "learning_rate": 8.965269030052187e-06, "loss": 0.9967, "step": 1204 }, { "epoch": 0.43, "learning_rate": 8.964369263991362e-06, "loss": 0.8997, "step": 1205 }, { "epoch": 0.43, "learning_rate": 8.963469497930539e-06, "loss": 0.8693, "step": 1206 }, { "epoch": 0.43, "learning_rate": 8.962569731869714e-06, "loss": 0.7784, "step": 1207 }, { "epoch": 0.43, "learning_rate": 8.96166996580889e-06, "loss": 0.924, "step": 1208 }, { "epoch": 0.43, "learning_rate": 8.960770199748067e-06, "loss": 0.8224, "step": 1209 }, { "epoch": 0.43, "learning_rate": 8.959870433687242e-06, "loss": 0.8757, "step": 1210 }, { "epoch": 0.43, "learning_rate": 8.958970667626417e-06, "loss": 0.8862, "step": 1211 }, { "epoch": 0.44, "learning_rate": 8.958070901565594e-06, "loss": 0.8052, "step": 1212 }, { "epoch": 0.44, "learning_rate": 8.95717113550477e-06, "loss": 0.7947, "step": 1213 }, { "epoch": 0.44, "learning_rate": 8.956271369443945e-06, "loss": 0.9462, "step": 1214 }, { "epoch": 0.44, "learning_rate": 8.95537160338312e-06, "loss": 0.8557, "step": 1215 }, { "epoch": 0.44, "learning_rate": 8.954471837322297e-06, "loss": 0.8606, "step": 1216 }, { "epoch": 0.44, "learning_rate": 8.953572071261473e-06, "loss": 0.9165, "step": 1217 }, { "epoch": 0.44, "learning_rate": 8.952672305200648e-06, "loss": 0.7568, "step": 1218 }, { "epoch": 0.44, "learning_rate": 8.951772539139825e-06, "loss": 0.7378, "step": 1219 }, { "epoch": 0.44, "learning_rate": 8.950872773079e-06, "loss": 0.7232, "step": 1220 }, { "epoch": 0.44, "learning_rate": 8.949973007018176e-06, "loss": 0.8643, "step": 1221 }, { "epoch": 0.44, "learning_rate": 8.949073240957353e-06, "loss": 0.8101, "step": 1222 }, { "epoch": 0.44, "learning_rate": 8.948173474896528e-06, "loss": 0.8839, "step": 1223 }, { "epoch": 0.44, "learning_rate": 8.947273708835703e-06, "loss": 0.7682, "step": 1224 }, { "epoch": 0.44, "learning_rate": 8.94637394277488e-06, "loss": 0.889, "step": 1225 }, { "epoch": 0.44, "learning_rate": 8.945474176714054e-06, "loss": 0.8003, "step": 1226 }, { "epoch": 0.44, "learning_rate": 8.944574410653231e-06, "loss": 0.9435, "step": 1227 }, { "epoch": 0.44, "learning_rate": 8.943674644592408e-06, "loss": 0.8724, "step": 1228 }, { "epoch": 0.44, "learning_rate": 8.942774878531583e-06, "loss": 0.8532, "step": 1229 }, { "epoch": 0.44, "learning_rate": 8.941875112470757e-06, "loss": 0.7726, "step": 1230 }, { "epoch": 0.44, "learning_rate": 8.940975346409934e-06, "loss": 0.8467, "step": 1231 }, { "epoch": 0.44, "learning_rate": 8.94007558034911e-06, "loss": 0.8299, "step": 1232 }, { "epoch": 0.44, "learning_rate": 8.939175814288286e-06, "loss": 0.9158, "step": 1233 }, { "epoch": 0.44, "learning_rate": 8.93827604822746e-06, "loss": 0.8658, "step": 1234 }, { "epoch": 0.44, "learning_rate": 8.937376282166637e-06, "loss": 0.9555, "step": 1235 }, { "epoch": 0.44, "learning_rate": 8.936476516105814e-06, "loss": 0.8343, "step": 1236 }, { "epoch": 0.44, "learning_rate": 8.935576750044989e-06, "loss": 0.9011, "step": 1237 }, { "epoch": 0.44, "learning_rate": 8.934676983984165e-06, "loss": 0.9912, "step": 1238 }, { "epoch": 0.44, "learning_rate": 8.93377721792334e-06, "loss": 0.7736, "step": 1239 }, { "epoch": 0.45, "learning_rate": 8.932877451862517e-06, "loss": 0.8083, "step": 1240 }, { "epoch": 0.45, "learning_rate": 8.931977685801693e-06, "loss": 0.7783, "step": 1241 }, { "epoch": 0.45, "learning_rate": 8.931077919740868e-06, "loss": 0.8582, "step": 1242 }, { "epoch": 0.45, "learning_rate": 8.930178153680043e-06, "loss": 0.8827, "step": 1243 }, { "epoch": 0.45, "learning_rate": 8.92927838761922e-06, "loss": 0.793, "step": 1244 }, { "epoch": 0.45, "learning_rate": 8.928378621558395e-06, "loss": 0.8482, "step": 1245 }, { "epoch": 0.45, "learning_rate": 8.927478855497571e-06, "loss": 0.8193, "step": 1246 }, { "epoch": 0.45, "learning_rate": 8.926579089436748e-06, "loss": 0.7974, "step": 1247 }, { "epoch": 0.45, "learning_rate": 8.925679323375923e-06, "loss": 0.8079, "step": 1248 }, { "epoch": 0.45, "learning_rate": 8.924779557315098e-06, "loss": 0.8475, "step": 1249 }, { "epoch": 0.45, "learning_rate": 8.923879791254275e-06, "loss": 0.8437, "step": 1250 }, { "epoch": 0.45, "learning_rate": 8.922980025193451e-06, "loss": 0.8481, "step": 1251 }, { "epoch": 0.45, "learning_rate": 8.922080259132626e-06, "loss": 0.9001, "step": 1252 }, { "epoch": 0.45, "learning_rate": 8.921180493071801e-06, "loss": 0.8862, "step": 1253 }, { "epoch": 0.45, "learning_rate": 8.920280727010978e-06, "loss": 0.9043, "step": 1254 }, { "epoch": 0.45, "learning_rate": 8.919380960950154e-06, "loss": 0.8871, "step": 1255 }, { "epoch": 0.45, "learning_rate": 8.918481194889329e-06, "loss": 0.8414, "step": 1256 }, { "epoch": 0.45, "learning_rate": 8.917581428828506e-06, "loss": 0.8931, "step": 1257 }, { "epoch": 0.45, "learning_rate": 8.91668166276768e-06, "loss": 1.0096, "step": 1258 }, { "epoch": 0.45, "learning_rate": 8.915781896706857e-06, "loss": 0.8563, "step": 1259 }, { "epoch": 0.45, "learning_rate": 8.914882130646034e-06, "loss": 0.8366, "step": 1260 }, { "epoch": 0.45, "learning_rate": 8.913982364585209e-06, "loss": 0.8595, "step": 1261 }, { "epoch": 0.45, "learning_rate": 8.913082598524384e-06, "loss": 0.778, "step": 1262 }, { "epoch": 0.45, "learning_rate": 8.91218283246356e-06, "loss": 0.7151, "step": 1263 }, { "epoch": 0.45, "learning_rate": 8.911283066402735e-06, "loss": 0.9124, "step": 1264 }, { "epoch": 0.45, "learning_rate": 8.910383300341912e-06, "loss": 0.8342, "step": 1265 }, { "epoch": 0.45, "learning_rate": 8.909483534281088e-06, "loss": 0.8633, "step": 1266 }, { "epoch": 0.45, "learning_rate": 8.908583768220263e-06, "loss": 0.8121, "step": 1267 }, { "epoch": 0.46, "learning_rate": 8.907684002159438e-06, "loss": 0.8337, "step": 1268 }, { "epoch": 0.46, "learning_rate": 8.906784236098615e-06, "loss": 0.8545, "step": 1269 }, { "epoch": 0.46, "learning_rate": 8.905884470037792e-06, "loss": 0.943, "step": 1270 }, { "epoch": 0.46, "learning_rate": 8.904984703976966e-06, "loss": 0.786, "step": 1271 }, { "epoch": 0.46, "learning_rate": 8.904084937916141e-06, "loss": 0.898, "step": 1272 }, { "epoch": 0.46, "learning_rate": 8.903185171855318e-06, "loss": 0.8463, "step": 1273 }, { "epoch": 0.46, "learning_rate": 8.902285405794495e-06, "loss": 0.878, "step": 1274 }, { "epoch": 0.46, "learning_rate": 8.90138563973367e-06, "loss": 0.9987, "step": 1275 }, { "epoch": 0.46, "learning_rate": 8.900485873672846e-06, "loss": 0.8444, "step": 1276 }, { "epoch": 0.46, "learning_rate": 8.899586107612021e-06, "loss": 0.9189, "step": 1277 }, { "epoch": 0.46, "learning_rate": 8.898686341551198e-06, "loss": 0.8886, "step": 1278 }, { "epoch": 0.46, "learning_rate": 8.897786575490374e-06, "loss": 0.9124, "step": 1279 }, { "epoch": 0.46, "learning_rate": 8.89688680942955e-06, "loss": 0.897, "step": 1280 }, { "epoch": 0.46, "learning_rate": 8.895987043368724e-06, "loss": 0.7556, "step": 1281 }, { "epoch": 0.46, "learning_rate": 8.8950872773079e-06, "loss": 0.9144, "step": 1282 }, { "epoch": 0.46, "learning_rate": 8.894187511247077e-06, "loss": 0.948, "step": 1283 }, { "epoch": 0.46, "learning_rate": 8.893287745186252e-06, "loss": 0.8757, "step": 1284 }, { "epoch": 0.46, "learning_rate": 8.892387979125429e-06, "loss": 0.8937, "step": 1285 }, { "epoch": 0.46, "learning_rate": 8.891488213064604e-06, "loss": 0.9465, "step": 1286 }, { "epoch": 0.46, "learning_rate": 8.890588447003779e-06, "loss": 0.7907, "step": 1287 }, { "epoch": 0.46, "learning_rate": 8.889688680942955e-06, "loss": 0.9647, "step": 1288 }, { "epoch": 0.46, "learning_rate": 8.888788914882132e-06, "loss": 0.8314, "step": 1289 }, { "epoch": 0.46, "learning_rate": 8.887889148821307e-06, "loss": 0.9271, "step": 1290 }, { "epoch": 0.46, "learning_rate": 8.886989382760482e-06, "loss": 0.9005, "step": 1291 }, { "epoch": 0.46, "learning_rate": 8.886089616699658e-06, "loss": 0.7207, "step": 1292 }, { "epoch": 0.46, "learning_rate": 8.885189850638835e-06, "loss": 0.7889, "step": 1293 }, { "epoch": 0.46, "learning_rate": 8.88429008457801e-06, "loss": 0.788, "step": 1294 }, { "epoch": 0.46, "learning_rate": 8.883390318517187e-06, "loss": 0.8267, "step": 1295 }, { "epoch": 0.47, "learning_rate": 8.882490552456362e-06, "loss": 0.9786, "step": 1296 }, { "epoch": 0.47, "learning_rate": 8.881590786395538e-06, "loss": 0.7749, "step": 1297 }, { "epoch": 0.47, "learning_rate": 8.880691020334715e-06, "loss": 0.8448, "step": 1298 }, { "epoch": 0.47, "learning_rate": 8.87979125427389e-06, "loss": 0.8471, "step": 1299 }, { "epoch": 0.47, "learning_rate": 8.878891488213065e-06, "loss": 0.8581, "step": 1300 }, { "epoch": 0.47, "learning_rate": 8.877991722152241e-06, "loss": 0.7892, "step": 1301 }, { "epoch": 0.47, "learning_rate": 8.877091956091418e-06, "loss": 0.8866, "step": 1302 }, { "epoch": 0.47, "learning_rate": 8.876192190030593e-06, "loss": 0.8559, "step": 1303 }, { "epoch": 0.47, "learning_rate": 8.87529242396977e-06, "loss": 0.7834, "step": 1304 }, { "epoch": 0.47, "learning_rate": 8.874392657908944e-06, "loss": 0.7844, "step": 1305 }, { "epoch": 0.47, "learning_rate": 8.87349289184812e-06, "loss": 0.9485, "step": 1306 }, { "epoch": 0.47, "learning_rate": 8.872593125787296e-06, "loss": 0.8764, "step": 1307 }, { "epoch": 0.47, "learning_rate": 8.871693359726472e-06, "loss": 0.8551, "step": 1308 }, { "epoch": 0.47, "learning_rate": 8.870793593665647e-06, "loss": 0.9535, "step": 1309 }, { "epoch": 0.47, "learning_rate": 8.869893827604822e-06, "loss": 0.8711, "step": 1310 }, { "epoch": 0.47, "learning_rate": 8.868994061543999e-06, "loss": 0.9551, "step": 1311 }, { "epoch": 0.47, "learning_rate": 8.868094295483176e-06, "loss": 0.992, "step": 1312 }, { "epoch": 0.47, "learning_rate": 8.86719452942235e-06, "loss": 0.7271, "step": 1313 }, { "epoch": 0.47, "learning_rate": 8.866294763361527e-06, "loss": 0.8335, "step": 1314 }, { "epoch": 0.47, "learning_rate": 8.865394997300702e-06, "loss": 0.8065, "step": 1315 }, { "epoch": 0.47, "learning_rate": 8.864495231239879e-06, "loss": 0.8775, "step": 1316 }, { "epoch": 0.47, "learning_rate": 8.863595465179055e-06, "loss": 0.9297, "step": 1317 }, { "epoch": 0.47, "learning_rate": 8.86269569911823e-06, "loss": 0.97, "step": 1318 }, { "epoch": 0.47, "learning_rate": 8.861795933057405e-06, "loss": 0.8873, "step": 1319 }, { "epoch": 0.47, "learning_rate": 8.860896166996582e-06, "loss": 0.8389, "step": 1320 }, { "epoch": 0.47, "learning_rate": 8.859996400935758e-06, "loss": 1.0123, "step": 1321 }, { "epoch": 0.47, "learning_rate": 8.859096634874933e-06, "loss": 0.8038, "step": 1322 }, { "epoch": 0.47, "learning_rate": 8.85819686881411e-06, "loss": 0.8464, "step": 1323 }, { "epoch": 0.48, "learning_rate": 8.857297102753285e-06, "loss": 0.8274, "step": 1324 }, { "epoch": 0.48, "learning_rate": 8.85639733669246e-06, "loss": 0.8537, "step": 1325 }, { "epoch": 0.48, "learning_rate": 8.855497570631636e-06, "loss": 0.7688, "step": 1326 }, { "epoch": 0.48, "learning_rate": 8.854597804570813e-06, "loss": 0.8017, "step": 1327 }, { "epoch": 0.48, "learning_rate": 8.853698038509988e-06, "loss": 0.8691, "step": 1328 }, { "epoch": 0.48, "learning_rate": 8.852798272449163e-06, "loss": 0.9568, "step": 1329 }, { "epoch": 0.48, "learning_rate": 8.85189850638834e-06, "loss": 0.7779, "step": 1330 }, { "epoch": 0.48, "learning_rate": 8.850998740327516e-06, "loss": 0.9456, "step": 1331 }, { "epoch": 0.48, "learning_rate": 8.850098974266691e-06, "loss": 0.7967, "step": 1332 }, { "epoch": 0.48, "learning_rate": 8.849199208205868e-06, "loss": 0.7785, "step": 1333 }, { "epoch": 0.48, "learning_rate": 8.848299442145042e-06, "loss": 0.7611, "step": 1334 }, { "epoch": 0.48, "learning_rate": 8.847399676084219e-06, "loss": 0.8587, "step": 1335 }, { "epoch": 0.48, "learning_rate": 8.846499910023396e-06, "loss": 0.9087, "step": 1336 }, { "epoch": 0.48, "learning_rate": 8.84560014396257e-06, "loss": 0.826, "step": 1337 }, { "epoch": 0.48, "learning_rate": 8.844700377901746e-06, "loss": 0.8914, "step": 1338 }, { "epoch": 0.48, "learning_rate": 8.843800611840922e-06, "loss": 0.8171, "step": 1339 }, { "epoch": 0.48, "learning_rate": 8.842900845780099e-06, "loss": 0.8662, "step": 1340 }, { "epoch": 0.48, "learning_rate": 8.842001079719274e-06, "loss": 0.8097, "step": 1341 }, { "epoch": 0.48, "learning_rate": 8.84110131365845e-06, "loss": 0.9223, "step": 1342 }, { "epoch": 0.48, "learning_rate": 8.840201547597625e-06, "loss": 0.8852, "step": 1343 }, { "epoch": 0.48, "learning_rate": 8.8393017815368e-06, "loss": 0.8327, "step": 1344 }, { "epoch": 0.48, "learning_rate": 8.838402015475977e-06, "loss": 0.8532, "step": 1345 }, { "epoch": 0.48, "learning_rate": 8.837502249415153e-06, "loss": 0.7414, "step": 1346 }, { "epoch": 0.48, "learning_rate": 8.836602483354328e-06, "loss": 0.8582, "step": 1347 }, { "epoch": 0.48, "learning_rate": 8.835702717293503e-06, "loss": 0.7488, "step": 1348 }, { "epoch": 0.48, "learning_rate": 8.83480295123268e-06, "loss": 0.7372, "step": 1349 }, { "epoch": 0.48, "learning_rate": 8.833903185171856e-06, "loss": 0.8708, "step": 1350 }, { "epoch": 0.48, "learning_rate": 8.833003419111031e-06, "loss": 0.7885, "step": 1351 }, { "epoch": 0.49, "learning_rate": 8.832103653050208e-06, "loss": 0.8746, "step": 1352 }, { "epoch": 0.49, "learning_rate": 8.831203886989383e-06, "loss": 0.7974, "step": 1353 }, { "epoch": 0.49, "learning_rate": 8.83030412092856e-06, "loss": 0.9379, "step": 1354 }, { "epoch": 0.49, "learning_rate": 8.829404354867736e-06, "loss": 0.9555, "step": 1355 }, { "epoch": 0.49, "learning_rate": 8.828504588806911e-06, "loss": 0.8874, "step": 1356 }, { "epoch": 0.49, "learning_rate": 8.827604822746086e-06, "loss": 1.0021, "step": 1357 }, { "epoch": 0.49, "learning_rate": 8.826705056685263e-06, "loss": 1.0378, "step": 1358 }, { "epoch": 0.49, "learning_rate": 8.82580529062444e-06, "loss": 1.0066, "step": 1359 }, { "epoch": 0.49, "learning_rate": 8.824905524563614e-06, "loss": 0.8328, "step": 1360 }, { "epoch": 0.49, "learning_rate": 8.82400575850279e-06, "loss": 0.8744, "step": 1361 }, { "epoch": 0.49, "learning_rate": 8.823105992441966e-06, "loss": 0.959, "step": 1362 }, { "epoch": 0.49, "learning_rate": 8.82220622638114e-06, "loss": 0.987, "step": 1363 }, { "epoch": 0.49, "learning_rate": 8.821306460320317e-06, "loss": 0.9547, "step": 1364 }, { "epoch": 0.49, "learning_rate": 8.820406694259494e-06, "loss": 0.958, "step": 1365 }, { "epoch": 0.49, "learning_rate": 8.819506928198669e-06, "loss": 0.757, "step": 1366 }, { "epoch": 0.49, "learning_rate": 8.818607162137844e-06, "loss": 0.7479, "step": 1367 }, { "epoch": 0.49, "learning_rate": 8.81770739607702e-06, "loss": 0.9351, "step": 1368 }, { "epoch": 0.49, "learning_rate": 8.816807630016197e-06, "loss": 0.8381, "step": 1369 }, { "epoch": 0.49, "learning_rate": 8.815907863955372e-06, "loss": 0.8486, "step": 1370 }, { "epoch": 0.49, "learning_rate": 8.815008097894548e-06, "loss": 0.8473, "step": 1371 }, { "epoch": 0.49, "learning_rate": 8.814108331833723e-06, "loss": 0.8082, "step": 1372 }, { "epoch": 0.49, "learning_rate": 8.8132085657729e-06, "loss": 0.8636, "step": 1373 }, { "epoch": 0.49, "learning_rate": 8.812308799712077e-06, "loss": 0.6939, "step": 1374 }, { "epoch": 0.49, "learning_rate": 8.811409033651252e-06, "loss": 0.9112, "step": 1375 }, { "epoch": 0.49, "learning_rate": 8.810509267590426e-06, "loss": 0.8631, "step": 1376 }, { "epoch": 0.49, "learning_rate": 8.809609501529603e-06, "loss": 0.8357, "step": 1377 }, { "epoch": 0.49, "learning_rate": 8.80870973546878e-06, "loss": 0.922, "step": 1378 }, { "epoch": 0.49, "learning_rate": 8.807809969407955e-06, "loss": 0.8358, "step": 1379 }, { "epoch": 0.5, "learning_rate": 8.806910203347131e-06, "loss": 0.8627, "step": 1380 }, { "epoch": 0.5, "learning_rate": 8.806010437286306e-06, "loss": 0.7839, "step": 1381 }, { "epoch": 0.5, "learning_rate": 8.805110671225481e-06, "loss": 0.7779, "step": 1382 }, { "epoch": 0.5, "learning_rate": 8.804210905164658e-06, "loss": 0.9216, "step": 1383 }, { "epoch": 0.5, "learning_rate": 8.803311139103834e-06, "loss": 0.8633, "step": 1384 }, { "epoch": 0.5, "learning_rate": 8.80241137304301e-06, "loss": 0.782, "step": 1385 }, { "epoch": 0.5, "learning_rate": 8.801511606982184e-06, "loss": 0.9397, "step": 1386 }, { "epoch": 0.5, "learning_rate": 8.80061184092136e-06, "loss": 0.7742, "step": 1387 }, { "epoch": 0.5, "learning_rate": 8.799712074860537e-06, "loss": 0.824, "step": 1388 }, { "epoch": 0.5, "learning_rate": 8.798812308799712e-06, "loss": 0.8036, "step": 1389 }, { "epoch": 0.5, "learning_rate": 8.797912542738889e-06, "loss": 0.8184, "step": 1390 }, { "epoch": 0.5, "learning_rate": 8.797012776678064e-06, "loss": 0.8363, "step": 1391 }, { "epoch": 0.5, "learning_rate": 8.79611301061724e-06, "loss": 0.8783, "step": 1392 }, { "epoch": 0.5, "learning_rate": 8.795213244556417e-06, "loss": 0.6756, "step": 1393 }, { "epoch": 0.5, "learning_rate": 8.794313478495592e-06, "loss": 0.8365, "step": 1394 }, { "epoch": 0.5, "learning_rate": 8.793413712434767e-06, "loss": 0.858, "step": 1395 }, { "epoch": 0.5, "learning_rate": 8.792513946373944e-06, "loss": 0.8307, "step": 1396 }, { "epoch": 0.5, "learning_rate": 8.79161418031312e-06, "loss": 0.8196, "step": 1397 }, { "epoch": 0.5, "learning_rate": 8.790714414252295e-06, "loss": 0.827, "step": 1398 }, { "epoch": 0.5, "learning_rate": 8.789814648191472e-06, "loss": 0.9766, "step": 1399 }, { "epoch": 0.5, "learning_rate": 8.788914882130647e-06, "loss": 0.9016, "step": 1400 }, { "epoch": 0.5, "learning_rate": 8.788015116069823e-06, "loss": 0.858, "step": 1401 }, { "epoch": 0.5, "learning_rate": 8.787115350008998e-06, "loss": 0.8182, "step": 1402 }, { "epoch": 0.5, "learning_rate": 8.786215583948175e-06, "loss": 0.8466, "step": 1403 }, { "epoch": 0.5, "learning_rate": 8.78531581788735e-06, "loss": 0.8834, "step": 1404 }, { "epoch": 0.5, "learning_rate": 8.784416051826525e-06, "loss": 0.8771, "step": 1405 }, { "epoch": 0.5, "learning_rate": 8.783516285765701e-06, "loss": 0.8252, "step": 1406 }, { "epoch": 0.51, "learning_rate": 8.782616519704878e-06, "loss": 0.845, "step": 1407 }, { "epoch": 0.51, "learning_rate": 8.781716753644053e-06, "loss": 0.8015, "step": 1408 }, { "epoch": 0.51, "learning_rate": 8.78081698758323e-06, "loss": 0.7679, "step": 1409 }, { "epoch": 0.51, "learning_rate": 8.779917221522404e-06, "loss": 0.8315, "step": 1410 }, { "epoch": 0.51, "learning_rate": 8.779017455461581e-06, "loss": 0.8226, "step": 1411 }, { "epoch": 0.51, "learning_rate": 8.778117689400758e-06, "loss": 0.8585, "step": 1412 }, { "epoch": 0.51, "learning_rate": 8.777217923339932e-06, "loss": 0.7675, "step": 1413 }, { "epoch": 0.51, "learning_rate": 8.776318157279107e-06, "loss": 0.7568, "step": 1414 }, { "epoch": 0.51, "learning_rate": 8.775418391218284e-06, "loss": 0.9148, "step": 1415 }, { "epoch": 0.51, "learning_rate": 8.77451862515746e-06, "loss": 0.8079, "step": 1416 }, { "epoch": 0.51, "learning_rate": 8.773618859096636e-06, "loss": 0.9139, "step": 1417 }, { "epoch": 0.51, "learning_rate": 8.772719093035812e-06, "loss": 0.801, "step": 1418 }, { "epoch": 0.51, "learning_rate": 8.771819326974987e-06, "loss": 0.8804, "step": 1419 }, { "epoch": 0.51, "learning_rate": 8.770919560914164e-06, "loss": 0.8092, "step": 1420 }, { "epoch": 0.51, "learning_rate": 8.770019794853339e-06, "loss": 0.748, "step": 1421 }, { "epoch": 0.51, "learning_rate": 8.769120028792515e-06, "loss": 0.7373, "step": 1422 }, { "epoch": 0.51, "learning_rate": 8.76822026273169e-06, "loss": 0.7463, "step": 1423 }, { "epoch": 0.51, "learning_rate": 8.767320496670865e-06, "loss": 0.8604, "step": 1424 }, { "epoch": 0.51, "learning_rate": 8.766420730610042e-06, "loss": 0.7809, "step": 1425 }, { "epoch": 0.51, "learning_rate": 8.765520964549218e-06, "loss": 0.6955, "step": 1426 }, { "epoch": 0.51, "learning_rate": 8.764621198488393e-06, "loss": 0.8957, "step": 1427 }, { "epoch": 0.51, "learning_rate": 8.76372143242757e-06, "loss": 0.7604, "step": 1428 }, { "epoch": 0.51, "learning_rate": 8.762821666366745e-06, "loss": 0.7568, "step": 1429 }, { "epoch": 0.51, "learning_rate": 8.761921900305921e-06, "loss": 0.914, "step": 1430 }, { "epoch": 0.51, "learning_rate": 8.761022134245098e-06, "loss": 0.8587, "step": 1431 }, { "epoch": 0.51, "learning_rate": 8.760122368184273e-06, "loss": 0.6938, "step": 1432 }, { "epoch": 0.51, "learning_rate": 8.759222602123448e-06, "loss": 0.9421, "step": 1433 }, { "epoch": 0.51, "learning_rate": 8.758322836062624e-06, "loss": 0.8911, "step": 1434 }, { "epoch": 0.52, "learning_rate": 8.757423070001801e-06, "loss": 0.8881, "step": 1435 }, { "epoch": 0.52, "learning_rate": 8.756523303940976e-06, "loss": 0.9669, "step": 1436 }, { "epoch": 0.52, "learning_rate": 8.755623537880153e-06, "loss": 0.9412, "step": 1437 }, { "epoch": 0.52, "learning_rate": 8.754723771819328e-06, "loss": 0.8075, "step": 1438 }, { "epoch": 0.52, "learning_rate": 8.753824005758504e-06, "loss": 0.7825, "step": 1439 }, { "epoch": 0.52, "learning_rate": 8.752924239697679e-06, "loss": 0.7929, "step": 1440 }, { "epoch": 0.52, "learning_rate": 8.752024473636856e-06, "loss": 0.8596, "step": 1441 }, { "epoch": 0.52, "learning_rate": 8.75112470757603e-06, "loss": 0.8154, "step": 1442 }, { "epoch": 0.52, "learning_rate": 8.750224941515206e-06, "loss": 0.8504, "step": 1443 }, { "epoch": 0.52, "learning_rate": 8.749325175454382e-06, "loss": 0.9524, "step": 1444 }, { "epoch": 0.52, "learning_rate": 8.748425409393559e-06, "loss": 0.7496, "step": 1445 }, { "epoch": 0.52, "learning_rate": 8.747525643332734e-06, "loss": 0.8953, "step": 1446 }, { "epoch": 0.52, "learning_rate": 8.74662587727191e-06, "loss": 0.9209, "step": 1447 }, { "epoch": 0.52, "learning_rate": 8.745726111211085e-06, "loss": 0.961, "step": 1448 }, { "epoch": 0.52, "learning_rate": 8.744826345150262e-06, "loss": 0.8052, "step": 1449 }, { "epoch": 0.52, "learning_rate": 8.743926579089438e-06, "loss": 0.8558, "step": 1450 }, { "epoch": 0.52, "learning_rate": 8.743026813028613e-06, "loss": 0.8887, "step": 1451 }, { "epoch": 0.52, "learning_rate": 8.742127046967788e-06, "loss": 0.9979, "step": 1452 }, { "epoch": 0.52, "learning_rate": 8.741227280906965e-06, "loss": 0.8791, "step": 1453 }, { "epoch": 0.52, "learning_rate": 8.740327514846142e-06, "loss": 0.9622, "step": 1454 }, { "epoch": 0.52, "learning_rate": 8.739427748785316e-06, "loss": 0.8626, "step": 1455 }, { "epoch": 0.52, "learning_rate": 8.738527982724493e-06, "loss": 0.7467, "step": 1456 }, { "epoch": 0.52, "learning_rate": 8.737628216663668e-06, "loss": 0.8014, "step": 1457 }, { "epoch": 0.52, "learning_rate": 8.736728450602845e-06, "loss": 0.868, "step": 1458 }, { "epoch": 0.52, "learning_rate": 8.73582868454202e-06, "loss": 0.8758, "step": 1459 }, { "epoch": 0.52, "learning_rate": 8.734928918481196e-06, "loss": 0.8817, "step": 1460 }, { "epoch": 0.52, "learning_rate": 8.734029152420371e-06, "loss": 0.8358, "step": 1461 }, { "epoch": 0.52, "learning_rate": 8.733129386359546e-06, "loss": 0.8659, "step": 1462 }, { "epoch": 0.53, "learning_rate": 8.732229620298723e-06, "loss": 0.8517, "step": 1463 }, { "epoch": 0.53, "learning_rate": 8.7313298542379e-06, "loss": 0.8625, "step": 1464 }, { "epoch": 0.53, "learning_rate": 8.730430088177074e-06, "loss": 0.8505, "step": 1465 }, { "epoch": 0.53, "learning_rate": 8.72953032211625e-06, "loss": 0.8673, "step": 1466 }, { "epoch": 0.53, "learning_rate": 8.728630556055426e-06, "loss": 0.802, "step": 1467 }, { "epoch": 0.53, "learning_rate": 8.727730789994602e-06, "loss": 0.9297, "step": 1468 }, { "epoch": 0.53, "learning_rate": 8.726831023933779e-06, "loss": 0.8016, "step": 1469 }, { "epoch": 0.53, "learning_rate": 8.725931257872954e-06, "loss": 0.9233, "step": 1470 }, { "epoch": 0.53, "learning_rate": 8.725031491812129e-06, "loss": 0.9261, "step": 1471 }, { "epoch": 0.53, "learning_rate": 8.724131725751305e-06, "loss": 0.8093, "step": 1472 }, { "epoch": 0.53, "learning_rate": 8.723231959690482e-06, "loss": 0.9107, "step": 1473 }, { "epoch": 0.53, "learning_rate": 8.722332193629657e-06, "loss": 0.8432, "step": 1474 }, { "epoch": 0.53, "learning_rate": 8.721432427568833e-06, "loss": 0.8715, "step": 1475 }, { "epoch": 0.53, "learning_rate": 8.720532661508008e-06, "loss": 0.9132, "step": 1476 }, { "epoch": 0.53, "learning_rate": 8.719632895447185e-06, "loss": 0.8229, "step": 1477 }, { "epoch": 0.53, "learning_rate": 8.71873312938636e-06, "loss": 0.7616, "step": 1478 }, { "epoch": 0.53, "learning_rate": 8.717833363325537e-06, "loss": 0.8037, "step": 1479 }, { "epoch": 0.53, "learning_rate": 8.716933597264711e-06, "loss": 0.9189, "step": 1480 }, { "epoch": 0.53, "learning_rate": 8.716033831203886e-06, "loss": 0.773, "step": 1481 }, { "epoch": 0.53, "learning_rate": 8.715134065143063e-06, "loss": 0.9392, "step": 1482 }, { "epoch": 0.53, "learning_rate": 8.71423429908224e-06, "loss": 0.9764, "step": 1483 }, { "epoch": 0.53, "learning_rate": 8.713334533021415e-06, "loss": 0.9365, "step": 1484 }, { "epoch": 0.53, "learning_rate": 8.712434766960591e-06, "loss": 0.8864, "step": 1485 }, { "epoch": 0.53, "learning_rate": 8.711535000899766e-06, "loss": 0.8426, "step": 1486 }, { "epoch": 0.53, "learning_rate": 8.710635234838943e-06, "loss": 0.8516, "step": 1487 }, { "epoch": 0.53, "learning_rate": 8.70973546877812e-06, "loss": 0.7652, "step": 1488 }, { "epoch": 0.53, "learning_rate": 8.708835702717294e-06, "loss": 0.8899, "step": 1489 }, { "epoch": 0.53, "learning_rate": 8.70793593665647e-06, "loss": 0.8619, "step": 1490 }, { "epoch": 0.54, "learning_rate": 8.707036170595646e-06, "loss": 0.8934, "step": 1491 }, { "epoch": 0.54, "learning_rate": 8.706136404534822e-06, "loss": 0.8389, "step": 1492 }, { "epoch": 0.54, "learning_rate": 8.705236638473997e-06, "loss": 0.8102, "step": 1493 }, { "epoch": 0.54, "learning_rate": 8.704336872413174e-06, "loss": 0.9507, "step": 1494 }, { "epoch": 0.54, "learning_rate": 8.703437106352349e-06, "loss": 0.8861, "step": 1495 }, { "epoch": 0.54, "learning_rate": 8.702537340291525e-06, "loss": 0.8107, "step": 1496 }, { "epoch": 0.54, "learning_rate": 8.7016375742307e-06, "loss": 0.9396, "step": 1497 }, { "epoch": 0.54, "learning_rate": 8.700737808169877e-06, "loss": 0.9686, "step": 1498 }, { "epoch": 0.54, "learning_rate": 8.699838042109052e-06, "loss": 0.9907, "step": 1499 }, { "epoch": 0.54, "learning_rate": 8.698938276048229e-06, "loss": 0.8445, "step": 1500 }, { "epoch": 0.54, "learning_rate": 8.698038509987403e-06, "loss": 0.8915, "step": 1501 }, { "epoch": 0.54, "learning_rate": 8.69713874392658e-06, "loss": 0.8223, "step": 1502 }, { "epoch": 0.54, "learning_rate": 8.696238977865755e-06, "loss": 0.8693, "step": 1503 }, { "epoch": 0.54, "learning_rate": 8.695339211804932e-06, "loss": 0.8269, "step": 1504 }, { "epoch": 0.54, "learning_rate": 8.694439445744107e-06, "loss": 0.7664, "step": 1505 }, { "epoch": 0.54, "learning_rate": 8.693539679683283e-06, "loss": 0.9745, "step": 1506 }, { "epoch": 0.54, "learning_rate": 8.69263991362246e-06, "loss": 0.8506, "step": 1507 }, { "epoch": 0.54, "learning_rate": 8.691740147561635e-06, "loss": 0.8267, "step": 1508 }, { "epoch": 0.54, "learning_rate": 8.69084038150081e-06, "loss": 0.8282, "step": 1509 }, { "epoch": 0.54, "learning_rate": 8.689940615439986e-06, "loss": 0.8799, "step": 1510 }, { "epoch": 0.54, "learning_rate": 8.689040849379163e-06, "loss": 0.8896, "step": 1511 }, { "epoch": 0.54, "learning_rate": 8.688141083318338e-06, "loss": 0.8639, "step": 1512 }, { "epoch": 0.54, "learning_rate": 8.687241317257514e-06, "loss": 0.7575, "step": 1513 }, { "epoch": 0.54, "learning_rate": 8.68634155119669e-06, "loss": 0.7875, "step": 1514 }, { "epoch": 0.54, "learning_rate": 8.685441785135866e-06, "loss": 0.7816, "step": 1515 }, { "epoch": 0.54, "learning_rate": 8.684542019075041e-06, "loss": 0.7931, "step": 1516 }, { "epoch": 0.54, "learning_rate": 8.683642253014217e-06, "loss": 0.8808, "step": 1517 }, { "epoch": 0.54, "learning_rate": 8.682742486953392e-06, "loss": 0.9854, "step": 1518 }, { "epoch": 0.55, "learning_rate": 8.681842720892569e-06, "loss": 0.7882, "step": 1519 }, { "epoch": 0.55, "learning_rate": 8.680942954831744e-06, "loss": 0.779, "step": 1520 }, { "epoch": 0.55, "learning_rate": 8.68004318877092e-06, "loss": 0.8634, "step": 1521 }, { "epoch": 0.55, "learning_rate": 8.679143422710095e-06, "loss": 0.9171, "step": 1522 }, { "epoch": 0.55, "learning_rate": 8.678243656649272e-06, "loss": 0.9141, "step": 1523 }, { "epoch": 0.55, "learning_rate": 8.677343890588447e-06, "loss": 0.8655, "step": 1524 }, { "epoch": 0.55, "learning_rate": 8.676444124527624e-06, "loss": 0.7672, "step": 1525 }, { "epoch": 0.55, "learning_rate": 8.6755443584668e-06, "loss": 0.8447, "step": 1526 }, { "epoch": 0.55, "learning_rate": 8.674644592405975e-06, "loss": 0.852, "step": 1527 }, { "epoch": 0.55, "learning_rate": 8.67374482634515e-06, "loss": 0.8737, "step": 1528 }, { "epoch": 0.55, "learning_rate": 8.672845060284327e-06, "loss": 0.7333, "step": 1529 }, { "epoch": 0.55, "learning_rate": 8.671945294223503e-06, "loss": 0.8126, "step": 1530 }, { "epoch": 0.55, "learning_rate": 8.671045528162678e-06, "loss": 0.8573, "step": 1531 }, { "epoch": 0.55, "learning_rate": 8.670145762101855e-06, "loss": 0.9016, "step": 1532 }, { "epoch": 0.55, "learning_rate": 8.66924599604103e-06, "loss": 0.9337, "step": 1533 }, { "epoch": 0.55, "learning_rate": 8.668346229980206e-06, "loss": 0.8788, "step": 1534 }, { "epoch": 0.55, "learning_rate": 8.667446463919381e-06, "loss": 0.9258, "step": 1535 }, { "epoch": 0.55, "learning_rate": 8.666546697858558e-06, "loss": 0.8527, "step": 1536 }, { "epoch": 0.55, "learning_rate": 8.665646931797733e-06, "loss": 0.8782, "step": 1537 }, { "epoch": 0.55, "learning_rate": 8.66474716573691e-06, "loss": 0.8527, "step": 1538 }, { "epoch": 0.55, "learning_rate": 8.663847399676084e-06, "loss": 0.7609, "step": 1539 }, { "epoch": 0.55, "learning_rate": 8.662947633615261e-06, "loss": 0.7986, "step": 1540 }, { "epoch": 0.55, "learning_rate": 8.662047867554436e-06, "loss": 0.8248, "step": 1541 }, { "epoch": 0.55, "learning_rate": 8.661148101493613e-06, "loss": 0.7845, "step": 1542 }, { "epoch": 0.55, "learning_rate": 8.660248335432787e-06, "loss": 0.8419, "step": 1543 }, { "epoch": 0.55, "learning_rate": 8.659348569371964e-06, "loss": 0.8292, "step": 1544 }, { "epoch": 0.55, "learning_rate": 8.65844880331114e-06, "loss": 0.7837, "step": 1545 }, { "epoch": 0.55, "learning_rate": 8.657549037250316e-06, "loss": 0.8676, "step": 1546 }, { "epoch": 0.56, "learning_rate": 8.65664927118949e-06, "loss": 0.7777, "step": 1547 }, { "epoch": 0.56, "learning_rate": 8.655749505128667e-06, "loss": 0.8982, "step": 1548 }, { "epoch": 0.56, "learning_rate": 8.654849739067844e-06, "loss": 0.8029, "step": 1549 }, { "epoch": 0.56, "learning_rate": 8.653949973007019e-06, "loss": 0.8805, "step": 1550 }, { "epoch": 0.56, "learning_rate": 8.653050206946195e-06, "loss": 0.917, "step": 1551 }, { "epoch": 0.56, "learning_rate": 8.65215044088537e-06, "loss": 0.8589, "step": 1552 }, { "epoch": 0.56, "learning_rate": 8.651250674824547e-06, "loss": 0.9623, "step": 1553 }, { "epoch": 0.56, "learning_rate": 8.650350908763722e-06, "loss": 0.7739, "step": 1554 }, { "epoch": 0.56, "learning_rate": 8.649451142702898e-06, "loss": 0.9725, "step": 1555 }, { "epoch": 0.56, "learning_rate": 8.648551376642073e-06, "loss": 0.91, "step": 1556 }, { "epoch": 0.56, "learning_rate": 8.64765161058125e-06, "loss": 0.8529, "step": 1557 }, { "epoch": 0.56, "learning_rate": 8.646751844520427e-06, "loss": 0.8867, "step": 1558 }, { "epoch": 0.56, "learning_rate": 8.645852078459601e-06, "loss": 0.8791, "step": 1559 }, { "epoch": 0.56, "learning_rate": 8.644952312398776e-06, "loss": 0.857, "step": 1560 }, { "epoch": 0.56, "learning_rate": 8.644052546337953e-06, "loss": 0.8029, "step": 1561 }, { "epoch": 0.56, "learning_rate": 8.643152780277128e-06, "loss": 1.0405, "step": 1562 }, { "epoch": 0.56, "learning_rate": 8.642253014216305e-06, "loss": 0.853, "step": 1563 }, { "epoch": 0.56, "learning_rate": 8.641353248155481e-06, "loss": 0.8913, "step": 1564 }, { "epoch": 0.56, "learning_rate": 8.640453482094656e-06, "loss": 0.8178, "step": 1565 }, { "epoch": 0.56, "learning_rate": 8.639553716033831e-06, "loss": 0.8493, "step": 1566 }, { "epoch": 0.56, "learning_rate": 8.638653949973008e-06, "loss": 0.9238, "step": 1567 }, { "epoch": 0.56, "learning_rate": 8.637754183912184e-06, "loss": 0.8106, "step": 1568 }, { "epoch": 0.56, "learning_rate": 8.636854417851359e-06, "loss": 0.9173, "step": 1569 }, { "epoch": 0.56, "learning_rate": 8.635954651790536e-06, "loss": 0.8715, "step": 1570 }, { "epoch": 0.56, "learning_rate": 8.63505488572971e-06, "loss": 0.8147, "step": 1571 }, { "epoch": 0.56, "learning_rate": 8.634155119668887e-06, "loss": 0.9426, "step": 1572 }, { "epoch": 0.56, "learning_rate": 8.633255353608062e-06, "loss": 0.8274, "step": 1573 }, { "epoch": 0.56, "learning_rate": 8.632355587547239e-06, "loss": 1.0492, "step": 1574 }, { "epoch": 0.57, "learning_rate": 8.631455821486414e-06, "loss": 0.836, "step": 1575 }, { "epoch": 0.57, "learning_rate": 8.63055605542559e-06, "loss": 0.8432, "step": 1576 }, { "epoch": 0.57, "learning_rate": 8.629656289364767e-06, "loss": 0.8477, "step": 1577 }, { "epoch": 0.57, "learning_rate": 8.628756523303942e-06, "loss": 0.8351, "step": 1578 }, { "epoch": 0.57, "learning_rate": 8.627856757243117e-06, "loss": 0.8309, "step": 1579 }, { "epoch": 0.57, "learning_rate": 8.626956991182293e-06, "loss": 0.751, "step": 1580 }, { "epoch": 0.57, "learning_rate": 8.626057225121468e-06, "loss": 0.904, "step": 1581 }, { "epoch": 0.57, "learning_rate": 8.625157459060645e-06, "loss": 0.8344, "step": 1582 }, { "epoch": 0.57, "learning_rate": 8.624257692999822e-06, "loss": 0.9288, "step": 1583 }, { "epoch": 0.57, "learning_rate": 8.623357926938997e-06, "loss": 0.8562, "step": 1584 }, { "epoch": 0.57, "learning_rate": 8.622458160878171e-06, "loss": 0.906, "step": 1585 }, { "epoch": 0.57, "learning_rate": 8.621558394817348e-06, "loss": 0.9286, "step": 1586 }, { "epoch": 0.57, "learning_rate": 8.620658628756525e-06, "loss": 0.9062, "step": 1587 }, { "epoch": 0.57, "learning_rate": 8.6197588626957e-06, "loss": 0.9554, "step": 1588 }, { "epoch": 0.57, "learning_rate": 8.618859096634876e-06, "loss": 0.8813, "step": 1589 }, { "epoch": 0.57, "learning_rate": 8.617959330574051e-06, "loss": 0.7445, "step": 1590 }, { "epoch": 0.57, "learning_rate": 8.617059564513228e-06, "loss": 0.8471, "step": 1591 }, { "epoch": 0.57, "learning_rate": 8.616159798452403e-06, "loss": 0.8415, "step": 1592 }, { "epoch": 0.57, "learning_rate": 8.61526003239158e-06, "loss": 0.9018, "step": 1593 }, { "epoch": 0.57, "learning_rate": 8.614360266330754e-06, "loss": 0.8744, "step": 1594 }, { "epoch": 0.57, "learning_rate": 8.61346050026993e-06, "loss": 0.9721, "step": 1595 }, { "epoch": 0.57, "learning_rate": 8.612560734209107e-06, "loss": 0.9672, "step": 1596 }, { "epoch": 0.57, "learning_rate": 8.611660968148282e-06, "loss": 0.8542, "step": 1597 }, { "epoch": 0.57, "learning_rate": 8.610761202087457e-06, "loss": 0.9523, "step": 1598 }, { "epoch": 0.57, "learning_rate": 8.609861436026634e-06, "loss": 0.9117, "step": 1599 }, { "epoch": 0.57, "learning_rate": 8.608961669965809e-06, "loss": 0.8303, "step": 1600 }, { "epoch": 0.57, "learning_rate": 8.608061903904985e-06, "loss": 0.8144, "step": 1601 }, { "epoch": 0.58, "learning_rate": 8.607162137844162e-06, "loss": 0.7854, "step": 1602 }, { "epoch": 0.58, "learning_rate": 8.606262371783337e-06, "loss": 0.9166, "step": 1603 }, { "epoch": 0.58, "learning_rate": 8.605362605722512e-06, "loss": 0.9277, "step": 1604 }, { "epoch": 0.58, "learning_rate": 8.604462839661689e-06, "loss": 0.8066, "step": 1605 }, { "epoch": 0.58, "learning_rate": 8.603563073600865e-06, "loss": 0.7914, "step": 1606 }, { "epoch": 0.58, "learning_rate": 8.60266330754004e-06, "loss": 0.8042, "step": 1607 }, { "epoch": 0.58, "learning_rate": 8.601763541479217e-06, "loss": 0.8675, "step": 1608 }, { "epoch": 0.58, "learning_rate": 8.600863775418392e-06, "loss": 0.8627, "step": 1609 }, { "epoch": 0.58, "learning_rate": 8.599964009357568e-06, "loss": 0.8693, "step": 1610 }, { "epoch": 0.58, "learning_rate": 8.599064243296743e-06, "loss": 0.8619, "step": 1611 }, { "epoch": 0.58, "learning_rate": 8.59816447723592e-06, "loss": 0.9394, "step": 1612 }, { "epoch": 0.58, "learning_rate": 8.597264711175095e-06, "loss": 0.7981, "step": 1613 }, { "epoch": 0.58, "learning_rate": 8.596364945114271e-06, "loss": 0.8716, "step": 1614 }, { "epoch": 0.58, "learning_rate": 8.595465179053448e-06, "loss": 0.7486, "step": 1615 }, { "epoch": 0.58, "learning_rate": 8.594565412992623e-06, "loss": 0.8744, "step": 1616 }, { "epoch": 0.58, "learning_rate": 8.593665646931798e-06, "loss": 0.829, "step": 1617 }, { "epoch": 0.58, "learning_rate": 8.592765880870974e-06, "loss": 0.8173, "step": 1618 }, { "epoch": 0.58, "learning_rate": 8.59186611481015e-06, "loss": 0.8467, "step": 1619 }, { "epoch": 0.58, "learning_rate": 8.590966348749326e-06, "loss": 0.892, "step": 1620 }, { "epoch": 0.58, "learning_rate": 8.590066582688503e-06, "loss": 0.9234, "step": 1621 }, { "epoch": 0.58, "learning_rate": 8.589166816627677e-06, "loss": 0.8734, "step": 1622 }, { "epoch": 0.58, "learning_rate": 8.588267050566852e-06, "loss": 0.9453, "step": 1623 }, { "epoch": 0.58, "learning_rate": 8.587367284506029e-06, "loss": 0.8594, "step": 1624 }, { "epoch": 0.58, "learning_rate": 8.586467518445206e-06, "loss": 0.8584, "step": 1625 }, { "epoch": 0.58, "learning_rate": 8.58556775238438e-06, "loss": 0.9156, "step": 1626 }, { "epoch": 0.58, "learning_rate": 8.584667986323555e-06, "loss": 0.9043, "step": 1627 }, { "epoch": 0.58, "learning_rate": 8.583768220262732e-06, "loss": 0.8879, "step": 1628 }, { "epoch": 0.58, "learning_rate": 8.582868454201909e-06, "loss": 0.8263, "step": 1629 }, { "epoch": 0.59, "learning_rate": 8.581968688141084e-06, "loss": 0.8093, "step": 1630 }, { "epoch": 0.59, "learning_rate": 8.58106892208026e-06, "loss": 0.9793, "step": 1631 }, { "epoch": 0.59, "learning_rate": 8.580169156019435e-06, "loss": 0.7833, "step": 1632 }, { "epoch": 0.59, "learning_rate": 8.579269389958612e-06, "loss": 1.0061, "step": 1633 }, { "epoch": 0.59, "learning_rate": 8.578369623897788e-06, "loss": 0.8769, "step": 1634 }, { "epoch": 0.59, "learning_rate": 8.577469857836963e-06, "loss": 0.991, "step": 1635 }, { "epoch": 0.59, "learning_rate": 8.576570091776138e-06, "loss": 0.7971, "step": 1636 }, { "epoch": 0.59, "learning_rate": 8.575670325715315e-06, "loss": 0.8462, "step": 1637 }, { "epoch": 0.59, "learning_rate": 8.57477055965449e-06, "loss": 0.8923, "step": 1638 }, { "epoch": 0.59, "learning_rate": 8.573870793593666e-06, "loss": 0.883, "step": 1639 }, { "epoch": 0.59, "learning_rate": 8.572971027532843e-06, "loss": 0.9362, "step": 1640 }, { "epoch": 0.59, "learning_rate": 8.572071261472018e-06, "loss": 0.8556, "step": 1641 }, { "epoch": 0.59, "learning_rate": 8.571171495411193e-06, "loss": 0.7812, "step": 1642 }, { "epoch": 0.59, "learning_rate": 8.57027172935037e-06, "loss": 0.96, "step": 1643 }, { "epoch": 0.59, "learning_rate": 8.569371963289546e-06, "loss": 0.8868, "step": 1644 }, { "epoch": 0.59, "learning_rate": 8.568472197228721e-06, "loss": 0.8794, "step": 1645 }, { "epoch": 0.59, "learning_rate": 8.567572431167896e-06, "loss": 0.7614, "step": 1646 }, { "epoch": 0.59, "learning_rate": 8.566672665107073e-06, "loss": 0.8616, "step": 1647 }, { "epoch": 0.59, "learning_rate": 8.565772899046249e-06, "loss": 0.8248, "step": 1648 }, { "epoch": 0.59, "learning_rate": 8.564873132985424e-06, "loss": 0.8343, "step": 1649 }, { "epoch": 0.59, "learning_rate": 8.5639733669246e-06, "loss": 0.7382, "step": 1650 }, { "epoch": 0.59, "learning_rate": 8.563073600863776e-06, "loss": 0.8584, "step": 1651 }, { "epoch": 0.59, "learning_rate": 8.562173834802952e-06, "loss": 0.8225, "step": 1652 }, { "epoch": 0.59, "learning_rate": 8.561274068742129e-06, "loss": 0.8902, "step": 1653 }, { "epoch": 0.59, "learning_rate": 8.560374302681304e-06, "loss": 0.888, "step": 1654 }, { "epoch": 0.59, "learning_rate": 8.559474536620479e-06, "loss": 0.8522, "step": 1655 }, { "epoch": 0.59, "learning_rate": 8.558574770559655e-06, "loss": 0.8521, "step": 1656 }, { "epoch": 0.59, "learning_rate": 8.557675004498832e-06, "loss": 0.9054, "step": 1657 }, { "epoch": 0.6, "learning_rate": 8.556775238438007e-06, "loss": 0.8638, "step": 1658 }, { "epoch": 0.6, "learning_rate": 8.555875472377183e-06, "loss": 0.8767, "step": 1659 }, { "epoch": 0.6, "learning_rate": 8.554975706316358e-06, "loss": 0.8424, "step": 1660 }, { "epoch": 0.6, "learning_rate": 8.554075940255533e-06, "loss": 0.8843, "step": 1661 }, { "epoch": 0.6, "learning_rate": 8.55317617419471e-06, "loss": 0.876, "step": 1662 }, { "epoch": 0.6, "learning_rate": 8.552276408133886e-06, "loss": 0.8382, "step": 1663 }, { "epoch": 0.6, "learning_rate": 8.551376642073061e-06, "loss": 1.0109, "step": 1664 }, { "epoch": 0.6, "learning_rate": 8.550476876012236e-06, "loss": 0.7652, "step": 1665 }, { "epoch": 0.6, "learning_rate": 8.549577109951413e-06, "loss": 0.8135, "step": 1666 }, { "epoch": 0.6, "learning_rate": 8.54867734389059e-06, "loss": 0.8889, "step": 1667 }, { "epoch": 0.6, "learning_rate": 8.547777577829764e-06, "loss": 0.9795, "step": 1668 }, { "epoch": 0.6, "learning_rate": 8.546877811768941e-06, "loss": 0.8348, "step": 1669 }, { "epoch": 0.6, "learning_rate": 8.545978045708116e-06, "loss": 0.8423, "step": 1670 }, { "epoch": 0.6, "learning_rate": 8.545078279647293e-06, "loss": 0.8866, "step": 1671 }, { "epoch": 0.6, "learning_rate": 8.54417851358647e-06, "loss": 0.8567, "step": 1672 }, { "epoch": 0.6, "learning_rate": 8.543278747525644e-06, "loss": 0.8288, "step": 1673 }, { "epoch": 0.6, "learning_rate": 8.542378981464819e-06, "loss": 0.8977, "step": 1674 }, { "epoch": 0.6, "learning_rate": 8.541479215403996e-06, "loss": 0.9362, "step": 1675 }, { "epoch": 0.6, "learning_rate": 8.540579449343172e-06, "loss": 0.776, "step": 1676 }, { "epoch": 0.6, "learning_rate": 8.539679683282347e-06, "loss": 0.7959, "step": 1677 }, { "epoch": 0.6, "learning_rate": 8.538779917221524e-06, "loss": 0.8683, "step": 1678 }, { "epoch": 0.6, "learning_rate": 8.537880151160699e-06, "loss": 0.8631, "step": 1679 }, { "epoch": 0.6, "learning_rate": 8.536980385099874e-06, "loss": 0.9389, "step": 1680 }, { "epoch": 0.6, "learning_rate": 8.53608061903905e-06, "loss": 0.867, "step": 1681 }, { "epoch": 0.6, "learning_rate": 8.535180852978227e-06, "loss": 0.9608, "step": 1682 }, { "epoch": 0.6, "learning_rate": 8.534281086917402e-06, "loss": 0.8115, "step": 1683 }, { "epoch": 0.6, "learning_rate": 8.533381320856577e-06, "loss": 0.8265, "step": 1684 }, { "epoch": 0.6, "learning_rate": 8.532481554795753e-06, "loss": 0.9258, "step": 1685 }, { "epoch": 0.61, "learning_rate": 8.53158178873493e-06, "loss": 0.9052, "step": 1686 }, { "epoch": 0.61, "learning_rate": 8.530682022674105e-06, "loss": 0.7442, "step": 1687 }, { "epoch": 0.61, "learning_rate": 8.529782256613282e-06, "loss": 0.8406, "step": 1688 }, { "epoch": 0.61, "learning_rate": 8.528882490552456e-06, "loss": 0.8469, "step": 1689 }, { "epoch": 0.61, "learning_rate": 8.527982724491633e-06, "loss": 0.854, "step": 1690 }, { "epoch": 0.61, "learning_rate": 8.52708295843081e-06, "loss": 0.806, "step": 1691 }, { "epoch": 0.61, "learning_rate": 8.526183192369985e-06, "loss": 0.881, "step": 1692 }, { "epoch": 0.61, "learning_rate": 8.52528342630916e-06, "loss": 0.8562, "step": 1693 }, { "epoch": 0.61, "learning_rate": 8.524383660248336e-06, "loss": 0.842, "step": 1694 }, { "epoch": 0.61, "learning_rate": 8.523483894187513e-06, "loss": 0.9168, "step": 1695 }, { "epoch": 0.61, "learning_rate": 8.522584128126688e-06, "loss": 0.8434, "step": 1696 }, { "epoch": 0.61, "learning_rate": 8.521684362065864e-06, "loss": 0.8839, "step": 1697 }, { "epoch": 0.61, "learning_rate": 8.52078459600504e-06, "loss": 0.8072, "step": 1698 }, { "epoch": 0.61, "learning_rate": 8.519884829944214e-06, "loss": 0.8442, "step": 1699 }, { "epoch": 0.61, "learning_rate": 8.51898506388339e-06, "loss": 0.9645, "step": 1700 }, { "epoch": 0.61, "learning_rate": 8.518085297822567e-06, "loss": 0.8972, "step": 1701 }, { "epoch": 0.61, "learning_rate": 8.517185531761742e-06, "loss": 0.9676, "step": 1702 }, { "epoch": 0.61, "learning_rate": 8.516285765700917e-06, "loss": 0.9206, "step": 1703 }, { "epoch": 0.61, "learning_rate": 8.515385999640094e-06, "loss": 0.8116, "step": 1704 }, { "epoch": 0.61, "learning_rate": 8.51448623357927e-06, "loss": 0.828, "step": 1705 }, { "epoch": 0.61, "learning_rate": 8.513586467518445e-06, "loss": 0.9584, "step": 1706 }, { "epoch": 0.61, "learning_rate": 8.512686701457622e-06, "loss": 0.9045, "step": 1707 }, { "epoch": 0.61, "learning_rate": 8.511786935396797e-06, "loss": 0.8729, "step": 1708 }, { "epoch": 0.61, "learning_rate": 8.510887169335974e-06, "loss": 0.8159, "step": 1709 }, { "epoch": 0.61, "learning_rate": 8.50998740327515e-06, "loss": 0.8564, "step": 1710 }, { "epoch": 0.61, "learning_rate": 8.509087637214325e-06, "loss": 0.939, "step": 1711 }, { "epoch": 0.61, "learning_rate": 8.5081878711535e-06, "loss": 0.8722, "step": 1712 }, { "epoch": 0.61, "learning_rate": 8.507288105092677e-06, "loss": 0.8528, "step": 1713 }, { "epoch": 0.62, "learning_rate": 8.506388339031853e-06, "loss": 0.9078, "step": 1714 }, { "epoch": 0.62, "learning_rate": 8.505488572971028e-06, "loss": 0.8271, "step": 1715 }, { "epoch": 0.62, "learning_rate": 8.504588806910205e-06, "loss": 0.8524, "step": 1716 }, { "epoch": 0.62, "learning_rate": 8.50368904084938e-06, "loss": 0.9198, "step": 1717 }, { "epoch": 0.62, "learning_rate": 8.502789274788555e-06, "loss": 0.8532, "step": 1718 }, { "epoch": 0.62, "learning_rate": 8.501889508727731e-06, "loss": 0.926, "step": 1719 }, { "epoch": 0.62, "learning_rate": 8.500989742666908e-06, "loss": 0.8824, "step": 1720 }, { "epoch": 0.62, "learning_rate": 8.500089976606083e-06, "loss": 0.7242, "step": 1721 }, { "epoch": 0.62, "learning_rate": 8.499190210545258e-06, "loss": 0.8913, "step": 1722 }, { "epoch": 0.62, "learning_rate": 8.498290444484434e-06, "loss": 0.8858, "step": 1723 }, { "epoch": 0.62, "learning_rate": 8.497390678423611e-06, "loss": 0.8069, "step": 1724 }, { "epoch": 0.62, "learning_rate": 8.496490912362786e-06, "loss": 0.8787, "step": 1725 }, { "epoch": 0.62, "learning_rate": 8.495591146301962e-06, "loss": 0.9199, "step": 1726 }, { "epoch": 0.62, "learning_rate": 8.494691380241137e-06, "loss": 0.7897, "step": 1727 }, { "epoch": 0.62, "learning_rate": 8.493791614180314e-06, "loss": 0.9065, "step": 1728 }, { "epoch": 0.62, "learning_rate": 8.49289184811949e-06, "loss": 0.879, "step": 1729 }, { "epoch": 0.62, "learning_rate": 8.491992082058666e-06, "loss": 0.8616, "step": 1730 }, { "epoch": 0.62, "learning_rate": 8.49109231599784e-06, "loss": 0.9279, "step": 1731 }, { "epoch": 0.62, "learning_rate": 8.490192549937017e-06, "loss": 0.8672, "step": 1732 }, { "epoch": 0.62, "learning_rate": 8.489292783876194e-06, "loss": 0.791, "step": 1733 }, { "epoch": 0.62, "learning_rate": 8.488393017815369e-06, "loss": 0.838, "step": 1734 }, { "epoch": 0.62, "learning_rate": 8.487493251754545e-06, "loss": 0.7833, "step": 1735 }, { "epoch": 0.62, "learning_rate": 8.48659348569372e-06, "loss": 0.7502, "step": 1736 }, { "epoch": 0.62, "learning_rate": 8.485693719632895e-06, "loss": 0.8972, "step": 1737 }, { "epoch": 0.62, "learning_rate": 8.484793953572072e-06, "loss": 0.9385, "step": 1738 }, { "epoch": 0.62, "learning_rate": 8.483894187511248e-06, "loss": 0.753, "step": 1739 }, { "epoch": 0.62, "learning_rate": 8.482994421450423e-06, "loss": 0.9104, "step": 1740 }, { "epoch": 0.62, "learning_rate": 8.482094655389598e-06, "loss": 0.8255, "step": 1741 }, { "epoch": 0.63, "learning_rate": 8.481194889328775e-06, "loss": 0.8808, "step": 1742 }, { "epoch": 0.63, "learning_rate": 8.480295123267951e-06, "loss": 0.7829, "step": 1743 }, { "epoch": 0.63, "learning_rate": 8.479395357207126e-06, "loss": 0.7593, "step": 1744 }, { "epoch": 0.63, "learning_rate": 8.478495591146303e-06, "loss": 0.8461, "step": 1745 }, { "epoch": 0.63, "learning_rate": 8.477595825085478e-06, "loss": 0.8407, "step": 1746 }, { "epoch": 0.63, "learning_rate": 8.476696059024654e-06, "loss": 0.8783, "step": 1747 }, { "epoch": 0.63, "learning_rate": 8.475796292963831e-06, "loss": 0.908, "step": 1748 }, { "epoch": 0.63, "learning_rate": 8.474896526903006e-06, "loss": 0.9246, "step": 1749 }, { "epoch": 0.63, "learning_rate": 8.473996760842181e-06, "loss": 0.9395, "step": 1750 }, { "epoch": 0.63, "learning_rate": 8.473096994781358e-06, "loss": 0.8336, "step": 1751 }, { "epoch": 0.63, "learning_rate": 8.472197228720534e-06, "loss": 0.8151, "step": 1752 }, { "epoch": 0.63, "learning_rate": 8.471297462659709e-06, "loss": 0.8751, "step": 1753 }, { "epoch": 0.63, "learning_rate": 8.470397696598886e-06, "loss": 0.8481, "step": 1754 }, { "epoch": 0.63, "learning_rate": 8.46949793053806e-06, "loss": 0.9588, "step": 1755 }, { "epoch": 0.63, "learning_rate": 8.468598164477236e-06, "loss": 0.8611, "step": 1756 }, { "epoch": 0.63, "learning_rate": 8.467698398416412e-06, "loss": 0.8152, "step": 1757 }, { "epoch": 0.63, "learning_rate": 8.466798632355589e-06, "loss": 0.9054, "step": 1758 }, { "epoch": 0.63, "learning_rate": 8.465898866294764e-06, "loss": 0.8847, "step": 1759 }, { "epoch": 0.63, "learning_rate": 8.464999100233939e-06, "loss": 0.7964, "step": 1760 }, { "epoch": 0.63, "learning_rate": 8.464099334173115e-06, "loss": 0.8136, "step": 1761 }, { "epoch": 0.63, "learning_rate": 8.463199568112292e-06, "loss": 0.9311, "step": 1762 }, { "epoch": 0.63, "learning_rate": 8.462299802051467e-06, "loss": 0.7339, "step": 1763 }, { "epoch": 0.63, "learning_rate": 8.461400035990643e-06, "loss": 0.8666, "step": 1764 }, { "epoch": 0.63, "learning_rate": 8.460500269929818e-06, "loss": 0.846, "step": 1765 }, { "epoch": 0.63, "learning_rate": 8.459600503868995e-06, "loss": 0.8794, "step": 1766 }, { "epoch": 0.63, "learning_rate": 8.458700737808172e-06, "loss": 0.9123, "step": 1767 }, { "epoch": 0.63, "learning_rate": 8.457800971747346e-06, "loss": 0.7377, "step": 1768 }, { "epoch": 0.63, "learning_rate": 8.456901205686521e-06, "loss": 0.9447, "step": 1769 }, { "epoch": 0.64, "learning_rate": 8.456001439625698e-06, "loss": 0.816, "step": 1770 }, { "epoch": 0.64, "learning_rate": 8.455101673564875e-06, "loss": 0.8868, "step": 1771 }, { "epoch": 0.64, "learning_rate": 8.45420190750405e-06, "loss": 0.9066, "step": 1772 }, { "epoch": 0.64, "learning_rate": 8.453302141443226e-06, "loss": 0.9581, "step": 1773 }, { "epoch": 0.64, "learning_rate": 8.452402375382401e-06, "loss": 0.8001, "step": 1774 }, { "epoch": 0.64, "learning_rate": 8.451502609321578e-06, "loss": 0.903, "step": 1775 }, { "epoch": 0.64, "learning_rate": 8.450602843260753e-06, "loss": 0.8173, "step": 1776 }, { "epoch": 0.64, "learning_rate": 8.44970307719993e-06, "loss": 0.8722, "step": 1777 }, { "epoch": 0.64, "learning_rate": 8.448803311139104e-06, "loss": 0.9697, "step": 1778 }, { "epoch": 0.64, "learning_rate": 8.447903545078279e-06, "loss": 0.7958, "step": 1779 }, { "epoch": 0.64, "learning_rate": 8.447003779017456e-06, "loss": 0.8008, "step": 1780 }, { "epoch": 0.64, "learning_rate": 8.446104012956632e-06, "loss": 0.8988, "step": 1781 }, { "epoch": 0.64, "learning_rate": 8.445204246895807e-06, "loss": 0.8967, "step": 1782 }, { "epoch": 0.64, "learning_rate": 8.444304480834984e-06, "loss": 0.7581, "step": 1783 }, { "epoch": 0.64, "learning_rate": 8.443404714774159e-06, "loss": 0.8453, "step": 1784 }, { "epoch": 0.64, "learning_rate": 8.442504948713335e-06, "loss": 0.7956, "step": 1785 }, { "epoch": 0.64, "learning_rate": 8.441605182652512e-06, "loss": 0.9075, "step": 1786 }, { "epoch": 0.64, "learning_rate": 8.440705416591687e-06, "loss": 0.7907, "step": 1787 }, { "epoch": 0.64, "learning_rate": 8.439805650530862e-06, "loss": 0.8365, "step": 1788 }, { "epoch": 0.64, "learning_rate": 8.438905884470038e-06, "loss": 0.8189, "step": 1789 }, { "epoch": 0.64, "learning_rate": 8.438006118409215e-06, "loss": 0.8084, "step": 1790 }, { "epoch": 0.64, "learning_rate": 8.43710635234839e-06, "loss": 0.802, "step": 1791 }, { "epoch": 0.64, "learning_rate": 8.436206586287567e-06, "loss": 0.8506, "step": 1792 }, { "epoch": 0.64, "learning_rate": 8.435306820226742e-06, "loss": 0.8816, "step": 1793 }, { "epoch": 0.64, "learning_rate": 8.434407054165918e-06, "loss": 0.8371, "step": 1794 }, { "epoch": 0.64, "learning_rate": 8.433507288105093e-06, "loss": 0.7506, "step": 1795 }, { "epoch": 0.64, "learning_rate": 8.43260752204427e-06, "loss": 0.9324, "step": 1796 }, { "epoch": 0.65, "learning_rate": 8.431707755983445e-06, "loss": 0.8634, "step": 1797 }, { "epoch": 0.65, "learning_rate": 8.43080798992262e-06, "loss": 0.804, "step": 1798 }, { "epoch": 0.65, "learning_rate": 8.429908223861796e-06, "loss": 0.807, "step": 1799 }, { "epoch": 0.65, "learning_rate": 8.429008457800973e-06, "loss": 0.7412, "step": 1800 }, { "epoch": 0.65, "learning_rate": 8.428108691740148e-06, "loss": 0.8345, "step": 1801 }, { "epoch": 0.65, "learning_rate": 8.427208925679324e-06, "loss": 0.8294, "step": 1802 }, { "epoch": 0.65, "learning_rate": 8.4263091596185e-06, "loss": 0.8702, "step": 1803 }, { "epoch": 0.65, "learning_rate": 8.425409393557676e-06, "loss": 0.7851, "step": 1804 }, { "epoch": 0.65, "learning_rate": 8.424509627496852e-06, "loss": 0.8664, "step": 1805 }, { "epoch": 0.65, "learning_rate": 8.423609861436027e-06, "loss": 0.8619, "step": 1806 }, { "epoch": 0.65, "learning_rate": 8.422710095375202e-06, "loss": 1.0189, "step": 1807 }, { "epoch": 0.65, "learning_rate": 8.421810329314379e-06, "loss": 0.8014, "step": 1808 }, { "epoch": 0.65, "learning_rate": 8.420910563253556e-06, "loss": 0.8092, "step": 1809 }, { "epoch": 0.65, "learning_rate": 8.42001079719273e-06, "loss": 0.8025, "step": 1810 }, { "epoch": 0.65, "learning_rate": 8.419111031131907e-06, "loss": 0.933, "step": 1811 }, { "epoch": 0.65, "learning_rate": 8.418211265071082e-06, "loss": 0.8628, "step": 1812 }, { "epoch": 0.65, "learning_rate": 8.417311499010259e-06, "loss": 0.8279, "step": 1813 }, { "epoch": 0.65, "learning_rate": 8.416411732949435e-06, "loss": 0.9036, "step": 1814 }, { "epoch": 0.65, "learning_rate": 8.41551196688861e-06, "loss": 0.8183, "step": 1815 }, { "epoch": 0.65, "learning_rate": 8.414612200827785e-06, "loss": 0.7465, "step": 1816 }, { "epoch": 0.65, "learning_rate": 8.41371243476696e-06, "loss": 0.9644, "step": 1817 }, { "epoch": 0.65, "learning_rate": 8.412812668706137e-06, "loss": 0.8358, "step": 1818 }, { "epoch": 0.65, "learning_rate": 8.411912902645313e-06, "loss": 0.7105, "step": 1819 }, { "epoch": 0.65, "learning_rate": 8.411013136584488e-06, "loss": 1.075, "step": 1820 }, { "epoch": 0.65, "learning_rate": 8.410113370523665e-06, "loss": 0.8697, "step": 1821 }, { "epoch": 0.65, "learning_rate": 8.40921360446284e-06, "loss": 0.8086, "step": 1822 }, { "epoch": 0.65, "learning_rate": 8.408313838402016e-06, "loss": 0.8461, "step": 1823 }, { "epoch": 0.65, "learning_rate": 8.407414072341193e-06, "loss": 0.9585, "step": 1824 }, { "epoch": 0.66, "learning_rate": 8.406514306280368e-06, "loss": 0.8381, "step": 1825 }, { "epoch": 0.66, "learning_rate": 8.405614540219543e-06, "loss": 0.8582, "step": 1826 }, { "epoch": 0.66, "learning_rate": 8.40471477415872e-06, "loss": 0.8455, "step": 1827 }, { "epoch": 0.66, "learning_rate": 8.403815008097896e-06, "loss": 0.8684, "step": 1828 }, { "epoch": 0.66, "learning_rate": 8.402915242037071e-06, "loss": 0.8184, "step": 1829 }, { "epoch": 0.66, "learning_rate": 8.402015475976248e-06, "loss": 0.9745, "step": 1830 }, { "epoch": 0.66, "learning_rate": 8.401115709915422e-06, "loss": 0.7825, "step": 1831 }, { "epoch": 0.66, "learning_rate": 8.400215943854599e-06, "loss": 0.9757, "step": 1832 }, { "epoch": 0.66, "learning_rate": 8.399316177793776e-06, "loss": 0.8905, "step": 1833 }, { "epoch": 0.66, "learning_rate": 8.39841641173295e-06, "loss": 0.7872, "step": 1834 }, { "epoch": 0.66, "learning_rate": 8.397516645672126e-06, "loss": 0.7706, "step": 1835 }, { "epoch": 0.66, "learning_rate": 8.3966168796113e-06, "loss": 0.8735, "step": 1836 }, { "epoch": 0.66, "learning_rate": 8.395717113550477e-06, "loss": 0.9219, "step": 1837 }, { "epoch": 0.66, "learning_rate": 8.394817347489654e-06, "loss": 0.8776, "step": 1838 }, { "epoch": 0.66, "learning_rate": 8.393917581428829e-06, "loss": 0.8057, "step": 1839 }, { "epoch": 0.66, "learning_rate": 8.393017815368005e-06, "loss": 0.8364, "step": 1840 }, { "epoch": 0.66, "learning_rate": 8.39211804930718e-06, "loss": 0.8292, "step": 1841 }, { "epoch": 0.66, "learning_rate": 8.391218283246357e-06, "loss": 0.8417, "step": 1842 }, { "epoch": 0.66, "learning_rate": 8.390318517185533e-06, "loss": 0.8941, "step": 1843 }, { "epoch": 0.66, "learning_rate": 8.389418751124708e-06, "loss": 0.7445, "step": 1844 }, { "epoch": 0.66, "learning_rate": 8.388518985063883e-06, "loss": 0.8663, "step": 1845 }, { "epoch": 0.66, "learning_rate": 8.38761921900306e-06, "loss": 0.7212, "step": 1846 }, { "epoch": 0.66, "learning_rate": 8.386719452942236e-06, "loss": 0.7542, "step": 1847 }, { "epoch": 0.66, "learning_rate": 8.385819686881411e-06, "loss": 0.802, "step": 1848 }, { "epoch": 0.66, "learning_rate": 8.384919920820588e-06, "loss": 0.7536, "step": 1849 }, { "epoch": 0.66, "learning_rate": 8.384020154759763e-06, "loss": 0.8094, "step": 1850 }, { "epoch": 0.66, "learning_rate": 8.38312038869894e-06, "loss": 0.8996, "step": 1851 }, { "epoch": 0.66, "learning_rate": 8.382220622638116e-06, "loss": 0.9421, "step": 1852 }, { "epoch": 0.67, "learning_rate": 8.381320856577291e-06, "loss": 0.8708, "step": 1853 }, { "epoch": 0.67, "learning_rate": 8.380421090516466e-06, "loss": 0.8663, "step": 1854 }, { "epoch": 0.67, "learning_rate": 8.379521324455641e-06, "loss": 0.8265, "step": 1855 }, { "epoch": 0.67, "learning_rate": 8.378621558394817e-06, "loss": 0.82, "step": 1856 }, { "epoch": 0.67, "learning_rate": 8.377721792333994e-06, "loss": 0.9267, "step": 1857 }, { "epoch": 0.67, "learning_rate": 8.376822026273169e-06, "loss": 0.9329, "step": 1858 }, { "epoch": 0.67, "learning_rate": 8.375922260212346e-06, "loss": 0.9818, "step": 1859 }, { "epoch": 0.67, "learning_rate": 8.37502249415152e-06, "loss": 0.7496, "step": 1860 }, { "epoch": 0.67, "learning_rate": 8.374122728090697e-06, "loss": 0.7942, "step": 1861 }, { "epoch": 0.67, "learning_rate": 8.373222962029874e-06, "loss": 0.8044, "step": 1862 }, { "epoch": 0.67, "learning_rate": 8.372323195969049e-06, "loss": 0.8471, "step": 1863 }, { "epoch": 0.67, "learning_rate": 8.371423429908224e-06, "loss": 0.9438, "step": 1864 }, { "epoch": 0.67, "learning_rate": 8.3705236638474e-06, "loss": 0.809, "step": 1865 }, { "epoch": 0.67, "learning_rate": 8.369623897786577e-06, "loss": 0.8029, "step": 1866 }, { "epoch": 0.67, "learning_rate": 8.368724131725752e-06, "loss": 0.9155, "step": 1867 }, { "epoch": 0.67, "learning_rate": 8.367824365664928e-06, "loss": 0.7285, "step": 1868 }, { "epoch": 0.67, "learning_rate": 8.366924599604103e-06, "loss": 0.7285, "step": 1869 }, { "epoch": 0.67, "learning_rate": 8.36602483354328e-06, "loss": 0.8077, "step": 1870 }, { "epoch": 0.67, "learning_rate": 8.365125067482457e-06, "loss": 0.7615, "step": 1871 }, { "epoch": 0.67, "learning_rate": 8.364225301421631e-06, "loss": 0.8641, "step": 1872 }, { "epoch": 0.67, "learning_rate": 8.363325535360806e-06, "loss": 0.888, "step": 1873 }, { "epoch": 0.67, "learning_rate": 8.362425769299981e-06, "loss": 0.7355, "step": 1874 }, { "epoch": 0.67, "learning_rate": 8.361526003239158e-06, "loss": 0.8511, "step": 1875 }, { "epoch": 0.67, "learning_rate": 8.360626237178335e-06, "loss": 0.8758, "step": 1876 }, { "epoch": 0.67, "learning_rate": 8.35972647111751e-06, "loss": 0.7904, "step": 1877 }, { "epoch": 0.67, "learning_rate": 8.358826705056686e-06, "loss": 0.9437, "step": 1878 }, { "epoch": 0.67, "learning_rate": 8.357926938995861e-06, "loss": 0.9624, "step": 1879 }, { "epoch": 0.67, "learning_rate": 8.357027172935038e-06, "loss": 0.7862, "step": 1880 }, { "epoch": 0.68, "learning_rate": 8.356127406874214e-06, "loss": 0.8508, "step": 1881 }, { "epoch": 0.68, "learning_rate": 8.35522764081339e-06, "loss": 0.9212, "step": 1882 }, { "epoch": 0.68, "learning_rate": 8.354327874752564e-06, "loss": 0.95, "step": 1883 }, { "epoch": 0.68, "learning_rate": 8.35342810869174e-06, "loss": 0.942, "step": 1884 }, { "epoch": 0.68, "learning_rate": 8.352528342630917e-06, "loss": 0.8579, "step": 1885 }, { "epoch": 0.68, "learning_rate": 8.351628576570092e-06, "loss": 0.8749, "step": 1886 }, { "epoch": 0.68, "learning_rate": 8.350728810509269e-06, "loss": 0.9084, "step": 1887 }, { "epoch": 0.68, "learning_rate": 8.349829044448444e-06, "loss": 0.8728, "step": 1888 }, { "epoch": 0.68, "learning_rate": 8.34892927838762e-06, "loss": 0.8229, "step": 1889 }, { "epoch": 0.68, "learning_rate": 8.348029512326797e-06, "loss": 0.8652, "step": 1890 }, { "epoch": 0.68, "learning_rate": 8.347129746265972e-06, "loss": 0.8701, "step": 1891 }, { "epoch": 0.68, "learning_rate": 8.346229980205147e-06, "loss": 0.9603, "step": 1892 }, { "epoch": 0.68, "learning_rate": 8.345330214144323e-06, "loss": 0.753, "step": 1893 }, { "epoch": 0.68, "learning_rate": 8.344430448083498e-06, "loss": 0.8547, "step": 1894 }, { "epoch": 0.68, "learning_rate": 8.343530682022675e-06, "loss": 0.8294, "step": 1895 }, { "epoch": 0.68, "learning_rate": 8.34263091596185e-06, "loss": 0.8316, "step": 1896 }, { "epoch": 0.68, "learning_rate": 8.341731149901027e-06, "loss": 0.8908, "step": 1897 }, { "epoch": 0.68, "learning_rate": 8.340831383840201e-06, "loss": 0.8196, "step": 1898 }, { "epoch": 0.68, "learning_rate": 8.339931617779378e-06, "loss": 0.7793, "step": 1899 }, { "epoch": 0.68, "learning_rate": 8.339031851718555e-06, "loss": 0.7807, "step": 1900 }, { "epoch": 0.68, "learning_rate": 8.33813208565773e-06, "loss": 0.8762, "step": 1901 }, { "epoch": 0.68, "learning_rate": 8.337232319596905e-06, "loss": 0.8705, "step": 1902 }, { "epoch": 0.68, "learning_rate": 8.336332553536081e-06, "loss": 0.8235, "step": 1903 }, { "epoch": 0.68, "learning_rate": 8.335432787475258e-06, "loss": 0.8167, "step": 1904 }, { "epoch": 0.68, "learning_rate": 8.334533021414433e-06, "loss": 0.8214, "step": 1905 }, { "epoch": 0.68, "learning_rate": 8.33363325535361e-06, "loss": 0.8507, "step": 1906 }, { "epoch": 0.68, "learning_rate": 8.332733489292784e-06, "loss": 0.8987, "step": 1907 }, { "epoch": 0.68, "learning_rate": 8.331833723231961e-06, "loss": 0.8418, "step": 1908 }, { "epoch": 0.69, "learning_rate": 8.330933957171137e-06, "loss": 0.8985, "step": 1909 }, { "epoch": 0.69, "learning_rate": 8.330034191110312e-06, "loss": 0.8055, "step": 1910 }, { "epoch": 0.69, "learning_rate": 8.329134425049487e-06, "loss": 0.9042, "step": 1911 }, { "epoch": 0.69, "learning_rate": 8.328234658988664e-06, "loss": 0.8423, "step": 1912 }, { "epoch": 0.69, "learning_rate": 8.327334892927839e-06, "loss": 0.9227, "step": 1913 }, { "epoch": 0.69, "learning_rate": 8.326435126867015e-06, "loss": 0.7921, "step": 1914 }, { "epoch": 0.69, "learning_rate": 8.32553536080619e-06, "loss": 0.791, "step": 1915 }, { "epoch": 0.69, "learning_rate": 8.324635594745367e-06, "loss": 0.8989, "step": 1916 }, { "epoch": 0.69, "learning_rate": 8.323735828684542e-06, "loss": 0.8934, "step": 1917 }, { "epoch": 0.69, "learning_rate": 8.322836062623719e-06, "loss": 0.7592, "step": 1918 }, { "epoch": 0.69, "learning_rate": 8.321936296562895e-06, "loss": 0.8902, "step": 1919 }, { "epoch": 0.69, "learning_rate": 8.32103653050207e-06, "loss": 0.7531, "step": 1920 }, { "epoch": 0.69, "learning_rate": 8.320136764441245e-06, "loss": 0.7358, "step": 1921 }, { "epoch": 0.69, "learning_rate": 8.319236998380422e-06, "loss": 0.9614, "step": 1922 }, { "epoch": 0.69, "learning_rate": 8.318337232319598e-06, "loss": 0.7651, "step": 1923 }, { "epoch": 0.69, "learning_rate": 8.317437466258773e-06, "loss": 0.721, "step": 1924 }, { "epoch": 0.69, "learning_rate": 8.31653770019795e-06, "loss": 0.7773, "step": 1925 }, { "epoch": 0.69, "learning_rate": 8.315637934137125e-06, "loss": 0.8575, "step": 1926 }, { "epoch": 0.69, "learning_rate": 8.314738168076301e-06, "loss": 0.9016, "step": 1927 }, { "epoch": 0.69, "learning_rate": 8.313838402015478e-06, "loss": 0.8288, "step": 1928 }, { "epoch": 0.69, "learning_rate": 8.312938635954653e-06, "loss": 0.7598, "step": 1929 }, { "epoch": 0.69, "learning_rate": 8.312038869893828e-06, "loss": 0.8372, "step": 1930 }, { "epoch": 0.69, "learning_rate": 8.311139103833004e-06, "loss": 0.8953, "step": 1931 }, { "epoch": 0.69, "learning_rate": 8.310239337772181e-06, "loss": 0.9059, "step": 1932 }, { "epoch": 0.69, "learning_rate": 8.309339571711356e-06, "loss": 0.8556, "step": 1933 }, { "epoch": 0.69, "learning_rate": 8.308439805650531e-06, "loss": 0.9, "step": 1934 }, { "epoch": 0.69, "learning_rate": 8.307540039589707e-06, "loss": 0.939, "step": 1935 }, { "epoch": 0.69, "learning_rate": 8.306640273528882e-06, "loss": 0.9321, "step": 1936 }, { "epoch": 0.7, "learning_rate": 8.305740507468059e-06, "loss": 0.8362, "step": 1937 }, { "epoch": 0.7, "learning_rate": 8.304840741407236e-06, "loss": 0.8824, "step": 1938 }, { "epoch": 0.7, "learning_rate": 8.30394097534641e-06, "loss": 0.8228, "step": 1939 }, { "epoch": 0.7, "learning_rate": 8.303041209285585e-06, "loss": 0.8531, "step": 1940 }, { "epoch": 0.7, "learning_rate": 8.302141443224762e-06, "loss": 1.0432, "step": 1941 }, { "epoch": 0.7, "learning_rate": 8.301241677163939e-06, "loss": 0.9161, "step": 1942 }, { "epoch": 0.7, "learning_rate": 8.300341911103114e-06, "loss": 0.9736, "step": 1943 }, { "epoch": 0.7, "learning_rate": 8.29944214504229e-06, "loss": 0.8051, "step": 1944 }, { "epoch": 0.7, "learning_rate": 8.298542378981465e-06, "loss": 0.8568, "step": 1945 }, { "epoch": 0.7, "learning_rate": 8.297642612920642e-06, "loss": 1.0188, "step": 1946 }, { "epoch": 0.7, "learning_rate": 8.296742846859818e-06, "loss": 0.8224, "step": 1947 }, { "epoch": 0.7, "learning_rate": 8.295843080798993e-06, "loss": 0.9523, "step": 1948 }, { "epoch": 0.7, "learning_rate": 8.294943314738168e-06, "loss": 0.7948, "step": 1949 }, { "epoch": 0.7, "learning_rate": 8.294043548677345e-06, "loss": 0.8806, "step": 1950 }, { "epoch": 0.7, "learning_rate": 8.293143782616521e-06, "loss": 0.8125, "step": 1951 }, { "epoch": 0.7, "learning_rate": 8.292244016555696e-06, "loss": 0.8264, "step": 1952 }, { "epoch": 0.7, "learning_rate": 8.291344250494871e-06, "loss": 0.8215, "step": 1953 }, { "epoch": 0.7, "learning_rate": 8.290444484434048e-06, "loss": 0.8264, "step": 1954 }, { "epoch": 0.7, "learning_rate": 8.289544718373223e-06, "loss": 0.751, "step": 1955 }, { "epoch": 0.7, "learning_rate": 8.2886449523124e-06, "loss": 0.8361, "step": 1956 }, { "epoch": 0.7, "learning_rate": 8.287745186251576e-06, "loss": 0.8847, "step": 1957 }, { "epoch": 0.7, "learning_rate": 8.286845420190751e-06, "loss": 0.8025, "step": 1958 }, { "epoch": 0.7, "learning_rate": 8.285945654129926e-06, "loss": 0.8045, "step": 1959 }, { "epoch": 0.7, "learning_rate": 8.285045888069103e-06, "loss": 1.137, "step": 1960 }, { "epoch": 0.7, "learning_rate": 8.284146122008279e-06, "loss": 0.8755, "step": 1961 }, { "epoch": 0.7, "learning_rate": 8.283246355947454e-06, "loss": 0.7794, "step": 1962 }, { "epoch": 0.7, "learning_rate": 8.28234658988663e-06, "loss": 0.7818, "step": 1963 }, { "epoch": 0.7, "learning_rate": 8.281446823825806e-06, "loss": 0.9657, "step": 1964 }, { "epoch": 0.71, "learning_rate": 8.280547057764982e-06, "loss": 0.9481, "step": 1965 }, { "epoch": 0.71, "learning_rate": 8.279647291704157e-06, "loss": 0.7876, "step": 1966 }, { "epoch": 0.71, "learning_rate": 8.278747525643334e-06, "loss": 0.8312, "step": 1967 }, { "epoch": 0.71, "learning_rate": 8.277847759582509e-06, "loss": 0.9149, "step": 1968 }, { "epoch": 0.71, "learning_rate": 8.276947993521685e-06, "loss": 0.9192, "step": 1969 }, { "epoch": 0.71, "learning_rate": 8.276048227460862e-06, "loss": 0.8802, "step": 1970 }, { "epoch": 0.71, "learning_rate": 8.275148461400037e-06, "loss": 0.8953, "step": 1971 }, { "epoch": 0.71, "learning_rate": 8.274248695339212e-06, "loss": 0.8433, "step": 1972 }, { "epoch": 0.71, "learning_rate": 8.273348929278388e-06, "loss": 0.7781, "step": 1973 }, { "epoch": 0.71, "learning_rate": 8.272449163217563e-06, "loss": 0.8472, "step": 1974 }, { "epoch": 0.71, "learning_rate": 8.27154939715674e-06, "loss": 0.805, "step": 1975 }, { "epoch": 0.71, "learning_rate": 8.270649631095917e-06, "loss": 0.8943, "step": 1976 }, { "epoch": 0.71, "learning_rate": 8.269749865035091e-06, "loss": 0.7864, "step": 1977 }, { "epoch": 0.71, "learning_rate": 8.268850098974266e-06, "loss": 0.676, "step": 1978 }, { "epoch": 0.71, "learning_rate": 8.267950332913443e-06, "loss": 0.8703, "step": 1979 }, { "epoch": 0.71, "learning_rate": 8.26705056685262e-06, "loss": 0.8011, "step": 1980 }, { "epoch": 0.71, "learning_rate": 8.266150800791795e-06, "loss": 0.8471, "step": 1981 }, { "epoch": 0.71, "learning_rate": 8.265251034730971e-06, "loss": 0.7653, "step": 1982 }, { "epoch": 0.71, "learning_rate": 8.264351268670146e-06, "loss": 0.8451, "step": 1983 }, { "epoch": 0.71, "learning_rate": 8.263451502609323e-06, "loss": 0.9169, "step": 1984 }, { "epoch": 0.71, "learning_rate": 8.262551736548498e-06, "loss": 0.8386, "step": 1985 }, { "epoch": 0.71, "learning_rate": 8.261651970487674e-06, "loss": 0.7944, "step": 1986 }, { "epoch": 0.71, "learning_rate": 8.260752204426849e-06, "loss": 0.8999, "step": 1987 }, { "epoch": 0.71, "learning_rate": 8.259852438366026e-06, "loss": 0.8468, "step": 1988 }, { "epoch": 0.71, "learning_rate": 8.258952672305202e-06, "loss": 0.7384, "step": 1989 }, { "epoch": 0.71, "learning_rate": 8.258052906244377e-06, "loss": 0.785, "step": 1990 }, { "epoch": 0.71, "learning_rate": 8.257153140183552e-06, "loss": 0.9549, "step": 1991 }, { "epoch": 0.72, "learning_rate": 8.256253374122729e-06, "loss": 0.8747, "step": 1992 }, { "epoch": 0.72, "learning_rate": 8.255353608061904e-06, "loss": 0.8237, "step": 1993 }, { "epoch": 0.72, "learning_rate": 8.25445384200108e-06, "loss": 0.8296, "step": 1994 }, { "epoch": 0.72, "learning_rate": 8.253554075940257e-06, "loss": 0.837, "step": 1995 }, { "epoch": 0.72, "learning_rate": 8.252654309879432e-06, "loss": 0.949, "step": 1996 }, { "epoch": 0.72, "learning_rate": 8.251754543818607e-06, "loss": 0.8004, "step": 1997 }, { "epoch": 0.72, "learning_rate": 8.250854777757783e-06, "loss": 0.8256, "step": 1998 }, { "epoch": 0.72, "learning_rate": 8.24995501169696e-06, "loss": 0.9768, "step": 1999 }, { "epoch": 0.72, "learning_rate": 8.249055245636135e-06, "loss": 1.0018, "step": 2000 }, { "epoch": 0.72, "learning_rate": 8.248155479575312e-06, "loss": 0.7787, "step": 2001 }, { "epoch": 0.72, "learning_rate": 8.247255713514487e-06, "loss": 0.8587, "step": 2002 }, { "epoch": 0.72, "learning_rate": 8.246355947453663e-06, "loss": 0.9863, "step": 2003 }, { "epoch": 0.72, "learning_rate": 8.245456181392838e-06, "loss": 0.7859, "step": 2004 }, { "epoch": 0.72, "learning_rate": 8.244556415332015e-06, "loss": 0.8706, "step": 2005 }, { "epoch": 0.72, "learning_rate": 8.24365664927119e-06, "loss": 0.9449, "step": 2006 }, { "epoch": 0.72, "learning_rate": 8.242756883210366e-06, "loss": 0.9021, "step": 2007 }, { "epoch": 0.72, "learning_rate": 8.241857117149543e-06, "loss": 0.933, "step": 2008 }, { "epoch": 0.72, "learning_rate": 8.240957351088718e-06, "loss": 0.8094, "step": 2009 }, { "epoch": 0.72, "learning_rate": 8.240057585027893e-06, "loss": 0.8268, "step": 2010 }, { "epoch": 0.72, "learning_rate": 8.23915781896707e-06, "loss": 0.911, "step": 2011 }, { "epoch": 0.72, "learning_rate": 8.238258052906244e-06, "loss": 0.8081, "step": 2012 }, { "epoch": 0.72, "learning_rate": 8.23735828684542e-06, "loss": 0.8029, "step": 2013 }, { "epoch": 0.72, "learning_rate": 8.236458520784597e-06, "loss": 0.8901, "step": 2014 }, { "epoch": 0.72, "learning_rate": 8.235558754723772e-06, "loss": 0.8654, "step": 2015 }, { "epoch": 0.72, "learning_rate": 8.234658988662947e-06, "loss": 0.8729, "step": 2016 }, { "epoch": 0.72, "learning_rate": 8.233759222602124e-06, "loss": 0.8111, "step": 2017 }, { "epoch": 0.72, "learning_rate": 8.2328594565413e-06, "loss": 0.9259, "step": 2018 }, { "epoch": 0.72, "learning_rate": 8.231959690480475e-06, "loss": 0.8269, "step": 2019 }, { "epoch": 0.73, "learning_rate": 8.231059924419652e-06, "loss": 0.8387, "step": 2020 }, { "epoch": 0.73, "learning_rate": 8.230160158358827e-06, "loss": 0.8348, "step": 2021 }, { "epoch": 0.73, "learning_rate": 8.229260392298004e-06, "loss": 0.8121, "step": 2022 }, { "epoch": 0.73, "learning_rate": 8.228360626237179e-06, "loss": 0.7665, "step": 2023 }, { "epoch": 0.73, "learning_rate": 8.227460860176355e-06, "loss": 0.8044, "step": 2024 }, { "epoch": 0.73, "learning_rate": 8.22656109411553e-06, "loss": 1.0054, "step": 2025 }, { "epoch": 0.73, "learning_rate": 8.225661328054707e-06, "loss": 0.848, "step": 2026 }, { "epoch": 0.73, "learning_rate": 8.224761561993883e-06, "loss": 0.9301, "step": 2027 }, { "epoch": 0.73, "learning_rate": 8.223861795933058e-06, "loss": 0.884, "step": 2028 }, { "epoch": 0.73, "learning_rate": 8.222962029872233e-06, "loss": 0.8338, "step": 2029 }, { "epoch": 0.73, "learning_rate": 8.22206226381141e-06, "loss": 0.8181, "step": 2030 }, { "epoch": 0.73, "learning_rate": 8.221162497750585e-06, "loss": 0.8713, "step": 2031 }, { "epoch": 0.73, "learning_rate": 8.220262731689761e-06, "loss": 0.872, "step": 2032 }, { "epoch": 0.73, "learning_rate": 8.219362965628938e-06, "loss": 0.7808, "step": 2033 }, { "epoch": 0.73, "learning_rate": 8.218463199568113e-06, "loss": 0.8749, "step": 2034 }, { "epoch": 0.73, "learning_rate": 8.217563433507288e-06, "loss": 0.9208, "step": 2035 }, { "epoch": 0.73, "learning_rate": 8.216663667446464e-06, "loss": 0.7799, "step": 2036 }, { "epoch": 0.73, "learning_rate": 8.215763901385641e-06, "loss": 0.8437, "step": 2037 }, { "epoch": 0.73, "learning_rate": 8.214864135324816e-06, "loss": 0.7407, "step": 2038 }, { "epoch": 0.73, "learning_rate": 8.213964369263993e-06, "loss": 0.8458, "step": 2039 }, { "epoch": 0.73, "learning_rate": 8.213064603203167e-06, "loss": 0.7861, "step": 2040 }, { "epoch": 0.73, "learning_rate": 8.212164837142344e-06, "loss": 0.754, "step": 2041 }, { "epoch": 0.73, "learning_rate": 8.211265071081519e-06, "loss": 0.9791, "step": 2042 }, { "epoch": 0.73, "learning_rate": 8.210365305020696e-06, "loss": 0.869, "step": 2043 }, { "epoch": 0.73, "learning_rate": 8.20946553895987e-06, "loss": 0.7959, "step": 2044 }, { "epoch": 0.73, "learning_rate": 8.208565772899047e-06, "loss": 0.8331, "step": 2045 }, { "epoch": 0.73, "learning_rate": 8.207666006838224e-06, "loss": 0.9073, "step": 2046 }, { "epoch": 0.73, "learning_rate": 8.206766240777399e-06, "loss": 0.8424, "step": 2047 }, { "epoch": 0.74, "learning_rate": 8.205866474716574e-06, "loss": 0.8245, "step": 2048 }, { "epoch": 0.74, "learning_rate": 8.20496670865575e-06, "loss": 0.8675, "step": 2049 }, { "epoch": 0.74, "learning_rate": 8.204066942594927e-06, "loss": 0.9194, "step": 2050 }, { "epoch": 0.74, "learning_rate": 8.203167176534102e-06, "loss": 0.8408, "step": 2051 }, { "epoch": 0.74, "learning_rate": 8.202267410473278e-06, "loss": 0.9644, "step": 2052 }, { "epoch": 0.74, "learning_rate": 8.201367644412453e-06, "loss": 0.8127, "step": 2053 }, { "epoch": 0.74, "learning_rate": 8.200467878351628e-06, "loss": 0.8153, "step": 2054 }, { "epoch": 0.74, "learning_rate": 8.199568112290805e-06, "loss": 0.8058, "step": 2055 }, { "epoch": 0.74, "learning_rate": 8.198668346229981e-06, "loss": 0.9007, "step": 2056 }, { "epoch": 0.74, "learning_rate": 8.197768580169156e-06, "loss": 0.8291, "step": 2057 }, { "epoch": 0.74, "learning_rate": 8.196868814108333e-06, "loss": 0.8372, "step": 2058 }, { "epoch": 0.74, "learning_rate": 8.195969048047508e-06, "loss": 0.8333, "step": 2059 }, { "epoch": 0.74, "learning_rate": 8.195069281986684e-06, "loss": 0.8315, "step": 2060 }, { "epoch": 0.74, "learning_rate": 8.19416951592586e-06, "loss": 0.7645, "step": 2061 }, { "epoch": 0.74, "learning_rate": 8.193269749865036e-06, "loss": 0.7792, "step": 2062 }, { "epoch": 0.74, "learning_rate": 8.192369983804211e-06, "loss": 0.8149, "step": 2063 }, { "epoch": 0.74, "learning_rate": 8.191470217743388e-06, "loss": 0.8846, "step": 2064 }, { "epoch": 0.74, "learning_rate": 8.190570451682564e-06, "loss": 0.8303, "step": 2065 }, { "epoch": 0.74, "learning_rate": 8.189670685621739e-06, "loss": 0.7782, "step": 2066 }, { "epoch": 0.74, "learning_rate": 8.188770919560914e-06, "loss": 0.8306, "step": 2067 }, { "epoch": 0.74, "learning_rate": 8.18787115350009e-06, "loss": 0.8665, "step": 2068 }, { "epoch": 0.74, "learning_rate": 8.186971387439267e-06, "loss": 0.7918, "step": 2069 }, { "epoch": 0.74, "learning_rate": 8.186071621378442e-06, "loss": 0.8359, "step": 2070 }, { "epoch": 0.74, "learning_rate": 8.185171855317619e-06, "loss": 0.7797, "step": 2071 }, { "epoch": 0.74, "learning_rate": 8.184272089256794e-06, "loss": 0.8882, "step": 2072 }, { "epoch": 0.74, "learning_rate": 8.183372323195969e-06, "loss": 0.9617, "step": 2073 }, { "epoch": 0.74, "learning_rate": 8.182472557135145e-06, "loss": 0.7523, "step": 2074 }, { "epoch": 0.74, "learning_rate": 8.181572791074322e-06, "loss": 0.8061, "step": 2075 }, { "epoch": 0.75, "learning_rate": 8.180673025013497e-06, "loss": 0.9257, "step": 2076 }, { "epoch": 0.75, "learning_rate": 8.179773258952673e-06, "loss": 0.8328, "step": 2077 }, { "epoch": 0.75, "learning_rate": 8.178873492891848e-06, "loss": 0.7761, "step": 2078 }, { "epoch": 0.75, "learning_rate": 8.177973726831025e-06, "loss": 0.8013, "step": 2079 }, { "epoch": 0.75, "learning_rate": 8.1770739607702e-06, "loss": 0.7566, "step": 2080 }, { "epoch": 0.75, "learning_rate": 8.176174194709376e-06, "loss": 0.7458, "step": 2081 }, { "epoch": 0.75, "learning_rate": 8.175274428648551e-06, "loss": 0.8173, "step": 2082 }, { "epoch": 0.75, "learning_rate": 8.174374662587728e-06, "loss": 0.8424, "step": 2083 }, { "epoch": 0.75, "learning_rate": 8.173474896526905e-06, "loss": 0.7528, "step": 2084 }, { "epoch": 0.75, "learning_rate": 8.17257513046608e-06, "loss": 0.7109, "step": 2085 }, { "epoch": 0.75, "learning_rate": 8.171675364405254e-06, "loss": 0.9058, "step": 2086 }, { "epoch": 0.75, "learning_rate": 8.170775598344431e-06, "loss": 0.8595, "step": 2087 }, { "epoch": 0.75, "learning_rate": 8.169875832283608e-06, "loss": 0.7937, "step": 2088 }, { "epoch": 0.75, "learning_rate": 8.168976066222783e-06, "loss": 0.7158, "step": 2089 }, { "epoch": 0.75, "learning_rate": 8.16807630016196e-06, "loss": 0.8324, "step": 2090 }, { "epoch": 0.75, "learning_rate": 8.167176534101134e-06, "loss": 0.8003, "step": 2091 }, { "epoch": 0.75, "learning_rate": 8.166276768040309e-06, "loss": 0.8262, "step": 2092 }, { "epoch": 0.75, "learning_rate": 8.165377001979486e-06, "loss": 0.8772, "step": 2093 }, { "epoch": 0.75, "learning_rate": 8.164477235918662e-06, "loss": 0.8683, "step": 2094 }, { "epoch": 0.75, "learning_rate": 8.163577469857837e-06, "loss": 0.8987, "step": 2095 }, { "epoch": 0.75, "learning_rate": 8.162677703797014e-06, "loss": 0.7994, "step": 2096 }, { "epoch": 0.75, "learning_rate": 8.161777937736189e-06, "loss": 0.8746, "step": 2097 }, { "epoch": 0.75, "learning_rate": 8.160878171675365e-06, "loss": 0.8351, "step": 2098 }, { "epoch": 0.75, "learning_rate": 8.15997840561454e-06, "loss": 0.8139, "step": 2099 }, { "epoch": 0.75, "learning_rate": 8.159078639553717e-06, "loss": 0.7685, "step": 2100 }, { "epoch": 0.75, "learning_rate": 8.158178873492892e-06, "loss": 0.8852, "step": 2101 }, { "epoch": 0.75, "learning_rate": 8.157279107432068e-06, "loss": 0.8016, "step": 2102 }, { "epoch": 0.75, "learning_rate": 8.156379341371245e-06, "loss": 0.8627, "step": 2103 }, { "epoch": 0.76, "learning_rate": 8.15547957531042e-06, "loss": 0.8416, "step": 2104 }, { "epoch": 0.76, "learning_rate": 8.154579809249595e-06, "loss": 0.8739, "step": 2105 }, { "epoch": 0.76, "learning_rate": 8.153680043188772e-06, "loss": 0.78, "step": 2106 }, { "epoch": 0.76, "learning_rate": 8.152780277127948e-06, "loss": 0.9008, "step": 2107 }, { "epoch": 0.76, "learning_rate": 8.151880511067123e-06, "loss": 0.7262, "step": 2108 }, { "epoch": 0.76, "learning_rate": 8.1509807450063e-06, "loss": 0.7806, "step": 2109 }, { "epoch": 0.76, "learning_rate": 8.150080978945475e-06, "loss": 0.9227, "step": 2110 }, { "epoch": 0.76, "learning_rate": 8.14918121288465e-06, "loss": 0.7806, "step": 2111 }, { "epoch": 0.76, "learning_rate": 8.148281446823826e-06, "loss": 0.8668, "step": 2112 }, { "epoch": 0.76, "learning_rate": 8.147381680763003e-06, "loss": 0.7958, "step": 2113 }, { "epoch": 0.76, "learning_rate": 8.146481914702178e-06, "loss": 0.7026, "step": 2114 }, { "epoch": 0.76, "learning_rate": 8.145582148641354e-06, "loss": 0.9217, "step": 2115 }, { "epoch": 0.76, "learning_rate": 8.14468238258053e-06, "loss": 0.8271, "step": 2116 }, { "epoch": 0.76, "learning_rate": 8.143782616519706e-06, "loss": 0.8641, "step": 2117 }, { "epoch": 0.76, "learning_rate": 8.14288285045888e-06, "loss": 0.7691, "step": 2118 }, { "epoch": 0.76, "learning_rate": 8.141983084398057e-06, "loss": 0.9167, "step": 2119 }, { "epoch": 0.76, "learning_rate": 8.141083318337232e-06, "loss": 0.7989, "step": 2120 }, { "epoch": 0.76, "learning_rate": 8.140183552276409e-06, "loss": 0.7545, "step": 2121 }, { "epoch": 0.76, "learning_rate": 8.139283786215586e-06, "loss": 0.8954, "step": 2122 }, { "epoch": 0.76, "learning_rate": 8.13838402015476e-06, "loss": 0.8581, "step": 2123 }, { "epoch": 0.76, "learning_rate": 8.137484254093935e-06, "loss": 0.8294, "step": 2124 }, { "epoch": 0.76, "learning_rate": 8.136584488033112e-06, "loss": 0.8708, "step": 2125 }, { "epoch": 0.76, "learning_rate": 8.135684721972289e-06, "loss": 0.8924, "step": 2126 }, { "epoch": 0.76, "learning_rate": 8.134784955911464e-06, "loss": 0.7606, "step": 2127 }, { "epoch": 0.76, "learning_rate": 8.13388518985064e-06, "loss": 0.7381, "step": 2128 }, { "epoch": 0.76, "learning_rate": 8.132985423789815e-06, "loss": 0.8173, "step": 2129 }, { "epoch": 0.76, "learning_rate": 8.13208565772899e-06, "loss": 0.809, "step": 2130 }, { "epoch": 0.76, "learning_rate": 8.131185891668167e-06, "loss": 0.9361, "step": 2131 }, { "epoch": 0.77, "learning_rate": 8.130286125607343e-06, "loss": 0.8279, "step": 2132 }, { "epoch": 0.77, "learning_rate": 8.129386359546518e-06, "loss": 0.8441, "step": 2133 }, { "epoch": 0.77, "learning_rate": 8.128486593485695e-06, "loss": 0.8057, "step": 2134 }, { "epoch": 0.77, "learning_rate": 8.12758682742487e-06, "loss": 0.7208, "step": 2135 }, { "epoch": 0.77, "learning_rate": 8.126687061364046e-06, "loss": 1.0614, "step": 2136 }, { "epoch": 0.77, "learning_rate": 8.125787295303221e-06, "loss": 0.7396, "step": 2137 }, { "epoch": 0.77, "learning_rate": 8.124887529242398e-06, "loss": 0.8357, "step": 2138 }, { "epoch": 0.77, "learning_rate": 8.123987763181573e-06, "loss": 0.7852, "step": 2139 }, { "epoch": 0.77, "learning_rate": 8.12308799712075e-06, "loss": 0.8029, "step": 2140 }, { "epoch": 0.77, "learning_rate": 8.122188231059926e-06, "loss": 0.7791, "step": 2141 }, { "epoch": 0.77, "learning_rate": 8.121288464999101e-06, "loss": 0.8065, "step": 2142 }, { "epoch": 0.77, "learning_rate": 8.120388698938276e-06, "loss": 0.8817, "step": 2143 }, { "epoch": 0.77, "learning_rate": 8.119488932877452e-06, "loss": 0.8821, "step": 2144 }, { "epoch": 0.77, "learning_rate": 8.118589166816629e-06, "loss": 0.8906, "step": 2145 }, { "epoch": 0.77, "learning_rate": 8.117689400755804e-06, "loss": 0.7437, "step": 2146 }, { "epoch": 0.77, "learning_rate": 8.11678963469498e-06, "loss": 1.0025, "step": 2147 }, { "epoch": 0.77, "learning_rate": 8.115889868634156e-06, "loss": 0.8819, "step": 2148 }, { "epoch": 0.77, "learning_rate": 8.114990102573332e-06, "loss": 0.9591, "step": 2149 }, { "epoch": 0.77, "learning_rate": 8.114090336512507e-06, "loss": 0.913, "step": 2150 }, { "epoch": 0.77, "learning_rate": 8.113190570451684e-06, "loss": 0.7749, "step": 2151 }, { "epoch": 0.77, "learning_rate": 8.112290804390859e-06, "loss": 0.9442, "step": 2152 }, { "epoch": 0.77, "learning_rate": 8.111391038330035e-06, "loss": 0.7831, "step": 2153 }, { "epoch": 0.77, "learning_rate": 8.11049127226921e-06, "loss": 0.7699, "step": 2154 }, { "epoch": 0.77, "learning_rate": 8.109591506208387e-06, "loss": 0.9611, "step": 2155 }, { "epoch": 0.77, "learning_rate": 8.108691740147562e-06, "loss": 0.7361, "step": 2156 }, { "epoch": 0.77, "learning_rate": 8.107791974086738e-06, "loss": 0.883, "step": 2157 }, { "epoch": 0.77, "learning_rate": 8.106892208025913e-06, "loss": 0.8987, "step": 2158 }, { "epoch": 0.77, "learning_rate": 8.10599244196509e-06, "loss": 0.8311, "step": 2159 }, { "epoch": 0.78, "learning_rate": 8.105092675904266e-06, "loss": 0.9262, "step": 2160 }, { "epoch": 0.78, "learning_rate": 8.104192909843441e-06, "loss": 0.8263, "step": 2161 }, { "epoch": 0.78, "learning_rate": 8.103293143782616e-06, "loss": 0.701, "step": 2162 }, { "epoch": 0.78, "learning_rate": 8.102393377721793e-06, "loss": 0.844, "step": 2163 }, { "epoch": 0.78, "learning_rate": 8.10149361166097e-06, "loss": 0.7638, "step": 2164 }, { "epoch": 0.78, "learning_rate": 8.100593845600144e-06, "loss": 0.7645, "step": 2165 }, { "epoch": 0.78, "learning_rate": 8.099694079539321e-06, "loss": 0.8653, "step": 2166 }, { "epoch": 0.78, "learning_rate": 8.098794313478496e-06, "loss": 0.7235, "step": 2167 }, { "epoch": 0.78, "learning_rate": 8.097894547417673e-06, "loss": 0.8301, "step": 2168 }, { "epoch": 0.78, "learning_rate": 8.096994781356848e-06, "loss": 0.8142, "step": 2169 }, { "epoch": 0.78, "learning_rate": 8.096095015296024e-06, "loss": 0.9835, "step": 2170 }, { "epoch": 0.78, "learning_rate": 8.095195249235199e-06, "loss": 0.8482, "step": 2171 }, { "epoch": 0.78, "learning_rate": 8.094295483174376e-06, "loss": 0.7629, "step": 2172 }, { "epoch": 0.78, "learning_rate": 8.09339571711355e-06, "loss": 0.8915, "step": 2173 }, { "epoch": 0.78, "learning_rate": 8.092495951052727e-06, "loss": 0.8118, "step": 2174 }, { "epoch": 0.78, "learning_rate": 8.091596184991902e-06, "loss": 0.7772, "step": 2175 }, { "epoch": 0.78, "learning_rate": 8.090696418931079e-06, "loss": 0.769, "step": 2176 }, { "epoch": 0.78, "learning_rate": 8.089796652870254e-06, "loss": 0.8166, "step": 2177 }, { "epoch": 0.78, "learning_rate": 8.08889688680943e-06, "loss": 0.9454, "step": 2178 }, { "epoch": 0.78, "learning_rate": 8.087997120748607e-06, "loss": 0.8704, "step": 2179 }, { "epoch": 0.78, "learning_rate": 8.087097354687782e-06, "loss": 0.8449, "step": 2180 }, { "epoch": 0.78, "learning_rate": 8.086197588626957e-06, "loss": 0.7894, "step": 2181 }, { "epoch": 0.78, "learning_rate": 8.085297822566133e-06, "loss": 0.8476, "step": 2182 }, { "epoch": 0.78, "learning_rate": 8.08439805650531e-06, "loss": 0.7776, "step": 2183 }, { "epoch": 0.78, "learning_rate": 8.083498290444485e-06, "loss": 0.782, "step": 2184 }, { "epoch": 0.78, "learning_rate": 8.082598524383662e-06, "loss": 0.7714, "step": 2185 }, { "epoch": 0.78, "learning_rate": 8.081698758322836e-06, "loss": 0.9373, "step": 2186 }, { "epoch": 0.78, "learning_rate": 8.080798992262013e-06, "loss": 0.8407, "step": 2187 }, { "epoch": 0.79, "learning_rate": 8.079899226201188e-06, "loss": 0.8854, "step": 2188 }, { "epoch": 0.79, "learning_rate": 8.078999460140365e-06, "loss": 0.6959, "step": 2189 }, { "epoch": 0.79, "learning_rate": 8.07809969407954e-06, "loss": 0.7769, "step": 2190 }, { "epoch": 0.79, "learning_rate": 8.077199928018716e-06, "loss": 0.8969, "step": 2191 }, { "epoch": 0.79, "learning_rate": 8.076300161957891e-06, "loss": 0.8712, "step": 2192 }, { "epoch": 0.79, "learning_rate": 8.075400395897068e-06, "loss": 0.8821, "step": 2193 }, { "epoch": 0.79, "learning_rate": 8.074500629836243e-06, "loss": 0.6942, "step": 2194 }, { "epoch": 0.79, "learning_rate": 8.07360086377542e-06, "loss": 0.8976, "step": 2195 }, { "epoch": 0.79, "learning_rate": 8.072701097714594e-06, "loss": 0.8678, "step": 2196 }, { "epoch": 0.79, "learning_rate": 8.07180133165377e-06, "loss": 0.7863, "step": 2197 }, { "epoch": 0.79, "learning_rate": 8.070901565592947e-06, "loss": 0.8344, "step": 2198 }, { "epoch": 0.79, "learning_rate": 8.070001799532122e-06, "loss": 0.8413, "step": 2199 }, { "epoch": 0.79, "learning_rate": 8.069102033471297e-06, "loss": 0.8068, "step": 2200 }, { "epoch": 0.79, "learning_rate": 8.068202267410474e-06, "loss": 0.7558, "step": 2201 }, { "epoch": 0.79, "learning_rate": 8.06730250134965e-06, "loss": 0.813, "step": 2202 }, { "epoch": 0.79, "learning_rate": 8.066402735288825e-06, "loss": 0.8902, "step": 2203 }, { "epoch": 0.79, "learning_rate": 8.065502969228002e-06, "loss": 0.9735, "step": 2204 }, { "epoch": 0.79, "learning_rate": 8.064603203167177e-06, "loss": 0.8583, "step": 2205 }, { "epoch": 0.79, "learning_rate": 8.063703437106354e-06, "loss": 0.8228, "step": 2206 }, { "epoch": 0.79, "learning_rate": 8.06280367104553e-06, "loss": 0.7657, "step": 2207 }, { "epoch": 0.79, "learning_rate": 8.061903904984705e-06, "loss": 0.8557, "step": 2208 }, { "epoch": 0.79, "learning_rate": 8.06100413892388e-06, "loss": 0.8157, "step": 2209 }, { "epoch": 0.79, "learning_rate": 8.060104372863057e-06, "loss": 0.8204, "step": 2210 }, { "epoch": 0.79, "learning_rate": 8.059204606802232e-06, "loss": 0.9405, "step": 2211 }, { "epoch": 0.79, "learning_rate": 8.058304840741408e-06, "loss": 0.7286, "step": 2212 }, { "epoch": 0.79, "learning_rate": 8.057405074680583e-06, "loss": 0.8857, "step": 2213 }, { "epoch": 0.79, "learning_rate": 8.05650530861976e-06, "loss": 0.8198, "step": 2214 }, { "epoch": 0.8, "learning_rate": 8.055605542558935e-06, "loss": 0.8265, "step": 2215 }, { "epoch": 0.8, "learning_rate": 8.054705776498111e-06, "loss": 0.9097, "step": 2216 }, { "epoch": 0.8, "learning_rate": 8.053806010437288e-06, "loss": 0.8965, "step": 2217 }, { "epoch": 0.8, "learning_rate": 8.052906244376463e-06, "loss": 0.8354, "step": 2218 }, { "epoch": 0.8, "learning_rate": 8.052006478315638e-06, "loss": 0.8569, "step": 2219 }, { "epoch": 0.8, "learning_rate": 8.051106712254814e-06, "loss": 0.872, "step": 2220 }, { "epoch": 0.8, "learning_rate": 8.050206946193991e-06, "loss": 0.8304, "step": 2221 }, { "epoch": 0.8, "learning_rate": 8.049307180133166e-06, "loss": 0.8042, "step": 2222 }, { "epoch": 0.8, "learning_rate": 8.048407414072342e-06, "loss": 0.8377, "step": 2223 }, { "epoch": 0.8, "learning_rate": 8.047507648011517e-06, "loss": 0.9116, "step": 2224 }, { "epoch": 0.8, "learning_rate": 8.046607881950694e-06, "loss": 0.9892, "step": 2225 }, { "epoch": 0.8, "learning_rate": 8.04570811588987e-06, "loss": 0.7957, "step": 2226 }, { "epoch": 0.8, "learning_rate": 8.044808349829046e-06, "loss": 0.7727, "step": 2227 }, { "epoch": 0.8, "learning_rate": 8.04390858376822e-06, "loss": 0.7775, "step": 2228 }, { "epoch": 0.8, "learning_rate": 8.043008817707397e-06, "loss": 0.8282, "step": 2229 }, { "epoch": 0.8, "learning_rate": 8.042109051646572e-06, "loss": 0.8605, "step": 2230 }, { "epoch": 0.8, "learning_rate": 8.041209285585749e-06, "loss": 0.7659, "step": 2231 }, { "epoch": 0.8, "learning_rate": 8.040309519524924e-06, "loss": 0.8693, "step": 2232 }, { "epoch": 0.8, "learning_rate": 8.0394097534641e-06, "loss": 0.8378, "step": 2233 }, { "epoch": 0.8, "learning_rate": 8.038509987403275e-06, "loss": 0.8682, "step": 2234 }, { "epoch": 0.8, "learning_rate": 8.037610221342452e-06, "loss": 0.7733, "step": 2235 }, { "epoch": 0.8, "learning_rate": 8.036710455281628e-06, "loss": 0.8107, "step": 2236 }, { "epoch": 0.8, "learning_rate": 8.035810689220803e-06, "loss": 0.8668, "step": 2237 }, { "epoch": 0.8, "learning_rate": 8.034910923159978e-06, "loss": 0.8901, "step": 2238 }, { "epoch": 0.8, "learning_rate": 8.034011157099155e-06, "loss": 0.8865, "step": 2239 }, { "epoch": 0.8, "learning_rate": 8.033111391038331e-06, "loss": 0.886, "step": 2240 }, { "epoch": 0.8, "learning_rate": 8.032211624977506e-06, "loss": 0.8471, "step": 2241 }, { "epoch": 0.8, "learning_rate": 8.031311858916683e-06, "loss": 0.8785, "step": 2242 }, { "epoch": 0.81, "learning_rate": 8.030412092855858e-06, "loss": 0.8392, "step": 2243 }, { "epoch": 0.81, "learning_rate": 8.029512326795034e-06, "loss": 0.8924, "step": 2244 }, { "epoch": 0.81, "learning_rate": 8.028612560734211e-06, "loss": 0.8103, "step": 2245 }, { "epoch": 0.81, "learning_rate": 8.027712794673386e-06, "loss": 0.8644, "step": 2246 }, { "epoch": 0.81, "learning_rate": 8.026813028612561e-06, "loss": 0.8662, "step": 2247 }, { "epoch": 0.81, "learning_rate": 8.025913262551737e-06, "loss": 0.8534, "step": 2248 }, { "epoch": 0.81, "learning_rate": 8.025013496490912e-06, "loss": 0.9558, "step": 2249 }, { "epoch": 0.81, "learning_rate": 8.024113730430089e-06, "loss": 0.7997, "step": 2250 }, { "epoch": 0.81, "learning_rate": 8.023213964369264e-06, "loss": 0.7573, "step": 2251 }, { "epoch": 0.81, "learning_rate": 8.02231419830844e-06, "loss": 0.7859, "step": 2252 }, { "epoch": 0.81, "learning_rate": 8.021414432247615e-06, "loss": 0.894, "step": 2253 }, { "epoch": 0.81, "learning_rate": 8.020514666186792e-06, "loss": 0.8, "step": 2254 }, { "epoch": 0.81, "learning_rate": 8.019614900125969e-06, "loss": 0.8521, "step": 2255 }, { "epoch": 0.81, "learning_rate": 8.019614900125969e-06, "loss": 0.878, "step": 2256 }, { "epoch": 0.81, "learning_rate": 8.018715134065144e-06, "loss": 0.7617, "step": 2257 }, { "epoch": 0.81, "learning_rate": 8.017815368004319e-06, "loss": 0.9368, "step": 2258 }, { "epoch": 0.81, "learning_rate": 8.016915601943495e-06, "loss": 0.9274, "step": 2259 }, { "epoch": 0.81, "learning_rate": 8.016015835882672e-06, "loss": 0.9017, "step": 2260 }, { "epoch": 0.81, "learning_rate": 8.015116069821847e-06, "loss": 0.8941, "step": 2261 }, { "epoch": 0.81, "learning_rate": 8.014216303761023e-06, "loss": 0.8431, "step": 2262 }, { "epoch": 0.81, "learning_rate": 8.013316537700198e-06, "loss": 0.8586, "step": 2263 }, { "epoch": 0.81, "learning_rate": 8.012416771639375e-06, "loss": 0.8457, "step": 2264 }, { "epoch": 0.81, "learning_rate": 8.011517005578551e-06, "loss": 0.7306, "step": 2265 }, { "epoch": 0.81, "learning_rate": 8.010617239517726e-06, "loss": 0.8887, "step": 2266 }, { "epoch": 0.81, "learning_rate": 8.009717473456901e-06, "loss": 0.8892, "step": 2267 }, { "epoch": 0.81, "learning_rate": 8.008817707396078e-06, "loss": 0.8243, "step": 2268 }, { "epoch": 0.81, "learning_rate": 8.007917941335253e-06, "loss": 0.987, "step": 2269 }, { "epoch": 0.81, "learning_rate": 8.00701817527443e-06, "loss": 0.9308, "step": 2270 }, { "epoch": 0.82, "learning_rate": 8.006118409213604e-06, "loss": 0.8625, "step": 2271 }, { "epoch": 0.82, "learning_rate": 8.005218643152781e-06, "loss": 0.8574, "step": 2272 }, { "epoch": 0.82, "learning_rate": 8.004318877091956e-06, "loss": 0.9357, "step": 2273 }, { "epoch": 0.82, "learning_rate": 8.003419111031133e-06, "loss": 1.028, "step": 2274 }, { "epoch": 0.82, "learning_rate": 8.00251934497031e-06, "loss": 0.9053, "step": 2275 }, { "epoch": 0.82, "learning_rate": 8.001619578909484e-06, "loss": 0.9819, "step": 2276 }, { "epoch": 0.82, "learning_rate": 8.000719812848659e-06, "loss": 0.7636, "step": 2277 }, { "epoch": 0.82, "learning_rate": 7.999820046787836e-06, "loss": 0.7905, "step": 2278 }, { "epoch": 0.82, "learning_rate": 7.998920280727012e-06, "loss": 0.8633, "step": 2279 }, { "epoch": 0.82, "learning_rate": 7.998020514666187e-06, "loss": 0.8379, "step": 2280 }, { "epoch": 0.82, "learning_rate": 7.997120748605364e-06, "loss": 0.8274, "step": 2281 }, { "epoch": 0.82, "learning_rate": 7.996220982544539e-06, "loss": 0.9289, "step": 2282 }, { "epoch": 0.82, "learning_rate": 7.995321216483715e-06, "loss": 0.8229, "step": 2283 }, { "epoch": 0.82, "learning_rate": 7.994421450422892e-06, "loss": 0.8687, "step": 2284 }, { "epoch": 0.82, "learning_rate": 7.993521684362067e-06, "loss": 0.8557, "step": 2285 }, { "epoch": 0.82, "learning_rate": 7.992621918301242e-06, "loss": 0.9322, "step": 2286 }, { "epoch": 0.82, "learning_rate": 7.991722152240418e-06, "loss": 0.8394, "step": 2287 }, { "epoch": 0.82, "learning_rate": 7.990822386179593e-06, "loss": 0.8184, "step": 2288 }, { "epoch": 0.82, "learning_rate": 7.98992262011877e-06, "loss": 0.9397, "step": 2289 }, { "epoch": 0.82, "learning_rate": 7.989022854057945e-06, "loss": 0.8119, "step": 2290 }, { "epoch": 0.82, "learning_rate": 7.988123087997121e-06, "loss": 0.8169, "step": 2291 }, { "epoch": 0.82, "learning_rate": 7.987223321936296e-06, "loss": 0.7885, "step": 2292 }, { "epoch": 0.82, "learning_rate": 7.986323555875473e-06, "loss": 0.8868, "step": 2293 }, { "epoch": 0.82, "learning_rate": 7.98542378981465e-06, "loss": 0.8385, "step": 2294 }, { "epoch": 0.82, "learning_rate": 7.984524023753825e-06, "loss": 0.7386, "step": 2295 }, { "epoch": 0.82, "learning_rate": 7.983624257693e-06, "loss": 0.7492, "step": 2296 }, { "epoch": 0.82, "learning_rate": 7.982724491632176e-06, "loss": 0.8206, "step": 2297 }, { "epoch": 0.82, "learning_rate": 7.981824725571353e-06, "loss": 0.7893, "step": 2298 }, { "epoch": 0.83, "learning_rate": 7.980924959510528e-06, "loss": 0.8634, "step": 2299 }, { "epoch": 0.83, "learning_rate": 7.980025193449704e-06, "loss": 0.9275, "step": 2300 }, { "epoch": 0.83, "learning_rate": 7.97912542738888e-06, "loss": 0.9035, "step": 2301 }, { "epoch": 0.83, "learning_rate": 7.978225661328056e-06, "loss": 0.8312, "step": 2302 }, { "epoch": 0.83, "learning_rate": 7.977325895267232e-06, "loss": 0.8706, "step": 2303 }, { "epoch": 0.83, "learning_rate": 7.976426129206407e-06, "loss": 0.8104, "step": 2304 }, { "epoch": 0.83, "learning_rate": 7.975526363145582e-06, "loss": 0.9667, "step": 2305 }, { "epoch": 0.83, "learning_rate": 7.974626597084759e-06, "loss": 0.857, "step": 2306 }, { "epoch": 0.83, "learning_rate": 7.973726831023935e-06, "loss": 0.8472, "step": 2307 }, { "epoch": 0.83, "learning_rate": 7.97282706496311e-06, "loss": 0.7926, "step": 2308 }, { "epoch": 0.83, "learning_rate": 7.971927298902285e-06, "loss": 0.8735, "step": 2309 }, { "epoch": 0.83, "learning_rate": 7.971027532841462e-06, "loss": 0.8807, "step": 2310 }, { "epoch": 0.83, "learning_rate": 7.970127766780637e-06, "loss": 0.9786, "step": 2311 }, { "epoch": 0.83, "learning_rate": 7.969228000719813e-06, "loss": 0.9865, "step": 2312 }, { "epoch": 0.83, "learning_rate": 7.96832823465899e-06, "loss": 0.9578, "step": 2313 }, { "epoch": 0.83, "learning_rate": 7.967428468598165e-06, "loss": 0.815, "step": 2314 }, { "epoch": 0.83, "learning_rate": 7.96652870253734e-06, "loss": 0.7862, "step": 2315 }, { "epoch": 0.83, "learning_rate": 7.965628936476517e-06, "loss": 0.7983, "step": 2316 }, { "epoch": 0.83, "learning_rate": 7.964729170415693e-06, "loss": 0.8376, "step": 2317 }, { "epoch": 0.83, "learning_rate": 7.963829404354868e-06, "loss": 0.8127, "step": 2318 }, { "epoch": 0.83, "learning_rate": 7.962929638294045e-06, "loss": 0.8655, "step": 2319 }, { "epoch": 0.83, "learning_rate": 7.96202987223322e-06, "loss": 0.7877, "step": 2320 }, { "epoch": 0.83, "learning_rate": 7.961130106172396e-06, "loss": 0.8743, "step": 2321 }, { "epoch": 0.83, "learning_rate": 7.960230340111573e-06, "loss": 0.8827, "step": 2322 }, { "epoch": 0.83, "learning_rate": 7.959330574050748e-06, "loss": 0.8424, "step": 2323 }, { "epoch": 0.83, "learning_rate": 7.958430807989923e-06, "loss": 0.9109, "step": 2324 }, { "epoch": 0.83, "learning_rate": 7.9575310419291e-06, "loss": 0.8723, "step": 2325 }, { "epoch": 0.83, "learning_rate": 7.956631275868276e-06, "loss": 0.85, "step": 2326 }, { "epoch": 0.84, "learning_rate": 7.955731509807451e-06, "loss": 0.8633, "step": 2327 }, { "epoch": 0.84, "learning_rate": 7.954831743746626e-06, "loss": 0.9032, "step": 2328 }, { "epoch": 0.84, "learning_rate": 7.953931977685802e-06, "loss": 0.8921, "step": 2329 }, { "epoch": 0.84, "learning_rate": 7.953032211624977e-06, "loss": 1.0146, "step": 2330 }, { "epoch": 0.84, "learning_rate": 7.952132445564154e-06, "loss": 0.9166, "step": 2331 }, { "epoch": 0.84, "learning_rate": 7.95123267950333e-06, "loss": 0.7258, "step": 2332 }, { "epoch": 0.84, "learning_rate": 7.950332913442505e-06, "loss": 0.864, "step": 2333 }, { "epoch": 0.84, "learning_rate": 7.94943314738168e-06, "loss": 0.9453, "step": 2334 }, { "epoch": 0.84, "learning_rate": 7.948533381320857e-06, "loss": 0.8909, "step": 2335 }, { "epoch": 0.84, "learning_rate": 7.947633615260034e-06, "loss": 0.8181, "step": 2336 }, { "epoch": 0.84, "learning_rate": 7.946733849199209e-06, "loss": 0.8742, "step": 2337 }, { "epoch": 0.84, "learning_rate": 7.945834083138385e-06, "loss": 0.93, "step": 2338 }, { "epoch": 0.84, "learning_rate": 7.94493431707756e-06, "loss": 0.891, "step": 2339 }, { "epoch": 0.84, "learning_rate": 7.944034551016737e-06, "loss": 0.93, "step": 2340 }, { "epoch": 0.84, "learning_rate": 7.943134784955913e-06, "loss": 0.8391, "step": 2341 }, { "epoch": 0.84, "learning_rate": 7.942235018895088e-06, "loss": 0.9384, "step": 2342 }, { "epoch": 0.84, "learning_rate": 7.941335252834263e-06, "loss": 0.9671, "step": 2343 }, { "epoch": 0.84, "learning_rate": 7.94043548677344e-06, "loss": 0.9113, "step": 2344 }, { "epoch": 0.84, "learning_rate": 7.939535720712616e-06, "loss": 0.8537, "step": 2345 }, { "epoch": 0.84, "learning_rate": 7.938635954651791e-06, "loss": 0.8633, "step": 2346 }, { "epoch": 0.84, "learning_rate": 7.937736188590966e-06, "loss": 0.7922, "step": 2347 }, { "epoch": 0.84, "learning_rate": 7.936836422530143e-06, "loss": 0.7353, "step": 2348 }, { "epoch": 0.84, "learning_rate": 7.935936656469318e-06, "loss": 0.761, "step": 2349 }, { "epoch": 0.84, "learning_rate": 7.935036890408494e-06, "loss": 0.7294, "step": 2350 }, { "epoch": 0.84, "learning_rate": 7.934137124347671e-06, "loss": 0.8405, "step": 2351 }, { "epoch": 0.84, "learning_rate": 7.933237358286846e-06, "loss": 0.8642, "step": 2352 }, { "epoch": 0.84, "learning_rate": 7.93233759222602e-06, "loss": 0.7839, "step": 2353 }, { "epoch": 0.84, "learning_rate": 7.931437826165197e-06, "loss": 0.8905, "step": 2354 }, { "epoch": 0.85, "learning_rate": 7.930538060104374e-06, "loss": 0.8018, "step": 2355 }, { "epoch": 0.85, "learning_rate": 7.929638294043549e-06, "loss": 0.8089, "step": 2356 }, { "epoch": 0.85, "learning_rate": 7.928738527982726e-06, "loss": 0.8099, "step": 2357 }, { "epoch": 0.85, "learning_rate": 7.9278387619219e-06, "loss": 0.7949, "step": 2358 }, { "epoch": 0.85, "learning_rate": 7.926938995861077e-06, "loss": 0.88, "step": 2359 }, { "epoch": 0.85, "learning_rate": 7.926039229800254e-06, "loss": 0.7957, "step": 2360 }, { "epoch": 0.85, "learning_rate": 7.925139463739429e-06, "loss": 0.8262, "step": 2361 }, { "epoch": 0.85, "learning_rate": 7.924239697678604e-06, "loss": 0.8641, "step": 2362 }, { "epoch": 0.85, "learning_rate": 7.92333993161778e-06, "loss": 0.8737, "step": 2363 }, { "epoch": 0.85, "learning_rate": 7.922440165556957e-06, "loss": 0.8145, "step": 2364 }, { "epoch": 0.85, "learning_rate": 7.921540399496132e-06, "loss": 0.7879, "step": 2365 }, { "epoch": 0.85, "learning_rate": 7.920640633435307e-06, "loss": 0.8739, "step": 2366 }, { "epoch": 0.85, "learning_rate": 7.919740867374483e-06, "loss": 0.7528, "step": 2367 }, { "epoch": 0.85, "learning_rate": 7.918841101313658e-06, "loss": 0.8802, "step": 2368 }, { "epoch": 0.85, "learning_rate": 7.917941335252835e-06, "loss": 0.7642, "step": 2369 }, { "epoch": 0.85, "learning_rate": 7.917041569192011e-06, "loss": 0.9409, "step": 2370 }, { "epoch": 0.85, "learning_rate": 7.916141803131186e-06, "loss": 0.9302, "step": 2371 }, { "epoch": 0.85, "learning_rate": 7.915242037070361e-06, "loss": 0.8642, "step": 2372 }, { "epoch": 0.85, "learning_rate": 7.914342271009538e-06, "loss": 0.9436, "step": 2373 }, { "epoch": 0.85, "learning_rate": 7.913442504948715e-06, "loss": 0.8944, "step": 2374 }, { "epoch": 0.85, "learning_rate": 7.91254273888789e-06, "loss": 0.869, "step": 2375 }, { "epoch": 0.85, "learning_rate": 7.911642972827066e-06, "loss": 0.8145, "step": 2376 }, { "epoch": 0.85, "learning_rate": 7.910743206766241e-06, "loss": 1.0237, "step": 2377 }, { "epoch": 0.85, "learning_rate": 7.909843440705418e-06, "loss": 0.7923, "step": 2378 }, { "epoch": 0.85, "learning_rate": 7.908943674644594e-06, "loss": 0.8059, "step": 2379 }, { "epoch": 0.85, "learning_rate": 7.908043908583769e-06, "loss": 0.9005, "step": 2380 }, { "epoch": 0.85, "learning_rate": 7.907144142522944e-06, "loss": 0.8978, "step": 2381 }, { "epoch": 0.85, "learning_rate": 7.90624437646212e-06, "loss": 0.8552, "step": 2382 }, { "epoch": 0.86, "learning_rate": 7.905344610401297e-06, "loss": 0.8999, "step": 2383 }, { "epoch": 0.86, "learning_rate": 7.904444844340472e-06, "loss": 0.7579, "step": 2384 }, { "epoch": 0.86, "learning_rate": 7.903545078279647e-06, "loss": 0.8351, "step": 2385 }, { "epoch": 0.86, "learning_rate": 7.902645312218824e-06, "loss": 0.8545, "step": 2386 }, { "epoch": 0.86, "learning_rate": 7.901745546157999e-06, "loss": 0.8383, "step": 2387 }, { "epoch": 0.86, "learning_rate": 7.900845780097175e-06, "loss": 0.7869, "step": 2388 }, { "epoch": 0.86, "learning_rate": 7.899946014036352e-06, "loss": 0.9516, "step": 2389 }, { "epoch": 0.86, "learning_rate": 7.899046247975527e-06, "loss": 0.7726, "step": 2390 }, { "epoch": 0.86, "learning_rate": 7.898146481914702e-06, "loss": 0.898, "step": 2391 }, { "epoch": 0.86, "learning_rate": 7.897246715853878e-06, "loss": 0.8667, "step": 2392 }, { "epoch": 0.86, "learning_rate": 7.896346949793055e-06, "loss": 0.8078, "step": 2393 }, { "epoch": 0.86, "learning_rate": 7.89544718373223e-06, "loss": 0.7479, "step": 2394 }, { "epoch": 0.86, "learning_rate": 7.894547417671407e-06, "loss": 0.8218, "step": 2395 }, { "epoch": 0.86, "learning_rate": 7.893647651610581e-06, "loss": 0.8971, "step": 2396 }, { "epoch": 0.86, "learning_rate": 7.892747885549758e-06, "loss": 0.8139, "step": 2397 }, { "epoch": 0.86, "learning_rate": 7.891848119488935e-06, "loss": 0.9115, "step": 2398 }, { "epoch": 0.86, "learning_rate": 7.89094835342811e-06, "loss": 0.8971, "step": 2399 }, { "epoch": 0.86, "learning_rate": 7.890048587367285e-06, "loss": 0.8856, "step": 2400 }, { "epoch": 0.86, "learning_rate": 7.889148821306461e-06, "loss": 0.7491, "step": 2401 }, { "epoch": 0.86, "learning_rate": 7.888249055245638e-06, "loss": 0.7688, "step": 2402 }, { "epoch": 0.86, "learning_rate": 7.887349289184813e-06, "loss": 0.8913, "step": 2403 }, { "epoch": 0.86, "learning_rate": 7.886449523123988e-06, "loss": 0.7614, "step": 2404 }, { "epoch": 0.86, "learning_rate": 7.885549757063164e-06, "loss": 0.9184, "step": 2405 }, { "epoch": 0.86, "learning_rate": 7.884649991002339e-06, "loss": 0.8985, "step": 2406 }, { "epoch": 0.86, "learning_rate": 7.883750224941516e-06, "loss": 0.8304, "step": 2407 }, { "epoch": 0.86, "learning_rate": 7.882850458880692e-06, "loss": 0.739, "step": 2408 }, { "epoch": 0.86, "learning_rate": 7.881950692819867e-06, "loss": 0.8383, "step": 2409 }, { "epoch": 0.87, "learning_rate": 7.881050926759042e-06, "loss": 0.7476, "step": 2410 }, { "epoch": 0.87, "learning_rate": 7.880151160698219e-06, "loss": 0.8852, "step": 2411 }, { "epoch": 0.87, "learning_rate": 7.879251394637395e-06, "loss": 0.759, "step": 2412 }, { "epoch": 0.87, "learning_rate": 7.87835162857657e-06, "loss": 0.8717, "step": 2413 }, { "epoch": 0.87, "learning_rate": 7.877451862515747e-06, "loss": 0.8989, "step": 2414 }, { "epoch": 0.87, "learning_rate": 7.876552096454922e-06, "loss": 0.8055, "step": 2415 }, { "epoch": 0.87, "learning_rate": 7.875652330394099e-06, "loss": 0.7629, "step": 2416 }, { "epoch": 0.87, "learning_rate": 7.874752564333275e-06, "loss": 0.8668, "step": 2417 }, { "epoch": 0.87, "learning_rate": 7.87385279827245e-06, "loss": 0.9192, "step": 2418 }, { "epoch": 0.87, "learning_rate": 7.872953032211625e-06, "loss": 0.82, "step": 2419 }, { "epoch": 0.87, "learning_rate": 7.872053266150802e-06, "loss": 0.8868, "step": 2420 }, { "epoch": 0.87, "learning_rate": 7.871153500089978e-06, "loss": 0.8995, "step": 2421 }, { "epoch": 0.87, "learning_rate": 7.870253734029153e-06, "loss": 0.8179, "step": 2422 }, { "epoch": 0.87, "learning_rate": 7.869353967968328e-06, "loss": 0.9829, "step": 2423 }, { "epoch": 0.87, "learning_rate": 7.868454201907505e-06, "loss": 0.9046, "step": 2424 }, { "epoch": 0.87, "learning_rate": 7.867554435846681e-06, "loss": 0.896, "step": 2425 }, { "epoch": 0.87, "learning_rate": 7.866654669785856e-06, "loss": 0.7634, "step": 2426 }, { "epoch": 0.87, "learning_rate": 7.865754903725033e-06, "loss": 0.8337, "step": 2427 }, { "epoch": 0.87, "learning_rate": 7.864855137664208e-06, "loss": 0.7068, "step": 2428 }, { "epoch": 0.87, "learning_rate": 7.863955371603383e-06, "loss": 0.8601, "step": 2429 }, { "epoch": 0.87, "learning_rate": 7.86305560554256e-06, "loss": 0.856, "step": 2430 }, { "epoch": 0.87, "learning_rate": 7.862155839481736e-06, "loss": 0.8533, "step": 2431 }, { "epoch": 0.87, "learning_rate": 7.86125607342091e-06, "loss": 0.7772, "step": 2432 }, { "epoch": 0.87, "learning_rate": 7.860356307360087e-06, "loss": 0.874, "step": 2433 }, { "epoch": 0.87, "learning_rate": 7.859456541299262e-06, "loss": 1.0453, "step": 2434 }, { "epoch": 0.87, "learning_rate": 7.858556775238439e-06, "loss": 0.828, "step": 2435 }, { "epoch": 0.87, "learning_rate": 7.857657009177616e-06, "loss": 0.9124, "step": 2436 }, { "epoch": 0.87, "learning_rate": 7.85675724311679e-06, "loss": 0.8552, "step": 2437 }, { "epoch": 0.88, "learning_rate": 7.855857477055965e-06, "loss": 0.7585, "step": 2438 }, { "epoch": 0.88, "learning_rate": 7.854957710995142e-06, "loss": 0.9715, "step": 2439 }, { "epoch": 0.88, "learning_rate": 7.854057944934319e-06, "loss": 0.7787, "step": 2440 }, { "epoch": 0.88, "learning_rate": 7.853158178873494e-06, "loss": 0.8773, "step": 2441 }, { "epoch": 0.88, "learning_rate": 7.852258412812668e-06, "loss": 0.9229, "step": 2442 }, { "epoch": 0.88, "learning_rate": 7.851358646751845e-06, "loss": 0.9663, "step": 2443 }, { "epoch": 0.88, "learning_rate": 7.850458880691022e-06, "loss": 0.8689, "step": 2444 }, { "epoch": 0.88, "learning_rate": 7.849559114630197e-06, "loss": 0.8265, "step": 2445 }, { "epoch": 0.88, "learning_rate": 7.848659348569373e-06, "loss": 0.8553, "step": 2446 }, { "epoch": 0.88, "learning_rate": 7.847759582508548e-06, "loss": 0.9063, "step": 2447 }, { "epoch": 0.88, "learning_rate": 7.846859816447723e-06, "loss": 0.8189, "step": 2448 }, { "epoch": 0.88, "learning_rate": 7.8459600503869e-06, "loss": 0.7784, "step": 2449 }, { "epoch": 0.88, "learning_rate": 7.845060284326076e-06, "loss": 0.8862, "step": 2450 }, { "epoch": 0.88, "learning_rate": 7.844160518265251e-06, "loss": 1.0618, "step": 2451 }, { "epoch": 0.88, "learning_rate": 7.843260752204428e-06, "loss": 0.7943, "step": 2452 }, { "epoch": 0.88, "learning_rate": 7.842360986143603e-06, "loss": 0.882, "step": 2453 }, { "epoch": 0.88, "learning_rate": 7.84146122008278e-06, "loss": 1.1121, "step": 2454 }, { "epoch": 0.88, "learning_rate": 7.840561454021956e-06, "loss": 0.8346, "step": 2455 }, { "epoch": 0.88, "learning_rate": 7.839661687961131e-06, "loss": 0.7079, "step": 2456 }, { "epoch": 0.88, "learning_rate": 7.838761921900306e-06, "loss": 0.8814, "step": 2457 }, { "epoch": 0.88, "learning_rate": 7.837862155839482e-06, "loss": 0.8603, "step": 2458 }, { "epoch": 0.88, "learning_rate": 7.836962389778659e-06, "loss": 0.9466, "step": 2459 }, { "epoch": 0.88, "learning_rate": 7.836062623717834e-06, "loss": 0.8119, "step": 2460 }, { "epoch": 0.88, "learning_rate": 7.835162857657009e-06, "loss": 0.9, "step": 2461 }, { "epoch": 0.88, "learning_rate": 7.834263091596186e-06, "loss": 0.8029, "step": 2462 }, { "epoch": 0.88, "learning_rate": 7.833363325535362e-06, "loss": 0.8451, "step": 2463 }, { "epoch": 0.88, "learning_rate": 7.832463559474537e-06, "loss": 0.9814, "step": 2464 }, { "epoch": 0.88, "learning_rate": 7.831563793413714e-06, "loss": 0.7824, "step": 2465 }, { "epoch": 0.89, "learning_rate": 7.830664027352889e-06, "loss": 0.9322, "step": 2466 }, { "epoch": 0.89, "learning_rate": 7.829764261292064e-06, "loss": 0.9182, "step": 2467 }, { "epoch": 0.89, "learning_rate": 7.82886449523124e-06, "loss": 0.8492, "step": 2468 }, { "epoch": 0.89, "learning_rate": 7.827964729170417e-06, "loss": 0.8466, "step": 2469 }, { "epoch": 0.89, "learning_rate": 7.827064963109592e-06, "loss": 0.8496, "step": 2470 }, { "epoch": 0.89, "learning_rate": 7.826165197048768e-06, "loss": 0.971, "step": 2471 }, { "epoch": 0.89, "learning_rate": 7.825265430987943e-06, "loss": 0.8634, "step": 2472 }, { "epoch": 0.89, "learning_rate": 7.82436566492712e-06, "loss": 0.9155, "step": 2473 }, { "epoch": 0.89, "learning_rate": 7.823465898866296e-06, "loss": 0.8667, "step": 2474 }, { "epoch": 0.89, "learning_rate": 7.822566132805471e-06, "loss": 0.9107, "step": 2475 }, { "epoch": 0.89, "learning_rate": 7.821666366744646e-06, "loss": 0.8481, "step": 2476 }, { "epoch": 0.89, "learning_rate": 7.820766600683823e-06, "loss": 0.9275, "step": 2477 }, { "epoch": 0.89, "learning_rate": 7.819866834623e-06, "loss": 0.7257, "step": 2478 }, { "epoch": 0.89, "learning_rate": 7.818967068562174e-06, "loss": 0.8009, "step": 2479 }, { "epoch": 0.89, "learning_rate": 7.81806730250135e-06, "loss": 0.8355, "step": 2480 }, { "epoch": 0.89, "learning_rate": 7.817167536440526e-06, "loss": 0.8064, "step": 2481 }, { "epoch": 0.89, "learning_rate": 7.816267770379703e-06, "loss": 0.8319, "step": 2482 }, { "epoch": 0.89, "learning_rate": 7.815368004318878e-06, "loss": 0.9056, "step": 2483 }, { "epoch": 0.89, "learning_rate": 7.814468238258054e-06, "loss": 0.8739, "step": 2484 }, { "epoch": 0.89, "learning_rate": 7.813568472197229e-06, "loss": 0.8978, "step": 2485 }, { "epoch": 0.89, "learning_rate": 7.812668706136404e-06, "loss": 1.0111, "step": 2486 }, { "epoch": 0.89, "learning_rate": 7.81176894007558e-06, "loss": 0.9505, "step": 2487 }, { "epoch": 0.89, "learning_rate": 7.810869174014757e-06, "loss": 0.943, "step": 2488 }, { "epoch": 0.89, "learning_rate": 7.809969407953932e-06, "loss": 0.7988, "step": 2489 }, { "epoch": 0.89, "learning_rate": 7.809069641893109e-06, "loss": 0.791, "step": 2490 }, { "epoch": 0.89, "learning_rate": 7.808169875832284e-06, "loss": 0.7572, "step": 2491 }, { "epoch": 0.89, "learning_rate": 7.80727010977146e-06, "loss": 0.7778, "step": 2492 }, { "epoch": 0.89, "learning_rate": 7.806370343710637e-06, "loss": 0.9248, "step": 2493 }, { "epoch": 0.9, "learning_rate": 7.805470577649812e-06, "loss": 0.8853, "step": 2494 }, { "epoch": 0.9, "learning_rate": 7.804570811588987e-06, "loss": 0.8701, "step": 2495 }, { "epoch": 0.9, "learning_rate": 7.803671045528163e-06, "loss": 0.8525, "step": 2496 }, { "epoch": 0.9, "learning_rate": 7.80277127946734e-06, "loss": 0.8479, "step": 2497 }, { "epoch": 0.9, "learning_rate": 7.801871513406515e-06, "loss": 0.8321, "step": 2498 }, { "epoch": 0.9, "learning_rate": 7.80097174734569e-06, "loss": 0.8401, "step": 2499 }, { "epoch": 0.9, "learning_rate": 7.800071981284866e-06, "loss": 0.8095, "step": 2500 }, { "epoch": 0.9, "learning_rate": 7.799172215224043e-06, "loss": 0.7946, "step": 2501 }, { "epoch": 0.9, "learning_rate": 7.798272449163218e-06, "loss": 0.9232, "step": 2502 }, { "epoch": 0.9, "learning_rate": 7.797372683102395e-06, "loss": 0.9482, "step": 2503 }, { "epoch": 0.9, "learning_rate": 7.79647291704157e-06, "loss": 0.8324, "step": 2504 }, { "epoch": 0.9, "learning_rate": 7.795573150980744e-06, "loss": 0.8502, "step": 2505 }, { "epoch": 0.9, "learning_rate": 7.794673384919921e-06, "loss": 0.8424, "step": 2506 }, { "epoch": 0.9, "learning_rate": 7.793773618859098e-06, "loss": 0.7891, "step": 2507 }, { "epoch": 0.9, "learning_rate": 7.792873852798273e-06, "loss": 0.8404, "step": 2508 }, { "epoch": 0.9, "learning_rate": 7.79197408673745e-06, "loss": 0.8182, "step": 2509 }, { "epoch": 0.9, "learning_rate": 7.791074320676624e-06, "loss": 0.8419, "step": 2510 }, { "epoch": 0.9, "learning_rate": 7.7901745546158e-06, "loss": 0.8488, "step": 2511 }, { "epoch": 0.9, "learning_rate": 7.789274788554977e-06, "loss": 0.8097, "step": 2512 }, { "epoch": 0.9, "learning_rate": 7.788375022494152e-06, "loss": 0.7848, "step": 2513 }, { "epoch": 0.9, "learning_rate": 7.787475256433327e-06, "loss": 0.92, "step": 2514 }, { "epoch": 0.9, "learning_rate": 7.786575490372504e-06, "loss": 0.7916, "step": 2515 }, { "epoch": 0.9, "learning_rate": 7.78567572431168e-06, "loss": 0.8308, "step": 2516 }, { "epoch": 0.9, "learning_rate": 7.784775958250855e-06, "loss": 0.8357, "step": 2517 }, { "epoch": 0.9, "learning_rate": 7.78387619219003e-06, "loss": 0.8721, "step": 2518 }, { "epoch": 0.9, "learning_rate": 7.782976426129207e-06, "loss": 0.8519, "step": 2519 }, { "epoch": 0.9, "learning_rate": 7.782076660068384e-06, "loss": 0.6931, "step": 2520 }, { "epoch": 0.9, "learning_rate": 7.781176894007558e-06, "loss": 0.738, "step": 2521 }, { "epoch": 0.91, "learning_rate": 7.780277127946735e-06, "loss": 0.7958, "step": 2522 }, { "epoch": 0.91, "learning_rate": 7.77937736188591e-06, "loss": 0.8911, "step": 2523 }, { "epoch": 0.91, "learning_rate": 7.778477595825085e-06, "loss": 0.9013, "step": 2524 }, { "epoch": 0.91, "learning_rate": 7.777577829764262e-06, "loss": 0.802, "step": 2525 }, { "epoch": 0.91, "learning_rate": 7.776678063703438e-06, "loss": 0.926, "step": 2526 }, { "epoch": 0.91, "learning_rate": 7.775778297642613e-06, "loss": 0.9077, "step": 2527 }, { "epoch": 0.91, "learning_rate": 7.77487853158179e-06, "loss": 0.8998, "step": 2528 }, { "epoch": 0.91, "learning_rate": 7.773978765520965e-06, "loss": 0.7855, "step": 2529 }, { "epoch": 0.91, "learning_rate": 7.773078999460141e-06, "loss": 0.8421, "step": 2530 }, { "epoch": 0.91, "learning_rate": 7.772179233399318e-06, "loss": 0.813, "step": 2531 }, { "epoch": 0.91, "learning_rate": 7.771279467338493e-06, "loss": 0.8564, "step": 2532 }, { "epoch": 0.91, "learning_rate": 7.770379701277668e-06, "loss": 0.8121, "step": 2533 }, { "epoch": 0.91, "learning_rate": 7.769479935216844e-06, "loss": 0.8045, "step": 2534 }, { "epoch": 0.91, "learning_rate": 7.768580169156021e-06, "loss": 0.8049, "step": 2535 }, { "epoch": 0.91, "learning_rate": 7.767680403095196e-06, "loss": 0.8668, "step": 2536 }, { "epoch": 0.91, "learning_rate": 7.76678063703437e-06, "loss": 0.7751, "step": 2537 }, { "epoch": 0.91, "learning_rate": 7.765880870973547e-06, "loss": 0.9025, "step": 2538 }, { "epoch": 0.91, "learning_rate": 7.764981104912724e-06, "loss": 0.8158, "step": 2539 }, { "epoch": 0.91, "learning_rate": 7.764081338851899e-06, "loss": 0.8355, "step": 2540 }, { "epoch": 0.91, "learning_rate": 7.763181572791076e-06, "loss": 0.8581, "step": 2541 }, { "epoch": 0.91, "learning_rate": 7.76228180673025e-06, "loss": 0.8016, "step": 2542 }, { "epoch": 0.91, "learning_rate": 7.761382040669427e-06, "loss": 0.7547, "step": 2543 }, { "epoch": 0.91, "learning_rate": 7.760482274608602e-06, "loss": 0.7956, "step": 2544 }, { "epoch": 0.91, "learning_rate": 7.759582508547779e-06, "loss": 0.7378, "step": 2545 }, { "epoch": 0.91, "learning_rate": 7.758682742486954e-06, "loss": 0.8682, "step": 2546 }, { "epoch": 0.91, "learning_rate": 7.75778297642613e-06, "loss": 0.7641, "step": 2547 }, { "epoch": 0.91, "learning_rate": 7.756883210365305e-06, "loss": 0.9277, "step": 2548 }, { "epoch": 0.91, "learning_rate": 7.755983444304482e-06, "loss": 0.7618, "step": 2549 }, { "epoch": 0.92, "learning_rate": 7.755083678243658e-06, "loss": 0.9047, "step": 2550 }, { "epoch": 0.92, "learning_rate": 7.754183912182833e-06, "loss": 0.9677, "step": 2551 }, { "epoch": 0.92, "learning_rate": 7.753284146122008e-06, "loss": 0.8219, "step": 2552 }, { "epoch": 0.92, "learning_rate": 7.752384380061185e-06, "loss": 0.8575, "step": 2553 }, { "epoch": 0.92, "learning_rate": 7.751484614000361e-06, "loss": 0.7746, "step": 2554 }, { "epoch": 0.92, "learning_rate": 7.750584847939536e-06, "loss": 0.8197, "step": 2555 }, { "epoch": 0.92, "learning_rate": 7.749685081878711e-06, "loss": 0.8694, "step": 2556 }, { "epoch": 0.92, "learning_rate": 7.748785315817888e-06, "loss": 0.863, "step": 2557 }, { "epoch": 0.92, "learning_rate": 7.747885549757064e-06, "loss": 0.9167, "step": 2558 }, { "epoch": 0.92, "learning_rate": 7.74698578369624e-06, "loss": 0.7893, "step": 2559 }, { "epoch": 0.92, "learning_rate": 7.746086017635416e-06, "loss": 0.7611, "step": 2560 }, { "epoch": 0.92, "learning_rate": 7.745186251574591e-06, "loss": 0.775, "step": 2561 }, { "epoch": 0.92, "learning_rate": 7.744286485513768e-06, "loss": 0.8751, "step": 2562 }, { "epoch": 0.92, "learning_rate": 7.743386719452942e-06, "loss": 0.8543, "step": 2563 }, { "epoch": 0.92, "learning_rate": 7.742486953392119e-06, "loss": 0.8085, "step": 2564 }, { "epoch": 0.92, "learning_rate": 7.741587187331294e-06, "loss": 0.7666, "step": 2565 }, { "epoch": 0.92, "learning_rate": 7.74068742127047e-06, "loss": 0.9277, "step": 2566 }, { "epoch": 0.92, "learning_rate": 7.739787655209646e-06, "loss": 0.8086, "step": 2567 }, { "epoch": 0.92, "learning_rate": 7.738887889148822e-06, "loss": 0.8056, "step": 2568 }, { "epoch": 0.92, "learning_rate": 7.737988123087999e-06, "loss": 0.8165, "step": 2569 }, { "epoch": 0.92, "learning_rate": 7.737088357027174e-06, "loss": 0.7804, "step": 2570 }, { "epoch": 0.92, "learning_rate": 7.736188590966349e-06, "loss": 0.8925, "step": 2571 }, { "epoch": 0.92, "learning_rate": 7.735288824905525e-06, "loss": 0.8717, "step": 2572 }, { "epoch": 0.92, "learning_rate": 7.734389058844702e-06, "loss": 0.801, "step": 2573 }, { "epoch": 0.92, "learning_rate": 7.733489292783877e-06, "loss": 0.7791, "step": 2574 }, { "epoch": 0.92, "learning_rate": 7.732589526723052e-06, "loss": 0.8673, "step": 2575 }, { "epoch": 0.92, "learning_rate": 7.731689760662228e-06, "loss": 0.7509, "step": 2576 }, { "epoch": 0.92, "learning_rate": 7.730789994601405e-06, "loss": 0.8277, "step": 2577 }, { "epoch": 0.93, "learning_rate": 7.72989022854058e-06, "loss": 1.0174, "step": 2578 }, { "epoch": 0.93, "learning_rate": 7.728990462479756e-06, "loss": 0.9877, "step": 2579 }, { "epoch": 0.93, "learning_rate": 7.728090696418931e-06, "loss": 0.9022, "step": 2580 }, { "epoch": 0.93, "learning_rate": 7.727190930358108e-06, "loss": 0.7616, "step": 2581 }, { "epoch": 0.93, "learning_rate": 7.726291164297285e-06, "loss": 0.8666, "step": 2582 }, { "epoch": 0.93, "learning_rate": 7.72539139823646e-06, "loss": 0.8577, "step": 2583 }, { "epoch": 0.93, "learning_rate": 7.724491632175634e-06, "loss": 0.8968, "step": 2584 }, { "epoch": 0.93, "learning_rate": 7.723591866114811e-06, "loss": 0.8794, "step": 2585 }, { "epoch": 0.93, "learning_rate": 7.722692100053986e-06, "loss": 0.8584, "step": 2586 }, { "epoch": 0.93, "learning_rate": 7.721792333993163e-06, "loss": 0.8336, "step": 2587 }, { "epoch": 0.93, "learning_rate": 7.72089256793234e-06, "loss": 1.0273, "step": 2588 }, { "epoch": 0.93, "learning_rate": 7.719992801871514e-06, "loss": 0.8901, "step": 2589 }, { "epoch": 0.93, "learning_rate": 7.719093035810689e-06, "loss": 0.812, "step": 2590 }, { "epoch": 0.93, "learning_rate": 7.718193269749866e-06, "loss": 0.9743, "step": 2591 }, { "epoch": 0.93, "learning_rate": 7.717293503689042e-06, "loss": 0.7777, "step": 2592 }, { "epoch": 0.93, "learning_rate": 7.716393737628217e-06, "loss": 0.9364, "step": 2593 }, { "epoch": 0.93, "learning_rate": 7.715493971567392e-06, "loss": 0.9529, "step": 2594 }, { "epoch": 0.93, "learning_rate": 7.714594205506569e-06, "loss": 0.8083, "step": 2595 }, { "epoch": 0.93, "learning_rate": 7.713694439445745e-06, "loss": 1.0036, "step": 2596 }, { "epoch": 0.93, "learning_rate": 7.71279467338492e-06, "loss": 0.7934, "step": 2597 }, { "epoch": 0.93, "learning_rate": 7.711894907324097e-06, "loss": 0.8701, "step": 2598 }, { "epoch": 0.93, "learning_rate": 7.710995141263272e-06, "loss": 0.8766, "step": 2599 }, { "epoch": 0.93, "learning_rate": 7.710095375202448e-06, "loss": 0.8442, "step": 2600 }, { "epoch": 0.93, "learning_rate": 7.709195609141625e-06, "loss": 0.8691, "step": 2601 }, { "epoch": 0.93, "learning_rate": 7.7082958430808e-06, "loss": 0.7791, "step": 2602 }, { "epoch": 0.93, "learning_rate": 7.707396077019975e-06, "loss": 0.8469, "step": 2603 }, { "epoch": 0.93, "learning_rate": 7.706496310959152e-06, "loss": 0.8536, "step": 2604 }, { "epoch": 0.94, "learning_rate": 7.705596544898326e-06, "loss": 0.8483, "step": 2605 }, { "epoch": 0.94, "learning_rate": 7.704696778837503e-06, "loss": 0.7904, "step": 2606 }, { "epoch": 0.94, "learning_rate": 7.70379701277668e-06, "loss": 0.8377, "step": 2607 }, { "epoch": 0.94, "learning_rate": 7.702897246715855e-06, "loss": 0.8609, "step": 2608 }, { "epoch": 0.94, "learning_rate": 7.70199748065503e-06, "loss": 0.7792, "step": 2609 }, { "epoch": 0.94, "learning_rate": 7.701097714594206e-06, "loss": 0.8473, "step": 2610 }, { "epoch": 0.94, "learning_rate": 7.700197948533383e-06, "loss": 0.7827, "step": 2611 }, { "epoch": 0.94, "learning_rate": 7.699298182472558e-06, "loss": 0.8059, "step": 2612 }, { "epoch": 0.94, "learning_rate": 7.698398416411733e-06, "loss": 0.8594, "step": 2613 }, { "epoch": 0.94, "learning_rate": 7.69749865035091e-06, "loss": 0.7875, "step": 2614 }, { "epoch": 0.94, "learning_rate": 7.696598884290086e-06, "loss": 0.8369, "step": 2615 }, { "epoch": 0.94, "learning_rate": 7.69569911822926e-06, "loss": 0.9241, "step": 2616 }, { "epoch": 0.94, "learning_rate": 7.694799352168437e-06, "loss": 0.8157, "step": 2617 }, { "epoch": 0.94, "learning_rate": 7.693899586107612e-06, "loss": 0.7811, "step": 2618 }, { "epoch": 0.94, "learning_rate": 7.692999820046789e-06, "loss": 0.8288, "step": 2619 }, { "epoch": 0.94, "learning_rate": 7.692100053985966e-06, "loss": 0.7722, "step": 2620 }, { "epoch": 0.94, "learning_rate": 7.69120028792514e-06, "loss": 0.8171, "step": 2621 }, { "epoch": 0.94, "learning_rate": 7.690300521864315e-06, "loss": 0.8475, "step": 2622 }, { "epoch": 0.94, "learning_rate": 7.689400755803492e-06, "loss": 0.7765, "step": 2623 }, { "epoch": 0.94, "learning_rate": 7.688500989742667e-06, "loss": 0.7617, "step": 2624 }, { "epoch": 0.94, "learning_rate": 7.687601223681844e-06, "loss": 0.8313, "step": 2625 }, { "epoch": 0.94, "learning_rate": 7.68670145762102e-06, "loss": 0.7928, "step": 2626 }, { "epoch": 0.94, "learning_rate": 7.685801691560195e-06, "loss": 0.8698, "step": 2627 }, { "epoch": 0.94, "learning_rate": 7.68490192549937e-06, "loss": 0.8656, "step": 2628 }, { "epoch": 0.94, "learning_rate": 7.684002159438547e-06, "loss": 0.9303, "step": 2629 }, { "epoch": 0.94, "learning_rate": 7.683102393377723e-06, "loss": 0.7802, "step": 2630 }, { "epoch": 0.94, "learning_rate": 7.682202627316898e-06, "loss": 0.7754, "step": 2631 }, { "epoch": 0.94, "learning_rate": 7.681302861256073e-06, "loss": 0.7242, "step": 2632 }, { "epoch": 0.95, "learning_rate": 7.68040309519525e-06, "loss": 0.8287, "step": 2633 }, { "epoch": 0.95, "learning_rate": 7.679503329134426e-06, "loss": 0.8119, "step": 2634 }, { "epoch": 0.95, "learning_rate": 7.678603563073601e-06, "loss": 0.7582, "step": 2635 }, { "epoch": 0.95, "learning_rate": 7.677703797012778e-06, "loss": 0.8345, "step": 2636 }, { "epoch": 0.95, "learning_rate": 7.676804030951953e-06, "loss": 0.8227, "step": 2637 }, { "epoch": 0.95, "learning_rate": 7.67590426489113e-06, "loss": 0.8521, "step": 2638 }, { "epoch": 0.95, "learning_rate": 7.675004498830306e-06, "loss": 0.8054, "step": 2639 }, { "epoch": 0.95, "learning_rate": 7.674104732769481e-06, "loss": 0.7293, "step": 2640 }, { "epoch": 0.95, "learning_rate": 7.673204966708656e-06, "loss": 0.8825, "step": 2641 }, { "epoch": 0.95, "learning_rate": 7.672305200647832e-06, "loss": 0.812, "step": 2642 }, { "epoch": 0.95, "learning_rate": 7.671405434587007e-06, "loss": 0.9178, "step": 2643 }, { "epoch": 0.95, "learning_rate": 7.670505668526184e-06, "loss": 0.8475, "step": 2644 }, { "epoch": 0.95, "learning_rate": 7.66960590246536e-06, "loss": 0.766, "step": 2645 }, { "epoch": 0.95, "learning_rate": 7.668706136404535e-06, "loss": 0.7985, "step": 2646 }, { "epoch": 0.95, "learning_rate": 7.66780637034371e-06, "loss": 0.9081, "step": 2647 }, { "epoch": 0.95, "learning_rate": 7.666906604282887e-06, "loss": 0.9164, "step": 2648 }, { "epoch": 0.95, "learning_rate": 7.666006838222064e-06, "loss": 0.783, "step": 2649 }, { "epoch": 0.95, "learning_rate": 7.665107072161239e-06, "loss": 0.9012, "step": 2650 }, { "epoch": 0.95, "learning_rate": 7.664207306100413e-06, "loss": 0.8935, "step": 2651 }, { "epoch": 0.95, "learning_rate": 7.66330754003959e-06, "loss": 0.826, "step": 2652 }, { "epoch": 0.95, "learning_rate": 7.662407773978767e-06, "loss": 0.9671, "step": 2653 }, { "epoch": 0.95, "learning_rate": 7.661508007917942e-06, "loss": 0.7461, "step": 2654 }, { "epoch": 0.95, "learning_rate": 7.660608241857118e-06, "loss": 0.842, "step": 2655 }, { "epoch": 0.95, "learning_rate": 7.659708475796293e-06, "loss": 0.871, "step": 2656 }, { "epoch": 0.95, "learning_rate": 7.65880870973547e-06, "loss": 0.8961, "step": 2657 }, { "epoch": 0.95, "learning_rate": 7.657908943674646e-06, "loss": 0.8466, "step": 2658 }, { "epoch": 0.95, "learning_rate": 7.657009177613821e-06, "loss": 0.8937, "step": 2659 }, { "epoch": 0.95, "learning_rate": 7.656109411552996e-06, "loss": 0.7187, "step": 2660 }, { "epoch": 0.96, "learning_rate": 7.655209645492173e-06, "loss": 0.8064, "step": 2661 }, { "epoch": 0.96, "learning_rate": 7.654309879431348e-06, "loss": 0.8043, "step": 2662 }, { "epoch": 0.96, "learning_rate": 7.653410113370524e-06, "loss": 0.8248, "step": 2663 }, { "epoch": 0.96, "learning_rate": 7.6525103473097e-06, "loss": 0.7846, "step": 2664 }, { "epoch": 0.96, "learning_rate": 7.651610581248876e-06, "loss": 0.848, "step": 2665 }, { "epoch": 0.96, "learning_rate": 7.650710815188051e-06, "loss": 0.8674, "step": 2666 }, { "epoch": 0.96, "learning_rate": 7.649811049127227e-06, "loss": 0.797, "step": 2667 }, { "epoch": 0.96, "learning_rate": 7.648911283066404e-06, "loss": 0.9177, "step": 2668 }, { "epoch": 0.96, "learning_rate": 7.648011517005579e-06, "loss": 0.8528, "step": 2669 }, { "epoch": 0.96, "learning_rate": 7.647111750944754e-06, "loss": 0.8371, "step": 2670 }, { "epoch": 0.96, "learning_rate": 7.64621198488393e-06, "loss": 0.8303, "step": 2671 }, { "epoch": 0.96, "learning_rate": 7.645312218823107e-06, "loss": 0.8829, "step": 2672 }, { "epoch": 0.96, "learning_rate": 7.644412452762282e-06, "loss": 0.9124, "step": 2673 }, { "epoch": 0.96, "learning_rate": 7.643512686701459e-06, "loss": 0.8429, "step": 2674 }, { "epoch": 0.96, "learning_rate": 7.642612920640634e-06, "loss": 0.8526, "step": 2675 }, { "epoch": 0.96, "learning_rate": 7.64171315457981e-06, "loss": 0.9074, "step": 2676 }, { "epoch": 0.96, "learning_rate": 7.640813388518987e-06, "loss": 1.0133, "step": 2677 }, { "epoch": 0.96, "learning_rate": 7.639913622458162e-06, "loss": 0.8174, "step": 2678 }, { "epoch": 0.96, "learning_rate": 7.639013856397337e-06, "loss": 0.8461, "step": 2679 }, { "epoch": 0.96, "learning_rate": 7.638114090336513e-06, "loss": 0.8289, "step": 2680 }, { "epoch": 0.96, "learning_rate": 7.637214324275688e-06, "loss": 0.9217, "step": 2681 }, { "epoch": 0.96, "learning_rate": 7.636314558214865e-06, "loss": 0.784, "step": 2682 }, { "epoch": 0.96, "learning_rate": 7.63541479215404e-06, "loss": 0.7926, "step": 2683 }, { "epoch": 0.96, "learning_rate": 7.634515026093216e-06, "loss": 0.8126, "step": 2684 }, { "epoch": 0.96, "learning_rate": 7.633615260032391e-06, "loss": 0.9011, "step": 2685 }, { "epoch": 0.96, "learning_rate": 7.632715493971568e-06, "loss": 0.8168, "step": 2686 }, { "epoch": 0.96, "learning_rate": 7.631815727910745e-06, "loss": 0.827, "step": 2687 }, { "epoch": 0.96, "learning_rate": 7.63091596184992e-06, "loss": 0.8886, "step": 2688 }, { "epoch": 0.97, "learning_rate": 7.630016195789094e-06, "loss": 0.9586, "step": 2689 }, { "epoch": 0.97, "learning_rate": 7.629116429728272e-06, "loss": 1.0003, "step": 2690 }, { "epoch": 0.97, "learning_rate": 7.628216663667447e-06, "loss": 0.866, "step": 2691 }, { "epoch": 0.97, "learning_rate": 7.6273168976066226e-06, "loss": 0.7541, "step": 2692 }, { "epoch": 0.97, "learning_rate": 7.626417131545799e-06, "loss": 0.8934, "step": 2693 }, { "epoch": 0.97, "learning_rate": 7.625517365484975e-06, "loss": 0.8338, "step": 2694 }, { "epoch": 0.97, "learning_rate": 7.62461759942415e-06, "loss": 0.8596, "step": 2695 }, { "epoch": 0.97, "learning_rate": 7.6237178333633265e-06, "loss": 0.8933, "step": 2696 }, { "epoch": 0.97, "learning_rate": 7.622818067302502e-06, "loss": 0.8662, "step": 2697 }, { "epoch": 0.97, "learning_rate": 7.621918301241677e-06, "loss": 0.8024, "step": 2698 }, { "epoch": 0.97, "learning_rate": 7.621018535180854e-06, "loss": 0.8823, "step": 2699 }, { "epoch": 0.97, "learning_rate": 7.6201187691200296e-06, "loss": 0.8055, "step": 2700 }, { "epoch": 0.97, "learning_rate": 7.619219003059205e-06, "loss": 0.7734, "step": 2701 }, { "epoch": 0.97, "learning_rate": 7.61831923699838e-06, "loss": 0.7669, "step": 2702 }, { "epoch": 0.97, "learning_rate": 7.617419470937557e-06, "loss": 0.8858, "step": 2703 }, { "epoch": 0.97, "learning_rate": 7.616519704876733e-06, "loss": 0.8305, "step": 2704 }, { "epoch": 0.97, "learning_rate": 7.615619938815908e-06, "loss": 0.8235, "step": 2705 }, { "epoch": 0.97, "learning_rate": 7.614720172755084e-06, "loss": 0.8854, "step": 2706 }, { "epoch": 0.97, "learning_rate": 7.61382040669426e-06, "loss": 0.7636, "step": 2707 }, { "epoch": 0.97, "learning_rate": 7.612920640633436e-06, "loss": 0.8601, "step": 2708 }, { "epoch": 0.97, "learning_rate": 7.612020874572612e-06, "loss": 0.7017, "step": 2709 }, { "epoch": 0.97, "learning_rate": 7.611121108511787e-06, "loss": 0.8123, "step": 2710 }, { "epoch": 0.97, "learning_rate": 7.610221342450963e-06, "loss": 0.9377, "step": 2711 }, { "epoch": 0.97, "learning_rate": 7.60932157639014e-06, "loss": 0.8891, "step": 2712 }, { "epoch": 0.97, "learning_rate": 7.608421810329315e-06, "loss": 0.73, "step": 2713 }, { "epoch": 0.97, "learning_rate": 7.60752204426849e-06, "loss": 0.932, "step": 2714 }, { "epoch": 0.97, "learning_rate": 7.606622278207667e-06, "loss": 0.9037, "step": 2715 }, { "epoch": 0.97, "learning_rate": 7.605722512146843e-06, "loss": 0.8678, "step": 2716 }, { "epoch": 0.98, "learning_rate": 7.604822746086018e-06, "loss": 0.8554, "step": 2717 }, { "epoch": 0.98, "learning_rate": 7.603922980025194e-06, "loss": 0.8942, "step": 2718 }, { "epoch": 0.98, "learning_rate": 7.60302321396437e-06, "loss": 0.8896, "step": 2719 }, { "epoch": 0.98, "learning_rate": 7.602123447903546e-06, "loss": 0.9224, "step": 2720 }, { "epoch": 0.98, "learning_rate": 7.601223681842721e-06, "loss": 0.7691, "step": 2721 }, { "epoch": 0.98, "learning_rate": 7.600323915781897e-06, "loss": 0.7682, "step": 2722 }, { "epoch": 0.98, "learning_rate": 7.599424149721073e-06, "loss": 0.8051, "step": 2723 }, { "epoch": 0.98, "learning_rate": 7.598524383660249e-06, "loss": 0.7848, "step": 2724 }, { "epoch": 0.98, "learning_rate": 7.5976246175994255e-06, "loss": 0.8002, "step": 2725 }, { "epoch": 0.98, "learning_rate": 7.5967248515386e-06, "loss": 0.7944, "step": 2726 }, { "epoch": 0.98, "learning_rate": 7.595825085477776e-06, "loss": 0.8397, "step": 2727 }, { "epoch": 0.98, "learning_rate": 7.594925319416953e-06, "loss": 0.8932, "step": 2728 }, { "epoch": 0.98, "learning_rate": 7.594025553356128e-06, "loss": 0.8582, "step": 2729 }, { "epoch": 0.98, "learning_rate": 7.5931257872953035e-06, "loss": 0.8306, "step": 2730 }, { "epoch": 0.98, "learning_rate": 7.59222602123448e-06, "loss": 0.8842, "step": 2731 }, { "epoch": 0.98, "learning_rate": 7.591326255173656e-06, "loss": 0.8338, "step": 2732 }, { "epoch": 0.98, "learning_rate": 7.590426489112831e-06, "loss": 0.8524, "step": 2733 }, { "epoch": 0.98, "learning_rate": 7.589526723052007e-06, "loss": 0.7666, "step": 2734 }, { "epoch": 0.98, "learning_rate": 7.588626956991183e-06, "loss": 0.9339, "step": 2735 }, { "epoch": 0.98, "learning_rate": 7.587727190930358e-06, "loss": 0.7906, "step": 2736 }, { "epoch": 0.98, "learning_rate": 7.586827424869535e-06, "loss": 0.8272, "step": 2737 }, { "epoch": 0.98, "learning_rate": 7.5859276588087105e-06, "loss": 0.8017, "step": 2738 }, { "epoch": 0.98, "learning_rate": 7.585027892747886e-06, "loss": 0.8391, "step": 2739 }, { "epoch": 0.98, "learning_rate": 7.584128126687061e-06, "loss": 0.913, "step": 2740 }, { "epoch": 0.98, "learning_rate": 7.583228360626238e-06, "loss": 0.7909, "step": 2741 }, { "epoch": 0.98, "learning_rate": 7.5823285945654135e-06, "loss": 0.7669, "step": 2742 }, { "epoch": 0.98, "learning_rate": 7.581428828504589e-06, "loss": 0.9316, "step": 2743 }, { "epoch": 0.98, "learning_rate": 7.580529062443766e-06, "loss": 0.9652, "step": 2744 }, { "epoch": 0.99, "learning_rate": 7.579629296382941e-06, "loss": 0.7765, "step": 2745 }, { "epoch": 0.99, "learning_rate": 7.578729530322117e-06, "loss": 0.7821, "step": 2746 }, { "epoch": 0.99, "learning_rate": 7.577829764261293e-06, "loss": 0.9528, "step": 2747 }, { "epoch": 0.99, "learning_rate": 7.576929998200468e-06, "loss": 0.844, "step": 2748 }, { "epoch": 0.99, "learning_rate": 7.576030232139644e-06, "loss": 0.8294, "step": 2749 }, { "epoch": 0.99, "learning_rate": 7.5751304660788205e-06, "loss": 0.8826, "step": 2750 }, { "epoch": 0.99, "learning_rate": 7.574230700017996e-06, "loss": 0.8914, "step": 2751 }, { "epoch": 0.99, "learning_rate": 7.573330933957171e-06, "loss": 0.8056, "step": 2752 }, { "epoch": 0.99, "learning_rate": 7.572431167896348e-06, "loss": 0.8902, "step": 2753 }, { "epoch": 0.99, "learning_rate": 7.571531401835524e-06, "loss": 0.7805, "step": 2754 }, { "epoch": 0.99, "learning_rate": 7.5706316357746985e-06, "loss": 0.6768, "step": 2755 }, { "epoch": 0.99, "learning_rate": 7.569731869713875e-06, "loss": 0.9051, "step": 2756 }, { "epoch": 0.99, "learning_rate": 7.568832103653051e-06, "loss": 0.8901, "step": 2757 }, { "epoch": 0.99, "learning_rate": 7.567932337592227e-06, "loss": 0.9803, "step": 2758 }, { "epoch": 0.99, "learning_rate": 7.567032571531402e-06, "loss": 0.9749, "step": 2759 }, { "epoch": 0.99, "learning_rate": 7.566132805470578e-06, "loss": 0.9169, "step": 2760 }, { "epoch": 0.99, "learning_rate": 7.565233039409754e-06, "loss": 0.7016, "step": 2761 }, { "epoch": 0.99, "learning_rate": 7.56433327334893e-06, "loss": 0.8287, "step": 2762 }, { "epoch": 0.99, "learning_rate": 7.563433507288106e-06, "loss": 0.8333, "step": 2763 }, { "epoch": 0.99, "learning_rate": 7.562533741227281e-06, "loss": 0.8953, "step": 2764 }, { "epoch": 0.99, "learning_rate": 7.561633975166457e-06, "loss": 0.9345, "step": 2765 }, { "epoch": 0.99, "learning_rate": 7.560734209105634e-06, "loss": 0.7498, "step": 2766 }, { "epoch": 0.99, "learning_rate": 7.559834443044809e-06, "loss": 0.8339, "step": 2767 }, { "epoch": 0.99, "learning_rate": 7.558934676983984e-06, "loss": 0.9392, "step": 2768 }, { "epoch": 0.99, "learning_rate": 7.558034910923161e-06, "loss": 0.8307, "step": 2769 }, { "epoch": 0.99, "learning_rate": 7.557135144862337e-06, "loss": 0.8772, "step": 2770 }, { "epoch": 0.99, "learning_rate": 7.556235378801512e-06, "loss": 0.9682, "step": 2771 }, { "epoch": 0.99, "learning_rate": 7.555335612740688e-06, "loss": 0.8592, "step": 2772 }, { "epoch": 1.0, "learning_rate": 7.554435846679864e-06, "loss": 0.8543, "step": 2773 }, { "epoch": 1.0, "learning_rate": 7.55353608061904e-06, "loss": 0.8099, "step": 2774 }, { "epoch": 1.0, "learning_rate": 7.552636314558216e-06, "loss": 0.8168, "step": 2775 }, { "epoch": 1.0, "learning_rate": 7.551736548497391e-06, "loss": 0.8837, "step": 2776 }, { "epoch": 1.0, "learning_rate": 7.550836782436567e-06, "loss": 0.8984, "step": 2777 }, { "epoch": 1.0, "learning_rate": 7.549937016375742e-06, "loss": 0.8831, "step": 2778 }, { "epoch": 1.0, "learning_rate": 7.549037250314919e-06, "loss": 0.7582, "step": 2779 }, { "epoch": 1.0, "learning_rate": 7.5481374842540944e-06, "loss": 0.9247, "step": 2780 }, { "epoch": 1.0, "learning_rate": 7.54723771819327e-06, "loss": 0.7767, "step": 2781 }, { "epoch": 1.0, "learning_rate": 7.546337952132447e-06, "loss": 0.8646, "step": 2782 }, { "epoch": 1.0, "learning_rate": 7.545438186071622e-06, "loss": 0.9519, "step": 2783 }, { "epoch": 1.0, "learning_rate": 7.5445384200107975e-06, "loss": 0.7231, "step": 2784 }, { "epoch": 1.0, "learning_rate": 7.543638653949974e-06, "loss": 0.9579, "step": 2785 }, { "epoch": 1.0, "learning_rate": 7.542738887889149e-06, "loss": 0.7701, "step": 2786 }, { "epoch": 1.0, "learning_rate": 7.541839121828325e-06, "loss": 0.6243, "step": 2787 }, { "epoch": 1.0, "learning_rate": 7.5409393557675014e-06, "loss": 0.6871, "step": 2788 }, { "epoch": 1.0, "learning_rate": 7.540039589706677e-06, "loss": 0.6429, "step": 2789 }, { "epoch": 1.0, "learning_rate": 7.539139823645852e-06, "loss": 0.6422, "step": 2790 }, { "epoch": 1.0, "learning_rate": 7.538240057585029e-06, "loss": 0.7302, "step": 2791 }, { "epoch": 1.0, "learning_rate": 7.5373402915242045e-06, "loss": 0.5757, "step": 2792 }, { "epoch": 1.0, "learning_rate": 7.53644052546338e-06, "loss": 0.5848, "step": 2793 }, { "epoch": 1.0, "learning_rate": 7.535540759402556e-06, "loss": 0.6091, "step": 2794 }, { "epoch": 1.0, "learning_rate": 7.534640993341732e-06, "loss": 0.5519, "step": 2795 }, { "epoch": 1.0, "learning_rate": 7.533741227280908e-06, "loss": 0.6993, "step": 2796 }, { "epoch": 1.0, "learning_rate": 7.5328414612200825e-06, "loss": 0.5448, "step": 2797 }, { "epoch": 1.0, "learning_rate": 7.531941695159259e-06, "loss": 0.684, "step": 2798 }, { "epoch": 1.0, "learning_rate": 7.531041929098435e-06, "loss": 0.5636, "step": 2799 }, { "epoch": 1.01, "learning_rate": 7.530142163037611e-06, "loss": 0.5919, "step": 2800 }, { "epoch": 1.01, "learning_rate": 7.529242396976787e-06, "loss": 0.5698, "step": 2801 }, { "epoch": 1.01, "learning_rate": 7.528342630915962e-06, "loss": 0.5806, "step": 2802 }, { "epoch": 1.01, "learning_rate": 7.527442864855138e-06, "loss": 0.6704, "step": 2803 }, { "epoch": 1.01, "learning_rate": 7.526543098794315e-06, "loss": 0.6061, "step": 2804 }, { "epoch": 1.01, "learning_rate": 7.5256433327334895e-06, "loss": 0.703, "step": 2805 }, { "epoch": 1.01, "learning_rate": 7.524743566672665e-06, "loss": 0.6048, "step": 2806 }, { "epoch": 1.01, "learning_rate": 7.523843800611842e-06, "loss": 0.5684, "step": 2807 }, { "epoch": 1.01, "learning_rate": 7.522944034551018e-06, "loss": 0.6398, "step": 2808 }, { "epoch": 1.01, "learning_rate": 7.522044268490193e-06, "loss": 0.5675, "step": 2809 }, { "epoch": 1.01, "learning_rate": 7.521144502429369e-06, "loss": 0.628, "step": 2810 }, { "epoch": 1.01, "learning_rate": 7.520244736368545e-06, "loss": 0.6632, "step": 2811 }, { "epoch": 1.01, "learning_rate": 7.519344970307721e-06, "loss": 0.6852, "step": 2812 }, { "epoch": 1.01, "learning_rate": 7.518445204246897e-06, "loss": 0.6605, "step": 2813 }, { "epoch": 1.01, "learning_rate": 7.517545438186072e-06, "loss": 0.5341, "step": 2814 }, { "epoch": 1.01, "learning_rate": 7.516645672125248e-06, "loss": 0.5753, "step": 2815 }, { "epoch": 1.01, "learning_rate": 7.515745906064423e-06, "loss": 0.5062, "step": 2816 }, { "epoch": 1.01, "learning_rate": 7.5148461400036e-06, "loss": 0.5199, "step": 2817 }, { "epoch": 1.01, "learning_rate": 7.513946373942775e-06, "loss": 0.5848, "step": 2818 }, { "epoch": 1.01, "learning_rate": 7.513046607881951e-06, "loss": 0.5934, "step": 2819 }, { "epoch": 1.01, "learning_rate": 7.512146841821128e-06, "loss": 0.5367, "step": 2820 }, { "epoch": 1.01, "learning_rate": 7.511247075760303e-06, "loss": 0.5314, "step": 2821 }, { "epoch": 1.01, "learning_rate": 7.5103473096994784e-06, "loss": 0.6913, "step": 2822 }, { "epoch": 1.01, "learning_rate": 7.509447543638655e-06, "loss": 0.7605, "step": 2823 }, { "epoch": 1.01, "learning_rate": 7.50854777757783e-06, "loss": 0.6244, "step": 2824 }, { "epoch": 1.01, "learning_rate": 7.507648011517006e-06, "loss": 0.6354, "step": 2825 }, { "epoch": 1.01, "learning_rate": 7.506748245456182e-06, "loss": 0.6653, "step": 2826 }, { "epoch": 1.01, "learning_rate": 7.505848479395358e-06, "loss": 0.7161, "step": 2827 }, { "epoch": 1.02, "learning_rate": 7.504948713334533e-06, "loss": 0.525, "step": 2828 }, { "epoch": 1.02, "learning_rate": 7.50404894727371e-06, "loss": 0.5697, "step": 2829 }, { "epoch": 1.02, "learning_rate": 7.503149181212885e-06, "loss": 0.6979, "step": 2830 }, { "epoch": 1.02, "learning_rate": 7.502249415152061e-06, "loss": 0.6079, "step": 2831 }, { "epoch": 1.02, "learning_rate": 7.501349649091238e-06, "loss": 0.6144, "step": 2832 }, { "epoch": 1.02, "learning_rate": 7.500449883030413e-06, "loss": 0.6471, "step": 2833 }, { "epoch": 1.02, "learning_rate": 7.4995501169695885e-06, "loss": 0.5551, "step": 2834 }, { "epoch": 1.02, "learning_rate": 7.498650350908763e-06, "loss": 0.5287, "step": 2835 }, { "epoch": 1.02, "learning_rate": 7.49775058484794e-06, "loss": 0.5076, "step": 2836 }, { "epoch": 1.02, "learning_rate": 7.496850818787116e-06, "loss": 0.6736, "step": 2837 }, { "epoch": 1.02, "learning_rate": 7.4959510527262916e-06, "loss": 0.5254, "step": 2838 }, { "epoch": 1.02, "learning_rate": 7.495051286665468e-06, "loss": 0.6043, "step": 2839 }, { "epoch": 1.02, "learning_rate": 7.494151520604643e-06, "loss": 0.6566, "step": 2840 }, { "epoch": 1.02, "learning_rate": 7.493251754543819e-06, "loss": 0.6021, "step": 2841 }, { "epoch": 1.02, "learning_rate": 7.4923519884829955e-06, "loss": 0.6055, "step": 2842 }, { "epoch": 1.02, "learning_rate": 7.491452222422171e-06, "loss": 0.647, "step": 2843 }, { "epoch": 1.02, "learning_rate": 7.490552456361346e-06, "loss": 0.5171, "step": 2844 }, { "epoch": 1.02, "learning_rate": 7.489652690300523e-06, "loss": 0.6114, "step": 2845 }, { "epoch": 1.02, "learning_rate": 7.4887529242396986e-06, "loss": 0.577, "step": 2846 }, { "epoch": 1.02, "learning_rate": 7.4878531581788735e-06, "loss": 0.6705, "step": 2847 }, { "epoch": 1.02, "learning_rate": 7.48695339211805e-06, "loss": 0.6037, "step": 2848 }, { "epoch": 1.02, "learning_rate": 7.486053626057226e-06, "loss": 0.6402, "step": 2849 }, { "epoch": 1.02, "learning_rate": 7.485153859996402e-06, "loss": 0.6811, "step": 2850 }, { "epoch": 1.02, "learning_rate": 7.484254093935578e-06, "loss": 0.6327, "step": 2851 }, { "epoch": 1.02, "learning_rate": 7.483354327874753e-06, "loss": 0.5739, "step": 2852 }, { "epoch": 1.02, "learning_rate": 7.482454561813929e-06, "loss": 0.6823, "step": 2853 }, { "epoch": 1.02, "learning_rate": 7.481554795753104e-06, "loss": 0.5482, "step": 2854 }, { "epoch": 1.02, "learning_rate": 7.4806550296922805e-06, "loss": 0.588, "step": 2855 }, { "epoch": 1.03, "learning_rate": 7.479755263631456e-06, "loss": 0.6451, "step": 2856 }, { "epoch": 1.03, "learning_rate": 7.478855497570632e-06, "loss": 0.6238, "step": 2857 }, { "epoch": 1.03, "learning_rate": 7.477955731509809e-06, "loss": 0.6711, "step": 2858 }, { "epoch": 1.03, "learning_rate": 7.4770559654489836e-06, "loss": 0.594, "step": 2859 }, { "epoch": 1.03, "learning_rate": 7.476156199388159e-06, "loss": 0.6146, "step": 2860 }, { "epoch": 1.03, "learning_rate": 7.475256433327336e-06, "loss": 0.63, "step": 2861 }, { "epoch": 1.03, "learning_rate": 7.474356667266512e-06, "loss": 0.59, "step": 2862 }, { "epoch": 1.03, "learning_rate": 7.473456901205687e-06, "loss": 0.5373, "step": 2863 }, { "epoch": 1.03, "learning_rate": 7.472557135144863e-06, "loss": 0.7597, "step": 2864 }, { "epoch": 1.03, "learning_rate": 7.471657369084039e-06, "loss": 0.6137, "step": 2865 }, { "epoch": 1.03, "learning_rate": 7.470757603023214e-06, "loss": 0.6046, "step": 2866 }, { "epoch": 1.03, "learning_rate": 7.4698578369623906e-06, "loss": 0.6151, "step": 2867 }, { "epoch": 1.03, "learning_rate": 7.468958070901566e-06, "loss": 0.7179, "step": 2868 }, { "epoch": 1.03, "learning_rate": 7.468058304840742e-06, "loss": 0.6212, "step": 2869 }, { "epoch": 1.03, "learning_rate": 7.467158538779919e-06, "loss": 0.5313, "step": 2870 }, { "epoch": 1.03, "learning_rate": 7.466258772719094e-06, "loss": 0.6239, "step": 2871 }, { "epoch": 1.03, "learning_rate": 7.465359006658269e-06, "loss": 0.6659, "step": 2872 }, { "epoch": 1.03, "learning_rate": 7.464459240597445e-06, "loss": 0.6455, "step": 2873 }, { "epoch": 1.03, "learning_rate": 7.463559474536621e-06, "loss": 0.6201, "step": 2874 }, { "epoch": 1.03, "learning_rate": 7.462659708475797e-06, "loss": 0.6571, "step": 2875 }, { "epoch": 1.03, "learning_rate": 7.4617599424149725e-06, "loss": 0.5742, "step": 2876 }, { "epoch": 1.03, "learning_rate": 7.460860176354149e-06, "loss": 0.6349, "step": 2877 }, { "epoch": 1.03, "learning_rate": 7.459960410293324e-06, "loss": 0.5998, "step": 2878 }, { "epoch": 1.03, "learning_rate": 7.4590606442325e-06, "loss": 0.6443, "step": 2879 }, { "epoch": 1.03, "learning_rate": 7.458160878171676e-06, "loss": 0.6144, "step": 2880 }, { "epoch": 1.03, "learning_rate": 7.457261112110852e-06, "loss": 0.6612, "step": 2881 }, { "epoch": 1.03, "learning_rate": 7.456361346050027e-06, "loss": 0.6423, "step": 2882 }, { "epoch": 1.03, "learning_rate": 7.455461579989204e-06, "loss": 0.6464, "step": 2883 }, { "epoch": 1.04, "learning_rate": 7.4545618139283795e-06, "loss": 0.6135, "step": 2884 }, { "epoch": 1.04, "learning_rate": 7.453662047867554e-06, "loss": 0.6327, "step": 2885 }, { "epoch": 1.04, "learning_rate": 7.452762281806731e-06, "loss": 0.6913, "step": 2886 }, { "epoch": 1.04, "learning_rate": 7.451862515745907e-06, "loss": 0.6864, "step": 2887 }, { "epoch": 1.04, "learning_rate": 7.4509627496850826e-06, "loss": 0.6161, "step": 2888 }, { "epoch": 1.04, "learning_rate": 7.450062983624259e-06, "loss": 0.6746, "step": 2889 }, { "epoch": 1.04, "learning_rate": 7.449163217563434e-06, "loss": 0.5825, "step": 2890 }, { "epoch": 1.04, "learning_rate": 7.44826345150261e-06, "loss": 0.6412, "step": 2891 }, { "epoch": 1.04, "learning_rate": 7.44826345150261e-06, "loss": 1.05, "step": 2892 }, { "epoch": 1.04, "learning_rate": 7.447363685441786e-06, "loss": 0.6384, "step": 2893 }, { "epoch": 1.04, "learning_rate": 7.446463919380961e-06, "loss": 0.541, "step": 2894 }, { "epoch": 1.04, "learning_rate": 7.445564153320137e-06, "loss": 0.5455, "step": 2895 }, { "epoch": 1.04, "learning_rate": 7.444664387259313e-06, "loss": 0.6251, "step": 2896 }, { "epoch": 1.04, "learning_rate": 7.4437646211984896e-06, "loss": 0.595, "step": 2897 }, { "epoch": 1.04, "learning_rate": 7.4428648551376645e-06, "loss": 0.7263, "step": 2898 }, { "epoch": 1.04, "learning_rate": 7.44196508907684e-06, "loss": 0.5252, "step": 2899 }, { "epoch": 1.04, "learning_rate": 7.441065323016017e-06, "loss": 0.6575, "step": 2900 }, { "epoch": 1.04, "learning_rate": 7.440165556955193e-06, "loss": 0.6111, "step": 2901 }, { "epoch": 1.04, "learning_rate": 7.4392657908943675e-06, "loss": 0.6324, "step": 2902 }, { "epoch": 1.04, "learning_rate": 7.438366024833544e-06, "loss": 0.5403, "step": 2903 }, { "epoch": 1.04, "learning_rate": 7.43746625877272e-06, "loss": 0.6537, "step": 2904 }, { "epoch": 1.04, "learning_rate": 7.436566492711895e-06, "loss": 0.6402, "step": 2905 }, { "epoch": 1.04, "learning_rate": 7.4356667266510715e-06, "loss": 0.6127, "step": 2906 }, { "epoch": 1.04, "learning_rate": 7.434766960590247e-06, "loss": 0.6976, "step": 2907 }, { "epoch": 1.04, "learning_rate": 7.433867194529423e-06, "loss": 0.6247, "step": 2908 }, { "epoch": 1.04, "learning_rate": 7.4329674284686e-06, "loss": 0.5214, "step": 2909 }, { "epoch": 1.04, "learning_rate": 7.4320676624077745e-06, "loss": 0.5166, "step": 2910 }, { "epoch": 1.04, "learning_rate": 7.43116789634695e-06, "loss": 0.6669, "step": 2911 }, { "epoch": 1.05, "learning_rate": 7.430268130286126e-06, "loss": 0.5902, "step": 2912 }, { "epoch": 1.05, "learning_rate": 7.429368364225303e-06, "loss": 0.534, "step": 2913 }, { "epoch": 1.05, "learning_rate": 7.428468598164478e-06, "loss": 0.6951, "step": 2914 }, { "epoch": 1.05, "learning_rate": 7.427568832103653e-06, "loss": 0.5777, "step": 2915 }, { "epoch": 1.05, "learning_rate": 7.42666906604283e-06, "loss": 0.6252, "step": 2916 }, { "epoch": 1.05, "learning_rate": 7.425769299982005e-06, "loss": 0.5867, "step": 2917 }, { "epoch": 1.05, "learning_rate": 7.424869533921181e-06, "loss": 0.5935, "step": 2918 }, { "epoch": 1.05, "learning_rate": 7.423969767860357e-06, "loss": 0.6017, "step": 2919 }, { "epoch": 1.05, "learning_rate": 7.423070001799533e-06, "loss": 0.6141, "step": 2920 }, { "epoch": 1.05, "learning_rate": 7.422170235738708e-06, "loss": 0.5521, "step": 2921 }, { "epoch": 1.05, "learning_rate": 7.421270469677885e-06, "loss": 0.5567, "step": 2922 }, { "epoch": 1.05, "learning_rate": 7.42037070361706e-06, "loss": 0.6129, "step": 2923 }, { "epoch": 1.05, "learning_rate": 7.419470937556235e-06, "loss": 0.5899, "step": 2924 }, { "epoch": 1.05, "learning_rate": 7.418571171495412e-06, "loss": 0.5396, "step": 2925 }, { "epoch": 1.05, "learning_rate": 7.417671405434588e-06, "loss": 0.6326, "step": 2926 }, { "epoch": 1.05, "learning_rate": 7.4167716393737635e-06, "loss": 0.656, "step": 2927 }, { "epoch": 1.05, "learning_rate": 7.41587187331294e-06, "loss": 0.5583, "step": 2928 }, { "epoch": 1.05, "learning_rate": 7.414972107252115e-06, "loss": 0.664, "step": 2929 }, { "epoch": 1.05, "learning_rate": 7.414072341191291e-06, "loss": 0.5705, "step": 2930 }, { "epoch": 1.05, "learning_rate": 7.4131725751304665e-06, "loss": 0.5622, "step": 2931 }, { "epoch": 1.05, "learning_rate": 7.412272809069643e-06, "loss": 0.578, "step": 2932 }, { "epoch": 1.05, "learning_rate": 7.411373043008818e-06, "loss": 0.584, "step": 2933 }, { "epoch": 1.05, "learning_rate": 7.410473276947994e-06, "loss": 0.6152, "step": 2934 }, { "epoch": 1.05, "learning_rate": 7.4095735108871705e-06, "loss": 0.6174, "step": 2935 }, { "epoch": 1.05, "learning_rate": 7.408673744826345e-06, "loss": 0.5077, "step": 2936 }, { "epoch": 1.05, "learning_rate": 7.407773978765521e-06, "loss": 0.5517, "step": 2937 }, { "epoch": 1.05, "learning_rate": 7.406874212704698e-06, "loss": 0.6359, "step": 2938 }, { "epoch": 1.05, "learning_rate": 7.4059744466438735e-06, "loss": 0.5946, "step": 2939 }, { "epoch": 1.06, "learning_rate": 7.4050746805830485e-06, "loss": 0.6432, "step": 2940 }, { "epoch": 1.06, "learning_rate": 7.404174914522225e-06, "loss": 0.5739, "step": 2941 }, { "epoch": 1.06, "learning_rate": 7.403275148461401e-06, "loss": 0.5241, "step": 2942 }, { "epoch": 1.06, "learning_rate": 7.402375382400576e-06, "loss": 0.5699, "step": 2943 }, { "epoch": 1.06, "learning_rate": 7.401475616339752e-06, "loss": 0.5392, "step": 2944 }, { "epoch": 1.06, "learning_rate": 7.400575850278928e-06, "loss": 0.6153, "step": 2945 }, { "epoch": 1.06, "learning_rate": 7.399676084218104e-06, "loss": 0.7091, "step": 2946 }, { "epoch": 1.06, "learning_rate": 7.3987763181572805e-06, "loss": 0.5786, "step": 2947 }, { "epoch": 1.06, "learning_rate": 7.3978765520964555e-06, "loss": 0.585, "step": 2948 }, { "epoch": 1.06, "learning_rate": 7.396976786035631e-06, "loss": 0.6049, "step": 2949 }, { "epoch": 1.06, "learning_rate": 7.396077019974807e-06, "loss": 0.5327, "step": 2950 }, { "epoch": 1.06, "learning_rate": 7.395177253913984e-06, "loss": 0.668, "step": 2951 }, { "epoch": 1.06, "learning_rate": 7.3942774878531585e-06, "loss": 0.6079, "step": 2952 }, { "epoch": 1.06, "learning_rate": 7.393377721792334e-06, "loss": 0.6107, "step": 2953 }, { "epoch": 1.06, "learning_rate": 7.392477955731511e-06, "loss": 0.5669, "step": 2954 }, { "epoch": 1.06, "learning_rate": 7.391578189670686e-06, "loss": 0.5326, "step": 2955 }, { "epoch": 1.06, "learning_rate": 7.390678423609862e-06, "loss": 0.5779, "step": 2956 }, { "epoch": 1.06, "learning_rate": 7.389778657549038e-06, "loss": 0.48, "step": 2957 }, { "epoch": 1.06, "learning_rate": 7.388878891488214e-06, "loss": 0.5653, "step": 2958 }, { "epoch": 1.06, "learning_rate": 7.387979125427389e-06, "loss": 0.675, "step": 2959 }, { "epoch": 1.06, "learning_rate": 7.3870793593665655e-06, "loss": 0.6326, "step": 2960 }, { "epoch": 1.06, "learning_rate": 7.386179593305741e-06, "loss": 0.556, "step": 2961 }, { "epoch": 1.06, "learning_rate": 7.385279827244917e-06, "loss": 0.5736, "step": 2962 }, { "epoch": 1.06, "learning_rate": 7.384380061184093e-06, "loss": 0.6757, "step": 2963 }, { "epoch": 1.06, "learning_rate": 7.383480295123269e-06, "loss": 0.6114, "step": 2964 }, { "epoch": 1.06, "learning_rate": 7.382580529062444e-06, "loss": 0.5902, "step": 2965 }, { "epoch": 1.06, "learning_rate": 7.381680763001621e-06, "loss": 0.6672, "step": 2966 }, { "epoch": 1.06, "learning_rate": 7.380780996940796e-06, "loss": 0.6218, "step": 2967 }, { "epoch": 1.07, "learning_rate": 7.379881230879972e-06, "loss": 0.5406, "step": 2968 }, { "epoch": 1.07, "learning_rate": 7.3789814648191474e-06, "loss": 0.5867, "step": 2969 }, { "epoch": 1.07, "learning_rate": 7.378081698758324e-06, "loss": 0.6707, "step": 2970 }, { "epoch": 1.07, "learning_rate": 7.377181932697499e-06, "loss": 0.7027, "step": 2971 }, { "epoch": 1.07, "learning_rate": 7.376282166636675e-06, "loss": 0.554, "step": 2972 }, { "epoch": 1.07, "learning_rate": 7.375382400575851e-06, "loss": 0.5168, "step": 2973 }, { "epoch": 1.07, "learning_rate": 7.374482634515026e-06, "loss": 0.503, "step": 2974 }, { "epoch": 1.07, "learning_rate": 7.373582868454202e-06, "loss": 0.7203, "step": 2975 }, { "epoch": 1.07, "learning_rate": 7.372683102393379e-06, "loss": 0.6051, "step": 2976 }, { "epoch": 1.07, "learning_rate": 7.3717833363325544e-06, "loss": 0.7012, "step": 2977 }, { "epoch": 1.07, "learning_rate": 7.370883570271729e-06, "loss": 0.5238, "step": 2978 }, { "epoch": 1.07, "learning_rate": 7.369983804210906e-06, "loss": 0.5971, "step": 2979 }, { "epoch": 1.07, "learning_rate": 7.369084038150082e-06, "loss": 0.6178, "step": 2980 }, { "epoch": 1.07, "learning_rate": 7.3681842720892575e-06, "loss": 0.6244, "step": 2981 }, { "epoch": 1.07, "learning_rate": 7.367284506028433e-06, "loss": 0.6374, "step": 2982 }, { "epoch": 1.07, "learning_rate": 7.366384739967609e-06, "loss": 0.6563, "step": 2983 }, { "epoch": 1.07, "learning_rate": 7.365484973906785e-06, "loss": 0.6305, "step": 2984 }, { "epoch": 1.07, "learning_rate": 7.3645852078459614e-06, "loss": 0.6036, "step": 2985 }, { "epoch": 1.07, "learning_rate": 7.363685441785136e-06, "loss": 0.6268, "step": 2986 }, { "epoch": 1.07, "learning_rate": 7.362785675724312e-06, "loss": 0.6905, "step": 2987 }, { "epoch": 1.07, "learning_rate": 7.361885909663488e-06, "loss": 0.6165, "step": 2988 }, { "epoch": 1.07, "learning_rate": 7.3609861436026645e-06, "loss": 0.6728, "step": 2989 }, { "epoch": 1.07, "learning_rate": 7.3600863775418394e-06, "loss": 0.6119, "step": 2990 }, { "epoch": 1.07, "learning_rate": 7.359186611481015e-06, "loss": 0.7153, "step": 2991 }, { "epoch": 1.07, "learning_rate": 7.358286845420192e-06, "loss": 0.5725, "step": 2992 }, { "epoch": 1.07, "learning_rate": 7.357387079359367e-06, "loss": 0.7358, "step": 2993 }, { "epoch": 1.07, "learning_rate": 7.3564873132985425e-06, "loss": 0.6064, "step": 2994 }, { "epoch": 1.08, "learning_rate": 7.355587547237719e-06, "loss": 0.5512, "step": 2995 }, { "epoch": 1.08, "learning_rate": 7.354687781176895e-06, "loss": 0.6238, "step": 2996 }, { "epoch": 1.08, "learning_rate": 7.35378801511607e-06, "loss": 0.6298, "step": 2997 }, { "epoch": 1.08, "learning_rate": 7.3528882490552464e-06, "loss": 0.57, "step": 2998 }, { "epoch": 1.08, "learning_rate": 7.351988482994422e-06, "loss": 0.5131, "step": 2999 }, { "epoch": 1.08, "learning_rate": 7.351088716933598e-06, "loss": 0.6257, "step": 3000 }, { "epoch": 1.08, "learning_rate": 7.350188950872775e-06, "loss": 0.6336, "step": 3001 }, { "epoch": 1.08, "learning_rate": 7.3492891848119495e-06, "loss": 0.63, "step": 3002 }, { "epoch": 1.08, "learning_rate": 7.348389418751125e-06, "loss": 0.6725, "step": 3003 }, { "epoch": 1.08, "learning_rate": 7.347489652690302e-06, "loss": 0.5944, "step": 3004 }, { "epoch": 1.08, "learning_rate": 7.346589886629477e-06, "loss": 0.5994, "step": 3005 }, { "epoch": 1.08, "learning_rate": 7.345690120568653e-06, "loss": 0.5952, "step": 3006 }, { "epoch": 1.08, "learning_rate": 7.344790354507828e-06, "loss": 0.6376, "step": 3007 }, { "epoch": 1.08, "learning_rate": 7.343890588447005e-06, "loss": 0.649, "step": 3008 }, { "epoch": 1.08, "learning_rate": 7.34299082238618e-06, "loss": 0.5546, "step": 3009 }, { "epoch": 1.08, "learning_rate": 7.342091056325356e-06, "loss": 0.6478, "step": 3010 }, { "epoch": 1.08, "learning_rate": 7.341191290264532e-06, "loss": 0.5901, "step": 3011 }, { "epoch": 1.08, "learning_rate": 7.340291524203707e-06, "loss": 0.7417, "step": 3012 }, { "epoch": 1.08, "learning_rate": 7.339391758142883e-06, "loss": 0.5545, "step": 3013 }, { "epoch": 1.08, "learning_rate": 7.33849199208206e-06, "loss": 0.6326, "step": 3014 }, { "epoch": 1.08, "learning_rate": 7.337592226021235e-06, "loss": 0.5469, "step": 3015 }, { "epoch": 1.08, "learning_rate": 7.33669245996041e-06, "loss": 0.6532, "step": 3016 }, { "epoch": 1.08, "learning_rate": 7.335792693899587e-06, "loss": 0.6518, "step": 3017 }, { "epoch": 1.08, "learning_rate": 7.334892927838763e-06, "loss": 0.5925, "step": 3018 }, { "epoch": 1.08, "learning_rate": 7.3339931617779384e-06, "loss": 0.647, "step": 3019 }, { "epoch": 1.08, "learning_rate": 7.333093395717115e-06, "loss": 0.553, "step": 3020 }, { "epoch": 1.08, "learning_rate": 7.33219362965629e-06, "loss": 0.5375, "step": 3021 }, { "epoch": 1.08, "learning_rate": 7.331293863595466e-06, "loss": 0.6012, "step": 3022 }, { "epoch": 1.09, "learning_rate": 7.330394097534641e-06, "loss": 0.4494, "step": 3023 }, { "epoch": 1.09, "learning_rate": 7.329494331473817e-06, "loss": 0.6042, "step": 3024 }, { "epoch": 1.09, "learning_rate": 7.328594565412993e-06, "loss": 0.6133, "step": 3025 }, { "epoch": 1.09, "learning_rate": 7.327694799352169e-06, "loss": 0.6233, "step": 3026 }, { "epoch": 1.09, "learning_rate": 7.326795033291345e-06, "loss": 0.6568, "step": 3027 }, { "epoch": 1.09, "learning_rate": 7.32589526723052e-06, "loss": 0.6497, "step": 3028 }, { "epoch": 1.09, "learning_rate": 7.324995501169696e-06, "loss": 0.604, "step": 3029 }, { "epoch": 1.09, "learning_rate": 7.324095735108873e-06, "loss": 0.6026, "step": 3030 }, { "epoch": 1.09, "learning_rate": 7.3231959690480485e-06, "loss": 0.5694, "step": 3031 }, { "epoch": 1.09, "learning_rate": 7.322296202987223e-06, "loss": 0.7053, "step": 3032 }, { "epoch": 1.09, "learning_rate": 7.3213964369264e-06, "loss": 0.6909, "step": 3033 }, { "epoch": 1.09, "learning_rate": 7.320496670865576e-06, "loss": 0.6834, "step": 3034 }, { "epoch": 1.09, "learning_rate": 7.319596904804751e-06, "loss": 0.6683, "step": 3035 }, { "epoch": 1.09, "learning_rate": 7.318697138743927e-06, "loss": 0.6039, "step": 3036 }, { "epoch": 1.09, "learning_rate": 7.317797372683103e-06, "loss": 0.5465, "step": 3037 }, { "epoch": 1.09, "learning_rate": 7.316897606622279e-06, "loss": 0.6135, "step": 3038 }, { "epoch": 1.09, "learning_rate": 7.3159978405614555e-06, "loss": 0.6213, "step": 3039 }, { "epoch": 1.09, "learning_rate": 7.31509807450063e-06, "loss": 0.6182, "step": 3040 }, { "epoch": 1.09, "learning_rate": 7.314198308439806e-06, "loss": 0.6413, "step": 3041 }, { "epoch": 1.09, "learning_rate": 7.313298542378981e-06, "loss": 0.5309, "step": 3042 }, { "epoch": 1.09, "learning_rate": 7.312398776318158e-06, "loss": 0.5868, "step": 3043 }, { "epoch": 1.09, "learning_rate": 7.3114990102573335e-06, "loss": 0.5469, "step": 3044 }, { "epoch": 1.09, "learning_rate": 7.310599244196509e-06, "loss": 0.5695, "step": 3045 }, { "epoch": 1.09, "learning_rate": 7.309699478135686e-06, "loss": 0.5572, "step": 3046 }, { "epoch": 1.09, "learning_rate": 7.308799712074861e-06, "loss": 0.6789, "step": 3047 }, { "epoch": 1.09, "learning_rate": 7.3078999460140366e-06, "loss": 0.5674, "step": 3048 }, { "epoch": 1.09, "learning_rate": 7.307000179953213e-06, "loss": 0.5953, "step": 3049 }, { "epoch": 1.09, "learning_rate": 7.306100413892389e-06, "loss": 0.5554, "step": 3050 }, { "epoch": 1.1, "learning_rate": 7.305200647831564e-06, "loss": 0.5648, "step": 3051 }, { "epoch": 1.1, "learning_rate": 7.3043008817707405e-06, "loss": 0.4878, "step": 3052 }, { "epoch": 1.1, "learning_rate": 7.303401115709916e-06, "loss": 0.6169, "step": 3053 }, { "epoch": 1.1, "learning_rate": 7.302501349649091e-06, "loss": 0.6085, "step": 3054 }, { "epoch": 1.1, "learning_rate": 7.301601583588268e-06, "loss": 0.5964, "step": 3055 }, { "epoch": 1.1, "learning_rate": 7.3007018175274436e-06, "loss": 0.5584, "step": 3056 }, { "epoch": 1.1, "learning_rate": 7.299802051466619e-06, "loss": 0.6285, "step": 3057 }, { "epoch": 1.1, "learning_rate": 7.298902285405796e-06, "loss": 0.5541, "step": 3058 }, { "epoch": 1.1, "learning_rate": 7.298002519344971e-06, "loss": 0.6312, "step": 3059 }, { "epoch": 1.1, "learning_rate": 7.297102753284147e-06, "loss": 0.6988, "step": 3060 }, { "epoch": 1.1, "learning_rate": 7.296202987223322e-06, "loss": 0.6275, "step": 3061 }, { "epoch": 1.1, "learning_rate": 7.295303221162498e-06, "loss": 0.6129, "step": 3062 }, { "epoch": 1.1, "learning_rate": 7.294403455101674e-06, "loss": 0.6354, "step": 3063 }, { "epoch": 1.1, "learning_rate": 7.29350368904085e-06, "loss": 0.6155, "step": 3064 }, { "epoch": 1.1, "learning_rate": 7.292603922980026e-06, "loss": 0.5322, "step": 3065 }, { "epoch": 1.1, "learning_rate": 7.291704156919201e-06, "loss": 0.6256, "step": 3066 }, { "epoch": 1.1, "learning_rate": 7.290804390858377e-06, "loss": 0.5756, "step": 3067 }, { "epoch": 1.1, "learning_rate": 7.289904624797554e-06, "loss": 0.5899, "step": 3068 }, { "epoch": 1.1, "learning_rate": 7.289004858736729e-06, "loss": 0.5081, "step": 3069 }, { "epoch": 1.1, "learning_rate": 7.288105092675904e-06, "loss": 0.6087, "step": 3070 }, { "epoch": 1.1, "learning_rate": 7.287205326615081e-06, "loss": 0.535, "step": 3071 }, { "epoch": 1.1, "learning_rate": 7.286305560554257e-06, "loss": 0.6464, "step": 3072 }, { "epoch": 1.1, "learning_rate": 7.285405794493432e-06, "loss": 0.6481, "step": 3073 }, { "epoch": 1.1, "learning_rate": 7.284506028432608e-06, "loss": 0.5625, "step": 3074 }, { "epoch": 1.1, "learning_rate": 7.283606262371784e-06, "loss": 0.5382, "step": 3075 }, { "epoch": 1.1, "learning_rate": 7.28270649631096e-06, "loss": 0.5595, "step": 3076 }, { "epoch": 1.1, "learning_rate": 7.281806730250136e-06, "loss": 0.6239, "step": 3077 }, { "epoch": 1.1, "learning_rate": 7.280906964189311e-06, "loss": 0.5785, "step": 3078 }, { "epoch": 1.11, "learning_rate": 7.280007198128487e-06, "loss": 0.6553, "step": 3079 }, { "epoch": 1.11, "learning_rate": 7.279107432067663e-06, "loss": 0.6155, "step": 3080 }, { "epoch": 1.11, "learning_rate": 7.278207666006839e-06, "loss": 0.6844, "step": 3081 }, { "epoch": 1.11, "learning_rate": 7.277307899946014e-06, "loss": 0.601, "step": 3082 }, { "epoch": 1.11, "learning_rate": 7.27640813388519e-06, "loss": 0.6107, "step": 3083 }, { "epoch": 1.11, "learning_rate": 7.275508367824367e-06, "loss": 0.6102, "step": 3084 }, { "epoch": 1.11, "learning_rate": 7.274608601763542e-06, "loss": 0.6681, "step": 3085 }, { "epoch": 1.11, "learning_rate": 7.2737088357027175e-06, "loss": 0.5033, "step": 3086 }, { "epoch": 1.11, "learning_rate": 7.272809069641894e-06, "loss": 0.6552, "step": 3087 }, { "epoch": 1.11, "learning_rate": 7.27190930358107e-06, "loss": 0.5767, "step": 3088 }, { "epoch": 1.11, "learning_rate": 7.271009537520245e-06, "loss": 0.5995, "step": 3089 }, { "epoch": 1.11, "learning_rate": 7.270109771459421e-06, "loss": 0.6141, "step": 3090 }, { "epoch": 1.11, "learning_rate": 7.269210005398597e-06, "loss": 0.6489, "step": 3091 }, { "epoch": 1.11, "learning_rate": 7.268310239337772e-06, "loss": 0.6713, "step": 3092 }, { "epoch": 1.11, "learning_rate": 7.267410473276949e-06, "loss": 0.6178, "step": 3093 }, { "epoch": 1.11, "learning_rate": 7.2665107072161245e-06, "loss": 0.6293, "step": 3094 }, { "epoch": 1.11, "learning_rate": 7.2656109411553e-06, "loss": 0.5688, "step": 3095 }, { "epoch": 1.11, "learning_rate": 7.264711175094477e-06, "loss": 0.6656, "step": 3096 }, { "epoch": 1.11, "learning_rate": 7.263811409033652e-06, "loss": 0.708, "step": 3097 }, { "epoch": 1.11, "learning_rate": 7.2629116429728275e-06, "loss": 0.5859, "step": 3098 }, { "epoch": 1.11, "learning_rate": 7.262011876912003e-06, "loss": 0.5616, "step": 3099 }, { "epoch": 1.11, "learning_rate": 7.261112110851179e-06, "loss": 0.5933, "step": 3100 }, { "epoch": 1.11, "learning_rate": 7.260212344790355e-06, "loss": 0.5583, "step": 3101 }, { "epoch": 1.11, "learning_rate": 7.259312578729531e-06, "loss": 0.6432, "step": 3102 }, { "epoch": 1.11, "learning_rate": 7.258412812668707e-06, "loss": 0.5693, "step": 3103 }, { "epoch": 1.11, "learning_rate": 7.257513046607882e-06, "loss": 0.6354, "step": 3104 }, { "epoch": 1.11, "learning_rate": 7.256613280547058e-06, "loss": 0.6536, "step": 3105 }, { "epoch": 1.11, "learning_rate": 7.2557135144862345e-06, "loss": 0.5749, "step": 3106 }, { "epoch": 1.12, "learning_rate": 7.25481374842541e-06, "loss": 0.5297, "step": 3107 }, { "epoch": 1.12, "learning_rate": 7.253913982364585e-06, "loss": 0.7164, "step": 3108 }, { "epoch": 1.12, "learning_rate": 7.253014216303762e-06, "loss": 0.6439, "step": 3109 }, { "epoch": 1.12, "learning_rate": 7.252114450242938e-06, "loss": 0.6229, "step": 3110 }, { "epoch": 1.12, "learning_rate": 7.2512146841821125e-06, "loss": 0.6982, "step": 3111 }, { "epoch": 1.12, "learning_rate": 7.250314918121289e-06, "loss": 0.6335, "step": 3112 }, { "epoch": 1.12, "learning_rate": 7.249415152060465e-06, "loss": 0.6281, "step": 3113 }, { "epoch": 1.12, "learning_rate": 7.248515385999641e-06, "loss": 0.5511, "step": 3114 }, { "epoch": 1.12, "learning_rate": 7.247615619938817e-06, "loss": 0.6024, "step": 3115 }, { "epoch": 1.12, "learning_rate": 7.246715853877992e-06, "loss": 0.6022, "step": 3116 }, { "epoch": 1.12, "learning_rate": 7.245816087817168e-06, "loss": 0.6308, "step": 3117 }, { "epoch": 1.12, "learning_rate": 7.244916321756344e-06, "loss": 0.5451, "step": 3118 }, { "epoch": 1.12, "learning_rate": 7.24401655569552e-06, "loss": 0.6976, "step": 3119 }, { "epoch": 1.12, "learning_rate": 7.243116789634695e-06, "loss": 0.5877, "step": 3120 }, { "epoch": 1.12, "learning_rate": 7.242217023573871e-06, "loss": 0.6229, "step": 3121 }, { "epoch": 1.12, "learning_rate": 7.241317257513048e-06, "loss": 0.6318, "step": 3122 }, { "epoch": 1.12, "learning_rate": 7.240417491452223e-06, "loss": 0.6491, "step": 3123 }, { "epoch": 1.12, "learning_rate": 7.239517725391398e-06, "loss": 0.6338, "step": 3124 }, { "epoch": 1.12, "learning_rate": 7.238617959330575e-06, "loss": 0.5983, "step": 3125 }, { "epoch": 1.12, "learning_rate": 7.237718193269751e-06, "loss": 0.7726, "step": 3126 }, { "epoch": 1.12, "learning_rate": 7.236818427208926e-06, "loss": 0.6066, "step": 3127 }, { "epoch": 1.12, "learning_rate": 7.235918661148102e-06, "loss": 0.6223, "step": 3128 }, { "epoch": 1.12, "learning_rate": 7.235018895087278e-06, "loss": 0.643, "step": 3129 }, { "epoch": 1.12, "learning_rate": 7.234119129026453e-06, "loss": 0.5582, "step": 3130 }, { "epoch": 1.12, "learning_rate": 7.23321936296563e-06, "loss": 0.6353, "step": 3131 }, { "epoch": 1.12, "learning_rate": 7.232319596904805e-06, "loss": 0.6172, "step": 3132 }, { "epoch": 1.12, "learning_rate": 7.231419830843981e-06, "loss": 0.5099, "step": 3133 }, { "epoch": 1.12, "learning_rate": 7.230520064783158e-06, "loss": 0.5972, "step": 3134 }, { "epoch": 1.13, "learning_rate": 7.229620298722333e-06, "loss": 0.5942, "step": 3135 }, { "epoch": 1.13, "learning_rate": 7.2287205326615085e-06, "loss": 0.6004, "step": 3136 }, { "epoch": 1.13, "learning_rate": 7.227820766600684e-06, "loss": 0.5707, "step": 3137 }, { "epoch": 1.13, "learning_rate": 7.226921000539861e-06, "loss": 0.5945, "step": 3138 }, { "epoch": 1.13, "learning_rate": 7.226021234479036e-06, "loss": 0.6055, "step": 3139 }, { "epoch": 1.13, "learning_rate": 7.2251214684182115e-06, "loss": 0.6975, "step": 3140 }, { "epoch": 1.13, "learning_rate": 7.224221702357388e-06, "loss": 0.5512, "step": 3141 }, { "epoch": 1.13, "learning_rate": 7.223321936296563e-06, "loss": 0.6725, "step": 3142 }, { "epoch": 1.13, "learning_rate": 7.222422170235739e-06, "loss": 0.5969, "step": 3143 }, { "epoch": 1.13, "learning_rate": 7.2215224041749155e-06, "loss": 0.5993, "step": 3144 }, { "epoch": 1.13, "learning_rate": 7.220622638114091e-06, "loss": 0.6434, "step": 3145 }, { "epoch": 1.13, "learning_rate": 7.219722872053266e-06, "loss": 0.728, "step": 3146 }, { "epoch": 1.13, "learning_rate": 7.218823105992443e-06, "loss": 0.564, "step": 3147 }, { "epoch": 1.13, "learning_rate": 7.2179233399316185e-06, "loss": 0.616, "step": 3148 }, { "epoch": 1.13, "learning_rate": 7.217023573870794e-06, "loss": 0.6316, "step": 3149 }, { "epoch": 1.13, "learning_rate": 7.21612380780997e-06, "loss": 0.6179, "step": 3150 }, { "epoch": 1.13, "learning_rate": 7.215224041749146e-06, "loss": 0.6142, "step": 3151 }, { "epoch": 1.13, "learning_rate": 7.214324275688322e-06, "loss": 0.5341, "step": 3152 }, { "epoch": 1.13, "learning_rate": 7.213424509627498e-06, "loss": 0.5812, "step": 3153 }, { "epoch": 1.13, "learning_rate": 7.212524743566673e-06, "loss": 0.5659, "step": 3154 }, { "epoch": 1.13, "learning_rate": 7.211624977505849e-06, "loss": 0.5602, "step": 3155 }, { "epoch": 1.13, "learning_rate": 7.210725211445025e-06, "loss": 0.6593, "step": 3156 }, { "epoch": 1.13, "learning_rate": 7.209825445384201e-06, "loss": 0.5736, "step": 3157 }, { "epoch": 1.13, "learning_rate": 7.208925679323376e-06, "loss": 0.6517, "step": 3158 }, { "epoch": 1.13, "learning_rate": 7.208025913262552e-06, "loss": 0.641, "step": 3159 }, { "epoch": 1.13, "learning_rate": 7.207126147201729e-06, "loss": 0.664, "step": 3160 }, { "epoch": 1.13, "learning_rate": 7.2062263811409035e-06, "loss": 0.6871, "step": 3161 }, { "epoch": 1.13, "learning_rate": 7.205326615080079e-06, "loss": 0.7555, "step": 3162 }, { "epoch": 1.14, "learning_rate": 7.204426849019256e-06, "loss": 0.6166, "step": 3163 }, { "epoch": 1.14, "learning_rate": 7.203527082958432e-06, "loss": 0.7321, "step": 3164 }, { "epoch": 1.14, "learning_rate": 7.202627316897607e-06, "loss": 0.7706, "step": 3165 }, { "epoch": 1.14, "learning_rate": 7.201727550836783e-06, "loss": 0.5691, "step": 3166 }, { "epoch": 1.14, "learning_rate": 7.200827784775959e-06, "loss": 0.637, "step": 3167 }, { "epoch": 1.14, "learning_rate": 7.199928018715135e-06, "loss": 0.5636, "step": 3168 }, { "epoch": 1.14, "learning_rate": 7.1990282526543105e-06, "loss": 0.5837, "step": 3169 }, { "epoch": 1.14, "learning_rate": 7.198128486593486e-06, "loss": 0.5647, "step": 3170 }, { "epoch": 1.14, "learning_rate": 7.197228720532662e-06, "loss": 0.5886, "step": 3171 }, { "epoch": 1.14, "learning_rate": 7.196328954471839e-06, "loss": 0.5376, "step": 3172 }, { "epoch": 1.14, "learning_rate": 7.195429188411014e-06, "loss": 0.6074, "step": 3173 }, { "epoch": 1.14, "learning_rate": 7.194529422350189e-06, "loss": 0.5723, "step": 3174 }, { "epoch": 1.14, "learning_rate": 7.193629656289365e-06, "loss": 0.6863, "step": 3175 }, { "epoch": 1.14, "learning_rate": 7.192729890228542e-06, "loss": 0.6003, "step": 3176 }, { "epoch": 1.14, "learning_rate": 7.191830124167717e-06, "loss": 0.6397, "step": 3177 }, { "epoch": 1.14, "learning_rate": 7.1909303581068924e-06, "loss": 0.6327, "step": 3178 }, { "epoch": 1.14, "learning_rate": 7.190030592046069e-06, "loss": 0.6015, "step": 3179 }, { "epoch": 1.14, "learning_rate": 7.189130825985244e-06, "loss": 0.6039, "step": 3180 }, { "epoch": 1.14, "learning_rate": 7.18823105992442e-06, "loss": 0.7192, "step": 3181 }, { "epoch": 1.14, "learning_rate": 7.187331293863596e-06, "loss": 0.5612, "step": 3182 }, { "epoch": 1.14, "learning_rate": 7.186431527802772e-06, "loss": 0.5746, "step": 3183 }, { "epoch": 1.14, "learning_rate": 7.185531761741947e-06, "loss": 0.6229, "step": 3184 }, { "epoch": 1.14, "learning_rate": 7.184631995681124e-06, "loss": 0.5628, "step": 3185 }, { "epoch": 1.14, "learning_rate": 7.1837322296202994e-06, "loss": 0.5892, "step": 3186 }, { "epoch": 1.14, "learning_rate": 7.182832463559475e-06, "loss": 0.7101, "step": 3187 }, { "epoch": 1.14, "learning_rate": 7.181932697498652e-06, "loss": 0.5287, "step": 3188 }, { "epoch": 1.14, "learning_rate": 7.181032931437827e-06, "loss": 0.5487, "step": 3189 }, { "epoch": 1.15, "learning_rate": 7.1801331653770025e-06, "loss": 0.6599, "step": 3190 }, { "epoch": 1.15, "learning_rate": 7.179233399316179e-06, "loss": 0.568, "step": 3191 }, { "epoch": 1.15, "learning_rate": 7.178333633255354e-06, "loss": 0.7218, "step": 3192 }, { "epoch": 1.15, "learning_rate": 7.17743386719453e-06, "loss": 0.6798, "step": 3193 }, { "epoch": 1.15, "learning_rate": 7.176534101133706e-06, "loss": 0.6245, "step": 3194 }, { "epoch": 1.15, "learning_rate": 7.175634335072882e-06, "loss": 0.5589, "step": 3195 }, { "epoch": 1.15, "learning_rate": 7.174734569012057e-06, "loss": 0.656, "step": 3196 }, { "epoch": 1.15, "learning_rate": 7.173834802951233e-06, "loss": 0.6197, "step": 3197 }, { "epoch": 1.15, "learning_rate": 7.1729350368904095e-06, "loss": 0.5357, "step": 3198 }, { "epoch": 1.15, "learning_rate": 7.1720352708295844e-06, "loss": 0.6787, "step": 3199 }, { "epoch": 1.15, "learning_rate": 7.17113550476876e-06, "loss": 0.6791, "step": 3200 }, { "epoch": 1.15, "learning_rate": 7.170235738707937e-06, "loss": 0.6494, "step": 3201 }, { "epoch": 1.15, "learning_rate": 7.169335972647113e-06, "loss": 0.5563, "step": 3202 }, { "epoch": 1.15, "learning_rate": 7.1684362065862875e-06, "loss": 0.6342, "step": 3203 }, { "epoch": 1.15, "learning_rate": 7.167536440525464e-06, "loss": 0.6125, "step": 3204 }, { "epoch": 1.15, "learning_rate": 7.16663667446464e-06, "loss": 0.5467, "step": 3205 }, { "epoch": 1.15, "learning_rate": 7.165736908403816e-06, "loss": 0.6296, "step": 3206 }, { "epoch": 1.15, "learning_rate": 7.164837142342992e-06, "loss": 0.6876, "step": 3207 }, { "epoch": 1.15, "learning_rate": 7.163937376282167e-06, "loss": 0.6562, "step": 3208 }, { "epoch": 1.15, "learning_rate": 7.163037610221343e-06, "loss": 0.6375, "step": 3209 }, { "epoch": 1.15, "learning_rate": 7.16213784416052e-06, "loss": 0.5528, "step": 3210 }, { "epoch": 1.15, "learning_rate": 7.1612380780996945e-06, "loss": 0.686, "step": 3211 }, { "epoch": 1.15, "learning_rate": 7.16033831203887e-06, "loss": 0.595, "step": 3212 }, { "epoch": 1.15, "learning_rate": 7.159438545978046e-06, "loss": 0.584, "step": 3213 }, { "epoch": 1.15, "learning_rate": 7.158538779917223e-06, "loss": 0.5788, "step": 3214 }, { "epoch": 1.15, "learning_rate": 7.157639013856398e-06, "loss": 0.589, "step": 3215 }, { "epoch": 1.15, "learning_rate": 7.156739247795573e-06, "loss": 0.5881, "step": 3216 }, { "epoch": 1.15, "learning_rate": 7.15583948173475e-06, "loss": 0.6501, "step": 3217 }, { "epoch": 1.16, "learning_rate": 7.154939715673926e-06, "loss": 0.6308, "step": 3218 }, { "epoch": 1.16, "learning_rate": 7.154039949613101e-06, "loss": 0.6098, "step": 3219 }, { "epoch": 1.16, "learning_rate": 7.153140183552277e-06, "loss": 0.555, "step": 3220 }, { "epoch": 1.16, "learning_rate": 7.152240417491453e-06, "loss": 0.5955, "step": 3221 }, { "epoch": 1.16, "learning_rate": 7.151340651430628e-06, "loss": 0.5774, "step": 3222 }, { "epoch": 1.16, "learning_rate": 7.150440885369805e-06, "loss": 0.6184, "step": 3223 }, { "epoch": 1.16, "learning_rate": 7.14954111930898e-06, "loss": 0.5924, "step": 3224 }, { "epoch": 1.16, "learning_rate": 7.148641353248156e-06, "loss": 0.5579, "step": 3225 }, { "epoch": 1.16, "learning_rate": 7.147741587187333e-06, "loss": 0.6696, "step": 3226 }, { "epoch": 1.16, "learning_rate": 7.146841821126508e-06, "loss": 0.4741, "step": 3227 }, { "epoch": 1.16, "learning_rate": 7.145942055065683e-06, "loss": 0.6114, "step": 3228 }, { "epoch": 1.16, "learning_rate": 7.14504228900486e-06, "loss": 0.6359, "step": 3229 }, { "epoch": 1.16, "learning_rate": 7.144142522944035e-06, "loss": 0.5926, "step": 3230 }, { "epoch": 1.16, "learning_rate": 7.143242756883211e-06, "loss": 0.6213, "step": 3231 }, { "epoch": 1.16, "learning_rate": 7.1423429908223865e-06, "loss": 0.719, "step": 3232 }, { "epoch": 1.16, "learning_rate": 7.141443224761563e-06, "loss": 0.6311, "step": 3233 }, { "epoch": 1.16, "learning_rate": 7.140543458700738e-06, "loss": 0.6008, "step": 3234 }, { "epoch": 1.16, "learning_rate": 7.139643692639914e-06, "loss": 0.6131, "step": 3235 }, { "epoch": 1.16, "learning_rate": 7.13874392657909e-06, "loss": 0.6158, "step": 3236 }, { "epoch": 1.16, "learning_rate": 7.137844160518266e-06, "loss": 0.5815, "step": 3237 }, { "epoch": 1.16, "learning_rate": 7.136944394457441e-06, "loss": 0.5845, "step": 3238 }, { "epoch": 1.16, "learning_rate": 7.136044628396618e-06, "loss": 0.6078, "step": 3239 }, { "epoch": 1.16, "learning_rate": 7.1351448623357935e-06, "loss": 0.7186, "step": 3240 }, { "epoch": 1.16, "learning_rate": 7.134245096274968e-06, "loss": 0.6408, "step": 3241 }, { "epoch": 1.16, "learning_rate": 7.133345330214145e-06, "loss": 0.5358, "step": 3242 }, { "epoch": 1.16, "learning_rate": 7.132445564153321e-06, "loss": 0.6337, "step": 3243 }, { "epoch": 1.16, "learning_rate": 7.1315457980924966e-06, "loss": 0.5948, "step": 3244 }, { "epoch": 1.16, "learning_rate": 7.130646032031673e-06, "loss": 0.6215, "step": 3245 }, { "epoch": 1.17, "learning_rate": 7.129746265970848e-06, "loss": 0.7243, "step": 3246 }, { "epoch": 1.17, "learning_rate": 7.128846499910024e-06, "loss": 0.6691, "step": 3247 }, { "epoch": 1.17, "learning_rate": 7.1279467338492005e-06, "loss": 0.5883, "step": 3248 }, { "epoch": 1.17, "learning_rate": 7.127046967788375e-06, "loss": 0.683, "step": 3249 }, { "epoch": 1.17, "learning_rate": 7.126147201727551e-06, "loss": 0.6281, "step": 3250 }, { "epoch": 1.17, "learning_rate": 7.125247435666727e-06, "loss": 0.61, "step": 3251 }, { "epoch": 1.17, "learning_rate": 7.1243476696059036e-06, "loss": 0.6084, "step": 3252 }, { "epoch": 1.17, "learning_rate": 7.1234479035450785e-06, "loss": 0.6764, "step": 3253 }, { "epoch": 1.17, "learning_rate": 7.122548137484254e-06, "loss": 0.6323, "step": 3254 }, { "epoch": 1.17, "learning_rate": 7.121648371423431e-06, "loss": 0.5656, "step": 3255 }, { "epoch": 1.17, "learning_rate": 7.120748605362607e-06, "loss": 0.5805, "step": 3256 }, { "epoch": 1.17, "learning_rate": 7.1198488393017816e-06, "loss": 0.5826, "step": 3257 }, { "epoch": 1.17, "learning_rate": 7.118949073240958e-06, "loss": 0.7029, "step": 3258 }, { "epoch": 1.17, "learning_rate": 7.118049307180134e-06, "loss": 0.6349, "step": 3259 }, { "epoch": 1.17, "learning_rate": 7.117149541119309e-06, "loss": 0.6664, "step": 3260 }, { "epoch": 1.17, "learning_rate": 7.1162497750584855e-06, "loss": 0.596, "step": 3261 }, { "epoch": 1.17, "learning_rate": 7.115350008997661e-06, "loss": 0.584, "step": 3262 }, { "epoch": 1.17, "learning_rate": 7.114450242936837e-06, "loss": 0.6081, "step": 3263 }, { "epoch": 1.17, "learning_rate": 7.113550476876014e-06, "loss": 0.5621, "step": 3264 }, { "epoch": 1.17, "learning_rate": 7.1126507108151886e-06, "loss": 0.6706, "step": 3265 }, { "epoch": 1.17, "learning_rate": 7.111750944754364e-06, "loss": 0.6023, "step": 3266 }, { "epoch": 1.17, "learning_rate": 7.110851178693541e-06, "loss": 0.6541, "step": 3267 }, { "epoch": 1.17, "learning_rate": 7.109951412632716e-06, "loss": 0.5962, "step": 3268 }, { "epoch": 1.17, "learning_rate": 7.109051646571892e-06, "loss": 0.6946, "step": 3269 }, { "epoch": 1.17, "learning_rate": 7.108151880511067e-06, "loss": 0.6643, "step": 3270 }, { "epoch": 1.17, "learning_rate": 7.107252114450244e-06, "loss": 0.511, "step": 3271 }, { "epoch": 1.17, "learning_rate": 7.106352348389419e-06, "loss": 0.6112, "step": 3272 }, { "epoch": 1.17, "learning_rate": 7.105452582328595e-06, "loss": 0.7644, "step": 3273 }, { "epoch": 1.18, "learning_rate": 7.104552816267771e-06, "loss": 0.7029, "step": 3274 }, { "epoch": 1.18, "learning_rate": 7.103653050206947e-06, "loss": 0.5829, "step": 3275 }, { "epoch": 1.18, "learning_rate": 7.102753284146122e-06, "loss": 0.5885, "step": 3276 }, { "epoch": 1.18, "learning_rate": 7.101853518085299e-06, "loss": 0.6096, "step": 3277 }, { "epoch": 1.18, "learning_rate": 7.100953752024474e-06, "loss": 0.6219, "step": 3278 }, { "epoch": 1.18, "learning_rate": 7.100053985963649e-06, "loss": 0.6651, "step": 3279 }, { "epoch": 1.18, "learning_rate": 7.099154219902826e-06, "loss": 0.6047, "step": 3280 }, { "epoch": 1.18, "learning_rate": 7.098254453842002e-06, "loss": 0.6681, "step": 3281 }, { "epoch": 1.18, "learning_rate": 7.0973546877811775e-06, "loss": 0.6389, "step": 3282 }, { "epoch": 1.18, "learning_rate": 7.096454921720354e-06, "loss": 0.5375, "step": 3283 }, { "epoch": 1.18, "learning_rate": 7.095555155659529e-06, "loss": 0.6498, "step": 3284 }, { "epoch": 1.18, "learning_rate": 7.094655389598705e-06, "loss": 0.6467, "step": 3285 }, { "epoch": 1.18, "learning_rate": 7.093755623537881e-06, "loss": 0.6791, "step": 3286 }, { "epoch": 1.18, "learning_rate": 7.092855857477056e-06, "loss": 0.5624, "step": 3287 }, { "epoch": 1.18, "learning_rate": 7.091956091416232e-06, "loss": 0.5522, "step": 3288 }, { "epoch": 1.18, "learning_rate": 7.091056325355408e-06, "loss": 0.593, "step": 3289 }, { "epoch": 1.18, "learning_rate": 7.0901565592945845e-06, "loss": 0.5041, "step": 3290 }, { "epoch": 1.18, "learning_rate": 7.089256793233759e-06, "loss": 0.5819, "step": 3291 }, { "epoch": 1.18, "learning_rate": 7.088357027172935e-06, "loss": 0.5995, "step": 3292 }, { "epoch": 1.18, "learning_rate": 7.087457261112112e-06, "loss": 0.5989, "step": 3293 }, { "epoch": 1.18, "learning_rate": 7.0865574950512875e-06, "loss": 0.6246, "step": 3294 }, { "epoch": 1.18, "learning_rate": 7.0856577289904625e-06, "loss": 0.6484, "step": 3295 }, { "epoch": 1.18, "learning_rate": 7.084757962929639e-06, "loss": 0.6173, "step": 3296 }, { "epoch": 1.18, "learning_rate": 7.083858196868815e-06, "loss": 0.6225, "step": 3297 }, { "epoch": 1.18, "learning_rate": 7.08295843080799e-06, "loss": 0.5467, "step": 3298 }, { "epoch": 1.18, "learning_rate": 7.082058664747166e-06, "loss": 0.6409, "step": 3299 }, { "epoch": 1.18, "learning_rate": 7.081158898686342e-06, "loss": 0.5855, "step": 3300 }, { "epoch": 1.18, "learning_rate": 7.080259132625518e-06, "loss": 0.7064, "step": 3301 }, { "epoch": 1.19, "learning_rate": 7.0793593665646945e-06, "loss": 0.5675, "step": 3302 }, { "epoch": 1.19, "learning_rate": 7.0784596005038695e-06, "loss": 0.5486, "step": 3303 }, { "epoch": 1.19, "learning_rate": 7.077559834443045e-06, "loss": 0.6437, "step": 3304 }, { "epoch": 1.19, "learning_rate": 7.076660068382222e-06, "loss": 0.6001, "step": 3305 }, { "epoch": 1.19, "learning_rate": 7.075760302321398e-06, "loss": 0.6272, "step": 3306 }, { "epoch": 1.19, "learning_rate": 7.0748605362605725e-06, "loss": 0.7062, "step": 3307 }, { "epoch": 1.19, "learning_rate": 7.073960770199748e-06, "loss": 0.6296, "step": 3308 }, { "epoch": 1.19, "learning_rate": 7.073061004138925e-06, "loss": 0.5848, "step": 3309 }, { "epoch": 1.19, "learning_rate": 7.0721612380781e-06, "loss": 0.5726, "step": 3310 }, { "epoch": 1.19, "learning_rate": 7.071261472017276e-06, "loss": 0.6989, "step": 3311 }, { "epoch": 1.19, "learning_rate": 7.070361705956452e-06, "loss": 0.6968, "step": 3312 }, { "epoch": 1.19, "learning_rate": 7.069461939895628e-06, "loss": 0.6379, "step": 3313 }, { "epoch": 1.19, "learning_rate": 7.068562173834803e-06, "loss": 0.547, "step": 3314 }, { "epoch": 1.19, "learning_rate": 7.0676624077739795e-06, "loss": 0.6685, "step": 3315 }, { "epoch": 1.19, "learning_rate": 7.066762641713155e-06, "loss": 0.6161, "step": 3316 }, { "epoch": 1.19, "learning_rate": 7.06586287565233e-06, "loss": 0.5962, "step": 3317 }, { "epoch": 1.19, "learning_rate": 7.064963109591507e-06, "loss": 0.6909, "step": 3318 }, { "epoch": 1.19, "learning_rate": 7.064063343530683e-06, "loss": 0.5319, "step": 3319 }, { "epoch": 1.19, "learning_rate": 7.063163577469858e-06, "loss": 0.698, "step": 3320 }, { "epoch": 1.19, "learning_rate": 7.062263811409035e-06, "loss": 0.6509, "step": 3321 }, { "epoch": 1.19, "learning_rate": 7.06136404534821e-06, "loss": 0.6332, "step": 3322 }, { "epoch": 1.19, "learning_rate": 7.060464279287386e-06, "loss": 0.6346, "step": 3323 }, { "epoch": 1.19, "learning_rate": 7.059564513226562e-06, "loss": 0.5982, "step": 3324 }, { "epoch": 1.19, "learning_rate": 7.058664747165738e-06, "loss": 0.6075, "step": 3325 }, { "epoch": 1.19, "learning_rate": 7.057764981104913e-06, "loss": 0.6215, "step": 3326 }, { "epoch": 1.19, "learning_rate": 7.056865215044089e-06, "loss": 0.5859, "step": 3327 }, { "epoch": 1.19, "learning_rate": 7.055965448983265e-06, "loss": 0.5194, "step": 3328 }, { "epoch": 1.19, "learning_rate": 7.05506568292244e-06, "loss": 0.7506, "step": 3329 }, { "epoch": 1.2, "learning_rate": 7.054165916861616e-06, "loss": 0.6306, "step": 3330 }, { "epoch": 1.2, "learning_rate": 7.053266150800793e-06, "loss": 0.668, "step": 3331 }, { "epoch": 1.2, "learning_rate": 7.0523663847399685e-06, "loss": 0.6529, "step": 3332 }, { "epoch": 1.2, "learning_rate": 7.051466618679143e-06, "loss": 0.6598, "step": 3333 }, { "epoch": 1.2, "learning_rate": 7.05056685261832e-06, "loss": 0.6321, "step": 3334 }, { "epoch": 1.2, "learning_rate": 7.049667086557496e-06, "loss": 0.5803, "step": 3335 }, { "epoch": 1.2, "learning_rate": 7.0487673204966715e-06, "loss": 0.6164, "step": 3336 }, { "epoch": 1.2, "learning_rate": 7.047867554435847e-06, "loss": 0.6462, "step": 3337 }, { "epoch": 1.2, "learning_rate": 7.046967788375023e-06, "loss": 0.689, "step": 3338 }, { "epoch": 1.2, "learning_rate": 7.046068022314199e-06, "loss": 0.7427, "step": 3339 }, { "epoch": 1.2, "learning_rate": 7.0451682562533755e-06, "loss": 0.6072, "step": 3340 }, { "epoch": 1.2, "learning_rate": 7.04426849019255e-06, "loss": 0.6477, "step": 3341 }, { "epoch": 1.2, "learning_rate": 7.043368724131726e-06, "loss": 0.5521, "step": 3342 }, { "epoch": 1.2, "learning_rate": 7.042468958070903e-06, "loss": 0.6254, "step": 3343 }, { "epoch": 1.2, "learning_rate": 7.0415691920100785e-06, "loss": 0.5885, "step": 3344 }, { "epoch": 1.2, "learning_rate": 7.0406694259492535e-06, "loss": 0.5443, "step": 3345 }, { "epoch": 1.2, "learning_rate": 7.039769659888429e-06, "loss": 0.5548, "step": 3346 }, { "epoch": 1.2, "learning_rate": 7.038869893827606e-06, "loss": 0.6018, "step": 3347 }, { "epoch": 1.2, "learning_rate": 7.037970127766781e-06, "loss": 0.5806, "step": 3348 }, { "epoch": 1.2, "learning_rate": 7.0370703617059565e-06, "loss": 0.6113, "step": 3349 }, { "epoch": 1.2, "learning_rate": 7.036170595645133e-06, "loss": 0.594, "step": 3350 }, { "epoch": 1.2, "learning_rate": 7.035270829584309e-06, "loss": 0.5987, "step": 3351 }, { "epoch": 1.2, "learning_rate": 7.034371063523484e-06, "loss": 0.6674, "step": 3352 }, { "epoch": 1.2, "learning_rate": 7.0334712974626604e-06, "loss": 0.6818, "step": 3353 }, { "epoch": 1.2, "learning_rate": 7.032571531401836e-06, "loss": 0.601, "step": 3354 }, { "epoch": 1.2, "learning_rate": 7.031671765341012e-06, "loss": 0.6508, "step": 3355 }, { "epoch": 1.2, "learning_rate": 7.030771999280188e-06, "loss": 0.6829, "step": 3356 }, { "epoch": 1.2, "learning_rate": 7.0298722332193635e-06, "loss": 0.6746, "step": 3357 }, { "epoch": 1.21, "learning_rate": 7.028972467158539e-06, "loss": 0.6924, "step": 3358 }, { "epoch": 1.21, "learning_rate": 7.028072701097716e-06, "loss": 0.5614, "step": 3359 }, { "epoch": 1.21, "learning_rate": 7.027172935036891e-06, "loss": 0.6965, "step": 3360 }, { "epoch": 1.21, "learning_rate": 7.026273168976067e-06, "loss": 0.6189, "step": 3361 }, { "epoch": 1.21, "learning_rate": 7.025373402915242e-06, "loss": 0.6169, "step": 3362 }, { "epoch": 1.21, "learning_rate": 7.024473636854419e-06, "loss": 0.5825, "step": 3363 }, { "epoch": 1.21, "learning_rate": 7.023573870793594e-06, "loss": 0.6338, "step": 3364 }, { "epoch": 1.21, "learning_rate": 7.02267410473277e-06, "loss": 0.7252, "step": 3365 }, { "epoch": 1.21, "learning_rate": 7.021774338671946e-06, "loss": 0.5725, "step": 3366 }, { "epoch": 1.21, "learning_rate": 7.020874572611121e-06, "loss": 0.6217, "step": 3367 }, { "epoch": 1.21, "learning_rate": 7.019974806550297e-06, "loss": 0.6827, "step": 3368 }, { "epoch": 1.21, "learning_rate": 7.019075040489474e-06, "loss": 0.7005, "step": 3369 }, { "epoch": 1.21, "learning_rate": 7.018175274428649e-06, "loss": 0.5735, "step": 3370 }, { "epoch": 1.21, "learning_rate": 7.017275508367824e-06, "loss": 0.5944, "step": 3371 }, { "epoch": 1.21, "learning_rate": 7.016375742307001e-06, "loss": 0.6395, "step": 3372 }, { "epoch": 1.21, "learning_rate": 7.015475976246177e-06, "loss": 0.586, "step": 3373 }, { "epoch": 1.21, "learning_rate": 7.0145762101853524e-06, "loss": 0.7602, "step": 3374 }, { "epoch": 1.21, "learning_rate": 7.013676444124529e-06, "loss": 0.5459, "step": 3375 }, { "epoch": 1.21, "learning_rate": 7.012776678063704e-06, "loss": 0.613, "step": 3376 }, { "epoch": 1.21, "learning_rate": 7.01187691200288e-06, "loss": 0.5668, "step": 3377 }, { "epoch": 1.21, "learning_rate": 7.010977145942056e-06, "loss": 0.555, "step": 3378 }, { "epoch": 1.21, "learning_rate": 7.010077379881231e-06, "loss": 0.5856, "step": 3379 }, { "epoch": 1.21, "learning_rate": 7.009177613820407e-06, "loss": 0.6159, "step": 3380 }, { "epoch": 1.21, "learning_rate": 7.008277847759583e-06, "loss": 0.6322, "step": 3381 }, { "epoch": 1.21, "learning_rate": 7.0073780816987594e-06, "loss": 0.62, "step": 3382 }, { "epoch": 1.21, "learning_rate": 7.006478315637934e-06, "loss": 0.6493, "step": 3383 }, { "epoch": 1.21, "learning_rate": 7.00557854957711e-06, "loss": 0.5311, "step": 3384 }, { "epoch": 1.22, "learning_rate": 7.004678783516287e-06, "loss": 0.5456, "step": 3385 }, { "epoch": 1.22, "learning_rate": 7.003779017455462e-06, "loss": 0.6095, "step": 3386 }, { "epoch": 1.22, "learning_rate": 7.0028792513946374e-06, "loss": 0.582, "step": 3387 }, { "epoch": 1.22, "learning_rate": 7.001979485333814e-06, "loss": 0.5522, "step": 3388 }, { "epoch": 1.22, "learning_rate": 7.00107971927299e-06, "loss": 0.6867, "step": 3389 }, { "epoch": 1.22, "learning_rate": 7.000179953212165e-06, "loss": 0.5834, "step": 3390 }, { "epoch": 1.22, "learning_rate": 6.999280187151341e-06, "loss": 0.624, "step": 3391 }, { "epoch": 1.22, "learning_rate": 6.998380421090517e-06, "loss": 0.5466, "step": 3392 }, { "epoch": 1.22, "learning_rate": 6.997480655029693e-06, "loss": 0.5791, "step": 3393 }, { "epoch": 1.22, "learning_rate": 6.9965808889688695e-06, "loss": 0.5897, "step": 3394 }, { "epoch": 1.22, "learning_rate": 6.9956811229080444e-06, "loss": 0.7702, "step": 3395 }, { "epoch": 1.22, "learning_rate": 6.99478135684722e-06, "loss": 0.6122, "step": 3396 }, { "epoch": 1.22, "learning_rate": 6.993881590786397e-06, "loss": 0.5201, "step": 3397 }, { "epoch": 1.22, "learning_rate": 6.992981824725572e-06, "loss": 0.6441, "step": 3398 }, { "epoch": 1.22, "learning_rate": 6.9920820586647475e-06, "loss": 0.6642, "step": 3399 }, { "epoch": 1.22, "learning_rate": 6.991182292603923e-06, "loss": 0.6736, "step": 3400 }, { "epoch": 1.22, "learning_rate": 6.9902825265431e-06, "loss": 0.5708, "step": 3401 }, { "epoch": 1.22, "learning_rate": 6.989382760482275e-06, "loss": 0.5983, "step": 3402 }, { "epoch": 1.22, "learning_rate": 6.988482994421451e-06, "loss": 0.4992, "step": 3403 }, { "epoch": 1.22, "learning_rate": 6.987583228360627e-06, "loss": 0.5417, "step": 3404 }, { "epoch": 1.22, "learning_rate": 6.986683462299803e-06, "loss": 0.6346, "step": 3405 }, { "epoch": 1.22, "learning_rate": 6.985783696238978e-06, "loss": 0.6072, "step": 3406 }, { "epoch": 1.22, "learning_rate": 6.9848839301781545e-06, "loss": 0.5936, "step": 3407 }, { "epoch": 1.22, "learning_rate": 6.98398416411733e-06, "loss": 0.5078, "step": 3408 }, { "epoch": 1.22, "learning_rate": 6.983084398056505e-06, "loss": 0.6587, "step": 3409 }, { "epoch": 1.22, "learning_rate": 6.982184631995682e-06, "loss": 0.5728, "step": 3410 }, { "epoch": 1.22, "learning_rate": 6.981284865934858e-06, "loss": 0.6195, "step": 3411 }, { "epoch": 1.22, "learning_rate": 6.980385099874033e-06, "loss": 0.5459, "step": 3412 }, { "epoch": 1.23, "learning_rate": 6.97948533381321e-06, "loss": 0.6147, "step": 3413 }, { "epoch": 1.23, "learning_rate": 6.978585567752385e-06, "loss": 0.5856, "step": 3414 }, { "epoch": 1.23, "learning_rate": 6.977685801691561e-06, "loss": 0.6294, "step": 3415 }, { "epoch": 1.23, "learning_rate": 6.976786035630737e-06, "loss": 0.5796, "step": 3416 }, { "epoch": 1.23, "learning_rate": 6.975886269569912e-06, "loss": 0.6214, "step": 3417 }, { "epoch": 1.23, "learning_rate": 6.974986503509088e-06, "loss": 0.5854, "step": 3418 }, { "epoch": 1.23, "learning_rate": 6.974086737448264e-06, "loss": 0.5681, "step": 3419 }, { "epoch": 1.23, "learning_rate": 6.97318697138744e-06, "loss": 0.5953, "step": 3420 }, { "epoch": 1.23, "learning_rate": 6.972287205326615e-06, "loss": 0.6183, "step": 3421 }, { "epoch": 1.23, "learning_rate": 6.971387439265791e-06, "loss": 0.5339, "step": 3422 }, { "epoch": 1.23, "learning_rate": 6.970487673204968e-06, "loss": 0.5613, "step": 3423 }, { "epoch": 1.23, "learning_rate": 6.969587907144143e-06, "loss": 0.6653, "step": 3424 }, { "epoch": 1.23, "learning_rate": 6.968688141083318e-06, "loss": 0.6769, "step": 3425 }, { "epoch": 1.23, "learning_rate": 6.967788375022495e-06, "loss": 0.5482, "step": 3426 }, { "epoch": 1.23, "learning_rate": 6.966888608961671e-06, "loss": 0.6428, "step": 3427 }, { "epoch": 1.23, "learning_rate": 6.965988842900846e-06, "loss": 0.6389, "step": 3428 }, { "epoch": 1.23, "learning_rate": 6.965089076840022e-06, "loss": 0.5782, "step": 3429 }, { "epoch": 1.23, "learning_rate": 6.964189310779198e-06, "loss": 0.5199, "step": 3430 }, { "epoch": 1.23, "learning_rate": 6.963289544718374e-06, "loss": 0.6284, "step": 3431 }, { "epoch": 1.23, "learning_rate": 6.96238977865755e-06, "loss": 0.644, "step": 3432 }, { "epoch": 1.23, "learning_rate": 6.961490012596725e-06, "loss": 0.6139, "step": 3433 }, { "epoch": 1.23, "learning_rate": 6.960590246535901e-06, "loss": 0.5697, "step": 3434 }, { "epoch": 1.23, "learning_rate": 6.959690480475078e-06, "loss": 0.59, "step": 3435 }, { "epoch": 1.23, "learning_rate": 6.958790714414253e-06, "loss": 0.5904, "step": 3436 }, { "epoch": 1.23, "learning_rate": 6.957890948353428e-06, "loss": 0.535, "step": 3437 }, { "epoch": 1.23, "learning_rate": 6.956991182292604e-06, "loss": 0.6656, "step": 3438 }, { "epoch": 1.23, "learning_rate": 6.956091416231781e-06, "loss": 0.6254, "step": 3439 }, { "epoch": 1.23, "learning_rate": 6.955191650170956e-06, "loss": 0.6317, "step": 3440 }, { "epoch": 1.24, "learning_rate": 6.9542918841101315e-06, "loss": 0.5306, "step": 3441 }, { "epoch": 1.24, "learning_rate": 6.953392118049308e-06, "loss": 0.6982, "step": 3442 }, { "epoch": 1.24, "learning_rate": 6.952492351988484e-06, "loss": 0.6844, "step": 3443 }, { "epoch": 1.24, "learning_rate": 6.951592585927659e-06, "loss": 0.5835, "step": 3444 }, { "epoch": 1.24, "learning_rate": 6.950692819866835e-06, "loss": 0.5241, "step": 3445 }, { "epoch": 1.24, "learning_rate": 6.949793053806011e-06, "loss": 0.7621, "step": 3446 }, { "epoch": 1.24, "learning_rate": 6.948893287745186e-06, "loss": 0.7422, "step": 3447 }, { "epoch": 1.24, "learning_rate": 6.947993521684363e-06, "loss": 0.6426, "step": 3448 }, { "epoch": 1.24, "learning_rate": 6.9470937556235385e-06, "loss": 0.5598, "step": 3449 }, { "epoch": 1.24, "learning_rate": 6.946193989562714e-06, "loss": 0.5381, "step": 3450 }, { "epoch": 1.24, "learning_rate": 6.945294223501891e-06, "loss": 0.6111, "step": 3451 }, { "epoch": 1.24, "learning_rate": 6.944394457441066e-06, "loss": 0.6534, "step": 3452 }, { "epoch": 1.24, "learning_rate": 6.9434946913802416e-06, "loss": 0.575, "step": 3453 }, { "epoch": 1.24, "learning_rate": 6.942594925319418e-06, "loss": 0.5949, "step": 3454 }, { "epoch": 1.24, "learning_rate": 6.941695159258593e-06, "loss": 0.6414, "step": 3455 }, { "epoch": 1.24, "learning_rate": 6.940795393197769e-06, "loss": 0.5686, "step": 3456 }, { "epoch": 1.24, "learning_rate": 6.939895627136945e-06, "loss": 0.6385, "step": 3457 }, { "epoch": 1.24, "learning_rate": 6.938995861076121e-06, "loss": 0.6275, "step": 3458 }, { "epoch": 1.24, "learning_rate": 6.938096095015296e-06, "loss": 0.6059, "step": 3459 }, { "epoch": 1.24, "learning_rate": 6.937196328954472e-06, "loss": 0.6625, "step": 3460 }, { "epoch": 1.24, "learning_rate": 6.9362965628936486e-06, "loss": 0.5978, "step": 3461 }, { "epoch": 1.24, "learning_rate": 6.935396796832824e-06, "loss": 0.5279, "step": 3462 }, { "epoch": 1.24, "learning_rate": 6.934497030771999e-06, "loss": 0.6022, "step": 3463 }, { "epoch": 1.24, "learning_rate": 6.933597264711176e-06, "loss": 0.6783, "step": 3464 }, { "epoch": 1.24, "learning_rate": 6.932697498650352e-06, "loss": 0.6161, "step": 3465 }, { "epoch": 1.24, "learning_rate": 6.9317977325895266e-06, "loss": 0.5736, "step": 3466 }, { "epoch": 1.24, "learning_rate": 6.930897966528703e-06, "loss": 0.6309, "step": 3467 }, { "epoch": 1.24, "learning_rate": 6.929998200467879e-06, "loss": 0.691, "step": 3468 }, { "epoch": 1.25, "learning_rate": 6.929098434407055e-06, "loss": 0.6666, "step": 3469 }, { "epoch": 1.25, "learning_rate": 6.928198668346231e-06, "loss": 0.6219, "step": 3470 }, { "epoch": 1.25, "learning_rate": 6.927298902285406e-06, "loss": 0.5936, "step": 3471 }, { "epoch": 1.25, "learning_rate": 6.926399136224582e-06, "loss": 0.618, "step": 3472 }, { "epoch": 1.25, "learning_rate": 6.925499370163759e-06, "loss": 0.64, "step": 3473 }, { "epoch": 1.25, "learning_rate": 6.9245996041029336e-06, "loss": 0.6123, "step": 3474 }, { "epoch": 1.25, "learning_rate": 6.923699838042109e-06, "loss": 0.5751, "step": 3475 }, { "epoch": 1.25, "learning_rate": 6.922800071981285e-06, "loss": 0.5995, "step": 3476 }, { "epoch": 1.25, "learning_rate": 6.921900305920462e-06, "loss": 0.6438, "step": 3477 }, { "epoch": 1.25, "learning_rate": 6.921000539859637e-06, "loss": 0.6133, "step": 3478 }, { "epoch": 1.25, "learning_rate": 6.920100773798812e-06, "loss": 0.5988, "step": 3479 }, { "epoch": 1.25, "learning_rate": 6.919201007737989e-06, "loss": 0.6194, "step": 3480 }, { "epoch": 1.25, "learning_rate": 6.918301241677165e-06, "loss": 0.6315, "step": 3481 }, { "epoch": 1.25, "learning_rate": 6.91740147561634e-06, "loss": 0.6609, "step": 3482 }, { "epoch": 1.25, "learning_rate": 6.916501709555516e-06, "loss": 0.6563, "step": 3483 }, { "epoch": 1.25, "learning_rate": 6.915601943494692e-06, "loss": 0.5109, "step": 3484 }, { "epoch": 1.25, "learning_rate": 6.914702177433867e-06, "loss": 0.6067, "step": 3485 }, { "epoch": 1.25, "learning_rate": 6.913802411373044e-06, "loss": 0.6893, "step": 3486 }, { "epoch": 1.25, "learning_rate": 6.912902645312219e-06, "loss": 0.5981, "step": 3487 }, { "epoch": 1.25, "learning_rate": 6.912002879251395e-06, "loss": 0.6192, "step": 3488 }, { "epoch": 1.25, "learning_rate": 6.911103113190572e-06, "loss": 0.6167, "step": 3489 }, { "epoch": 1.25, "learning_rate": 6.910203347129747e-06, "loss": 0.6182, "step": 3490 }, { "epoch": 1.25, "learning_rate": 6.9093035810689225e-06, "loss": 0.6289, "step": 3491 }, { "epoch": 1.25, "learning_rate": 6.908403815008099e-06, "loss": 0.6058, "step": 3492 }, { "epoch": 1.25, "learning_rate": 6.907504048947275e-06, "loss": 0.6581, "step": 3493 }, { "epoch": 1.25, "learning_rate": 6.90660428288645e-06, "loss": 0.6511, "step": 3494 }, { "epoch": 1.25, "learning_rate": 6.9057045168256255e-06, "loss": 0.5523, "step": 3495 }, { "epoch": 1.25, "learning_rate": 6.904804750764802e-06, "loss": 0.6223, "step": 3496 }, { "epoch": 1.26, "learning_rate": 6.903904984703977e-06, "loss": 0.5838, "step": 3497 }, { "epoch": 1.26, "learning_rate": 6.903005218643153e-06, "loss": 0.7074, "step": 3498 }, { "epoch": 1.26, "learning_rate": 6.9021054525823295e-06, "loss": 0.7418, "step": 3499 }, { "epoch": 1.26, "learning_rate": 6.901205686521505e-06, "loss": 0.6446, "step": 3500 }, { "epoch": 1.26, "learning_rate": 6.90030592046068e-06, "loss": 0.5633, "step": 3501 }, { "epoch": 1.26, "learning_rate": 6.899406154399857e-06, "loss": 0.6411, "step": 3502 }, { "epoch": 1.26, "learning_rate": 6.8985063883390325e-06, "loss": 0.6015, "step": 3503 }, { "epoch": 1.26, "learning_rate": 6.8976066222782075e-06, "loss": 0.5935, "step": 3504 }, { "epoch": 1.26, "learning_rate": 6.896706856217384e-06, "loss": 0.6152, "step": 3505 }, { "epoch": 1.26, "learning_rate": 6.89580709015656e-06, "loss": 0.6064, "step": 3506 }, { "epoch": 1.26, "learning_rate": 6.894907324095736e-06, "loss": 0.6723, "step": 3507 }, { "epoch": 1.26, "learning_rate": 6.894007558034912e-06, "loss": 0.6899, "step": 3508 }, { "epoch": 1.26, "learning_rate": 6.893107791974087e-06, "loss": 0.6348, "step": 3509 }, { "epoch": 1.26, "learning_rate": 6.892208025913263e-06, "loss": 0.5969, "step": 3510 }, { "epoch": 1.26, "learning_rate": 6.8913082598524395e-06, "loss": 0.6328, "step": 3511 }, { "epoch": 1.26, "learning_rate": 6.890408493791615e-06, "loss": 0.7074, "step": 3512 }, { "epoch": 1.26, "learning_rate": 6.88950872773079e-06, "loss": 0.5586, "step": 3513 }, { "epoch": 1.26, "learning_rate": 6.888608961669966e-06, "loss": 0.6151, "step": 3514 }, { "epoch": 1.26, "learning_rate": 6.887709195609143e-06, "loss": 0.5999, "step": 3515 }, { "epoch": 1.26, "learning_rate": 6.8868094295483175e-06, "loss": 0.6343, "step": 3516 }, { "epoch": 1.26, "learning_rate": 6.885909663487493e-06, "loss": 0.5825, "step": 3517 }, { "epoch": 1.26, "learning_rate": 6.88500989742667e-06, "loss": 0.5364, "step": 3518 }, { "epoch": 1.26, "learning_rate": 6.884110131365846e-06, "loss": 0.5981, "step": 3519 }, { "epoch": 1.26, "learning_rate": 6.883210365305021e-06, "loss": 0.5746, "step": 3520 }, { "epoch": 1.26, "learning_rate": 6.882310599244197e-06, "loss": 0.7533, "step": 3521 }, { "epoch": 1.26, "learning_rate": 6.881410833183373e-06, "loss": 0.6826, "step": 3522 }, { "epoch": 1.26, "learning_rate": 6.880511067122549e-06, "loss": 0.7224, "step": 3523 }, { "epoch": 1.26, "learning_rate": 6.8796113010617245e-06, "loss": 0.6238, "step": 3524 }, { "epoch": 1.27, "learning_rate": 6.8787115350009e-06, "loss": 0.7023, "step": 3525 }, { "epoch": 1.27, "learning_rate": 6.877811768940076e-06, "loss": 0.7275, "step": 3526 }, { "epoch": 1.27, "learning_rate": 6.876912002879253e-06, "loss": 0.6634, "step": 3527 }, { "epoch": 1.27, "learning_rate": 6.876012236818428e-06, "loss": 0.6624, "step": 3528 }, { "epoch": 1.27, "learning_rate": 6.875112470757603e-06, "loss": 0.5101, "step": 3529 }, { "epoch": 1.27, "learning_rate": 6.87421270469678e-06, "loss": 0.5444, "step": 3530 }, { "epoch": 1.27, "learning_rate": 6.873312938635956e-06, "loss": 0.6347, "step": 3531 }, { "epoch": 1.27, "learning_rate": 6.872413172575131e-06, "loss": 0.6785, "step": 3532 }, { "epoch": 1.27, "learning_rate": 6.8715134065143065e-06, "loss": 0.5674, "step": 3533 }, { "epoch": 1.27, "learning_rate": 6.870613640453483e-06, "loss": 0.6635, "step": 3534 }, { "epoch": 1.27, "learning_rate": 6.869713874392658e-06, "loss": 0.4412, "step": 3535 }, { "epoch": 1.27, "learning_rate": 6.868814108331834e-06, "loss": 0.6922, "step": 3536 }, { "epoch": 1.27, "learning_rate": 6.86791434227101e-06, "loss": 0.6376, "step": 3537 }, { "epoch": 1.27, "learning_rate": 6.867014576210186e-06, "loss": 0.6195, "step": 3538 }, { "epoch": 1.27, "learning_rate": 6.866114810149361e-06, "loss": 0.6005, "step": 3539 }, { "epoch": 1.27, "learning_rate": 6.865215044088538e-06, "loss": 0.5364, "step": 3540 }, { "epoch": 1.27, "learning_rate": 6.8643152780277135e-06, "loss": 0.6968, "step": 3541 }, { "epoch": 1.27, "learning_rate": 6.863415511966889e-06, "loss": 0.6068, "step": 3542 }, { "epoch": 1.27, "learning_rate": 6.862515745906065e-06, "loss": 0.5051, "step": 3543 }, { "epoch": 1.27, "learning_rate": 6.861615979845241e-06, "loss": 0.638, "step": 3544 }, { "epoch": 1.27, "learning_rate": 6.8607162137844165e-06, "loss": 0.6115, "step": 3545 }, { "epoch": 1.27, "learning_rate": 6.859816447723593e-06, "loss": 0.6625, "step": 3546 }, { "epoch": 1.27, "learning_rate": 6.858916681662768e-06, "loss": 0.6702, "step": 3547 }, { "epoch": 1.27, "learning_rate": 6.858016915601944e-06, "loss": 0.676, "step": 3548 }, { "epoch": 1.27, "learning_rate": 6.8571171495411204e-06, "loss": 0.6245, "step": 3549 }, { "epoch": 1.27, "learning_rate": 6.856217383480296e-06, "loss": 0.6756, "step": 3550 }, { "epoch": 1.27, "learning_rate": 6.855317617419471e-06, "loss": 0.5525, "step": 3551 }, { "epoch": 1.27, "learning_rate": 6.854417851358647e-06, "loss": 0.6095, "step": 3552 }, { "epoch": 1.28, "learning_rate": 6.8535180852978235e-06, "loss": 0.6044, "step": 3553 }, { "epoch": 1.28, "learning_rate": 6.8526183192369984e-06, "loss": 0.5578, "step": 3554 }, { "epoch": 1.28, "learning_rate": 6.851718553176174e-06, "loss": 0.5934, "step": 3555 }, { "epoch": 1.28, "learning_rate": 6.850818787115351e-06, "loss": 0.6308, "step": 3556 }, { "epoch": 1.28, "learning_rate": 6.849919021054527e-06, "loss": 0.608, "step": 3557 }, { "epoch": 1.28, "learning_rate": 6.8490192549937015e-06, "loss": 0.6102, "step": 3558 }, { "epoch": 1.28, "learning_rate": 6.848119488932878e-06, "loss": 0.6218, "step": 3559 }, { "epoch": 1.28, "learning_rate": 6.847219722872054e-06, "loss": 0.6655, "step": 3560 }, { "epoch": 1.28, "learning_rate": 6.84631995681123e-06, "loss": 0.6277, "step": 3561 }, { "epoch": 1.28, "learning_rate": 6.845420190750406e-06, "loss": 0.5866, "step": 3562 }, { "epoch": 1.28, "learning_rate": 6.844520424689581e-06, "loss": 0.6647, "step": 3563 }, { "epoch": 1.28, "learning_rate": 6.843620658628757e-06, "loss": 0.6001, "step": 3564 }, { "epoch": 1.28, "learning_rate": 6.842720892567934e-06, "loss": 0.6942, "step": 3565 }, { "epoch": 1.28, "learning_rate": 6.8418211265071085e-06, "loss": 0.6366, "step": 3566 }, { "epoch": 1.28, "learning_rate": 6.840921360446284e-06, "loss": 0.6352, "step": 3567 }, { "epoch": 1.28, "learning_rate": 6.840021594385461e-06, "loss": 0.6064, "step": 3568 }, { "epoch": 1.28, "learning_rate": 6.839121828324637e-06, "loss": 0.6021, "step": 3569 }, { "epoch": 1.28, "learning_rate": 6.838222062263812e-06, "loss": 0.5299, "step": 3570 }, { "epoch": 1.28, "learning_rate": 6.837322296202987e-06, "loss": 0.5905, "step": 3571 }, { "epoch": 1.28, "learning_rate": 6.836422530142164e-06, "loss": 0.6461, "step": 3572 }, { "epoch": 1.28, "learning_rate": 6.835522764081339e-06, "loss": 0.635, "step": 3573 }, { "epoch": 1.28, "learning_rate": 6.834622998020515e-06, "loss": 0.5702, "step": 3574 }, { "epoch": 1.28, "learning_rate": 6.833723231959691e-06, "loss": 0.6259, "step": 3575 }, { "epoch": 1.28, "learning_rate": 6.832823465898867e-06, "loss": 0.6149, "step": 3576 }, { "epoch": 1.28, "learning_rate": 6.831923699838042e-06, "loss": 0.6264, "step": 3577 }, { "epoch": 1.28, "learning_rate": 6.831023933777219e-06, "loss": 0.5771, "step": 3578 }, { "epoch": 1.28, "learning_rate": 6.830124167716394e-06, "loss": 0.6169, "step": 3579 }, { "epoch": 1.28, "learning_rate": 6.82922440165557e-06, "loss": 0.6697, "step": 3580 }, { "epoch": 1.29, "learning_rate": 6.828324635594747e-06, "loss": 0.6674, "step": 3581 }, { "epoch": 1.29, "learning_rate": 6.827424869533922e-06, "loss": 0.7284, "step": 3582 }, { "epoch": 1.29, "learning_rate": 6.8265251034730974e-06, "loss": 0.6392, "step": 3583 }, { "epoch": 1.29, "learning_rate": 6.825625337412274e-06, "loss": 0.6681, "step": 3584 }, { "epoch": 1.29, "learning_rate": 6.824725571351449e-06, "loss": 0.7133, "step": 3585 }, { "epoch": 1.29, "learning_rate": 6.823825805290625e-06, "loss": 0.6046, "step": 3586 }, { "epoch": 1.29, "learning_rate": 6.822926039229801e-06, "loss": 0.6872, "step": 3587 }, { "epoch": 1.29, "learning_rate": 6.822026273168977e-06, "loss": 0.5916, "step": 3588 }, { "epoch": 1.29, "learning_rate": 6.821126507108152e-06, "loss": 0.6123, "step": 3589 }, { "epoch": 1.29, "learning_rate": 6.820226741047328e-06, "loss": 0.5615, "step": 3590 }, { "epoch": 1.29, "learning_rate": 6.8193269749865044e-06, "loss": 0.5659, "step": 3591 }, { "epoch": 1.29, "learning_rate": 6.818427208925679e-06, "loss": 0.6069, "step": 3592 }, { "epoch": 1.29, "learning_rate": 6.817527442864855e-06, "loss": 0.5856, "step": 3593 }, { "epoch": 1.29, "learning_rate": 6.816627676804032e-06, "loss": 0.6416, "step": 3594 }, { "epoch": 1.29, "learning_rate": 6.8157279107432075e-06, "loss": 0.6444, "step": 3595 }, { "epoch": 1.29, "learning_rate": 6.8148281446823824e-06, "loss": 0.6453, "step": 3596 }, { "epoch": 1.29, "learning_rate": 6.813928378621559e-06, "loss": 0.7319, "step": 3597 }, { "epoch": 1.29, "learning_rate": 6.813028612560735e-06, "loss": 0.5718, "step": 3598 }, { "epoch": 1.29, "learning_rate": 6.812128846499911e-06, "loss": 0.6679, "step": 3599 }, { "epoch": 1.29, "learning_rate": 6.811229080439087e-06, "loss": 0.6217, "step": 3600 }, { "epoch": 1.29, "learning_rate": 6.810329314378262e-06, "loss": 0.594, "step": 3601 }, { "epoch": 1.29, "learning_rate": 6.809429548317438e-06, "loss": 0.6685, "step": 3602 }, { "epoch": 1.29, "learning_rate": 6.8085297822566145e-06, "loss": 0.6737, "step": 3603 }, { "epoch": 1.29, "learning_rate": 6.8076300161957894e-06, "loss": 0.5897, "step": 3604 }, { "epoch": 1.29, "learning_rate": 6.806730250134965e-06, "loss": 0.6015, "step": 3605 }, { "epoch": 1.29, "learning_rate": 6.805830484074142e-06, "loss": 0.6263, "step": 3606 }, { "epoch": 1.29, "learning_rate": 6.804930718013318e-06, "loss": 0.5067, "step": 3607 }, { "epoch": 1.3, "learning_rate": 6.8040309519524925e-06, "loss": 0.6518, "step": 3608 }, { "epoch": 1.3, "learning_rate": 6.803131185891668e-06, "loss": 0.5891, "step": 3609 }, { "epoch": 1.3, "learning_rate": 6.802231419830845e-06, "loss": 0.5911, "step": 3610 }, { "epoch": 1.3, "learning_rate": 6.801331653770021e-06, "loss": 0.6613, "step": 3611 }, { "epoch": 1.3, "learning_rate": 6.800431887709196e-06, "loss": 0.579, "step": 3612 }, { "epoch": 1.3, "learning_rate": 6.799532121648372e-06, "loss": 0.65, "step": 3613 }, { "epoch": 1.3, "learning_rate": 6.798632355587548e-06, "loss": 0.6242, "step": 3614 }, { "epoch": 1.3, "learning_rate": 6.797732589526723e-06, "loss": 0.5668, "step": 3615 }, { "epoch": 1.3, "learning_rate": 6.7968328234658995e-06, "loss": 0.6523, "step": 3616 }, { "epoch": 1.3, "learning_rate": 6.795933057405075e-06, "loss": 0.6159, "step": 3617 }, { "epoch": 1.3, "learning_rate": 6.795033291344251e-06, "loss": 0.6671, "step": 3618 }, { "epoch": 1.3, "learning_rate": 6.794133525283428e-06, "loss": 0.6774, "step": 3619 }, { "epoch": 1.3, "learning_rate": 6.7932337592226026e-06, "loss": 0.527, "step": 3620 }, { "epoch": 1.3, "learning_rate": 6.792333993161778e-06, "loss": 0.5942, "step": 3621 }, { "epoch": 1.3, "learning_rate": 6.791434227100955e-06, "loss": 0.6548, "step": 3622 }, { "epoch": 1.3, "learning_rate": 6.79053446104013e-06, "loss": 0.6391, "step": 3623 }, { "epoch": 1.3, "learning_rate": 6.789634694979306e-06, "loss": 0.5827, "step": 3624 }, { "epoch": 1.3, "learning_rate": 6.788734928918482e-06, "loss": 0.6269, "step": 3625 }, { "epoch": 1.3, "learning_rate": 6.787835162857658e-06, "loss": 0.5825, "step": 3626 }, { "epoch": 1.3, "learning_rate": 6.786935396796833e-06, "loss": 0.5948, "step": 3627 }, { "epoch": 1.3, "learning_rate": 6.786035630736009e-06, "loss": 0.5706, "step": 3628 }, { "epoch": 1.3, "learning_rate": 6.785135864675185e-06, "loss": 0.5555, "step": 3629 }, { "epoch": 1.3, "learning_rate": 6.784236098614361e-06, "loss": 0.6101, "step": 3630 }, { "epoch": 1.3, "learning_rate": 6.783336332553536e-06, "loss": 0.6336, "step": 3631 }, { "epoch": 1.3, "learning_rate": 6.782436566492713e-06, "loss": 0.6653, "step": 3632 }, { "epoch": 1.3, "learning_rate": 6.781536800431888e-06, "loss": 0.6922, "step": 3633 }, { "epoch": 1.3, "learning_rate": 6.780637034371063e-06, "loss": 0.6329, "step": 3634 }, { "epoch": 1.3, "learning_rate": 6.77973726831024e-06, "loss": 0.5953, "step": 3635 }, { "epoch": 1.31, "learning_rate": 6.778837502249416e-06, "loss": 0.6932, "step": 3636 }, { "epoch": 1.31, "learning_rate": 6.7779377361885915e-06, "loss": 0.5961, "step": 3637 }, { "epoch": 1.31, "learning_rate": 6.777037970127768e-06, "loss": 0.7506, "step": 3638 }, { "epoch": 1.31, "learning_rate": 6.776138204066943e-06, "loss": 0.6436, "step": 3639 }, { "epoch": 1.31, "learning_rate": 6.775238438006119e-06, "loss": 0.5353, "step": 3640 }, { "epoch": 1.31, "learning_rate": 6.774338671945295e-06, "loss": 0.5601, "step": 3641 }, { "epoch": 1.31, "learning_rate": 6.77343890588447e-06, "loss": 0.6392, "step": 3642 }, { "epoch": 1.31, "learning_rate": 6.772539139823646e-06, "loss": 0.6231, "step": 3643 }, { "epoch": 1.31, "learning_rate": 6.771639373762823e-06, "loss": 0.5808, "step": 3644 }, { "epoch": 1.31, "learning_rate": 6.7707396077019985e-06, "loss": 0.7409, "step": 3645 }, { "epoch": 1.31, "learning_rate": 6.769839841641173e-06, "loss": 0.5756, "step": 3646 }, { "epoch": 1.31, "learning_rate": 6.768940075580349e-06, "loss": 0.5713, "step": 3647 }, { "epoch": 1.31, "learning_rate": 6.768040309519526e-06, "loss": 0.5908, "step": 3648 }, { "epoch": 1.31, "learning_rate": 6.7671405434587016e-06, "loss": 0.628, "step": 3649 }, { "epoch": 1.31, "learning_rate": 6.7662407773978765e-06, "loss": 0.6654, "step": 3650 }, { "epoch": 1.31, "learning_rate": 6.765341011337053e-06, "loss": 0.6035, "step": 3651 }, { "epoch": 1.31, "learning_rate": 6.764441245276229e-06, "loss": 0.6572, "step": 3652 }, { "epoch": 1.31, "learning_rate": 6.763541479215404e-06, "loss": 0.5985, "step": 3653 }, { "epoch": 1.31, "learning_rate": 6.76264171315458e-06, "loss": 0.6283, "step": 3654 }, { "epoch": 1.31, "learning_rate": 6.761741947093756e-06, "loss": 0.6987, "step": 3655 }, { "epoch": 1.31, "learning_rate": 6.760842181032932e-06, "loss": 0.6351, "step": 3656 }, { "epoch": 1.31, "learning_rate": 6.7599424149721086e-06, "loss": 0.6159, "step": 3657 }, { "epoch": 1.31, "learning_rate": 6.7590426489112835e-06, "loss": 0.5528, "step": 3658 }, { "epoch": 1.31, "learning_rate": 6.758142882850459e-06, "loss": 0.6209, "step": 3659 }, { "epoch": 1.31, "learning_rate": 6.757243116789636e-06, "loss": 0.6448, "step": 3660 }, { "epoch": 1.31, "learning_rate": 6.756343350728811e-06, "loss": 0.5709, "step": 3661 }, { "epoch": 1.31, "learning_rate": 6.7554435846679866e-06, "loss": 0.6239, "step": 3662 }, { "epoch": 1.31, "learning_rate": 6.754543818607163e-06, "loss": 0.5926, "step": 3663 }, { "epoch": 1.32, "learning_rate": 6.753644052546339e-06, "loss": 0.557, "step": 3664 }, { "epoch": 1.32, "learning_rate": 6.752744286485514e-06, "loss": 0.6315, "step": 3665 }, { "epoch": 1.32, "learning_rate": 6.75184452042469e-06, "loss": 0.6139, "step": 3666 }, { "epoch": 1.32, "learning_rate": 6.750944754363866e-06, "loss": 0.5923, "step": 3667 }, { "epoch": 1.32, "learning_rate": 6.750044988303042e-06, "loss": 0.6218, "step": 3668 }, { "epoch": 1.32, "learning_rate": 6.749145222242217e-06, "loss": 0.5816, "step": 3669 }, { "epoch": 1.32, "learning_rate": 6.7482454561813936e-06, "loss": 0.6783, "step": 3670 }, { "epoch": 1.32, "learning_rate": 6.747345690120569e-06, "loss": 0.6193, "step": 3671 }, { "epoch": 1.32, "learning_rate": 6.746445924059744e-06, "loss": 0.6213, "step": 3672 }, { "epoch": 1.32, "learning_rate": 6.745546157998921e-06, "loss": 0.6342, "step": 3673 }, { "epoch": 1.32, "learning_rate": 6.744646391938097e-06, "loss": 0.5809, "step": 3674 }, { "epoch": 1.32, "learning_rate": 6.743746625877272e-06, "loss": 0.6235, "step": 3675 }, { "epoch": 1.32, "learning_rate": 6.742846859816449e-06, "loss": 0.627, "step": 3676 }, { "epoch": 1.32, "learning_rate": 6.741947093755624e-06, "loss": 0.6032, "step": 3677 }, { "epoch": 1.32, "learning_rate": 6.7410473276948e-06, "loss": 0.5804, "step": 3678 }, { "epoch": 1.32, "learning_rate": 6.740147561633976e-06, "loss": 0.6464, "step": 3679 }, { "epoch": 1.32, "learning_rate": 6.739247795573152e-06, "loss": 0.635, "step": 3680 }, { "epoch": 1.32, "learning_rate": 6.738348029512327e-06, "loss": 0.6371, "step": 3681 }, { "epoch": 1.32, "learning_rate": 6.737448263451504e-06, "loss": 0.5691, "step": 3682 }, { "epoch": 1.32, "learning_rate": 6.736548497390679e-06, "loss": 0.5642, "step": 3683 }, { "epoch": 1.32, "learning_rate": 6.735648731329854e-06, "loss": 0.6132, "step": 3684 }, { "epoch": 1.32, "learning_rate": 6.73474896526903e-06, "loss": 0.5833, "step": 3685 }, { "epoch": 1.32, "learning_rate": 6.733849199208207e-06, "loss": 0.7002, "step": 3686 }, { "epoch": 1.32, "learning_rate": 6.7329494331473825e-06, "loss": 0.6423, "step": 3687 }, { "epoch": 1.32, "learning_rate": 6.732049667086557e-06, "loss": 0.6201, "step": 3688 }, { "epoch": 1.32, "learning_rate": 6.731149901025734e-06, "loss": 0.6186, "step": 3689 }, { "epoch": 1.32, "learning_rate": 6.73025013496491e-06, "loss": 0.6637, "step": 3690 }, { "epoch": 1.32, "learning_rate": 6.729350368904085e-06, "loss": 0.5916, "step": 3691 }, { "epoch": 1.33, "learning_rate": 6.728450602843261e-06, "loss": 0.6499, "step": 3692 }, { "epoch": 1.33, "learning_rate": 6.727550836782437e-06, "loss": 0.5844, "step": 3693 }, { "epoch": 1.33, "learning_rate": 6.726651070721613e-06, "loss": 0.6263, "step": 3694 }, { "epoch": 1.33, "learning_rate": 6.7257513046607895e-06, "loss": 0.5942, "step": 3695 }, { "epoch": 1.33, "learning_rate": 6.724851538599964e-06, "loss": 0.5896, "step": 3696 }, { "epoch": 1.33, "learning_rate": 6.72395177253914e-06, "loss": 0.7091, "step": 3697 }, { "epoch": 1.33, "learning_rate": 6.723052006478317e-06, "loss": 0.62, "step": 3698 }, { "epoch": 1.33, "learning_rate": 6.7221522404174925e-06, "loss": 0.7248, "step": 3699 }, { "epoch": 1.33, "learning_rate": 6.7212524743566675e-06, "loss": 0.6193, "step": 3700 }, { "epoch": 1.33, "learning_rate": 6.720352708295844e-06, "loss": 0.5943, "step": 3701 }, { "epoch": 1.33, "learning_rate": 6.71945294223502e-06, "loss": 0.6051, "step": 3702 }, { "epoch": 1.33, "learning_rate": 6.718553176174195e-06, "loss": 0.652, "step": 3703 }, { "epoch": 1.33, "learning_rate": 6.7176534101133705e-06, "loss": 0.628, "step": 3704 }, { "epoch": 1.33, "learning_rate": 6.716753644052547e-06, "loss": 0.5586, "step": 3705 }, { "epoch": 1.33, "learning_rate": 6.715853877991723e-06, "loss": 0.6415, "step": 3706 }, { "epoch": 1.33, "learning_rate": 6.714954111930898e-06, "loss": 0.6447, "step": 3707 }, { "epoch": 1.33, "learning_rate": 6.7140543458700745e-06, "loss": 0.6048, "step": 3708 }, { "epoch": 1.33, "learning_rate": 6.71315457980925e-06, "loss": 0.7365, "step": 3709 }, { "epoch": 1.33, "learning_rate": 6.712254813748426e-06, "loss": 0.6044, "step": 3710 }, { "epoch": 1.33, "learning_rate": 6.711355047687602e-06, "loss": 0.6114, "step": 3711 }, { "epoch": 1.33, "learning_rate": 6.7104552816267775e-06, "loss": 0.735, "step": 3712 }, { "epoch": 1.33, "learning_rate": 6.709555515565953e-06, "loss": 0.5882, "step": 3713 }, { "epoch": 1.33, "learning_rate": 6.70865574950513e-06, "loss": 0.6962, "step": 3714 }, { "epoch": 1.33, "learning_rate": 6.707755983444305e-06, "loss": 0.5702, "step": 3715 }, { "epoch": 1.33, "learning_rate": 6.706856217383481e-06, "loss": 0.6046, "step": 3716 }, { "epoch": 1.33, "learning_rate": 6.705956451322657e-06, "loss": 0.5757, "step": 3717 }, { "epoch": 1.33, "learning_rate": 6.705056685261833e-06, "loss": 0.6209, "step": 3718 }, { "epoch": 1.33, "learning_rate": 6.704156919201008e-06, "loss": 0.6053, "step": 3719 }, { "epoch": 1.34, "learning_rate": 6.703257153140184e-06, "loss": 0.6705, "step": 3720 }, { "epoch": 1.34, "learning_rate": 6.70235738707936e-06, "loss": 0.6317, "step": 3721 }, { "epoch": 1.34, "learning_rate": 6.701457621018535e-06, "loss": 0.5635, "step": 3722 }, { "epoch": 1.34, "learning_rate": 6.700557854957711e-06, "loss": 0.594, "step": 3723 }, { "epoch": 1.34, "learning_rate": 6.699658088896888e-06, "loss": 0.5848, "step": 3724 }, { "epoch": 1.34, "learning_rate": 6.698758322836063e-06, "loss": 0.6855, "step": 3725 }, { "epoch": 1.34, "learning_rate": 6.697858556775238e-06, "loss": 0.51, "step": 3726 }, { "epoch": 1.34, "learning_rate": 6.696958790714415e-06, "loss": 0.6261, "step": 3727 }, { "epoch": 1.34, "learning_rate": 6.696059024653591e-06, "loss": 0.6143, "step": 3728 }, { "epoch": 1.34, "learning_rate": 6.6951592585927665e-06, "loss": 0.6302, "step": 3729 }, { "epoch": 1.34, "learning_rate": 6.694259492531942e-06, "loss": 0.6292, "step": 3730 }, { "epoch": 1.34, "learning_rate": 6.693359726471118e-06, "loss": 0.6257, "step": 3731 }, { "epoch": 1.34, "learning_rate": 6.692459960410294e-06, "loss": 0.6578, "step": 3732 }, { "epoch": 1.34, "learning_rate": 6.69156019434947e-06, "loss": 0.6459, "step": 3733 }, { "epoch": 1.34, "learning_rate": 6.690660428288645e-06, "loss": 0.5869, "step": 3734 }, { "epoch": 1.34, "learning_rate": 6.689760662227821e-06, "loss": 0.618, "step": 3735 }, { "epoch": 1.34, "learning_rate": 6.688860896166998e-06, "loss": 0.7711, "step": 3736 }, { "epoch": 1.34, "learning_rate": 6.6879611301061735e-06, "loss": 0.5177, "step": 3737 }, { "epoch": 1.34, "learning_rate": 6.687061364045348e-06, "loss": 0.5712, "step": 3738 }, { "epoch": 1.34, "learning_rate": 6.686161597984524e-06, "loss": 0.5688, "step": 3739 }, { "epoch": 1.34, "learning_rate": 6.685261831923701e-06, "loss": 0.5453, "step": 3740 }, { "epoch": 1.34, "learning_rate": 6.684362065862876e-06, "loss": 0.6174, "step": 3741 }, { "epoch": 1.34, "learning_rate": 6.6834622998020514e-06, "loss": 0.578, "step": 3742 }, { "epoch": 1.34, "learning_rate": 6.682562533741228e-06, "loss": 0.7007, "step": 3743 }, { "epoch": 1.34, "learning_rate": 6.681662767680404e-06, "loss": 0.6533, "step": 3744 }, { "epoch": 1.34, "learning_rate": 6.680763001619579e-06, "loss": 0.6553, "step": 3745 }, { "epoch": 1.34, "learning_rate": 6.679863235558755e-06, "loss": 0.6454, "step": 3746 }, { "epoch": 1.34, "learning_rate": 6.678963469497931e-06, "loss": 0.5745, "step": 3747 }, { "epoch": 1.35, "learning_rate": 6.678063703437107e-06, "loss": 0.5732, "step": 3748 }, { "epoch": 1.35, "learning_rate": 6.677163937376283e-06, "loss": 0.6379, "step": 3749 }, { "epoch": 1.35, "learning_rate": 6.6762641713154584e-06, "loss": 0.6269, "step": 3750 }, { "epoch": 1.35, "learning_rate": 6.675364405254634e-06, "loss": 0.575, "step": 3751 }, { "epoch": 1.35, "learning_rate": 6.674464639193811e-06, "loss": 0.5698, "step": 3752 }, { "epoch": 1.35, "learning_rate": 6.673564873132986e-06, "loss": 0.6323, "step": 3753 }, { "epoch": 1.35, "learning_rate": 6.6726651070721615e-06, "loss": 0.716, "step": 3754 }, { "epoch": 1.35, "learning_rate": 6.671765341011338e-06, "loss": 0.6029, "step": 3755 }, { "epoch": 1.35, "learning_rate": 6.670865574950514e-06, "loss": 0.7214, "step": 3756 }, { "epoch": 1.35, "learning_rate": 6.669965808889689e-06, "loss": 0.6849, "step": 3757 }, { "epoch": 1.35, "learning_rate": 6.669066042828865e-06, "loss": 0.6304, "step": 3758 }, { "epoch": 1.35, "learning_rate": 6.668166276768041e-06, "loss": 0.5723, "step": 3759 }, { "epoch": 1.35, "learning_rate": 6.667266510707216e-06, "loss": 0.6156, "step": 3760 }, { "epoch": 1.35, "learning_rate": 6.666366744646392e-06, "loss": 0.676, "step": 3761 }, { "epoch": 1.35, "learning_rate": 6.6654669785855685e-06, "loss": 0.5781, "step": 3762 }, { "epoch": 1.35, "learning_rate": 6.664567212524744e-06, "loss": 0.6884, "step": 3763 }, { "epoch": 1.35, "learning_rate": 6.663667446463919e-06, "loss": 0.6543, "step": 3764 }, { "epoch": 1.35, "learning_rate": 6.662767680403096e-06, "loss": 0.616, "step": 3765 }, { "epoch": 1.35, "learning_rate": 6.661867914342272e-06, "loss": 0.6409, "step": 3766 }, { "epoch": 1.35, "learning_rate": 6.660968148281447e-06, "loss": 0.5691, "step": 3767 }, { "epoch": 1.35, "learning_rate": 6.660068382220624e-06, "loss": 0.6, "step": 3768 }, { "epoch": 1.35, "learning_rate": 6.659168616159799e-06, "loss": 0.4968, "step": 3769 }, { "epoch": 1.35, "learning_rate": 6.658268850098975e-06, "loss": 0.5433, "step": 3770 }, { "epoch": 1.35, "learning_rate": 6.657369084038151e-06, "loss": 0.5474, "step": 3771 }, { "epoch": 1.35, "learning_rate": 6.656469317977326e-06, "loss": 0.5612, "step": 3772 }, { "epoch": 1.35, "learning_rate": 6.655569551916502e-06, "loss": 0.6637, "step": 3773 }, { "epoch": 1.35, "learning_rate": 6.654669785855679e-06, "loss": 0.6243, "step": 3774 }, { "epoch": 1.35, "learning_rate": 6.653770019794854e-06, "loss": 0.5539, "step": 3775 }, { "epoch": 1.36, "learning_rate": 6.652870253734029e-06, "loss": 0.5665, "step": 3776 }, { "epoch": 1.36, "learning_rate": 6.651970487673205e-06, "loss": 0.636, "step": 3777 }, { "epoch": 1.36, "learning_rate": 6.651070721612382e-06, "loss": 0.5755, "step": 3778 }, { "epoch": 1.36, "learning_rate": 6.650170955551557e-06, "loss": 0.5162, "step": 3779 }, { "epoch": 1.36, "learning_rate": 6.649271189490732e-06, "loss": 0.5621, "step": 3780 }, { "epoch": 1.36, "learning_rate": 6.648371423429909e-06, "loss": 0.6177, "step": 3781 }, { "epoch": 1.36, "learning_rate": 6.647471657369085e-06, "loss": 0.7527, "step": 3782 }, { "epoch": 1.36, "learning_rate": 6.64657189130826e-06, "loss": 0.6624, "step": 3783 }, { "epoch": 1.36, "learning_rate": 6.645672125247436e-06, "loss": 0.6398, "step": 3784 }, { "epoch": 1.36, "learning_rate": 6.644772359186612e-06, "loss": 0.6232, "step": 3785 }, { "epoch": 1.36, "learning_rate": 6.643872593125788e-06, "loss": 0.6364, "step": 3786 }, { "epoch": 1.36, "learning_rate": 6.6429728270649644e-06, "loss": 0.5247, "step": 3787 }, { "epoch": 1.36, "learning_rate": 6.642073061004139e-06, "loss": 0.6017, "step": 3788 }, { "epoch": 1.36, "learning_rate": 6.641173294943315e-06, "loss": 0.575, "step": 3789 }, { "epoch": 1.36, "learning_rate": 6.640273528882492e-06, "loss": 0.5484, "step": 3790 }, { "epoch": 1.36, "learning_rate": 6.639373762821667e-06, "loss": 0.5851, "step": 3791 }, { "epoch": 1.36, "learning_rate": 6.6384739967608424e-06, "loss": 0.6108, "step": 3792 }, { "epoch": 1.36, "learning_rate": 6.637574230700019e-06, "loss": 0.6057, "step": 3793 }, { "epoch": 1.36, "learning_rate": 6.636674464639195e-06, "loss": 0.5558, "step": 3794 }, { "epoch": 1.36, "learning_rate": 6.63577469857837e-06, "loss": 0.5648, "step": 3795 }, { "epoch": 1.36, "learning_rate": 6.6348749325175455e-06, "loss": 0.5949, "step": 3796 }, { "epoch": 1.36, "learning_rate": 6.633975166456722e-06, "loss": 0.6334, "step": 3797 }, { "epoch": 1.36, "learning_rate": 6.633075400395898e-06, "loss": 0.5576, "step": 3798 }, { "epoch": 1.36, "learning_rate": 6.632175634335073e-06, "loss": 0.6292, "step": 3799 }, { "epoch": 1.36, "learning_rate": 6.6312758682742494e-06, "loss": 0.6569, "step": 3800 }, { "epoch": 1.36, "learning_rate": 6.630376102213425e-06, "loss": 0.5584, "step": 3801 }, { "epoch": 1.36, "learning_rate": 6.6294763361526e-06, "loss": 0.6301, "step": 3802 }, { "epoch": 1.37, "learning_rate": 6.628576570091777e-06, "loss": 0.6109, "step": 3803 }, { "epoch": 1.37, "learning_rate": 6.6276768040309525e-06, "loss": 0.6122, "step": 3804 }, { "epoch": 1.37, "learning_rate": 6.626777037970128e-06, "loss": 0.5747, "step": 3805 }, { "epoch": 1.37, "learning_rate": 6.625877271909305e-06, "loss": 0.5599, "step": 3806 }, { "epoch": 1.37, "learning_rate": 6.62497750584848e-06, "loss": 0.489, "step": 3807 }, { "epoch": 1.37, "learning_rate": 6.624077739787656e-06, "loss": 0.6283, "step": 3808 }, { "epoch": 1.37, "learning_rate": 6.623177973726832e-06, "loss": 0.731, "step": 3809 }, { "epoch": 1.37, "learning_rate": 6.622278207666007e-06, "loss": 0.6773, "step": 3810 }, { "epoch": 1.37, "learning_rate": 6.621378441605183e-06, "loss": 0.6207, "step": 3811 }, { "epoch": 1.37, "learning_rate": 6.6204786755443595e-06, "loss": 0.5808, "step": 3812 }, { "epoch": 1.37, "learning_rate": 6.619578909483535e-06, "loss": 0.7392, "step": 3813 }, { "epoch": 1.37, "learning_rate": 6.61867914342271e-06, "loss": 0.577, "step": 3814 }, { "epoch": 1.37, "learning_rate": 6.617779377361886e-06, "loss": 0.8017, "step": 3815 }, { "epoch": 1.37, "learning_rate": 6.6168796113010626e-06, "loss": 0.4886, "step": 3816 }, { "epoch": 1.37, "learning_rate": 6.615979845240238e-06, "loss": 0.7311, "step": 3817 }, { "epoch": 1.37, "learning_rate": 6.615080079179413e-06, "loss": 0.6139, "step": 3818 }, { "epoch": 1.37, "learning_rate": 6.61418031311859e-06, "loss": 0.7221, "step": 3819 }, { "epoch": 1.37, "learning_rate": 6.613280547057766e-06, "loss": 0.6024, "step": 3820 }, { "epoch": 1.37, "learning_rate": 6.6123807809969406e-06, "loss": 0.5501, "step": 3821 }, { "epoch": 1.37, "learning_rate": 6.611481014936117e-06, "loss": 0.6629, "step": 3822 }, { "epoch": 1.37, "learning_rate": 6.610581248875293e-06, "loss": 0.592, "step": 3823 }, { "epoch": 1.37, "learning_rate": 6.609681482814469e-06, "loss": 0.7781, "step": 3824 }, { "epoch": 1.37, "learning_rate": 6.608781716753645e-06, "loss": 0.5702, "step": 3825 }, { "epoch": 1.37, "learning_rate": 6.60788195069282e-06, "loss": 0.6257, "step": 3826 }, { "epoch": 1.37, "learning_rate": 6.606982184631996e-06, "loss": 0.6362, "step": 3827 }, { "epoch": 1.37, "learning_rate": 6.606082418571173e-06, "loss": 0.6555, "step": 3828 }, { "epoch": 1.37, "learning_rate": 6.6051826525103476e-06, "loss": 0.757, "step": 3829 }, { "epoch": 1.37, "learning_rate": 6.604282886449523e-06, "loss": 0.5849, "step": 3830 }, { "epoch": 1.38, "learning_rate": 6.6033831203887e-06, "loss": 0.5982, "step": 3831 }, { "epoch": 1.38, "learning_rate": 6.602483354327876e-06, "loss": 0.6211, "step": 3832 }, { "epoch": 1.38, "learning_rate": 6.601583588267051e-06, "loss": 0.5974, "step": 3833 }, { "epoch": 1.38, "learning_rate": 6.600683822206226e-06, "loss": 0.5987, "step": 3834 }, { "epoch": 1.38, "learning_rate": 6.599784056145403e-06, "loss": 0.6076, "step": 3835 }, { "epoch": 1.38, "learning_rate": 6.598884290084579e-06, "loss": 0.5628, "step": 3836 }, { "epoch": 1.38, "learning_rate": 6.597984524023754e-06, "loss": 0.5966, "step": 3837 }, { "epoch": 1.38, "learning_rate": 6.59708475796293e-06, "loss": 0.5637, "step": 3838 }, { "epoch": 1.38, "learning_rate": 6.596184991902106e-06, "loss": 0.6685, "step": 3839 }, { "epoch": 1.38, "learning_rate": 6.595285225841281e-06, "loss": 0.6693, "step": 3840 }, { "epoch": 1.38, "learning_rate": 6.594385459780458e-06, "loss": 0.5625, "step": 3841 }, { "epoch": 1.38, "learning_rate": 6.593485693719633e-06, "loss": 0.5554, "step": 3842 }, { "epoch": 1.38, "learning_rate": 6.592585927658809e-06, "loss": 0.7139, "step": 3843 }, { "epoch": 1.38, "learning_rate": 6.591686161597986e-06, "loss": 0.5789, "step": 3844 }, { "epoch": 1.38, "learning_rate": 6.590786395537161e-06, "loss": 0.6056, "step": 3845 }, { "epoch": 1.38, "learning_rate": 6.5898866294763365e-06, "loss": 0.697, "step": 3846 }, { "epoch": 1.38, "learning_rate": 6.588986863415513e-06, "loss": 0.5399, "step": 3847 }, { "epoch": 1.38, "learning_rate": 6.588087097354688e-06, "loss": 0.6346, "step": 3848 }, { "epoch": 1.38, "learning_rate": 6.587187331293864e-06, "loss": 0.5793, "step": 3849 }, { "epoch": 1.38, "learning_rate": 6.58628756523304e-06, "loss": 0.5367, "step": 3850 }, { "epoch": 1.38, "learning_rate": 6.585387799172216e-06, "loss": 0.6727, "step": 3851 }, { "epoch": 1.38, "learning_rate": 6.584488033111391e-06, "loss": 0.5316, "step": 3852 }, { "epoch": 1.38, "learning_rate": 6.583588267050567e-06, "loss": 0.6017, "step": 3853 }, { "epoch": 1.38, "learning_rate": 6.5826885009897435e-06, "loss": 0.6584, "step": 3854 }, { "epoch": 1.38, "learning_rate": 6.581788734928919e-06, "loss": 0.6142, "step": 3855 }, { "epoch": 1.38, "learning_rate": 6.580888968868094e-06, "loss": 0.588, "step": 3856 }, { "epoch": 1.38, "learning_rate": 6.579989202807271e-06, "loss": 0.6964, "step": 3857 }, { "epoch": 1.38, "learning_rate": 6.5790894367464466e-06, "loss": 0.5537, "step": 3858 }, { "epoch": 1.39, "learning_rate": 6.5781896706856215e-06, "loss": 0.6731, "step": 3859 }, { "epoch": 1.39, "learning_rate": 6.577289904624798e-06, "loss": 0.6892, "step": 3860 }, { "epoch": 1.39, "learning_rate": 6.576390138563974e-06, "loss": 0.6383, "step": 3861 }, { "epoch": 1.39, "learning_rate": 6.57549037250315e-06, "loss": 0.6244, "step": 3862 }, { "epoch": 1.39, "learning_rate": 6.574590606442326e-06, "loss": 0.6084, "step": 3863 }, { "epoch": 1.39, "learning_rate": 6.573690840381501e-06, "loss": 0.6433, "step": 3864 }, { "epoch": 1.39, "learning_rate": 6.572791074320677e-06, "loss": 0.5262, "step": 3865 }, { "epoch": 1.39, "learning_rate": 6.5718913082598536e-06, "loss": 0.5923, "step": 3866 }, { "epoch": 1.39, "learning_rate": 6.570991542199029e-06, "loss": 0.727, "step": 3867 }, { "epoch": 1.39, "learning_rate": 6.570091776138204e-06, "loss": 0.7362, "step": 3868 }, { "epoch": 1.39, "learning_rate": 6.569192010077381e-06, "loss": 0.6067, "step": 3869 }, { "epoch": 1.39, "learning_rate": 6.568292244016557e-06, "loss": 0.5719, "step": 3870 }, { "epoch": 1.39, "learning_rate": 6.5673924779557316e-06, "loss": 0.7304, "step": 3871 }, { "epoch": 1.39, "learning_rate": 6.566492711894907e-06, "loss": 0.6602, "step": 3872 }, { "epoch": 1.39, "learning_rate": 6.565592945834084e-06, "loss": 0.627, "step": 3873 }, { "epoch": 1.39, "learning_rate": 6.56469317977326e-06, "loss": 0.6038, "step": 3874 }, { "epoch": 1.39, "learning_rate": 6.563793413712435e-06, "loss": 0.5777, "step": 3875 }, { "epoch": 1.39, "learning_rate": 6.562893647651611e-06, "loss": 0.6007, "step": 3876 }, { "epoch": 1.39, "learning_rate": 6.561993881590787e-06, "loss": 0.6638, "step": 3877 }, { "epoch": 1.39, "learning_rate": 6.561094115529962e-06, "loss": 0.5863, "step": 3878 }, { "epoch": 1.39, "learning_rate": 6.5601943494691386e-06, "loss": 0.6437, "step": 3879 }, { "epoch": 1.39, "learning_rate": 6.559294583408314e-06, "loss": 0.5837, "step": 3880 }, { "epoch": 1.39, "learning_rate": 6.55839481734749e-06, "loss": 0.6704, "step": 3881 }, { "epoch": 1.39, "learning_rate": 6.557495051286667e-06, "loss": 0.585, "step": 3882 }, { "epoch": 1.39, "learning_rate": 6.556595285225842e-06, "loss": 0.6123, "step": 3883 }, { "epoch": 1.39, "learning_rate": 6.555695519165017e-06, "loss": 0.6193, "step": 3884 }, { "epoch": 1.39, "learning_rate": 6.554795753104194e-06, "loss": 0.6195, "step": 3885 }, { "epoch": 1.39, "learning_rate": 6.55389598704337e-06, "loss": 0.6464, "step": 3886 }, { "epoch": 1.4, "learning_rate": 6.552996220982545e-06, "loss": 0.597, "step": 3887 }, { "epoch": 1.4, "learning_rate": 6.552096454921721e-06, "loss": 0.6217, "step": 3888 }, { "epoch": 1.4, "learning_rate": 6.551196688860897e-06, "loss": 0.6255, "step": 3889 }, { "epoch": 1.4, "learning_rate": 6.550296922800072e-06, "loss": 0.6486, "step": 3890 }, { "epoch": 1.4, "learning_rate": 6.549397156739248e-06, "loss": 0.5958, "step": 3891 }, { "epoch": 1.4, "learning_rate": 6.548497390678424e-06, "loss": 0.7245, "step": 3892 }, { "epoch": 1.4, "learning_rate": 6.5475976246176e-06, "loss": 0.6573, "step": 3893 }, { "epoch": 1.4, "learning_rate": 6.546697858556775e-06, "loss": 0.58, "step": 3894 }, { "epoch": 1.4, "learning_rate": 6.545798092495952e-06, "loss": 0.5924, "step": 3895 }, { "epoch": 1.4, "learning_rate": 6.5448983264351275e-06, "loss": 0.4907, "step": 3896 }, { "epoch": 1.4, "learning_rate": 6.543998560374302e-06, "loss": 0.6701, "step": 3897 }, { "epoch": 1.4, "learning_rate": 6.543098794313479e-06, "loss": 0.5465, "step": 3898 }, { "epoch": 1.4, "learning_rate": 6.542199028252655e-06, "loss": 0.649, "step": 3899 }, { "epoch": 1.4, "learning_rate": 6.5412992621918305e-06, "loss": 0.5772, "step": 3900 }, { "epoch": 1.4, "learning_rate": 6.540399496131007e-06, "loss": 0.5712, "step": 3901 }, { "epoch": 1.4, "learning_rate": 6.539499730070182e-06, "loss": 0.676, "step": 3902 }, { "epoch": 1.4, "learning_rate": 6.538599964009358e-06, "loss": 0.6261, "step": 3903 }, { "epoch": 1.4, "learning_rate": 6.5377001979485345e-06, "loss": 0.5404, "step": 3904 }, { "epoch": 1.4, "learning_rate": 6.53680043188771e-06, "loss": 0.7752, "step": 3905 }, { "epoch": 1.4, "learning_rate": 6.535900665826885e-06, "loss": 0.645, "step": 3906 }, { "epoch": 1.4, "learning_rate": 6.535000899766062e-06, "loss": 0.6313, "step": 3907 }, { "epoch": 1.4, "learning_rate": 6.5341011337052375e-06, "loss": 0.6422, "step": 3908 }, { "epoch": 1.4, "learning_rate": 6.5332013676444125e-06, "loss": 0.5995, "step": 3909 }, { "epoch": 1.4, "learning_rate": 6.532301601583588e-06, "loss": 0.5767, "step": 3910 }, { "epoch": 1.4, "learning_rate": 6.531401835522765e-06, "loss": 0.6322, "step": 3911 }, { "epoch": 1.4, "learning_rate": 6.530502069461941e-06, "loss": 0.6167, "step": 3912 }, { "epoch": 1.4, "learning_rate": 6.5296023034011155e-06, "loss": 0.6844, "step": 3913 }, { "epoch": 1.4, "learning_rate": 6.528702537340292e-06, "loss": 0.6575, "step": 3914 }, { "epoch": 1.41, "learning_rate": 6.527802771279468e-06, "loss": 0.6119, "step": 3915 }, { "epoch": 1.41, "learning_rate": 6.526903005218644e-06, "loss": 0.6609, "step": 3916 }, { "epoch": 1.41, "learning_rate": 6.5260032391578195e-06, "loss": 0.6778, "step": 3917 }, { "epoch": 1.41, "learning_rate": 6.525103473096995e-06, "loss": 0.6067, "step": 3918 }, { "epoch": 1.41, "learning_rate": 6.524203707036171e-06, "loss": 0.5787, "step": 3919 }, { "epoch": 1.41, "learning_rate": 6.523303940975348e-06, "loss": 0.6192, "step": 3920 }, { "epoch": 1.41, "learning_rate": 6.5224041749145225e-06, "loss": 0.6286, "step": 3921 }, { "epoch": 1.41, "learning_rate": 6.521504408853698e-06, "loss": 0.7003, "step": 3922 }, { "epoch": 1.41, "learning_rate": 6.520604642792875e-06, "loss": 0.6756, "step": 3923 }, { "epoch": 1.41, "learning_rate": 6.519704876732051e-06, "loss": 0.651, "step": 3924 }, { "epoch": 1.41, "learning_rate": 6.518805110671226e-06, "loss": 0.6153, "step": 3925 }, { "epoch": 1.41, "learning_rate": 6.517905344610402e-06, "loss": 0.6253, "step": 3926 }, { "epoch": 1.41, "learning_rate": 6.517005578549578e-06, "loss": 0.6635, "step": 3927 }, { "epoch": 1.41, "learning_rate": 6.516105812488753e-06, "loss": 0.6442, "step": 3928 }, { "epoch": 1.41, "learning_rate": 6.515206046427929e-06, "loss": 0.6227, "step": 3929 }, { "epoch": 1.41, "learning_rate": 6.514306280367105e-06, "loss": 0.6466, "step": 3930 }, { "epoch": 1.41, "learning_rate": 6.513406514306281e-06, "loss": 0.5864, "step": 3931 }, { "epoch": 1.41, "learning_rate": 6.512506748245456e-06, "loss": 0.6814, "step": 3932 }, { "epoch": 1.41, "learning_rate": 6.511606982184633e-06, "loss": 0.6312, "step": 3933 }, { "epoch": 1.41, "learning_rate": 6.510707216123808e-06, "loss": 0.6309, "step": 3934 }, { "epoch": 1.41, "learning_rate": 6.509807450062984e-06, "loss": 0.547, "step": 3935 }, { "epoch": 1.41, "learning_rate": 6.50890768400216e-06, "loss": 0.5269, "step": 3936 }, { "epoch": 1.41, "learning_rate": 6.508007917941336e-06, "loss": 0.6087, "step": 3937 }, { "epoch": 1.41, "learning_rate": 6.5071081518805115e-06, "loss": 0.6389, "step": 3938 }, { "epoch": 1.41, "learning_rate": 6.506208385819688e-06, "loss": 0.663, "step": 3939 }, { "epoch": 1.41, "learning_rate": 6.505308619758863e-06, "loss": 0.688, "step": 3940 }, { "epoch": 1.41, "learning_rate": 6.504408853698039e-06, "loss": 0.697, "step": 3941 }, { "epoch": 1.41, "learning_rate": 6.503509087637215e-06, "loss": 0.6023, "step": 3942 }, { "epoch": 1.42, "learning_rate": 6.502609321576391e-06, "loss": 0.6008, "step": 3943 }, { "epoch": 1.42, "learning_rate": 6.501709555515566e-06, "loss": 0.7295, "step": 3944 }, { "epoch": 1.42, "learning_rate": 6.500809789454743e-06, "loss": 0.661, "step": 3945 }, { "epoch": 1.42, "learning_rate": 6.4999100233939184e-06, "loss": 0.6234, "step": 3946 }, { "epoch": 1.42, "learning_rate": 6.499010257333093e-06, "loss": 0.7134, "step": 3947 }, { "epoch": 1.42, "learning_rate": 6.498110491272269e-06, "loss": 0.5576, "step": 3948 }, { "epoch": 1.42, "learning_rate": 6.497210725211446e-06, "loss": 0.6768, "step": 3949 }, { "epoch": 1.42, "learning_rate": 6.4963109591506215e-06, "loss": 0.5404, "step": 3950 }, { "epoch": 1.42, "learning_rate": 6.4954111930897964e-06, "loss": 0.6111, "step": 3951 }, { "epoch": 1.42, "learning_rate": 6.494511427028973e-06, "loss": 0.6117, "step": 3952 }, { "epoch": 1.42, "learning_rate": 6.493611660968149e-06, "loss": 0.6089, "step": 3953 }, { "epoch": 1.42, "learning_rate": 6.492711894907325e-06, "loss": 0.6526, "step": 3954 }, { "epoch": 1.42, "learning_rate": 6.491812128846501e-06, "loss": 0.6324, "step": 3955 }, { "epoch": 1.42, "learning_rate": 6.490912362785676e-06, "loss": 0.6931, "step": 3956 }, { "epoch": 1.42, "learning_rate": 6.490012596724852e-06, "loss": 0.6682, "step": 3957 }, { "epoch": 1.42, "learning_rate": 6.4891128306640285e-06, "loss": 0.6801, "step": 3958 }, { "epoch": 1.42, "learning_rate": 6.4882130646032034e-06, "loss": 0.5329, "step": 3959 }, { "epoch": 1.42, "learning_rate": 6.487313298542379e-06, "loss": 0.5645, "step": 3960 }, { "epoch": 1.42, "learning_rate": 6.486413532481556e-06, "loss": 0.6341, "step": 3961 }, { "epoch": 1.42, "learning_rate": 6.485513766420732e-06, "loss": 0.6548, "step": 3962 }, { "epoch": 1.42, "learning_rate": 6.4846140003599065e-06, "loss": 0.609, "step": 3963 }, { "epoch": 1.42, "learning_rate": 6.483714234299083e-06, "loss": 0.6193, "step": 3964 }, { "epoch": 1.42, "learning_rate": 6.482814468238259e-06, "loss": 0.5911, "step": 3965 }, { "epoch": 1.42, "learning_rate": 6.481914702177434e-06, "loss": 0.6679, "step": 3966 }, { "epoch": 1.42, "learning_rate": 6.48101493611661e-06, "loss": 0.6866, "step": 3967 }, { "epoch": 1.42, "learning_rate": 6.480115170055786e-06, "loss": 0.5991, "step": 3968 }, { "epoch": 1.42, "learning_rate": 6.479215403994962e-06, "loss": 0.6533, "step": 3969 }, { "epoch": 1.42, "learning_rate": 6.478315637934137e-06, "loss": 0.5508, "step": 3970 }, { "epoch": 1.43, "learning_rate": 6.4774158718733135e-06, "loss": 0.5446, "step": 3971 }, { "epoch": 1.43, "learning_rate": 6.476516105812489e-06, "loss": 0.622, "step": 3972 }, { "epoch": 1.43, "learning_rate": 6.475616339751665e-06, "loss": 0.5725, "step": 3973 }, { "epoch": 1.43, "learning_rate": 6.474716573690842e-06, "loss": 0.5778, "step": 3974 }, { "epoch": 1.43, "learning_rate": 6.473816807630017e-06, "loss": 0.5931, "step": 3975 }, { "epoch": 1.43, "learning_rate": 6.472917041569192e-06, "loss": 0.6192, "step": 3976 }, { "epoch": 1.43, "learning_rate": 6.472017275508369e-06, "loss": 0.6403, "step": 3977 }, { "epoch": 1.43, "learning_rate": 6.471117509447544e-06, "loss": 0.6689, "step": 3978 }, { "epoch": 1.43, "learning_rate": 6.47021774338672e-06, "loss": 0.7096, "step": 3979 }, { "epoch": 1.43, "learning_rate": 6.469317977325896e-06, "loss": 0.7869, "step": 3980 }, { "epoch": 1.43, "learning_rate": 6.468418211265072e-06, "loss": 0.5484, "step": 3981 }, { "epoch": 1.43, "learning_rate": 6.467518445204247e-06, "loss": 0.5917, "step": 3982 }, { "epoch": 1.43, "learning_rate": 6.466618679143424e-06, "loss": 0.5602, "step": 3983 }, { "epoch": 1.43, "learning_rate": 6.465718913082599e-06, "loss": 0.6322, "step": 3984 }, { "epoch": 1.43, "learning_rate": 6.464819147021775e-06, "loss": 0.6213, "step": 3985 }, { "epoch": 1.43, "learning_rate": 6.46391938096095e-06, "loss": 0.6124, "step": 3986 }, { "epoch": 1.43, "learning_rate": 6.463019614900127e-06, "loss": 0.5948, "step": 3987 }, { "epoch": 1.43, "learning_rate": 6.4621198488393024e-06, "loss": 0.5809, "step": 3988 }, { "epoch": 1.43, "learning_rate": 6.461220082778477e-06, "loss": 0.6053, "step": 3989 }, { "epoch": 1.43, "learning_rate": 6.460320316717654e-06, "loss": 0.6633, "step": 3990 }, { "epoch": 1.43, "learning_rate": 6.45942055065683e-06, "loss": 0.5473, "step": 3991 }, { "epoch": 1.43, "learning_rate": 6.4585207845960055e-06, "loss": 0.5763, "step": 3992 }, { "epoch": 1.43, "learning_rate": 6.457621018535182e-06, "loss": 0.5437, "step": 3993 }, { "epoch": 1.43, "learning_rate": 6.456721252474357e-06, "loss": 0.7044, "step": 3994 }, { "epoch": 1.43, "learning_rate": 6.455821486413533e-06, "loss": 0.538, "step": 3995 }, { "epoch": 1.43, "learning_rate": 6.4549217203527094e-06, "loss": 0.6527, "step": 3996 }, { "epoch": 1.43, "learning_rate": 6.454021954291884e-06, "loss": 0.4892, "step": 3997 }, { "epoch": 1.44, "learning_rate": 6.45312218823106e-06, "loss": 0.6199, "step": 3998 }, { "epoch": 1.44, "learning_rate": 6.452222422170237e-06, "loss": 0.5009, "step": 3999 }, { "epoch": 1.44, "learning_rate": 6.4513226561094125e-06, "loss": 0.5461, "step": 4000 }, { "epoch": 1.44, "learning_rate": 6.4504228900485874e-06, "loss": 0.6127, "step": 4001 }, { "epoch": 1.44, "learning_rate": 6.449523123987764e-06, "loss": 0.6855, "step": 4002 }, { "epoch": 1.44, "learning_rate": 6.44862335792694e-06, "loss": 0.6076, "step": 4003 }, { "epoch": 1.44, "learning_rate": 6.447723591866116e-06, "loss": 0.6355, "step": 4004 }, { "epoch": 1.44, "learning_rate": 6.4468238258052905e-06, "loss": 0.5814, "step": 4005 }, { "epoch": 1.44, "learning_rate": 6.445924059744467e-06, "loss": 0.5614, "step": 4006 }, { "epoch": 1.44, "learning_rate": 6.445024293683643e-06, "loss": 0.5948, "step": 4007 }, { "epoch": 1.44, "learning_rate": 6.444124527622818e-06, "loss": 0.7023, "step": 4008 }, { "epoch": 1.44, "learning_rate": 6.443224761561994e-06, "loss": 0.6986, "step": 4009 }, { "epoch": 1.44, "learning_rate": 6.44232499550117e-06, "loss": 0.5498, "step": 4010 }, { "epoch": 1.44, "learning_rate": 6.441425229440346e-06, "loss": 0.6299, "step": 4011 }, { "epoch": 1.44, "learning_rate": 6.4405254633795226e-06, "loss": 0.5898, "step": 4012 }, { "epoch": 1.44, "learning_rate": 6.4396256973186975e-06, "loss": 0.6677, "step": 4013 }, { "epoch": 1.44, "learning_rate": 6.438725931257873e-06, "loss": 0.5742, "step": 4014 }, { "epoch": 1.44, "learning_rate": 6.43782616519705e-06, "loss": 0.7231, "step": 4015 }, { "epoch": 1.44, "learning_rate": 6.436926399136225e-06, "loss": 0.5783, "step": 4016 }, { "epoch": 1.44, "learning_rate": 6.4360266330754006e-06, "loss": 0.582, "step": 4017 }, { "epoch": 1.44, "learning_rate": 6.435126867014577e-06, "loss": 0.5601, "step": 4018 }, { "epoch": 1.44, "learning_rate": 6.434227100953753e-06, "loss": 0.6598, "step": 4019 }, { "epoch": 1.44, "learning_rate": 6.433327334892928e-06, "loss": 0.6049, "step": 4020 }, { "epoch": 1.44, "learning_rate": 6.4324275688321045e-06, "loss": 0.597, "step": 4021 }, { "epoch": 1.44, "learning_rate": 6.43152780277128e-06, "loss": 0.6427, "step": 4022 }, { "epoch": 1.44, "learning_rate": 6.430628036710456e-06, "loss": 0.6531, "step": 4023 }, { "epoch": 1.44, "learning_rate": 6.429728270649631e-06, "loss": 0.595, "step": 4024 }, { "epoch": 1.44, "learning_rate": 6.4288285045888076e-06, "loss": 0.6001, "step": 4025 }, { "epoch": 1.45, "learning_rate": 6.427928738527983e-06, "loss": 0.5912, "step": 4026 }, { "epoch": 1.45, "learning_rate": 6.427028972467158e-06, "loss": 0.5972, "step": 4027 }, { "epoch": 1.45, "learning_rate": 6.426129206406335e-06, "loss": 0.6159, "step": 4028 }, { "epoch": 1.45, "learning_rate": 6.425229440345511e-06, "loss": 0.662, "step": 4029 }, { "epoch": 1.45, "learning_rate": 6.424329674284686e-06, "loss": 0.6671, "step": 4030 }, { "epoch": 1.45, "learning_rate": 6.423429908223863e-06, "loss": 0.5046, "step": 4031 }, { "epoch": 1.45, "learning_rate": 6.422530142163038e-06, "loss": 0.6248, "step": 4032 }, { "epoch": 1.45, "learning_rate": 6.421630376102214e-06, "loss": 0.6278, "step": 4033 }, { "epoch": 1.45, "learning_rate": 6.42073061004139e-06, "loss": 0.5738, "step": 4034 }, { "epoch": 1.45, "learning_rate": 6.419830843980565e-06, "loss": 0.6466, "step": 4035 }, { "epoch": 1.45, "learning_rate": 6.418931077919741e-06, "loss": 0.6065, "step": 4036 }, { "epoch": 1.45, "learning_rate": 6.418031311858918e-06, "loss": 0.6029, "step": 4037 }, { "epoch": 1.45, "learning_rate": 6.417131545798093e-06, "loss": 0.5572, "step": 4038 }, { "epoch": 1.45, "learning_rate": 6.416231779737268e-06, "loss": 0.6417, "step": 4039 }, { "epoch": 1.45, "learning_rate": 6.415332013676445e-06, "loss": 0.6104, "step": 4040 }, { "epoch": 1.45, "learning_rate": 6.414432247615621e-06, "loss": 0.6407, "step": 4041 }, { "epoch": 1.45, "learning_rate": 6.4135324815547965e-06, "loss": 0.6502, "step": 4042 }, { "epoch": 1.45, "learning_rate": 6.412632715493971e-06, "loss": 0.6359, "step": 4043 }, { "epoch": 1.45, "learning_rate": 6.411732949433148e-06, "loss": 0.6761, "step": 4044 }, { "epoch": 1.45, "learning_rate": 6.410833183372324e-06, "loss": 0.5718, "step": 4045 }, { "epoch": 1.45, "learning_rate": 6.409933417311499e-06, "loss": 0.6585, "step": 4046 }, { "epoch": 1.45, "learning_rate": 6.409033651250675e-06, "loss": 0.6333, "step": 4047 }, { "epoch": 1.45, "learning_rate": 6.408133885189851e-06, "loss": 0.6205, "step": 4048 }, { "epoch": 1.45, "learning_rate": 6.407234119129027e-06, "loss": 0.5338, "step": 4049 }, { "epoch": 1.45, "learning_rate": 6.4063343530682035e-06, "loss": 0.6965, "step": 4050 }, { "epoch": 1.45, "learning_rate": 6.405434587007378e-06, "loss": 0.6216, "step": 4051 }, { "epoch": 1.45, "learning_rate": 6.404534820946554e-06, "loss": 0.6206, "step": 4052 }, { "epoch": 1.45, "learning_rate": 6.403635054885731e-06, "loss": 0.5442, "step": 4053 }, { "epoch": 1.46, "learning_rate": 6.4027352888249066e-06, "loss": 0.6572, "step": 4054 }, { "epoch": 1.46, "learning_rate": 6.4018355227640815e-06, "loss": 0.6543, "step": 4055 }, { "epoch": 1.46, "learning_rate": 6.400935756703258e-06, "loss": 0.7409, "step": 4056 }, { "epoch": 1.46, "learning_rate": 6.400035990642434e-06, "loss": 0.5433, "step": 4057 }, { "epoch": 1.46, "learning_rate": 6.399136224581609e-06, "loss": 0.7629, "step": 4058 }, { "epoch": 1.46, "learning_rate": 6.3982364585207846e-06, "loss": 0.6732, "step": 4059 }, { "epoch": 1.46, "learning_rate": 6.397336692459961e-06, "loss": 0.5866, "step": 4060 }, { "epoch": 1.46, "learning_rate": 6.396436926399137e-06, "loss": 0.5973, "step": 4061 }, { "epoch": 1.46, "learning_rate": 6.395537160338312e-06, "loss": 0.5915, "step": 4062 }, { "epoch": 1.46, "learning_rate": 6.3946373942774885e-06, "loss": 0.6018, "step": 4063 }, { "epoch": 1.46, "learning_rate": 6.393737628216664e-06, "loss": 0.5686, "step": 4064 }, { "epoch": 1.46, "learning_rate": 6.392837862155839e-06, "loss": 0.6109, "step": 4065 }, { "epoch": 1.46, "learning_rate": 6.391938096095016e-06, "loss": 0.5908, "step": 4066 }, { "epoch": 1.46, "learning_rate": 6.3910383300341916e-06, "loss": 0.6564, "step": 4067 }, { "epoch": 1.46, "learning_rate": 6.390138563973367e-06, "loss": 0.5573, "step": 4068 }, { "epoch": 1.46, "learning_rate": 6.389238797912544e-06, "loss": 0.5735, "step": 4069 }, { "epoch": 1.46, "learning_rate": 6.388339031851719e-06, "loss": 0.6515, "step": 4070 }, { "epoch": 1.46, "learning_rate": 6.387439265790895e-06, "loss": 0.5662, "step": 4071 }, { "epoch": 1.46, "learning_rate": 6.386539499730071e-06, "loss": 0.5997, "step": 4072 }, { "epoch": 1.46, "learning_rate": 6.385639733669247e-06, "loss": 0.6513, "step": 4073 }, { "epoch": 1.46, "learning_rate": 6.384739967608422e-06, "loss": 0.505, "step": 4074 }, { "epoch": 1.46, "learning_rate": 6.3838402015475986e-06, "loss": 0.6175, "step": 4075 }, { "epoch": 1.46, "learning_rate": 6.382940435486774e-06, "loss": 0.6089, "step": 4076 }, { "epoch": 1.46, "learning_rate": 6.382040669425949e-06, "loss": 0.6576, "step": 4077 }, { "epoch": 1.46, "learning_rate": 6.381140903365125e-06, "loss": 0.6429, "step": 4078 }, { "epoch": 1.46, "learning_rate": 6.380241137304302e-06, "loss": 0.691, "step": 4079 }, { "epoch": 1.46, "learning_rate": 6.379341371243477e-06, "loss": 0.6247, "step": 4080 }, { "epoch": 1.46, "learning_rate": 6.378441605182652e-06, "loss": 0.6777, "step": 4081 }, { "epoch": 1.47, "learning_rate": 6.377541839121829e-06, "loss": 0.6799, "step": 4082 }, { "epoch": 1.47, "learning_rate": 6.376642073061005e-06, "loss": 0.5594, "step": 4083 }, { "epoch": 1.47, "learning_rate": 6.37574230700018e-06, "loss": 0.6032, "step": 4084 }, { "epoch": 1.47, "learning_rate": 6.374842540939356e-06, "loss": 0.6034, "step": 4085 }, { "epoch": 1.47, "learning_rate": 6.373942774878532e-06, "loss": 0.59, "step": 4086 }, { "epoch": 1.47, "learning_rate": 6.373043008817708e-06, "loss": 0.611, "step": 4087 }, { "epoch": 1.47, "learning_rate": 6.372143242756884e-06, "loss": 0.6504, "step": 4088 }, { "epoch": 1.47, "learning_rate": 6.371243476696059e-06, "loss": 0.6402, "step": 4089 }, { "epoch": 1.47, "learning_rate": 6.370343710635235e-06, "loss": 0.644, "step": 4090 }, { "epoch": 1.47, "learning_rate": 6.369443944574412e-06, "loss": 0.6609, "step": 4091 }, { "epoch": 1.47, "learning_rate": 6.3685441785135875e-06, "loss": 0.6783, "step": 4092 }, { "epoch": 1.47, "learning_rate": 6.367644412452762e-06, "loss": 0.6512, "step": 4093 }, { "epoch": 1.47, "learning_rate": 6.366744646391939e-06, "loss": 0.5799, "step": 4094 }, { "epoch": 1.47, "learning_rate": 6.365844880331115e-06, "loss": 0.607, "step": 4095 }, { "epoch": 1.47, "learning_rate": 6.36494511427029e-06, "loss": 0.6124, "step": 4096 }, { "epoch": 1.47, "learning_rate": 6.3640453482094655e-06, "loss": 0.6095, "step": 4097 }, { "epoch": 1.47, "learning_rate": 6.363145582148642e-06, "loss": 0.6292, "step": 4098 }, { "epoch": 1.47, "learning_rate": 6.362245816087818e-06, "loss": 0.6156, "step": 4099 }, { "epoch": 1.47, "learning_rate": 6.361346050026993e-06, "loss": 0.6514, "step": 4100 }, { "epoch": 1.47, "learning_rate": 6.360446283966169e-06, "loss": 0.6286, "step": 4101 }, { "epoch": 1.47, "learning_rate": 6.359546517905345e-06, "loss": 0.6615, "step": 4102 }, { "epoch": 1.47, "learning_rate": 6.358646751844521e-06, "loss": 0.6167, "step": 4103 }, { "epoch": 1.47, "learning_rate": 6.357746985783697e-06, "loss": 0.6658, "step": 4104 }, { "epoch": 1.47, "learning_rate": 6.3568472197228725e-06, "loss": 0.6176, "step": 4105 }, { "epoch": 1.47, "learning_rate": 6.355947453662048e-06, "loss": 0.6585, "step": 4106 }, { "epoch": 1.47, "learning_rate": 6.355047687601225e-06, "loss": 0.6308, "step": 4107 }, { "epoch": 1.47, "learning_rate": 6.3541479215404e-06, "loss": 0.6346, "step": 4108 }, { "epoch": 1.47, "learning_rate": 6.3532481554795755e-06, "loss": 0.5798, "step": 4109 }, { "epoch": 1.48, "learning_rate": 6.352348389418752e-06, "loss": 0.596, "step": 4110 }, { "epoch": 1.48, "learning_rate": 6.351448623357928e-06, "loss": 0.6346, "step": 4111 }, { "epoch": 1.48, "learning_rate": 6.350548857297103e-06, "loss": 0.5998, "step": 4112 }, { "epoch": 1.48, "learning_rate": 6.3496490912362795e-06, "loss": 0.5661, "step": 4113 }, { "epoch": 1.48, "learning_rate": 6.348749325175455e-06, "loss": 0.5714, "step": 4114 }, { "epoch": 1.48, "learning_rate": 6.34784955911463e-06, "loss": 0.5619, "step": 4115 }, { "epoch": 1.48, "learning_rate": 6.346949793053806e-06, "loss": 0.5838, "step": 4116 }, { "epoch": 1.48, "learning_rate": 6.3460500269929825e-06, "loss": 0.7183, "step": 4117 }, { "epoch": 1.48, "learning_rate": 6.345150260932158e-06, "loss": 0.5853, "step": 4118 }, { "epoch": 1.48, "learning_rate": 6.344250494871333e-06, "loss": 0.5971, "step": 4119 }, { "epoch": 1.48, "learning_rate": 6.34335072881051e-06, "loss": 0.6661, "step": 4120 }, { "epoch": 1.48, "learning_rate": 6.342450962749686e-06, "loss": 0.5438, "step": 4121 }, { "epoch": 1.48, "learning_rate": 6.342450962749686e-06, "loss": 0.6767, "step": 4122 }, { "epoch": 1.48, "learning_rate": 6.341551196688861e-06, "loss": 0.5569, "step": 4123 }, { "epoch": 1.48, "learning_rate": 6.340651430628037e-06, "loss": 0.5742, "step": 4124 }, { "epoch": 1.48, "learning_rate": 6.339751664567213e-06, "loss": 0.549, "step": 4125 }, { "epoch": 1.48, "learning_rate": 6.338851898506389e-06, "loss": 0.6595, "step": 4126 }, { "epoch": 1.48, "learning_rate": 6.337952132445565e-06, "loss": 0.6187, "step": 4127 }, { "epoch": 1.48, "learning_rate": 6.33705236638474e-06, "loss": 0.585, "step": 4128 }, { "epoch": 1.48, "learning_rate": 6.336152600323916e-06, "loss": 0.6475, "step": 4129 }, { "epoch": 1.48, "learning_rate": 6.335252834263093e-06, "loss": 0.5845, "step": 4130 }, { "epoch": 1.48, "learning_rate": 6.334353068202268e-06, "loss": 0.5734, "step": 4131 }, { "epoch": 1.48, "learning_rate": 6.333453302141443e-06, "loss": 0.6676, "step": 4132 }, { "epoch": 1.48, "learning_rate": 6.33255353608062e-06, "loss": 0.6752, "step": 4133 }, { "epoch": 1.48, "learning_rate": 6.331653770019796e-06, "loss": 0.6254, "step": 4134 }, { "epoch": 1.48, "learning_rate": 6.330754003958971e-06, "loss": 0.5959, "step": 4135 }, { "epoch": 1.48, "learning_rate": 6.329854237898146e-06, "loss": 0.6678, "step": 4136 }, { "epoch": 1.48, "learning_rate": 6.328954471837323e-06, "loss": 0.5537, "step": 4137 }, { "epoch": 1.49, "learning_rate": 6.328054705776499e-06, "loss": 0.5638, "step": 4138 }, { "epoch": 1.49, "learning_rate": 6.327154939715674e-06, "loss": 0.7289, "step": 4139 }, { "epoch": 1.49, "learning_rate": 6.32625517365485e-06, "loss": 0.6789, "step": 4140 }, { "epoch": 1.49, "learning_rate": 6.325355407594026e-06, "loss": 0.6156, "step": 4141 }, { "epoch": 1.49, "learning_rate": 6.324455641533202e-06, "loss": 0.6613, "step": 4142 }, { "epoch": 1.49, "learning_rate": 6.3235558754723784e-06, "loss": 0.7894, "step": 4143 }, { "epoch": 1.49, "learning_rate": 6.322656109411553e-06, "loss": 0.6177, "step": 4144 }, { "epoch": 1.49, "learning_rate": 6.321756343350729e-06, "loss": 0.5724, "step": 4145 }, { "epoch": 1.49, "learning_rate": 6.320856577289906e-06, "loss": 0.5978, "step": 4146 }, { "epoch": 1.49, "learning_rate": 6.319956811229081e-06, "loss": 0.588, "step": 4147 }, { "epoch": 1.49, "learning_rate": 6.3190570451682564e-06, "loss": 0.6232, "step": 4148 }, { "epoch": 1.49, "learning_rate": 6.318157279107433e-06, "loss": 0.6203, "step": 4149 }, { "epoch": 1.49, "learning_rate": 6.317257513046609e-06, "loss": 0.5165, "step": 4150 }, { "epoch": 1.49, "learning_rate": 6.316357746985784e-06, "loss": 0.6776, "step": 4151 }, { "epoch": 1.49, "learning_rate": 6.31545798092496e-06, "loss": 0.6512, "step": 4152 }, { "epoch": 1.49, "learning_rate": 6.314558214864136e-06, "loss": 0.6603, "step": 4153 }, { "epoch": 1.49, "learning_rate": 6.313658448803311e-06, "loss": 0.5818, "step": 4154 }, { "epoch": 1.49, "learning_rate": 6.312758682742487e-06, "loss": 0.611, "step": 4155 }, { "epoch": 1.49, "learning_rate": 6.3118589166816634e-06, "loss": 0.5601, "step": 4156 }, { "epoch": 1.49, "learning_rate": 6.310959150620839e-06, "loss": 0.6775, "step": 4157 }, { "epoch": 1.49, "learning_rate": 6.310059384560014e-06, "loss": 0.6593, "step": 4158 }, { "epoch": 1.49, "learning_rate": 6.309159618499191e-06, "loss": 0.6234, "step": 4159 }, { "epoch": 1.49, "learning_rate": 6.3082598524383665e-06, "loss": 0.6366, "step": 4160 }, { "epoch": 1.49, "learning_rate": 6.307360086377542e-06, "loss": 0.6459, "step": 4161 }, { "epoch": 1.49, "learning_rate": 6.306460320316719e-06, "loss": 0.6485, "step": 4162 }, { "epoch": 1.49, "learning_rate": 6.305560554255894e-06, "loss": 0.667, "step": 4163 }, { "epoch": 1.49, "learning_rate": 6.30466078819507e-06, "loss": 0.5993, "step": 4164 }, { "epoch": 1.49, "learning_rate": 6.303761022134246e-06, "loss": 0.6976, "step": 4165 }, { "epoch": 1.5, "learning_rate": 6.302861256073421e-06, "loss": 0.6115, "step": 4166 }, { "epoch": 1.5, "learning_rate": 6.301961490012597e-06, "loss": 0.6286, "step": 4167 }, { "epoch": 1.5, "learning_rate": 6.3010617239517735e-06, "loss": 0.6042, "step": 4168 }, { "epoch": 1.5, "learning_rate": 6.300161957890949e-06, "loss": 0.7548, "step": 4169 }, { "epoch": 1.5, "learning_rate": 6.299262191830124e-06, "loss": 0.5842, "step": 4170 }, { "epoch": 1.5, "learning_rate": 6.298362425769301e-06, "loss": 0.5576, "step": 4171 }, { "epoch": 1.5, "learning_rate": 6.297462659708477e-06, "loss": 0.6113, "step": 4172 }, { "epoch": 1.5, "learning_rate": 6.296562893647652e-06, "loss": 0.5706, "step": 4173 }, { "epoch": 1.5, "learning_rate": 6.295663127586827e-06, "loss": 0.6618, "step": 4174 }, { "epoch": 1.5, "learning_rate": 6.294763361526004e-06, "loss": 0.6174, "step": 4175 }, { "epoch": 1.5, "learning_rate": 6.29386359546518e-06, "loss": 0.5558, "step": 4176 }, { "epoch": 1.5, "learning_rate": 6.292963829404355e-06, "loss": 0.61, "step": 4177 }, { "epoch": 1.5, "learning_rate": 6.292064063343531e-06, "loss": 0.6062, "step": 4178 }, { "epoch": 1.5, "learning_rate": 6.291164297282707e-06, "loss": 0.6481, "step": 4179 }, { "epoch": 1.5, "learning_rate": 6.290264531221883e-06, "loss": 0.666, "step": 4180 }, { "epoch": 1.5, "learning_rate": 6.289364765161059e-06, "loss": 0.5699, "step": 4181 }, { "epoch": 1.5, "learning_rate": 6.288464999100234e-06, "loss": 0.5904, "step": 4182 }, { "epoch": 1.5, "learning_rate": 6.28756523303941e-06, "loss": 0.5774, "step": 4183 }, { "epoch": 1.5, "learning_rate": 6.286665466978587e-06, "loss": 0.548, "step": 4184 }, { "epoch": 1.5, "learning_rate": 6.285765700917762e-06, "loss": 0.5545, "step": 4185 }, { "epoch": 1.5, "learning_rate": 6.284865934856937e-06, "loss": 0.5836, "step": 4186 }, { "epoch": 1.5, "learning_rate": 6.283966168796114e-06, "loss": 0.6364, "step": 4187 }, { "epoch": 1.5, "learning_rate": 6.28306640273529e-06, "loss": 0.5133, "step": 4188 }, { "epoch": 1.5, "learning_rate": 6.282166636674465e-06, "loss": 0.6422, "step": 4189 }, { "epoch": 1.5, "learning_rate": 6.281266870613641e-06, "loss": 0.671, "step": 4190 }, { "epoch": 1.5, "learning_rate": 6.280367104552817e-06, "loss": 0.6954, "step": 4191 }, { "epoch": 1.5, "learning_rate": 6.279467338491993e-06, "loss": 0.6466, "step": 4192 }, { "epoch": 1.51, "learning_rate": 6.278567572431168e-06, "loss": 0.5679, "step": 4193 }, { "epoch": 1.51, "learning_rate": 6.277667806370344e-06, "loss": 0.594, "step": 4194 }, { "epoch": 1.51, "learning_rate": 6.27676804030952e-06, "loss": 0.5624, "step": 4195 }, { "epoch": 1.51, "learning_rate": 6.275868274248695e-06, "loss": 0.5578, "step": 4196 }, { "epoch": 1.51, "learning_rate": 6.274968508187872e-06, "loss": 0.6046, "step": 4197 }, { "epoch": 1.51, "learning_rate": 6.2740687421270474e-06, "loss": 0.5757, "step": 4198 }, { "epoch": 1.51, "learning_rate": 6.273168976066223e-06, "loss": 0.7114, "step": 4199 }, { "epoch": 1.51, "learning_rate": 6.2722692100054e-06, "loss": 0.5834, "step": 4200 }, { "epoch": 1.51, "learning_rate": 6.271369443944575e-06, "loss": 0.6506, "step": 4201 }, { "epoch": 1.51, "learning_rate": 6.2704696778837505e-06, "loss": 0.6598, "step": 4202 }, { "epoch": 1.51, "learning_rate": 6.269569911822927e-06, "loss": 0.6111, "step": 4203 }, { "epoch": 1.51, "learning_rate": 6.268670145762102e-06, "loss": 0.6076, "step": 4204 }, { "epoch": 1.51, "learning_rate": 6.267770379701278e-06, "loss": 0.549, "step": 4205 }, { "epoch": 1.51, "learning_rate": 6.266870613640454e-06, "loss": 0.6415, "step": 4206 }, { "epoch": 1.51, "learning_rate": 6.26597084757963e-06, "loss": 0.6093, "step": 4207 }, { "epoch": 1.51, "learning_rate": 6.265071081518805e-06, "loss": 0.6764, "step": 4208 }, { "epoch": 1.51, "learning_rate": 6.264171315457982e-06, "loss": 0.6469, "step": 4209 }, { "epoch": 1.51, "learning_rate": 6.2632715493971575e-06, "loss": 0.6151, "step": 4210 }, { "epoch": 1.51, "learning_rate": 6.262371783336333e-06, "loss": 0.5953, "step": 4211 }, { "epoch": 1.51, "learning_rate": 6.261472017275508e-06, "loss": 0.5307, "step": 4212 }, { "epoch": 1.51, "learning_rate": 6.260572251214685e-06, "loss": 0.6703, "step": 4213 }, { "epoch": 1.51, "learning_rate": 6.2596724851538606e-06, "loss": 0.6706, "step": 4214 }, { "epoch": 1.51, "learning_rate": 6.2587727190930355e-06, "loss": 0.5652, "step": 4215 }, { "epoch": 1.51, "learning_rate": 6.257872953032212e-06, "loss": 0.5796, "step": 4216 }, { "epoch": 1.51, "learning_rate": 6.256973186971388e-06, "loss": 0.5922, "step": 4217 }, { "epoch": 1.51, "learning_rate": 6.256073420910564e-06, "loss": 0.5987, "step": 4218 }, { "epoch": 1.51, "learning_rate": 6.25517365484974e-06, "loss": 0.6669, "step": 4219 }, { "epoch": 1.51, "learning_rate": 6.254273888788915e-06, "loss": 0.6302, "step": 4220 }, { "epoch": 1.52, "learning_rate": 6.253374122728091e-06, "loss": 0.5667, "step": 4221 }, { "epoch": 1.52, "learning_rate": 6.2524743566672676e-06, "loss": 0.6026, "step": 4222 }, { "epoch": 1.52, "learning_rate": 6.2515745906064425e-06, "loss": 0.6825, "step": 4223 }, { "epoch": 1.52, "learning_rate": 6.250674824545618e-06, "loss": 0.6164, "step": 4224 }, { "epoch": 1.52, "learning_rate": 6.249775058484795e-06, "loss": 0.6754, "step": 4225 }, { "epoch": 1.52, "learning_rate": 6.248875292423971e-06, "loss": 0.5566, "step": 4226 }, { "epoch": 1.52, "learning_rate": 6.2479755263631456e-06, "loss": 0.5853, "step": 4227 }, { "epoch": 1.52, "learning_rate": 6.247075760302322e-06, "loss": 0.5919, "step": 4228 }, { "epoch": 1.52, "learning_rate": 6.246175994241498e-06, "loss": 0.6673, "step": 4229 }, { "epoch": 1.52, "learning_rate": 6.245276228180674e-06, "loss": 0.5855, "step": 4230 }, { "epoch": 1.52, "learning_rate": 6.244376462119849e-06, "loss": 0.5308, "step": 4231 }, { "epoch": 1.52, "learning_rate": 6.243476696059025e-06, "loss": 0.5858, "step": 4232 }, { "epoch": 1.52, "learning_rate": 6.242576929998201e-06, "loss": 0.6549, "step": 4233 }, { "epoch": 1.52, "learning_rate": 6.241677163937376e-06, "loss": 0.6321, "step": 4234 }, { "epoch": 1.52, "learning_rate": 6.2407773978765526e-06, "loss": 0.5935, "step": 4235 }, { "epoch": 1.52, "learning_rate": 6.239877631815728e-06, "loss": 0.6142, "step": 4236 }, { "epoch": 1.52, "learning_rate": 6.238977865754904e-06, "loss": 0.5709, "step": 4237 }, { "epoch": 1.52, "learning_rate": 6.238078099694081e-06, "loss": 0.5956, "step": 4238 }, { "epoch": 1.52, "learning_rate": 6.237178333633256e-06, "loss": 0.6416, "step": 4239 }, { "epoch": 1.52, "learning_rate": 6.236278567572431e-06, "loss": 0.6109, "step": 4240 }, { "epoch": 1.52, "learning_rate": 6.235378801511608e-06, "loss": 0.6202, "step": 4241 }, { "epoch": 1.52, "learning_rate": 6.234479035450783e-06, "loss": 0.7144, "step": 4242 }, { "epoch": 1.52, "learning_rate": 6.233579269389959e-06, "loss": 0.6224, "step": 4243 }, { "epoch": 1.52, "learning_rate": 6.232679503329135e-06, "loss": 0.598, "step": 4244 }, { "epoch": 1.52, "learning_rate": 6.231779737268311e-06, "loss": 0.6023, "step": 4245 }, { "epoch": 1.52, "learning_rate": 6.230879971207486e-06, "loss": 0.6211, "step": 4246 }, { "epoch": 1.52, "learning_rate": 6.229980205146663e-06, "loss": 0.6078, "step": 4247 }, { "epoch": 1.52, "learning_rate": 6.229080439085838e-06, "loss": 0.6982, "step": 4248 }, { "epoch": 1.53, "learning_rate": 6.228180673025014e-06, "loss": 0.723, "step": 4249 }, { "epoch": 1.53, "learning_rate": 6.227280906964189e-06, "loss": 0.5095, "step": 4250 }, { "epoch": 1.53, "learning_rate": 6.226381140903366e-06, "loss": 0.59, "step": 4251 }, { "epoch": 1.53, "learning_rate": 6.2254813748425415e-06, "loss": 0.6808, "step": 4252 }, { "epoch": 1.53, "learning_rate": 6.224581608781716e-06, "loss": 0.6374, "step": 4253 }, { "epoch": 1.53, "learning_rate": 6.223681842720893e-06, "loss": 0.6777, "step": 4254 }, { "epoch": 1.53, "learning_rate": 6.222782076660069e-06, "loss": 0.5777, "step": 4255 }, { "epoch": 1.53, "learning_rate": 6.2218823105992446e-06, "loss": 0.641, "step": 4256 }, { "epoch": 1.53, "learning_rate": 6.220982544538421e-06, "loss": 0.5976, "step": 4257 }, { "epoch": 1.53, "learning_rate": 6.220082778477596e-06, "loss": 0.587, "step": 4258 }, { "epoch": 1.53, "learning_rate": 6.219183012416772e-06, "loss": 0.705, "step": 4259 }, { "epoch": 1.53, "learning_rate": 6.2182832463559485e-06, "loss": 0.5941, "step": 4260 }, { "epoch": 1.53, "learning_rate": 6.217383480295124e-06, "loss": 0.6245, "step": 4261 }, { "epoch": 1.53, "learning_rate": 6.216483714234299e-06, "loss": 0.7146, "step": 4262 }, { "epoch": 1.53, "learning_rate": 6.215583948173476e-06, "loss": 0.5658, "step": 4263 }, { "epoch": 1.53, "learning_rate": 6.2146841821126516e-06, "loss": 0.6493, "step": 4264 }, { "epoch": 1.53, "learning_rate": 6.2137844160518265e-06, "loss": 0.7051, "step": 4265 }, { "epoch": 1.53, "learning_rate": 6.212884649991003e-06, "loss": 0.579, "step": 4266 }, { "epoch": 1.53, "learning_rate": 6.211984883930179e-06, "loss": 0.6864, "step": 4267 }, { "epoch": 1.53, "learning_rate": 6.211085117869355e-06, "loss": 0.6712, "step": 4268 }, { "epoch": 1.53, "learning_rate": 6.2101853518085296e-06, "loss": 0.6158, "step": 4269 }, { "epoch": 1.53, "learning_rate": 6.209285585747706e-06, "loss": 0.5393, "step": 4270 }, { "epoch": 1.53, "learning_rate": 6.208385819686882e-06, "loss": 0.6424, "step": 4271 }, { "epoch": 1.53, "learning_rate": 6.207486053626057e-06, "loss": 0.6761, "step": 4272 }, { "epoch": 1.53, "learning_rate": 6.2065862875652335e-06, "loss": 0.75, "step": 4273 }, { "epoch": 1.53, "learning_rate": 6.205686521504409e-06, "loss": 0.566, "step": 4274 }, { "epoch": 1.53, "learning_rate": 6.204786755443585e-06, "loss": 0.6334, "step": 4275 }, { "epoch": 1.53, "learning_rate": 6.203886989382762e-06, "loss": 0.6985, "step": 4276 }, { "epoch": 1.54, "learning_rate": 6.2029872233219365e-06, "loss": 0.6158, "step": 4277 }, { "epoch": 1.54, "learning_rate": 6.202087457261112e-06, "loss": 0.5569, "step": 4278 }, { "epoch": 1.54, "learning_rate": 6.201187691200289e-06, "loss": 0.6344, "step": 4279 }, { "epoch": 1.54, "learning_rate": 6.200287925139465e-06, "loss": 0.6595, "step": 4280 }, { "epoch": 1.54, "learning_rate": 6.19938815907864e-06, "loss": 0.5817, "step": 4281 }, { "epoch": 1.54, "learning_rate": 6.198488393017816e-06, "loss": 0.5826, "step": 4282 }, { "epoch": 1.54, "learning_rate": 6.197588626956992e-06, "loss": 0.6045, "step": 4283 }, { "epoch": 1.54, "learning_rate": 6.196688860896167e-06, "loss": 0.6053, "step": 4284 }, { "epoch": 1.54, "learning_rate": 6.1957890948353435e-06, "loss": 0.6271, "step": 4285 }, { "epoch": 1.54, "learning_rate": 6.194889328774519e-06, "loss": 0.6356, "step": 4286 }, { "epoch": 1.54, "learning_rate": 6.193989562713695e-06, "loss": 0.6172, "step": 4287 }, { "epoch": 1.54, "learning_rate": 6.19308979665287e-06, "loss": 0.64, "step": 4288 }, { "epoch": 1.54, "learning_rate": 6.192190030592047e-06, "loss": 0.5023, "step": 4289 }, { "epoch": 1.54, "learning_rate": 6.191290264531222e-06, "loss": 0.534, "step": 4290 }, { "epoch": 1.54, "learning_rate": 6.190390498470398e-06, "loss": 0.533, "step": 4291 }, { "epoch": 1.54, "learning_rate": 6.189490732409574e-06, "loss": 0.6024, "step": 4292 }, { "epoch": 1.54, "learning_rate": 6.18859096634875e-06, "loss": 0.5931, "step": 4293 }, { "epoch": 1.54, "learning_rate": 6.1876912002879255e-06, "loss": 0.6653, "step": 4294 }, { "epoch": 1.54, "learning_rate": 6.186791434227102e-06, "loss": 0.5717, "step": 4295 }, { "epoch": 1.54, "learning_rate": 6.185891668166277e-06, "loss": 0.6814, "step": 4296 }, { "epoch": 1.54, "learning_rate": 6.184991902105453e-06, "loss": 0.6015, "step": 4297 }, { "epoch": 1.54, "learning_rate": 6.184092136044629e-06, "loss": 0.6425, "step": 4298 }, { "epoch": 1.54, "learning_rate": 6.183192369983805e-06, "loss": 0.5965, "step": 4299 }, { "epoch": 1.54, "learning_rate": 6.18229260392298e-06, "loss": 0.5875, "step": 4300 }, { "epoch": 1.54, "learning_rate": 6.181392837862157e-06, "loss": 0.5597, "step": 4301 }, { "epoch": 1.54, "learning_rate": 6.1804930718013325e-06, "loss": 0.658, "step": 4302 }, { "epoch": 1.54, "learning_rate": 6.179593305740507e-06, "loss": 0.6928, "step": 4303 }, { "epoch": 1.54, "learning_rate": 6.178693539679684e-06, "loss": 0.675, "step": 4304 }, { "epoch": 1.55, "learning_rate": 6.17779377361886e-06, "loss": 0.692, "step": 4305 }, { "epoch": 1.55, "learning_rate": 6.1768940075580355e-06, "loss": 0.5994, "step": 4306 }, { "epoch": 1.55, "learning_rate": 6.1759942414972105e-06, "loss": 0.6311, "step": 4307 }, { "epoch": 1.55, "learning_rate": 6.175094475436387e-06, "loss": 0.5679, "step": 4308 }, { "epoch": 1.55, "learning_rate": 6.174194709375563e-06, "loss": 0.6353, "step": 4309 }, { "epoch": 1.55, "learning_rate": 6.173294943314739e-06, "loss": 0.6544, "step": 4310 }, { "epoch": 1.55, "learning_rate": 6.172395177253914e-06, "loss": 0.5655, "step": 4311 }, { "epoch": 1.55, "learning_rate": 6.17149541119309e-06, "loss": 0.5745, "step": 4312 }, { "epoch": 1.55, "learning_rate": 6.170595645132266e-06, "loss": 0.6292, "step": 4313 }, { "epoch": 1.55, "learning_rate": 6.1696958790714425e-06, "loss": 0.7601, "step": 4314 }, { "epoch": 1.55, "learning_rate": 6.1687961130106175e-06, "loss": 0.6237, "step": 4315 }, { "epoch": 1.55, "learning_rate": 6.167896346949793e-06, "loss": 0.6792, "step": 4316 }, { "epoch": 1.55, "learning_rate": 6.16699658088897e-06, "loss": 0.6939, "step": 4317 }, { "epoch": 1.55, "learning_rate": 6.166096814828146e-06, "loss": 0.6169, "step": 4318 }, { "epoch": 1.55, "learning_rate": 6.1651970487673205e-06, "loss": 0.5813, "step": 4319 }, { "epoch": 1.55, "learning_rate": 6.164297282706497e-06, "loss": 0.6747, "step": 4320 }, { "epoch": 1.55, "learning_rate": 6.163397516645673e-06, "loss": 0.6718, "step": 4321 }, { "epoch": 1.55, "learning_rate": 6.162497750584848e-06, "loss": 0.6145, "step": 4322 }, { "epoch": 1.55, "learning_rate": 6.1615979845240245e-06, "loss": 0.6724, "step": 4323 }, { "epoch": 1.55, "learning_rate": 6.1606982184632e-06, "loss": 0.5696, "step": 4324 }, { "epoch": 1.55, "learning_rate": 6.159798452402376e-06, "loss": 0.6392, "step": 4325 }, { "epoch": 1.55, "learning_rate": 6.158898686341551e-06, "loss": 0.6628, "step": 4326 }, { "epoch": 1.55, "learning_rate": 6.1579989202807275e-06, "loss": 0.5812, "step": 4327 }, { "epoch": 1.55, "learning_rate": 6.157099154219903e-06, "loss": 0.5984, "step": 4328 }, { "epoch": 1.55, "learning_rate": 6.156199388159079e-06, "loss": 0.6791, "step": 4329 }, { "epoch": 1.55, "learning_rate": 6.155299622098256e-06, "loss": 0.6704, "step": 4330 }, { "epoch": 1.55, "learning_rate": 6.154399856037431e-06, "loss": 0.6394, "step": 4331 }, { "epoch": 1.55, "learning_rate": 6.153500089976606e-06, "loss": 0.6017, "step": 4332 }, { "epoch": 1.56, "learning_rate": 6.152600323915783e-06, "loss": 0.5489, "step": 4333 }, { "epoch": 1.56, "learning_rate": 6.151700557854958e-06, "loss": 0.7179, "step": 4334 }, { "epoch": 1.56, "learning_rate": 6.150800791794134e-06, "loss": 0.5856, "step": 4335 }, { "epoch": 1.56, "learning_rate": 6.14990102573331e-06, "loss": 0.6359, "step": 4336 }, { "epoch": 1.56, "learning_rate": 6.149001259672486e-06, "loss": 0.6267, "step": 4337 }, { "epoch": 1.56, "learning_rate": 6.148101493611661e-06, "loss": 0.6179, "step": 4338 }, { "epoch": 1.56, "learning_rate": 6.147201727550838e-06, "loss": 0.5872, "step": 4339 }, { "epoch": 1.56, "learning_rate": 6.146301961490013e-06, "loss": 0.6172, "step": 4340 }, { "epoch": 1.56, "learning_rate": 6.145402195429188e-06, "loss": 0.4945, "step": 4341 }, { "epoch": 1.56, "learning_rate": 6.144502429368365e-06, "loss": 0.6594, "step": 4342 }, { "epoch": 1.56, "learning_rate": 6.143602663307541e-06, "loss": 0.5975, "step": 4343 }, { "epoch": 1.56, "learning_rate": 6.1427028972467164e-06, "loss": 0.6175, "step": 4344 }, { "epoch": 1.56, "learning_rate": 6.141803131185891e-06, "loss": 0.6781, "step": 4345 }, { "epoch": 1.56, "learning_rate": 6.140903365125068e-06, "loss": 0.6557, "step": 4346 }, { "epoch": 1.56, "learning_rate": 6.140003599064244e-06, "loss": 0.7138, "step": 4347 }, { "epoch": 1.56, "learning_rate": 6.1391038330034195e-06, "loss": 0.6716, "step": 4348 }, { "epoch": 1.56, "learning_rate": 6.138204066942596e-06, "loss": 0.6389, "step": 4349 }, { "epoch": 1.56, "learning_rate": 6.137304300881771e-06, "loss": 0.5934, "step": 4350 }, { "epoch": 1.56, "learning_rate": 6.136404534820947e-06, "loss": 0.6999, "step": 4351 }, { "epoch": 1.56, "learning_rate": 6.1355047687601234e-06, "loss": 0.6432, "step": 4352 }, { "epoch": 1.56, "learning_rate": 6.134605002699298e-06, "loss": 0.58, "step": 4353 }, { "epoch": 1.56, "learning_rate": 6.133705236638474e-06, "loss": 0.5368, "step": 4354 }, { "epoch": 1.56, "learning_rate": 6.132805470577651e-06, "loss": 0.5839, "step": 4355 }, { "epoch": 1.56, "learning_rate": 6.1319057045168265e-06, "loss": 0.6447, "step": 4356 }, { "epoch": 1.56, "learning_rate": 6.1310059384560014e-06, "loss": 0.7112, "step": 4357 }, { "epoch": 1.56, "learning_rate": 6.130106172395178e-06, "loss": 0.7189, "step": 4358 }, { "epoch": 1.56, "learning_rate": 6.129206406334354e-06, "loss": 0.6745, "step": 4359 }, { "epoch": 1.56, "learning_rate": 6.12830664027353e-06, "loss": 0.5974, "step": 4360 }, { "epoch": 1.57, "learning_rate": 6.127406874212705e-06, "loss": 0.6356, "step": 4361 }, { "epoch": 1.57, "learning_rate": 6.126507108151881e-06, "loss": 0.5422, "step": 4362 }, { "epoch": 1.57, "learning_rate": 6.125607342091057e-06, "loss": 0.6099, "step": 4363 }, { "epoch": 1.57, "learning_rate": 6.124707576030232e-06, "loss": 0.6015, "step": 4364 }, { "epoch": 1.57, "learning_rate": 6.1238078099694084e-06, "loss": 0.6257, "step": 4365 }, { "epoch": 1.57, "learning_rate": 6.122908043908584e-06, "loss": 0.5803, "step": 4366 }, { "epoch": 1.57, "learning_rate": 6.12200827784776e-06, "loss": 0.5775, "step": 4367 }, { "epoch": 1.57, "learning_rate": 6.121108511786937e-06, "loss": 0.5615, "step": 4368 }, { "epoch": 1.57, "learning_rate": 6.1202087457261115e-06, "loss": 0.5507, "step": 4369 }, { "epoch": 1.57, "learning_rate": 6.119308979665287e-06, "loss": 0.5366, "step": 4370 }, { "epoch": 1.57, "learning_rate": 6.118409213604464e-06, "loss": 0.6203, "step": 4371 }, { "epoch": 1.57, "learning_rate": 6.117509447543639e-06, "loss": 0.5708, "step": 4372 }, { "epoch": 1.57, "learning_rate": 6.116609681482815e-06, "loss": 0.6202, "step": 4373 }, { "epoch": 1.57, "learning_rate": 6.115709915421991e-06, "loss": 0.6264, "step": 4374 }, { "epoch": 1.57, "learning_rate": 6.114810149361167e-06, "loss": 0.5739, "step": 4375 }, { "epoch": 1.57, "learning_rate": 6.113910383300342e-06, "loss": 0.5771, "step": 4376 }, { "epoch": 1.57, "learning_rate": 6.1130106172395185e-06, "loss": 0.5483, "step": 4377 }, { "epoch": 1.57, "learning_rate": 6.112110851178694e-06, "loss": 0.5848, "step": 4378 }, { "epoch": 1.57, "learning_rate": 6.11121108511787e-06, "loss": 0.6947, "step": 4379 }, { "epoch": 1.57, "learning_rate": 6.110311319057046e-06, "loss": 0.6786, "step": 4380 }, { "epoch": 1.57, "learning_rate": 6.109411552996222e-06, "loss": 0.5792, "step": 4381 }, { "epoch": 1.57, "learning_rate": 6.108511786935397e-06, "loss": 0.7373, "step": 4382 }, { "epoch": 1.57, "learning_rate": 6.107612020874572e-06, "loss": 0.6733, "step": 4383 }, { "epoch": 1.57, "learning_rate": 6.106712254813749e-06, "loss": 0.5661, "step": 4384 }, { "epoch": 1.57, "learning_rate": 6.105812488752925e-06, "loss": 0.5908, "step": 4385 }, { "epoch": 1.57, "learning_rate": 6.1049127226921004e-06, "loss": 0.6376, "step": 4386 }, { "epoch": 1.57, "learning_rate": 6.104012956631277e-06, "loss": 0.6832, "step": 4387 }, { "epoch": 1.58, "learning_rate": 6.103113190570452e-06, "loss": 0.5831, "step": 4388 }, { "epoch": 1.58, "learning_rate": 6.102213424509628e-06, "loss": 0.6103, "step": 4389 }, { "epoch": 1.58, "learning_rate": 6.101313658448804e-06, "loss": 0.6978, "step": 4390 }, { "epoch": 1.58, "learning_rate": 6.100413892387979e-06, "loss": 0.725, "step": 4391 }, { "epoch": 1.58, "learning_rate": 6.099514126327155e-06, "loss": 0.6208, "step": 4392 }, { "epoch": 1.58, "learning_rate": 6.098614360266332e-06, "loss": 0.7092, "step": 4393 }, { "epoch": 1.58, "learning_rate": 6.0977145942055074e-06, "loss": 0.6407, "step": 4394 }, { "epoch": 1.58, "learning_rate": 6.096814828144682e-06, "loss": 0.6215, "step": 4395 }, { "epoch": 1.58, "learning_rate": 6.095915062083859e-06, "loss": 0.6695, "step": 4396 }, { "epoch": 1.58, "learning_rate": 6.095015296023035e-06, "loss": 0.4989, "step": 4397 }, { "epoch": 1.58, "learning_rate": 6.0941155299622105e-06, "loss": 0.6559, "step": 4398 }, { "epoch": 1.58, "learning_rate": 6.093215763901385e-06, "loss": 0.5648, "step": 4399 }, { "epoch": 1.58, "learning_rate": 6.092315997840562e-06, "loss": 0.4881, "step": 4400 }, { "epoch": 1.58, "learning_rate": 6.091416231779738e-06, "loss": 0.5613, "step": 4401 }, { "epoch": 1.58, "learning_rate": 6.090516465718913e-06, "loss": 0.5985, "step": 4402 }, { "epoch": 1.58, "learning_rate": 6.089616699658089e-06, "loss": 0.5283, "step": 4403 }, { "epoch": 1.58, "learning_rate": 6.088716933597265e-06, "loss": 0.6261, "step": 4404 }, { "epoch": 1.58, "learning_rate": 6.087817167536441e-06, "loss": 0.6781, "step": 4405 }, { "epoch": 1.58, "learning_rate": 6.0869174014756175e-06, "loss": 0.5348, "step": 4406 }, { "epoch": 1.58, "learning_rate": 6.086017635414792e-06, "loss": 0.6483, "step": 4407 }, { "epoch": 1.58, "learning_rate": 6.085117869353968e-06, "loss": 0.6438, "step": 4408 }, { "epoch": 1.58, "learning_rate": 6.084218103293145e-06, "loss": 0.5382, "step": 4409 }, { "epoch": 1.58, "learning_rate": 6.08331833723232e-06, "loss": 0.6263, "step": 4410 }, { "epoch": 1.58, "learning_rate": 6.0824185711714955e-06, "loss": 0.5964, "step": 4411 }, { "epoch": 1.58, "learning_rate": 6.081518805110672e-06, "loss": 0.5644, "step": 4412 }, { "epoch": 1.58, "learning_rate": 6.080619039049848e-06, "loss": 0.5214, "step": 4413 }, { "epoch": 1.58, "learning_rate": 6.079719272989023e-06, "loss": 0.6379, "step": 4414 }, { "epoch": 1.58, "learning_rate": 6.078819506928199e-06, "loss": 0.6906, "step": 4415 }, { "epoch": 1.59, "learning_rate": 6.077919740867375e-06, "loss": 0.4956, "step": 4416 }, { "epoch": 1.59, "learning_rate": 6.077019974806551e-06, "loss": 0.6299, "step": 4417 }, { "epoch": 1.59, "learning_rate": 6.076120208745726e-06, "loss": 0.7127, "step": 4418 }, { "epoch": 1.59, "learning_rate": 6.0752204426849025e-06, "loss": 0.661, "step": 4419 }, { "epoch": 1.59, "learning_rate": 6.074320676624078e-06, "loss": 0.6218, "step": 4420 }, { "epoch": 1.59, "learning_rate": 6.073420910563253e-06, "loss": 0.6342, "step": 4421 }, { "epoch": 1.59, "learning_rate": 6.07252114450243e-06, "loss": 0.6037, "step": 4422 }, { "epoch": 1.59, "learning_rate": 6.0716213784416056e-06, "loss": 0.6378, "step": 4423 }, { "epoch": 1.59, "learning_rate": 6.070721612380781e-06, "loss": 0.6923, "step": 4424 }, { "epoch": 1.59, "learning_rate": 6.069821846319958e-06, "loss": 0.7549, "step": 4425 }, { "epoch": 1.59, "learning_rate": 6.068922080259133e-06, "loss": 0.6888, "step": 4426 }, { "epoch": 1.59, "learning_rate": 6.068022314198309e-06, "loss": 0.6696, "step": 4427 }, { "epoch": 1.59, "learning_rate": 6.067122548137485e-06, "loss": 0.6037, "step": 4428 }, { "epoch": 1.59, "learning_rate": 6.06622278207666e-06, "loss": 0.7077, "step": 4429 }, { "epoch": 1.59, "learning_rate": 6.065323016015836e-06, "loss": 0.6317, "step": 4430 }, { "epoch": 1.59, "learning_rate": 6.0644232499550126e-06, "loss": 0.5809, "step": 4431 }, { "epoch": 1.59, "learning_rate": 6.063523483894188e-06, "loss": 0.6575, "step": 4432 }, { "epoch": 1.59, "learning_rate": 6.062623717833363e-06, "loss": 0.6253, "step": 4433 }, { "epoch": 1.59, "learning_rate": 6.06172395177254e-06, "loss": 0.6769, "step": 4434 }, { "epoch": 1.59, "learning_rate": 6.060824185711716e-06, "loss": 0.724, "step": 4435 }, { "epoch": 1.59, "learning_rate": 6.059924419650891e-06, "loss": 0.7187, "step": 4436 }, { "epoch": 1.59, "learning_rate": 6.059024653590066e-06, "loss": 0.6075, "step": 4437 }, { "epoch": 1.59, "learning_rate": 6.058124887529243e-06, "loss": 0.5703, "step": 4438 }, { "epoch": 1.59, "learning_rate": 6.057225121468419e-06, "loss": 0.5985, "step": 4439 }, { "epoch": 1.59, "learning_rate": 6.056325355407594e-06, "loss": 0.5641, "step": 4440 }, { "epoch": 1.59, "learning_rate": 6.05542558934677e-06, "loss": 0.643, "step": 4441 }, { "epoch": 1.59, "learning_rate": 6.054525823285946e-06, "loss": 0.6427, "step": 4442 }, { "epoch": 1.59, "learning_rate": 6.053626057225122e-06, "loss": 0.6102, "step": 4443 }, { "epoch": 1.6, "learning_rate": 6.052726291164298e-06, "loss": 0.5742, "step": 4444 }, { "epoch": 1.6, "learning_rate": 6.051826525103473e-06, "loss": 0.6614, "step": 4445 }, { "epoch": 1.6, "learning_rate": 6.050926759042649e-06, "loss": 0.6288, "step": 4446 }, { "epoch": 1.6, "learning_rate": 6.050026992981826e-06, "loss": 0.5826, "step": 4447 }, { "epoch": 1.6, "learning_rate": 6.0491272269210015e-06, "loss": 0.6791, "step": 4448 }, { "epoch": 1.6, "learning_rate": 6.048227460860176e-06, "loss": 0.5536, "step": 4449 }, { "epoch": 1.6, "learning_rate": 6.047327694799353e-06, "loss": 0.5383, "step": 4450 }, { "epoch": 1.6, "learning_rate": 6.046427928738529e-06, "loss": 0.591, "step": 4451 }, { "epoch": 1.6, "learning_rate": 6.045528162677704e-06, "loss": 0.7289, "step": 4452 }, { "epoch": 1.6, "learning_rate": 6.04462839661688e-06, "loss": 0.6666, "step": 4453 }, { "epoch": 1.6, "learning_rate": 6.043728630556056e-06, "loss": 0.603, "step": 4454 }, { "epoch": 1.6, "learning_rate": 6.042828864495232e-06, "loss": 0.6335, "step": 4455 }, { "epoch": 1.6, "learning_rate": 6.041929098434407e-06, "loss": 0.6701, "step": 4456 }, { "epoch": 1.6, "learning_rate": 6.041029332373583e-06, "loss": 0.625, "step": 4457 }, { "epoch": 1.6, "learning_rate": 6.040129566312759e-06, "loss": 0.6719, "step": 4458 }, { "epoch": 1.6, "learning_rate": 6.039229800251934e-06, "loss": 0.627, "step": 4459 }, { "epoch": 1.6, "learning_rate": 6.038330034191111e-06, "loss": 0.6241, "step": 4460 }, { "epoch": 1.6, "learning_rate": 6.0374302681302865e-06, "loss": 0.6199, "step": 4461 }, { "epoch": 1.6, "learning_rate": 6.036530502069462e-06, "loss": 0.574, "step": 4462 }, { "epoch": 1.6, "learning_rate": 6.035630736008639e-06, "loss": 0.6638, "step": 4463 }, { "epoch": 1.6, "learning_rate": 6.034730969947814e-06, "loss": 0.7004, "step": 4464 }, { "epoch": 1.6, "learning_rate": 6.0338312038869896e-06, "loss": 0.6339, "step": 4465 }, { "epoch": 1.6, "learning_rate": 6.032931437826166e-06, "loss": 0.609, "step": 4466 }, { "epoch": 1.6, "learning_rate": 6.032031671765342e-06, "loss": 0.7201, "step": 4467 }, { "epoch": 1.6, "learning_rate": 6.031131905704517e-06, "loss": 0.6241, "step": 4468 }, { "epoch": 1.6, "learning_rate": 6.0302321396436935e-06, "loss": 0.5645, "step": 4469 }, { "epoch": 1.6, "learning_rate": 6.029332373582869e-06, "loss": 0.6093, "step": 4470 }, { "epoch": 1.6, "learning_rate": 6.028432607522044e-06, "loss": 0.5847, "step": 4471 }, { "epoch": 1.61, "learning_rate": 6.027532841461221e-06, "loss": 0.6648, "step": 4472 }, { "epoch": 1.61, "learning_rate": 6.0266330754003965e-06, "loss": 0.6553, "step": 4473 }, { "epoch": 1.61, "learning_rate": 6.025733309339572e-06, "loss": 0.547, "step": 4474 }, { "epoch": 1.61, "learning_rate": 6.024833543278747e-06, "loss": 0.5482, "step": 4475 }, { "epoch": 1.61, "learning_rate": 6.023933777217924e-06, "loss": 0.6301, "step": 4476 }, { "epoch": 1.61, "learning_rate": 6.0230340111571e-06, "loss": 0.6085, "step": 4477 }, { "epoch": 1.61, "learning_rate": 6.022134245096275e-06, "loss": 0.5841, "step": 4478 }, { "epoch": 1.61, "learning_rate": 6.021234479035451e-06, "loss": 0.681, "step": 4479 }, { "epoch": 1.61, "learning_rate": 6.020334712974627e-06, "loss": 0.6264, "step": 4480 }, { "epoch": 1.61, "learning_rate": 6.019434946913803e-06, "loss": 0.5907, "step": 4481 }, { "epoch": 1.61, "learning_rate": 6.018535180852979e-06, "loss": 0.6497, "step": 4482 }, { "epoch": 1.61, "learning_rate": 6.017635414792154e-06, "loss": 0.5968, "step": 4483 }, { "epoch": 1.61, "learning_rate": 6.01673564873133e-06, "loss": 0.6727, "step": 4484 }, { "epoch": 1.61, "learning_rate": 6.015835882670507e-06, "loss": 0.657, "step": 4485 }, { "epoch": 1.61, "learning_rate": 6.014936116609682e-06, "loss": 0.6199, "step": 4486 }, { "epoch": 1.61, "learning_rate": 6.014036350548857e-06, "loss": 0.599, "step": 4487 }, { "epoch": 1.61, "learning_rate": 6.013136584488034e-06, "loss": 0.5816, "step": 4488 }, { "epoch": 1.61, "learning_rate": 6.01223681842721e-06, "loss": 0.581, "step": 4489 }, { "epoch": 1.61, "learning_rate": 6.011337052366385e-06, "loss": 0.6925, "step": 4490 }, { "epoch": 1.61, "learning_rate": 6.010437286305561e-06, "loss": 0.5444, "step": 4491 }, { "epoch": 1.61, "learning_rate": 6.009537520244737e-06, "loss": 0.5777, "step": 4492 }, { "epoch": 1.61, "learning_rate": 6.008637754183913e-06, "loss": 0.5843, "step": 4493 }, { "epoch": 1.61, "learning_rate": 6.007737988123088e-06, "loss": 0.5994, "step": 4494 }, { "epoch": 1.61, "learning_rate": 6.006838222062264e-06, "loss": 0.6142, "step": 4495 }, { "epoch": 1.61, "learning_rate": 6.00593845600144e-06, "loss": 0.6202, "step": 4496 }, { "epoch": 1.61, "learning_rate": 6.005038689940616e-06, "loss": 0.6461, "step": 4497 }, { "epoch": 1.61, "learning_rate": 6.004138923879792e-06, "loss": 0.6077, "step": 4498 }, { "epoch": 1.61, "learning_rate": 6.003239157818967e-06, "loss": 0.5911, "step": 4499 }, { "epoch": 1.62, "learning_rate": 6.002339391758143e-06, "loss": 0.642, "step": 4500 }, { "epoch": 1.62, "learning_rate": 6.00143962569732e-06, "loss": 0.5724, "step": 4501 }, { "epoch": 1.62, "learning_rate": 6.000539859636495e-06, "loss": 0.6318, "step": 4502 }, { "epoch": 1.62, "learning_rate": 5.9996400935756705e-06, "loss": 0.6692, "step": 4503 }, { "epoch": 1.62, "learning_rate": 5.998740327514847e-06, "loss": 0.7174, "step": 4504 }, { "epoch": 1.62, "learning_rate": 5.997840561454023e-06, "loss": 0.6558, "step": 4505 }, { "epoch": 1.62, "learning_rate": 5.996940795393198e-06, "loss": 0.7018, "step": 4506 }, { "epoch": 1.62, "learning_rate": 5.996041029332374e-06, "loss": 0.562, "step": 4507 }, { "epoch": 1.62, "learning_rate": 5.99514126327155e-06, "loss": 0.6485, "step": 4508 }, { "epoch": 1.62, "learning_rate": 5.994241497210725e-06, "loss": 0.6328, "step": 4509 }, { "epoch": 1.62, "learning_rate": 5.993341731149902e-06, "loss": 0.6069, "step": 4510 }, { "epoch": 1.62, "learning_rate": 5.9924419650890775e-06, "loss": 0.6958, "step": 4511 }, { "epoch": 1.62, "learning_rate": 5.991542199028253e-06, "loss": 0.6365, "step": 4512 }, { "epoch": 1.62, "learning_rate": 5.990642432967428e-06, "loss": 0.5409, "step": 4513 }, { "epoch": 1.62, "learning_rate": 5.989742666906605e-06, "loss": 0.6114, "step": 4514 }, { "epoch": 1.62, "learning_rate": 5.9888429008457805e-06, "loss": 0.6044, "step": 4515 }, { "epoch": 1.62, "learning_rate": 5.987943134784956e-06, "loss": 0.6785, "step": 4516 }, { "epoch": 1.62, "learning_rate": 5.987043368724133e-06, "loss": 0.6615, "step": 4517 }, { "epoch": 1.62, "learning_rate": 5.986143602663308e-06, "loss": 0.6384, "step": 4518 }, { "epoch": 1.62, "learning_rate": 5.985243836602484e-06, "loss": 0.6314, "step": 4519 }, { "epoch": 1.62, "learning_rate": 5.98434407054166e-06, "loss": 0.6018, "step": 4520 }, { "epoch": 1.62, "learning_rate": 5.983444304480835e-06, "loss": 0.5901, "step": 4521 }, { "epoch": 1.62, "learning_rate": 5.982544538420011e-06, "loss": 0.5435, "step": 4522 }, { "epoch": 1.62, "learning_rate": 5.9816447723591875e-06, "loss": 0.6192, "step": 4523 }, { "epoch": 1.62, "learning_rate": 5.980745006298363e-06, "loss": 0.5033, "step": 4524 }, { "epoch": 1.62, "learning_rate": 5.979845240237538e-06, "loss": 0.5381, "step": 4525 }, { "epoch": 1.62, "learning_rate": 5.978945474176715e-06, "loss": 0.5778, "step": 4526 }, { "epoch": 1.62, "learning_rate": 5.978045708115891e-06, "loss": 0.6131, "step": 4527 }, { "epoch": 1.63, "learning_rate": 5.9771459420550655e-06, "loss": 0.6014, "step": 4528 }, { "epoch": 1.63, "learning_rate": 5.976246175994242e-06, "loss": 0.6432, "step": 4529 }, { "epoch": 1.63, "learning_rate": 5.975346409933418e-06, "loss": 0.5894, "step": 4530 }, { "epoch": 1.63, "learning_rate": 5.974446643872594e-06, "loss": 0.7018, "step": 4531 }, { "epoch": 1.63, "learning_rate": 5.973546877811769e-06, "loss": 0.6202, "step": 4532 }, { "epoch": 1.63, "learning_rate": 5.972647111750945e-06, "loss": 0.6357, "step": 4533 }, { "epoch": 1.63, "learning_rate": 5.971747345690121e-06, "loss": 0.6381, "step": 4534 }, { "epoch": 1.63, "learning_rate": 5.970847579629297e-06, "loss": 0.6371, "step": 4535 }, { "epoch": 1.63, "learning_rate": 5.969947813568473e-06, "loss": 0.6557, "step": 4536 }, { "epoch": 1.63, "learning_rate": 5.969048047507648e-06, "loss": 0.6286, "step": 4537 }, { "epoch": 1.63, "learning_rate": 5.968148281446824e-06, "loss": 0.541, "step": 4538 }, { "epoch": 1.63, "learning_rate": 5.967248515386001e-06, "loss": 0.626, "step": 4539 }, { "epoch": 1.63, "learning_rate": 5.966348749325176e-06, "loss": 0.5941, "step": 4540 }, { "epoch": 1.63, "learning_rate": 5.965448983264351e-06, "loss": 0.6546, "step": 4541 }, { "epoch": 1.63, "learning_rate": 5.964549217203528e-06, "loss": 0.6693, "step": 4542 }, { "epoch": 1.63, "learning_rate": 5.963649451142704e-06, "loss": 0.6276, "step": 4543 }, { "epoch": 1.63, "learning_rate": 5.962749685081879e-06, "loss": 0.6152, "step": 4544 }, { "epoch": 1.63, "learning_rate": 5.961849919021055e-06, "loss": 0.5709, "step": 4545 }, { "epoch": 1.63, "learning_rate": 5.960950152960231e-06, "loss": 0.5772, "step": 4546 }, { "epoch": 1.63, "learning_rate": 5.960050386899407e-06, "loss": 0.631, "step": 4547 }, { "epoch": 1.63, "learning_rate": 5.959150620838583e-06, "loss": 0.6178, "step": 4548 }, { "epoch": 1.63, "learning_rate": 5.958250854777758e-06, "loss": 0.5723, "step": 4549 }, { "epoch": 1.63, "learning_rate": 5.957351088716934e-06, "loss": 0.6169, "step": 4550 }, { "epoch": 1.63, "learning_rate": 5.956451322656109e-06, "loss": 0.6894, "step": 4551 }, { "epoch": 1.63, "learning_rate": 5.955551556595286e-06, "loss": 0.6111, "step": 4552 }, { "epoch": 1.63, "learning_rate": 5.9546517905344614e-06, "loss": 0.5889, "step": 4553 }, { "epoch": 1.63, "learning_rate": 5.953752024473637e-06, "loss": 0.6453, "step": 4554 }, { "epoch": 1.63, "learning_rate": 5.952852258412814e-06, "loss": 0.6723, "step": 4555 }, { "epoch": 1.64, "learning_rate": 5.951952492351989e-06, "loss": 0.5337, "step": 4556 }, { "epoch": 1.64, "learning_rate": 5.9510527262911645e-06, "loss": 0.575, "step": 4557 }, { "epoch": 1.64, "learning_rate": 5.950152960230341e-06, "loss": 0.6197, "step": 4558 }, { "epoch": 1.64, "learning_rate": 5.949253194169516e-06, "loss": 0.6113, "step": 4559 }, { "epoch": 1.64, "learning_rate": 5.948353428108692e-06, "loss": 0.6776, "step": 4560 }, { "epoch": 1.64, "learning_rate": 5.9474536620478684e-06, "loss": 0.6157, "step": 4561 }, { "epoch": 1.64, "learning_rate": 5.946553895987044e-06, "loss": 0.5853, "step": 4562 }, { "epoch": 1.64, "learning_rate": 5.945654129926219e-06, "loss": 0.6588, "step": 4563 }, { "epoch": 1.64, "learning_rate": 5.944754363865396e-06, "loss": 0.6119, "step": 4564 }, { "epoch": 1.64, "learning_rate": 5.9438545978045715e-06, "loss": 0.6439, "step": 4565 }, { "epoch": 1.64, "learning_rate": 5.942954831743747e-06, "loss": 0.6222, "step": 4566 }, { "epoch": 1.64, "learning_rate": 5.942055065682923e-06, "loss": 0.6309, "step": 4567 }, { "epoch": 1.64, "learning_rate": 5.941155299622099e-06, "loss": 0.6341, "step": 4568 }, { "epoch": 1.64, "learning_rate": 5.940255533561275e-06, "loss": 0.6371, "step": 4569 }, { "epoch": 1.64, "learning_rate": 5.9393557675004495e-06, "loss": 0.5521, "step": 4570 }, { "epoch": 1.64, "learning_rate": 5.938456001439626e-06, "loss": 0.6092, "step": 4571 }, { "epoch": 1.64, "learning_rate": 5.937556235378802e-06, "loss": 0.66, "step": 4572 }, { "epoch": 1.64, "learning_rate": 5.936656469317978e-06, "loss": 0.7063, "step": 4573 }, { "epoch": 1.64, "learning_rate": 5.935756703257154e-06, "loss": 0.6539, "step": 4574 }, { "epoch": 1.64, "learning_rate": 5.934856937196329e-06, "loss": 0.5728, "step": 4575 }, { "epoch": 1.64, "learning_rate": 5.933957171135505e-06, "loss": 0.576, "step": 4576 }, { "epoch": 1.64, "learning_rate": 5.933057405074682e-06, "loss": 0.6419, "step": 4577 }, { "epoch": 1.64, "learning_rate": 5.9321576390138565e-06, "loss": 0.6274, "step": 4578 }, { "epoch": 1.64, "learning_rate": 5.931257872953032e-06, "loss": 0.5771, "step": 4579 }, { "epoch": 1.64, "learning_rate": 5.930358106892209e-06, "loss": 0.5792, "step": 4580 }, { "epoch": 1.64, "learning_rate": 5.929458340831385e-06, "loss": 0.5814, "step": 4581 }, { "epoch": 1.64, "learning_rate": 5.92855857477056e-06, "loss": 0.6766, "step": 4582 }, { "epoch": 1.65, "learning_rate": 5.927658808709736e-06, "loss": 0.5789, "step": 4583 }, { "epoch": 1.65, "learning_rate": 5.926759042648912e-06, "loss": 0.5788, "step": 4584 }, { "epoch": 1.65, "learning_rate": 5.925859276588088e-06, "loss": 0.5982, "step": 4585 }, { "epoch": 1.65, "learning_rate": 5.9249595105272635e-06, "loss": 0.6196, "step": 4586 }, { "epoch": 1.65, "learning_rate": 5.924059744466439e-06, "loss": 0.7473, "step": 4587 }, { "epoch": 1.65, "learning_rate": 5.923159978405615e-06, "loss": 0.6528, "step": 4588 }, { "epoch": 1.65, "learning_rate": 5.92226021234479e-06, "loss": 0.5644, "step": 4589 }, { "epoch": 1.65, "learning_rate": 5.921360446283967e-06, "loss": 0.5517, "step": 4590 }, { "epoch": 1.65, "learning_rate": 5.920460680223142e-06, "loss": 0.6387, "step": 4591 }, { "epoch": 1.65, "learning_rate": 5.919560914162318e-06, "loss": 0.6938, "step": 4592 }, { "epoch": 1.65, "learning_rate": 5.918661148101495e-06, "loss": 0.6546, "step": 4593 }, { "epoch": 1.65, "learning_rate": 5.91776138204067e-06, "loss": 0.5703, "step": 4594 }, { "epoch": 1.65, "learning_rate": 5.9168616159798454e-06, "loss": 0.673, "step": 4595 }, { "epoch": 1.65, "learning_rate": 5.915961849919022e-06, "loss": 0.6142, "step": 4596 }, { "epoch": 1.65, "learning_rate": 5.915062083858197e-06, "loss": 0.596, "step": 4597 }, { "epoch": 1.65, "learning_rate": 5.914162317797373e-06, "loss": 0.5594, "step": 4598 }, { "epoch": 1.65, "learning_rate": 5.913262551736549e-06, "loss": 0.6866, "step": 4599 }, { "epoch": 1.65, "learning_rate": 5.912362785675725e-06, "loss": 0.5267, "step": 4600 }, { "epoch": 1.65, "learning_rate": 5.9114630196149e-06, "loss": 0.6318, "step": 4601 }, { "epoch": 1.65, "learning_rate": 5.910563253554077e-06, "loss": 0.5319, "step": 4602 }, { "epoch": 1.65, "learning_rate": 5.909663487493252e-06, "loss": 0.6104, "step": 4603 }, { "epoch": 1.65, "learning_rate": 5.908763721432428e-06, "loss": 0.6102, "step": 4604 }, { "epoch": 1.65, "learning_rate": 5.907863955371605e-06, "loss": 0.6857, "step": 4605 }, { "epoch": 1.65, "learning_rate": 5.90696418931078e-06, "loss": 0.7655, "step": 4606 }, { "epoch": 1.65, "learning_rate": 5.9060644232499555e-06, "loss": 0.6366, "step": 4607 }, { "epoch": 1.65, "learning_rate": 5.90516465718913e-06, "loss": 0.5794, "step": 4608 }, { "epoch": 1.65, "learning_rate": 5.904264891128307e-06, "loss": 0.6425, "step": 4609 }, { "epoch": 1.65, "learning_rate": 5.903365125067483e-06, "loss": 0.5878, "step": 4610 }, { "epoch": 1.66, "learning_rate": 5.9024653590066586e-06, "loss": 0.5271, "step": 4611 }, { "epoch": 1.66, "learning_rate": 5.901565592945835e-06, "loss": 0.6287, "step": 4612 }, { "epoch": 1.66, "learning_rate": 5.90066582688501e-06, "loss": 0.6917, "step": 4613 }, { "epoch": 1.66, "learning_rate": 5.899766060824186e-06, "loss": 0.6148, "step": 4614 }, { "epoch": 1.66, "learning_rate": 5.8988662947633625e-06, "loss": 0.6935, "step": 4615 }, { "epoch": 1.66, "learning_rate": 5.897966528702537e-06, "loss": 0.6626, "step": 4616 }, { "epoch": 1.66, "learning_rate": 5.897066762641713e-06, "loss": 0.6944, "step": 4617 }, { "epoch": 1.66, "learning_rate": 5.89616699658089e-06, "loss": 0.588, "step": 4618 }, { "epoch": 1.66, "learning_rate": 5.8952672305200656e-06, "loss": 0.6508, "step": 4619 }, { "epoch": 1.66, "learning_rate": 5.8943674644592405e-06, "loss": 0.5775, "step": 4620 }, { "epoch": 1.66, "learning_rate": 5.893467698398417e-06, "loss": 0.6626, "step": 4621 }, { "epoch": 1.66, "learning_rate": 5.892567932337593e-06, "loss": 0.6109, "step": 4622 }, { "epoch": 1.66, "learning_rate": 5.891668166276769e-06, "loss": 0.6581, "step": 4623 }, { "epoch": 1.66, "learning_rate": 5.890768400215945e-06, "loss": 0.736, "step": 4624 }, { "epoch": 1.66, "learning_rate": 5.88986863415512e-06, "loss": 0.6489, "step": 4625 }, { "epoch": 1.66, "learning_rate": 5.888968868094296e-06, "loss": 0.554, "step": 4626 }, { "epoch": 1.66, "learning_rate": 5.888069102033471e-06, "loss": 0.5845, "step": 4627 }, { "epoch": 1.66, "learning_rate": 5.8871693359726475e-06, "loss": 0.6443, "step": 4628 }, { "epoch": 1.66, "learning_rate": 5.886269569911823e-06, "loss": 0.5608, "step": 4629 }, { "epoch": 1.66, "learning_rate": 5.885369803850999e-06, "loss": 0.6305, "step": 4630 }, { "epoch": 1.66, "learning_rate": 5.884470037790176e-06, "loss": 0.5697, "step": 4631 }, { "epoch": 1.66, "learning_rate": 5.8835702717293506e-06, "loss": 0.5898, "step": 4632 }, { "epoch": 1.66, "learning_rate": 5.882670505668526e-06, "loss": 0.5883, "step": 4633 }, { "epoch": 1.66, "learning_rate": 5.881770739607703e-06, "loss": 0.6412, "step": 4634 }, { "epoch": 1.66, "learning_rate": 5.880870973546879e-06, "loss": 0.6897, "step": 4635 }, { "epoch": 1.66, "learning_rate": 5.879971207486054e-06, "loss": 0.6269, "step": 4636 }, { "epoch": 1.66, "learning_rate": 5.87907144142523e-06, "loss": 0.7442, "step": 4637 }, { "epoch": 1.66, "learning_rate": 5.878171675364406e-06, "loss": 0.7042, "step": 4638 }, { "epoch": 1.67, "learning_rate": 5.877271909303581e-06, "loss": 0.6081, "step": 4639 }, { "epoch": 1.67, "learning_rate": 5.8763721432427576e-06, "loss": 0.628, "step": 4640 }, { "epoch": 1.67, "learning_rate": 5.875472377181933e-06, "loss": 0.6321, "step": 4641 }, { "epoch": 1.67, "learning_rate": 5.874572611121109e-06, "loss": 0.7041, "step": 4642 }, { "epoch": 1.67, "learning_rate": 5.873672845060286e-06, "loss": 0.551, "step": 4643 }, { "epoch": 1.67, "learning_rate": 5.872773078999461e-06, "loss": 0.636, "step": 4644 }, { "epoch": 1.67, "learning_rate": 5.871873312938636e-06, "loss": 0.575, "step": 4645 }, { "epoch": 1.67, "learning_rate": 5.870973546877811e-06, "loss": 0.5909, "step": 4646 }, { "epoch": 1.67, "learning_rate": 5.870073780816988e-06, "loss": 0.6602, "step": 4647 }, { "epoch": 1.67, "learning_rate": 5.869174014756164e-06, "loss": 0.7014, "step": 4648 }, { "epoch": 1.67, "learning_rate": 5.8682742486953395e-06, "loss": 0.6638, "step": 4649 }, { "epoch": 1.67, "learning_rate": 5.867374482634516e-06, "loss": 0.543, "step": 4650 }, { "epoch": 1.67, "learning_rate": 5.866474716573691e-06, "loss": 0.6212, "step": 4651 }, { "epoch": 1.67, "learning_rate": 5.865574950512867e-06, "loss": 0.7532, "step": 4652 }, { "epoch": 1.67, "learning_rate": 5.864675184452043e-06, "loss": 0.7603, "step": 4653 }, { "epoch": 1.67, "learning_rate": 5.863775418391219e-06, "loss": 0.7109, "step": 4654 }, { "epoch": 1.67, "learning_rate": 5.862875652330394e-06, "loss": 0.6813, "step": 4655 }, { "epoch": 1.67, "learning_rate": 5.861975886269571e-06, "loss": 0.552, "step": 4656 }, { "epoch": 1.67, "learning_rate": 5.8610761202087465e-06, "loss": 0.5873, "step": 4657 }, { "epoch": 1.67, "learning_rate": 5.860176354147921e-06, "loss": 0.5972, "step": 4658 }, { "epoch": 1.67, "learning_rate": 5.859276588087098e-06, "loss": 0.6278, "step": 4659 }, { "epoch": 1.67, "learning_rate": 5.858376822026274e-06, "loss": 0.6724, "step": 4660 }, { "epoch": 1.67, "learning_rate": 5.8574770559654496e-06, "loss": 0.605, "step": 4661 }, { "epoch": 1.67, "learning_rate": 5.856577289904626e-06, "loss": 0.6573, "step": 4662 }, { "epoch": 1.67, "learning_rate": 5.855677523843801e-06, "loss": 0.549, "step": 4663 }, { "epoch": 1.67, "learning_rate": 5.854777757782977e-06, "loss": 0.6346, "step": 4664 }, { "epoch": 1.67, "learning_rate": 5.853877991722153e-06, "loss": 0.595, "step": 4665 }, { "epoch": 1.67, "learning_rate": 5.852978225661328e-06, "loss": 0.545, "step": 4666 }, { "epoch": 1.68, "learning_rate": 5.852078459600504e-06, "loss": 0.632, "step": 4667 }, { "epoch": 1.68, "learning_rate": 5.85117869353968e-06, "loss": 0.5907, "step": 4668 }, { "epoch": 1.68, "learning_rate": 5.8502789274788565e-06, "loss": 0.5598, "step": 4669 }, { "epoch": 1.68, "learning_rate": 5.8493791614180315e-06, "loss": 0.7532, "step": 4670 }, { "epoch": 1.68, "learning_rate": 5.848479395357207e-06, "loss": 0.5403, "step": 4671 }, { "epoch": 1.68, "learning_rate": 5.847579629296384e-06, "loss": 0.6378, "step": 4672 }, { "epoch": 1.68, "learning_rate": 5.84667986323556e-06, "loss": 0.567, "step": 4673 }, { "epoch": 1.68, "learning_rate": 5.8457800971747345e-06, "loss": 0.5384, "step": 4674 }, { "epoch": 1.68, "learning_rate": 5.844880331113911e-06, "loss": 0.617, "step": 4675 }, { "epoch": 1.68, "learning_rate": 5.843980565053087e-06, "loss": 0.626, "step": 4676 }, { "epoch": 1.68, "learning_rate": 5.843080798992262e-06, "loss": 0.5806, "step": 4677 }, { "epoch": 1.68, "learning_rate": 5.8421810329314385e-06, "loss": 0.5747, "step": 4678 }, { "epoch": 1.68, "learning_rate": 5.841281266870614e-06, "loss": 0.5637, "step": 4679 }, { "epoch": 1.68, "learning_rate": 5.84038150080979e-06, "loss": 0.5858, "step": 4680 }, { "epoch": 1.68, "learning_rate": 5.839481734748967e-06, "loss": 0.6269, "step": 4681 }, { "epoch": 1.68, "learning_rate": 5.8385819686881415e-06, "loss": 0.6522, "step": 4682 }, { "epoch": 1.68, "learning_rate": 5.837682202627317e-06, "loss": 0.5586, "step": 4683 }, { "epoch": 1.68, "learning_rate": 5.836782436566493e-06, "loss": 0.5233, "step": 4684 }, { "epoch": 1.68, "learning_rate": 5.835882670505669e-06, "loss": 0.5514, "step": 4685 }, { "epoch": 1.68, "learning_rate": 5.834982904444845e-06, "loss": 0.5393, "step": 4686 }, { "epoch": 1.68, "learning_rate": 5.83408313838402e-06, "loss": 0.5903, "step": 4687 }, { "epoch": 1.68, "learning_rate": 5.833183372323197e-06, "loss": 0.5728, "step": 4688 }, { "epoch": 1.68, "learning_rate": 5.832283606262372e-06, "loss": 0.6261, "step": 4689 }, { "epoch": 1.68, "learning_rate": 5.831383840201548e-06, "loss": 0.6126, "step": 4690 }, { "epoch": 1.68, "learning_rate": 5.830484074140724e-06, "loss": 0.5681, "step": 4691 }, { "epoch": 1.68, "learning_rate": 5.8295843080799e-06, "loss": 0.7638, "step": 4692 }, { "epoch": 1.68, "learning_rate": 5.828684542019075e-06, "loss": 0.6022, "step": 4693 }, { "epoch": 1.68, "learning_rate": 5.827784775958252e-06, "loss": 0.6323, "step": 4694 }, { "epoch": 1.69, "learning_rate": 5.826885009897427e-06, "loss": 0.6602, "step": 4695 }, { "epoch": 1.69, "learning_rate": 5.825985243836602e-06, "loss": 0.7101, "step": 4696 }, { "epoch": 1.69, "learning_rate": 5.825085477775779e-06, "loss": 0.5977, "step": 4697 }, { "epoch": 1.69, "learning_rate": 5.824185711714955e-06, "loss": 0.5753, "step": 4698 }, { "epoch": 1.69, "learning_rate": 5.8232859456541305e-06, "loss": 0.6122, "step": 4699 }, { "epoch": 1.69, "learning_rate": 5.822386179593307e-06, "loss": 0.591, "step": 4700 }, { "epoch": 1.69, "learning_rate": 5.821486413532482e-06, "loss": 0.5538, "step": 4701 }, { "epoch": 1.69, "learning_rate": 5.820586647471658e-06, "loss": 0.5942, "step": 4702 }, { "epoch": 1.69, "learning_rate": 5.8196868814108335e-06, "loss": 0.5445, "step": 4703 }, { "epoch": 1.69, "learning_rate": 5.81878711535001e-06, "loss": 0.6301, "step": 4704 }, { "epoch": 1.69, "learning_rate": 5.817887349289185e-06, "loss": 0.5664, "step": 4705 }, { "epoch": 1.69, "learning_rate": 5.816987583228361e-06, "loss": 0.5485, "step": 4706 }, { "epoch": 1.69, "learning_rate": 5.8160878171675375e-06, "loss": 0.6142, "step": 4707 }, { "epoch": 1.69, "learning_rate": 5.815188051106712e-06, "loss": 0.6041, "step": 4708 }, { "epoch": 1.69, "learning_rate": 5.814288285045888e-06, "loss": 0.6465, "step": 4709 }, { "epoch": 1.69, "learning_rate": 5.813388518985065e-06, "loss": 0.7003, "step": 4710 }, { "epoch": 1.69, "learning_rate": 5.8124887529242405e-06, "loss": 0.6624, "step": 4711 }, { "epoch": 1.69, "learning_rate": 5.8115889868634155e-06, "loss": 0.5072, "step": 4712 }, { "epoch": 1.69, "learning_rate": 5.810689220802592e-06, "loss": 0.665, "step": 4713 }, { "epoch": 1.69, "learning_rate": 5.809789454741768e-06, "loss": 0.655, "step": 4714 }, { "epoch": 1.69, "learning_rate": 5.808889688680943e-06, "loss": 0.6584, "step": 4715 }, { "epoch": 1.69, "learning_rate": 5.807989922620119e-06, "loss": 0.6963, "step": 4716 }, { "epoch": 1.69, "learning_rate": 5.807090156559295e-06, "loss": 0.6597, "step": 4717 }, { "epoch": 1.69, "learning_rate": 5.806190390498471e-06, "loss": 0.6135, "step": 4718 }, { "epoch": 1.69, "learning_rate": 5.8052906244376475e-06, "loss": 0.5132, "step": 4719 }, { "epoch": 1.69, "learning_rate": 5.8043908583768225e-06, "loss": 0.6418, "step": 4720 }, { "epoch": 1.69, "learning_rate": 5.803491092315998e-06, "loss": 0.5699, "step": 4721 }, { "epoch": 1.69, "learning_rate": 5.802591326255174e-06, "loss": 0.6163, "step": 4722 }, { "epoch": 1.7, "learning_rate": 5.801691560194351e-06, "loss": 0.6238, "step": 4723 }, { "epoch": 1.7, "learning_rate": 5.8007917941335255e-06, "loss": 0.5926, "step": 4724 }, { "epoch": 1.7, "learning_rate": 5.799892028072701e-06, "loss": 0.5718, "step": 4725 }, { "epoch": 1.7, "learning_rate": 5.798992262011878e-06, "loss": 0.5715, "step": 4726 }, { "epoch": 1.7, "learning_rate": 5.798092495951053e-06, "loss": 0.6275, "step": 4727 }, { "epoch": 1.7, "learning_rate": 5.797192729890229e-06, "loss": 0.5671, "step": 4728 }, { "epoch": 1.7, "learning_rate": 5.796292963829405e-06, "loss": 0.6289, "step": 4729 }, { "epoch": 1.7, "learning_rate": 5.795393197768581e-06, "loss": 0.6862, "step": 4730 }, { "epoch": 1.7, "learning_rate": 5.794493431707756e-06, "loss": 0.7129, "step": 4731 }, { "epoch": 1.7, "learning_rate": 5.7935936656469325e-06, "loss": 0.6669, "step": 4732 }, { "epoch": 1.7, "learning_rate": 5.792693899586108e-06, "loss": 0.6334, "step": 4733 }, { "epoch": 1.7, "learning_rate": 5.791794133525283e-06, "loss": 0.6263, "step": 4734 }, { "epoch": 1.7, "learning_rate": 5.79089436746446e-06, "loss": 0.5797, "step": 4735 }, { "epoch": 1.7, "learning_rate": 5.789994601403636e-06, "loss": 0.6359, "step": 4736 }, { "epoch": 1.7, "learning_rate": 5.789094835342811e-06, "loss": 0.4793, "step": 4737 }, { "epoch": 1.7, "learning_rate": 5.788195069281988e-06, "loss": 0.6066, "step": 4738 }, { "epoch": 1.7, "learning_rate": 5.787295303221163e-06, "loss": 0.6398, "step": 4739 }, { "epoch": 1.7, "learning_rate": 5.786395537160339e-06, "loss": 0.5822, "step": 4740 }, { "epoch": 1.7, "learning_rate": 5.7854957710995144e-06, "loss": 0.532, "step": 4741 }, { "epoch": 1.7, "learning_rate": 5.784596005038691e-06, "loss": 0.6434, "step": 4742 }, { "epoch": 1.7, "learning_rate": 5.783696238977866e-06, "loss": 0.5941, "step": 4743 }, { "epoch": 1.7, "learning_rate": 5.782796472917042e-06, "loss": 0.5138, "step": 4744 }, { "epoch": 1.7, "learning_rate": 5.781896706856218e-06, "loss": 0.598, "step": 4745 }, { "epoch": 1.7, "learning_rate": 5.780996940795393e-06, "loss": 0.5835, "step": 4746 }, { "epoch": 1.7, "learning_rate": 5.780097174734569e-06, "loss": 0.6997, "step": 4747 }, { "epoch": 1.7, "learning_rate": 5.779197408673746e-06, "loss": 0.6508, "step": 4748 }, { "epoch": 1.7, "learning_rate": 5.7782976426129214e-06, "loss": 0.7377, "step": 4749 }, { "epoch": 1.7, "learning_rate": 5.777397876552096e-06, "loss": 0.6821, "step": 4750 }, { "epoch": 1.71, "learning_rate": 5.776498110491273e-06, "loss": 0.6801, "step": 4751 }, { "epoch": 1.71, "learning_rate": 5.775598344430449e-06, "loss": 0.5903, "step": 4752 }, { "epoch": 1.71, "learning_rate": 5.7746985783696245e-06, "loss": 0.5841, "step": 4753 }, { "epoch": 1.71, "learning_rate": 5.7737988123088e-06, "loss": 0.5513, "step": 4754 }, { "epoch": 1.71, "learning_rate": 5.772899046247976e-06, "loss": 0.5819, "step": 4755 }, { "epoch": 1.71, "learning_rate": 5.771999280187152e-06, "loss": 0.5166, "step": 4756 }, { "epoch": 1.71, "learning_rate": 5.771099514126327e-06, "loss": 0.6884, "step": 4757 }, { "epoch": 1.71, "learning_rate": 5.770199748065503e-06, "loss": 0.5695, "step": 4758 }, { "epoch": 1.71, "learning_rate": 5.769299982004679e-06, "loss": 0.7162, "step": 4759 }, { "epoch": 1.71, "learning_rate": 5.768400215943855e-06, "loss": 0.5671, "step": 4760 }, { "epoch": 1.71, "learning_rate": 5.7675004498830315e-06, "loss": 0.5893, "step": 4761 }, { "epoch": 1.71, "learning_rate": 5.7666006838222064e-06, "loss": 0.7097, "step": 4762 }, { "epoch": 1.71, "learning_rate": 5.765700917761382e-06, "loss": 0.5737, "step": 4763 }, { "epoch": 1.71, "learning_rate": 5.764801151700559e-06, "loss": 0.56, "step": 4764 }, { "epoch": 1.71, "learning_rate": 5.763901385639734e-06, "loss": 0.6141, "step": 4765 }, { "epoch": 1.71, "learning_rate": 5.7630016195789095e-06, "loss": 0.6339, "step": 4766 }, { "epoch": 1.71, "learning_rate": 5.762101853518086e-06, "loss": 0.6992, "step": 4767 }, { "epoch": 1.71, "learning_rate": 5.761202087457262e-06, "loss": 0.6645, "step": 4768 }, { "epoch": 1.71, "learning_rate": 5.760302321396437e-06, "loss": 0.6678, "step": 4769 }, { "epoch": 1.71, "learning_rate": 5.7594025553356134e-06, "loss": 0.6295, "step": 4770 }, { "epoch": 1.71, "learning_rate": 5.758502789274789e-06, "loss": 0.618, "step": 4771 }, { "epoch": 1.71, "learning_rate": 5.757603023213965e-06, "loss": 0.6616, "step": 4772 }, { "epoch": 1.71, "learning_rate": 5.756703257153141e-06, "loss": 0.5152, "step": 4773 }, { "epoch": 1.71, "learning_rate": 5.7558034910923165e-06, "loss": 0.5653, "step": 4774 }, { "epoch": 1.71, "learning_rate": 5.754903725031492e-06, "loss": 0.7314, "step": 4775 }, { "epoch": 1.71, "learning_rate": 5.754003958970667e-06, "loss": 0.6688, "step": 4776 }, { "epoch": 1.71, "learning_rate": 5.753104192909844e-06, "loss": 0.6967, "step": 4777 }, { "epoch": 1.72, "learning_rate": 5.75220442684902e-06, "loss": 0.595, "step": 4778 }, { "epoch": 1.72, "learning_rate": 5.751304660788195e-06, "loss": 0.6205, "step": 4779 }, { "epoch": 1.72, "learning_rate": 5.750404894727372e-06, "loss": 0.6229, "step": 4780 }, { "epoch": 1.72, "learning_rate": 5.749505128666547e-06, "loss": 0.6533, "step": 4781 }, { "epoch": 1.72, "learning_rate": 5.748605362605723e-06, "loss": 0.7089, "step": 4782 }, { "epoch": 1.72, "learning_rate": 5.747705596544899e-06, "loss": 0.5164, "step": 4783 }, { "epoch": 1.72, "learning_rate": 5.746805830484074e-06, "loss": 0.6997, "step": 4784 }, { "epoch": 1.72, "learning_rate": 5.74590606442325e-06, "loss": 0.5619, "step": 4785 }, { "epoch": 1.72, "learning_rate": 5.745006298362427e-06, "loss": 0.5536, "step": 4786 }, { "epoch": 1.72, "learning_rate": 5.744106532301602e-06, "loss": 0.6087, "step": 4787 }, { "epoch": 1.72, "learning_rate": 5.743206766240777e-06, "loss": 0.6194, "step": 4788 }, { "epoch": 1.72, "learning_rate": 5.742307000179954e-06, "loss": 0.5747, "step": 4789 }, { "epoch": 1.72, "learning_rate": 5.74140723411913e-06, "loss": 0.624, "step": 4790 }, { "epoch": 1.72, "learning_rate": 5.7405074680583054e-06, "loss": 0.7286, "step": 4791 }, { "epoch": 1.72, "learning_rate": 5.739607701997482e-06, "loss": 0.6602, "step": 4792 }, { "epoch": 1.72, "learning_rate": 5.738707935936657e-06, "loss": 0.5794, "step": 4793 }, { "epoch": 1.72, "learning_rate": 5.737808169875833e-06, "loss": 0.5545, "step": 4794 }, { "epoch": 1.72, "learning_rate": 5.736908403815008e-06, "loss": 0.5671, "step": 4795 }, { "epoch": 1.72, "learning_rate": 5.736008637754184e-06, "loss": 0.5772, "step": 4796 }, { "epoch": 1.72, "learning_rate": 5.73510887169336e-06, "loss": 0.6411, "step": 4797 }, { "epoch": 1.72, "learning_rate": 5.734209105632536e-06, "loss": 0.6115, "step": 4798 }, { "epoch": 1.72, "learning_rate": 5.733309339571712e-06, "loss": 0.7112, "step": 4799 }, { "epoch": 1.72, "learning_rate": 5.732409573510887e-06, "loss": 0.5505, "step": 4800 }, { "epoch": 1.72, "learning_rate": 5.731509807450063e-06, "loss": 0.6567, "step": 4801 }, { "epoch": 1.72, "learning_rate": 5.73061004138924e-06, "loss": 0.6019, "step": 4802 }, { "epoch": 1.72, "learning_rate": 5.729710275328415e-06, "loss": 0.5002, "step": 4803 }, { "epoch": 1.72, "learning_rate": 5.72881050926759e-06, "loss": 0.5654, "step": 4804 }, { "epoch": 1.72, "learning_rate": 5.727910743206767e-06, "loss": 0.5042, "step": 4805 }, { "epoch": 1.73, "learning_rate": 5.727010977145943e-06, "loss": 0.6041, "step": 4806 }, { "epoch": 1.73, "learning_rate": 5.726111211085118e-06, "loss": 0.6122, "step": 4807 }, { "epoch": 1.73, "learning_rate": 5.725211445024294e-06, "loss": 0.5622, "step": 4808 }, { "epoch": 1.73, "learning_rate": 5.72431167896347e-06, "loss": 0.6665, "step": 4809 }, { "epoch": 1.73, "learning_rate": 5.723411912902646e-06, "loss": 0.7846, "step": 4810 }, { "epoch": 1.73, "learning_rate": 5.7225121468418225e-06, "loss": 0.552, "step": 4811 }, { "epoch": 1.73, "learning_rate": 5.721612380780997e-06, "loss": 0.6151, "step": 4812 }, { "epoch": 1.73, "learning_rate": 5.720712614720173e-06, "loss": 0.5834, "step": 4813 }, { "epoch": 1.73, "learning_rate": 5.719812848659348e-06, "loss": 0.6049, "step": 4814 }, { "epoch": 1.73, "learning_rate": 5.718913082598525e-06, "loss": 0.5835, "step": 4815 }, { "epoch": 1.73, "learning_rate": 5.7180133165377005e-06, "loss": 0.6455, "step": 4816 }, { "epoch": 1.73, "learning_rate": 5.717113550476876e-06, "loss": 0.6005, "step": 4817 }, { "epoch": 1.73, "learning_rate": 5.716213784416053e-06, "loss": 0.5592, "step": 4818 }, { "epoch": 1.73, "learning_rate": 5.715314018355228e-06, "loss": 0.6178, "step": 4819 }, { "epoch": 1.73, "learning_rate": 5.7144142522944036e-06, "loss": 0.6582, "step": 4820 }, { "epoch": 1.73, "learning_rate": 5.71351448623358e-06, "loss": 0.6723, "step": 4821 }, { "epoch": 1.73, "learning_rate": 5.712614720172756e-06, "loss": 0.6367, "step": 4822 }, { "epoch": 1.73, "learning_rate": 5.711714954111931e-06, "loss": 0.6232, "step": 4823 }, { "epoch": 1.73, "learning_rate": 5.7108151880511075e-06, "loss": 0.6519, "step": 4824 }, { "epoch": 1.73, "learning_rate": 5.709915421990283e-06, "loss": 0.5441, "step": 4825 }, { "epoch": 1.73, "learning_rate": 5.709015655929458e-06, "loss": 0.5415, "step": 4826 }, { "epoch": 1.73, "learning_rate": 5.708115889868635e-06, "loss": 0.6897, "step": 4827 }, { "epoch": 1.73, "learning_rate": 5.7072161238078106e-06, "loss": 0.5806, "step": 4828 }, { "epoch": 1.73, "learning_rate": 5.706316357746986e-06, "loss": 0.6345, "step": 4829 }, { "epoch": 1.73, "learning_rate": 5.705416591686163e-06, "loss": 0.603, "step": 4830 }, { "epoch": 1.73, "learning_rate": 5.704516825625338e-06, "loss": 0.5717, "step": 4831 }, { "epoch": 1.73, "learning_rate": 5.703617059564514e-06, "loss": 0.7841, "step": 4832 }, { "epoch": 1.73, "learning_rate": 5.7027172935036886e-06, "loss": 0.5965, "step": 4833 }, { "epoch": 1.74, "learning_rate": 5.701817527442865e-06, "loss": 0.6639, "step": 4834 }, { "epoch": 1.74, "learning_rate": 5.700917761382041e-06, "loss": 0.5857, "step": 4835 }, { "epoch": 1.74, "learning_rate": 5.700017995321217e-06, "loss": 0.6391, "step": 4836 }, { "epoch": 1.74, "learning_rate": 5.699118229260393e-06, "loss": 0.6421, "step": 4837 }, { "epoch": 1.74, "learning_rate": 5.698218463199568e-06, "loss": 0.7203, "step": 4838 }, { "epoch": 1.74, "learning_rate": 5.697318697138744e-06, "loss": 0.6205, "step": 4839 }, { "epoch": 1.74, "learning_rate": 5.696418931077921e-06, "loss": 0.6405, "step": 4840 }, { "epoch": 1.74, "learning_rate": 5.695519165017096e-06, "loss": 0.5422, "step": 4841 }, { "epoch": 1.74, "learning_rate": 5.694619398956271e-06, "loss": 0.5708, "step": 4842 }, { "epoch": 1.74, "learning_rate": 5.693719632895448e-06, "loss": 0.5277, "step": 4843 }, { "epoch": 1.74, "learning_rate": 5.692819866834624e-06, "loss": 0.6808, "step": 4844 }, { "epoch": 1.74, "learning_rate": 5.691920100773799e-06, "loss": 0.659, "step": 4845 }, { "epoch": 1.74, "learning_rate": 5.691020334712975e-06, "loss": 0.6261, "step": 4846 }, { "epoch": 1.74, "learning_rate": 5.690120568652151e-06, "loss": 0.5667, "step": 4847 }, { "epoch": 1.74, "learning_rate": 5.689220802591327e-06, "loss": 0.5853, "step": 4848 }, { "epoch": 1.74, "learning_rate": 5.688321036530503e-06, "loss": 0.6774, "step": 4849 }, { "epoch": 1.74, "learning_rate": 5.687421270469678e-06, "loss": 0.574, "step": 4850 }, { "epoch": 1.74, "learning_rate": 5.686521504408854e-06, "loss": 0.5808, "step": 4851 }, { "epoch": 1.74, "learning_rate": 5.68562173834803e-06, "loss": 0.6482, "step": 4852 }, { "epoch": 1.74, "learning_rate": 5.684721972287206e-06, "loss": 0.6275, "step": 4853 }, { "epoch": 1.74, "learning_rate": 5.683822206226381e-06, "loss": 0.674, "step": 4854 }, { "epoch": 1.74, "learning_rate": 5.682922440165557e-06, "loss": 0.6726, "step": 4855 }, { "epoch": 1.74, "learning_rate": 5.682022674104734e-06, "loss": 0.5035, "step": 4856 }, { "epoch": 1.74, "learning_rate": 5.681122908043909e-06, "loss": 0.6941, "step": 4857 }, { "epoch": 1.74, "learning_rate": 5.6802231419830845e-06, "loss": 0.5759, "step": 4858 }, { "epoch": 1.74, "learning_rate": 5.679323375922261e-06, "loss": 0.5941, "step": 4859 }, { "epoch": 1.74, "learning_rate": 5.678423609861437e-06, "loss": 0.637, "step": 4860 }, { "epoch": 1.74, "learning_rate": 5.677523843800612e-06, "loss": 0.5652, "step": 4861 }, { "epoch": 1.75, "learning_rate": 5.676624077739788e-06, "loss": 0.6566, "step": 4862 }, { "epoch": 1.75, "learning_rate": 5.675724311678964e-06, "loss": 0.6161, "step": 4863 }, { "epoch": 1.75, "learning_rate": 5.674824545618139e-06, "loss": 0.6866, "step": 4864 }, { "epoch": 1.75, "learning_rate": 5.673924779557316e-06, "loss": 0.7351, "step": 4865 }, { "epoch": 1.75, "learning_rate": 5.6730250134964915e-06, "loss": 0.549, "step": 4866 }, { "epoch": 1.75, "learning_rate": 5.672125247435667e-06, "loss": 0.5541, "step": 4867 }, { "epoch": 1.75, "learning_rate": 5.671225481374844e-06, "loss": 0.6932, "step": 4868 }, { "epoch": 1.75, "learning_rate": 5.670325715314019e-06, "loss": 0.6277, "step": 4869 }, { "epoch": 1.75, "learning_rate": 5.6694259492531945e-06, "loss": 0.5139, "step": 4870 }, { "epoch": 1.75, "learning_rate": 5.66852618319237e-06, "loss": 0.6445, "step": 4871 }, { "epoch": 1.75, "learning_rate": 5.667626417131546e-06, "loss": 0.5573, "step": 4872 }, { "epoch": 1.75, "learning_rate": 5.666726651070722e-06, "loss": 0.719, "step": 4873 }, { "epoch": 1.75, "learning_rate": 5.665826885009898e-06, "loss": 0.6633, "step": 4874 }, { "epoch": 1.75, "learning_rate": 5.664927118949074e-06, "loss": 0.6135, "step": 4875 }, { "epoch": 1.75, "learning_rate": 5.664027352888249e-06, "loss": 0.5852, "step": 4876 }, { "epoch": 1.75, "learning_rate": 5.663127586827425e-06, "loss": 0.6108, "step": 4877 }, { "epoch": 1.75, "learning_rate": 5.6622278207666015e-06, "loss": 0.5486, "step": 4878 }, { "epoch": 1.75, "learning_rate": 5.661328054705777e-06, "loss": 0.7224, "step": 4879 }, { "epoch": 1.75, "learning_rate": 5.660428288644952e-06, "loss": 0.6473, "step": 4880 }, { "epoch": 1.75, "learning_rate": 5.659528522584129e-06, "loss": 0.7317, "step": 4881 }, { "epoch": 1.75, "learning_rate": 5.658628756523305e-06, "loss": 0.6123, "step": 4882 }, { "epoch": 1.75, "learning_rate": 5.6577289904624795e-06, "loss": 0.6263, "step": 4883 }, { "epoch": 1.75, "learning_rate": 5.656829224401656e-06, "loss": 0.5348, "step": 4884 }, { "epoch": 1.75, "learning_rate": 5.655929458340832e-06, "loss": 0.5574, "step": 4885 }, { "epoch": 1.75, "learning_rate": 5.655029692280008e-06, "loss": 0.5264, "step": 4886 }, { "epoch": 1.75, "learning_rate": 5.654129926219184e-06, "loss": 0.5818, "step": 4887 }, { "epoch": 1.75, "learning_rate": 5.653230160158359e-06, "loss": 0.6006, "step": 4888 }, { "epoch": 1.75, "learning_rate": 5.652330394097535e-06, "loss": 0.6576, "step": 4889 }, { "epoch": 1.76, "learning_rate": 5.651430628036711e-06, "loss": 0.5597, "step": 4890 }, { "epoch": 1.76, "learning_rate": 5.6505308619758865e-06, "loss": 0.6198, "step": 4891 }, { "epoch": 1.76, "learning_rate": 5.649631095915062e-06, "loss": 0.5652, "step": 4892 }, { "epoch": 1.76, "learning_rate": 5.648731329854238e-06, "loss": 0.6452, "step": 4893 }, { "epoch": 1.76, "learning_rate": 5.647831563793415e-06, "loss": 0.6122, "step": 4894 }, { "epoch": 1.76, "learning_rate": 5.64693179773259e-06, "loss": 0.6403, "step": 4895 }, { "epoch": 1.76, "learning_rate": 5.646032031671765e-06, "loss": 0.671, "step": 4896 }, { "epoch": 1.76, "learning_rate": 5.645132265610942e-06, "loss": 0.6113, "step": 4897 }, { "epoch": 1.76, "learning_rate": 5.644232499550118e-06, "loss": 0.6382, "step": 4898 }, { "epoch": 1.76, "learning_rate": 5.643332733489293e-06, "loss": 0.5744, "step": 4899 }, { "epoch": 1.76, "learning_rate": 5.642432967428469e-06, "loss": 0.6536, "step": 4900 }, { "epoch": 1.76, "learning_rate": 5.641533201367645e-06, "loss": 0.6857, "step": 4901 }, { "epoch": 1.76, "learning_rate": 5.64063343530682e-06, "loss": 0.5834, "step": 4902 }, { "epoch": 1.76, "learning_rate": 5.639733669245997e-06, "loss": 0.5708, "step": 4903 }, { "epoch": 1.76, "learning_rate": 5.638833903185172e-06, "loss": 0.6828, "step": 4904 }, { "epoch": 1.76, "learning_rate": 5.637934137124348e-06, "loss": 0.7439, "step": 4905 }, { "epoch": 1.76, "learning_rate": 5.637034371063525e-06, "loss": 0.697, "step": 4906 }, { "epoch": 1.76, "learning_rate": 5.6361346050027e-06, "loss": 0.6729, "step": 4907 }, { "epoch": 1.76, "learning_rate": 5.6352348389418755e-06, "loss": 0.6117, "step": 4908 }, { "epoch": 1.76, "learning_rate": 5.634335072881051e-06, "loss": 0.6811, "step": 4909 }, { "epoch": 1.76, "learning_rate": 5.633435306820228e-06, "loss": 0.5453, "step": 4910 }, { "epoch": 1.76, "learning_rate": 5.632535540759403e-06, "loss": 0.6714, "step": 4911 }, { "epoch": 1.76, "learning_rate": 5.6316357746985785e-06, "loss": 0.5519, "step": 4912 }, { "epoch": 1.76, "learning_rate": 5.630736008637755e-06, "loss": 0.6483, "step": 4913 }, { "epoch": 1.76, "learning_rate": 5.62983624257693e-06, "loss": 0.5886, "step": 4914 }, { "epoch": 1.76, "learning_rate": 5.628936476516106e-06, "loss": 0.575, "step": 4915 }, { "epoch": 1.76, "learning_rate": 5.6280367104552825e-06, "loss": 0.6123, "step": 4916 }, { "epoch": 1.76, "learning_rate": 5.627136944394458e-06, "loss": 0.6216, "step": 4917 }, { "epoch": 1.77, "learning_rate": 5.626237178333633e-06, "loss": 0.6829, "step": 4918 }, { "epoch": 1.77, "learning_rate": 5.62533741227281e-06, "loss": 0.5622, "step": 4919 }, { "epoch": 1.77, "learning_rate": 5.6244376462119855e-06, "loss": 0.6459, "step": 4920 }, { "epoch": 1.77, "learning_rate": 5.6235378801511604e-06, "loss": 0.6295, "step": 4921 }, { "epoch": 1.77, "learning_rate": 5.622638114090337e-06, "loss": 0.6409, "step": 4922 }, { "epoch": 1.77, "learning_rate": 5.621738348029513e-06, "loss": 0.6937, "step": 4923 }, { "epoch": 1.77, "learning_rate": 5.620838581968689e-06, "loss": 0.6685, "step": 4924 }, { "epoch": 1.77, "learning_rate": 5.619938815907865e-06, "loss": 0.6051, "step": 4925 }, { "epoch": 1.77, "learning_rate": 5.61903904984704e-06, "loss": 0.5379, "step": 4926 }, { "epoch": 1.77, "learning_rate": 5.618139283786216e-06, "loss": 0.6568, "step": 4927 }, { "epoch": 1.77, "learning_rate": 5.617239517725392e-06, "loss": 0.6522, "step": 4928 }, { "epoch": 1.77, "learning_rate": 5.616339751664568e-06, "loss": 0.6189, "step": 4929 }, { "epoch": 1.77, "learning_rate": 5.615439985603743e-06, "loss": 0.6565, "step": 4930 }, { "epoch": 1.77, "learning_rate": 5.614540219542919e-06, "loss": 0.7038, "step": 4931 }, { "epoch": 1.77, "learning_rate": 5.613640453482096e-06, "loss": 0.6101, "step": 4932 }, { "epoch": 1.77, "learning_rate": 5.6127406874212705e-06, "loss": 0.5481, "step": 4933 }, { "epoch": 1.77, "learning_rate": 5.611840921360446e-06, "loss": 0.6099, "step": 4934 }, { "epoch": 1.77, "learning_rate": 5.610941155299623e-06, "loss": 0.6063, "step": 4935 }, { "epoch": 1.77, "learning_rate": 5.610041389238799e-06, "loss": 0.6756, "step": 4936 }, { "epoch": 1.77, "learning_rate": 5.609141623177974e-06, "loss": 0.6573, "step": 4937 }, { "epoch": 1.77, "learning_rate": 5.60824185711715e-06, "loss": 0.6524, "step": 4938 }, { "epoch": 1.77, "learning_rate": 5.607342091056326e-06, "loss": 0.5954, "step": 4939 }, { "epoch": 1.77, "learning_rate": 5.606442324995502e-06, "loss": 0.6347, "step": 4940 }, { "epoch": 1.77, "learning_rate": 5.6055425589346775e-06, "loss": 0.5914, "step": 4941 }, { "epoch": 1.77, "learning_rate": 5.604642792873853e-06, "loss": 0.6662, "step": 4942 }, { "epoch": 1.77, "learning_rate": 5.603743026813029e-06, "loss": 0.5633, "step": 4943 }, { "epoch": 1.77, "learning_rate": 5.602843260752206e-06, "loss": 0.6711, "step": 4944 }, { "epoch": 1.77, "learning_rate": 5.601943494691381e-06, "loss": 0.5298, "step": 4945 }, { "epoch": 1.78, "learning_rate": 5.601043728630556e-06, "loss": 0.6512, "step": 4946 }, { "epoch": 1.78, "learning_rate": 5.600143962569732e-06, "loss": 0.5963, "step": 4947 }, { "epoch": 1.78, "learning_rate": 5.599244196508909e-06, "loss": 0.6488, "step": 4948 }, { "epoch": 1.78, "learning_rate": 5.598344430448084e-06, "loss": 0.5788, "step": 4949 }, { "epoch": 1.78, "learning_rate": 5.5974446643872594e-06, "loss": 0.619, "step": 4950 }, { "epoch": 1.78, "learning_rate": 5.596544898326436e-06, "loss": 0.6379, "step": 4951 }, { "epoch": 1.78, "learning_rate": 5.595645132265611e-06, "loss": 0.5952, "step": 4952 }, { "epoch": 1.78, "learning_rate": 5.594745366204787e-06, "loss": 0.5842, "step": 4953 }, { "epoch": 1.78, "learning_rate": 5.593845600143963e-06, "loss": 0.6419, "step": 4954 }, { "epoch": 1.78, "learning_rate": 5.592945834083139e-06, "loss": 0.5957, "step": 4955 }, { "epoch": 1.78, "learning_rate": 5.592046068022314e-06, "loss": 0.5436, "step": 4956 }, { "epoch": 1.78, "learning_rate": 5.591146301961491e-06, "loss": 0.7001, "step": 4957 }, { "epoch": 1.78, "learning_rate": 5.5902465359006664e-06, "loss": 0.518, "step": 4958 }, { "epoch": 1.78, "learning_rate": 5.589346769839842e-06, "loss": 0.5957, "step": 4959 }, { "epoch": 1.78, "learning_rate": 5.588447003779018e-06, "loss": 0.6173, "step": 4960 }, { "epoch": 1.78, "learning_rate": 5.587547237718194e-06, "loss": 0.7066, "step": 4961 }, { "epoch": 1.78, "learning_rate": 5.5866474716573695e-06, "loss": 0.658, "step": 4962 }, { "epoch": 1.78, "learning_rate": 5.585747705596546e-06, "loss": 0.6524, "step": 4963 }, { "epoch": 1.78, "learning_rate": 5.584847939535721e-06, "loss": 0.6401, "step": 4964 }, { "epoch": 1.78, "learning_rate": 5.583948173474897e-06, "loss": 0.5878, "step": 4965 }, { "epoch": 1.78, "learning_rate": 5.583048407414073e-06, "loss": 0.5382, "step": 4966 }, { "epoch": 1.78, "learning_rate": 5.582148641353249e-06, "loss": 0.6531, "step": 4967 }, { "epoch": 1.78, "learning_rate": 5.581248875292424e-06, "loss": 0.631, "step": 4968 }, { "epoch": 1.78, "learning_rate": 5.5803491092316e-06, "loss": 0.591, "step": 4969 }, { "epoch": 1.78, "learning_rate": 5.5794493431707765e-06, "loss": 0.5413, "step": 4970 }, { "epoch": 1.78, "learning_rate": 5.5785495771099514e-06, "loss": 0.5728, "step": 4971 }, { "epoch": 1.78, "learning_rate": 5.577649811049127e-06, "loss": 0.646, "step": 4972 }, { "epoch": 1.78, "learning_rate": 5.576750044988304e-06, "loss": 0.6013, "step": 4973 }, { "epoch": 1.79, "learning_rate": 5.57585027892748e-06, "loss": 0.6007, "step": 4974 }, { "epoch": 1.79, "learning_rate": 5.5749505128666545e-06, "loss": 0.6589, "step": 4975 }, { "epoch": 1.79, "learning_rate": 5.574050746805831e-06, "loss": 0.608, "step": 4976 }, { "epoch": 1.79, "learning_rate": 5.573150980745007e-06, "loss": 0.6325, "step": 4977 }, { "epoch": 1.79, "learning_rate": 5.572251214684183e-06, "loss": 0.711, "step": 4978 }, { "epoch": 1.79, "learning_rate": 5.571351448623359e-06, "loss": 0.625, "step": 4979 }, { "epoch": 1.79, "learning_rate": 5.570451682562534e-06, "loss": 0.6265, "step": 4980 }, { "epoch": 1.79, "learning_rate": 5.56955191650171e-06, "loss": 0.6255, "step": 4981 }, { "epoch": 1.79, "learning_rate": 5.568652150440887e-06, "loss": 0.6721, "step": 4982 }, { "epoch": 1.79, "learning_rate": 5.5677523843800615e-06, "loss": 0.6268, "step": 4983 }, { "epoch": 1.79, "learning_rate": 5.566852618319237e-06, "loss": 0.5884, "step": 4984 }, { "epoch": 1.79, "learning_rate": 5.565952852258413e-06, "loss": 0.6085, "step": 4985 }, { "epoch": 1.79, "learning_rate": 5.56505308619759e-06, "loss": 0.576, "step": 4986 }, { "epoch": 1.79, "learning_rate": 5.564153320136765e-06, "loss": 0.6581, "step": 4987 }, { "epoch": 1.79, "learning_rate": 5.56325355407594e-06, "loss": 0.5835, "step": 4988 }, { "epoch": 1.79, "learning_rate": 5.562353788015117e-06, "loss": 0.5604, "step": 4989 }, { "epoch": 1.79, "learning_rate": 5.561454021954292e-06, "loss": 0.6081, "step": 4990 }, { "epoch": 1.79, "learning_rate": 5.560554255893468e-06, "loss": 0.5583, "step": 4991 }, { "epoch": 1.79, "learning_rate": 5.559654489832644e-06, "loss": 0.5944, "step": 4992 }, { "epoch": 1.79, "learning_rate": 5.55875472377182e-06, "loss": 0.7339, "step": 4993 }, { "epoch": 1.79, "learning_rate": 5.557854957710995e-06, "loss": 0.6424, "step": 4994 }, { "epoch": 1.79, "learning_rate": 5.5569551916501716e-06, "loss": 0.6005, "step": 4995 }, { "epoch": 1.79, "learning_rate": 5.556055425589347e-06, "loss": 0.5982, "step": 4996 }, { "epoch": 1.79, "learning_rate": 5.555155659528523e-06, "loss": 0.6737, "step": 4997 }, { "epoch": 1.79, "learning_rate": 5.5542558934677e-06, "loss": 0.5299, "step": 4998 }, { "epoch": 1.79, "learning_rate": 5.553356127406875e-06, "loss": 0.6336, "step": 4999 }, { "epoch": 1.79, "learning_rate": 5.55245636134605e-06, "loss": 0.6972, "step": 5000 }, { "epoch": 1.8, "learning_rate": 5.551556595285227e-06, "loss": 0.6113, "step": 5001 }, { "epoch": 1.8, "learning_rate": 5.550656829224402e-06, "loss": 0.704, "step": 5002 }, { "epoch": 1.8, "learning_rate": 5.549757063163578e-06, "loss": 0.6497, "step": 5003 }, { "epoch": 1.8, "learning_rate": 5.5488572971027535e-06, "loss": 0.5496, "step": 5004 }, { "epoch": 1.8, "learning_rate": 5.54795753104193e-06, "loss": 0.5811, "step": 5005 }, { "epoch": 1.8, "learning_rate": 5.547057764981105e-06, "loss": 0.5513, "step": 5006 }, { "epoch": 1.8, "learning_rate": 5.546157998920281e-06, "loss": 0.6227, "step": 5007 }, { "epoch": 1.8, "learning_rate": 5.545258232859457e-06, "loss": 0.6705, "step": 5008 }, { "epoch": 1.8, "learning_rate": 5.544358466798633e-06, "loss": 0.5572, "step": 5009 }, { "epoch": 1.8, "learning_rate": 5.543458700737808e-06, "loss": 0.6032, "step": 5010 }, { "epoch": 1.8, "learning_rate": 5.542558934676985e-06, "loss": 0.6854, "step": 5011 }, { "epoch": 1.8, "learning_rate": 5.5416591686161605e-06, "loss": 0.6063, "step": 5012 }, { "epoch": 1.8, "learning_rate": 5.540759402555335e-06, "loss": 0.5271, "step": 5013 }, { "epoch": 1.8, "learning_rate": 5.539859636494512e-06, "loss": 0.6746, "step": 5014 }, { "epoch": 1.8, "learning_rate": 5.538959870433688e-06, "loss": 0.6402, "step": 5015 }, { "epoch": 1.8, "learning_rate": 5.5380601043728636e-06, "loss": 0.696, "step": 5016 }, { "epoch": 1.8, "learning_rate": 5.53716033831204e-06, "loss": 0.6058, "step": 5017 }, { "epoch": 1.8, "learning_rate": 5.536260572251215e-06, "loss": 0.4705, "step": 5018 }, { "epoch": 1.8, "learning_rate": 5.535360806190391e-06, "loss": 0.6301, "step": 5019 }, { "epoch": 1.8, "learning_rate": 5.5344610401295675e-06, "loss": 0.5272, "step": 5020 }, { "epoch": 1.8, "learning_rate": 5.533561274068742e-06, "loss": 0.6762, "step": 5021 }, { "epoch": 1.8, "learning_rate": 5.532661508007918e-06, "loss": 0.684, "step": 5022 }, { "epoch": 1.8, "learning_rate": 5.531761741947094e-06, "loss": 0.6889, "step": 5023 }, { "epoch": 1.8, "learning_rate": 5.5308619758862706e-06, "loss": 0.658, "step": 5024 }, { "epoch": 1.8, "learning_rate": 5.5299622098254455e-06, "loss": 0.6074, "step": 5025 }, { "epoch": 1.8, "learning_rate": 5.529062443764621e-06, "loss": 0.6146, "step": 5026 }, { "epoch": 1.8, "learning_rate": 5.528162677703798e-06, "loss": 0.5762, "step": 5027 }, { "epoch": 1.8, "learning_rate": 5.527262911642974e-06, "loss": 0.7374, "step": 5028 }, { "epoch": 1.81, "learning_rate": 5.5263631455821486e-06, "loss": 0.5777, "step": 5029 }, { "epoch": 1.81, "learning_rate": 5.525463379521325e-06, "loss": 0.5936, "step": 5030 }, { "epoch": 1.81, "learning_rate": 5.524563613460501e-06, "loss": 0.6866, "step": 5031 }, { "epoch": 1.81, "learning_rate": 5.523663847399676e-06, "loss": 0.549, "step": 5032 }, { "epoch": 1.81, "learning_rate": 5.5227640813388525e-06, "loss": 0.6042, "step": 5033 }, { "epoch": 1.81, "learning_rate": 5.521864315278028e-06, "loss": 0.5646, "step": 5034 }, { "epoch": 1.81, "learning_rate": 5.520964549217204e-06, "loss": 0.5927, "step": 5035 }, { "epoch": 1.81, "learning_rate": 5.520064783156381e-06, "loss": 0.6365, "step": 5036 }, { "epoch": 1.81, "learning_rate": 5.5191650170955556e-06, "loss": 0.6432, "step": 5037 }, { "epoch": 1.81, "learning_rate": 5.518265251034731e-06, "loss": 0.5879, "step": 5038 }, { "epoch": 1.81, "learning_rate": 5.517365484973908e-06, "loss": 0.6198, "step": 5039 }, { "epoch": 1.81, "learning_rate": 5.516465718913083e-06, "loss": 0.5632, "step": 5040 }, { "epoch": 1.81, "learning_rate": 5.515565952852259e-06, "loss": 0.4857, "step": 5041 }, { "epoch": 1.81, "learning_rate": 5.514666186791434e-06, "loss": 0.6281, "step": 5042 }, { "epoch": 1.81, "learning_rate": 5.513766420730611e-06, "loss": 0.6679, "step": 5043 }, { "epoch": 1.81, "learning_rate": 5.512866654669786e-06, "loss": 0.587, "step": 5044 }, { "epoch": 1.81, "learning_rate": 5.511966888608962e-06, "loss": 0.6058, "step": 5045 }, { "epoch": 1.81, "learning_rate": 5.511067122548138e-06, "loss": 0.6911, "step": 5046 }, { "epoch": 1.81, "learning_rate": 5.510167356487314e-06, "loss": 0.5859, "step": 5047 }, { "epoch": 1.81, "learning_rate": 5.509267590426489e-06, "loss": 0.6069, "step": 5048 }, { "epoch": 1.81, "learning_rate": 5.508367824365666e-06, "loss": 0.6022, "step": 5049 }, { "epoch": 1.81, "learning_rate": 5.507468058304841e-06, "loss": 0.6348, "step": 5050 }, { "epoch": 1.81, "learning_rate": 5.506568292244016e-06, "loss": 0.4436, "step": 5051 }, { "epoch": 1.81, "learning_rate": 5.505668526183193e-06, "loss": 0.5568, "step": 5052 }, { "epoch": 1.81, "learning_rate": 5.504768760122369e-06, "loss": 0.6598, "step": 5053 }, { "epoch": 1.81, "learning_rate": 5.5038689940615445e-06, "loss": 0.5694, "step": 5054 }, { "epoch": 1.81, "learning_rate": 5.502969228000721e-06, "loss": 0.685, "step": 5055 }, { "epoch": 1.81, "learning_rate": 5.502069461939896e-06, "loss": 0.6215, "step": 5056 }, { "epoch": 1.82, "learning_rate": 5.501169695879072e-06, "loss": 0.6839, "step": 5057 }, { "epoch": 1.82, "learning_rate": 5.500269929818248e-06, "loss": 0.6366, "step": 5058 }, { "epoch": 1.82, "learning_rate": 5.499370163757423e-06, "loss": 0.6238, "step": 5059 }, { "epoch": 1.82, "learning_rate": 5.498470397696599e-06, "loss": 0.5639, "step": 5060 }, { "epoch": 1.82, "learning_rate": 5.497570631635775e-06, "loss": 0.6613, "step": 5061 }, { "epoch": 1.82, "learning_rate": 5.4966708655749515e-06, "loss": 0.6267, "step": 5062 }, { "epoch": 1.82, "learning_rate": 5.495771099514126e-06, "loss": 0.6464, "step": 5063 }, { "epoch": 1.82, "learning_rate": 5.494871333453302e-06, "loss": 0.6689, "step": 5064 }, { "epoch": 1.82, "learning_rate": 5.493971567392479e-06, "loss": 0.666, "step": 5065 }, { "epoch": 1.82, "learning_rate": 5.4930718013316545e-06, "loss": 0.6022, "step": 5066 }, { "epoch": 1.82, "learning_rate": 5.4921720352708295e-06, "loss": 0.6536, "step": 5067 }, { "epoch": 1.82, "learning_rate": 5.491272269210006e-06, "loss": 0.5716, "step": 5068 }, { "epoch": 1.82, "learning_rate": 5.490372503149182e-06, "loss": 0.5462, "step": 5069 }, { "epoch": 1.82, "learning_rate": 5.489472737088357e-06, "loss": 0.5557, "step": 5070 }, { "epoch": 1.82, "learning_rate": 5.488572971027533e-06, "loss": 0.5811, "step": 5071 }, { "epoch": 1.82, "learning_rate": 5.487673204966709e-06, "loss": 0.6813, "step": 5072 }, { "epoch": 1.82, "learning_rate": 5.486773438905885e-06, "loss": 0.6163, "step": 5073 }, { "epoch": 1.82, "learning_rate": 5.4858736728450615e-06, "loss": 0.6522, "step": 5074 }, { "epoch": 1.82, "learning_rate": 5.4849739067842365e-06, "loss": 0.737, "step": 5075 }, { "epoch": 1.82, "learning_rate": 5.484074140723412e-06, "loss": 0.5721, "step": 5076 }, { "epoch": 1.82, "learning_rate": 5.483174374662589e-06, "loss": 0.5769, "step": 5077 }, { "epoch": 1.82, "learning_rate": 5.482274608601764e-06, "loss": 0.5711, "step": 5078 }, { "epoch": 1.82, "learning_rate": 5.4813748425409395e-06, "loss": 0.614, "step": 5079 }, { "epoch": 1.82, "learning_rate": 5.480475076480115e-06, "loss": 0.6016, "step": 5080 }, { "epoch": 1.82, "learning_rate": 5.479575310419292e-06, "loss": 0.7179, "step": 5081 }, { "epoch": 1.82, "learning_rate": 5.478675544358467e-06, "loss": 0.6967, "step": 5082 }, { "epoch": 1.82, "learning_rate": 5.477775778297643e-06, "loss": 0.6369, "step": 5083 }, { "epoch": 1.82, "learning_rate": 5.476876012236819e-06, "loss": 0.6623, "step": 5084 }, { "epoch": 1.83, "learning_rate": 5.475976246175995e-06, "loss": 0.623, "step": 5085 }, { "epoch": 1.83, "learning_rate": 5.47507648011517e-06, "loss": 0.5844, "step": 5086 }, { "epoch": 1.83, "learning_rate": 5.4741767140543465e-06, "loss": 0.5938, "step": 5087 }, { "epoch": 1.83, "learning_rate": 5.473276947993522e-06, "loss": 0.5173, "step": 5088 }, { "epoch": 1.83, "learning_rate": 5.472377181932697e-06, "loss": 0.6176, "step": 5089 }, { "epoch": 1.83, "learning_rate": 5.471477415871874e-06, "loss": 0.5952, "step": 5090 }, { "epoch": 1.83, "learning_rate": 5.47057764981105e-06, "loss": 0.5525, "step": 5091 }, { "epoch": 1.83, "learning_rate": 5.469677883750225e-06, "loss": 0.5183, "step": 5092 }, { "epoch": 1.83, "learning_rate": 5.468778117689402e-06, "loss": 0.657, "step": 5093 }, { "epoch": 1.83, "learning_rate": 5.467878351628577e-06, "loss": 0.6307, "step": 5094 }, { "epoch": 1.83, "learning_rate": 5.466978585567753e-06, "loss": 0.6528, "step": 5095 }, { "epoch": 1.83, "learning_rate": 5.4660788195069285e-06, "loss": 0.6769, "step": 5096 }, { "epoch": 1.83, "learning_rate": 5.465179053446105e-06, "loss": 0.5541, "step": 5097 }, { "epoch": 1.83, "learning_rate": 5.46427928738528e-06, "loss": 0.6639, "step": 5098 }, { "epoch": 1.83, "learning_rate": 5.463379521324456e-06, "loss": 0.657, "step": 5099 }, { "epoch": 1.83, "learning_rate": 5.462479755263632e-06, "loss": 0.4861, "step": 5100 }, { "epoch": 1.83, "learning_rate": 5.461579989202807e-06, "loss": 0.6358, "step": 5101 }, { "epoch": 1.83, "learning_rate": 5.460680223141983e-06, "loss": 0.6509, "step": 5102 }, { "epoch": 1.83, "learning_rate": 5.45978045708116e-06, "loss": 0.5839, "step": 5103 }, { "epoch": 1.83, "learning_rate": 5.4588806910203355e-06, "loss": 0.6172, "step": 5104 }, { "epoch": 1.83, "learning_rate": 5.45798092495951e-06, "loss": 0.6434, "step": 5105 }, { "epoch": 1.83, "learning_rate": 5.457081158898687e-06, "loss": 0.4996, "step": 5106 }, { "epoch": 1.83, "learning_rate": 5.456181392837863e-06, "loss": 0.5773, "step": 5107 }, { "epoch": 1.83, "learning_rate": 5.455281626777038e-06, "loss": 0.6808, "step": 5108 }, { "epoch": 1.83, "learning_rate": 5.454381860716214e-06, "loss": 0.6518, "step": 5109 }, { "epoch": 1.83, "learning_rate": 5.45348209465539e-06, "loss": 0.5827, "step": 5110 }, { "epoch": 1.83, "learning_rate": 5.452582328594566e-06, "loss": 0.6531, "step": 5111 }, { "epoch": 1.83, "learning_rate": 5.4516825625337425e-06, "loss": 0.6439, "step": 5112 }, { "epoch": 1.84, "learning_rate": 5.450782796472917e-06, "loss": 0.7114, "step": 5113 }, { "epoch": 1.84, "learning_rate": 5.449883030412093e-06, "loss": 0.6908, "step": 5114 }, { "epoch": 1.84, "learning_rate": 5.448983264351269e-06, "loss": 0.5968, "step": 5115 }, { "epoch": 1.84, "learning_rate": 5.4480834982904455e-06, "loss": 0.6285, "step": 5116 }, { "epoch": 1.84, "learning_rate": 5.4471837322296204e-06, "loss": 0.6782, "step": 5117 }, { "epoch": 1.84, "learning_rate": 5.446283966168796e-06, "loss": 0.7075, "step": 5118 }, { "epoch": 1.84, "learning_rate": 5.445384200107973e-06, "loss": 0.5132, "step": 5119 }, { "epoch": 1.84, "learning_rate": 5.444484434047148e-06, "loss": 0.647, "step": 5120 }, { "epoch": 1.84, "learning_rate": 5.4435846679863235e-06, "loss": 0.5212, "step": 5121 }, { "epoch": 1.84, "learning_rate": 5.4426849019255e-06, "loss": 0.5936, "step": 5122 }, { "epoch": 1.84, "learning_rate": 5.441785135864676e-06, "loss": 0.5475, "step": 5123 }, { "epoch": 1.84, "learning_rate": 5.440885369803851e-06, "loss": 0.6584, "step": 5124 }, { "epoch": 1.84, "learning_rate": 5.4399856037430274e-06, "loss": 0.6015, "step": 5125 }, { "epoch": 1.84, "learning_rate": 5.439085837682203e-06, "loss": 0.5775, "step": 5126 }, { "epoch": 1.84, "learning_rate": 5.438186071621379e-06, "loss": 0.58, "step": 5127 }, { "epoch": 1.84, "learning_rate": 5.437286305560555e-06, "loss": 0.592, "step": 5128 }, { "epoch": 1.84, "learning_rate": 5.4363865394997305e-06, "loss": 0.6242, "step": 5129 }, { "epoch": 1.84, "learning_rate": 5.435486773438906e-06, "loss": 0.7267, "step": 5130 }, { "epoch": 1.84, "learning_rate": 5.434587007378083e-06, "loss": 0.7575, "step": 5131 }, { "epoch": 1.84, "learning_rate": 5.433687241317258e-06, "loss": 0.6792, "step": 5132 }, { "epoch": 1.84, "learning_rate": 5.432787475256434e-06, "loss": 0.5886, "step": 5133 }, { "epoch": 1.84, "learning_rate": 5.431887709195609e-06, "loss": 0.639, "step": 5134 }, { "epoch": 1.84, "learning_rate": 5.430987943134786e-06, "loss": 0.6233, "step": 5135 }, { "epoch": 1.84, "learning_rate": 5.430088177073961e-06, "loss": 0.6014, "step": 5136 }, { "epoch": 1.84, "learning_rate": 5.429188411013137e-06, "loss": 0.5142, "step": 5137 }, { "epoch": 1.84, "learning_rate": 5.428288644952313e-06, "loss": 0.5454, "step": 5138 }, { "epoch": 1.84, "learning_rate": 5.427388878891488e-06, "loss": 0.6143, "step": 5139 }, { "epoch": 1.84, "learning_rate": 5.426489112830664e-06, "loss": 0.7269, "step": 5140 }, { "epoch": 1.85, "learning_rate": 5.425589346769841e-06, "loss": 0.6519, "step": 5141 }, { "epoch": 1.85, "learning_rate": 5.424689580709016e-06, "loss": 0.5197, "step": 5142 }, { "epoch": 1.85, "learning_rate": 5.423789814648191e-06, "loss": 0.5632, "step": 5143 }, { "epoch": 1.85, "learning_rate": 5.422890048587368e-06, "loss": 0.6367, "step": 5144 }, { "epoch": 1.85, "learning_rate": 5.421990282526544e-06, "loss": 0.7065, "step": 5145 }, { "epoch": 1.85, "learning_rate": 5.4210905164657194e-06, "loss": 0.6341, "step": 5146 }, { "epoch": 1.85, "learning_rate": 5.420190750404895e-06, "loss": 0.7041, "step": 5147 }, { "epoch": 1.85, "learning_rate": 5.419290984344071e-06, "loss": 0.5938, "step": 5148 }, { "epoch": 1.85, "learning_rate": 5.418391218283247e-06, "loss": 0.6643, "step": 5149 }, { "epoch": 1.85, "learning_rate": 5.417491452222423e-06, "loss": 0.746, "step": 5150 }, { "epoch": 1.85, "learning_rate": 5.416591686161598e-06, "loss": 0.5919, "step": 5151 }, { "epoch": 1.85, "learning_rate": 5.415691920100774e-06, "loss": 0.5434, "step": 5152 }, { "epoch": 1.85, "learning_rate": 5.41479215403995e-06, "loss": 0.5484, "step": 5153 }, { "epoch": 1.85, "learning_rate": 5.4138923879791264e-06, "loss": 0.5928, "step": 5154 }, { "epoch": 1.85, "learning_rate": 5.412992621918301e-06, "loss": 0.6464, "step": 5155 }, { "epoch": 1.85, "learning_rate": 5.412092855857477e-06, "loss": 0.5934, "step": 5156 }, { "epoch": 1.85, "learning_rate": 5.411193089796654e-06, "loss": 0.6286, "step": 5157 }, { "epoch": 1.85, "learning_rate": 5.410293323735829e-06, "loss": 0.6741, "step": 5158 }, { "epoch": 1.85, "learning_rate": 5.4093935576750044e-06, "loss": 0.7207, "step": 5159 }, { "epoch": 1.85, "learning_rate": 5.408493791614181e-06, "loss": 0.6149, "step": 5160 }, { "epoch": 1.85, "learning_rate": 5.407594025553357e-06, "loss": 0.6532, "step": 5161 }, { "epoch": 1.85, "learning_rate": 5.406694259492532e-06, "loss": 0.5859, "step": 5162 }, { "epoch": 1.85, "learning_rate": 5.405794493431708e-06, "loss": 0.6163, "step": 5163 }, { "epoch": 1.85, "learning_rate": 5.404894727370884e-06, "loss": 0.536, "step": 5164 }, { "epoch": 1.85, "learning_rate": 5.40399496131006e-06, "loss": 0.5822, "step": 5165 }, { "epoch": 1.85, "learning_rate": 5.4030951952492365e-06, "loss": 0.6331, "step": 5166 }, { "epoch": 1.85, "learning_rate": 5.4021954291884114e-06, "loss": 0.6531, "step": 5167 }, { "epoch": 1.85, "learning_rate": 5.401295663127587e-06, "loss": 0.6328, "step": 5168 }, { "epoch": 1.86, "learning_rate": 5.400395897066764e-06, "loss": 0.5565, "step": 5169 }, { "epoch": 1.86, "learning_rate": 5.399496131005939e-06, "loss": 0.6313, "step": 5170 }, { "epoch": 1.86, "learning_rate": 5.3985963649451145e-06, "loss": 0.6215, "step": 5171 }, { "epoch": 1.86, "learning_rate": 5.39769659888429e-06, "loss": 0.5945, "step": 5172 }, { "epoch": 1.86, "learning_rate": 5.396796832823467e-06, "loss": 0.6554, "step": 5173 }, { "epoch": 1.86, "learning_rate": 5.395897066762642e-06, "loss": 0.6082, "step": 5174 }, { "epoch": 1.86, "learning_rate": 5.394997300701818e-06, "loss": 0.6545, "step": 5175 }, { "epoch": 1.86, "learning_rate": 5.394097534640994e-06, "loss": 0.576, "step": 5176 }, { "epoch": 1.86, "learning_rate": 5.393197768580169e-06, "loss": 0.6457, "step": 5177 }, { "epoch": 1.86, "learning_rate": 5.392298002519345e-06, "loss": 0.6569, "step": 5178 }, { "epoch": 1.86, "learning_rate": 5.3913982364585215e-06, "loss": 0.621, "step": 5179 }, { "epoch": 1.86, "learning_rate": 5.390498470397697e-06, "loss": 0.5233, "step": 5180 }, { "epoch": 1.86, "learning_rate": 5.389598704336872e-06, "loss": 0.615, "step": 5181 }, { "epoch": 1.86, "learning_rate": 5.388698938276049e-06, "loss": 0.6178, "step": 5182 }, { "epoch": 1.86, "learning_rate": 5.387799172215225e-06, "loss": 0.608, "step": 5183 }, { "epoch": 1.86, "learning_rate": 5.3868994061544e-06, "loss": 0.6548, "step": 5184 }, { "epoch": 1.86, "learning_rate": 5.385999640093577e-06, "loss": 0.7167, "step": 5185 }, { "epoch": 1.86, "learning_rate": 5.385099874032752e-06, "loss": 0.6075, "step": 5186 }, { "epoch": 1.86, "learning_rate": 5.384200107971928e-06, "loss": 0.6621, "step": 5187 }, { "epoch": 1.86, "learning_rate": 5.383300341911104e-06, "loss": 0.5926, "step": 5188 }, { "epoch": 1.86, "learning_rate": 5.382400575850279e-06, "loss": 0.7303, "step": 5189 }, { "epoch": 1.86, "learning_rate": 5.381500809789455e-06, "loss": 0.587, "step": 5190 }, { "epoch": 1.86, "learning_rate": 5.380601043728631e-06, "loss": 0.6474, "step": 5191 }, { "epoch": 1.86, "learning_rate": 5.379701277667807e-06, "loss": 0.6645, "step": 5192 }, { "epoch": 1.86, "learning_rate": 5.378801511606982e-06, "loss": 0.6612, "step": 5193 }, { "epoch": 1.86, "learning_rate": 5.377901745546158e-06, "loss": 0.6277, "step": 5194 }, { "epoch": 1.86, "learning_rate": 5.377001979485335e-06, "loss": 0.5489, "step": 5195 }, { "epoch": 1.87, "learning_rate": 5.37610221342451e-06, "loss": 0.5862, "step": 5196 }, { "epoch": 1.87, "learning_rate": 5.375202447363685e-06, "loss": 0.6485, "step": 5197 }, { "epoch": 1.87, "learning_rate": 5.374302681302862e-06, "loss": 0.5874, "step": 5198 }, { "epoch": 1.87, "learning_rate": 5.373402915242038e-06, "loss": 0.6182, "step": 5199 }, { "epoch": 1.87, "learning_rate": 5.372503149181213e-06, "loss": 0.676, "step": 5200 }, { "epoch": 1.87, "learning_rate": 5.371603383120389e-06, "loss": 0.6752, "step": 5201 }, { "epoch": 1.87, "learning_rate": 5.370703617059565e-06, "loss": 0.692, "step": 5202 }, { "epoch": 1.87, "learning_rate": 5.369803850998741e-06, "loss": 0.502, "step": 5203 }, { "epoch": 1.87, "learning_rate": 5.368904084937917e-06, "loss": 0.573, "step": 5204 }, { "epoch": 1.87, "learning_rate": 5.368004318877092e-06, "loss": 0.6116, "step": 5205 }, { "epoch": 1.87, "learning_rate": 5.367104552816268e-06, "loss": 0.6847, "step": 5206 }, { "epoch": 1.87, "learning_rate": 5.366204786755445e-06, "loss": 0.6984, "step": 5207 }, { "epoch": 1.87, "learning_rate": 5.36530502069462e-06, "loss": 0.5816, "step": 5208 }, { "epoch": 1.87, "learning_rate": 5.364405254633795e-06, "loss": 0.5639, "step": 5209 }, { "epoch": 1.87, "learning_rate": 5.363505488572971e-06, "loss": 0.675, "step": 5210 }, { "epoch": 1.87, "learning_rate": 5.362605722512148e-06, "loss": 0.6577, "step": 5211 }, { "epoch": 1.87, "learning_rate": 5.361705956451323e-06, "loss": 0.6144, "step": 5212 }, { "epoch": 1.87, "learning_rate": 5.3608061903904985e-06, "loss": 0.6618, "step": 5213 }, { "epoch": 1.87, "learning_rate": 5.359906424329675e-06, "loss": 0.6933, "step": 5214 }, { "epoch": 1.87, "learning_rate": 5.359006658268851e-06, "loss": 0.6195, "step": 5215 }, { "epoch": 1.87, "learning_rate": 5.358106892208026e-06, "loss": 0.6196, "step": 5216 }, { "epoch": 1.87, "learning_rate": 5.357207126147202e-06, "loss": 0.6351, "step": 5217 }, { "epoch": 1.87, "learning_rate": 5.356307360086378e-06, "loss": 0.6554, "step": 5218 }, { "epoch": 1.87, "learning_rate": 5.355407594025553e-06, "loss": 0.6483, "step": 5219 }, { "epoch": 1.87, "learning_rate": 5.35450782796473e-06, "loss": 0.6858, "step": 5220 }, { "epoch": 1.87, "learning_rate": 5.3536080619039055e-06, "loss": 0.6167, "step": 5221 }, { "epoch": 1.87, "learning_rate": 5.352708295843081e-06, "loss": 0.6599, "step": 5222 }, { "epoch": 1.87, "learning_rate": 5.351808529782258e-06, "loss": 0.6114, "step": 5223 }, { "epoch": 1.88, "learning_rate": 5.350908763721433e-06, "loss": 0.5635, "step": 5224 }, { "epoch": 1.88, "learning_rate": 5.3500089976606086e-06, "loss": 0.6813, "step": 5225 }, { "epoch": 1.88, "learning_rate": 5.349109231599785e-06, "loss": 0.5264, "step": 5226 }, { "epoch": 1.88, "learning_rate": 5.34820946553896e-06, "loss": 0.636, "step": 5227 }, { "epoch": 1.88, "learning_rate": 5.347309699478136e-06, "loss": 0.6536, "step": 5228 }, { "epoch": 1.88, "learning_rate": 5.346409933417312e-06, "loss": 0.6504, "step": 5229 }, { "epoch": 1.88, "learning_rate": 5.345510167356488e-06, "loss": 0.5756, "step": 5230 }, { "epoch": 1.88, "learning_rate": 5.344610401295663e-06, "loss": 0.5957, "step": 5231 }, { "epoch": 1.88, "learning_rate": 5.343710635234839e-06, "loss": 0.6323, "step": 5232 }, { "epoch": 1.88, "learning_rate": 5.3428108691740156e-06, "loss": 0.676, "step": 5233 }, { "epoch": 1.88, "learning_rate": 5.341911103113191e-06, "loss": 0.6161, "step": 5234 }, { "epoch": 1.88, "learning_rate": 5.341011337052366e-06, "loss": 0.5783, "step": 5235 }, { "epoch": 1.88, "learning_rate": 5.340111570991543e-06, "loss": 0.613, "step": 5236 }, { "epoch": 1.88, "learning_rate": 5.339211804930719e-06, "loss": 0.6737, "step": 5237 }, { "epoch": 1.88, "learning_rate": 5.3383120388698936e-06, "loss": 0.6041, "step": 5238 }, { "epoch": 1.88, "learning_rate": 5.33741227280907e-06, "loss": 0.6823, "step": 5239 }, { "epoch": 1.88, "learning_rate": 5.336512506748246e-06, "loss": 0.5697, "step": 5240 }, { "epoch": 1.88, "learning_rate": 5.335612740687422e-06, "loss": 0.5602, "step": 5241 }, { "epoch": 1.88, "learning_rate": 5.334712974626598e-06, "loss": 0.5702, "step": 5242 }, { "epoch": 1.88, "learning_rate": 5.333813208565773e-06, "loss": 0.5399, "step": 5243 }, { "epoch": 1.88, "learning_rate": 5.332913442504949e-06, "loss": 0.6955, "step": 5244 }, { "epoch": 1.88, "learning_rate": 5.332013676444126e-06, "loss": 0.5591, "step": 5245 }, { "epoch": 1.88, "learning_rate": 5.3311139103833006e-06, "loss": 0.6086, "step": 5246 }, { "epoch": 1.88, "learning_rate": 5.330214144322476e-06, "loss": 0.5965, "step": 5247 }, { "epoch": 1.88, "learning_rate": 5.329314378261652e-06, "loss": 0.6669, "step": 5248 }, { "epoch": 1.88, "learning_rate": 5.328414612200829e-06, "loss": 0.5676, "step": 5249 }, { "epoch": 1.88, "learning_rate": 5.327514846140004e-06, "loss": 0.6018, "step": 5250 }, { "epoch": 1.88, "learning_rate": 5.326615080079179e-06, "loss": 0.5786, "step": 5251 }, { "epoch": 1.89, "learning_rate": 5.325715314018356e-06, "loss": 0.6642, "step": 5252 }, { "epoch": 1.89, "learning_rate": 5.324815547957532e-06, "loss": 0.6662, "step": 5253 }, { "epoch": 1.89, "learning_rate": 5.323915781896707e-06, "loss": 0.6591, "step": 5254 }, { "epoch": 1.89, "learning_rate": 5.323016015835883e-06, "loss": 0.6133, "step": 5255 }, { "epoch": 1.89, "learning_rate": 5.322116249775059e-06, "loss": 0.5608, "step": 5256 }, { "epoch": 1.89, "learning_rate": 5.321216483714234e-06, "loss": 0.5665, "step": 5257 }, { "epoch": 1.89, "learning_rate": 5.320316717653411e-06, "loss": 0.6854, "step": 5258 }, { "epoch": 1.89, "learning_rate": 5.319416951592586e-06, "loss": 0.4901, "step": 5259 }, { "epoch": 1.89, "learning_rate": 5.318517185531762e-06, "loss": 0.6213, "step": 5260 }, { "epoch": 1.89, "learning_rate": 5.317617419470939e-06, "loss": 0.6357, "step": 5261 }, { "epoch": 1.89, "learning_rate": 5.316717653410114e-06, "loss": 0.5884, "step": 5262 }, { "epoch": 1.89, "learning_rate": 5.3158178873492895e-06, "loss": 0.6695, "step": 5263 }, { "epoch": 1.89, "learning_rate": 5.314918121288466e-06, "loss": 0.6243, "step": 5264 }, { "epoch": 1.89, "learning_rate": 5.314018355227641e-06, "loss": 0.6657, "step": 5265 }, { "epoch": 1.89, "learning_rate": 5.313118589166817e-06, "loss": 0.6135, "step": 5266 }, { "epoch": 1.89, "learning_rate": 5.3122188231059925e-06, "loss": 0.6462, "step": 5267 }, { "epoch": 1.89, "learning_rate": 5.311319057045169e-06, "loss": 0.569, "step": 5268 }, { "epoch": 1.89, "learning_rate": 5.310419290984344e-06, "loss": 0.5665, "step": 5269 }, { "epoch": 1.89, "learning_rate": 5.30951952492352e-06, "loss": 0.7033, "step": 5270 }, { "epoch": 1.89, "learning_rate": 5.3086197588626965e-06, "loss": 0.5929, "step": 5271 }, { "epoch": 1.89, "learning_rate": 5.307719992801872e-06, "loss": 0.7313, "step": 5272 }, { "epoch": 1.89, "learning_rate": 5.306820226741047e-06, "loss": 0.5761, "step": 5273 }, { "epoch": 1.89, "learning_rate": 5.305920460680224e-06, "loss": 0.7565, "step": 5274 }, { "epoch": 1.89, "learning_rate": 5.3050206946193995e-06, "loss": 0.6239, "step": 5275 }, { "epoch": 1.89, "learning_rate": 5.3041209285585745e-06, "loss": 0.6195, "step": 5276 }, { "epoch": 1.89, "learning_rate": 5.303221162497751e-06, "loss": 0.6943, "step": 5277 }, { "epoch": 1.89, "learning_rate": 5.302321396436927e-06, "loss": 0.5629, "step": 5278 }, { "epoch": 1.89, "learning_rate": 5.301421630376103e-06, "loss": 0.7286, "step": 5279 }, { "epoch": 1.9, "learning_rate": 5.300521864315279e-06, "loss": 0.556, "step": 5280 }, { "epoch": 1.9, "learning_rate": 5.299622098254454e-06, "loss": 0.6391, "step": 5281 }, { "epoch": 1.9, "learning_rate": 5.29872233219363e-06, "loss": 0.6005, "step": 5282 }, { "epoch": 1.9, "learning_rate": 5.2978225661328065e-06, "loss": 0.6495, "step": 5283 }, { "epoch": 1.9, "learning_rate": 5.296922800071982e-06, "loss": 0.5548, "step": 5284 }, { "epoch": 1.9, "learning_rate": 5.296023034011157e-06, "loss": 0.5907, "step": 5285 }, { "epoch": 1.9, "learning_rate": 5.295123267950333e-06, "loss": 0.6734, "step": 5286 }, { "epoch": 1.9, "learning_rate": 5.29422350188951e-06, "loss": 0.631, "step": 5287 }, { "epoch": 1.9, "learning_rate": 5.2933237358286845e-06, "loss": 0.6514, "step": 5288 }, { "epoch": 1.9, "learning_rate": 5.29242396976786e-06, "loss": 0.5933, "step": 5289 }, { "epoch": 1.9, "learning_rate": 5.291524203707037e-06, "loss": 0.6303, "step": 5290 }, { "epoch": 1.9, "learning_rate": 5.290624437646213e-06, "loss": 0.5663, "step": 5291 }, { "epoch": 1.9, "learning_rate": 5.289724671585388e-06, "loss": 0.5787, "step": 5292 }, { "epoch": 1.9, "learning_rate": 5.288824905524564e-06, "loss": 0.5946, "step": 5293 }, { "epoch": 1.9, "learning_rate": 5.28792513946374e-06, "loss": 0.6366, "step": 5294 }, { "epoch": 1.9, "learning_rate": 5.287025373402915e-06, "loss": 0.6822, "step": 5295 }, { "epoch": 1.9, "learning_rate": 5.2861256073420915e-06, "loss": 0.7105, "step": 5296 }, { "epoch": 1.9, "learning_rate": 5.285225841281267e-06, "loss": 0.6169, "step": 5297 }, { "epoch": 1.9, "learning_rate": 5.284326075220443e-06, "loss": 0.5945, "step": 5298 }, { "epoch": 1.9, "learning_rate": 5.28342630915962e-06, "loss": 0.5387, "step": 5299 }, { "epoch": 1.9, "learning_rate": 5.282526543098795e-06, "loss": 0.6687, "step": 5300 }, { "epoch": 1.9, "learning_rate": 5.28162677703797e-06, "loss": 0.5141, "step": 5301 }, { "epoch": 1.9, "learning_rate": 5.280727010977147e-06, "loss": 0.5776, "step": 5302 }, { "epoch": 1.9, "learning_rate": 5.279827244916323e-06, "loss": 0.59, "step": 5303 }, { "epoch": 1.9, "learning_rate": 5.278927478855498e-06, "loss": 0.641, "step": 5304 }, { "epoch": 1.9, "learning_rate": 5.2780277127946735e-06, "loss": 0.6585, "step": 5305 }, { "epoch": 1.9, "learning_rate": 5.27712794673385e-06, "loss": 0.6514, "step": 5306 }, { "epoch": 1.9, "learning_rate": 5.276228180673025e-06, "loss": 0.5824, "step": 5307 }, { "epoch": 1.91, "learning_rate": 5.275328414612201e-06, "loss": 0.5884, "step": 5308 }, { "epoch": 1.91, "learning_rate": 5.274428648551377e-06, "loss": 0.6281, "step": 5309 }, { "epoch": 1.91, "learning_rate": 5.273528882490553e-06, "loss": 0.598, "step": 5310 }, { "epoch": 1.91, "learning_rate": 5.272629116429728e-06, "loss": 0.6479, "step": 5311 }, { "epoch": 1.91, "learning_rate": 5.271729350368905e-06, "loss": 0.6734, "step": 5312 }, { "epoch": 1.91, "learning_rate": 5.2708295843080805e-06, "loss": 0.702, "step": 5313 }, { "epoch": 1.91, "learning_rate": 5.269929818247256e-06, "loss": 0.6148, "step": 5314 }, { "epoch": 1.91, "learning_rate": 5.269030052186432e-06, "loss": 0.5597, "step": 5315 }, { "epoch": 1.91, "learning_rate": 5.268130286125608e-06, "loss": 0.6447, "step": 5316 }, { "epoch": 1.91, "learning_rate": 5.2672305200647835e-06, "loss": 0.6816, "step": 5317 }, { "epoch": 1.91, "learning_rate": 5.26633075400396e-06, "loss": 0.5732, "step": 5318 }, { "epoch": 1.91, "learning_rate": 5.265430987943135e-06, "loss": 0.5571, "step": 5319 }, { "epoch": 1.91, "learning_rate": 5.264531221882311e-06, "loss": 0.5944, "step": 5320 }, { "epoch": 1.91, "learning_rate": 5.2636314558214874e-06, "loss": 0.6062, "step": 5321 }, { "epoch": 1.91, "learning_rate": 5.262731689760663e-06, "loss": 0.5878, "step": 5322 }, { "epoch": 1.91, "learning_rate": 5.261831923699838e-06, "loss": 0.5625, "step": 5323 }, { "epoch": 1.91, "learning_rate": 5.260932157639014e-06, "loss": 0.6399, "step": 5324 }, { "epoch": 1.91, "learning_rate": 5.2600323915781905e-06, "loss": 0.5701, "step": 5325 }, { "epoch": 1.91, "learning_rate": 5.2591326255173654e-06, "loss": 0.601, "step": 5326 }, { "epoch": 1.91, "learning_rate": 5.258232859456541e-06, "loss": 0.6222, "step": 5327 }, { "epoch": 1.91, "learning_rate": 5.257333093395718e-06, "loss": 0.6776, "step": 5328 }, { "epoch": 1.91, "learning_rate": 5.256433327334894e-06, "loss": 0.6601, "step": 5329 }, { "epoch": 1.91, "learning_rate": 5.2555335612740685e-06, "loss": 0.5912, "step": 5330 }, { "epoch": 1.91, "learning_rate": 5.254633795213245e-06, "loss": 0.6845, "step": 5331 }, { "epoch": 1.91, "learning_rate": 5.253734029152421e-06, "loss": 0.6342, "step": 5332 }, { "epoch": 1.91, "learning_rate": 5.252834263091597e-06, "loss": 0.6135, "step": 5333 }, { "epoch": 1.91, "learning_rate": 5.2519344970307724e-06, "loss": 0.5873, "step": 5334 }, { "epoch": 1.91, "learning_rate": 5.251034730969948e-06, "loss": 0.5816, "step": 5335 }, { "epoch": 1.92, "learning_rate": 5.250134964909124e-06, "loss": 0.5828, "step": 5336 }, { "epoch": 1.92, "learning_rate": 5.249235198848301e-06, "loss": 0.6085, "step": 5337 }, { "epoch": 1.92, "learning_rate": 5.2483354327874755e-06, "loss": 0.5433, "step": 5338 }, { "epoch": 1.92, "learning_rate": 5.247435666726651e-06, "loss": 0.6269, "step": 5339 }, { "epoch": 1.92, "learning_rate": 5.246535900665828e-06, "loss": 0.5933, "step": 5340 }, { "epoch": 1.92, "learning_rate": 5.245636134605004e-06, "loss": 0.5542, "step": 5341 }, { "epoch": 1.92, "learning_rate": 5.244736368544179e-06, "loss": 0.612, "step": 5342 }, { "epoch": 1.92, "learning_rate": 5.243836602483354e-06, "loss": 0.6508, "step": 5343 }, { "epoch": 1.92, "learning_rate": 5.242936836422531e-06, "loss": 0.6975, "step": 5344 }, { "epoch": 1.92, "learning_rate": 5.242037070361706e-06, "loss": 0.634, "step": 5345 }, { "epoch": 1.92, "learning_rate": 5.241137304300882e-06, "loss": 0.6652, "step": 5346 }, { "epoch": 1.92, "learning_rate": 5.240237538240058e-06, "loss": 0.6393, "step": 5347 }, { "epoch": 1.92, "learning_rate": 5.239337772179234e-06, "loss": 0.5379, "step": 5348 }, { "epoch": 1.92, "learning_rate": 5.238438006118409e-06, "loss": 0.5675, "step": 5349 }, { "epoch": 1.92, "learning_rate": 5.237538240057586e-06, "loss": 0.573, "step": 5350 }, { "epoch": 1.92, "learning_rate": 5.236638473996761e-06, "loss": 0.5809, "step": 5351 }, { "epoch": 1.92, "learning_rate": 5.235738707935937e-06, "loss": 0.6083, "step": 5352 }, { "epoch": 1.92, "learning_rate": 5.234838941875114e-06, "loss": 0.5294, "step": 5353 }, { "epoch": 1.92, "learning_rate": 5.233939175814289e-06, "loss": 0.675, "step": 5354 }, { "epoch": 1.92, "learning_rate": 5.2330394097534644e-06, "loss": 0.6398, "step": 5355 }, { "epoch": 1.92, "learning_rate": 5.232139643692641e-06, "loss": 0.5756, "step": 5356 }, { "epoch": 1.92, "learning_rate": 5.231239877631816e-06, "loss": 0.6249, "step": 5357 }, { "epoch": 1.92, "learning_rate": 5.230340111570992e-06, "loss": 0.6757, "step": 5358 }, { "epoch": 1.92, "learning_rate": 5.229440345510168e-06, "loss": 0.6466, "step": 5359 }, { "epoch": 1.92, "learning_rate": 5.228540579449344e-06, "loss": 0.5268, "step": 5360 }, { "epoch": 1.92, "learning_rate": 5.227640813388519e-06, "loss": 0.6085, "step": 5361 }, { "epoch": 1.92, "learning_rate": 5.226741047327695e-06, "loss": 0.6478, "step": 5362 }, { "epoch": 1.92, "learning_rate": 5.2258412812668714e-06, "loss": 0.5059, "step": 5363 }, { "epoch": 1.93, "learning_rate": 5.224941515206046e-06, "loss": 0.6095, "step": 5364 }, { "epoch": 1.93, "learning_rate": 5.224041749145222e-06, "loss": 0.608, "step": 5365 }, { "epoch": 1.93, "learning_rate": 5.223141983084399e-06, "loss": 0.626, "step": 5366 }, { "epoch": 1.93, "learning_rate": 5.2222422170235745e-06, "loss": 0.6179, "step": 5367 }, { "epoch": 1.93, "learning_rate": 5.2213424509627494e-06, "loss": 0.7214, "step": 5368 }, { "epoch": 1.93, "learning_rate": 5.220442684901926e-06, "loss": 0.6631, "step": 5369 }, { "epoch": 1.93, "learning_rate": 5.219542918841102e-06, "loss": 0.6977, "step": 5370 }, { "epoch": 1.93, "learning_rate": 5.218643152780278e-06, "loss": 0.6218, "step": 5371 }, { "epoch": 1.93, "learning_rate": 5.217743386719454e-06, "loss": 0.566, "step": 5372 }, { "epoch": 1.93, "learning_rate": 5.216843620658629e-06, "loss": 0.5833, "step": 5373 }, { "epoch": 1.93, "learning_rate": 5.215943854597805e-06, "loss": 0.6652, "step": 5374 }, { "epoch": 1.93, "learning_rate": 5.2150440885369815e-06, "loss": 0.6781, "step": 5375 }, { "epoch": 1.93, "learning_rate": 5.2141443224761564e-06, "loss": 0.5012, "step": 5376 }, { "epoch": 1.93, "learning_rate": 5.213244556415332e-06, "loss": 0.6526, "step": 5377 }, { "epoch": 1.93, "learning_rate": 5.212344790354509e-06, "loss": 0.5803, "step": 5378 }, { "epoch": 1.93, "learning_rate": 5.211445024293685e-06, "loss": 0.7005, "step": 5379 }, { "epoch": 1.93, "learning_rate": 5.2105452582328595e-06, "loss": 0.5757, "step": 5380 }, { "epoch": 1.93, "learning_rate": 5.209645492172035e-06, "loss": 0.6627, "step": 5381 }, { "epoch": 1.93, "learning_rate": 5.208745726111212e-06, "loss": 0.6208, "step": 5382 }, { "epoch": 1.93, "learning_rate": 5.207845960050387e-06, "loss": 0.5883, "step": 5383 }, { "epoch": 1.93, "learning_rate": 5.2069461939895626e-06, "loss": 0.5509, "step": 5384 }, { "epoch": 1.93, "learning_rate": 5.206046427928739e-06, "loss": 0.6674, "step": 5385 }, { "epoch": 1.93, "learning_rate": 5.205146661867915e-06, "loss": 0.6118, "step": 5386 }, { "epoch": 1.93, "learning_rate": 5.20424689580709e-06, "loss": 0.561, "step": 5387 }, { "epoch": 1.93, "learning_rate": 5.2033471297462665e-06, "loss": 0.6649, "step": 5388 }, { "epoch": 1.93, "learning_rate": 5.202447363685442e-06, "loss": 0.5544, "step": 5389 }, { "epoch": 1.93, "learning_rate": 5.201547597624618e-06, "loss": 0.6604, "step": 5390 }, { "epoch": 1.94, "learning_rate": 5.200647831563795e-06, "loss": 0.5317, "step": 5391 }, { "epoch": 1.94, "learning_rate": 5.1997480655029696e-06, "loss": 0.5779, "step": 5392 }, { "epoch": 1.94, "learning_rate": 5.198848299442145e-06, "loss": 0.696, "step": 5393 }, { "epoch": 1.94, "learning_rate": 5.197948533381322e-06, "loss": 0.5676, "step": 5394 }, { "epoch": 1.94, "learning_rate": 5.197048767320497e-06, "loss": 0.6958, "step": 5395 }, { "epoch": 1.94, "learning_rate": 5.196149001259673e-06, "loss": 0.6318, "step": 5396 }, { "epoch": 1.94, "learning_rate": 5.195249235198849e-06, "loss": 0.628, "step": 5397 }, { "epoch": 1.94, "learning_rate": 5.194349469138025e-06, "loss": 0.5948, "step": 5398 }, { "epoch": 1.94, "learning_rate": 5.1934497030772e-06, "loss": 0.5316, "step": 5399 }, { "epoch": 1.94, "learning_rate": 5.192549937016376e-06, "loss": 0.4869, "step": 5400 }, { "epoch": 1.94, "learning_rate": 5.191650170955552e-06, "loss": 0.7003, "step": 5401 }, { "epoch": 1.94, "learning_rate": 5.190750404894728e-06, "loss": 0.6776, "step": 5402 }, { "epoch": 1.94, "learning_rate": 5.189850638833903e-06, "loss": 0.5918, "step": 5403 }, { "epoch": 1.94, "learning_rate": 5.18895087277308e-06, "loss": 0.6677, "step": 5404 }, { "epoch": 1.94, "learning_rate": 5.188051106712255e-06, "loss": 0.8004, "step": 5405 }, { "epoch": 1.94, "learning_rate": 5.18715134065143e-06, "loss": 0.654, "step": 5406 }, { "epoch": 1.94, "learning_rate": 5.186251574590607e-06, "loss": 0.512, "step": 5407 }, { "epoch": 1.94, "learning_rate": 5.185351808529783e-06, "loss": 0.6884, "step": 5408 }, { "epoch": 1.94, "learning_rate": 5.1844520424689585e-06, "loss": 0.6197, "step": 5409 }, { "epoch": 1.94, "learning_rate": 5.183552276408135e-06, "loss": 0.6348, "step": 5410 }, { "epoch": 1.94, "learning_rate": 5.18265251034731e-06, "loss": 0.5912, "step": 5411 }, { "epoch": 1.94, "learning_rate": 5.181752744286486e-06, "loss": 0.5122, "step": 5412 }, { "epoch": 1.94, "learning_rate": 5.180852978225662e-06, "loss": 0.5538, "step": 5413 }, { "epoch": 1.94, "learning_rate": 5.179953212164837e-06, "loss": 0.6125, "step": 5414 }, { "epoch": 1.94, "learning_rate": 5.179053446104013e-06, "loss": 0.6167, "step": 5415 }, { "epoch": 1.94, "learning_rate": 5.17815368004319e-06, "loss": 0.5851, "step": 5416 }, { "epoch": 1.94, "learning_rate": 5.1772539139823655e-06, "loss": 0.7195, "step": 5417 }, { "epoch": 1.94, "learning_rate": 5.17635414792154e-06, "loss": 0.6434, "step": 5418 }, { "epoch": 1.95, "learning_rate": 5.175454381860716e-06, "loss": 0.6487, "step": 5419 }, { "epoch": 1.95, "learning_rate": 5.174554615799893e-06, "loss": 0.7212, "step": 5420 }, { "epoch": 1.95, "learning_rate": 5.1736548497390686e-06, "loss": 0.6493, "step": 5421 }, { "epoch": 1.95, "learning_rate": 5.1727550836782435e-06, "loss": 0.6113, "step": 5422 }, { "epoch": 1.95, "learning_rate": 5.17185531761742e-06, "loss": 0.7763, "step": 5423 }, { "epoch": 1.95, "learning_rate": 5.170955551556596e-06, "loss": 0.5552, "step": 5424 }, { "epoch": 1.95, "learning_rate": 5.170055785495771e-06, "loss": 0.6972, "step": 5425 }, { "epoch": 1.95, "learning_rate": 5.169156019434947e-06, "loss": 0.5957, "step": 5426 }, { "epoch": 1.95, "learning_rate": 5.168256253374123e-06, "loss": 0.669, "step": 5427 }, { "epoch": 1.95, "learning_rate": 5.167356487313299e-06, "loss": 0.6917, "step": 5428 }, { "epoch": 1.95, "learning_rate": 5.1664567212524756e-06, "loss": 0.713, "step": 5429 }, { "epoch": 1.95, "learning_rate": 5.1655569551916505e-06, "loss": 0.6028, "step": 5430 }, { "epoch": 1.95, "learning_rate": 5.164657189130826e-06, "loss": 0.6165, "step": 5431 }, { "epoch": 1.95, "learning_rate": 5.163757423070003e-06, "loss": 0.5626, "step": 5432 }, { "epoch": 1.95, "learning_rate": 5.162857657009178e-06, "loss": 0.5515, "step": 5433 }, { "epoch": 1.95, "learning_rate": 5.1619578909483536e-06, "loss": 0.6161, "step": 5434 }, { "epoch": 1.95, "learning_rate": 5.16105812488753e-06, "loss": 0.6609, "step": 5435 }, { "epoch": 1.95, "learning_rate": 5.160158358826706e-06, "loss": 0.6569, "step": 5436 }, { "epoch": 1.95, "learning_rate": 5.159258592765881e-06, "loss": 0.5858, "step": 5437 }, { "epoch": 1.95, "learning_rate": 5.158358826705057e-06, "loss": 0.6456, "step": 5438 }, { "epoch": 1.95, "learning_rate": 5.157459060644233e-06, "loss": 0.5437, "step": 5439 }, { "epoch": 1.95, "learning_rate": 5.156559294583409e-06, "loss": 0.5356, "step": 5440 }, { "epoch": 1.95, "learning_rate": 5.155659528522584e-06, "loss": 0.7074, "step": 5441 }, { "epoch": 1.95, "learning_rate": 5.1547597624617606e-06, "loss": 0.634, "step": 5442 }, { "epoch": 1.95, "learning_rate": 5.153859996400936e-06, "loss": 0.6072, "step": 5443 }, { "epoch": 1.95, "learning_rate": 5.152960230340111e-06, "loss": 0.6868, "step": 5444 }, { "epoch": 1.95, "learning_rate": 5.152060464279288e-06, "loss": 0.5488, "step": 5445 }, { "epoch": 1.95, "learning_rate": 5.151160698218464e-06, "loss": 0.5382, "step": 5446 }, { "epoch": 1.96, "learning_rate": 5.150260932157639e-06, "loss": 0.6712, "step": 5447 }, { "epoch": 1.96, "learning_rate": 5.149361166096816e-06, "loss": 0.6229, "step": 5448 }, { "epoch": 1.96, "learning_rate": 5.148461400035991e-06, "loss": 0.6504, "step": 5449 }, { "epoch": 1.96, "learning_rate": 5.147561633975167e-06, "loss": 0.6537, "step": 5450 }, { "epoch": 1.96, "learning_rate": 5.146661867914343e-06, "loss": 0.5708, "step": 5451 }, { "epoch": 1.96, "learning_rate": 5.145762101853518e-06, "loss": 0.5693, "step": 5452 }, { "epoch": 1.96, "learning_rate": 5.144862335792694e-06, "loss": 0.6412, "step": 5453 }, { "epoch": 1.96, "learning_rate": 5.14396256973187e-06, "loss": 0.729, "step": 5454 }, { "epoch": 1.96, "learning_rate": 5.143062803671046e-06, "loss": 0.6248, "step": 5455 }, { "epoch": 1.96, "learning_rate": 5.142163037610221e-06, "loss": 0.6369, "step": 5456 }, { "epoch": 1.96, "learning_rate": 5.141263271549397e-06, "loss": 0.5723, "step": 5457 }, { "epoch": 1.96, "learning_rate": 5.140363505488574e-06, "loss": 0.6106, "step": 5458 }, { "epoch": 1.96, "learning_rate": 5.1394637394277495e-06, "loss": 0.6398, "step": 5459 }, { "epoch": 1.96, "learning_rate": 5.138563973366924e-06, "loss": 0.5572, "step": 5460 }, { "epoch": 1.96, "learning_rate": 5.137664207306101e-06, "loss": 0.6874, "step": 5461 }, { "epoch": 1.96, "learning_rate": 5.136764441245277e-06, "loss": 0.661, "step": 5462 }, { "epoch": 1.96, "learning_rate": 5.135864675184452e-06, "loss": 0.6259, "step": 5463 }, { "epoch": 1.96, "learning_rate": 5.134964909123628e-06, "loss": 0.601, "step": 5464 }, { "epoch": 1.96, "learning_rate": 5.134065143062804e-06, "loss": 0.5768, "step": 5465 }, { "epoch": 1.96, "learning_rate": 5.13316537700198e-06, "loss": 0.6806, "step": 5466 }, { "epoch": 1.96, "learning_rate": 5.1322656109411565e-06, "loss": 0.6605, "step": 5467 }, { "epoch": 1.96, "learning_rate": 5.131365844880331e-06, "loss": 0.6829, "step": 5468 }, { "epoch": 1.96, "learning_rate": 5.130466078819507e-06, "loss": 0.5674, "step": 5469 }, { "epoch": 1.96, "learning_rate": 5.129566312758684e-06, "loss": 0.6404, "step": 5470 }, { "epoch": 1.96, "learning_rate": 5.1286665466978595e-06, "loss": 0.5593, "step": 5471 }, { "epoch": 1.96, "learning_rate": 5.1277667806370345e-06, "loss": 0.5636, "step": 5472 }, { "epoch": 1.96, "learning_rate": 5.12686701457621e-06, "loss": 0.5948, "step": 5473 }, { "epoch": 1.96, "learning_rate": 5.125967248515387e-06, "loss": 0.7246, "step": 5474 }, { "epoch": 1.97, "learning_rate": 5.125067482454562e-06, "loss": 0.5986, "step": 5475 }, { "epoch": 1.97, "learning_rate": 5.1241677163937375e-06, "loss": 0.6799, "step": 5476 }, { "epoch": 1.97, "learning_rate": 5.123267950332914e-06, "loss": 0.6881, "step": 5477 }, { "epoch": 1.97, "learning_rate": 5.12236818427209e-06, "loss": 0.5922, "step": 5478 }, { "epoch": 1.97, "learning_rate": 5.121468418211265e-06, "loss": 0.6159, "step": 5479 }, { "epoch": 1.97, "learning_rate": 5.1205686521504415e-06, "loss": 0.6266, "step": 5480 }, { "epoch": 1.97, "learning_rate": 5.119668886089617e-06, "loss": 0.6574, "step": 5481 }, { "epoch": 1.97, "learning_rate": 5.118769120028792e-06, "loss": 0.6491, "step": 5482 }, { "epoch": 1.97, "learning_rate": 5.117869353967969e-06, "loss": 0.5628, "step": 5483 }, { "epoch": 1.97, "learning_rate": 5.1169695879071445e-06, "loss": 0.6733, "step": 5484 }, { "epoch": 1.97, "learning_rate": 5.11606982184632e-06, "loss": 0.5762, "step": 5485 }, { "epoch": 1.97, "learning_rate": 5.115170055785497e-06, "loss": 0.6438, "step": 5486 }, { "epoch": 1.97, "learning_rate": 5.114270289724672e-06, "loss": 0.5689, "step": 5487 }, { "epoch": 1.97, "learning_rate": 5.113370523663848e-06, "loss": 0.5833, "step": 5488 }, { "epoch": 1.97, "learning_rate": 5.112470757603024e-06, "loss": 0.5438, "step": 5489 }, { "epoch": 1.97, "learning_rate": 5.1115709915422e-06, "loss": 0.6686, "step": 5490 }, { "epoch": 1.97, "learning_rate": 5.110671225481375e-06, "loss": 0.6377, "step": 5491 }, { "epoch": 1.97, "learning_rate": 5.109771459420551e-06, "loss": 0.6474, "step": 5492 }, { "epoch": 1.97, "learning_rate": 5.108871693359727e-06, "loss": 0.6896, "step": 5493 }, { "epoch": 1.97, "learning_rate": 5.107971927298902e-06, "loss": 0.6735, "step": 5494 }, { "epoch": 1.97, "learning_rate": 5.107072161238078e-06, "loss": 0.678, "step": 5495 }, { "epoch": 1.97, "learning_rate": 5.106172395177255e-06, "loss": 0.5461, "step": 5496 }, { "epoch": 1.97, "learning_rate": 5.10527262911643e-06, "loss": 0.6323, "step": 5497 }, { "epoch": 1.97, "learning_rate": 5.104372863055605e-06, "loss": 0.6072, "step": 5498 }, { "epoch": 1.97, "learning_rate": 5.103473096994782e-06, "loss": 0.5744, "step": 5499 }, { "epoch": 1.97, "learning_rate": 5.102573330933958e-06, "loss": 0.5973, "step": 5500 }, { "epoch": 1.97, "learning_rate": 5.1016735648731335e-06, "loss": 0.597, "step": 5501 }, { "epoch": 1.97, "learning_rate": 5.100773798812309e-06, "loss": 0.5799, "step": 5502 }, { "epoch": 1.98, "learning_rate": 5.099874032751485e-06, "loss": 0.5775, "step": 5503 }, { "epoch": 1.98, "learning_rate": 5.098974266690661e-06, "loss": 0.6271, "step": 5504 }, { "epoch": 1.98, "learning_rate": 5.098074500629837e-06, "loss": 0.5496, "step": 5505 }, { "epoch": 1.98, "learning_rate": 5.097174734569012e-06, "loss": 0.5696, "step": 5506 }, { "epoch": 1.98, "learning_rate": 5.096274968508188e-06, "loss": 0.564, "step": 5507 }, { "epoch": 1.98, "learning_rate": 5.095375202447365e-06, "loss": 0.5782, "step": 5508 }, { "epoch": 1.98, "learning_rate": 5.0944754363865405e-06, "loss": 0.6167, "step": 5509 }, { "epoch": 1.98, "learning_rate": 5.093575670325715e-06, "loss": 0.5605, "step": 5510 }, { "epoch": 1.98, "learning_rate": 5.092675904264891e-06, "loss": 0.5663, "step": 5511 }, { "epoch": 1.98, "learning_rate": 5.091776138204068e-06, "loss": 0.6471, "step": 5512 }, { "epoch": 1.98, "learning_rate": 5.090876372143243e-06, "loss": 0.5151, "step": 5513 }, { "epoch": 1.98, "learning_rate": 5.0899766060824184e-06, "loss": 0.596, "step": 5514 }, { "epoch": 1.98, "learning_rate": 5.089076840021595e-06, "loss": 0.6954, "step": 5515 }, { "epoch": 1.98, "learning_rate": 5.088177073960771e-06, "loss": 0.6584, "step": 5516 }, { "epoch": 1.98, "learning_rate": 5.087277307899946e-06, "loss": 0.6294, "step": 5517 }, { "epoch": 1.98, "learning_rate": 5.086377541839122e-06, "loss": 0.6138, "step": 5518 }, { "epoch": 1.98, "learning_rate": 5.085477775778298e-06, "loss": 0.5802, "step": 5519 }, { "epoch": 1.98, "learning_rate": 5.084578009717474e-06, "loss": 0.6122, "step": 5520 }, { "epoch": 1.98, "learning_rate": 5.08367824365665e-06, "loss": 0.708, "step": 5521 }, { "epoch": 1.98, "learning_rate": 5.0827784775958254e-06, "loss": 0.5881, "step": 5522 }, { "epoch": 1.98, "learning_rate": 5.081878711535001e-06, "loss": 0.5498, "step": 5523 }, { "epoch": 1.98, "learning_rate": 5.080978945474178e-06, "loss": 0.5959, "step": 5524 }, { "epoch": 1.98, "learning_rate": 5.080079179413353e-06, "loss": 0.5691, "step": 5525 }, { "epoch": 1.98, "learning_rate": 5.0791794133525285e-06, "loss": 0.7251, "step": 5526 }, { "epoch": 1.98, "learning_rate": 5.078279647291705e-06, "loss": 0.6426, "step": 5527 }, { "epoch": 1.98, "learning_rate": 5.077379881230881e-06, "loss": 0.6164, "step": 5528 }, { "epoch": 1.98, "learning_rate": 5.076480115170056e-06, "loss": 0.6463, "step": 5529 }, { "epoch": 1.98, "learning_rate": 5.075580349109232e-06, "loss": 0.5727, "step": 5530 }, { "epoch": 1.99, "learning_rate": 5.074680583048408e-06, "loss": 0.499, "step": 5531 }, { "epoch": 1.99, "learning_rate": 5.073780816987583e-06, "loss": 0.6551, "step": 5532 }, { "epoch": 1.99, "learning_rate": 5.072881050926759e-06, "loss": 0.6731, "step": 5533 }, { "epoch": 1.99, "learning_rate": 5.0719812848659355e-06, "loss": 0.656, "step": 5534 }, { "epoch": 1.99, "learning_rate": 5.071081518805111e-06, "loss": 0.6302, "step": 5535 }, { "epoch": 1.99, "learning_rate": 5.070181752744286e-06, "loss": 0.5865, "step": 5536 }, { "epoch": 1.99, "learning_rate": 5.069281986683463e-06, "loss": 0.6265, "step": 5537 }, { "epoch": 1.99, "learning_rate": 5.068382220622639e-06, "loss": 0.6176, "step": 5538 }, { "epoch": 1.99, "learning_rate": 5.067482454561814e-06, "loss": 0.6548, "step": 5539 }, { "epoch": 1.99, "learning_rate": 5.06658268850099e-06, "loss": 0.5415, "step": 5540 }, { "epoch": 1.99, "learning_rate": 5.065682922440166e-06, "loss": 0.6829, "step": 5541 }, { "epoch": 1.99, "learning_rate": 5.064783156379342e-06, "loss": 0.5942, "step": 5542 }, { "epoch": 1.99, "learning_rate": 5.063883390318518e-06, "loss": 0.5256, "step": 5543 }, { "epoch": 1.99, "learning_rate": 5.062983624257693e-06, "loss": 0.6145, "step": 5544 }, { "epoch": 1.99, "learning_rate": 5.062083858196869e-06, "loss": 0.6508, "step": 5545 }, { "epoch": 1.99, "learning_rate": 5.061184092136046e-06, "loss": 0.6272, "step": 5546 }, { "epoch": 1.99, "learning_rate": 5.060284326075221e-06, "loss": 0.4861, "step": 5547 }, { "epoch": 1.99, "learning_rate": 5.059384560014396e-06, "loss": 0.5656, "step": 5548 }, { "epoch": 1.99, "learning_rate": 5.058484793953572e-06, "loss": 0.6215, "step": 5549 }, { "epoch": 1.99, "learning_rate": 5.057585027892749e-06, "loss": 0.661, "step": 5550 }, { "epoch": 1.99, "learning_rate": 5.056685261831924e-06, "loss": 0.5498, "step": 5551 }, { "epoch": 1.99, "learning_rate": 5.055785495771099e-06, "loss": 0.6059, "step": 5552 }, { "epoch": 1.99, "learning_rate": 5.054885729710276e-06, "loss": 0.7134, "step": 5553 }, { "epoch": 1.99, "learning_rate": 5.053985963649452e-06, "loss": 0.6116, "step": 5554 }, { "epoch": 1.99, "learning_rate": 5.053086197588627e-06, "loss": 0.6263, "step": 5555 }, { "epoch": 1.99, "learning_rate": 5.052186431527803e-06, "loss": 0.5694, "step": 5556 }, { "epoch": 1.99, "learning_rate": 5.051286665466979e-06, "loss": 0.6142, "step": 5557 }, { "epoch": 1.99, "learning_rate": 5.050386899406155e-06, "loss": 0.634, "step": 5558 }, { "epoch": 2.0, "learning_rate": 5.0494871333453314e-06, "loss": 0.6507, "step": 5559 }, { "epoch": 2.0, "learning_rate": 5.048587367284506e-06, "loss": 0.6418, "step": 5560 }, { "epoch": 2.0, "learning_rate": 5.047687601223682e-06, "loss": 0.6663, "step": 5561 }, { "epoch": 2.0, "learning_rate": 5.046787835162859e-06, "loss": 0.5766, "step": 5562 }, { "epoch": 2.0, "learning_rate": 5.045888069102034e-06, "loss": 0.6468, "step": 5563 }, { "epoch": 2.0, "learning_rate": 5.0449883030412094e-06, "loss": 0.573, "step": 5564 }, { "epoch": 2.0, "learning_rate": 5.044088536980386e-06, "loss": 0.5582, "step": 5565 }, { "epoch": 2.0, "learning_rate": 5.043188770919562e-06, "loss": 0.6528, "step": 5566 }, { "epoch": 2.0, "learning_rate": 5.042289004858737e-06, "loss": 0.6051, "step": 5567 }, { "epoch": 2.0, "learning_rate": 5.0413892387979125e-06, "loss": 0.5977, "step": 5568 }, { "epoch": 2.0, "learning_rate": 5.040489472737089e-06, "loss": 0.5875, "step": 5569 }, { "epoch": 2.0, "learning_rate": 5.039589706676264e-06, "loss": 0.5624, "step": 5570 }, { "epoch": 2.0, "learning_rate": 5.03868994061544e-06, "loss": 0.6093, "step": 5571 }, { "epoch": 2.0, "learning_rate": 5.0377901745546164e-06, "loss": 0.59, "step": 5572 }, { "epoch": 2.0, "learning_rate": 5.036890408493792e-06, "loss": 0.3883, "step": 5573 }, { "epoch": 2.0, "learning_rate": 5.035990642432967e-06, "loss": 0.3772, "step": 5574 }, { "epoch": 2.0, "learning_rate": 5.035090876372144e-06, "loss": 0.4457, "step": 5575 }, { "epoch": 2.0, "learning_rate": 5.0341911103113195e-06, "loss": 0.3178, "step": 5576 }, { "epoch": 2.0, "learning_rate": 5.033291344250495e-06, "loss": 0.3424, "step": 5577 }, { "epoch": 2.0, "learning_rate": 5.032391578189672e-06, "loss": 0.3445, "step": 5578 }, { "epoch": 2.0, "learning_rate": 5.031491812128847e-06, "loss": 0.3231, "step": 5579 }, { "epoch": 2.0, "learning_rate": 5.0305920460680226e-06, "loss": 0.3289, "step": 5580 }, { "epoch": 2.0, "learning_rate": 5.029692280007199e-06, "loss": 0.2466, "step": 5581 }, { "epoch": 2.0, "learning_rate": 5.028792513946374e-06, "loss": 0.3474, "step": 5582 }, { "epoch": 2.0, "learning_rate": 5.02789274788555e-06, "loss": 0.3508, "step": 5583 }, { "epoch": 2.0, "learning_rate": 5.0269929818247265e-06, "loss": 0.3751, "step": 5584 }, { "epoch": 2.0, "learning_rate": 5.026093215763902e-06, "loss": 0.3746, "step": 5585 }, { "epoch": 2.01, "learning_rate": 5.025193449703077e-06, "loss": 0.3559, "step": 5586 }, { "epoch": 2.01, "learning_rate": 5.024293683642253e-06, "loss": 0.3506, "step": 5587 }, { "epoch": 2.01, "learning_rate": 5.0233939175814296e-06, "loss": 0.3372, "step": 5588 }, { "epoch": 2.01, "learning_rate": 5.022494151520605e-06, "loss": 0.3976, "step": 5589 }, { "epoch": 2.01, "learning_rate": 5.02159438545978e-06, "loss": 0.3276, "step": 5590 }, { "epoch": 2.01, "learning_rate": 5.020694619398957e-06, "loss": 0.3774, "step": 5591 }, { "epoch": 2.01, "learning_rate": 5.019794853338133e-06, "loss": 0.3823, "step": 5592 }, { "epoch": 2.01, "learning_rate": 5.0188950872773076e-06, "loss": 0.3255, "step": 5593 }, { "epoch": 2.01, "learning_rate": 5.017995321216484e-06, "loss": 0.3862, "step": 5594 }, { "epoch": 2.01, "learning_rate": 5.01709555515566e-06, "loss": 0.3304, "step": 5595 }, { "epoch": 2.01, "learning_rate": 5.016195789094836e-06, "loss": 0.4453, "step": 5596 }, { "epoch": 2.01, "learning_rate": 5.015296023034012e-06, "loss": 0.3284, "step": 5597 }, { "epoch": 2.01, "learning_rate": 5.014396256973187e-06, "loss": 0.3373, "step": 5598 }, { "epoch": 2.01, "learning_rate": 5.013496490912363e-06, "loss": 0.356, "step": 5599 }, { "epoch": 2.01, "learning_rate": 5.01259672485154e-06, "loss": 0.4031, "step": 5600 }, { "epoch": 2.01, "learning_rate": 5.0116969587907146e-06, "loss": 0.3762, "step": 5601 }, { "epoch": 2.01, "learning_rate": 5.01079719272989e-06, "loss": 0.289, "step": 5602 }, { "epoch": 2.01, "learning_rate": 5.009897426669067e-06, "loss": 0.3155, "step": 5603 }, { "epoch": 2.01, "learning_rate": 5.008997660608243e-06, "loss": 0.3274, "step": 5604 }, { "epoch": 2.01, "learning_rate": 5.008097894547418e-06, "loss": 0.3623, "step": 5605 }, { "epoch": 2.01, "learning_rate": 5.007198128486593e-06, "loss": 0.3046, "step": 5606 }, { "epoch": 2.01, "learning_rate": 5.00629836242577e-06, "loss": 0.2926, "step": 5607 }, { "epoch": 2.01, "learning_rate": 5.005398596364946e-06, "loss": 0.3298, "step": 5608 }, { "epoch": 2.01, "learning_rate": 5.004498830304121e-06, "loss": 0.3264, "step": 5609 }, { "epoch": 2.01, "learning_rate": 5.003599064243297e-06, "loss": 0.3207, "step": 5610 }, { "epoch": 2.01, "learning_rate": 5.002699298182473e-06, "loss": 0.3101, "step": 5611 }, { "epoch": 2.01, "learning_rate": 5.001799532121648e-06, "loss": 0.2802, "step": 5612 }, { "epoch": 2.01, "learning_rate": 5.000899766060825e-06, "loss": 0.3494, "step": 5613 }, { "epoch": 2.02, "learning_rate": 5e-06, "loss": 0.3154, "step": 5614 }, { "epoch": 2.02, "learning_rate": 4.999100233939176e-06, "loss": 0.3442, "step": 5615 }, { "epoch": 2.02, "learning_rate": 4.998200467878352e-06, "loss": 0.4062, "step": 5616 }, { "epoch": 2.02, "learning_rate": 4.997300701817528e-06, "loss": 0.3609, "step": 5617 }, { "epoch": 2.02, "learning_rate": 4.996400935756704e-06, "loss": 0.3417, "step": 5618 }, { "epoch": 2.02, "learning_rate": 4.995501169695879e-06, "loss": 0.3006, "step": 5619 }, { "epoch": 2.02, "learning_rate": 4.994601403635055e-06, "loss": 0.3746, "step": 5620 }, { "epoch": 2.02, "learning_rate": 4.993701637574231e-06, "loss": 0.3198, "step": 5621 }, { "epoch": 2.02, "learning_rate": 4.9928018715134066e-06, "loss": 0.3517, "step": 5622 }, { "epoch": 2.02, "learning_rate": 4.991902105452583e-06, "loss": 0.3779, "step": 5623 }, { "epoch": 2.02, "learning_rate": 4.991002339391758e-06, "loss": 0.3001, "step": 5624 }, { "epoch": 2.02, "learning_rate": 4.990102573330935e-06, "loss": 0.3255, "step": 5625 }, { "epoch": 2.02, "learning_rate": 4.9892028072701105e-06, "loss": 0.3247, "step": 5626 }, { "epoch": 2.02, "learning_rate": 4.988303041209286e-06, "loss": 0.2899, "step": 5627 }, { "epoch": 2.02, "learning_rate": 4.987403275148462e-06, "loss": 0.3071, "step": 5628 }, { "epoch": 2.02, "learning_rate": 4.986503509087637e-06, "loss": 0.3235, "step": 5629 }, { "epoch": 2.02, "learning_rate": 4.9856037430268136e-06, "loss": 0.3891, "step": 5630 }, { "epoch": 2.02, "learning_rate": 4.984703976965989e-06, "loss": 0.3356, "step": 5631 }, { "epoch": 2.02, "learning_rate": 4.983804210905165e-06, "loss": 0.3768, "step": 5632 }, { "epoch": 2.02, "learning_rate": 4.982904444844341e-06, "loss": 0.3052, "step": 5633 }, { "epoch": 2.02, "learning_rate": 4.982004678783517e-06, "loss": 0.3441, "step": 5634 }, { "epoch": 2.02, "learning_rate": 4.981104912722692e-06, "loss": 0.3421, "step": 5635 }, { "epoch": 2.02, "learning_rate": 4.980205146661868e-06, "loss": 0.2341, "step": 5636 }, { "epoch": 2.02, "learning_rate": 4.979305380601045e-06, "loss": 0.3114, "step": 5637 }, { "epoch": 2.02, "learning_rate": 4.97840561454022e-06, "loss": 0.3844, "step": 5638 }, { "epoch": 2.02, "learning_rate": 4.9775058484793955e-06, "loss": 0.3463, "step": 5639 }, { "epoch": 2.02, "learning_rate": 4.976606082418571e-06, "loss": 0.2851, "step": 5640 }, { "epoch": 2.02, "learning_rate": 4.975706316357747e-06, "loss": 0.4084, "step": 5641 }, { "epoch": 2.03, "learning_rate": 4.974806550296924e-06, "loss": 0.3637, "step": 5642 }, { "epoch": 2.03, "learning_rate": 4.9739067842360986e-06, "loss": 0.3973, "step": 5643 }, { "epoch": 2.03, "learning_rate": 4.973007018175275e-06, "loss": 0.4016, "step": 5644 }, { "epoch": 2.03, "learning_rate": 4.972107252114451e-06, "loss": 0.394, "step": 5645 }, { "epoch": 2.03, "learning_rate": 4.971207486053627e-06, "loss": 0.3208, "step": 5646 }, { "epoch": 2.03, "learning_rate": 4.9703077199928025e-06, "loss": 0.2971, "step": 5647 }, { "epoch": 2.03, "learning_rate": 4.969407953931978e-06, "loss": 0.4458, "step": 5648 }, { "epoch": 2.03, "learning_rate": 4.968508187871154e-06, "loss": 0.3423, "step": 5649 }, { "epoch": 2.03, "learning_rate": 4.96760842181033e-06, "loss": 0.3609, "step": 5650 }, { "epoch": 2.03, "learning_rate": 4.9667086557495055e-06, "loss": 0.4178, "step": 5651 }, { "epoch": 2.03, "learning_rate": 4.965808889688681e-06, "loss": 0.3642, "step": 5652 }, { "epoch": 2.03, "learning_rate": 4.964909123627857e-06, "loss": 0.3779, "step": 5653 }, { "epoch": 2.03, "learning_rate": 4.964009357567033e-06, "loss": 0.3052, "step": 5654 }, { "epoch": 2.03, "learning_rate": 4.963109591506209e-06, "loss": 0.4019, "step": 5655 }, { "epoch": 2.03, "learning_rate": 4.962209825445385e-06, "loss": 0.3338, "step": 5656 }, { "epoch": 2.03, "learning_rate": 4.96131005938456e-06, "loss": 0.2845, "step": 5657 }, { "epoch": 2.03, "learning_rate": 4.960410293323737e-06, "loss": 0.3042, "step": 5658 }, { "epoch": 2.03, "learning_rate": 4.959510527262912e-06, "loss": 0.3904, "step": 5659 }, { "epoch": 2.03, "learning_rate": 4.9586107612020875e-06, "loss": 0.3388, "step": 5660 }, { "epoch": 2.03, "learning_rate": 4.957710995141264e-06, "loss": 0.2779, "step": 5661 }, { "epoch": 2.03, "learning_rate": 4.956811229080439e-06, "loss": 0.3381, "step": 5662 }, { "epoch": 2.03, "learning_rate": 4.955911463019616e-06, "loss": 0.4502, "step": 5663 }, { "epoch": 2.03, "learning_rate": 4.955011696958791e-06, "loss": 0.3297, "step": 5664 }, { "epoch": 2.03, "learning_rate": 4.954111930897967e-06, "loss": 0.3835, "step": 5665 }, { "epoch": 2.03, "learning_rate": 4.953212164837143e-06, "loss": 0.3319, "step": 5666 }, { "epoch": 2.03, "learning_rate": 4.952312398776319e-06, "loss": 0.3863, "step": 5667 }, { "epoch": 2.03, "learning_rate": 4.9514126327154945e-06, "loss": 0.3311, "step": 5668 }, { "epoch": 2.03, "learning_rate": 4.95051286665467e-06, "loss": 0.3292, "step": 5669 }, { "epoch": 2.04, "learning_rate": 4.949613100593846e-06, "loss": 0.2886, "step": 5670 }, { "epoch": 2.04, "learning_rate": 4.948713334533022e-06, "loss": 0.3392, "step": 5671 }, { "epoch": 2.04, "learning_rate": 4.9478135684721975e-06, "loss": 0.319, "step": 5672 }, { "epoch": 2.04, "learning_rate": 4.9478135684721975e-06, "loss": 0.4409, "step": 5673 }, { "epoch": 2.04, "learning_rate": 4.946913802411373e-06, "loss": 0.2816, "step": 5674 }, { "epoch": 2.04, "learning_rate": 4.946014036350549e-06, "loss": 0.3193, "step": 5675 }, { "epoch": 2.04, "learning_rate": 4.945114270289726e-06, "loss": 0.295, "step": 5676 }, { "epoch": 2.04, "learning_rate": 4.944214504228901e-06, "loss": 0.3626, "step": 5677 }, { "epoch": 2.04, "learning_rate": 4.943314738168077e-06, "loss": 0.3382, "step": 5678 }, { "epoch": 2.04, "learning_rate": 4.942414972107252e-06, "loss": 0.3254, "step": 5679 }, { "epoch": 2.04, "learning_rate": 4.941515206046428e-06, "loss": 0.3369, "step": 5680 }, { "epoch": 2.04, "learning_rate": 4.9406154399856045e-06, "loss": 0.3041, "step": 5681 }, { "epoch": 2.04, "learning_rate": 4.9397156739247795e-06, "loss": 0.3599, "step": 5682 }, { "epoch": 2.04, "learning_rate": 4.938815907863956e-06, "loss": 0.3809, "step": 5683 }, { "epoch": 2.04, "learning_rate": 4.937916141803132e-06, "loss": 0.3369, "step": 5684 }, { "epoch": 2.04, "learning_rate": 4.937016375742308e-06, "loss": 0.3951, "step": 5685 }, { "epoch": 2.04, "learning_rate": 4.936116609681483e-06, "loss": 0.2733, "step": 5686 }, { "epoch": 2.04, "learning_rate": 4.935216843620659e-06, "loss": 0.303, "step": 5687 }, { "epoch": 2.04, "learning_rate": 4.934317077559835e-06, "loss": 0.3157, "step": 5688 }, { "epoch": 2.04, "learning_rate": 4.933417311499011e-06, "loss": 0.3501, "step": 5689 }, { "epoch": 2.04, "learning_rate": 4.9325175454381865e-06, "loss": 0.3534, "step": 5690 }, { "epoch": 2.04, "learning_rate": 4.931617779377362e-06, "loss": 0.3337, "step": 5691 }, { "epoch": 2.04, "learning_rate": 4.930718013316538e-06, "loss": 0.3147, "step": 5692 }, { "epoch": 2.04, "learning_rate": 4.929818247255714e-06, "loss": 0.359, "step": 5693 }, { "epoch": 2.04, "learning_rate": 4.9289184811948895e-06, "loss": 0.2811, "step": 5694 }, { "epoch": 2.04, "learning_rate": 4.928018715134066e-06, "loss": 0.3535, "step": 5695 }, { "epoch": 2.04, "learning_rate": 4.927118949073241e-06, "loss": 0.299, "step": 5696 }, { "epoch": 2.04, "learning_rate": 4.926219183012418e-06, "loss": 0.3387, "step": 5697 }, { "epoch": 2.05, "learning_rate": 4.925319416951593e-06, "loss": 0.3893, "step": 5698 }, { "epoch": 2.05, "learning_rate": 4.924419650890768e-06, "loss": 0.3455, "step": 5699 }, { "epoch": 2.05, "learning_rate": 4.923519884829945e-06, "loss": 0.305, "step": 5700 }, { "epoch": 2.05, "learning_rate": 4.92262011876912e-06, "loss": 0.3303, "step": 5701 }, { "epoch": 2.05, "learning_rate": 4.9217203527082965e-06, "loss": 0.3224, "step": 5702 }, { "epoch": 2.05, "learning_rate": 4.920820586647472e-06, "loss": 0.3377, "step": 5703 }, { "epoch": 2.05, "learning_rate": 4.919920820586648e-06, "loss": 0.2686, "step": 5704 }, { "epoch": 2.05, "learning_rate": 4.919021054525824e-06, "loss": 0.3023, "step": 5705 }, { "epoch": 2.05, "learning_rate": 4.918121288465e-06, "loss": 0.3125, "step": 5706 }, { "epoch": 2.05, "learning_rate": 4.917221522404175e-06, "loss": 0.3316, "step": 5707 }, { "epoch": 2.05, "learning_rate": 4.916321756343351e-06, "loss": 0.3019, "step": 5708 }, { "epoch": 2.05, "learning_rate": 4.915421990282527e-06, "loss": 0.3389, "step": 5709 }, { "epoch": 2.05, "learning_rate": 4.914522224221703e-06, "loss": 0.3583, "step": 5710 }, { "epoch": 2.05, "learning_rate": 4.9136224581608784e-06, "loss": 0.3088, "step": 5711 }, { "epoch": 2.05, "learning_rate": 4.912722692100054e-06, "loss": 0.2821, "step": 5712 }, { "epoch": 2.05, "learning_rate": 4.91182292603923e-06, "loss": 0.2794, "step": 5713 }, { "epoch": 2.05, "learning_rate": 4.910923159978406e-06, "loss": 0.3384, "step": 5714 }, { "epoch": 2.05, "learning_rate": 4.9100233939175815e-06, "loss": 0.3583, "step": 5715 }, { "epoch": 2.05, "learning_rate": 4.909123627856758e-06, "loss": 0.3592, "step": 5716 }, { "epoch": 2.05, "learning_rate": 4.908223861795933e-06, "loss": 0.3734, "step": 5717 }, { "epoch": 2.05, "learning_rate": 4.90732409573511e-06, "loss": 0.422, "step": 5718 }, { "epoch": 2.05, "learning_rate": 4.9064243296742854e-06, "loss": 0.3489, "step": 5719 }, { "epoch": 2.05, "learning_rate": 4.90552456361346e-06, "loss": 0.3001, "step": 5720 }, { "epoch": 2.05, "learning_rate": 4.904624797552637e-06, "loss": 0.3398, "step": 5721 }, { "epoch": 2.05, "learning_rate": 4.903725031491813e-06, "loss": 0.3295, "step": 5722 }, { "epoch": 2.05, "learning_rate": 4.9028252654309885e-06, "loss": 0.3865, "step": 5723 }, { "epoch": 2.05, "learning_rate": 4.901925499370164e-06, "loss": 0.3305, "step": 5724 }, { "epoch": 2.05, "learning_rate": 4.90102573330934e-06, "loss": 0.3748, "step": 5725 }, { "epoch": 2.06, "learning_rate": 4.900125967248516e-06, "loss": 0.3026, "step": 5726 }, { "epoch": 2.06, "learning_rate": 4.899226201187692e-06, "loss": 0.2835, "step": 5727 }, { "epoch": 2.06, "learning_rate": 4.898326435126867e-06, "loss": 0.332, "step": 5728 }, { "epoch": 2.06, "learning_rate": 4.897426669066043e-06, "loss": 0.3659, "step": 5729 }, { "epoch": 2.06, "learning_rate": 4.896526903005219e-06, "loss": 0.4227, "step": 5730 }, { "epoch": 2.06, "learning_rate": 4.895627136944395e-06, "loss": 0.383, "step": 5731 }, { "epoch": 2.06, "learning_rate": 4.8947273708835704e-06, "loss": 0.3358, "step": 5732 }, { "epoch": 2.06, "learning_rate": 4.893827604822746e-06, "loss": 0.3071, "step": 5733 }, { "epoch": 2.06, "learning_rate": 4.892927838761922e-06, "loss": 0.302, "step": 5734 }, { "epoch": 2.06, "learning_rate": 4.892028072701099e-06, "loss": 0.4041, "step": 5735 }, { "epoch": 2.06, "learning_rate": 4.8911283066402735e-06, "loss": 0.35, "step": 5736 }, { "epoch": 2.06, "learning_rate": 4.89022854057945e-06, "loss": 0.3154, "step": 5737 }, { "epoch": 2.06, "learning_rate": 4.889328774518626e-06, "loss": 0.3417, "step": 5738 }, { "epoch": 2.06, "learning_rate": 4.888429008457801e-06, "loss": 0.3797, "step": 5739 }, { "epoch": 2.06, "learning_rate": 4.8875292423969774e-06, "loss": 0.3036, "step": 5740 }, { "epoch": 2.06, "learning_rate": 4.886629476336153e-06, "loss": 0.3515, "step": 5741 }, { "epoch": 2.06, "learning_rate": 4.885729710275329e-06, "loss": 0.3944, "step": 5742 }, { "epoch": 2.06, "learning_rate": 4.884829944214505e-06, "loss": 0.3352, "step": 5743 }, { "epoch": 2.06, "learning_rate": 4.8839301781536805e-06, "loss": 0.3148, "step": 5744 }, { "epoch": 2.06, "learning_rate": 4.883030412092856e-06, "loss": 0.3287, "step": 5745 }, { "epoch": 2.06, "learning_rate": 4.882130646032032e-06, "loss": 0.3614, "step": 5746 }, { "epoch": 2.06, "learning_rate": 4.881230879971208e-06, "loss": 0.395, "step": 5747 }, { "epoch": 2.06, "learning_rate": 4.880331113910384e-06, "loss": 0.3234, "step": 5748 }, { "epoch": 2.06, "learning_rate": 4.879431347849559e-06, "loss": 0.3388, "step": 5749 }, { "epoch": 2.06, "learning_rate": 4.878531581788735e-06, "loss": 0.3605, "step": 5750 }, { "epoch": 2.06, "learning_rate": 4.877631815727911e-06, "loss": 0.2996, "step": 5751 }, { "epoch": 2.06, "learning_rate": 4.876732049667087e-06, "loss": 0.3609, "step": 5752 }, { "epoch": 2.06, "learning_rate": 4.8758322836062624e-06, "loss": 0.3166, "step": 5753 }, { "epoch": 2.07, "learning_rate": 4.874932517545439e-06, "loss": 0.3437, "step": 5754 }, { "epoch": 2.07, "learning_rate": 4.874032751484614e-06, "loss": 0.2698, "step": 5755 }, { "epoch": 2.07, "learning_rate": 4.873132985423791e-06, "loss": 0.3096, "step": 5756 }, { "epoch": 2.07, "learning_rate": 4.872233219362966e-06, "loss": 0.3306, "step": 5757 }, { "epoch": 2.07, "learning_rate": 4.871333453302141e-06, "loss": 0.3274, "step": 5758 }, { "epoch": 2.07, "learning_rate": 4.870433687241318e-06, "loss": 0.3057, "step": 5759 }, { "epoch": 2.07, "learning_rate": 4.869533921180494e-06, "loss": 0.2889, "step": 5760 }, { "epoch": 2.07, "learning_rate": 4.8686341551196694e-06, "loss": 0.3585, "step": 5761 }, { "epoch": 2.07, "learning_rate": 4.867734389058845e-06, "loss": 0.4065, "step": 5762 }, { "epoch": 2.07, "learning_rate": 4.866834622998021e-06, "loss": 0.4139, "step": 5763 }, { "epoch": 2.07, "learning_rate": 4.865934856937197e-06, "loss": 0.3528, "step": 5764 }, { "epoch": 2.07, "learning_rate": 4.8650350908763725e-06, "loss": 0.2926, "step": 5765 }, { "epoch": 2.07, "learning_rate": 4.864135324815548e-06, "loss": 0.3046, "step": 5766 }, { "epoch": 2.07, "learning_rate": 4.863235558754724e-06, "loss": 0.3072, "step": 5767 }, { "epoch": 2.07, "learning_rate": 4.8623357926939e-06, "loss": 0.3742, "step": 5768 }, { "epoch": 2.07, "learning_rate": 4.861436026633076e-06, "loss": 0.3915, "step": 5769 }, { "epoch": 2.07, "learning_rate": 4.860536260572251e-06, "loss": 0.3773, "step": 5770 }, { "epoch": 2.07, "learning_rate": 4.859636494511427e-06, "loss": 0.3689, "step": 5771 }, { "epoch": 2.07, "learning_rate": 4.858736728450603e-06, "loss": 0.3259, "step": 5772 }, { "epoch": 2.07, "learning_rate": 4.8578369623897795e-06, "loss": 0.2739, "step": 5773 }, { "epoch": 2.07, "learning_rate": 4.856937196328954e-06, "loss": 0.385, "step": 5774 }, { "epoch": 2.07, "learning_rate": 4.856037430268131e-06, "loss": 0.3194, "step": 5775 }, { "epoch": 2.07, "learning_rate": 4.855137664207307e-06, "loss": 0.285, "step": 5776 }, { "epoch": 2.07, "learning_rate": 4.854237898146483e-06, "loss": 0.3701, "step": 5777 }, { "epoch": 2.07, "learning_rate": 4.853338132085658e-06, "loss": 0.3743, "step": 5778 }, { "epoch": 2.07, "learning_rate": 4.852438366024834e-06, "loss": 0.3561, "step": 5779 }, { "epoch": 2.07, "learning_rate": 4.85153859996401e-06, "loss": 0.3122, "step": 5780 }, { "epoch": 2.08, "learning_rate": 4.850638833903186e-06, "loss": 0.3674, "step": 5781 }, { "epoch": 2.08, "learning_rate": 4.849739067842361e-06, "loss": 0.3161, "step": 5782 }, { "epoch": 2.08, "learning_rate": 4.848839301781537e-06, "loss": 0.2973, "step": 5783 }, { "epoch": 2.08, "learning_rate": 4.847939535720713e-06, "loss": 0.3033, "step": 5784 }, { "epoch": 2.08, "learning_rate": 4.847039769659889e-06, "loss": 0.3586, "step": 5785 }, { "epoch": 2.08, "learning_rate": 4.8461400035990645e-06, "loss": 0.3733, "step": 5786 }, { "epoch": 2.08, "learning_rate": 4.84524023753824e-06, "loss": 0.2603, "step": 5787 }, { "epoch": 2.08, "learning_rate": 4.844340471477416e-06, "loss": 0.3647, "step": 5788 }, { "epoch": 2.08, "learning_rate": 4.843440705416592e-06, "loss": 0.3424, "step": 5789 }, { "epoch": 2.08, "learning_rate": 4.8425409393557676e-06, "loss": 0.2678, "step": 5790 }, { "epoch": 2.08, "learning_rate": 4.841641173294943e-06, "loss": 0.3693, "step": 5791 }, { "epoch": 2.08, "learning_rate": 4.84074140723412e-06, "loss": 0.321, "step": 5792 }, { "epoch": 2.08, "learning_rate": 4.839841641173295e-06, "loss": 0.2894, "step": 5793 }, { "epoch": 2.08, "learning_rate": 4.8389418751124715e-06, "loss": 0.3666, "step": 5794 }, { "epoch": 2.08, "learning_rate": 4.838042109051647e-06, "loss": 0.3803, "step": 5795 }, { "epoch": 2.08, "learning_rate": 4.837142342990823e-06, "loss": 0.346, "step": 5796 }, { "epoch": 2.08, "learning_rate": 4.836242576929999e-06, "loss": 0.3099, "step": 5797 }, { "epoch": 2.08, "learning_rate": 4.8353428108691746e-06, "loss": 0.3803, "step": 5798 }, { "epoch": 2.08, "learning_rate": 4.83444304480835e-06, "loss": 0.3295, "step": 5799 }, { "epoch": 2.08, "learning_rate": 4.833543278747526e-06, "loss": 0.2647, "step": 5800 }, { "epoch": 2.08, "learning_rate": 4.832643512686702e-06, "loss": 0.3899, "step": 5801 }, { "epoch": 2.08, "learning_rate": 4.831743746625878e-06, "loss": 0.3466, "step": 5802 }, { "epoch": 2.08, "learning_rate": 4.830843980565053e-06, "loss": 0.3672, "step": 5803 }, { "epoch": 2.08, "learning_rate": 4.829944214504229e-06, "loss": 0.3845, "step": 5804 }, { "epoch": 2.08, "learning_rate": 4.829044448443405e-06, "loss": 0.3056, "step": 5805 }, { "epoch": 2.08, "learning_rate": 4.8281446823825816e-06, "loss": 0.3323, "step": 5806 }, { "epoch": 2.08, "learning_rate": 4.8272449163217565e-06, "loss": 0.3081, "step": 5807 }, { "epoch": 2.08, "learning_rate": 4.826345150260932e-06, "loss": 0.3418, "step": 5808 }, { "epoch": 2.09, "learning_rate": 4.825445384200108e-06, "loss": 0.3445, "step": 5809 }, { "epoch": 2.09, "learning_rate": 4.824545618139284e-06, "loss": 0.3684, "step": 5810 }, { "epoch": 2.09, "learning_rate": 4.82364585207846e-06, "loss": 0.3217, "step": 5811 }, { "epoch": 2.09, "learning_rate": 4.822746086017635e-06, "loss": 0.3385, "step": 5812 }, { "epoch": 2.09, "learning_rate": 4.821846319956812e-06, "loss": 0.3809, "step": 5813 }, { "epoch": 2.09, "learning_rate": 4.820946553895988e-06, "loss": 0.3331, "step": 5814 }, { "epoch": 2.09, "learning_rate": 4.8200467878351635e-06, "loss": 0.3536, "step": 5815 }, { "epoch": 2.09, "learning_rate": 4.819147021774339e-06, "loss": 0.345, "step": 5816 }, { "epoch": 2.09, "learning_rate": 4.818247255713515e-06, "loss": 0.3192, "step": 5817 }, { "epoch": 2.09, "learning_rate": 4.817347489652691e-06, "loss": 0.324, "step": 5818 }, { "epoch": 2.09, "learning_rate": 4.8164477235918666e-06, "loss": 0.3739, "step": 5819 }, { "epoch": 2.09, "learning_rate": 4.815547957531042e-06, "loss": 0.2449, "step": 5820 }, { "epoch": 2.09, "learning_rate": 4.814648191470218e-06, "loss": 0.3179, "step": 5821 }, { "epoch": 2.09, "learning_rate": 4.813748425409394e-06, "loss": 0.2837, "step": 5822 }, { "epoch": 2.09, "learning_rate": 4.81284865934857e-06, "loss": 0.3679, "step": 5823 }, { "epoch": 2.09, "learning_rate": 4.811948893287745e-06, "loss": 0.344, "step": 5824 }, { "epoch": 2.09, "learning_rate": 4.811049127226922e-06, "loss": 0.3641, "step": 5825 }, { "epoch": 2.09, "learning_rate": 4.810149361166097e-06, "loss": 0.3895, "step": 5826 }, { "epoch": 2.09, "learning_rate": 4.809249595105273e-06, "loss": 0.3179, "step": 5827 }, { "epoch": 2.09, "learning_rate": 4.8083498290444485e-06, "loss": 0.3467, "step": 5828 }, { "epoch": 2.09, "learning_rate": 4.807450062983624e-06, "loss": 0.3288, "step": 5829 }, { "epoch": 2.09, "learning_rate": 4.806550296922801e-06, "loss": 0.3325, "step": 5830 }, { "epoch": 2.09, "learning_rate": 4.805650530861976e-06, "loss": 0.3336, "step": 5831 }, { "epoch": 2.09, "learning_rate": 4.804750764801152e-06, "loss": 0.3616, "step": 5832 }, { "epoch": 2.09, "learning_rate": 4.803850998740328e-06, "loss": 0.3066, "step": 5833 }, { "epoch": 2.09, "learning_rate": 4.802951232679504e-06, "loss": 0.304, "step": 5834 }, { "epoch": 2.09, "learning_rate": 4.80205146661868e-06, "loss": 0.3163, "step": 5835 }, { "epoch": 2.09, "learning_rate": 4.8011517005578555e-06, "loss": 0.35, "step": 5836 }, { "epoch": 2.1, "learning_rate": 4.800251934497031e-06, "loss": 0.3398, "step": 5837 }, { "epoch": 2.1, "learning_rate": 4.799352168436207e-06, "loss": 0.3672, "step": 5838 }, { "epoch": 2.1, "learning_rate": 4.798452402375383e-06, "loss": 0.3283, "step": 5839 }, { "epoch": 2.1, "learning_rate": 4.7975526363145586e-06, "loss": 0.3001, "step": 5840 }, { "epoch": 2.1, "learning_rate": 4.796652870253734e-06, "loss": 0.3196, "step": 5841 }, { "epoch": 2.1, "learning_rate": 4.79575310419291e-06, "loss": 0.4027, "step": 5842 }, { "epoch": 2.1, "learning_rate": 4.794853338132086e-06, "loss": 0.3823, "step": 5843 }, { "epoch": 2.1, "learning_rate": 4.7939535720712625e-06, "loss": 0.2664, "step": 5844 }, { "epoch": 2.1, "learning_rate": 4.793053806010437e-06, "loss": 0.307, "step": 5845 }, { "epoch": 2.1, "learning_rate": 4.792154039949614e-06, "loss": 0.3272, "step": 5846 }, { "epoch": 2.1, "learning_rate": 4.791254273888789e-06, "loss": 0.3102, "step": 5847 }, { "epoch": 2.1, "learning_rate": 4.790354507827965e-06, "loss": 0.4424, "step": 5848 }, { "epoch": 2.1, "learning_rate": 4.789454741767141e-06, "loss": 0.3192, "step": 5849 }, { "epoch": 2.1, "learning_rate": 4.788554975706316e-06, "loss": 0.3189, "step": 5850 }, { "epoch": 2.1, "learning_rate": 4.787655209645493e-06, "loss": 0.2957, "step": 5851 }, { "epoch": 2.1, "learning_rate": 4.786755443584669e-06, "loss": 0.2447, "step": 5852 }, { "epoch": 2.1, "learning_rate": 4.785855677523844e-06, "loss": 0.3627, "step": 5853 }, { "epoch": 2.1, "learning_rate": 4.78495591146302e-06, "loss": 0.3111, "step": 5854 }, { "epoch": 2.1, "learning_rate": 4.784056145402196e-06, "loss": 0.3847, "step": 5855 }, { "epoch": 2.1, "learning_rate": 4.783156379341372e-06, "loss": 0.3609, "step": 5856 }, { "epoch": 2.1, "learning_rate": 4.7822566132805475e-06, "loss": 0.3892, "step": 5857 }, { "epoch": 2.1, "learning_rate": 4.781356847219723e-06, "loss": 0.2906, "step": 5858 }, { "epoch": 2.1, "learning_rate": 4.780457081158899e-06, "loss": 0.4165, "step": 5859 }, { "epoch": 2.1, "learning_rate": 4.779557315098075e-06, "loss": 0.282, "step": 5860 }, { "epoch": 2.1, "learning_rate": 4.7786575490372505e-06, "loss": 0.3157, "step": 5861 }, { "epoch": 2.1, "learning_rate": 4.777757782976426e-06, "loss": 0.3866, "step": 5862 }, { "epoch": 2.1, "learning_rate": 4.776858016915603e-06, "loss": 0.3854, "step": 5863 }, { "epoch": 2.1, "learning_rate": 4.775958250854778e-06, "loss": 0.3194, "step": 5864 }, { "epoch": 2.11, "learning_rate": 4.7750584847939545e-06, "loss": 0.3529, "step": 5865 }, { "epoch": 2.11, "learning_rate": 4.774158718733129e-06, "loss": 0.3647, "step": 5866 }, { "epoch": 2.11, "learning_rate": 4.773258952672305e-06, "loss": 0.2824, "step": 5867 }, { "epoch": 2.11, "learning_rate": 4.772359186611482e-06, "loss": 0.4167, "step": 5868 }, { "epoch": 2.11, "learning_rate": 4.771459420550657e-06, "loss": 0.2928, "step": 5869 }, { "epoch": 2.11, "learning_rate": 4.770559654489833e-06, "loss": 0.3427, "step": 5870 }, { "epoch": 2.11, "learning_rate": 4.769659888429009e-06, "loss": 0.3658, "step": 5871 }, { "epoch": 2.11, "learning_rate": 4.768760122368185e-06, "loss": 0.4037, "step": 5872 }, { "epoch": 2.11, "learning_rate": 4.767860356307361e-06, "loss": 0.2841, "step": 5873 }, { "epoch": 2.11, "learning_rate": 4.766960590246536e-06, "loss": 0.3446, "step": 5874 }, { "epoch": 2.11, "learning_rate": 4.766060824185712e-06, "loss": 0.382, "step": 5875 }, { "epoch": 2.11, "learning_rate": 4.765161058124888e-06, "loss": 0.3183, "step": 5876 }, { "epoch": 2.11, "learning_rate": 4.764261292064064e-06, "loss": 0.3933, "step": 5877 }, { "epoch": 2.11, "learning_rate": 4.7633615260032395e-06, "loss": 0.331, "step": 5878 }, { "epoch": 2.11, "learning_rate": 4.762461759942415e-06, "loss": 0.3092, "step": 5879 }, { "epoch": 2.11, "learning_rate": 4.761561993881591e-06, "loss": 0.3195, "step": 5880 }, { "epoch": 2.11, "learning_rate": 4.760662227820767e-06, "loss": 0.3289, "step": 5881 }, { "epoch": 2.11, "learning_rate": 4.759762461759943e-06, "loss": 0.354, "step": 5882 }, { "epoch": 2.11, "learning_rate": 4.758862695699118e-06, "loss": 0.2927, "step": 5883 }, { "epoch": 2.11, "learning_rate": 4.757962929638295e-06, "loss": 0.2894, "step": 5884 }, { "epoch": 2.11, "learning_rate": 4.75706316357747e-06, "loss": 0.3282, "step": 5885 }, { "epoch": 2.11, "learning_rate": 4.756163397516646e-06, "loss": 0.3519, "step": 5886 }, { "epoch": 2.11, "learning_rate": 4.755263631455822e-06, "loss": 0.3017, "step": 5887 }, { "epoch": 2.11, "learning_rate": 4.754363865394997e-06, "loss": 0.3503, "step": 5888 }, { "epoch": 2.11, "learning_rate": 4.753464099334174e-06, "loss": 0.3102, "step": 5889 }, { "epoch": 2.11, "learning_rate": 4.7525643332733495e-06, "loss": 0.3907, "step": 5890 }, { "epoch": 2.11, "learning_rate": 4.751664567212525e-06, "loss": 0.3373, "step": 5891 }, { "epoch": 2.11, "learning_rate": 4.750764801151701e-06, "loss": 0.3289, "step": 5892 }, { "epoch": 2.12, "learning_rate": 4.749865035090877e-06, "loss": 0.3183, "step": 5893 }, { "epoch": 2.12, "learning_rate": 4.748965269030053e-06, "loss": 0.3676, "step": 5894 }, { "epoch": 2.12, "learning_rate": 4.748065502969228e-06, "loss": 0.3761, "step": 5895 }, { "epoch": 2.12, "learning_rate": 4.747165736908404e-06, "loss": 0.3006, "step": 5896 }, { "epoch": 2.12, "learning_rate": 4.74626597084758e-06, "loss": 0.2772, "step": 5897 }, { "epoch": 2.12, "learning_rate": 4.745366204786756e-06, "loss": 0.3168, "step": 5898 }, { "epoch": 2.12, "learning_rate": 4.7444664387259315e-06, "loss": 0.3402, "step": 5899 }, { "epoch": 2.12, "learning_rate": 4.743566672665107e-06, "loss": 0.3247, "step": 5900 }, { "epoch": 2.12, "learning_rate": 4.742666906604284e-06, "loss": 0.3134, "step": 5901 }, { "epoch": 2.12, "learning_rate": 4.741767140543459e-06, "loss": 0.3325, "step": 5902 }, { "epoch": 2.12, "learning_rate": 4.740867374482635e-06, "loss": 0.3525, "step": 5903 }, { "epoch": 2.12, "learning_rate": 4.73996760842181e-06, "loss": 0.2965, "step": 5904 }, { "epoch": 2.12, "learning_rate": 4.739067842360987e-06, "loss": 0.3629, "step": 5905 }, { "epoch": 2.12, "learning_rate": 4.738168076300163e-06, "loss": 0.3474, "step": 5906 }, { "epoch": 2.12, "learning_rate": 4.737268310239338e-06, "loss": 0.3275, "step": 5907 }, { "epoch": 2.12, "learning_rate": 4.736368544178514e-06, "loss": 0.3265, "step": 5908 }, { "epoch": 2.12, "learning_rate": 4.73546877811769e-06, "loss": 0.3719, "step": 5909 }, { "epoch": 2.12, "learning_rate": 4.734569012056866e-06, "loss": 0.3401, "step": 5910 }, { "epoch": 2.12, "learning_rate": 4.7336692459960415e-06, "loss": 0.3705, "step": 5911 }, { "epoch": 2.12, "learning_rate": 4.732769479935217e-06, "loss": 0.3509, "step": 5912 }, { "epoch": 2.12, "learning_rate": 4.731869713874393e-06, "loss": 0.3222, "step": 5913 }, { "epoch": 2.12, "learning_rate": 4.730969947813569e-06, "loss": 0.3389, "step": 5914 }, { "epoch": 2.12, "learning_rate": 4.730070181752745e-06, "loss": 0.3679, "step": 5915 }, { "epoch": 2.12, "learning_rate": 4.72917041569192e-06, "loss": 0.3376, "step": 5916 }, { "epoch": 2.12, "learning_rate": 4.728270649631096e-06, "loss": 0.3244, "step": 5917 }, { "epoch": 2.12, "learning_rate": 4.727370883570272e-06, "loss": 0.3466, "step": 5918 }, { "epoch": 2.12, "learning_rate": 4.726471117509448e-06, "loss": 0.3414, "step": 5919 }, { "epoch": 2.12, "learning_rate": 4.725571351448624e-06, "loss": 0.4249, "step": 5920 }, { "epoch": 2.13, "learning_rate": 4.724671585387799e-06, "loss": 0.3727, "step": 5921 }, { "epoch": 2.13, "learning_rate": 4.723771819326976e-06, "loss": 0.342, "step": 5922 }, { "epoch": 2.13, "learning_rate": 4.722872053266151e-06, "loss": 0.3073, "step": 5923 }, { "epoch": 2.13, "learning_rate": 4.721972287205327e-06, "loss": 0.3272, "step": 5924 }, { "epoch": 2.13, "learning_rate": 4.721072521144503e-06, "loss": 0.2939, "step": 5925 }, { "epoch": 2.13, "learning_rate": 4.720172755083678e-06, "loss": 0.3575, "step": 5926 }, { "epoch": 2.13, "learning_rate": 4.719272989022855e-06, "loss": 0.2903, "step": 5927 }, { "epoch": 2.13, "learning_rate": 4.7183732229620304e-06, "loss": 0.4048, "step": 5928 }, { "epoch": 2.13, "learning_rate": 4.717473456901206e-06, "loss": 0.4202, "step": 5929 }, { "epoch": 2.13, "learning_rate": 4.716573690840382e-06, "loss": 0.3152, "step": 5930 }, { "epoch": 2.13, "learning_rate": 4.715673924779558e-06, "loss": 0.3146, "step": 5931 }, { "epoch": 2.13, "learning_rate": 4.7147741587187335e-06, "loss": 0.432, "step": 5932 }, { "epoch": 2.13, "learning_rate": 4.713874392657909e-06, "loss": 0.3525, "step": 5933 }, { "epoch": 2.13, "learning_rate": 4.712974626597085e-06, "loss": 0.3104, "step": 5934 }, { "epoch": 2.13, "learning_rate": 4.712074860536261e-06, "loss": 0.3731, "step": 5935 }, { "epoch": 2.13, "learning_rate": 4.711175094475437e-06, "loss": 0.3605, "step": 5936 }, { "epoch": 2.13, "learning_rate": 4.710275328414612e-06, "loss": 0.3252, "step": 5937 }, { "epoch": 2.13, "learning_rate": 4.709375562353788e-06, "loss": 0.3152, "step": 5938 }, { "epoch": 2.13, "learning_rate": 4.708475796292965e-06, "loss": 0.3299, "step": 5939 }, { "epoch": 2.13, "learning_rate": 4.70757603023214e-06, "loss": 0.3433, "step": 5940 }, { "epoch": 2.13, "learning_rate": 4.706676264171316e-06, "loss": 0.3117, "step": 5941 }, { "epoch": 2.13, "learning_rate": 4.705776498110491e-06, "loss": 0.381, "step": 5942 }, { "epoch": 2.13, "learning_rate": 4.704876732049668e-06, "loss": 0.355, "step": 5943 }, { "epoch": 2.13, "learning_rate": 4.703976965988844e-06, "loss": 0.3633, "step": 5944 }, { "epoch": 2.13, "learning_rate": 4.7030771999280185e-06, "loss": 0.341, "step": 5945 }, { "epoch": 2.13, "learning_rate": 4.702177433867195e-06, "loss": 0.353, "step": 5946 }, { "epoch": 2.13, "learning_rate": 4.701277667806371e-06, "loss": 0.2822, "step": 5947 }, { "epoch": 2.13, "learning_rate": 4.700377901745547e-06, "loss": 0.386, "step": 5948 }, { "epoch": 2.14, "learning_rate": 4.6994781356847224e-06, "loss": 0.3116, "step": 5949 }, { "epoch": 2.14, "learning_rate": 4.698578369623898e-06, "loss": 0.3955, "step": 5950 }, { "epoch": 2.14, "learning_rate": 4.697678603563074e-06, "loss": 0.3883, "step": 5951 }, { "epoch": 2.14, "learning_rate": 4.69677883750225e-06, "loss": 0.5165, "step": 5952 }, { "epoch": 2.14, "learning_rate": 4.6958790714414255e-06, "loss": 0.3533, "step": 5953 }, { "epoch": 2.14, "learning_rate": 4.694979305380601e-06, "loss": 0.3554, "step": 5954 }, { "epoch": 2.14, "learning_rate": 4.694079539319777e-06, "loss": 0.2644, "step": 5955 }, { "epoch": 2.14, "learning_rate": 4.693179773258953e-06, "loss": 0.3235, "step": 5956 }, { "epoch": 2.14, "learning_rate": 4.692280007198129e-06, "loss": 0.3114, "step": 5957 }, { "epoch": 2.14, "learning_rate": 4.691380241137305e-06, "loss": 0.3188, "step": 5958 }, { "epoch": 2.14, "learning_rate": 4.69048047507648e-06, "loss": 0.341, "step": 5959 }, { "epoch": 2.14, "learning_rate": 4.689580709015657e-06, "loss": 0.2731, "step": 5960 }, { "epoch": 2.14, "learning_rate": 4.688680942954832e-06, "loss": 0.3715, "step": 5961 }, { "epoch": 2.14, "learning_rate": 4.687781176894008e-06, "loss": 0.337, "step": 5962 }, { "epoch": 2.14, "learning_rate": 4.686881410833184e-06, "loss": 0.2687, "step": 5963 }, { "epoch": 2.14, "learning_rate": 4.68598164477236e-06, "loss": 0.3828, "step": 5964 }, { "epoch": 2.14, "learning_rate": 4.685081878711536e-06, "loss": 0.3338, "step": 5965 }, { "epoch": 2.14, "learning_rate": 4.684182112650711e-06, "loss": 0.4041, "step": 5966 }, { "epoch": 2.14, "learning_rate": 4.683282346589887e-06, "loss": 0.2854, "step": 5967 }, { "epoch": 2.14, "learning_rate": 4.682382580529063e-06, "loss": 0.3015, "step": 5968 }, { "epoch": 2.14, "learning_rate": 4.681482814468239e-06, "loss": 0.301, "step": 5969 }, { "epoch": 2.14, "learning_rate": 4.6805830484074144e-06, "loss": 0.3185, "step": 5970 }, { "epoch": 2.14, "learning_rate": 4.67968328234659e-06, "loss": 0.2886, "step": 5971 }, { "epoch": 2.14, "learning_rate": 4.678783516285766e-06, "loss": 0.34, "step": 5972 }, { "epoch": 2.14, "learning_rate": 4.677883750224942e-06, "loss": 0.353, "step": 5973 }, { "epoch": 2.14, "learning_rate": 4.6769839841641175e-06, "loss": 0.3859, "step": 5974 }, { "epoch": 2.14, "learning_rate": 4.676084218103293e-06, "loss": 0.3301, "step": 5975 }, { "epoch": 2.15, "learning_rate": 4.675184452042469e-06, "loss": 0.3227, "step": 5976 }, { "epoch": 2.15, "learning_rate": 4.674284685981646e-06, "loss": 0.3172, "step": 5977 }, { "epoch": 2.15, "learning_rate": 4.6733849199208206e-06, "loss": 0.3706, "step": 5978 }, { "epoch": 2.15, "learning_rate": 4.672485153859997e-06, "loss": 0.3649, "step": 5979 }, { "epoch": 2.15, "learning_rate": 4.671585387799172e-06, "loss": 0.3266, "step": 5980 }, { "epoch": 2.15, "learning_rate": 4.670685621738349e-06, "loss": 0.4041, "step": 5981 }, { "epoch": 2.15, "learning_rate": 4.6697858556775245e-06, "loss": 0.3825, "step": 5982 }, { "epoch": 2.15, "learning_rate": 4.6688860896167e-06, "loss": 0.413, "step": 5983 }, { "epoch": 2.15, "learning_rate": 4.667986323555876e-06, "loss": 0.2425, "step": 5984 }, { "epoch": 2.15, "learning_rate": 4.667086557495052e-06, "loss": 0.2989, "step": 5985 }, { "epoch": 2.15, "learning_rate": 4.6661867914342276e-06, "loss": 0.3883, "step": 5986 }, { "epoch": 2.15, "learning_rate": 4.665287025373403e-06, "loss": 0.407, "step": 5987 }, { "epoch": 2.15, "learning_rate": 4.664387259312579e-06, "loss": 0.3546, "step": 5988 }, { "epoch": 2.15, "learning_rate": 4.663487493251755e-06, "loss": 0.3484, "step": 5989 }, { "epoch": 2.15, "learning_rate": 4.662587727190931e-06, "loss": 0.3707, "step": 5990 }, { "epoch": 2.15, "learning_rate": 4.661687961130106e-06, "loss": 0.2967, "step": 5991 }, { "epoch": 2.15, "learning_rate": 4.660788195069282e-06, "loss": 0.2761, "step": 5992 }, { "epoch": 2.15, "learning_rate": 4.659888429008459e-06, "loss": 0.3093, "step": 5993 }, { "epoch": 2.15, "learning_rate": 4.658988662947634e-06, "loss": 0.385, "step": 5994 }, { "epoch": 2.15, "learning_rate": 4.6580888968868095e-06, "loss": 0.3892, "step": 5995 }, { "epoch": 2.15, "learning_rate": 4.657189130825986e-06, "loss": 0.3002, "step": 5996 }, { "epoch": 2.15, "learning_rate": 4.656289364765161e-06, "loss": 0.3196, "step": 5997 }, { "epoch": 2.15, "learning_rate": 4.655389598704338e-06, "loss": 0.3119, "step": 5998 }, { "epoch": 2.15, "learning_rate": 4.6544898326435126e-06, "loss": 0.3387, "step": 5999 }, { "epoch": 2.15, "learning_rate": 4.653590066582689e-06, "loss": 0.3175, "step": 6000 }, { "epoch": 2.15, "learning_rate": 4.652690300521865e-06, "loss": 0.3142, "step": 6001 }, { "epoch": 2.15, "learning_rate": 4.651790534461041e-06, "loss": 0.2552, "step": 6002 }, { "epoch": 2.15, "learning_rate": 4.6508907684002165e-06, "loss": 0.301, "step": 6003 }, { "epoch": 2.16, "learning_rate": 4.649991002339392e-06, "loss": 0.4086, "step": 6004 }, { "epoch": 2.16, "learning_rate": 4.649091236278568e-06, "loss": 0.2755, "step": 6005 }, { "epoch": 2.16, "learning_rate": 4.648191470217744e-06, "loss": 0.3229, "step": 6006 }, { "epoch": 2.16, "learning_rate": 4.6472917041569196e-06, "loss": 0.3564, "step": 6007 }, { "epoch": 2.16, "learning_rate": 4.646391938096095e-06, "loss": 0.3254, "step": 6008 }, { "epoch": 2.16, "learning_rate": 4.645492172035271e-06, "loss": 0.4312, "step": 6009 }, { "epoch": 2.16, "learning_rate": 4.644592405974447e-06, "loss": 0.3266, "step": 6010 }, { "epoch": 2.16, "learning_rate": 4.643692639913623e-06, "loss": 0.3194, "step": 6011 }, { "epoch": 2.16, "learning_rate": 4.642792873852799e-06, "loss": 0.2877, "step": 6012 }, { "epoch": 2.16, "learning_rate": 4.641893107791974e-06, "loss": 0.4384, "step": 6013 }, { "epoch": 2.16, "learning_rate": 4.64099334173115e-06, "loss": 0.3694, "step": 6014 }, { "epoch": 2.16, "learning_rate": 4.6400935756703266e-06, "loss": 0.4061, "step": 6015 }, { "epoch": 2.16, "learning_rate": 4.6391938096095015e-06, "loss": 0.2703, "step": 6016 }, { "epoch": 2.16, "learning_rate": 4.638294043548678e-06, "loss": 0.3067, "step": 6017 }, { "epoch": 2.16, "learning_rate": 4.637394277487853e-06, "loss": 0.3029, "step": 6018 }, { "epoch": 2.16, "learning_rate": 4.63649451142703e-06, "loss": 0.3465, "step": 6019 }, { "epoch": 2.16, "learning_rate": 4.635594745366205e-06, "loss": 0.4244, "step": 6020 }, { "epoch": 2.16, "learning_rate": 4.634694979305381e-06, "loss": 0.3232, "step": 6021 }, { "epoch": 2.16, "learning_rate": 4.633795213244557e-06, "loss": 0.3346, "step": 6022 }, { "epoch": 2.16, "learning_rate": 4.632895447183733e-06, "loss": 0.3248, "step": 6023 }, { "epoch": 2.16, "learning_rate": 4.6319956811229085e-06, "loss": 0.3997, "step": 6024 }, { "epoch": 2.16, "learning_rate": 4.631095915062084e-06, "loss": 0.3644, "step": 6025 }, { "epoch": 2.16, "learning_rate": 4.63019614900126e-06, "loss": 0.2865, "step": 6026 }, { "epoch": 2.16, "learning_rate": 4.629296382940436e-06, "loss": 0.3472, "step": 6027 }, { "epoch": 2.16, "learning_rate": 4.6283966168796116e-06, "loss": 0.3377, "step": 6028 }, { "epoch": 2.16, "learning_rate": 4.627496850818787e-06, "loss": 0.3492, "step": 6029 }, { "epoch": 2.16, "learning_rate": 4.626597084757963e-06, "loss": 0.2807, "step": 6030 }, { "epoch": 2.16, "learning_rate": 4.62569731869714e-06, "loss": 0.3502, "step": 6031 }, { "epoch": 2.17, "learning_rate": 4.624797552636315e-06, "loss": 0.3936, "step": 6032 }, { "epoch": 2.17, "learning_rate": 4.62389778657549e-06, "loss": 0.3237, "step": 6033 }, { "epoch": 2.17, "learning_rate": 4.622998020514667e-06, "loss": 0.3916, "step": 6034 }, { "epoch": 2.17, "learning_rate": 4.622098254453842e-06, "loss": 0.3089, "step": 6035 }, { "epoch": 2.17, "learning_rate": 4.6211984883930186e-06, "loss": 0.3486, "step": 6036 }, { "epoch": 2.17, "learning_rate": 4.6202987223321935e-06, "loss": 0.3263, "step": 6037 }, { "epoch": 2.17, "learning_rate": 4.61939895627137e-06, "loss": 0.4035, "step": 6038 }, { "epoch": 2.17, "learning_rate": 4.618499190210546e-06, "loss": 0.3452, "step": 6039 }, { "epoch": 2.17, "learning_rate": 4.617599424149722e-06, "loss": 0.3721, "step": 6040 }, { "epoch": 2.17, "learning_rate": 4.616699658088897e-06, "loss": 0.3238, "step": 6041 }, { "epoch": 2.17, "learning_rate": 4.615799892028073e-06, "loss": 0.3303, "step": 6042 }, { "epoch": 2.17, "learning_rate": 4.614900125967249e-06, "loss": 0.3639, "step": 6043 }, { "epoch": 2.17, "learning_rate": 4.614000359906425e-06, "loss": 0.4005, "step": 6044 }, { "epoch": 2.17, "learning_rate": 4.6131005938456005e-06, "loss": 0.3733, "step": 6045 }, { "epoch": 2.17, "learning_rate": 4.612200827784776e-06, "loss": 0.3195, "step": 6046 }, { "epoch": 2.17, "learning_rate": 4.611301061723952e-06, "loss": 0.33, "step": 6047 }, { "epoch": 2.17, "learning_rate": 4.610401295663128e-06, "loss": 0.3133, "step": 6048 }, { "epoch": 2.17, "learning_rate": 4.6095015296023035e-06, "loss": 0.3366, "step": 6049 }, { "epoch": 2.17, "learning_rate": 4.60860176354148e-06, "loss": 0.4082, "step": 6050 }, { "epoch": 2.17, "learning_rate": 4.607701997480655e-06, "loss": 0.3357, "step": 6051 }, { "epoch": 2.17, "learning_rate": 4.606802231419832e-06, "loss": 0.3861, "step": 6052 }, { "epoch": 2.17, "learning_rate": 4.605902465359007e-06, "loss": 0.3544, "step": 6053 }, { "epoch": 2.17, "learning_rate": 4.605002699298182e-06, "loss": 0.2834, "step": 6054 }, { "epoch": 2.17, "learning_rate": 4.604102933237359e-06, "loss": 0.2925, "step": 6055 }, { "epoch": 2.17, "learning_rate": 4.603203167176534e-06, "loss": 0.4095, "step": 6056 }, { "epoch": 2.17, "learning_rate": 4.6023034011157105e-06, "loss": 0.3085, "step": 6057 }, { "epoch": 2.17, "learning_rate": 4.601403635054886e-06, "loss": 0.3438, "step": 6058 }, { "epoch": 2.17, "learning_rate": 4.600503868994062e-06, "loss": 0.297, "step": 6059 }, { "epoch": 2.18, "learning_rate": 4.599604102933238e-06, "loss": 0.3663, "step": 6060 }, { "epoch": 2.18, "learning_rate": 4.598704336872414e-06, "loss": 0.3256, "step": 6061 }, { "epoch": 2.18, "learning_rate": 4.597804570811589e-06, "loss": 0.3305, "step": 6062 }, { "epoch": 2.18, "learning_rate": 4.596904804750765e-06, "loss": 0.3013, "step": 6063 }, { "epoch": 2.18, "learning_rate": 4.596005038689941e-06, "loss": 0.3235, "step": 6064 }, { "epoch": 2.18, "learning_rate": 4.595105272629117e-06, "loss": 0.437, "step": 6065 }, { "epoch": 2.18, "learning_rate": 4.5942055065682925e-06, "loss": 0.3709, "step": 6066 }, { "epoch": 2.18, "learning_rate": 4.593305740507468e-06, "loss": 0.2898, "step": 6067 }, { "epoch": 2.18, "learning_rate": 4.592405974446644e-06, "loss": 0.3055, "step": 6068 }, { "epoch": 2.18, "learning_rate": 4.591506208385821e-06, "loss": 0.3373, "step": 6069 }, { "epoch": 2.18, "learning_rate": 4.5906064423249955e-06, "loss": 0.3769, "step": 6070 }, { "epoch": 2.18, "learning_rate": 4.589706676264172e-06, "loss": 0.3429, "step": 6071 }, { "epoch": 2.18, "learning_rate": 4.588806910203347e-06, "loss": 0.3344, "step": 6072 }, { "epoch": 2.18, "learning_rate": 4.587907144142523e-06, "loss": 0.3129, "step": 6073 }, { "epoch": 2.18, "learning_rate": 4.5870073780816995e-06, "loss": 0.3644, "step": 6074 }, { "epoch": 2.18, "learning_rate": 4.586107612020874e-06, "loss": 0.3117, "step": 6075 }, { "epoch": 2.18, "learning_rate": 4.585207845960051e-06, "loss": 0.3249, "step": 6076 }, { "epoch": 2.18, "learning_rate": 4.584308079899227e-06, "loss": 0.345, "step": 6077 }, { "epoch": 2.18, "learning_rate": 4.5834083138384025e-06, "loss": 0.2998, "step": 6078 }, { "epoch": 2.18, "learning_rate": 4.582508547777578e-06, "loss": 0.3628, "step": 6079 }, { "epoch": 2.18, "learning_rate": 4.581608781716754e-06, "loss": 0.3043, "step": 6080 }, { "epoch": 2.18, "learning_rate": 4.58070901565593e-06, "loss": 0.3077, "step": 6081 }, { "epoch": 2.18, "learning_rate": 4.579809249595106e-06, "loss": 0.4187, "step": 6082 }, { "epoch": 2.18, "learning_rate": 4.578909483534281e-06, "loss": 0.3007, "step": 6083 }, { "epoch": 2.18, "learning_rate": 4.578009717473457e-06, "loss": 0.3828, "step": 6084 }, { "epoch": 2.18, "learning_rate": 4.577109951412633e-06, "loss": 0.3092, "step": 6085 }, { "epoch": 2.18, "learning_rate": 4.576210185351809e-06, "loss": 0.3534, "step": 6086 }, { "epoch": 2.18, "learning_rate": 4.5753104192909845e-06, "loss": 0.3468, "step": 6087 }, { "epoch": 2.19, "learning_rate": 4.574410653230161e-06, "loss": 0.3622, "step": 6088 }, { "epoch": 2.19, "learning_rate": 4.573510887169336e-06, "loss": 0.351, "step": 6089 }, { "epoch": 2.19, "learning_rate": 4.572611121108513e-06, "loss": 0.4232, "step": 6090 }, { "epoch": 2.19, "learning_rate": 4.5717113550476875e-06, "loss": 0.3568, "step": 6091 }, { "epoch": 2.19, "learning_rate": 4.570811588986864e-06, "loss": 0.3717, "step": 6092 }, { "epoch": 2.19, "learning_rate": 4.56991182292604e-06, "loss": 0.3285, "step": 6093 }, { "epoch": 2.19, "learning_rate": 4.569012056865215e-06, "loss": 0.355, "step": 6094 }, { "epoch": 2.19, "learning_rate": 4.5681122908043915e-06, "loss": 0.3949, "step": 6095 }, { "epoch": 2.19, "learning_rate": 4.567212524743567e-06, "loss": 0.4255, "step": 6096 }, { "epoch": 2.19, "learning_rate": 4.566312758682743e-06, "loss": 0.3368, "step": 6097 }, { "epoch": 2.19, "learning_rate": 4.565412992621919e-06, "loss": 0.3947, "step": 6098 }, { "epoch": 2.19, "learning_rate": 4.5645132265610945e-06, "loss": 0.3421, "step": 6099 }, { "epoch": 2.19, "learning_rate": 4.56361346050027e-06, "loss": 0.3023, "step": 6100 }, { "epoch": 2.19, "learning_rate": 4.562713694439446e-06, "loss": 0.32, "step": 6101 }, { "epoch": 2.19, "learning_rate": 4.561813928378622e-06, "loss": 0.353, "step": 6102 }, { "epoch": 2.19, "learning_rate": 4.560914162317798e-06, "loss": 0.3311, "step": 6103 }, { "epoch": 2.19, "learning_rate": 4.560014396256973e-06, "loss": 0.4093, "step": 6104 }, { "epoch": 2.19, "learning_rate": 4.559114630196149e-06, "loss": 0.2821, "step": 6105 }, { "epoch": 2.19, "learning_rate": 4.558214864135325e-06, "loss": 0.3644, "step": 6106 }, { "epoch": 2.19, "learning_rate": 4.5573150980745015e-06, "loss": 0.4308, "step": 6107 }, { "epoch": 2.19, "learning_rate": 4.5564153320136764e-06, "loss": 0.3936, "step": 6108 }, { "epoch": 2.19, "learning_rate": 4.555515565952853e-06, "loss": 0.3568, "step": 6109 }, { "epoch": 2.19, "learning_rate": 4.554615799892028e-06, "loss": 0.3675, "step": 6110 }, { "epoch": 2.19, "learning_rate": 4.553716033831205e-06, "loss": 0.3569, "step": 6111 }, { "epoch": 2.19, "learning_rate": 4.55281626777038e-06, "loss": 0.3151, "step": 6112 }, { "epoch": 2.19, "learning_rate": 4.551916501709555e-06, "loss": 0.3059, "step": 6113 }, { "epoch": 2.19, "learning_rate": 4.551016735648732e-06, "loss": 0.3944, "step": 6114 }, { "epoch": 2.19, "learning_rate": 4.550116969587908e-06, "loss": 0.3305, "step": 6115 }, { "epoch": 2.2, "learning_rate": 4.5492172035270834e-06, "loss": 0.3248, "step": 6116 }, { "epoch": 2.2, "learning_rate": 4.548317437466259e-06, "loss": 0.3825, "step": 6117 }, { "epoch": 2.2, "learning_rate": 4.547417671405435e-06, "loss": 0.3827, "step": 6118 }, { "epoch": 2.2, "learning_rate": 4.546517905344611e-06, "loss": 0.3191, "step": 6119 }, { "epoch": 2.2, "learning_rate": 4.5456181392837865e-06, "loss": 0.3684, "step": 6120 }, { "epoch": 2.2, "learning_rate": 4.544718373222962e-06, "loss": 0.3906, "step": 6121 }, { "epoch": 2.2, "learning_rate": 4.543818607162138e-06, "loss": 0.4132, "step": 6122 }, { "epoch": 2.2, "learning_rate": 4.542918841101314e-06, "loss": 0.3253, "step": 6123 }, { "epoch": 2.2, "learning_rate": 4.54201907504049e-06, "loss": 0.3669, "step": 6124 }, { "epoch": 2.2, "learning_rate": 4.541119308979665e-06, "loss": 0.3071, "step": 6125 }, { "epoch": 2.2, "learning_rate": 4.540219542918842e-06, "loss": 0.3628, "step": 6126 }, { "epoch": 2.2, "learning_rate": 4.539319776858017e-06, "loss": 0.3614, "step": 6127 }, { "epoch": 2.2, "learning_rate": 4.5384200107971935e-06, "loss": 0.3475, "step": 6128 }, { "epoch": 2.2, "learning_rate": 4.5375202447363684e-06, "loss": 0.3514, "step": 6129 }, { "epoch": 2.2, "learning_rate": 4.536620478675545e-06, "loss": 0.3395, "step": 6130 }, { "epoch": 2.2, "learning_rate": 4.535720712614721e-06, "loss": 0.4038, "step": 6131 }, { "epoch": 2.2, "learning_rate": 4.534820946553896e-06, "loss": 0.3395, "step": 6132 }, { "epoch": 2.2, "learning_rate": 4.533921180493072e-06, "loss": 0.3435, "step": 6133 }, { "epoch": 2.2, "learning_rate": 4.533021414432248e-06, "loss": 0.4249, "step": 6134 }, { "epoch": 2.2, "learning_rate": 4.532121648371424e-06, "loss": 0.3655, "step": 6135 }, { "epoch": 2.2, "learning_rate": 4.5312218823106e-06, "loss": 0.4532, "step": 6136 }, { "epoch": 2.2, "learning_rate": 4.5303221162497754e-06, "loss": 0.3132, "step": 6137 }, { "epoch": 2.2, "learning_rate": 4.529422350188951e-06, "loss": 0.2912, "step": 6138 }, { "epoch": 2.2, "learning_rate": 4.528522584128127e-06, "loss": 0.4207, "step": 6139 }, { "epoch": 2.2, "learning_rate": 4.527622818067303e-06, "loss": 0.3395, "step": 6140 }, { "epoch": 2.2, "learning_rate": 4.5267230520064785e-06, "loss": 0.2883, "step": 6141 }, { "epoch": 2.2, "learning_rate": 4.525823285945654e-06, "loss": 0.329, "step": 6142 }, { "epoch": 2.2, "learning_rate": 4.52492351988483e-06, "loss": 0.3511, "step": 6143 }, { "epoch": 2.21, "learning_rate": 4.524023753824006e-06, "loss": 0.2634, "step": 6144 }, { "epoch": 2.21, "learning_rate": 4.5231239877631824e-06, "loss": 0.3726, "step": 6145 }, { "epoch": 2.21, "learning_rate": 4.522224221702357e-06, "loss": 0.3255, "step": 6146 }, { "epoch": 2.21, "learning_rate": 4.521324455641534e-06, "loss": 0.3764, "step": 6147 }, { "epoch": 2.21, "learning_rate": 4.520424689580709e-06, "loss": 0.2933, "step": 6148 }, { "epoch": 2.21, "learning_rate": 4.5195249235198855e-06, "loss": 0.4273, "step": 6149 }, { "epoch": 2.21, "learning_rate": 4.518625157459061e-06, "loss": 0.4132, "step": 6150 }, { "epoch": 2.21, "learning_rate": 4.517725391398237e-06, "loss": 0.3611, "step": 6151 }, { "epoch": 2.21, "learning_rate": 4.516825625337413e-06, "loss": 0.3494, "step": 6152 }, { "epoch": 2.21, "learning_rate": 4.515925859276589e-06, "loss": 0.335, "step": 6153 }, { "epoch": 2.21, "learning_rate": 4.515026093215764e-06, "loss": 0.3333, "step": 6154 }, { "epoch": 2.21, "learning_rate": 4.51412632715494e-06, "loss": 0.3628, "step": 6155 }, { "epoch": 2.21, "learning_rate": 4.513226561094116e-06, "loss": 0.3404, "step": 6156 }, { "epoch": 2.21, "learning_rate": 4.512326795033292e-06, "loss": 0.2986, "step": 6157 }, { "epoch": 2.21, "learning_rate": 4.5114270289724674e-06, "loss": 0.4112, "step": 6158 }, { "epoch": 2.21, "learning_rate": 4.510527262911643e-06, "loss": 0.3295, "step": 6159 }, { "epoch": 2.21, "learning_rate": 4.509627496850819e-06, "loss": 0.3011, "step": 6160 }, { "epoch": 2.21, "learning_rate": 4.508727730789995e-06, "loss": 0.3065, "step": 6161 }, { "epoch": 2.21, "learning_rate": 4.5078279647291705e-06, "loss": 0.3746, "step": 6162 }, { "epoch": 2.21, "learning_rate": 4.506928198668346e-06, "loss": 0.2722, "step": 6163 }, { "epoch": 2.21, "learning_rate": 4.506028432607523e-06, "loss": 0.4191, "step": 6164 }, { "epoch": 2.21, "learning_rate": 4.505128666546698e-06, "loss": 0.3347, "step": 6165 }, { "epoch": 2.21, "learning_rate": 4.5042289004858744e-06, "loss": 0.3196, "step": 6166 }, { "epoch": 2.21, "learning_rate": 4.503329134425049e-06, "loss": 0.319, "step": 6167 }, { "epoch": 2.21, "learning_rate": 4.502429368364226e-06, "loss": 0.3934, "step": 6168 }, { "epoch": 2.21, "learning_rate": 4.501529602303402e-06, "loss": 0.288, "step": 6169 }, { "epoch": 2.21, "learning_rate": 4.5006298362425775e-06, "loss": 0.3264, "step": 6170 }, { "epoch": 2.22, "learning_rate": 4.499730070181753e-06, "loss": 0.3065, "step": 6171 }, { "epoch": 2.22, "learning_rate": 4.498830304120929e-06, "loss": 0.3772, "step": 6172 }, { "epoch": 2.22, "learning_rate": 4.497930538060105e-06, "loss": 0.3956, "step": 6173 }, { "epoch": 2.22, "learning_rate": 4.4970307719992806e-06, "loss": 0.3658, "step": 6174 }, { "epoch": 2.22, "learning_rate": 4.496131005938456e-06, "loss": 0.2893, "step": 6175 }, { "epoch": 2.22, "learning_rate": 4.495231239877632e-06, "loss": 0.3153, "step": 6176 }, { "epoch": 2.22, "learning_rate": 4.494331473816808e-06, "loss": 0.2764, "step": 6177 }, { "epoch": 2.22, "learning_rate": 4.493431707755984e-06, "loss": 0.3201, "step": 6178 }, { "epoch": 2.22, "learning_rate": 4.492531941695159e-06, "loss": 0.3401, "step": 6179 }, { "epoch": 2.22, "learning_rate": 4.491632175634336e-06, "loss": 0.3532, "step": 6180 }, { "epoch": 2.22, "learning_rate": 4.490732409573511e-06, "loss": 0.3067, "step": 6181 }, { "epoch": 2.22, "learning_rate": 4.489832643512687e-06, "loss": 0.3245, "step": 6182 }, { "epoch": 2.22, "learning_rate": 4.488932877451863e-06, "loss": 0.3148, "step": 6183 }, { "epoch": 2.22, "learning_rate": 4.488033111391038e-06, "loss": 0.3741, "step": 6184 }, { "epoch": 2.22, "learning_rate": 4.487133345330215e-06, "loss": 0.2986, "step": 6185 }, { "epoch": 2.22, "learning_rate": 4.48623357926939e-06, "loss": 0.2804, "step": 6186 }, { "epoch": 2.22, "learning_rate": 4.485333813208566e-06, "loss": 0.3337, "step": 6187 }, { "epoch": 2.22, "learning_rate": 4.484434047147742e-06, "loss": 0.4225, "step": 6188 }, { "epoch": 2.22, "learning_rate": 4.483534281086918e-06, "loss": 0.3287, "step": 6189 }, { "epoch": 2.22, "learning_rate": 4.482634515026094e-06, "loss": 0.347, "step": 6190 }, { "epoch": 2.22, "learning_rate": 4.4817347489652695e-06, "loss": 0.3031, "step": 6191 }, { "epoch": 2.22, "learning_rate": 4.480834982904445e-06, "loss": 0.3443, "step": 6192 }, { "epoch": 2.22, "learning_rate": 4.479935216843621e-06, "loss": 0.3587, "step": 6193 }, { "epoch": 2.22, "learning_rate": 4.479035450782797e-06, "loss": 0.2481, "step": 6194 }, { "epoch": 2.22, "learning_rate": 4.4781356847219726e-06, "loss": 0.3359, "step": 6195 }, { "epoch": 2.22, "learning_rate": 4.477235918661148e-06, "loss": 0.3028, "step": 6196 }, { "epoch": 2.22, "learning_rate": 4.476336152600324e-06, "loss": 0.2802, "step": 6197 }, { "epoch": 2.22, "learning_rate": 4.4754363865395e-06, "loss": 0.3303, "step": 6198 }, { "epoch": 2.23, "learning_rate": 4.4745366204786765e-06, "loss": 0.3309, "step": 6199 }, { "epoch": 2.23, "learning_rate": 4.473636854417851e-06, "loss": 0.2888, "step": 6200 }, { "epoch": 2.23, "learning_rate": 4.472737088357027e-06, "loss": 0.3115, "step": 6201 }, { "epoch": 2.23, "learning_rate": 4.471837322296204e-06, "loss": 0.3701, "step": 6202 }, { "epoch": 2.23, "learning_rate": 4.470937556235379e-06, "loss": 0.4343, "step": 6203 }, { "epoch": 2.23, "learning_rate": 4.470037790174555e-06, "loss": 0.3652, "step": 6204 }, { "epoch": 2.23, "learning_rate": 4.46913802411373e-06, "loss": 0.33, "step": 6205 }, { "epoch": 2.23, "learning_rate": 4.468238258052907e-06, "loss": 0.3319, "step": 6206 }, { "epoch": 2.23, "learning_rate": 4.467338491992083e-06, "loss": 0.3369, "step": 6207 }, { "epoch": 2.23, "learning_rate": 4.466438725931258e-06, "loss": 0.2979, "step": 6208 }, { "epoch": 2.23, "learning_rate": 4.465538959870434e-06, "loss": 0.3252, "step": 6209 }, { "epoch": 2.23, "learning_rate": 4.46463919380961e-06, "loss": 0.3805, "step": 6210 }, { "epoch": 2.23, "learning_rate": 4.463739427748786e-06, "loss": 0.3484, "step": 6211 }, { "epoch": 2.23, "learning_rate": 4.4628396616879615e-06, "loss": 0.3004, "step": 6212 }, { "epoch": 2.23, "learning_rate": 4.461939895627137e-06, "loss": 0.4266, "step": 6213 }, { "epoch": 2.23, "learning_rate": 4.461040129566313e-06, "loss": 0.3615, "step": 6214 }, { "epoch": 2.23, "learning_rate": 4.460140363505489e-06, "loss": 0.3047, "step": 6215 }, { "epoch": 2.23, "learning_rate": 4.4592405974446646e-06, "loss": 0.3445, "step": 6216 }, { "epoch": 2.23, "learning_rate": 4.45834083138384e-06, "loss": 0.3059, "step": 6217 }, { "epoch": 2.23, "learning_rate": 4.457441065323017e-06, "loss": 0.2747, "step": 6218 }, { "epoch": 2.23, "learning_rate": 4.456541299262192e-06, "loss": 0.2849, "step": 6219 }, { "epoch": 2.23, "learning_rate": 4.455641533201368e-06, "loss": 0.394, "step": 6220 }, { "epoch": 2.23, "learning_rate": 4.454741767140544e-06, "loss": 0.341, "step": 6221 }, { "epoch": 2.23, "learning_rate": 4.453842001079719e-06, "loss": 0.3387, "step": 6222 }, { "epoch": 2.23, "learning_rate": 4.452942235018896e-06, "loss": 0.326, "step": 6223 }, { "epoch": 2.23, "learning_rate": 4.452042468958071e-06, "loss": 0.3986, "step": 6224 }, { "epoch": 2.23, "learning_rate": 4.451142702897247e-06, "loss": 0.3396, "step": 6225 }, { "epoch": 2.23, "learning_rate": 4.450242936836423e-06, "loss": 0.3302, "step": 6226 }, { "epoch": 2.24, "learning_rate": 4.449343170775599e-06, "loss": 0.3766, "step": 6227 }, { "epoch": 2.24, "learning_rate": 4.448443404714775e-06, "loss": 0.338, "step": 6228 }, { "epoch": 2.24, "learning_rate": 4.44754363865395e-06, "loss": 0.2876, "step": 6229 }, { "epoch": 2.24, "learning_rate": 4.446643872593126e-06, "loss": 0.375, "step": 6230 }, { "epoch": 2.24, "learning_rate": 4.445744106532302e-06, "loss": 0.3139, "step": 6231 }, { "epoch": 2.24, "learning_rate": 4.444844340471478e-06, "loss": 0.2999, "step": 6232 }, { "epoch": 2.24, "learning_rate": 4.4439445744106535e-06, "loss": 0.3538, "step": 6233 }, { "epoch": 2.24, "learning_rate": 4.443044808349829e-06, "loss": 0.346, "step": 6234 }, { "epoch": 2.24, "learning_rate": 4.442145042289005e-06, "loss": 0.3084, "step": 6235 }, { "epoch": 2.24, "learning_rate": 4.441245276228181e-06, "loss": 0.3499, "step": 6236 }, { "epoch": 2.24, "learning_rate": 4.440345510167357e-06, "loss": 0.3343, "step": 6237 }, { "epoch": 2.24, "learning_rate": 4.439445744106532e-06, "loss": 0.3826, "step": 6238 }, { "epoch": 2.24, "learning_rate": 4.438545978045709e-06, "loss": 0.3452, "step": 6239 }, { "epoch": 2.24, "learning_rate": 4.437646211984885e-06, "loss": 0.3216, "step": 6240 }, { "epoch": 2.24, "learning_rate": 4.43674644592406e-06, "loss": 0.2765, "step": 6241 }, { "epoch": 2.24, "learning_rate": 4.435846679863236e-06, "loss": 0.3147, "step": 6242 }, { "epoch": 2.24, "learning_rate": 4.434946913802411e-06, "loss": 0.3472, "step": 6243 }, { "epoch": 2.24, "learning_rate": 4.434047147741588e-06, "loss": 0.3016, "step": 6244 }, { "epoch": 2.24, "learning_rate": 4.4331473816807635e-06, "loss": 0.3006, "step": 6245 }, { "epoch": 2.24, "learning_rate": 4.432247615619939e-06, "loss": 0.3258, "step": 6246 }, { "epoch": 2.24, "learning_rate": 4.431347849559115e-06, "loss": 0.2968, "step": 6247 }, { "epoch": 2.24, "learning_rate": 4.430448083498291e-06, "loss": 0.3248, "step": 6248 }, { "epoch": 2.24, "learning_rate": 4.429548317437467e-06, "loss": 0.3531, "step": 6249 }, { "epoch": 2.24, "learning_rate": 4.428648551376642e-06, "loss": 0.3087, "step": 6250 }, { "epoch": 2.24, "learning_rate": 4.427748785315818e-06, "loss": 0.3412, "step": 6251 }, { "epoch": 2.24, "learning_rate": 4.426849019254994e-06, "loss": 0.3695, "step": 6252 }, { "epoch": 2.24, "learning_rate": 4.42594925319417e-06, "loss": 0.3701, "step": 6253 }, { "epoch": 2.24, "learning_rate": 4.4250494871333455e-06, "loss": 0.3364, "step": 6254 }, { "epoch": 2.25, "learning_rate": 4.424149721072521e-06, "loss": 0.3767, "step": 6255 }, { "epoch": 2.25, "learning_rate": 4.423249955011698e-06, "loss": 0.3309, "step": 6256 }, { "epoch": 2.25, "learning_rate": 4.422350188950873e-06, "loss": 0.3202, "step": 6257 }, { "epoch": 2.25, "learning_rate": 4.421450422890049e-06, "loss": 0.3314, "step": 6258 }, { "epoch": 2.25, "learning_rate": 4.420550656829225e-06, "loss": 0.2713, "step": 6259 }, { "epoch": 2.25, "learning_rate": 4.4196508907684e-06, "loss": 0.2982, "step": 6260 }, { "epoch": 2.25, "learning_rate": 4.418751124707577e-06, "loss": 0.3522, "step": 6261 }, { "epoch": 2.25, "learning_rate": 4.417851358646752e-06, "loss": 0.3511, "step": 6262 }, { "epoch": 2.25, "learning_rate": 4.416951592585928e-06, "loss": 0.3809, "step": 6263 }, { "epoch": 2.25, "learning_rate": 4.416051826525104e-06, "loss": 0.3073, "step": 6264 }, { "epoch": 2.25, "learning_rate": 4.41515206046428e-06, "loss": 0.3754, "step": 6265 }, { "epoch": 2.25, "learning_rate": 4.4142522944034555e-06, "loss": 0.3145, "step": 6266 }, { "epoch": 2.25, "learning_rate": 4.413352528342631e-06, "loss": 0.3198, "step": 6267 }, { "epoch": 2.25, "learning_rate": 4.412452762281807e-06, "loss": 0.284, "step": 6268 }, { "epoch": 2.25, "learning_rate": 4.411552996220983e-06, "loss": 0.3134, "step": 6269 }, { "epoch": 2.25, "learning_rate": 4.410653230160159e-06, "loss": 0.3559, "step": 6270 }, { "epoch": 2.25, "learning_rate": 4.409753464099334e-06, "loss": 0.2874, "step": 6271 }, { "epoch": 2.25, "learning_rate": 4.40885369803851e-06, "loss": 0.3286, "step": 6272 }, { "epoch": 2.25, "learning_rate": 4.407953931977686e-06, "loss": 0.3958, "step": 6273 }, { "epoch": 2.25, "learning_rate": 4.407054165916862e-06, "loss": 0.3346, "step": 6274 }, { "epoch": 2.25, "learning_rate": 4.406154399856038e-06, "loss": 0.3468, "step": 6275 }, { "epoch": 2.25, "learning_rate": 4.405254633795213e-06, "loss": 0.2668, "step": 6276 }, { "epoch": 2.25, "learning_rate": 4.40435486773439e-06, "loss": 0.3372, "step": 6277 }, { "epoch": 2.25, "learning_rate": 4.403455101673566e-06, "loss": 0.355, "step": 6278 }, { "epoch": 2.25, "learning_rate": 4.4025553356127405e-06, "loss": 0.3503, "step": 6279 }, { "epoch": 2.25, "learning_rate": 4.401655569551917e-06, "loss": 0.3003, "step": 6280 }, { "epoch": 2.25, "learning_rate": 4.400755803491092e-06, "loss": 0.3756, "step": 6281 }, { "epoch": 2.25, "learning_rate": 4.399856037430269e-06, "loss": 0.3809, "step": 6282 }, { "epoch": 2.26, "learning_rate": 4.3989562713694445e-06, "loss": 0.3643, "step": 6283 }, { "epoch": 2.26, "learning_rate": 4.39805650530862e-06, "loss": 0.3765, "step": 6284 }, { "epoch": 2.26, "learning_rate": 4.397156739247796e-06, "loss": 0.3407, "step": 6285 }, { "epoch": 2.26, "learning_rate": 4.396256973186972e-06, "loss": 0.358, "step": 6286 }, { "epoch": 2.26, "learning_rate": 4.3953572071261475e-06, "loss": 0.3216, "step": 6287 }, { "epoch": 2.26, "learning_rate": 4.394457441065323e-06, "loss": 0.4026, "step": 6288 }, { "epoch": 2.26, "learning_rate": 4.393557675004499e-06, "loss": 0.3818, "step": 6289 }, { "epoch": 2.26, "learning_rate": 4.392657908943675e-06, "loss": 0.4003, "step": 6290 }, { "epoch": 2.26, "learning_rate": 4.391758142882851e-06, "loss": 0.4037, "step": 6291 }, { "epoch": 2.26, "learning_rate": 4.390858376822026e-06, "loss": 0.3588, "step": 6292 }, { "epoch": 2.26, "learning_rate": 4.389958610761202e-06, "loss": 0.3254, "step": 6293 }, { "epoch": 2.26, "learning_rate": 4.389058844700379e-06, "loss": 0.2308, "step": 6294 }, { "epoch": 2.26, "learning_rate": 4.388159078639554e-06, "loss": 0.3197, "step": 6295 }, { "epoch": 2.26, "learning_rate": 4.38725931257873e-06, "loss": 0.3177, "step": 6296 }, { "epoch": 2.26, "learning_rate": 4.386359546517906e-06, "loss": 0.3515, "step": 6297 }, { "epoch": 2.26, "learning_rate": 4.385459780457082e-06, "loss": 0.3632, "step": 6298 }, { "epoch": 2.26, "learning_rate": 4.384560014396258e-06, "loss": 0.2843, "step": 6299 }, { "epoch": 2.26, "learning_rate": 4.3836602483354325e-06, "loss": 0.3278, "step": 6300 }, { "epoch": 2.26, "learning_rate": 4.382760482274609e-06, "loss": 0.3224, "step": 6301 }, { "epoch": 2.26, "learning_rate": 4.381860716213785e-06, "loss": 0.3951, "step": 6302 }, { "epoch": 2.26, "learning_rate": 4.380960950152961e-06, "loss": 0.3774, "step": 6303 }, { "epoch": 2.26, "learning_rate": 4.3800611840921364e-06, "loss": 0.372, "step": 6304 }, { "epoch": 2.26, "learning_rate": 4.379161418031312e-06, "loss": 0.3941, "step": 6305 }, { "epoch": 2.26, "learning_rate": 4.378261651970488e-06, "loss": 0.3081, "step": 6306 }, { "epoch": 2.26, "learning_rate": 4.377361885909664e-06, "loss": 0.396, "step": 6307 }, { "epoch": 2.26, "learning_rate": 4.3764621198488395e-06, "loss": 0.375, "step": 6308 }, { "epoch": 2.26, "learning_rate": 4.375562353788015e-06, "loss": 0.3966, "step": 6309 }, { "epoch": 2.26, "learning_rate": 4.374662587727191e-06, "loss": 0.3077, "step": 6310 }, { "epoch": 2.27, "learning_rate": 4.373762821666367e-06, "loss": 0.4332, "step": 6311 }, { "epoch": 2.27, "learning_rate": 4.372863055605543e-06, "loss": 0.3039, "step": 6312 }, { "epoch": 2.27, "learning_rate": 4.371963289544719e-06, "loss": 0.3616, "step": 6313 }, { "epoch": 2.27, "learning_rate": 4.371063523483894e-06, "loss": 0.3508, "step": 6314 }, { "epoch": 2.27, "learning_rate": 4.370163757423071e-06, "loss": 0.306, "step": 6315 }, { "epoch": 2.27, "learning_rate": 4.3692639913622465e-06, "loss": 0.3165, "step": 6316 }, { "epoch": 2.27, "learning_rate": 4.368364225301422e-06, "loss": 0.3858, "step": 6317 }, { "epoch": 2.27, "learning_rate": 4.367464459240598e-06, "loss": 0.3527, "step": 6318 }, { "epoch": 2.27, "learning_rate": 4.366564693179773e-06, "loss": 0.3067, "step": 6319 }, { "epoch": 2.27, "learning_rate": 4.36566492711895e-06, "loss": 0.3781, "step": 6320 }, { "epoch": 2.27, "learning_rate": 4.364765161058125e-06, "loss": 0.3355, "step": 6321 }, { "epoch": 2.27, "learning_rate": 4.363865394997301e-06, "loss": 0.2889, "step": 6322 }, { "epoch": 2.27, "learning_rate": 4.362965628936477e-06, "loss": 0.3871, "step": 6323 }, { "epoch": 2.27, "learning_rate": 4.362065862875653e-06, "loss": 0.3636, "step": 6324 }, { "epoch": 2.27, "learning_rate": 4.3611660968148284e-06, "loss": 0.3791, "step": 6325 }, { "epoch": 2.27, "learning_rate": 4.360266330754004e-06, "loss": 0.3675, "step": 6326 }, { "epoch": 2.27, "learning_rate": 4.35936656469318e-06, "loss": 0.3483, "step": 6327 }, { "epoch": 2.27, "learning_rate": 4.358466798632356e-06, "loss": 0.5519, "step": 6328 }, { "epoch": 2.27, "learning_rate": 4.3575670325715315e-06, "loss": 0.3023, "step": 6329 }, { "epoch": 2.27, "learning_rate": 4.356667266510707e-06, "loss": 0.3232, "step": 6330 }, { "epoch": 2.27, "learning_rate": 4.355767500449883e-06, "loss": 0.3515, "step": 6331 }, { "epoch": 2.27, "learning_rate": 4.35486773438906e-06, "loss": 0.3609, "step": 6332 }, { "epoch": 2.27, "learning_rate": 4.353967968328235e-06, "loss": 0.3562, "step": 6333 }, { "epoch": 2.27, "learning_rate": 4.353068202267411e-06, "loss": 0.3509, "step": 6334 }, { "epoch": 2.27, "learning_rate": 4.352168436206587e-06, "loss": 0.3433, "step": 6335 }, { "epoch": 2.27, "learning_rate": 4.351268670145763e-06, "loss": 0.2899, "step": 6336 }, { "epoch": 2.27, "learning_rate": 4.3503689040849385e-06, "loss": 0.3429, "step": 6337 }, { "epoch": 2.27, "learning_rate": 4.349469138024114e-06, "loss": 0.3138, "step": 6338 }, { "epoch": 2.28, "learning_rate": 4.34856937196329e-06, "loss": 0.3366, "step": 6339 }, { "epoch": 2.28, "learning_rate": 4.347669605902466e-06, "loss": 0.2944, "step": 6340 }, { "epoch": 2.28, "learning_rate": 4.346769839841642e-06, "loss": 0.346, "step": 6341 }, { "epoch": 2.28, "learning_rate": 4.345870073780817e-06, "loss": 0.2947, "step": 6342 }, { "epoch": 2.28, "learning_rate": 4.344970307719993e-06, "loss": 0.3523, "step": 6343 }, { "epoch": 2.28, "learning_rate": 4.344070541659169e-06, "loss": 0.3595, "step": 6344 }, { "epoch": 2.28, "learning_rate": 4.343170775598345e-06, "loss": 0.3152, "step": 6345 }, { "epoch": 2.28, "learning_rate": 4.3422710095375204e-06, "loss": 0.363, "step": 6346 }, { "epoch": 2.28, "learning_rate": 4.341371243476696e-06, "loss": 0.363, "step": 6347 }, { "epoch": 2.28, "learning_rate": 4.340471477415872e-06, "loss": 0.3907, "step": 6348 }, { "epoch": 2.28, "learning_rate": 4.339571711355048e-06, "loss": 0.3265, "step": 6349 }, { "epoch": 2.28, "learning_rate": 4.3386719452942235e-06, "loss": 0.3716, "step": 6350 }, { "epoch": 2.28, "learning_rate": 4.3377721792334e-06, "loss": 0.3762, "step": 6351 }, { "epoch": 2.28, "learning_rate": 4.336872413172575e-06, "loss": 0.3192, "step": 6352 }, { "epoch": 2.28, "learning_rate": 4.335972647111752e-06, "loss": 0.3364, "step": 6353 }, { "epoch": 2.28, "learning_rate": 4.3350728810509274e-06, "loss": 0.3373, "step": 6354 }, { "epoch": 2.28, "learning_rate": 4.334173114990103e-06, "loss": 0.2817, "step": 6355 }, { "epoch": 2.28, "learning_rate": 4.333273348929279e-06, "loss": 0.3969, "step": 6356 }, { "epoch": 2.28, "learning_rate": 4.332373582868455e-06, "loss": 0.4239, "step": 6357 }, { "epoch": 2.28, "learning_rate": 4.3314738168076305e-06, "loss": 0.3628, "step": 6358 }, { "epoch": 2.28, "learning_rate": 4.330574050746806e-06, "loss": 0.4386, "step": 6359 }, { "epoch": 2.28, "learning_rate": 4.329674284685982e-06, "loss": 0.3014, "step": 6360 }, { "epoch": 2.28, "learning_rate": 4.328774518625158e-06, "loss": 0.3397, "step": 6361 }, { "epoch": 2.28, "learning_rate": 4.327874752564334e-06, "loss": 0.2846, "step": 6362 }, { "epoch": 2.28, "learning_rate": 4.326974986503509e-06, "loss": 0.3483, "step": 6363 }, { "epoch": 2.28, "learning_rate": 4.326075220442685e-06, "loss": 0.2516, "step": 6364 }, { "epoch": 2.28, "learning_rate": 4.325175454381861e-06, "loss": 0.3554, "step": 6365 }, { "epoch": 2.28, "learning_rate": 4.324275688321037e-06, "loss": 0.3872, "step": 6366 }, { "epoch": 2.29, "learning_rate": 4.323375922260213e-06, "loss": 0.3354, "step": 6367 }, { "epoch": 2.29, "learning_rate": 4.322476156199388e-06, "loss": 0.3949, "step": 6368 }, { "epoch": 2.29, "learning_rate": 4.321576390138564e-06, "loss": 0.3055, "step": 6369 }, { "epoch": 2.29, "learning_rate": 4.3206766240777406e-06, "loss": 0.3631, "step": 6370 }, { "epoch": 2.29, "learning_rate": 4.3197768580169155e-06, "loss": 0.3683, "step": 6371 }, { "epoch": 2.29, "learning_rate": 4.318877091956092e-06, "loss": 0.3387, "step": 6372 }, { "epoch": 2.29, "learning_rate": 4.317977325895268e-06, "loss": 0.3129, "step": 6373 }, { "epoch": 2.29, "learning_rate": 4.317077559834444e-06, "loss": 0.316, "step": 6374 }, { "epoch": 2.29, "learning_rate": 4.316177793773619e-06, "loss": 0.2837, "step": 6375 }, { "epoch": 2.29, "learning_rate": 4.315278027712795e-06, "loss": 0.3459, "step": 6376 }, { "epoch": 2.29, "learning_rate": 4.314378261651971e-06, "loss": 0.3569, "step": 6377 }, { "epoch": 2.29, "learning_rate": 4.313478495591147e-06, "loss": 0.3813, "step": 6378 }, { "epoch": 2.29, "learning_rate": 4.3125787295303225e-06, "loss": 0.3809, "step": 6379 }, { "epoch": 2.29, "learning_rate": 4.311678963469498e-06, "loss": 0.3172, "step": 6380 }, { "epoch": 2.29, "learning_rate": 4.310779197408674e-06, "loss": 0.3639, "step": 6381 }, { "epoch": 2.29, "learning_rate": 4.30987943134785e-06, "loss": 0.3542, "step": 6382 }, { "epoch": 2.29, "learning_rate": 4.3089796652870256e-06, "loss": 0.3073, "step": 6383 }, { "epoch": 2.29, "learning_rate": 4.308079899226201e-06, "loss": 0.3138, "step": 6384 }, { "epoch": 2.29, "learning_rate": 4.307180133165377e-06, "loss": 0.3729, "step": 6385 }, { "epoch": 2.29, "learning_rate": 4.306280367104554e-06, "loss": 0.2691, "step": 6386 }, { "epoch": 2.29, "learning_rate": 4.305380601043729e-06, "loss": 0.3168, "step": 6387 }, { "epoch": 2.29, "learning_rate": 4.304480834982904e-06, "loss": 0.2889, "step": 6388 }, { "epoch": 2.29, "learning_rate": 4.303581068922081e-06, "loss": 0.3696, "step": 6389 }, { "epoch": 2.29, "learning_rate": 4.302681302861256e-06, "loss": 0.3546, "step": 6390 }, { "epoch": 2.29, "learning_rate": 4.3017815368004326e-06, "loss": 0.3397, "step": 6391 }, { "epoch": 2.29, "learning_rate": 4.300881770739608e-06, "loss": 0.3525, "step": 6392 }, { "epoch": 2.29, "learning_rate": 4.299982004678784e-06, "loss": 0.2628, "step": 6393 }, { "epoch": 2.3, "learning_rate": 4.29908223861796e-06, "loss": 0.3752, "step": 6394 }, { "epoch": 2.3, "learning_rate": 4.298182472557136e-06, "loss": 0.3458, "step": 6395 }, { "epoch": 2.3, "learning_rate": 4.297282706496311e-06, "loss": 0.2821, "step": 6396 }, { "epoch": 2.3, "learning_rate": 4.296382940435487e-06, "loss": 0.2876, "step": 6397 }, { "epoch": 2.3, "learning_rate": 4.295483174374663e-06, "loss": 0.4212, "step": 6398 }, { "epoch": 2.3, "learning_rate": 4.294583408313839e-06, "loss": 0.3114, "step": 6399 }, { "epoch": 2.3, "learning_rate": 4.2936836422530145e-06, "loss": 0.3627, "step": 6400 }, { "epoch": 2.3, "learning_rate": 4.29278387619219e-06, "loss": 0.3649, "step": 6401 }, { "epoch": 2.3, "learning_rate": 4.291884110131366e-06, "loss": 0.3225, "step": 6402 }, { "epoch": 2.3, "learning_rate": 4.290984344070542e-06, "loss": 0.356, "step": 6403 }, { "epoch": 2.3, "learning_rate": 4.2900845780097176e-06, "loss": 0.3119, "step": 6404 }, { "epoch": 2.3, "learning_rate": 4.289184811948894e-06, "loss": 0.4054, "step": 6405 }, { "epoch": 2.3, "learning_rate": 4.288285045888069e-06, "loss": 0.3319, "step": 6406 }, { "epoch": 2.3, "learning_rate": 4.287385279827245e-06, "loss": 0.2481, "step": 6407 }, { "epoch": 2.3, "learning_rate": 4.2864855137664215e-06, "loss": 0.354, "step": 6408 }, { "epoch": 2.3, "learning_rate": 4.285585747705596e-06, "loss": 0.3528, "step": 6409 }, { "epoch": 2.3, "learning_rate": 4.284685981644773e-06, "loss": 0.3962, "step": 6410 }, { "epoch": 2.3, "learning_rate": 4.283786215583948e-06, "loss": 0.3628, "step": 6411 }, { "epoch": 2.3, "learning_rate": 4.2828864495231246e-06, "loss": 0.3081, "step": 6412 }, { "epoch": 2.3, "learning_rate": 4.2819866834623e-06, "loss": 0.3391, "step": 6413 }, { "epoch": 2.3, "learning_rate": 4.281086917401476e-06, "loss": 0.3239, "step": 6414 }, { "epoch": 2.3, "learning_rate": 4.280187151340652e-06, "loss": 0.3271, "step": 6415 }, { "epoch": 2.3, "learning_rate": 4.279287385279828e-06, "loss": 0.3459, "step": 6416 }, { "epoch": 2.3, "learning_rate": 4.278387619219003e-06, "loss": 0.3626, "step": 6417 }, { "epoch": 2.3, "learning_rate": 4.277487853158179e-06, "loss": 0.3489, "step": 6418 }, { "epoch": 2.3, "learning_rate": 4.276588087097355e-06, "loss": 0.3131, "step": 6419 }, { "epoch": 2.3, "learning_rate": 4.275688321036531e-06, "loss": 0.3991, "step": 6420 }, { "epoch": 2.3, "learning_rate": 4.2747885549757065e-06, "loss": 0.3935, "step": 6421 }, { "epoch": 2.31, "learning_rate": 4.273888788914882e-06, "loss": 0.3629, "step": 6422 }, { "epoch": 2.31, "learning_rate": 4.272989022854058e-06, "loss": 0.2926, "step": 6423 }, { "epoch": 2.31, "learning_rate": 4.272089256793235e-06, "loss": 0.2964, "step": 6424 }, { "epoch": 2.31, "learning_rate": 4.2711894907324096e-06, "loss": 0.3378, "step": 6425 }, { "epoch": 2.31, "learning_rate": 4.270289724671586e-06, "loss": 0.4156, "step": 6426 }, { "epoch": 2.31, "learning_rate": 4.269389958610762e-06, "loss": 0.3086, "step": 6427 }, { "epoch": 2.31, "learning_rate": 4.268490192549937e-06, "loss": 0.4559, "step": 6428 }, { "epoch": 2.31, "learning_rate": 4.2675904264891135e-06, "loss": 0.3637, "step": 6429 }, { "epoch": 2.31, "learning_rate": 4.266690660428288e-06, "loss": 0.3248, "step": 6430 }, { "epoch": 2.31, "learning_rate": 4.265790894367465e-06, "loss": 0.2902, "step": 6431 }, { "epoch": 2.31, "learning_rate": 4.264891128306641e-06, "loss": 0.2917, "step": 6432 }, { "epoch": 2.31, "learning_rate": 4.2639913622458166e-06, "loss": 0.2978, "step": 6433 }, { "epoch": 2.31, "learning_rate": 4.263091596184992e-06, "loss": 0.3305, "step": 6434 }, { "epoch": 2.31, "learning_rate": 4.262191830124168e-06, "loss": 0.2853, "step": 6435 }, { "epoch": 2.31, "learning_rate": 4.261292064063344e-06, "loss": 0.3801, "step": 6436 }, { "epoch": 2.31, "learning_rate": 4.26039229800252e-06, "loss": 0.3055, "step": 6437 }, { "epoch": 2.31, "learning_rate": 4.259492531941695e-06, "loss": 0.296, "step": 6438 }, { "epoch": 2.31, "learning_rate": 4.258592765880871e-06, "loss": 0.3299, "step": 6439 }, { "epoch": 2.31, "learning_rate": 4.257692999820047e-06, "loss": 0.3481, "step": 6440 }, { "epoch": 2.31, "learning_rate": 4.256793233759223e-06, "loss": 0.3718, "step": 6441 }, { "epoch": 2.31, "learning_rate": 4.2558934676983985e-06, "loss": 0.3258, "step": 6442 }, { "epoch": 2.31, "learning_rate": 4.254993701637575e-06, "loss": 0.3372, "step": 6443 }, { "epoch": 2.31, "learning_rate": 4.25409393557675e-06, "loss": 0.2594, "step": 6444 }, { "epoch": 2.31, "learning_rate": 4.253194169515927e-06, "loss": 0.3207, "step": 6445 }, { "epoch": 2.31, "learning_rate": 4.252294403455102e-06, "loss": 0.3062, "step": 6446 }, { "epoch": 2.31, "learning_rate": 4.251394637394277e-06, "loss": 0.3305, "step": 6447 }, { "epoch": 2.31, "learning_rate": 4.250494871333454e-06, "loss": 0.356, "step": 6448 }, { "epoch": 2.31, "learning_rate": 4.249595105272629e-06, "loss": 0.3043, "step": 6449 }, { "epoch": 2.32, "learning_rate": 4.2486953392118055e-06, "loss": 0.3533, "step": 6450 }, { "epoch": 2.32, "learning_rate": 4.247795573150981e-06, "loss": 0.3488, "step": 6451 }, { "epoch": 2.32, "learning_rate": 4.246895807090157e-06, "loss": 0.4109, "step": 6452 }, { "epoch": 2.32, "learning_rate": 4.245996041029333e-06, "loss": 0.2968, "step": 6453 }, { "epoch": 2.32, "learning_rate": 4.2450962749685085e-06, "loss": 0.3526, "step": 6454 }, { "epoch": 2.32, "learning_rate": 4.244196508907684e-06, "loss": 0.3893, "step": 6455 }, { "epoch": 2.32, "learning_rate": 4.24329674284686e-06, "loss": 0.3307, "step": 6456 }, { "epoch": 2.32, "learning_rate": 4.242396976786036e-06, "loss": 0.274, "step": 6457 }, { "epoch": 2.32, "learning_rate": 4.241497210725212e-06, "loss": 0.3667, "step": 6458 }, { "epoch": 2.32, "learning_rate": 4.240597444664387e-06, "loss": 0.3262, "step": 6459 }, { "epoch": 2.32, "learning_rate": 4.239697678603563e-06, "loss": 0.3222, "step": 6460 }, { "epoch": 2.32, "learning_rate": 4.238797912542739e-06, "loss": 0.3657, "step": 6461 }, { "epoch": 2.32, "learning_rate": 4.2378981464819155e-06, "loss": 0.3181, "step": 6462 }, { "epoch": 2.32, "learning_rate": 4.2369983804210905e-06, "loss": 0.366, "step": 6463 }, { "epoch": 2.32, "learning_rate": 4.236098614360267e-06, "loss": 0.3166, "step": 6464 }, { "epoch": 2.32, "learning_rate": 4.235198848299443e-06, "loss": 0.3562, "step": 6465 }, { "epoch": 2.32, "learning_rate": 4.234299082238618e-06, "loss": 0.3587, "step": 6466 }, { "epoch": 2.32, "learning_rate": 4.233399316177794e-06, "loss": 0.2914, "step": 6467 }, { "epoch": 2.32, "learning_rate": 4.232499550116969e-06, "loss": 0.2917, "step": 6468 }, { "epoch": 2.32, "learning_rate": 4.231599784056146e-06, "loss": 0.3458, "step": 6469 }, { "epoch": 2.32, "learning_rate": 4.230700017995322e-06, "loss": 0.385, "step": 6470 }, { "epoch": 2.32, "learning_rate": 4.2298002519344975e-06, "loss": 0.3024, "step": 6471 }, { "epoch": 2.32, "learning_rate": 4.228900485873673e-06, "loss": 0.3799, "step": 6472 }, { "epoch": 2.32, "learning_rate": 4.228000719812849e-06, "loss": 0.3558, "step": 6473 }, { "epoch": 2.32, "learning_rate": 4.227100953752025e-06, "loss": 0.3577, "step": 6474 }, { "epoch": 2.32, "learning_rate": 4.2262011876912005e-06, "loss": 0.2573, "step": 6475 }, { "epoch": 2.32, "learning_rate": 4.225301421630376e-06, "loss": 0.3676, "step": 6476 }, { "epoch": 2.32, "learning_rate": 4.224401655569552e-06, "loss": 0.2719, "step": 6477 }, { "epoch": 2.33, "learning_rate": 4.223501889508728e-06, "loss": 0.286, "step": 6478 }, { "epoch": 2.33, "learning_rate": 4.222602123447904e-06, "loss": 0.3014, "step": 6479 }, { "epoch": 2.33, "learning_rate": 4.221702357387079e-06, "loss": 0.3541, "step": 6480 }, { "epoch": 2.33, "learning_rate": 4.220802591326256e-06, "loss": 0.283, "step": 6481 }, { "epoch": 2.33, "learning_rate": 4.219902825265431e-06, "loss": 0.3751, "step": 6482 }, { "epoch": 2.33, "learning_rate": 4.2190030592046075e-06, "loss": 0.346, "step": 6483 }, { "epoch": 2.33, "learning_rate": 4.218103293143783e-06, "loss": 0.4025, "step": 6484 }, { "epoch": 2.33, "learning_rate": 4.217203527082959e-06, "loss": 0.3566, "step": 6485 }, { "epoch": 2.33, "learning_rate": 4.216303761022135e-06, "loss": 0.3931, "step": 6486 }, { "epoch": 2.33, "learning_rate": 4.21540399496131e-06, "loss": 0.3093, "step": 6487 }, { "epoch": 2.33, "learning_rate": 4.214504228900486e-06, "loss": 0.3517, "step": 6488 }, { "epoch": 2.33, "learning_rate": 4.213604462839662e-06, "loss": 0.3524, "step": 6489 }, { "epoch": 2.33, "learning_rate": 4.212704696778838e-06, "loss": 0.4403, "step": 6490 }, { "epoch": 2.33, "learning_rate": 4.211804930718014e-06, "loss": 0.3623, "step": 6491 }, { "epoch": 2.33, "learning_rate": 4.2109051646571894e-06, "loss": 0.3238, "step": 6492 }, { "epoch": 2.33, "learning_rate": 4.210005398596365e-06, "loss": 0.3608, "step": 6493 }, { "epoch": 2.33, "learning_rate": 4.209105632535541e-06, "loss": 0.3212, "step": 6494 }, { "epoch": 2.33, "learning_rate": 4.208205866474718e-06, "loss": 0.4131, "step": 6495 }, { "epoch": 2.33, "learning_rate": 4.2073061004138925e-06, "loss": 0.3736, "step": 6496 }, { "epoch": 2.33, "learning_rate": 4.206406334353068e-06, "loss": 0.3799, "step": 6497 }, { "epoch": 2.33, "learning_rate": 4.205506568292244e-06, "loss": 0.2675, "step": 6498 }, { "epoch": 2.33, "learning_rate": 4.20460680223142e-06, "loss": 0.3622, "step": 6499 }, { "epoch": 2.33, "learning_rate": 4.2037070361705964e-06, "loss": 0.3412, "step": 6500 }, { "epoch": 2.33, "learning_rate": 4.202807270109771e-06, "loss": 0.3524, "step": 6501 }, { "epoch": 2.33, "learning_rate": 4.201907504048948e-06, "loss": 0.3317, "step": 6502 }, { "epoch": 2.33, "learning_rate": 4.201007737988124e-06, "loss": 0.3483, "step": 6503 }, { "epoch": 2.33, "learning_rate": 4.2001079719272995e-06, "loss": 0.3459, "step": 6504 }, { "epoch": 2.33, "learning_rate": 4.199208205866475e-06, "loss": 0.3664, "step": 6505 }, { "epoch": 2.34, "learning_rate": 4.19830843980565e-06, "loss": 0.3369, "step": 6506 }, { "epoch": 2.34, "learning_rate": 4.197408673744827e-06, "loss": 0.3935, "step": 6507 }, { "epoch": 2.34, "learning_rate": 4.196508907684003e-06, "loss": 0.3408, "step": 6508 }, { "epoch": 2.34, "learning_rate": 4.195609141623178e-06, "loss": 0.3119, "step": 6509 }, { "epoch": 2.34, "learning_rate": 4.194709375562354e-06, "loss": 0.3639, "step": 6510 }, { "epoch": 2.34, "learning_rate": 4.19380960950153e-06, "loss": 0.3312, "step": 6511 }, { "epoch": 2.34, "learning_rate": 4.192909843440706e-06, "loss": 0.3288, "step": 6512 }, { "epoch": 2.34, "learning_rate": 4.1920100773798814e-06, "loss": 0.3091, "step": 6513 }, { "epoch": 2.34, "learning_rate": 4.191110311319058e-06, "loss": 0.3229, "step": 6514 }, { "epoch": 2.34, "learning_rate": 4.190210545258233e-06, "loss": 0.304, "step": 6515 }, { "epoch": 2.34, "learning_rate": 4.189310779197409e-06, "loss": 0.3319, "step": 6516 }, { "epoch": 2.34, "learning_rate": 4.1884110131365845e-06, "loss": 0.2955, "step": 6517 }, { "epoch": 2.34, "learning_rate": 4.18751124707576e-06, "loss": 0.3306, "step": 6518 }, { "epoch": 2.34, "learning_rate": 4.186611481014937e-06, "loss": 0.3919, "step": 6519 }, { "epoch": 2.34, "learning_rate": 4.185711714954112e-06, "loss": 0.3617, "step": 6520 }, { "epoch": 2.34, "learning_rate": 4.1848119488932884e-06, "loss": 0.3662, "step": 6521 }, { "epoch": 2.34, "learning_rate": 4.183912182832464e-06, "loss": 0.3533, "step": 6522 }, { "epoch": 2.34, "learning_rate": 4.18301241677164e-06, "loss": 0.305, "step": 6523 }, { "epoch": 2.34, "learning_rate": 4.182112650710816e-06, "loss": 0.3605, "step": 6524 }, { "epoch": 2.34, "learning_rate": 4.181212884649991e-06, "loss": 0.3082, "step": 6525 }, { "epoch": 2.34, "learning_rate": 4.180313118589167e-06, "loss": 0.3616, "step": 6526 }, { "epoch": 2.34, "learning_rate": 4.179413352528343e-06, "loss": 0.2915, "step": 6527 }, { "epoch": 2.34, "learning_rate": 4.178513586467519e-06, "loss": 0.2598, "step": 6528 }, { "epoch": 2.34, "learning_rate": 4.177613820406695e-06, "loss": 0.3525, "step": 6529 }, { "epoch": 2.34, "learning_rate": 4.17671405434587e-06, "loss": 0.3542, "step": 6530 }, { "epoch": 2.34, "learning_rate": 4.175814288285046e-06, "loss": 0.3174, "step": 6531 }, { "epoch": 2.34, "learning_rate": 4.174914522224222e-06, "loss": 0.3628, "step": 6532 }, { "epoch": 2.34, "learning_rate": 4.1740147561633985e-06, "loss": 0.321, "step": 6533 }, { "epoch": 2.35, "learning_rate": 4.1731149901025734e-06, "loss": 0.3557, "step": 6534 }, { "epoch": 2.35, "learning_rate": 4.172215224041749e-06, "loss": 0.3273, "step": 6535 }, { "epoch": 2.35, "learning_rate": 4.171315457980925e-06, "loss": 0.3092, "step": 6536 }, { "epoch": 2.35, "learning_rate": 4.170415691920101e-06, "loss": 0.314, "step": 6537 }, { "epoch": 2.35, "learning_rate": 4.169515925859277e-06, "loss": 0.2798, "step": 6538 }, { "epoch": 2.35, "learning_rate": 4.168616159798452e-06, "loss": 0.2596, "step": 6539 }, { "epoch": 2.35, "learning_rate": 4.167716393737629e-06, "loss": 0.3725, "step": 6540 }, { "epoch": 2.35, "learning_rate": 4.166816627676805e-06, "loss": 0.416, "step": 6541 }, { "epoch": 2.35, "learning_rate": 4.1659168616159804e-06, "loss": 0.4049, "step": 6542 }, { "epoch": 2.35, "learning_rate": 4.165017095555156e-06, "loss": 0.3492, "step": 6543 }, { "epoch": 2.35, "learning_rate": 4.164117329494332e-06, "loss": 0.3072, "step": 6544 }, { "epoch": 2.35, "learning_rate": 4.163217563433508e-06, "loss": 0.3321, "step": 6545 }, { "epoch": 2.35, "learning_rate": 4.1623177973726835e-06, "loss": 0.3065, "step": 6546 }, { "epoch": 2.35, "learning_rate": 4.161418031311859e-06, "loss": 0.3754, "step": 6547 }, { "epoch": 2.35, "learning_rate": 4.160518265251035e-06, "loss": 0.3337, "step": 6548 }, { "epoch": 2.35, "learning_rate": 4.159618499190211e-06, "loss": 0.3248, "step": 6549 }, { "epoch": 2.35, "learning_rate": 4.158718733129387e-06, "loss": 0.3878, "step": 6550 }, { "epoch": 2.35, "learning_rate": 4.157818967068562e-06, "loss": 0.3308, "step": 6551 }, { "epoch": 2.35, "learning_rate": 4.156919201007739e-06, "loss": 0.3222, "step": 6552 }, { "epoch": 2.35, "learning_rate": 4.156019434946914e-06, "loss": 0.3449, "step": 6553 }, { "epoch": 2.35, "learning_rate": 4.1551196688860905e-06, "loss": 0.3091, "step": 6554 }, { "epoch": 2.35, "learning_rate": 4.1542199028252654e-06, "loss": 0.3103, "step": 6555 }, { "epoch": 2.35, "learning_rate": 4.153320136764441e-06, "loss": 0.4002, "step": 6556 }, { "epoch": 2.35, "learning_rate": 4.152420370703618e-06, "loss": 0.396, "step": 6557 }, { "epoch": 2.35, "learning_rate": 4.151520604642793e-06, "loss": 0.3702, "step": 6558 }, { "epoch": 2.35, "learning_rate": 4.150620838581969e-06, "loss": 0.3669, "step": 6559 }, { "epoch": 2.35, "learning_rate": 4.149721072521145e-06, "loss": 0.2935, "step": 6560 }, { "epoch": 2.35, "learning_rate": 4.148821306460321e-06, "loss": 0.3145, "step": 6561 }, { "epoch": 2.36, "learning_rate": 4.147921540399497e-06, "loss": 0.3883, "step": 6562 }, { "epoch": 2.36, "learning_rate": 4.147021774338672e-06, "loss": 0.3849, "step": 6563 }, { "epoch": 2.36, "learning_rate": 4.146122008277848e-06, "loss": 0.3076, "step": 6564 }, { "epoch": 2.36, "learning_rate": 4.145222242217024e-06, "loss": 0.3112, "step": 6565 }, { "epoch": 2.36, "learning_rate": 4.1443224761562e-06, "loss": 0.3171, "step": 6566 }, { "epoch": 2.36, "learning_rate": 4.1434227100953755e-06, "loss": 0.3755, "step": 6567 }, { "epoch": 2.36, "learning_rate": 4.142522944034551e-06, "loss": 0.323, "step": 6568 }, { "epoch": 2.36, "learning_rate": 4.141623177973727e-06, "loss": 0.3042, "step": 6569 }, { "epoch": 2.36, "learning_rate": 4.140723411912903e-06, "loss": 0.3432, "step": 6570 }, { "epoch": 2.36, "learning_rate": 4.1398236458520786e-06, "loss": 0.3256, "step": 6571 }, { "epoch": 2.36, "learning_rate": 4.138923879791254e-06, "loss": 0.3428, "step": 6572 }, { "epoch": 2.36, "learning_rate": 4.138024113730431e-06, "loss": 0.3396, "step": 6573 }, { "epoch": 2.36, "learning_rate": 4.137124347669606e-06, "loss": 0.3054, "step": 6574 }, { "epoch": 2.36, "learning_rate": 4.136224581608782e-06, "loss": 0.3442, "step": 6575 }, { "epoch": 2.36, "learning_rate": 4.135324815547958e-06, "loss": 0.3028, "step": 6576 }, { "epoch": 2.36, "learning_rate": 4.134425049487133e-06, "loss": 0.2896, "step": 6577 }, { "epoch": 2.36, "learning_rate": 4.13352528342631e-06, "loss": 0.3055, "step": 6578 }, { "epoch": 2.36, "learning_rate": 4.1326255173654856e-06, "loss": 0.3937, "step": 6579 }, { "epoch": 2.36, "learning_rate": 4.131725751304661e-06, "loss": 0.3806, "step": 6580 }, { "epoch": 2.36, "learning_rate": 4.130825985243837e-06, "loss": 0.3581, "step": 6581 }, { "epoch": 2.36, "learning_rate": 4.129926219183013e-06, "loss": 0.3831, "step": 6582 }, { "epoch": 2.36, "learning_rate": 4.129026453122189e-06, "loss": 0.2657, "step": 6583 }, { "epoch": 2.36, "learning_rate": 4.128126687061364e-06, "loss": 0.3241, "step": 6584 }, { "epoch": 2.36, "learning_rate": 4.12722692100054e-06, "loss": 0.403, "step": 6585 }, { "epoch": 2.36, "learning_rate": 4.126327154939716e-06, "loss": 0.2887, "step": 6586 }, { "epoch": 2.36, "learning_rate": 4.125427388878892e-06, "loss": 0.3136, "step": 6587 }, { "epoch": 2.36, "learning_rate": 4.1245276228180675e-06, "loss": 0.3991, "step": 6588 }, { "epoch": 2.37, "learning_rate": 4.123627856757243e-06, "loss": 0.3193, "step": 6589 }, { "epoch": 2.37, "learning_rate": 4.122728090696419e-06, "loss": 0.3327, "step": 6590 }, { "epoch": 2.37, "learning_rate": 4.121828324635595e-06, "loss": 0.3769, "step": 6591 }, { "epoch": 2.37, "learning_rate": 4.120928558574771e-06, "loss": 0.3058, "step": 6592 }, { "epoch": 2.37, "learning_rate": 4.120028792513946e-06, "loss": 0.4112, "step": 6593 }, { "epoch": 2.37, "learning_rate": 4.119129026453122e-06, "loss": 0.3344, "step": 6594 }, { "epoch": 2.37, "learning_rate": 4.118229260392299e-06, "loss": 0.287, "step": 6595 }, { "epoch": 2.37, "learning_rate": 4.117329494331474e-06, "loss": 0.3395, "step": 6596 }, { "epoch": 2.37, "learning_rate": 4.11642972827065e-06, "loss": 0.3108, "step": 6597 }, { "epoch": 2.37, "learning_rate": 4.115529962209826e-06, "loss": 0.3138, "step": 6598 }, { "epoch": 2.37, "learning_rate": 4.114630196149002e-06, "loss": 0.3747, "step": 6599 }, { "epoch": 2.37, "learning_rate": 4.1137304300881776e-06, "loss": 0.3781, "step": 6600 }, { "epoch": 2.37, "learning_rate": 4.112830664027353e-06, "loss": 0.3479, "step": 6601 }, { "epoch": 2.37, "learning_rate": 4.111930897966529e-06, "loss": 0.3835, "step": 6602 }, { "epoch": 2.37, "learning_rate": 4.111031131905705e-06, "loss": 0.2471, "step": 6603 }, { "epoch": 2.37, "learning_rate": 4.110131365844881e-06, "loss": 0.3518, "step": 6604 }, { "epoch": 2.37, "learning_rate": 4.109231599784056e-06, "loss": 0.3026, "step": 6605 }, { "epoch": 2.37, "learning_rate": 4.108331833723232e-06, "loss": 0.3515, "step": 6606 }, { "epoch": 2.37, "learning_rate": 4.107432067662408e-06, "loss": 0.3801, "step": 6607 }, { "epoch": 2.37, "learning_rate": 4.106532301601584e-06, "loss": 0.3223, "step": 6608 }, { "epoch": 2.37, "learning_rate": 4.1056325355407595e-06, "loss": 0.3501, "step": 6609 }, { "epoch": 2.37, "learning_rate": 4.104732769479935e-06, "loss": 0.3259, "step": 6610 }, { "epoch": 2.37, "learning_rate": 4.103833003419112e-06, "loss": 0.2949, "step": 6611 }, { "epoch": 2.37, "learning_rate": 4.102933237358287e-06, "loss": 0.3582, "step": 6612 }, { "epoch": 2.37, "learning_rate": 4.102033471297463e-06, "loss": 0.3192, "step": 6613 }, { "epoch": 2.37, "learning_rate": 4.101133705236639e-06, "loss": 0.3539, "step": 6614 }, { "epoch": 2.37, "learning_rate": 4.100233939175814e-06, "loss": 0.3993, "step": 6615 }, { "epoch": 2.37, "learning_rate": 4.099334173114991e-06, "loss": 0.3229, "step": 6616 }, { "epoch": 2.38, "learning_rate": 4.0984344070541665e-06, "loss": 0.3563, "step": 6617 }, { "epoch": 2.38, "learning_rate": 4.097534640993342e-06, "loss": 0.2885, "step": 6618 }, { "epoch": 2.38, "learning_rate": 4.096634874932518e-06, "loss": 0.313, "step": 6619 }, { "epoch": 2.38, "learning_rate": 4.095735108871694e-06, "loss": 0.3197, "step": 6620 }, { "epoch": 2.38, "learning_rate": 4.0948353428108696e-06, "loss": 0.3268, "step": 6621 }, { "epoch": 2.38, "learning_rate": 4.093935576750045e-06, "loss": 0.3704, "step": 6622 }, { "epoch": 2.38, "learning_rate": 4.093035810689221e-06, "loss": 0.3502, "step": 6623 }, { "epoch": 2.38, "learning_rate": 4.092136044628397e-06, "loss": 0.3618, "step": 6624 }, { "epoch": 2.38, "learning_rate": 4.091236278567573e-06, "loss": 0.3058, "step": 6625 }, { "epoch": 2.38, "learning_rate": 4.090336512506748e-06, "loss": 0.3327, "step": 6626 }, { "epoch": 2.38, "learning_rate": 4.089436746445924e-06, "loss": 0.3349, "step": 6627 }, { "epoch": 2.38, "learning_rate": 4.0885369803851e-06, "loss": 0.3089, "step": 6628 }, { "epoch": 2.38, "learning_rate": 4.087637214324276e-06, "loss": 0.3277, "step": 6629 }, { "epoch": 2.38, "learning_rate": 4.086737448263452e-06, "loss": 0.3026, "step": 6630 }, { "epoch": 2.38, "learning_rate": 4.085837682202627e-06, "loss": 0.3211, "step": 6631 }, { "epoch": 2.38, "learning_rate": 4.084937916141804e-06, "loss": 0.3299, "step": 6632 }, { "epoch": 2.38, "learning_rate": 4.08403815008098e-06, "loss": 0.4054, "step": 6633 }, { "epoch": 2.38, "learning_rate": 4.0831383840201545e-06, "loss": 0.387, "step": 6634 }, { "epoch": 2.38, "learning_rate": 4.082238617959331e-06, "loss": 0.3543, "step": 6635 }, { "epoch": 2.38, "learning_rate": 4.081338851898507e-06, "loss": 0.3597, "step": 6636 }, { "epoch": 2.38, "learning_rate": 4.080439085837683e-06, "loss": 0.3335, "step": 6637 }, { "epoch": 2.38, "learning_rate": 4.0795393197768585e-06, "loss": 0.3498, "step": 6638 }, { "epoch": 2.38, "learning_rate": 4.078639553716034e-06, "loss": 0.3689, "step": 6639 }, { "epoch": 2.38, "learning_rate": 4.07773978765521e-06, "loss": 0.2954, "step": 6640 }, { "epoch": 2.38, "learning_rate": 4.076840021594386e-06, "loss": 0.4074, "step": 6641 }, { "epoch": 2.38, "learning_rate": 4.0759402555335615e-06, "loss": 0.3834, "step": 6642 }, { "epoch": 2.38, "learning_rate": 4.075040489472737e-06, "loss": 0.4019, "step": 6643 }, { "epoch": 2.38, "learning_rate": 4.074140723411913e-06, "loss": 0.2721, "step": 6644 }, { "epoch": 2.39, "learning_rate": 4.073240957351089e-06, "loss": 0.2915, "step": 6645 }, { "epoch": 2.39, "learning_rate": 4.072341191290265e-06, "loss": 0.3042, "step": 6646 }, { "epoch": 2.39, "learning_rate": 4.07144142522944e-06, "loss": 0.3601, "step": 6647 }, { "epoch": 2.39, "learning_rate": 4.070541659168616e-06, "loss": 0.2558, "step": 6648 }, { "epoch": 2.39, "learning_rate": 4.069641893107793e-06, "loss": 0.2985, "step": 6649 }, { "epoch": 2.39, "learning_rate": 4.068742127046968e-06, "loss": 0.3006, "step": 6650 }, { "epoch": 2.39, "learning_rate": 4.067842360986144e-06, "loss": 0.255, "step": 6651 }, { "epoch": 2.39, "learning_rate": 4.06694259492532e-06, "loss": 0.3106, "step": 6652 }, { "epoch": 2.39, "learning_rate": 4.066042828864495e-06, "loss": 0.3156, "step": 6653 }, { "epoch": 2.39, "learning_rate": 4.065143062803672e-06, "loss": 0.3593, "step": 6654 }, { "epoch": 2.39, "learning_rate": 4.064243296742847e-06, "loss": 0.3101, "step": 6655 }, { "epoch": 2.39, "learning_rate": 4.063343530682023e-06, "loss": 0.324, "step": 6656 }, { "epoch": 2.39, "learning_rate": 4.062443764621199e-06, "loss": 0.3047, "step": 6657 }, { "epoch": 2.39, "learning_rate": 4.061543998560375e-06, "loss": 0.3301, "step": 6658 }, { "epoch": 2.39, "learning_rate": 4.0606442324995505e-06, "loss": 0.3464, "step": 6659 }, { "epoch": 2.39, "learning_rate": 4.059744466438726e-06, "loss": 0.357, "step": 6660 }, { "epoch": 2.39, "learning_rate": 4.058844700377902e-06, "loss": 0.2835, "step": 6661 }, { "epoch": 2.39, "learning_rate": 4.057944934317078e-06, "loss": 0.3521, "step": 6662 }, { "epoch": 2.39, "learning_rate": 4.0570451682562535e-06, "loss": 0.3727, "step": 6663 }, { "epoch": 2.39, "learning_rate": 4.056145402195429e-06, "loss": 0.3202, "step": 6664 }, { "epoch": 2.39, "learning_rate": 4.055245636134605e-06, "loss": 0.3044, "step": 6665 }, { "epoch": 2.39, "learning_rate": 4.054345870073781e-06, "loss": 0.2797, "step": 6666 }, { "epoch": 2.39, "learning_rate": 4.053446104012957e-06, "loss": 0.3404, "step": 6667 }, { "epoch": 2.39, "learning_rate": 4.052546337952133e-06, "loss": 0.3283, "step": 6668 }, { "epoch": 2.39, "learning_rate": 4.051646571891308e-06, "loss": 0.2802, "step": 6669 }, { "epoch": 2.39, "learning_rate": 4.050746805830485e-06, "loss": 0.324, "step": 6670 }, { "epoch": 2.39, "learning_rate": 4.0498470397696605e-06, "loss": 0.3298, "step": 6671 }, { "epoch": 2.39, "learning_rate": 4.048947273708836e-06, "loss": 0.2827, "step": 6672 }, { "epoch": 2.4, "learning_rate": 4.048047507648012e-06, "loss": 0.2919, "step": 6673 }, { "epoch": 2.4, "learning_rate": 4.047147741587188e-06, "loss": 0.3292, "step": 6674 }, { "epoch": 2.4, "learning_rate": 4.046247975526364e-06, "loss": 0.3134, "step": 6675 }, { "epoch": 2.4, "learning_rate": 4.045348209465539e-06, "loss": 0.364, "step": 6676 }, { "epoch": 2.4, "learning_rate": 4.044448443404715e-06, "loss": 0.312, "step": 6677 }, { "epoch": 2.4, "learning_rate": 4.043548677343891e-06, "loss": 0.318, "step": 6678 }, { "epoch": 2.4, "learning_rate": 4.043548677343891e-06, "loss": 0.3236, "step": 6679 }, { "epoch": 2.4, "learning_rate": 4.042648911283067e-06, "loss": 0.3702, "step": 6680 }, { "epoch": 2.4, "learning_rate": 4.0417491452222425e-06, "loss": 0.3639, "step": 6681 }, { "epoch": 2.4, "learning_rate": 4.040849379161418e-06, "loss": 0.3172, "step": 6682 }, { "epoch": 2.4, "learning_rate": 4.039949613100594e-06, "loss": 0.3562, "step": 6683 }, { "epoch": 2.4, "learning_rate": 4.03904984703977e-06, "loss": 0.2884, "step": 6684 }, { "epoch": 2.4, "learning_rate": 4.0381500809789455e-06, "loss": 0.2854, "step": 6685 }, { "epoch": 2.4, "learning_rate": 4.037250314918121e-06, "loss": 0.29, "step": 6686 }, { "epoch": 2.4, "learning_rate": 4.036350548857297e-06, "loss": 0.3039, "step": 6687 }, { "epoch": 2.4, "learning_rate": 4.035450782796474e-06, "loss": 0.2803, "step": 6688 }, { "epoch": 2.4, "learning_rate": 4.034551016735649e-06, "loss": 0.3276, "step": 6689 }, { "epoch": 2.4, "learning_rate": 4.033651250674825e-06, "loss": 0.3224, "step": 6690 }, { "epoch": 2.4, "learning_rate": 4.032751484614001e-06, "loss": 0.385, "step": 6691 }, { "epoch": 2.4, "learning_rate": 4.031851718553177e-06, "loss": 0.3059, "step": 6692 }, { "epoch": 2.4, "learning_rate": 4.0309519524923525e-06, "loss": 0.3051, "step": 6693 }, { "epoch": 2.4, "learning_rate": 4.030052186431528e-06, "loss": 0.2788, "step": 6694 }, { "epoch": 2.4, "learning_rate": 4.029152420370704e-06, "loss": 0.3237, "step": 6695 }, { "epoch": 2.4, "learning_rate": 4.02825265430988e-06, "loss": 0.3633, "step": 6696 }, { "epoch": 2.4, "learning_rate": 4.027352888249056e-06, "loss": 0.3407, "step": 6697 }, { "epoch": 2.4, "learning_rate": 4.026453122188231e-06, "loss": 0.4038, "step": 6698 }, { "epoch": 2.4, "learning_rate": 4.025553356127407e-06, "loss": 0.3679, "step": 6699 }, { "epoch": 2.4, "learning_rate": 4.024653590066583e-06, "loss": 0.3305, "step": 6700 }, { "epoch": 2.41, "learning_rate": 4.023753824005759e-06, "loss": 0.3505, "step": 6701 }, { "epoch": 2.41, "learning_rate": 4.022854057944935e-06, "loss": 0.2785, "step": 6702 }, { "epoch": 2.41, "learning_rate": 4.02195429188411e-06, "loss": 0.3655, "step": 6703 }, { "epoch": 2.41, "learning_rate": 4.021054525823286e-06, "loss": 0.3987, "step": 6704 }, { "epoch": 2.41, "learning_rate": 4.020154759762462e-06, "loss": 0.2728, "step": 6705 }, { "epoch": 2.41, "learning_rate": 4.0192549937016375e-06, "loss": 0.3184, "step": 6706 }, { "epoch": 2.41, "learning_rate": 4.018355227640814e-06, "loss": 0.3523, "step": 6707 }, { "epoch": 2.41, "learning_rate": 4.017455461579989e-06, "loss": 0.3569, "step": 6708 }, { "epoch": 2.41, "learning_rate": 4.016555695519166e-06, "loss": 0.2495, "step": 6709 }, { "epoch": 2.41, "learning_rate": 4.0156559294583414e-06, "loss": 0.3504, "step": 6710 }, { "epoch": 2.41, "learning_rate": 4.014756163397517e-06, "loss": 0.3791, "step": 6711 }, { "epoch": 2.41, "learning_rate": 4.013856397336693e-06, "loss": 0.2323, "step": 6712 }, { "epoch": 2.41, "learning_rate": 4.012956631275869e-06, "loss": 0.3538, "step": 6713 }, { "epoch": 2.41, "learning_rate": 4.0120568652150445e-06, "loss": 0.3455, "step": 6714 }, { "epoch": 2.41, "learning_rate": 4.01115709915422e-06, "loss": 0.3008, "step": 6715 }, { "epoch": 2.41, "learning_rate": 4.010257333093396e-06, "loss": 0.286, "step": 6716 }, { "epoch": 2.41, "learning_rate": 4.009357567032572e-06, "loss": 0.3329, "step": 6717 }, { "epoch": 2.41, "learning_rate": 4.008457800971748e-06, "loss": 0.3053, "step": 6718 }, { "epoch": 2.41, "learning_rate": 4.007558034910923e-06, "loss": 0.3265, "step": 6719 }, { "epoch": 2.41, "learning_rate": 4.007558034910923e-06, "loss": 0.3537, "step": 6720 }, { "epoch": 2.41, "learning_rate": 4.006658268850099e-06, "loss": 0.3801, "step": 6721 }, { "epoch": 2.41, "learning_rate": 4.005758502789276e-06, "loss": 0.3476, "step": 6722 }, { "epoch": 2.41, "learning_rate": 4.004858736728451e-06, "loss": 0.3096, "step": 6723 }, { "epoch": 2.41, "learning_rate": 4.0039589706676264e-06, "loss": 0.3445, "step": 6724 }, { "epoch": 2.41, "learning_rate": 4.003059204606802e-06, "loss": 0.3587, "step": 6725 }, { "epoch": 2.41, "learning_rate": 4.002159438545978e-06, "loss": 0.3834, "step": 6726 }, { "epoch": 2.41, "learning_rate": 4.001259672485155e-06, "loss": 0.3587, "step": 6727 }, { "epoch": 2.41, "learning_rate": 4.0003599064243295e-06, "loss": 0.3719, "step": 6728 }, { "epoch": 2.42, "learning_rate": 3.999460140363506e-06, "loss": 0.3564, "step": 6729 }, { "epoch": 2.42, "learning_rate": 3.998560374302682e-06, "loss": 0.398, "step": 6730 }, { "epoch": 2.42, "learning_rate": 3.997660608241858e-06, "loss": 0.3327, "step": 6731 }, { "epoch": 2.42, "learning_rate": 3.9967608421810334e-06, "loss": 0.2915, "step": 6732 }, { "epoch": 2.42, "learning_rate": 3.995861076120209e-06, "loss": 0.2873, "step": 6733 }, { "epoch": 2.42, "learning_rate": 3.994961310059385e-06, "loss": 0.3553, "step": 6734 }, { "epoch": 2.42, "learning_rate": 3.994061543998561e-06, "loss": 0.3542, "step": 6735 }, { "epoch": 2.42, "learning_rate": 3.9931617779377365e-06, "loss": 0.3351, "step": 6736 }, { "epoch": 2.42, "learning_rate": 3.992262011876912e-06, "loss": 0.3996, "step": 6737 }, { "epoch": 2.42, "learning_rate": 3.991362245816088e-06, "loss": 0.2923, "step": 6738 }, { "epoch": 2.42, "learning_rate": 3.990462479755264e-06, "loss": 0.3586, "step": 6739 }, { "epoch": 2.42, "learning_rate": 3.98956271369444e-06, "loss": 0.3784, "step": 6740 }, { "epoch": 2.42, "learning_rate": 3.988662947633616e-06, "loss": 0.3568, "step": 6741 }, { "epoch": 2.42, "learning_rate": 3.987763181572791e-06, "loss": 0.3543, "step": 6742 }, { "epoch": 2.42, "learning_rate": 3.986863415511968e-06, "loss": 0.3214, "step": 6743 }, { "epoch": 2.42, "learning_rate": 3.985963649451143e-06, "loss": 0.3604, "step": 6744 }, { "epoch": 2.42, "learning_rate": 3.9850638833903184e-06, "loss": 0.37, "step": 6745 }, { "epoch": 2.42, "learning_rate": 3.984164117329495e-06, "loss": 0.3799, "step": 6746 }, { "epoch": 2.42, "learning_rate": 3.98326435126867e-06, "loss": 0.3541, "step": 6747 }, { "epoch": 2.42, "learning_rate": 3.982364585207847e-06, "loss": 0.3041, "step": 6748 }, { "epoch": 2.42, "learning_rate": 3.981464819147022e-06, "loss": 0.3145, "step": 6749 }, { "epoch": 2.42, "learning_rate": 3.980565053086198e-06, "loss": 0.3692, "step": 6750 }, { "epoch": 2.42, "learning_rate": 3.979665287025374e-06, "loss": 0.4168, "step": 6751 }, { "epoch": 2.42, "learning_rate": 3.97876552096455e-06, "loss": 0.3482, "step": 6752 }, { "epoch": 2.42, "learning_rate": 3.9778657549037254e-06, "loss": 0.2914, "step": 6753 }, { "epoch": 2.42, "learning_rate": 3.976965988842901e-06, "loss": 0.3208, "step": 6754 }, { "epoch": 2.42, "learning_rate": 3.976066222782077e-06, "loss": 0.2848, "step": 6755 }, { "epoch": 2.42, "learning_rate": 3.975166456721253e-06, "loss": 0.4409, "step": 6756 }, { "epoch": 2.43, "learning_rate": 3.975166456721253e-06, "loss": 0.5143, "step": 6757 }, { "epoch": 2.43, "learning_rate": 3.9742666906604285e-06, "loss": 0.3294, "step": 6758 }, { "epoch": 2.43, "learning_rate": 3.973366924599604e-06, "loss": 0.353, "step": 6759 }, { "epoch": 2.43, "learning_rate": 3.97246715853878e-06, "loss": 0.3207, "step": 6760 }, { "epoch": 2.43, "learning_rate": 3.971567392477957e-06, "loss": 0.3762, "step": 6761 }, { "epoch": 2.43, "learning_rate": 3.9706676264171316e-06, "loss": 0.3671, "step": 6762 }, { "epoch": 2.43, "learning_rate": 3.969767860356308e-06, "loss": 0.3857, "step": 6763 }, { "epoch": 2.43, "learning_rate": 3.968868094295483e-06, "loss": 0.3565, "step": 6764 }, { "epoch": 2.43, "learning_rate": 3.967968328234659e-06, "loss": 0.3814, "step": 6765 }, { "epoch": 2.43, "learning_rate": 3.9670685621738355e-06, "loss": 0.3412, "step": 6766 }, { "epoch": 2.43, "learning_rate": 3.96616879611301e-06, "loss": 0.3181, "step": 6767 }, { "epoch": 2.43, "learning_rate": 3.965269030052187e-06, "loss": 0.2395, "step": 6768 }, { "epoch": 2.43, "learning_rate": 3.964369263991363e-06, "loss": 0.2774, "step": 6769 }, { "epoch": 2.43, "learning_rate": 3.9634694979305386e-06, "loss": 0.3587, "step": 6770 }, { "epoch": 2.43, "learning_rate": 3.962569731869714e-06, "loss": 0.3382, "step": 6771 }, { "epoch": 2.43, "learning_rate": 3.96166996580889e-06, "loss": 0.3127, "step": 6772 }, { "epoch": 2.43, "learning_rate": 3.960770199748066e-06, "loss": 0.3141, "step": 6773 }, { "epoch": 2.43, "learning_rate": 3.959870433687242e-06, "loss": 0.3045, "step": 6774 }, { "epoch": 2.43, "learning_rate": 3.958970667626417e-06, "loss": 0.3008, "step": 6775 }, { "epoch": 2.43, "learning_rate": 3.958070901565593e-06, "loss": 0.3356, "step": 6776 }, { "epoch": 2.43, "learning_rate": 3.957171135504769e-06, "loss": 0.3592, "step": 6777 }, { "epoch": 2.43, "learning_rate": 3.956271369443945e-06, "loss": 0.3867, "step": 6778 }, { "epoch": 2.43, "learning_rate": 3.9553716033831205e-06, "loss": 0.324, "step": 6779 }, { "epoch": 2.43, "learning_rate": 3.954471837322297e-06, "loss": 0.384, "step": 6780 }, { "epoch": 2.43, "learning_rate": 3.953572071261472e-06, "loss": 0.3941, "step": 6781 }, { "epoch": 2.43, "learning_rate": 3.952672305200649e-06, "loss": 0.3358, "step": 6782 }, { "epoch": 2.43, "learning_rate": 3.9517725391398236e-06, "loss": 0.3746, "step": 6783 }, { "epoch": 2.44, "learning_rate": 3.950872773078999e-06, "loss": 0.3231, "step": 6784 }, { "epoch": 2.44, "learning_rate": 3.949973007018176e-06, "loss": 0.3456, "step": 6785 }, { "epoch": 2.44, "learning_rate": 3.949073240957351e-06, "loss": 0.365, "step": 6786 }, { "epoch": 2.44, "learning_rate": 3.9481734748965275e-06, "loss": 0.3367, "step": 6787 }, { "epoch": 2.44, "learning_rate": 3.947273708835703e-06, "loss": 0.3394, "step": 6788 }, { "epoch": 2.44, "learning_rate": 3.946373942774879e-06, "loss": 0.3296, "step": 6789 }, { "epoch": 2.44, "learning_rate": 3.945474176714055e-06, "loss": 0.3467, "step": 6790 }, { "epoch": 2.44, "learning_rate": 3.9445744106532306e-06, "loss": 0.3012, "step": 6791 }, { "epoch": 2.44, "learning_rate": 3.943674644592406e-06, "loss": 0.2766, "step": 6792 }, { "epoch": 2.44, "learning_rate": 3.942774878531582e-06, "loss": 0.3527, "step": 6793 }, { "epoch": 2.44, "learning_rate": 3.941875112470758e-06, "loss": 0.3464, "step": 6794 }, { "epoch": 2.44, "learning_rate": 3.940975346409934e-06, "loss": 0.3502, "step": 6795 }, { "epoch": 2.44, "learning_rate": 3.940075580349109e-06, "loss": 0.3237, "step": 6796 }, { "epoch": 2.44, "learning_rate": 3.939175814288285e-06, "loss": 0.3917, "step": 6797 }, { "epoch": 2.44, "learning_rate": 3.938276048227461e-06, "loss": 0.4252, "step": 6798 }, { "epoch": 2.44, "learning_rate": 3.9373762821666376e-06, "loss": 0.3344, "step": 6799 }, { "epoch": 2.44, "learning_rate": 3.9364765161058125e-06, "loss": 0.3343, "step": 6800 }, { "epoch": 2.44, "learning_rate": 3.935576750044989e-06, "loss": 0.2679, "step": 6801 }, { "epoch": 2.44, "learning_rate": 3.934676983984164e-06, "loss": 0.3329, "step": 6802 }, { "epoch": 2.44, "learning_rate": 3.933777217923341e-06, "loss": 0.2864, "step": 6803 }, { "epoch": 2.44, "learning_rate": 3.932877451862516e-06, "loss": 0.3099, "step": 6804 }, { "epoch": 2.44, "learning_rate": 3.931977685801691e-06, "loss": 0.3081, "step": 6805 }, { "epoch": 2.44, "learning_rate": 3.931077919740868e-06, "loss": 0.3496, "step": 6806 }, { "epoch": 2.44, "learning_rate": 3.930178153680044e-06, "loss": 0.3421, "step": 6807 }, { "epoch": 2.44, "learning_rate": 3.9292783876192195e-06, "loss": 0.3552, "step": 6808 }, { "epoch": 2.44, "learning_rate": 3.928378621558395e-06, "loss": 0.2921, "step": 6809 }, { "epoch": 2.44, "learning_rate": 3.927478855497571e-06, "loss": 0.3458, "step": 6810 }, { "epoch": 2.44, "learning_rate": 3.926579089436747e-06, "loss": 0.3142, "step": 6811 }, { "epoch": 2.45, "learning_rate": 3.9256793233759226e-06, "loss": 0.3276, "step": 6812 }, { "epoch": 2.45, "learning_rate": 3.924779557315098e-06, "loss": 0.3385, "step": 6813 }, { "epoch": 2.45, "learning_rate": 3.923879791254274e-06, "loss": 0.3145, "step": 6814 }, { "epoch": 2.45, "learning_rate": 3.92298002519345e-06, "loss": 0.3463, "step": 6815 }, { "epoch": 2.45, "learning_rate": 3.922080259132626e-06, "loss": 0.3875, "step": 6816 }, { "epoch": 2.45, "learning_rate": 3.921180493071801e-06, "loss": 0.3254, "step": 6817 }, { "epoch": 2.45, "learning_rate": 3.920280727010978e-06, "loss": 0.3527, "step": 6818 }, { "epoch": 2.45, "learning_rate": 3.919380960950153e-06, "loss": 0.3589, "step": 6819 }, { "epoch": 2.45, "learning_rate": 3.9184811948893296e-06, "loss": 0.3224, "step": 6820 }, { "epoch": 2.45, "learning_rate": 3.9175814288285045e-06, "loss": 0.3285, "step": 6821 }, { "epoch": 2.45, "learning_rate": 3.916681662767681e-06, "loss": 0.3088, "step": 6822 }, { "epoch": 2.45, "learning_rate": 3.915781896706857e-06, "loss": 0.3393, "step": 6823 }, { "epoch": 2.45, "learning_rate": 3.914882130646032e-06, "loss": 0.3372, "step": 6824 }, { "epoch": 2.45, "learning_rate": 3.913982364585208e-06, "loss": 0.3041, "step": 6825 }, { "epoch": 2.45, "learning_rate": 3.913082598524384e-06, "loss": 0.3278, "step": 6826 }, { "epoch": 2.45, "learning_rate": 3.91218283246356e-06, "loss": 0.3069, "step": 6827 }, { "epoch": 2.45, "learning_rate": 3.911283066402736e-06, "loss": 0.3723, "step": 6828 }, { "epoch": 2.45, "learning_rate": 3.9103833003419115e-06, "loss": 0.4673, "step": 6829 }, { "epoch": 2.45, "learning_rate": 3.909483534281087e-06, "loss": 0.3351, "step": 6830 }, { "epoch": 2.45, "learning_rate": 3.908583768220263e-06, "loss": 0.3352, "step": 6831 }, { "epoch": 2.45, "learning_rate": 3.907684002159439e-06, "loss": 0.3163, "step": 6832 }, { "epoch": 2.45, "learning_rate": 3.9067842360986145e-06, "loss": 0.375, "step": 6833 }, { "epoch": 2.45, "learning_rate": 3.90588447003779e-06, "loss": 0.4321, "step": 6834 }, { "epoch": 2.45, "learning_rate": 3.904984703976966e-06, "loss": 0.3776, "step": 6835 }, { "epoch": 2.45, "learning_rate": 3.904084937916142e-06, "loss": 0.2931, "step": 6836 }, { "epoch": 2.45, "learning_rate": 3.9031851718553185e-06, "loss": 0.3251, "step": 6837 }, { "epoch": 2.45, "learning_rate": 3.902285405794493e-06, "loss": 0.3429, "step": 6838 }, { "epoch": 2.45, "learning_rate": 3.90138563973367e-06, "loss": 0.3127, "step": 6839 }, { "epoch": 2.46, "learning_rate": 3.900485873672845e-06, "loss": 0.3949, "step": 6840 }, { "epoch": 2.46, "learning_rate": 3.8995861076120215e-06, "loss": 0.3202, "step": 6841 }, { "epoch": 2.46, "learning_rate": 3.898686341551197e-06, "loss": 0.3332, "step": 6842 }, { "epoch": 2.46, "learning_rate": 3.897786575490372e-06, "loss": 0.361, "step": 6843 }, { "epoch": 2.46, "learning_rate": 3.896886809429549e-06, "loss": 0.3079, "step": 6844 }, { "epoch": 2.46, "learning_rate": 3.895987043368725e-06, "loss": 0.3133, "step": 6845 }, { "epoch": 2.46, "learning_rate": 3.8950872773079e-06, "loss": 0.4301, "step": 6846 }, { "epoch": 2.46, "learning_rate": 3.894187511247076e-06, "loss": 0.3018, "step": 6847 }, { "epoch": 2.46, "learning_rate": 3.893287745186252e-06, "loss": 0.3098, "step": 6848 }, { "epoch": 2.46, "learning_rate": 3.892387979125428e-06, "loss": 0.3426, "step": 6849 }, { "epoch": 2.46, "learning_rate": 3.8914882130646035e-06, "loss": 0.382, "step": 6850 }, { "epoch": 2.46, "learning_rate": 3.890588447003779e-06, "loss": 0.3444, "step": 6851 }, { "epoch": 2.46, "learning_rate": 3.889688680942955e-06, "loss": 0.3315, "step": 6852 }, { "epoch": 2.46, "learning_rate": 3.888788914882131e-06, "loss": 0.3484, "step": 6853 }, { "epoch": 2.46, "learning_rate": 3.8878891488213065e-06, "loss": 0.3214, "step": 6854 }, { "epoch": 2.46, "learning_rate": 3.886989382760482e-06, "loss": 0.3124, "step": 6855 }, { "epoch": 2.46, "learning_rate": 3.886089616699659e-06, "loss": 0.3233, "step": 6856 }, { "epoch": 2.46, "learning_rate": 3.885189850638834e-06, "loss": 0.2722, "step": 6857 }, { "epoch": 2.46, "learning_rate": 3.8842900845780105e-06, "loss": 0.3275, "step": 6858 }, { "epoch": 2.46, "learning_rate": 3.883390318517185e-06, "loss": 0.3564, "step": 6859 }, { "epoch": 2.46, "learning_rate": 3.882490552456362e-06, "loss": 0.3124, "step": 6860 }, { "epoch": 2.46, "learning_rate": 3.881590786395538e-06, "loss": 0.4237, "step": 6861 }, { "epoch": 2.46, "learning_rate": 3.8806910203347135e-06, "loss": 0.2776, "step": 6862 }, { "epoch": 2.46, "learning_rate": 3.879791254273889e-06, "loss": 0.399, "step": 6863 }, { "epoch": 2.46, "learning_rate": 3.878891488213065e-06, "loss": 0.3088, "step": 6864 }, { "epoch": 2.46, "learning_rate": 3.877991722152241e-06, "loss": 0.4011, "step": 6865 }, { "epoch": 2.46, "learning_rate": 3.877091956091417e-06, "loss": 0.2967, "step": 6866 }, { "epoch": 2.46, "learning_rate": 3.876192190030592e-06, "loss": 0.3163, "step": 6867 }, { "epoch": 2.47, "learning_rate": 3.875292423969768e-06, "loss": 0.2571, "step": 6868 }, { "epoch": 2.47, "learning_rate": 3.874392657908944e-06, "loss": 0.343, "step": 6869 }, { "epoch": 2.47, "learning_rate": 3.87349289184812e-06, "loss": 0.3544, "step": 6870 }, { "epoch": 2.47, "learning_rate": 3.8725931257872955e-06, "loss": 0.3182, "step": 6871 }, { "epoch": 2.47, "learning_rate": 3.871693359726471e-06, "loss": 0.3972, "step": 6872 }, { "epoch": 2.47, "learning_rate": 3.870793593665647e-06, "loss": 0.3926, "step": 6873 }, { "epoch": 2.47, "learning_rate": 3.869893827604823e-06, "loss": 0.3158, "step": 6874 }, { "epoch": 2.47, "learning_rate": 3.868994061543999e-06, "loss": 0.3312, "step": 6875 }, { "epoch": 2.47, "learning_rate": 3.868094295483174e-06, "loss": 0.3283, "step": 6876 }, { "epoch": 2.47, "learning_rate": 3.867194529422351e-06, "loss": 0.3385, "step": 6877 }, { "epoch": 2.47, "learning_rate": 3.866294763361526e-06, "loss": 0.3278, "step": 6878 }, { "epoch": 2.47, "learning_rate": 3.8653949973007025e-06, "loss": 0.3168, "step": 6879 }, { "epoch": 2.47, "learning_rate": 3.864495231239878e-06, "loss": 0.3625, "step": 6880 }, { "epoch": 2.47, "learning_rate": 3.863595465179054e-06, "loss": 0.3505, "step": 6881 }, { "epoch": 2.47, "learning_rate": 3.86269569911823e-06, "loss": 0.3221, "step": 6882 }, { "epoch": 2.47, "learning_rate": 3.8617959330574055e-06, "loss": 0.3577, "step": 6883 }, { "epoch": 2.47, "learning_rate": 3.860896166996581e-06, "loss": 0.2909, "step": 6884 }, { "epoch": 2.47, "learning_rate": 3.859996400935757e-06, "loss": 0.3626, "step": 6885 }, { "epoch": 2.47, "learning_rate": 3.859096634874933e-06, "loss": 0.3646, "step": 6886 }, { "epoch": 2.47, "learning_rate": 3.858196868814109e-06, "loss": 0.358, "step": 6887 }, { "epoch": 2.47, "learning_rate": 3.857297102753284e-06, "loss": 0.3844, "step": 6888 }, { "epoch": 2.47, "learning_rate": 3.85639733669246e-06, "loss": 0.3221, "step": 6889 }, { "epoch": 2.47, "learning_rate": 3.855497570631636e-06, "loss": 0.3455, "step": 6890 }, { "epoch": 2.47, "learning_rate": 3.8545978045708125e-06, "loss": 0.3051, "step": 6891 }, { "epoch": 2.47, "learning_rate": 3.8536980385099874e-06, "loss": 0.3037, "step": 6892 }, { "epoch": 2.47, "learning_rate": 3.852798272449163e-06, "loss": 0.3474, "step": 6893 }, { "epoch": 2.47, "learning_rate": 3.85189850638834e-06, "loss": 0.3274, "step": 6894 }, { "epoch": 2.47, "learning_rate": 3.850998740327515e-06, "loss": 0.3666, "step": 6895 }, { "epoch": 2.48, "learning_rate": 3.850098974266691e-06, "loss": 0.3787, "step": 6896 }, { "epoch": 2.48, "learning_rate": 3.849199208205866e-06, "loss": 0.3606, "step": 6897 }, { "epoch": 2.48, "learning_rate": 3.848299442145043e-06, "loss": 0.358, "step": 6898 }, { "epoch": 2.48, "learning_rate": 3.847399676084219e-06, "loss": 0.3008, "step": 6899 }, { "epoch": 2.48, "learning_rate": 3.8464999100233944e-06, "loss": 0.3086, "step": 6900 }, { "epoch": 2.48, "learning_rate": 3.84560014396257e-06, "loss": 0.318, "step": 6901 }, { "epoch": 2.48, "learning_rate": 3.844700377901746e-06, "loss": 0.3289, "step": 6902 }, { "epoch": 2.48, "learning_rate": 3.843800611840922e-06, "loss": 0.3124, "step": 6903 }, { "epoch": 2.48, "learning_rate": 3.8429008457800975e-06, "loss": 0.311, "step": 6904 }, { "epoch": 2.48, "learning_rate": 3.842001079719273e-06, "loss": 0.3402, "step": 6905 }, { "epoch": 2.48, "learning_rate": 3.841101313658449e-06, "loss": 0.2922, "step": 6906 }, { "epoch": 2.48, "learning_rate": 3.840201547597625e-06, "loss": 0.3412, "step": 6907 }, { "epoch": 2.48, "learning_rate": 3.839301781536801e-06, "loss": 0.4022, "step": 6908 }, { "epoch": 2.48, "learning_rate": 3.838402015475976e-06, "loss": 0.3149, "step": 6909 }, { "epoch": 2.48, "learning_rate": 3.837502249415153e-06, "loss": 0.2784, "step": 6910 }, { "epoch": 2.48, "learning_rate": 3.836602483354328e-06, "loss": 0.3378, "step": 6911 }, { "epoch": 2.48, "learning_rate": 3.835702717293504e-06, "loss": 0.2348, "step": 6912 }, { "epoch": 2.48, "learning_rate": 3.83480295123268e-06, "loss": 0.3473, "step": 6913 }, { "epoch": 2.48, "learning_rate": 3.833903185171855e-06, "loss": 0.326, "step": 6914 }, { "epoch": 2.48, "learning_rate": 3.833003419111032e-06, "loss": 0.358, "step": 6915 }, { "epoch": 2.48, "learning_rate": 3.832103653050207e-06, "loss": 0.3691, "step": 6916 }, { "epoch": 2.48, "learning_rate": 3.831203886989383e-06, "loss": 0.3605, "step": 6917 }, { "epoch": 2.48, "learning_rate": 3.830304120928559e-06, "loss": 0.3775, "step": 6918 }, { "epoch": 2.48, "learning_rate": 3.829404354867735e-06, "loss": 0.3035, "step": 6919 }, { "epoch": 2.48, "learning_rate": 3.828504588806911e-06, "loss": 0.2794, "step": 6920 }, { "epoch": 2.48, "learning_rate": 3.8276048227460864e-06, "loss": 0.2804, "step": 6921 }, { "epoch": 2.48, "learning_rate": 3.826705056685262e-06, "loss": 0.4074, "step": 6922 }, { "epoch": 2.48, "learning_rate": 3.825805290624438e-06, "loss": 0.3489, "step": 6923 }, { "epoch": 2.49, "learning_rate": 3.824905524563614e-06, "loss": 0.3361, "step": 6924 }, { "epoch": 2.49, "learning_rate": 3.8240057585027895e-06, "loss": 0.3923, "step": 6925 }, { "epoch": 2.49, "learning_rate": 3.823105992441965e-06, "loss": 0.3923, "step": 6926 }, { "epoch": 2.49, "learning_rate": 3.822206226381141e-06, "loss": 0.3092, "step": 6927 }, { "epoch": 2.49, "learning_rate": 3.821306460320317e-06, "loss": 0.2767, "step": 6928 }, { "epoch": 2.49, "learning_rate": 3.8204066942594934e-06, "loss": 0.3783, "step": 6929 }, { "epoch": 2.49, "learning_rate": 3.819506928198668e-06, "loss": 0.3014, "step": 6930 }, { "epoch": 2.49, "learning_rate": 3.818607162137844e-06, "loss": 0.3666, "step": 6931 }, { "epoch": 2.49, "learning_rate": 3.81770739607702e-06, "loss": 0.3521, "step": 6932 }, { "epoch": 2.49, "learning_rate": 3.816807630016196e-06, "loss": 0.2835, "step": 6933 }, { "epoch": 2.49, "learning_rate": 3.815907863955372e-06, "loss": 0.3654, "step": 6934 }, { "epoch": 2.49, "learning_rate": 3.815008097894547e-06, "loss": 0.3225, "step": 6935 }, { "epoch": 2.49, "learning_rate": 3.8141083318337234e-06, "loss": 0.4027, "step": 6936 }, { "epoch": 2.49, "learning_rate": 3.8132085657728996e-06, "loss": 0.354, "step": 6937 }, { "epoch": 2.49, "learning_rate": 3.812308799712075e-06, "loss": 0.3018, "step": 6938 }, { "epoch": 2.49, "learning_rate": 3.811409033651251e-06, "loss": 0.2917, "step": 6939 }, { "epoch": 2.49, "learning_rate": 3.810509267590427e-06, "loss": 0.34, "step": 6940 }, { "epoch": 2.49, "learning_rate": 3.8096095015296027e-06, "loss": 0.3369, "step": 6941 }, { "epoch": 2.49, "learning_rate": 3.8087097354687784e-06, "loss": 0.324, "step": 6942 }, { "epoch": 2.49, "learning_rate": 3.807809969407954e-06, "loss": 0.3073, "step": 6943 }, { "epoch": 2.49, "learning_rate": 3.80691020334713e-06, "loss": 0.4195, "step": 6944 }, { "epoch": 2.49, "learning_rate": 3.806010437286306e-06, "loss": 0.3185, "step": 6945 }, { "epoch": 2.49, "learning_rate": 3.8051106712254815e-06, "loss": 0.3769, "step": 6946 }, { "epoch": 2.49, "learning_rate": 3.8042109051646577e-06, "loss": 0.2763, "step": 6947 }, { "epoch": 2.49, "learning_rate": 3.8033111391038335e-06, "loss": 0.2928, "step": 6948 }, { "epoch": 2.49, "learning_rate": 3.802411373043009e-06, "loss": 0.2938, "step": 6949 }, { "epoch": 2.49, "learning_rate": 3.801511606982185e-06, "loss": 0.3563, "step": 6950 }, { "epoch": 2.49, "learning_rate": 3.8006118409213603e-06, "loss": 0.2931, "step": 6951 }, { "epoch": 2.5, "learning_rate": 3.7997120748605365e-06, "loss": 0.3391, "step": 6952 }, { "epoch": 2.5, "learning_rate": 3.7988123087997127e-06, "loss": 0.305, "step": 6953 }, { "epoch": 2.5, "learning_rate": 3.797912542738888e-06, "loss": 0.3429, "step": 6954 }, { "epoch": 2.5, "learning_rate": 3.797012776678064e-06, "loss": 0.3581, "step": 6955 }, { "epoch": 2.5, "learning_rate": 3.79611301061724e-06, "loss": 0.3404, "step": 6956 }, { "epoch": 2.5, "learning_rate": 3.7952132445564154e-06, "loss": 0.3366, "step": 6957 }, { "epoch": 2.5, "learning_rate": 3.7943134784955916e-06, "loss": 0.3654, "step": 6958 }, { "epoch": 2.5, "learning_rate": 3.7934137124347673e-06, "loss": 0.322, "step": 6959 }, { "epoch": 2.5, "learning_rate": 3.792513946373943e-06, "loss": 0.3351, "step": 6960 }, { "epoch": 2.5, "learning_rate": 3.791614180313119e-06, "loss": 0.3248, "step": 6961 }, { "epoch": 2.5, "learning_rate": 3.7907144142522947e-06, "loss": 0.3348, "step": 6962 }, { "epoch": 2.5, "learning_rate": 3.7898146481914704e-06, "loss": 0.3796, "step": 6963 }, { "epoch": 2.5, "learning_rate": 3.7889148821306466e-06, "loss": 0.3373, "step": 6964 }, { "epoch": 2.5, "learning_rate": 3.788015116069822e-06, "loss": 0.2934, "step": 6965 }, { "epoch": 2.5, "learning_rate": 3.787115350008998e-06, "loss": 0.3189, "step": 6966 }, { "epoch": 2.5, "learning_rate": 3.786215583948174e-06, "loss": 0.3726, "step": 6967 }, { "epoch": 2.5, "learning_rate": 3.7853158178873493e-06, "loss": 0.3031, "step": 6968 }, { "epoch": 2.5, "learning_rate": 3.7844160518265255e-06, "loss": 0.3315, "step": 6969 }, { "epoch": 2.5, "learning_rate": 3.783516285765701e-06, "loss": 0.3639, "step": 6970 }, { "epoch": 2.5, "learning_rate": 3.782616519704877e-06, "loss": 0.3102, "step": 6971 }, { "epoch": 2.5, "learning_rate": 3.781716753644053e-06, "loss": 0.3769, "step": 6972 }, { "epoch": 2.5, "learning_rate": 3.7808169875832285e-06, "loss": 0.3637, "step": 6973 }, { "epoch": 2.5, "learning_rate": 3.7799172215224043e-06, "loss": 0.3388, "step": 6974 }, { "epoch": 2.5, "learning_rate": 3.7790174554615805e-06, "loss": 0.3185, "step": 6975 }, { "epoch": 2.5, "learning_rate": 3.778117689400756e-06, "loss": 0.3845, "step": 6976 }, { "epoch": 2.5, "learning_rate": 3.777217923339932e-06, "loss": 0.2991, "step": 6977 }, { "epoch": 2.5, "learning_rate": 3.776318157279108e-06, "loss": 0.3379, "step": 6978 }, { "epoch": 2.51, "learning_rate": 3.7754183912182836e-06, "loss": 0.3682, "step": 6979 }, { "epoch": 2.51, "learning_rate": 3.7745186251574593e-06, "loss": 0.3698, "step": 6980 }, { "epoch": 2.51, "learning_rate": 3.773618859096635e-06, "loss": 0.4045, "step": 6981 }, { "epoch": 2.51, "learning_rate": 3.772719093035811e-06, "loss": 0.3259, "step": 6982 }, { "epoch": 2.51, "learning_rate": 3.771819326974987e-06, "loss": 0.2891, "step": 6983 }, { "epoch": 2.51, "learning_rate": 3.7709195609141624e-06, "loss": 0.3428, "step": 6984 }, { "epoch": 2.51, "learning_rate": 3.7700197948533386e-06, "loss": 0.4087, "step": 6985 }, { "epoch": 2.51, "learning_rate": 3.7691200287925144e-06, "loss": 0.3659, "step": 6986 }, { "epoch": 2.51, "learning_rate": 3.76822026273169e-06, "loss": 0.3203, "step": 6987 }, { "epoch": 2.51, "learning_rate": 3.767320496670866e-06, "loss": 0.3779, "step": 6988 }, { "epoch": 2.51, "learning_rate": 3.7664207306100413e-06, "loss": 0.4334, "step": 6989 }, { "epoch": 2.51, "learning_rate": 3.7655209645492174e-06, "loss": 0.3664, "step": 6990 }, { "epoch": 2.51, "learning_rate": 3.7646211984883936e-06, "loss": 0.282, "step": 6991 }, { "epoch": 2.51, "learning_rate": 3.763721432427569e-06, "loss": 0.3153, "step": 6992 }, { "epoch": 2.51, "learning_rate": 3.7628216663667448e-06, "loss": 0.3232, "step": 6993 }, { "epoch": 2.51, "learning_rate": 3.761921900305921e-06, "loss": 0.3029, "step": 6994 }, { "epoch": 2.51, "learning_rate": 3.7610221342450963e-06, "loss": 0.311, "step": 6995 }, { "epoch": 2.51, "learning_rate": 3.7601223681842725e-06, "loss": 0.2887, "step": 6996 }, { "epoch": 2.51, "learning_rate": 3.7592226021234487e-06, "loss": 0.3769, "step": 6997 }, { "epoch": 2.51, "learning_rate": 3.758322836062624e-06, "loss": 0.2788, "step": 6998 }, { "epoch": 2.51, "learning_rate": 3.7574230700018e-06, "loss": 0.3383, "step": 6999 }, { "epoch": 2.51, "learning_rate": 3.7565233039409756e-06, "loss": 0.3529, "step": 7000 }, { "epoch": 2.51, "learning_rate": 3.7556235378801513e-06, "loss": 0.2673, "step": 7001 }, { "epoch": 2.51, "learning_rate": 3.7547237718193275e-06, "loss": 0.3822, "step": 7002 }, { "epoch": 2.51, "learning_rate": 3.753824005758503e-06, "loss": 0.2447, "step": 7003 }, { "epoch": 2.51, "learning_rate": 3.752924239697679e-06, "loss": 0.3116, "step": 7004 }, { "epoch": 2.51, "learning_rate": 3.752024473636855e-06, "loss": 0.347, "step": 7005 }, { "epoch": 2.51, "learning_rate": 3.7511247075760306e-06, "loss": 0.3344, "step": 7006 }, { "epoch": 2.52, "learning_rate": 3.7502249415152064e-06, "loss": 0.3525, "step": 7007 }, { "epoch": 2.52, "learning_rate": 3.7493251754543817e-06, "loss": 0.3341, "step": 7008 }, { "epoch": 2.52, "learning_rate": 3.748425409393558e-06, "loss": 0.3279, "step": 7009 }, { "epoch": 2.52, "learning_rate": 3.747525643332734e-06, "loss": 0.3703, "step": 7010 }, { "epoch": 2.52, "learning_rate": 3.7466258772719094e-06, "loss": 0.3856, "step": 7011 }, { "epoch": 2.52, "learning_rate": 3.7457261112110856e-06, "loss": 0.3085, "step": 7012 }, { "epoch": 2.52, "learning_rate": 3.7448263451502614e-06, "loss": 0.3525, "step": 7013 }, { "epoch": 2.52, "learning_rate": 3.7439265790894367e-06, "loss": 0.3429, "step": 7014 }, { "epoch": 2.52, "learning_rate": 3.743026813028613e-06, "loss": 0.3478, "step": 7015 }, { "epoch": 2.52, "learning_rate": 3.742127046967789e-06, "loss": 0.3326, "step": 7016 }, { "epoch": 2.52, "learning_rate": 3.7412272809069645e-06, "loss": 0.3582, "step": 7017 }, { "epoch": 2.52, "learning_rate": 3.7403275148461402e-06, "loss": 0.2824, "step": 7018 }, { "epoch": 2.52, "learning_rate": 3.739427748785316e-06, "loss": 0.3139, "step": 7019 }, { "epoch": 2.52, "learning_rate": 3.7385279827244918e-06, "loss": 0.3246, "step": 7020 }, { "epoch": 2.52, "learning_rate": 3.737628216663668e-06, "loss": 0.3301, "step": 7021 }, { "epoch": 2.52, "learning_rate": 3.7367284506028433e-06, "loss": 0.2902, "step": 7022 }, { "epoch": 2.52, "learning_rate": 3.7358286845420195e-06, "loss": 0.4447, "step": 7023 }, { "epoch": 2.52, "learning_rate": 3.7349289184811953e-06, "loss": 0.3377, "step": 7024 }, { "epoch": 2.52, "learning_rate": 3.734029152420371e-06, "loss": 0.3048, "step": 7025 }, { "epoch": 2.52, "learning_rate": 3.733129386359547e-06, "loss": 0.255, "step": 7026 }, { "epoch": 2.52, "learning_rate": 3.7322296202987226e-06, "loss": 0.3689, "step": 7027 }, { "epoch": 2.52, "learning_rate": 3.7313298542378984e-06, "loss": 0.3443, "step": 7028 }, { "epoch": 2.52, "learning_rate": 3.7304300881770745e-06, "loss": 0.2976, "step": 7029 }, { "epoch": 2.52, "learning_rate": 3.72953032211625e-06, "loss": 0.2951, "step": 7030 }, { "epoch": 2.52, "learning_rate": 3.728630556055426e-06, "loss": 0.3407, "step": 7031 }, { "epoch": 2.52, "learning_rate": 3.727730789994602e-06, "loss": 0.2606, "step": 7032 }, { "epoch": 2.52, "learning_rate": 3.726831023933777e-06, "loss": 0.3717, "step": 7033 }, { "epoch": 2.52, "learning_rate": 3.7259312578729534e-06, "loss": 0.3417, "step": 7034 }, { "epoch": 2.53, "learning_rate": 3.7250314918121296e-06, "loss": 0.3164, "step": 7035 }, { "epoch": 2.53, "learning_rate": 3.724131725751305e-06, "loss": 0.3211, "step": 7036 }, { "epoch": 2.53, "learning_rate": 3.7232319596904807e-06, "loss": 0.3588, "step": 7037 }, { "epoch": 2.53, "learning_rate": 3.7223321936296565e-06, "loss": 0.2991, "step": 7038 }, { "epoch": 2.53, "learning_rate": 3.7214324275688322e-06, "loss": 0.3157, "step": 7039 }, { "epoch": 2.53, "learning_rate": 3.7205326615080084e-06, "loss": 0.3117, "step": 7040 }, { "epoch": 2.53, "learning_rate": 3.7196328954471838e-06, "loss": 0.3534, "step": 7041 }, { "epoch": 2.53, "learning_rate": 3.71873312938636e-06, "loss": 0.2753, "step": 7042 }, { "epoch": 2.53, "learning_rate": 3.7178333633255357e-06, "loss": 0.3406, "step": 7043 }, { "epoch": 2.53, "learning_rate": 3.7169335972647115e-06, "loss": 0.3524, "step": 7044 }, { "epoch": 2.53, "learning_rate": 3.7160338312038873e-06, "loss": 0.3766, "step": 7045 }, { "epoch": 2.53, "learning_rate": 3.715134065143063e-06, "loss": 0.2745, "step": 7046 }, { "epoch": 2.53, "learning_rate": 3.714234299082239e-06, "loss": 0.3703, "step": 7047 }, { "epoch": 2.53, "learning_rate": 3.713334533021415e-06, "loss": 0.3603, "step": 7048 }, { "epoch": 2.53, "learning_rate": 3.7124347669605903e-06, "loss": 0.3206, "step": 7049 }, { "epoch": 2.53, "learning_rate": 3.7115350008997665e-06, "loss": 0.3527, "step": 7050 }, { "epoch": 2.53, "learning_rate": 3.7106352348389423e-06, "loss": 0.3151, "step": 7051 }, { "epoch": 2.53, "learning_rate": 3.7097354687781177e-06, "loss": 0.4136, "step": 7052 }, { "epoch": 2.53, "learning_rate": 3.708835702717294e-06, "loss": 0.371, "step": 7053 }, { "epoch": 2.53, "learning_rate": 3.70793593665647e-06, "loss": 0.3706, "step": 7054 }, { "epoch": 2.53, "learning_rate": 3.7070361705956454e-06, "loss": 0.3042, "step": 7055 }, { "epoch": 2.53, "learning_rate": 3.7061364045348216e-06, "loss": 0.3836, "step": 7056 }, { "epoch": 2.53, "learning_rate": 3.705236638473997e-06, "loss": 0.4699, "step": 7057 }, { "epoch": 2.53, "learning_rate": 3.7043368724131727e-06, "loss": 0.408, "step": 7058 }, { "epoch": 2.53, "learning_rate": 3.703437106352349e-06, "loss": 0.3286, "step": 7059 }, { "epoch": 2.53, "learning_rate": 3.7025373402915242e-06, "loss": 0.3079, "step": 7060 }, { "epoch": 2.53, "learning_rate": 3.7016375742307004e-06, "loss": 0.3692, "step": 7061 }, { "epoch": 2.53, "learning_rate": 3.700737808169876e-06, "loss": 0.3625, "step": 7062 }, { "epoch": 2.54, "learning_rate": 3.699838042109052e-06, "loss": 0.3222, "step": 7063 }, { "epoch": 2.54, "learning_rate": 3.6989382760482277e-06, "loss": 0.3923, "step": 7064 }, { "epoch": 2.54, "learning_rate": 3.6980385099874035e-06, "loss": 0.2647, "step": 7065 }, { "epoch": 2.54, "learning_rate": 3.6971387439265793e-06, "loss": 0.359, "step": 7066 }, { "epoch": 2.54, "learning_rate": 3.6962389778657555e-06, "loss": 0.3739, "step": 7067 }, { "epoch": 2.54, "learning_rate": 3.695339211804931e-06, "loss": 0.4398, "step": 7068 }, { "epoch": 2.54, "learning_rate": 3.694439445744107e-06, "loss": 0.3494, "step": 7069 }, { "epoch": 2.54, "learning_rate": 3.6935396796832828e-06, "loss": 0.3433, "step": 7070 }, { "epoch": 2.54, "learning_rate": 3.6926399136224585e-06, "loss": 0.3638, "step": 7071 }, { "epoch": 2.54, "learning_rate": 3.6917401475616343e-06, "loss": 0.3156, "step": 7072 }, { "epoch": 2.54, "learning_rate": 3.6908403815008105e-06, "loss": 0.2779, "step": 7073 }, { "epoch": 2.54, "learning_rate": 3.689940615439986e-06, "loss": 0.3899, "step": 7074 }, { "epoch": 2.54, "learning_rate": 3.689040849379162e-06, "loss": 0.258, "step": 7075 }, { "epoch": 2.54, "learning_rate": 3.6881410833183374e-06, "loss": 0.3306, "step": 7076 }, { "epoch": 2.54, "learning_rate": 3.687241317257513e-06, "loss": 0.3018, "step": 7077 }, { "epoch": 2.54, "learning_rate": 3.6863415511966893e-06, "loss": 0.3543, "step": 7078 }, { "epoch": 2.54, "learning_rate": 3.6854417851358647e-06, "loss": 0.2996, "step": 7079 }, { "epoch": 2.54, "learning_rate": 3.684542019075041e-06, "loss": 0.3885, "step": 7080 }, { "epoch": 2.54, "learning_rate": 3.6836422530142166e-06, "loss": 0.3773, "step": 7081 }, { "epoch": 2.54, "learning_rate": 3.6827424869533924e-06, "loss": 0.3089, "step": 7082 }, { "epoch": 2.54, "learning_rate": 3.681842720892568e-06, "loss": 0.396, "step": 7083 }, { "epoch": 2.54, "learning_rate": 3.680942954831744e-06, "loss": 0.352, "step": 7084 }, { "epoch": 2.54, "learning_rate": 3.6800431887709197e-06, "loss": 0.2962, "step": 7085 }, { "epoch": 2.54, "learning_rate": 3.679143422710096e-06, "loss": 0.3602, "step": 7086 }, { "epoch": 2.54, "learning_rate": 3.6782436566492713e-06, "loss": 0.3188, "step": 7087 }, { "epoch": 2.54, "learning_rate": 3.6773438905884474e-06, "loss": 0.3204, "step": 7088 }, { "epoch": 2.54, "learning_rate": 3.6764441245276232e-06, "loss": 0.2632, "step": 7089 }, { "epoch": 2.54, "learning_rate": 3.675544358466799e-06, "loss": 0.3504, "step": 7090 }, { "epoch": 2.55, "learning_rate": 3.6746445924059748e-06, "loss": 0.3546, "step": 7091 }, { "epoch": 2.55, "learning_rate": 3.673744826345151e-06, "loss": 0.2797, "step": 7092 }, { "epoch": 2.55, "learning_rate": 3.6728450602843263e-06, "loss": 0.328, "step": 7093 }, { "epoch": 2.55, "learning_rate": 3.6719452942235025e-06, "loss": 0.3669, "step": 7094 }, { "epoch": 2.55, "learning_rate": 3.671045528162678e-06, "loss": 0.3561, "step": 7095 }, { "epoch": 2.55, "learning_rate": 3.6701457621018536e-06, "loss": 0.339, "step": 7096 }, { "epoch": 2.55, "learning_rate": 3.66924599604103e-06, "loss": 0.3154, "step": 7097 }, { "epoch": 2.55, "learning_rate": 3.668346229980205e-06, "loss": 0.34, "step": 7098 }, { "epoch": 2.55, "learning_rate": 3.6674464639193813e-06, "loss": 0.3444, "step": 7099 }, { "epoch": 2.55, "learning_rate": 3.6665466978585575e-06, "loss": 0.3828, "step": 7100 }, { "epoch": 2.55, "learning_rate": 3.665646931797733e-06, "loss": 0.3505, "step": 7101 }, { "epoch": 2.55, "learning_rate": 3.6647471657369086e-06, "loss": 0.3218, "step": 7102 }, { "epoch": 2.55, "learning_rate": 3.6638473996760844e-06, "loss": 0.3349, "step": 7103 }, { "epoch": 2.55, "learning_rate": 3.66294763361526e-06, "loss": 0.3278, "step": 7104 }, { "epoch": 2.55, "learning_rate": 3.6620478675544364e-06, "loss": 0.3229, "step": 7105 }, { "epoch": 2.55, "learning_rate": 3.6611481014936117e-06, "loss": 0.3848, "step": 7106 }, { "epoch": 2.55, "learning_rate": 3.660248335432788e-06, "loss": 0.4279, "step": 7107 }, { "epoch": 2.55, "learning_rate": 3.6593485693719637e-06, "loss": 0.3423, "step": 7108 }, { "epoch": 2.55, "learning_rate": 3.6584488033111394e-06, "loss": 0.3159, "step": 7109 }, { "epoch": 2.55, "learning_rate": 3.657549037250315e-06, "loss": 0.3002, "step": 7110 }, { "epoch": 2.55, "learning_rate": 3.6566492711894906e-06, "loss": 0.2741, "step": 7111 }, { "epoch": 2.55, "learning_rate": 3.6557495051286667e-06, "loss": 0.3423, "step": 7112 }, { "epoch": 2.55, "learning_rate": 3.654849739067843e-06, "loss": 0.4219, "step": 7113 }, { "epoch": 2.55, "learning_rate": 3.6539499730070183e-06, "loss": 0.3553, "step": 7114 }, { "epoch": 2.55, "learning_rate": 3.6530502069461945e-06, "loss": 0.4294, "step": 7115 }, { "epoch": 2.55, "learning_rate": 3.6521504408853702e-06, "loss": 0.3588, "step": 7116 }, { "epoch": 2.55, "learning_rate": 3.6512506748245456e-06, "loss": 0.3456, "step": 7117 }, { "epoch": 2.55, "learning_rate": 3.6503509087637218e-06, "loss": 0.3378, "step": 7118 }, { "epoch": 2.56, "learning_rate": 3.649451142702898e-06, "loss": 0.3035, "step": 7119 }, { "epoch": 2.56, "learning_rate": 3.6485513766420733e-06, "loss": 0.356, "step": 7120 }, { "epoch": 2.56, "learning_rate": 3.647651610581249e-06, "loss": 0.3865, "step": 7121 }, { "epoch": 2.56, "learning_rate": 3.646751844520425e-06, "loss": 0.3704, "step": 7122 }, { "epoch": 2.56, "learning_rate": 3.6458520784596006e-06, "loss": 0.3871, "step": 7123 }, { "epoch": 2.56, "learning_rate": 3.644952312398777e-06, "loss": 0.3379, "step": 7124 }, { "epoch": 2.56, "learning_rate": 3.644052546337952e-06, "loss": 0.3981, "step": 7125 }, { "epoch": 2.56, "learning_rate": 3.6431527802771284e-06, "loss": 0.2677, "step": 7126 }, { "epoch": 2.56, "learning_rate": 3.642253014216304e-06, "loss": 0.3485, "step": 7127 }, { "epoch": 2.56, "learning_rate": 3.64135324815548e-06, "loss": 0.3468, "step": 7128 }, { "epoch": 2.56, "learning_rate": 3.6404534820946557e-06, "loss": 0.2954, "step": 7129 }, { "epoch": 2.56, "learning_rate": 3.6395537160338314e-06, "loss": 0.3089, "step": 7130 }, { "epoch": 2.56, "learning_rate": 3.638653949973007e-06, "loss": 0.3095, "step": 7131 }, { "epoch": 2.56, "learning_rate": 3.6377541839121834e-06, "loss": 0.3942, "step": 7132 }, { "epoch": 2.56, "learning_rate": 3.6368544178513587e-06, "loss": 0.4102, "step": 7133 }, { "epoch": 2.56, "learning_rate": 3.635954651790535e-06, "loss": 0.3348, "step": 7134 }, { "epoch": 2.56, "learning_rate": 3.6350548857297107e-06, "loss": 0.3043, "step": 7135 }, { "epoch": 2.56, "learning_rate": 3.634155119668886e-06, "loss": 0.3141, "step": 7136 }, { "epoch": 2.56, "learning_rate": 3.6332553536080622e-06, "loss": 0.3093, "step": 7137 }, { "epoch": 2.56, "learning_rate": 3.6323555875472384e-06, "loss": 0.3573, "step": 7138 }, { "epoch": 2.56, "learning_rate": 3.6314558214864138e-06, "loss": 0.3691, "step": 7139 }, { "epoch": 2.56, "learning_rate": 3.6305560554255895e-06, "loss": 0.3188, "step": 7140 }, { "epoch": 2.56, "learning_rate": 3.6296562893647653e-06, "loss": 0.2749, "step": 7141 }, { "epoch": 2.56, "learning_rate": 3.628756523303941e-06, "loss": 0.3158, "step": 7142 }, { "epoch": 2.56, "learning_rate": 3.6278567572431173e-06, "loss": 0.3434, "step": 7143 }, { "epoch": 2.56, "learning_rate": 3.6269569911822926e-06, "loss": 0.366, "step": 7144 }, { "epoch": 2.56, "learning_rate": 3.626057225121469e-06, "loss": 0.4173, "step": 7145 }, { "epoch": 2.56, "learning_rate": 3.6251574590606446e-06, "loss": 0.3168, "step": 7146 }, { "epoch": 2.57, "learning_rate": 3.6242576929998203e-06, "loss": 0.3073, "step": 7147 }, { "epoch": 2.57, "learning_rate": 3.623357926938996e-06, "loss": 0.3518, "step": 7148 }, { "epoch": 2.57, "learning_rate": 3.622458160878172e-06, "loss": 0.3593, "step": 7149 }, { "epoch": 2.57, "learning_rate": 3.6215583948173477e-06, "loss": 0.3459, "step": 7150 }, { "epoch": 2.57, "learning_rate": 3.620658628756524e-06, "loss": 0.3628, "step": 7151 }, { "epoch": 2.57, "learning_rate": 3.619758862695699e-06, "loss": 0.338, "step": 7152 }, { "epoch": 2.57, "learning_rate": 3.6188590966348754e-06, "loss": 0.2952, "step": 7153 }, { "epoch": 2.57, "learning_rate": 3.617959330574051e-06, "loss": 0.3701, "step": 7154 }, { "epoch": 2.57, "learning_rate": 3.6170595645132265e-06, "loss": 0.3868, "step": 7155 }, { "epoch": 2.57, "learning_rate": 3.6161597984524027e-06, "loss": 0.3386, "step": 7156 }, { "epoch": 2.57, "learning_rate": 3.615260032391579e-06, "loss": 0.3366, "step": 7157 }, { "epoch": 2.57, "learning_rate": 3.6143602663307542e-06, "loss": 0.3095, "step": 7158 }, { "epoch": 2.57, "learning_rate": 3.6134605002699304e-06, "loss": 0.2998, "step": 7159 }, { "epoch": 2.57, "learning_rate": 3.6125607342091058e-06, "loss": 0.3377, "step": 7160 }, { "epoch": 2.57, "learning_rate": 3.6116609681482815e-06, "loss": 0.3449, "step": 7161 }, { "epoch": 2.57, "learning_rate": 3.6107612020874577e-06, "loss": 0.3519, "step": 7162 }, { "epoch": 2.57, "learning_rate": 3.609861436026633e-06, "loss": 0.3693, "step": 7163 }, { "epoch": 2.57, "learning_rate": 3.6089616699658093e-06, "loss": 0.314, "step": 7164 }, { "epoch": 2.57, "learning_rate": 3.608061903904985e-06, "loss": 0.2441, "step": 7165 }, { "epoch": 2.57, "learning_rate": 3.607162137844161e-06, "loss": 0.3039, "step": 7166 }, { "epoch": 2.57, "learning_rate": 3.6062623717833366e-06, "loss": 0.3295, "step": 7167 }, { "epoch": 2.57, "learning_rate": 3.6053626057225123e-06, "loss": 0.3374, "step": 7168 }, { "epoch": 2.57, "learning_rate": 3.604462839661688e-06, "loss": 0.3102, "step": 7169 }, { "epoch": 2.57, "learning_rate": 3.6035630736008643e-06, "loss": 0.3841, "step": 7170 }, { "epoch": 2.57, "learning_rate": 3.6026633075400396e-06, "loss": 0.3921, "step": 7171 }, { "epoch": 2.57, "learning_rate": 3.601763541479216e-06, "loss": 0.3681, "step": 7172 }, { "epoch": 2.57, "learning_rate": 3.6008637754183916e-06, "loss": 0.3238, "step": 7173 }, { "epoch": 2.58, "learning_rate": 3.5999640093575674e-06, "loss": 0.3214, "step": 7174 }, { "epoch": 2.58, "learning_rate": 3.599064243296743e-06, "loss": 0.3085, "step": 7175 }, { "epoch": 2.58, "learning_rate": 3.5981644772359193e-06, "loss": 0.3971, "step": 7176 }, { "epoch": 2.58, "learning_rate": 3.5972647111750947e-06, "loss": 0.3718, "step": 7177 }, { "epoch": 2.58, "learning_rate": 3.596364945114271e-06, "loss": 0.2996, "step": 7178 }, { "epoch": 2.58, "learning_rate": 3.5954651790534462e-06, "loss": 0.3187, "step": 7179 }, { "epoch": 2.58, "learning_rate": 3.594565412992622e-06, "loss": 0.343, "step": 7180 }, { "epoch": 2.58, "learning_rate": 3.593665646931798e-06, "loss": 0.3286, "step": 7181 }, { "epoch": 2.58, "learning_rate": 3.5927658808709735e-06, "loss": 0.3605, "step": 7182 }, { "epoch": 2.58, "learning_rate": 3.5918661148101497e-06, "loss": 0.3785, "step": 7183 }, { "epoch": 2.58, "learning_rate": 3.590966348749326e-06, "loss": 0.365, "step": 7184 }, { "epoch": 2.58, "learning_rate": 3.5900665826885013e-06, "loss": 0.3755, "step": 7185 }, { "epoch": 2.58, "learning_rate": 3.589166816627677e-06, "loss": 0.3132, "step": 7186 }, { "epoch": 2.58, "learning_rate": 3.588267050566853e-06, "loss": 0.3799, "step": 7187 }, { "epoch": 2.58, "learning_rate": 3.5873672845060286e-06, "loss": 0.3369, "step": 7188 }, { "epoch": 2.58, "learning_rate": 3.5864675184452048e-06, "loss": 0.2821, "step": 7189 }, { "epoch": 2.58, "learning_rate": 3.58556775238438e-06, "loss": 0.3512, "step": 7190 }, { "epoch": 2.58, "learning_rate": 3.5846679863235563e-06, "loss": 0.3452, "step": 7191 }, { "epoch": 2.58, "learning_rate": 3.583768220262732e-06, "loss": 0.3647, "step": 7192 }, { "epoch": 2.58, "learning_rate": 3.582868454201908e-06, "loss": 0.3497, "step": 7193 }, { "epoch": 2.58, "learning_rate": 3.5819686881410836e-06, "loss": 0.3119, "step": 7194 }, { "epoch": 2.58, "learning_rate": 3.58106892208026e-06, "loss": 0.3418, "step": 7195 }, { "epoch": 2.58, "learning_rate": 3.580169156019435e-06, "loss": 0.3771, "step": 7196 }, { "epoch": 2.58, "learning_rate": 3.5792693899586113e-06, "loss": 0.3473, "step": 7197 }, { "epoch": 2.58, "learning_rate": 3.5783696238977867e-06, "loss": 0.3416, "step": 7198 }, { "epoch": 2.58, "learning_rate": 3.577469857836963e-06, "loss": 0.3122, "step": 7199 }, { "epoch": 2.58, "learning_rate": 3.5765700917761386e-06, "loss": 0.3662, "step": 7200 }, { "epoch": 2.58, "learning_rate": 3.575670325715314e-06, "loss": 0.35, "step": 7201 }, { "epoch": 2.59, "learning_rate": 3.57477055965449e-06, "loss": 0.3521, "step": 7202 }, { "epoch": 2.59, "learning_rate": 3.5738707935936664e-06, "loss": 0.266, "step": 7203 }, { "epoch": 2.59, "learning_rate": 3.5729710275328417e-06, "loss": 0.3207, "step": 7204 }, { "epoch": 2.59, "learning_rate": 3.5720712614720175e-06, "loss": 0.2972, "step": 7205 }, { "epoch": 2.59, "learning_rate": 3.5711714954111932e-06, "loss": 0.2245, "step": 7206 }, { "epoch": 2.59, "learning_rate": 3.570271729350369e-06, "loss": 0.3368, "step": 7207 }, { "epoch": 2.59, "learning_rate": 3.569371963289545e-06, "loss": 0.364, "step": 7208 }, { "epoch": 2.59, "learning_rate": 3.5684721972287206e-06, "loss": 0.3133, "step": 7209 }, { "epoch": 2.59, "learning_rate": 3.5675724311678967e-06, "loss": 0.3, "step": 7210 }, { "epoch": 2.59, "learning_rate": 3.5666726651070725e-06, "loss": 0.3542, "step": 7211 }, { "epoch": 2.59, "learning_rate": 3.5657728990462483e-06, "loss": 0.3509, "step": 7212 }, { "epoch": 2.59, "learning_rate": 3.564873132985424e-06, "loss": 0.3123, "step": 7213 }, { "epoch": 2.59, "learning_rate": 3.5639733669246002e-06, "loss": 0.3319, "step": 7214 }, { "epoch": 2.59, "learning_rate": 3.5630736008637756e-06, "loss": 0.3938, "step": 7215 }, { "epoch": 2.59, "learning_rate": 3.5621738348029518e-06, "loss": 0.3009, "step": 7216 }, { "epoch": 2.59, "learning_rate": 3.561274068742127e-06, "loss": 0.3459, "step": 7217 }, { "epoch": 2.59, "learning_rate": 3.5603743026813033e-06, "loss": 0.2728, "step": 7218 }, { "epoch": 2.59, "learning_rate": 3.559474536620479e-06, "loss": 0.359, "step": 7219 }, { "epoch": 2.59, "learning_rate": 3.5585747705596544e-06, "loss": 0.2973, "step": 7220 }, { "epoch": 2.59, "learning_rate": 3.5576750044988306e-06, "loss": 0.3149, "step": 7221 }, { "epoch": 2.59, "learning_rate": 3.556775238438007e-06, "loss": 0.3078, "step": 7222 }, { "epoch": 2.59, "learning_rate": 3.555875472377182e-06, "loss": 0.3857, "step": 7223 }, { "epoch": 2.59, "learning_rate": 3.554975706316358e-06, "loss": 0.3049, "step": 7224 }, { "epoch": 2.59, "learning_rate": 3.5540759402555337e-06, "loss": 0.3673, "step": 7225 }, { "epoch": 2.59, "learning_rate": 3.5531761741947095e-06, "loss": 0.3412, "step": 7226 }, { "epoch": 2.59, "learning_rate": 3.5522764081338857e-06, "loss": 0.3067, "step": 7227 }, { "epoch": 2.59, "learning_rate": 3.551376642073061e-06, "loss": 0.3094, "step": 7228 }, { "epoch": 2.59, "learning_rate": 3.550476876012237e-06, "loss": 0.2844, "step": 7229 }, { "epoch": 2.6, "learning_rate": 3.549577109951413e-06, "loss": 0.3166, "step": 7230 }, { "epoch": 2.6, "learning_rate": 3.5486773438905887e-06, "loss": 0.3957, "step": 7231 }, { "epoch": 2.6, "learning_rate": 3.5477775778297645e-06, "loss": 0.271, "step": 7232 }, { "epoch": 2.6, "learning_rate": 3.5468778117689407e-06, "loss": 0.318, "step": 7233 }, { "epoch": 2.6, "learning_rate": 3.545978045708116e-06, "loss": 0.3677, "step": 7234 }, { "epoch": 2.6, "learning_rate": 3.5450782796472922e-06, "loss": 0.3457, "step": 7235 }, { "epoch": 2.6, "learning_rate": 3.5441785135864676e-06, "loss": 0.3964, "step": 7236 }, { "epoch": 2.6, "learning_rate": 3.5432787475256438e-06, "loss": 0.3442, "step": 7237 }, { "epoch": 2.6, "learning_rate": 3.5423789814648195e-06, "loss": 0.3753, "step": 7238 }, { "epoch": 2.6, "learning_rate": 3.541479215403995e-06, "loss": 0.3742, "step": 7239 }, { "epoch": 2.6, "learning_rate": 3.540579449343171e-06, "loss": 0.272, "step": 7240 }, { "epoch": 2.6, "learning_rate": 3.5396796832823473e-06, "loss": 0.3149, "step": 7241 }, { "epoch": 2.6, "learning_rate": 3.5387799172215226e-06, "loss": 0.2933, "step": 7242 }, { "epoch": 2.6, "learning_rate": 3.537880151160699e-06, "loss": 0.3487, "step": 7243 }, { "epoch": 2.6, "learning_rate": 3.536980385099874e-06, "loss": 0.3542, "step": 7244 }, { "epoch": 2.6, "learning_rate": 3.53608061903905e-06, "loss": 0.3325, "step": 7245 }, { "epoch": 2.6, "learning_rate": 3.535180852978226e-06, "loss": 0.374, "step": 7246 }, { "epoch": 2.6, "learning_rate": 3.5342810869174015e-06, "loss": 0.3105, "step": 7247 }, { "epoch": 2.6, "learning_rate": 3.5333813208565777e-06, "loss": 0.3327, "step": 7248 }, { "epoch": 2.6, "learning_rate": 3.5324815547957534e-06, "loss": 0.3311, "step": 7249 }, { "epoch": 2.6, "learning_rate": 3.531581788734929e-06, "loss": 0.3778, "step": 7250 }, { "epoch": 2.6, "learning_rate": 3.530682022674105e-06, "loss": 0.3216, "step": 7251 }, { "epoch": 2.6, "learning_rate": 3.529782256613281e-06, "loss": 0.3476, "step": 7252 }, { "epoch": 2.6, "learning_rate": 3.5288824905524565e-06, "loss": 0.3906, "step": 7253 }, { "epoch": 2.6, "learning_rate": 3.5279827244916327e-06, "loss": 0.2952, "step": 7254 }, { "epoch": 2.6, "learning_rate": 3.527082958430808e-06, "loss": 0.3022, "step": 7255 }, { "epoch": 2.6, "learning_rate": 3.5261831923699842e-06, "loss": 0.3072, "step": 7256 }, { "epoch": 2.6, "learning_rate": 3.52528342630916e-06, "loss": 0.3185, "step": 7257 }, { "epoch": 2.61, "learning_rate": 3.5243836602483358e-06, "loss": 0.3364, "step": 7258 }, { "epoch": 2.61, "learning_rate": 3.5234838941875115e-06, "loss": 0.3989, "step": 7259 }, { "epoch": 2.61, "learning_rate": 3.5225841281266877e-06, "loss": 0.3075, "step": 7260 }, { "epoch": 2.61, "learning_rate": 3.521684362065863e-06, "loss": 0.3116, "step": 7261 }, { "epoch": 2.61, "learning_rate": 3.5207845960050393e-06, "loss": 0.3044, "step": 7262 }, { "epoch": 2.61, "learning_rate": 3.5198848299442146e-06, "loss": 0.2909, "step": 7263 }, { "epoch": 2.61, "learning_rate": 3.5189850638833904e-06, "loss": 0.3379, "step": 7264 }, { "epoch": 2.61, "learning_rate": 3.5180852978225666e-06, "loss": 0.3257, "step": 7265 }, { "epoch": 2.61, "learning_rate": 3.517185531761742e-06, "loss": 0.2613, "step": 7266 }, { "epoch": 2.61, "learning_rate": 3.516285765700918e-06, "loss": 0.3526, "step": 7267 }, { "epoch": 2.61, "learning_rate": 3.515385999640094e-06, "loss": 0.3453, "step": 7268 }, { "epoch": 2.61, "learning_rate": 3.5144862335792696e-06, "loss": 0.3538, "step": 7269 }, { "epoch": 2.61, "learning_rate": 3.5135864675184454e-06, "loss": 0.332, "step": 7270 }, { "epoch": 2.61, "learning_rate": 3.512686701457621e-06, "loss": 0.3794, "step": 7271 }, { "epoch": 2.61, "learning_rate": 3.511786935396797e-06, "loss": 0.2844, "step": 7272 }, { "epoch": 2.61, "learning_rate": 3.510887169335973e-06, "loss": 0.3205, "step": 7273 }, { "epoch": 2.61, "learning_rate": 3.5099874032751485e-06, "loss": 0.3117, "step": 7274 }, { "epoch": 2.61, "learning_rate": 3.5090876372143247e-06, "loss": 0.3656, "step": 7275 }, { "epoch": 2.61, "learning_rate": 3.5081878711535005e-06, "loss": 0.3328, "step": 7276 }, { "epoch": 2.61, "learning_rate": 3.5072881050926762e-06, "loss": 0.3439, "step": 7277 }, { "epoch": 2.61, "learning_rate": 3.506388339031852e-06, "loss": 0.3649, "step": 7278 }, { "epoch": 2.61, "learning_rate": 3.505488572971028e-06, "loss": 0.3078, "step": 7279 }, { "epoch": 2.61, "learning_rate": 3.5045888069102035e-06, "loss": 0.37, "step": 7280 }, { "epoch": 2.61, "learning_rate": 3.5036890408493797e-06, "loss": 0.3458, "step": 7281 }, { "epoch": 2.61, "learning_rate": 3.502789274788555e-06, "loss": 0.36, "step": 7282 }, { "epoch": 2.61, "learning_rate": 3.501889508727731e-06, "loss": 0.3408, "step": 7283 }, { "epoch": 2.61, "learning_rate": 3.500989742666907e-06, "loss": 0.3113, "step": 7284 }, { "epoch": 2.61, "learning_rate": 3.5000899766060824e-06, "loss": 0.3025, "step": 7285 }, { "epoch": 2.62, "learning_rate": 3.4991902105452586e-06, "loss": 0.3032, "step": 7286 }, { "epoch": 2.62, "learning_rate": 3.4982904444844348e-06, "loss": 0.2761, "step": 7287 }, { "epoch": 2.62, "learning_rate": 3.49739067842361e-06, "loss": 0.3847, "step": 7288 }, { "epoch": 2.62, "learning_rate": 3.496490912362786e-06, "loss": 0.3815, "step": 7289 }, { "epoch": 2.62, "learning_rate": 3.4955911463019616e-06, "loss": 0.3627, "step": 7290 }, { "epoch": 2.62, "learning_rate": 3.4946913802411374e-06, "loss": 0.3736, "step": 7291 }, { "epoch": 2.62, "learning_rate": 3.4937916141803136e-06, "loss": 0.3918, "step": 7292 }, { "epoch": 2.62, "learning_rate": 3.492891848119489e-06, "loss": 0.327, "step": 7293 }, { "epoch": 2.62, "learning_rate": 3.491992082058665e-06, "loss": 0.316, "step": 7294 }, { "epoch": 2.62, "learning_rate": 3.491092315997841e-06, "loss": 0.3103, "step": 7295 }, { "epoch": 2.62, "learning_rate": 3.4901925499370167e-06, "loss": 0.3621, "step": 7296 }, { "epoch": 2.62, "learning_rate": 3.4892927838761924e-06, "loss": 0.3509, "step": 7297 }, { "epoch": 2.62, "learning_rate": 3.4883930178153686e-06, "loss": 0.2873, "step": 7298 }, { "epoch": 2.62, "learning_rate": 3.487493251754544e-06, "loss": 0.3681, "step": 7299 }, { "epoch": 2.62, "learning_rate": 3.48659348569372e-06, "loss": 0.3343, "step": 7300 }, { "epoch": 2.62, "learning_rate": 3.4856937196328955e-06, "loss": 0.3569, "step": 7301 }, { "epoch": 2.62, "learning_rate": 3.4847939535720717e-06, "loss": 0.3952, "step": 7302 }, { "epoch": 2.62, "learning_rate": 3.4838941875112475e-06, "loss": 0.3563, "step": 7303 }, { "epoch": 2.62, "learning_rate": 3.482994421450423e-06, "loss": 0.3549, "step": 7304 }, { "epoch": 2.62, "learning_rate": 3.482094655389599e-06, "loss": 0.3471, "step": 7305 }, { "epoch": 2.62, "learning_rate": 3.481194889328775e-06, "loss": 0.3248, "step": 7306 }, { "epoch": 2.62, "learning_rate": 3.4802951232679506e-06, "loss": 0.2757, "step": 7307 }, { "epoch": 2.62, "learning_rate": 3.4793953572071263e-06, "loss": 0.3339, "step": 7308 }, { "epoch": 2.62, "learning_rate": 3.478495591146302e-06, "loss": 0.2598, "step": 7309 }, { "epoch": 2.62, "learning_rate": 3.477595825085478e-06, "loss": 0.3525, "step": 7310 }, { "epoch": 2.62, "learning_rate": 3.476696059024654e-06, "loss": 0.3136, "step": 7311 }, { "epoch": 2.62, "learning_rate": 3.4757962929638294e-06, "loss": 0.3212, "step": 7312 }, { "epoch": 2.62, "learning_rate": 3.4748965269030056e-06, "loss": 0.3543, "step": 7313 }, { "epoch": 2.63, "learning_rate": 3.4739967608421814e-06, "loss": 0.3542, "step": 7314 }, { "epoch": 2.63, "learning_rate": 3.473096994781357e-06, "loss": 0.3577, "step": 7315 }, { "epoch": 2.63, "learning_rate": 3.472197228720533e-06, "loss": 0.3678, "step": 7316 }, { "epoch": 2.63, "learning_rate": 3.471297462659709e-06, "loss": 0.3583, "step": 7317 }, { "epoch": 2.63, "learning_rate": 3.4703976965988844e-06, "loss": 0.3511, "step": 7318 }, { "epoch": 2.63, "learning_rate": 3.4694979305380606e-06, "loss": 0.3608, "step": 7319 }, { "epoch": 2.63, "learning_rate": 3.468598164477236e-06, "loss": 0.401, "step": 7320 }, { "epoch": 2.63, "learning_rate": 3.467698398416412e-06, "loss": 0.2997, "step": 7321 }, { "epoch": 2.63, "learning_rate": 3.466798632355588e-06, "loss": 0.372, "step": 7322 }, { "epoch": 2.63, "learning_rate": 3.4658988662947633e-06, "loss": 0.2739, "step": 7323 }, { "epoch": 2.63, "learning_rate": 3.4649991002339395e-06, "loss": 0.3241, "step": 7324 }, { "epoch": 2.63, "learning_rate": 3.4640993341731157e-06, "loss": 0.3538, "step": 7325 }, { "epoch": 2.63, "learning_rate": 3.463199568112291e-06, "loss": 0.3832, "step": 7326 }, { "epoch": 2.63, "learning_rate": 3.4622998020514668e-06, "loss": 0.3233, "step": 7327 }, { "epoch": 2.63, "learning_rate": 3.4614000359906425e-06, "loss": 0.3333, "step": 7328 }, { "epoch": 2.63, "learning_rate": 3.4605002699298183e-06, "loss": 0.319, "step": 7329 }, { "epoch": 2.63, "learning_rate": 3.4596005038689945e-06, "loss": 0.2984, "step": 7330 }, { "epoch": 2.63, "learning_rate": 3.45870073780817e-06, "loss": 0.3276, "step": 7331 }, { "epoch": 2.63, "learning_rate": 3.457800971747346e-06, "loss": 0.3315, "step": 7332 }, { "epoch": 2.63, "learning_rate": 3.456901205686522e-06, "loss": 0.2835, "step": 7333 }, { "epoch": 2.63, "learning_rate": 3.4560014396256976e-06, "loss": 0.2824, "step": 7334 }, { "epoch": 2.63, "learning_rate": 3.4551016735648734e-06, "loss": 0.3283, "step": 7335 }, { "epoch": 2.63, "learning_rate": 3.4542019075040495e-06, "loss": 0.3488, "step": 7336 }, { "epoch": 2.63, "learning_rate": 3.453302141443225e-06, "loss": 0.3158, "step": 7337 }, { "epoch": 2.63, "learning_rate": 3.452402375382401e-06, "loss": 0.3309, "step": 7338 }, { "epoch": 2.63, "learning_rate": 3.4515026093215764e-06, "loss": 0.3385, "step": 7339 }, { "epoch": 2.63, "learning_rate": 3.4506028432607526e-06, "loss": 0.3642, "step": 7340 }, { "epoch": 2.63, "learning_rate": 3.4497030771999284e-06, "loss": 0.3896, "step": 7341 }, { "epoch": 2.64, "learning_rate": 3.4488033111391037e-06, "loss": 0.3412, "step": 7342 }, { "epoch": 2.64, "learning_rate": 3.44790354507828e-06, "loss": 0.3552, "step": 7343 }, { "epoch": 2.64, "learning_rate": 3.447003779017456e-06, "loss": 0.3381, "step": 7344 }, { "epoch": 2.64, "learning_rate": 3.4461040129566315e-06, "loss": 0.4197, "step": 7345 }, { "epoch": 2.64, "learning_rate": 3.4452042468958077e-06, "loss": 0.3334, "step": 7346 }, { "epoch": 2.64, "learning_rate": 3.444304480834983e-06, "loss": 0.2964, "step": 7347 }, { "epoch": 2.64, "learning_rate": 3.4434047147741588e-06, "loss": 0.2738, "step": 7348 }, { "epoch": 2.64, "learning_rate": 3.442504948713335e-06, "loss": 0.3446, "step": 7349 }, { "epoch": 2.64, "learning_rate": 3.4416051826525103e-06, "loss": 0.3167, "step": 7350 }, { "epoch": 2.64, "learning_rate": 3.4407054165916865e-06, "loss": 0.3128, "step": 7351 }, { "epoch": 2.64, "learning_rate": 3.4398056505308623e-06, "loss": 0.2823, "step": 7352 }, { "epoch": 2.64, "learning_rate": 3.438905884470038e-06, "loss": 0.3297, "step": 7353 }, { "epoch": 2.64, "learning_rate": 3.438006118409214e-06, "loss": 0.3033, "step": 7354 }, { "epoch": 2.64, "learning_rate": 3.43710635234839e-06, "loss": 0.3421, "step": 7355 }, { "epoch": 2.64, "learning_rate": 3.4362065862875653e-06, "loss": 0.3207, "step": 7356 }, { "epoch": 2.64, "learning_rate": 3.4353068202267415e-06, "loss": 0.3299, "step": 7357 }, { "epoch": 2.64, "learning_rate": 3.434407054165917e-06, "loss": 0.3376, "step": 7358 }, { "epoch": 2.64, "learning_rate": 3.433507288105093e-06, "loss": 0.3777, "step": 7359 }, { "epoch": 2.64, "learning_rate": 3.432607522044269e-06, "loss": 0.2734, "step": 7360 }, { "epoch": 2.64, "learning_rate": 3.4317077559834446e-06, "loss": 0.3018, "step": 7361 }, { "epoch": 2.64, "learning_rate": 3.4308079899226204e-06, "loss": 0.2896, "step": 7362 }, { "epoch": 2.64, "learning_rate": 3.4299082238617966e-06, "loss": 0.2912, "step": 7363 }, { "epoch": 2.64, "learning_rate": 3.429008457800972e-06, "loss": 0.3335, "step": 7364 }, { "epoch": 2.64, "learning_rate": 3.428108691740148e-06, "loss": 0.3177, "step": 7365 }, { "epoch": 2.64, "learning_rate": 3.4272089256793235e-06, "loss": 0.3943, "step": 7366 }, { "epoch": 2.64, "learning_rate": 3.4263091596184992e-06, "loss": 0.3047, "step": 7367 }, { "epoch": 2.64, "learning_rate": 3.4254093935576754e-06, "loss": 0.3729, "step": 7368 }, { "epoch": 2.65, "learning_rate": 3.4245096274968508e-06, "loss": 0.341, "step": 7369 }, { "epoch": 2.65, "learning_rate": 3.423609861436027e-06, "loss": 0.3372, "step": 7370 }, { "epoch": 2.65, "learning_rate": 3.422710095375203e-06, "loss": 0.3452, "step": 7371 }, { "epoch": 2.65, "learning_rate": 3.4218103293143785e-06, "loss": 0.3846, "step": 7372 }, { "epoch": 2.65, "learning_rate": 3.4209105632535543e-06, "loss": 0.3028, "step": 7373 }, { "epoch": 2.65, "learning_rate": 3.4200107971927305e-06, "loss": 0.3298, "step": 7374 }, { "epoch": 2.65, "learning_rate": 3.419111031131906e-06, "loss": 0.354, "step": 7375 }, { "epoch": 2.65, "learning_rate": 3.418211265071082e-06, "loss": 0.2945, "step": 7376 }, { "epoch": 2.65, "learning_rate": 3.4173114990102573e-06, "loss": 0.3112, "step": 7377 }, { "epoch": 2.65, "learning_rate": 3.4164117329494335e-06, "loss": 0.3581, "step": 7378 }, { "epoch": 2.65, "learning_rate": 3.4155119668886093e-06, "loss": 0.3171, "step": 7379 }, { "epoch": 2.65, "learning_rate": 3.414612200827785e-06, "loss": 0.3365, "step": 7380 }, { "epoch": 2.65, "learning_rate": 3.413712434766961e-06, "loss": 0.3605, "step": 7381 }, { "epoch": 2.65, "learning_rate": 3.412812668706137e-06, "loss": 0.36, "step": 7382 }, { "epoch": 2.65, "learning_rate": 3.4119129026453124e-06, "loss": 0.3089, "step": 7383 }, { "epoch": 2.65, "learning_rate": 3.4110131365844886e-06, "loss": 0.371, "step": 7384 }, { "epoch": 2.65, "learning_rate": 3.410113370523664e-06, "loss": 0.3245, "step": 7385 }, { "epoch": 2.65, "learning_rate": 3.4092136044628397e-06, "loss": 0.3415, "step": 7386 }, { "epoch": 2.65, "learning_rate": 3.408313838402016e-06, "loss": 0.2891, "step": 7387 }, { "epoch": 2.65, "learning_rate": 3.4074140723411912e-06, "loss": 0.372, "step": 7388 }, { "epoch": 2.65, "learning_rate": 3.4065143062803674e-06, "loss": 0.3786, "step": 7389 }, { "epoch": 2.65, "learning_rate": 3.4056145402195436e-06, "loss": 0.3271, "step": 7390 }, { "epoch": 2.65, "learning_rate": 3.404714774158719e-06, "loss": 0.3303, "step": 7391 }, { "epoch": 2.65, "learning_rate": 3.4038150080978947e-06, "loss": 0.3658, "step": 7392 }, { "epoch": 2.65, "learning_rate": 3.402915242037071e-06, "loss": 0.3689, "step": 7393 }, { "epoch": 2.65, "learning_rate": 3.4020154759762463e-06, "loss": 0.3603, "step": 7394 }, { "epoch": 2.65, "learning_rate": 3.4011157099154224e-06, "loss": 0.3022, "step": 7395 }, { "epoch": 2.65, "learning_rate": 3.400215943854598e-06, "loss": 0.3901, "step": 7396 }, { "epoch": 2.66, "learning_rate": 3.399316177793774e-06, "loss": 0.3215, "step": 7397 }, { "epoch": 2.66, "learning_rate": 3.3984164117329497e-06, "loss": 0.2863, "step": 7398 }, { "epoch": 2.66, "learning_rate": 3.3975166456721255e-06, "loss": 0.3161, "step": 7399 }, { "epoch": 2.66, "learning_rate": 3.3966168796113013e-06, "loss": 0.2879, "step": 7400 }, { "epoch": 2.66, "learning_rate": 3.3957171135504775e-06, "loss": 0.3827, "step": 7401 }, { "epoch": 2.66, "learning_rate": 3.394817347489653e-06, "loss": 0.3711, "step": 7402 }, { "epoch": 2.66, "learning_rate": 3.393917581428829e-06, "loss": 0.3143, "step": 7403 }, { "epoch": 2.66, "learning_rate": 3.3930178153680044e-06, "loss": 0.3643, "step": 7404 }, { "epoch": 2.66, "learning_rate": 3.3921180493071806e-06, "loss": 0.3703, "step": 7405 }, { "epoch": 2.66, "learning_rate": 3.3912182832463563e-06, "loss": 0.2998, "step": 7406 }, { "epoch": 2.66, "learning_rate": 3.3903185171855317e-06, "loss": 0.3856, "step": 7407 }, { "epoch": 2.66, "learning_rate": 3.389418751124708e-06, "loss": 0.4081, "step": 7408 }, { "epoch": 2.66, "learning_rate": 3.388518985063884e-06, "loss": 0.3683, "step": 7409 }, { "epoch": 2.66, "learning_rate": 3.3876192190030594e-06, "loss": 0.3838, "step": 7410 }, { "epoch": 2.66, "learning_rate": 3.386719452942235e-06, "loss": 0.3552, "step": 7411 }, { "epoch": 2.66, "learning_rate": 3.3858196868814114e-06, "loss": 0.384, "step": 7412 }, { "epoch": 2.66, "learning_rate": 3.3849199208205867e-06, "loss": 0.2664, "step": 7413 }, { "epoch": 2.66, "learning_rate": 3.384020154759763e-06, "loss": 0.3144, "step": 7414 }, { "epoch": 2.66, "learning_rate": 3.3831203886989382e-06, "loss": 0.346, "step": 7415 }, { "epoch": 2.66, "learning_rate": 3.3822206226381144e-06, "loss": 0.3147, "step": 7416 }, { "epoch": 2.66, "learning_rate": 3.38132085657729e-06, "loss": 0.3863, "step": 7417 }, { "epoch": 2.66, "learning_rate": 3.380421090516466e-06, "loss": 0.3116, "step": 7418 }, { "epoch": 2.66, "learning_rate": 3.3795213244556417e-06, "loss": 0.3787, "step": 7419 }, { "epoch": 2.66, "learning_rate": 3.378621558394818e-06, "loss": 0.3305, "step": 7420 }, { "epoch": 2.66, "learning_rate": 3.3777217923339933e-06, "loss": 0.3302, "step": 7421 }, { "epoch": 2.66, "learning_rate": 3.3768220262731695e-06, "loss": 0.3202, "step": 7422 }, { "epoch": 2.66, "learning_rate": 3.375922260212345e-06, "loss": 0.3109, "step": 7423 }, { "epoch": 2.66, "learning_rate": 3.375022494151521e-06, "loss": 0.2811, "step": 7424 }, { "epoch": 2.67, "learning_rate": 3.3741227280906968e-06, "loss": 0.3236, "step": 7425 }, { "epoch": 2.67, "learning_rate": 3.373222962029872e-06, "loss": 0.2911, "step": 7426 }, { "epoch": 2.67, "learning_rate": 3.3723231959690483e-06, "loss": 0.3746, "step": 7427 }, { "epoch": 2.67, "learning_rate": 3.3714234299082245e-06, "loss": 0.3636, "step": 7428 }, { "epoch": 2.67, "learning_rate": 3.3705236638474e-06, "loss": 0.4271, "step": 7429 }, { "epoch": 2.67, "learning_rate": 3.369623897786576e-06, "loss": 0.3098, "step": 7430 }, { "epoch": 2.67, "learning_rate": 3.368724131725752e-06, "loss": 0.334, "step": 7431 }, { "epoch": 2.67, "learning_rate": 3.367824365664927e-06, "loss": 0.3241, "step": 7432 }, { "epoch": 2.67, "learning_rate": 3.3669245996041034e-06, "loss": 0.2937, "step": 7433 }, { "epoch": 2.67, "learning_rate": 3.3660248335432787e-06, "loss": 0.3141, "step": 7434 }, { "epoch": 2.67, "learning_rate": 3.365125067482455e-06, "loss": 0.2837, "step": 7435 }, { "epoch": 2.67, "learning_rate": 3.3642253014216307e-06, "loss": 0.3454, "step": 7436 }, { "epoch": 2.67, "learning_rate": 3.3633255353608064e-06, "loss": 0.3345, "step": 7437 }, { "epoch": 2.67, "learning_rate": 3.362425769299982e-06, "loss": 0.3184, "step": 7438 }, { "epoch": 2.67, "learning_rate": 3.3615260032391584e-06, "loss": 0.3587, "step": 7439 }, { "epoch": 2.67, "learning_rate": 3.3606262371783337e-06, "loss": 0.3833, "step": 7440 }, { "epoch": 2.67, "learning_rate": 3.35972647111751e-06, "loss": 0.3677, "step": 7441 }, { "epoch": 2.67, "learning_rate": 3.3588267050566853e-06, "loss": 0.3596, "step": 7442 }, { "epoch": 2.67, "learning_rate": 3.3579269389958615e-06, "loss": 0.3201, "step": 7443 }, { "epoch": 2.67, "learning_rate": 3.3570271729350372e-06, "loss": 0.3224, "step": 7444 }, { "epoch": 2.67, "learning_rate": 3.356127406874213e-06, "loss": 0.2638, "step": 7445 }, { "epoch": 2.67, "learning_rate": 3.3552276408133888e-06, "loss": 0.3044, "step": 7446 }, { "epoch": 2.67, "learning_rate": 3.354327874752565e-06, "loss": 0.3492, "step": 7447 }, { "epoch": 2.67, "learning_rate": 3.3534281086917403e-06, "loss": 0.302, "step": 7448 }, { "epoch": 2.67, "learning_rate": 3.3525283426309165e-06, "loss": 0.4189, "step": 7449 }, { "epoch": 2.67, "learning_rate": 3.351628576570092e-06, "loss": 0.3672, "step": 7450 }, { "epoch": 2.67, "learning_rate": 3.3507288105092676e-06, "loss": 0.3911, "step": 7451 }, { "epoch": 2.67, "learning_rate": 3.349829044448444e-06, "loss": 0.3229, "step": 7452 }, { "epoch": 2.68, "learning_rate": 3.348929278387619e-06, "loss": 0.3442, "step": 7453 }, { "epoch": 2.68, "learning_rate": 3.3480295123267953e-06, "loss": 0.2756, "step": 7454 }, { "epoch": 2.68, "learning_rate": 3.347129746265971e-06, "loss": 0.3585, "step": 7455 }, { "epoch": 2.68, "learning_rate": 3.346229980205147e-06, "loss": 0.3342, "step": 7456 }, { "epoch": 2.68, "learning_rate": 3.3453302141443226e-06, "loss": 0.3598, "step": 7457 }, { "epoch": 2.68, "learning_rate": 3.344430448083499e-06, "loss": 0.2897, "step": 7458 }, { "epoch": 2.68, "learning_rate": 3.343530682022674e-06, "loss": 0.372, "step": 7459 }, { "epoch": 2.68, "learning_rate": 3.3426309159618504e-06, "loss": 0.3888, "step": 7460 }, { "epoch": 2.68, "learning_rate": 3.3417311499010257e-06, "loss": 0.3725, "step": 7461 }, { "epoch": 2.68, "learning_rate": 3.340831383840202e-06, "loss": 0.2823, "step": 7462 }, { "epoch": 2.68, "learning_rate": 3.3399316177793777e-06, "loss": 0.3298, "step": 7463 }, { "epoch": 2.68, "learning_rate": 3.3390318517185535e-06, "loss": 0.3434, "step": 7464 }, { "epoch": 2.68, "learning_rate": 3.3381320856577292e-06, "loss": 0.352, "step": 7465 }, { "epoch": 2.68, "learning_rate": 3.3372323195969054e-06, "loss": 0.3635, "step": 7466 }, { "epoch": 2.68, "learning_rate": 3.3363325535360808e-06, "loss": 0.3492, "step": 7467 }, { "epoch": 2.68, "learning_rate": 3.335432787475257e-06, "loss": 0.3029, "step": 7468 }, { "epoch": 2.68, "learning_rate": 3.3345330214144323e-06, "loss": 0.3618, "step": 7469 }, { "epoch": 2.68, "learning_rate": 3.333633255353608e-06, "loss": 0.3243, "step": 7470 }, { "epoch": 2.68, "learning_rate": 3.3327334892927843e-06, "loss": 0.286, "step": 7471 }, { "epoch": 2.68, "learning_rate": 3.3318337232319596e-06, "loss": 0.2833, "step": 7472 }, { "epoch": 2.68, "learning_rate": 3.330933957171136e-06, "loss": 0.33, "step": 7473 }, { "epoch": 2.68, "learning_rate": 3.330034191110312e-06, "loss": 0.3243, "step": 7474 }, { "epoch": 2.68, "learning_rate": 3.3291344250494873e-06, "loss": 0.3384, "step": 7475 }, { "epoch": 2.68, "learning_rate": 3.328234658988663e-06, "loss": 0.2722, "step": 7476 }, { "epoch": 2.68, "learning_rate": 3.3273348929278393e-06, "loss": 0.3127, "step": 7477 }, { "epoch": 2.68, "learning_rate": 3.3264351268670146e-06, "loss": 0.3671, "step": 7478 }, { "epoch": 2.68, "learning_rate": 3.325535360806191e-06, "loss": 0.3244, "step": 7479 }, { "epoch": 2.68, "learning_rate": 3.324635594745366e-06, "loss": 0.3559, "step": 7480 }, { "epoch": 2.69, "learning_rate": 3.3237358286845424e-06, "loss": 0.2882, "step": 7481 }, { "epoch": 2.69, "learning_rate": 3.322836062623718e-06, "loss": 0.3588, "step": 7482 }, { "epoch": 2.69, "learning_rate": 3.322836062623718e-06, "loss": 0.5509, "step": 7483 }, { "epoch": 2.69, "learning_rate": 3.321936296562894e-06, "loss": 0.3632, "step": 7484 }, { "epoch": 2.69, "learning_rate": 3.3210365305020697e-06, "loss": 0.3557, "step": 7485 }, { "epoch": 2.69, "learning_rate": 3.320136764441246e-06, "loss": 0.3067, "step": 7486 }, { "epoch": 2.69, "learning_rate": 3.3192369983804212e-06, "loss": 0.394, "step": 7487 }, { "epoch": 2.69, "learning_rate": 3.3183372323195974e-06, "loss": 0.3306, "step": 7488 }, { "epoch": 2.69, "learning_rate": 3.3174374662587728e-06, "loss": 0.338, "step": 7489 }, { "epoch": 2.69, "learning_rate": 3.316537700197949e-06, "loss": 0.3542, "step": 7490 }, { "epoch": 2.69, "learning_rate": 3.3156379341371247e-06, "loss": 0.3073, "step": 7491 }, { "epoch": 2.69, "learning_rate": 3.3147381680763e-06, "loss": 0.3346, "step": 7492 }, { "epoch": 2.69, "learning_rate": 3.3138384020154763e-06, "loss": 0.3151, "step": 7493 }, { "epoch": 2.69, "learning_rate": 3.3129386359546524e-06, "loss": 0.3406, "step": 7494 }, { "epoch": 2.69, "learning_rate": 3.312038869893828e-06, "loss": 0.2864, "step": 7495 }, { "epoch": 2.69, "learning_rate": 3.3111391038330036e-06, "loss": 0.3402, "step": 7496 }, { "epoch": 2.69, "learning_rate": 3.3102393377721798e-06, "loss": 0.4059, "step": 7497 }, { "epoch": 2.69, "learning_rate": 3.309339571711355e-06, "loss": 0.3092, "step": 7498 }, { "epoch": 2.69, "learning_rate": 3.3084398056505313e-06, "loss": 0.329, "step": 7499 }, { "epoch": 2.69, "learning_rate": 3.3075400395897066e-06, "loss": 0.3286, "step": 7500 }, { "epoch": 2.69, "learning_rate": 3.306640273528883e-06, "loss": 0.3371, "step": 7501 }, { "epoch": 2.69, "learning_rate": 3.3057405074680586e-06, "loss": 0.3787, "step": 7502 }, { "epoch": 2.69, "learning_rate": 3.3048407414072344e-06, "loss": 0.314, "step": 7503 }, { "epoch": 2.69, "learning_rate": 3.30394097534641e-06, "loss": 0.3271, "step": 7504 }, { "epoch": 2.69, "learning_rate": 3.3030412092855863e-06, "loss": 0.3005, "step": 7505 }, { "epoch": 2.69, "learning_rate": 3.3021414432247617e-06, "loss": 0.2925, "step": 7506 }, { "epoch": 2.69, "learning_rate": 3.301241677163938e-06, "loss": 0.3862, "step": 7507 }, { "epoch": 2.69, "learning_rate": 3.300341911103113e-06, "loss": 0.2919, "step": 7508 }, { "epoch": 2.7, "learning_rate": 3.2994421450422894e-06, "loss": 0.2952, "step": 7509 }, { "epoch": 2.7, "learning_rate": 3.298542378981465e-06, "loss": 0.3459, "step": 7510 }, { "epoch": 2.7, "learning_rate": 3.2976426129206405e-06, "loss": 0.4145, "step": 7511 }, { "epoch": 2.7, "learning_rate": 3.2967428468598167e-06, "loss": 0.2919, "step": 7512 }, { "epoch": 2.7, "learning_rate": 3.295843080798993e-06, "loss": 0.3206, "step": 7513 }, { "epoch": 2.7, "learning_rate": 3.2949433147381682e-06, "loss": 0.3662, "step": 7514 }, { "epoch": 2.7, "learning_rate": 3.294043548677344e-06, "loss": 0.3303, "step": 7515 }, { "epoch": 2.7, "learning_rate": 3.29314378261652e-06, "loss": 0.318, "step": 7516 }, { "epoch": 2.7, "learning_rate": 3.2922440165556955e-06, "loss": 0.2999, "step": 7517 }, { "epoch": 2.7, "learning_rate": 3.2913442504948717e-06, "loss": 0.3444, "step": 7518 }, { "epoch": 2.7, "learning_rate": 3.290444484434047e-06, "loss": 0.3462, "step": 7519 }, { "epoch": 2.7, "learning_rate": 3.2895447183732233e-06, "loss": 0.3365, "step": 7520 }, { "epoch": 2.7, "learning_rate": 3.288644952312399e-06, "loss": 0.3238, "step": 7521 }, { "epoch": 2.7, "learning_rate": 3.287745186251575e-06, "loss": 0.3142, "step": 7522 }, { "epoch": 2.7, "learning_rate": 3.2868454201907506e-06, "loss": 0.4418, "step": 7523 }, { "epoch": 2.7, "learning_rate": 3.2859456541299268e-06, "loss": 0.3721, "step": 7524 }, { "epoch": 2.7, "learning_rate": 3.285045888069102e-06, "loss": 0.3129, "step": 7525 }, { "epoch": 2.7, "learning_rate": 3.2841461220082783e-06, "loss": 0.2711, "step": 7526 }, { "epoch": 2.7, "learning_rate": 3.2832463559474537e-06, "loss": 0.3579, "step": 7527 }, { "epoch": 2.7, "learning_rate": 3.28234658988663e-06, "loss": 0.3458, "step": 7528 }, { "epoch": 2.7, "learning_rate": 3.2814468238258056e-06, "loss": 0.3373, "step": 7529 }, { "epoch": 2.7, "learning_rate": 3.280547057764981e-06, "loss": 0.3534, "step": 7530 }, { "epoch": 2.7, "learning_rate": 3.279647291704157e-06, "loss": 0.3197, "step": 7531 }, { "epoch": 2.7, "learning_rate": 3.2787475256433334e-06, "loss": 0.3303, "step": 7532 }, { "epoch": 2.7, "learning_rate": 3.2778477595825087e-06, "loss": 0.3488, "step": 7533 }, { "epoch": 2.7, "learning_rate": 3.276947993521685e-06, "loss": 0.3142, "step": 7534 }, { "epoch": 2.7, "learning_rate": 3.2760482274608607e-06, "loss": 0.2862, "step": 7535 }, { "epoch": 2.7, "learning_rate": 3.275148461400036e-06, "loss": 0.3412, "step": 7536 }, { "epoch": 2.71, "learning_rate": 3.274248695339212e-06, "loss": 0.2798, "step": 7537 }, { "epoch": 2.71, "learning_rate": 3.2733489292783875e-06, "loss": 0.3485, "step": 7538 }, { "epoch": 2.71, "learning_rate": 3.2724491632175637e-06, "loss": 0.3935, "step": 7539 }, { "epoch": 2.71, "learning_rate": 3.2715493971567395e-06, "loss": 0.3272, "step": 7540 }, { "epoch": 2.71, "learning_rate": 3.2706496310959153e-06, "loss": 0.3322, "step": 7541 }, { "epoch": 2.71, "learning_rate": 3.269749865035091e-06, "loss": 0.3975, "step": 7542 }, { "epoch": 2.71, "learning_rate": 3.2688500989742672e-06, "loss": 0.3081, "step": 7543 }, { "epoch": 2.71, "learning_rate": 3.2679503329134426e-06, "loss": 0.3379, "step": 7544 }, { "epoch": 2.71, "learning_rate": 3.2670505668526188e-06, "loss": 0.2973, "step": 7545 }, { "epoch": 2.71, "learning_rate": 3.266150800791794e-06, "loss": 0.3718, "step": 7546 }, { "epoch": 2.71, "learning_rate": 3.2652510347309703e-06, "loss": 0.3308, "step": 7547 }, { "epoch": 2.71, "learning_rate": 3.264351268670146e-06, "loss": 0.3508, "step": 7548 }, { "epoch": 2.71, "learning_rate": 3.263451502609322e-06, "loss": 0.3363, "step": 7549 }, { "epoch": 2.71, "learning_rate": 3.2625517365484976e-06, "loss": 0.2955, "step": 7550 }, { "epoch": 2.71, "learning_rate": 3.261651970487674e-06, "loss": 0.3593, "step": 7551 }, { "epoch": 2.71, "learning_rate": 3.260752204426849e-06, "loss": 0.2963, "step": 7552 }, { "epoch": 2.71, "learning_rate": 3.2598524383660253e-06, "loss": 0.3042, "step": 7553 }, { "epoch": 2.71, "learning_rate": 3.258952672305201e-06, "loss": 0.3782, "step": 7554 }, { "epoch": 2.71, "learning_rate": 3.2580529062443765e-06, "loss": 0.4186, "step": 7555 }, { "epoch": 2.71, "learning_rate": 3.2571531401835526e-06, "loss": 0.3381, "step": 7556 }, { "epoch": 2.71, "learning_rate": 3.256253374122728e-06, "loss": 0.366, "step": 7557 }, { "epoch": 2.71, "learning_rate": 3.255353608061904e-06, "loss": 0.3091, "step": 7558 }, { "epoch": 2.71, "learning_rate": 3.25445384200108e-06, "loss": 0.3384, "step": 7559 }, { "epoch": 2.71, "learning_rate": 3.2535540759402557e-06, "loss": 0.3422, "step": 7560 }, { "epoch": 2.71, "learning_rate": 3.2526543098794315e-06, "loss": 0.3254, "step": 7561 }, { "epoch": 2.71, "learning_rate": 3.2517545438186077e-06, "loss": 0.3096, "step": 7562 }, { "epoch": 2.71, "learning_rate": 3.250854777757783e-06, "loss": 0.2737, "step": 7563 }, { "epoch": 2.72, "learning_rate": 3.2499550116969592e-06, "loss": 0.3205, "step": 7564 }, { "epoch": 2.72, "learning_rate": 3.2490552456361346e-06, "loss": 0.3677, "step": 7565 }, { "epoch": 2.72, "learning_rate": 3.2481554795753108e-06, "loss": 0.351, "step": 7566 }, { "epoch": 2.72, "learning_rate": 3.2472557135144865e-06, "loss": 0.2571, "step": 7567 }, { "epoch": 2.72, "learning_rate": 3.2463559474536623e-06, "loss": 0.3275, "step": 7568 }, { "epoch": 2.72, "learning_rate": 3.245456181392838e-06, "loss": 0.3821, "step": 7569 }, { "epoch": 2.72, "learning_rate": 3.2445564153320143e-06, "loss": 0.3088, "step": 7570 }, { "epoch": 2.72, "learning_rate": 3.2436566492711896e-06, "loss": 0.2916, "step": 7571 }, { "epoch": 2.72, "learning_rate": 3.242756883210366e-06, "loss": 0.376, "step": 7572 }, { "epoch": 2.72, "learning_rate": 3.2418571171495416e-06, "loss": 0.296, "step": 7573 }, { "epoch": 2.72, "learning_rate": 3.240957351088717e-06, "loss": 0.2956, "step": 7574 }, { "epoch": 2.72, "learning_rate": 3.240057585027893e-06, "loss": 0.3394, "step": 7575 }, { "epoch": 2.72, "learning_rate": 3.2391578189670684e-06, "loss": 0.3333, "step": 7576 }, { "epoch": 2.72, "learning_rate": 3.2382580529062446e-06, "loss": 0.3777, "step": 7577 }, { "epoch": 2.72, "learning_rate": 3.237358286845421e-06, "loss": 0.3544, "step": 7578 }, { "epoch": 2.72, "learning_rate": 3.236458520784596e-06, "loss": 0.3708, "step": 7579 }, { "epoch": 2.72, "learning_rate": 3.235558754723772e-06, "loss": 0.3513, "step": 7580 }, { "epoch": 2.72, "learning_rate": 3.234658988662948e-06, "loss": 0.3413, "step": 7581 }, { "epoch": 2.72, "learning_rate": 3.2337592226021235e-06, "loss": 0.3796, "step": 7582 }, { "epoch": 2.72, "learning_rate": 3.2328594565412997e-06, "loss": 0.3267, "step": 7583 }, { "epoch": 2.72, "learning_rate": 3.231959690480475e-06, "loss": 0.3173, "step": 7584 }, { "epoch": 2.72, "learning_rate": 3.2310599244196512e-06, "loss": 0.3538, "step": 7585 }, { "epoch": 2.72, "learning_rate": 3.230160158358827e-06, "loss": 0.3233, "step": 7586 }, { "epoch": 2.72, "learning_rate": 3.2292603922980028e-06, "loss": 0.3504, "step": 7587 }, { "epoch": 2.72, "learning_rate": 3.2283606262371785e-06, "loss": 0.3485, "step": 7588 }, { "epoch": 2.72, "learning_rate": 3.2274608601763547e-06, "loss": 0.4038, "step": 7589 }, { "epoch": 2.72, "learning_rate": 3.22656109411553e-06, "loss": 0.3207, "step": 7590 }, { "epoch": 2.72, "learning_rate": 3.2256613280547063e-06, "loss": 0.2605, "step": 7591 }, { "epoch": 2.73, "learning_rate": 3.224761561993882e-06, "loss": 0.3308, "step": 7592 }, { "epoch": 2.73, "learning_rate": 3.223861795933058e-06, "loss": 0.3702, "step": 7593 }, { "epoch": 2.73, "learning_rate": 3.2229620298722336e-06, "loss": 0.3744, "step": 7594 }, { "epoch": 2.73, "learning_rate": 3.222062263811409e-06, "loss": 0.3629, "step": 7595 }, { "epoch": 2.73, "learning_rate": 3.221162497750585e-06, "loss": 0.35, "step": 7596 }, { "epoch": 2.73, "learning_rate": 3.2202627316897613e-06, "loss": 0.3022, "step": 7597 }, { "epoch": 2.73, "learning_rate": 3.2193629656289366e-06, "loss": 0.2976, "step": 7598 }, { "epoch": 2.73, "learning_rate": 3.2184631995681124e-06, "loss": 0.3551, "step": 7599 }, { "epoch": 2.73, "learning_rate": 3.2175634335072886e-06, "loss": 0.3782, "step": 7600 }, { "epoch": 2.73, "learning_rate": 3.216663667446464e-06, "loss": 0.348, "step": 7601 }, { "epoch": 2.73, "learning_rate": 3.21576390138564e-06, "loss": 0.3936, "step": 7602 }, { "epoch": 2.73, "learning_rate": 3.2148641353248155e-06, "loss": 0.3461, "step": 7603 }, { "epoch": 2.73, "learning_rate": 3.2139643692639917e-06, "loss": 0.3637, "step": 7604 }, { "epoch": 2.73, "learning_rate": 3.2130646032031674e-06, "loss": 0.3105, "step": 7605 }, { "epoch": 2.73, "learning_rate": 3.212164837142343e-06, "loss": 0.3448, "step": 7606 }, { "epoch": 2.73, "learning_rate": 3.211265071081519e-06, "loss": 0.3135, "step": 7607 }, { "epoch": 2.73, "learning_rate": 3.210365305020695e-06, "loss": 0.3536, "step": 7608 }, { "epoch": 2.73, "learning_rate": 3.2094655389598705e-06, "loss": 0.289, "step": 7609 }, { "epoch": 2.73, "learning_rate": 3.2085657728990467e-06, "loss": 0.3341, "step": 7610 }, { "epoch": 2.73, "learning_rate": 3.2076660068382225e-06, "loss": 0.5866, "step": 7611 }, { "epoch": 2.73, "learning_rate": 3.2067662407773982e-06, "loss": 0.3177, "step": 7612 }, { "epoch": 2.73, "learning_rate": 3.205866474716574e-06, "loss": 0.3374, "step": 7613 }, { "epoch": 2.73, "learning_rate": 3.2049667086557494e-06, "loss": 0.3279, "step": 7614 }, { "epoch": 2.73, "learning_rate": 3.2040669425949255e-06, "loss": 0.3797, "step": 7615 }, { "epoch": 2.73, "learning_rate": 3.2031671765341017e-06, "loss": 0.3267, "step": 7616 }, { "epoch": 2.73, "learning_rate": 3.202267410473277e-06, "loss": 0.311, "step": 7617 }, { "epoch": 2.73, "learning_rate": 3.2013676444124533e-06, "loss": 0.3651, "step": 7618 }, { "epoch": 2.73, "learning_rate": 3.200467878351629e-06, "loss": 0.3412, "step": 7619 }, { "epoch": 2.74, "learning_rate": 3.1995681122908044e-06, "loss": 0.3638, "step": 7620 }, { "epoch": 2.74, "learning_rate": 3.1986683462299806e-06, "loss": 0.3092, "step": 7621 }, { "epoch": 2.74, "learning_rate": 3.197768580169156e-06, "loss": 0.358, "step": 7622 }, { "epoch": 2.74, "learning_rate": 3.196868814108332e-06, "loss": 0.3699, "step": 7623 }, { "epoch": 2.74, "learning_rate": 3.195969048047508e-06, "loss": 0.3537, "step": 7624 }, { "epoch": 2.74, "learning_rate": 3.1950692819866837e-06, "loss": 0.3418, "step": 7625 }, { "epoch": 2.74, "learning_rate": 3.1941695159258594e-06, "loss": 0.3246, "step": 7626 }, { "epoch": 2.74, "learning_rate": 3.1932697498650356e-06, "loss": 0.2803, "step": 7627 }, { "epoch": 2.74, "learning_rate": 3.192369983804211e-06, "loss": 0.3166, "step": 7628 }, { "epoch": 2.74, "learning_rate": 3.191470217743387e-06, "loss": 0.33, "step": 7629 }, { "epoch": 2.74, "learning_rate": 3.1905704516825625e-06, "loss": 0.3005, "step": 7630 }, { "epoch": 2.74, "learning_rate": 3.1896706856217387e-06, "loss": 0.3828, "step": 7631 }, { "epoch": 2.74, "learning_rate": 3.1887709195609145e-06, "loss": 0.3971, "step": 7632 }, { "epoch": 2.74, "learning_rate": 3.18787115350009e-06, "loss": 0.3393, "step": 7633 }, { "epoch": 2.74, "learning_rate": 3.186971387439266e-06, "loss": 0.3316, "step": 7634 }, { "epoch": 2.74, "learning_rate": 3.186071621378442e-06, "loss": 0.3984, "step": 7635 }, { "epoch": 2.74, "learning_rate": 3.1851718553176175e-06, "loss": 0.3134, "step": 7636 }, { "epoch": 2.74, "learning_rate": 3.1842720892567937e-06, "loss": 0.3538, "step": 7637 }, { "epoch": 2.74, "learning_rate": 3.1833723231959695e-06, "loss": 0.3197, "step": 7638 }, { "epoch": 2.74, "learning_rate": 3.182472557135145e-06, "loss": 0.3504, "step": 7639 }, { "epoch": 2.74, "learning_rate": 3.181572791074321e-06, "loss": 0.3194, "step": 7640 }, { "epoch": 2.74, "learning_rate": 3.1806730250134964e-06, "loss": 0.3352, "step": 7641 }, { "epoch": 2.74, "learning_rate": 3.1797732589526726e-06, "loss": 0.346, "step": 7642 }, { "epoch": 2.74, "learning_rate": 3.1788734928918483e-06, "loss": 0.3496, "step": 7643 }, { "epoch": 2.74, "learning_rate": 3.177973726831024e-06, "loss": 0.2963, "step": 7644 }, { "epoch": 2.74, "learning_rate": 3.1770739607702e-06, "loss": 0.373, "step": 7645 }, { "epoch": 2.74, "learning_rate": 3.176174194709376e-06, "loss": 0.3237, "step": 7646 }, { "epoch": 2.74, "learning_rate": 3.1752744286485514e-06, "loss": 0.2991, "step": 7647 }, { "epoch": 2.75, "learning_rate": 3.1743746625877276e-06, "loss": 0.3323, "step": 7648 }, { "epoch": 2.75, "learning_rate": 3.173474896526903e-06, "loss": 0.3333, "step": 7649 }, { "epoch": 2.75, "learning_rate": 3.172575130466079e-06, "loss": 0.3175, "step": 7650 }, { "epoch": 2.75, "learning_rate": 3.171675364405255e-06, "loss": 0.3353, "step": 7651 }, { "epoch": 2.75, "learning_rate": 3.1707755983444307e-06, "loss": 0.3621, "step": 7652 }, { "epoch": 2.75, "learning_rate": 3.1698758322836065e-06, "loss": 0.3134, "step": 7653 }, { "epoch": 2.75, "learning_rate": 3.1689760662227827e-06, "loss": 0.4407, "step": 7654 }, { "epoch": 2.75, "learning_rate": 3.168076300161958e-06, "loss": 0.3325, "step": 7655 }, { "epoch": 2.75, "learning_rate": 3.167176534101134e-06, "loss": 0.3225, "step": 7656 }, { "epoch": 2.75, "learning_rate": 3.16627676804031e-06, "loss": 0.3275, "step": 7657 }, { "epoch": 2.75, "learning_rate": 3.1653770019794853e-06, "loss": 0.3646, "step": 7658 }, { "epoch": 2.75, "learning_rate": 3.1644772359186615e-06, "loss": 0.3441, "step": 7659 }, { "epoch": 2.75, "learning_rate": 3.163577469857837e-06, "loss": 0.3067, "step": 7660 }, { "epoch": 2.75, "learning_rate": 3.162677703797013e-06, "loss": 0.3432, "step": 7661 }, { "epoch": 2.75, "learning_rate": 3.1617779377361892e-06, "loss": 0.3201, "step": 7662 }, { "epoch": 2.75, "learning_rate": 3.1608781716753646e-06, "loss": 0.3065, "step": 7663 }, { "epoch": 2.75, "learning_rate": 3.1599784056145403e-06, "loss": 0.4021, "step": 7664 }, { "epoch": 2.75, "learning_rate": 3.1590786395537165e-06, "loss": 0.3743, "step": 7665 }, { "epoch": 2.75, "learning_rate": 3.158178873492892e-06, "loss": 0.2531, "step": 7666 }, { "epoch": 2.75, "learning_rate": 3.157279107432068e-06, "loss": 0.3188, "step": 7667 }, { "epoch": 2.75, "learning_rate": 3.1563793413712434e-06, "loss": 0.3662, "step": 7668 }, { "epoch": 2.75, "learning_rate": 3.1554795753104196e-06, "loss": 0.3412, "step": 7669 }, { "epoch": 2.75, "learning_rate": 3.1545798092495954e-06, "loss": 0.3394, "step": 7670 }, { "epoch": 2.75, "learning_rate": 3.153680043188771e-06, "loss": 0.3063, "step": 7671 }, { "epoch": 2.75, "learning_rate": 3.152780277127947e-06, "loss": 0.29, "step": 7672 }, { "epoch": 2.75, "learning_rate": 3.151880511067123e-06, "loss": 0.3606, "step": 7673 }, { "epoch": 2.75, "learning_rate": 3.1509807450062984e-06, "loss": 0.2981, "step": 7674 }, { "epoch": 2.75, "learning_rate": 3.1500809789454746e-06, "loss": 0.2772, "step": 7675 }, { "epoch": 2.76, "learning_rate": 3.1491812128846504e-06, "loss": 0.3095, "step": 7676 }, { "epoch": 2.76, "learning_rate": 3.148281446823826e-06, "loss": 0.2792, "step": 7677 }, { "epoch": 2.76, "learning_rate": 3.147381680763002e-06, "loss": 0.3451, "step": 7678 }, { "epoch": 2.76, "learning_rate": 3.1464819147021773e-06, "loss": 0.3212, "step": 7679 }, { "epoch": 2.76, "learning_rate": 3.1455821486413535e-06, "loss": 0.3609, "step": 7680 }, { "epoch": 2.76, "learning_rate": 3.1446823825805297e-06, "loss": 0.2832, "step": 7681 }, { "epoch": 2.76, "learning_rate": 3.143782616519705e-06, "loss": 0.3099, "step": 7682 }, { "epoch": 2.76, "learning_rate": 3.142882850458881e-06, "loss": 0.3027, "step": 7683 }, { "epoch": 2.76, "learning_rate": 3.141983084398057e-06, "loss": 0.3197, "step": 7684 }, { "epoch": 2.76, "learning_rate": 3.1410833183372323e-06, "loss": 0.3664, "step": 7685 }, { "epoch": 2.76, "learning_rate": 3.1401835522764085e-06, "loss": 0.33, "step": 7686 }, { "epoch": 2.76, "learning_rate": 3.139283786215584e-06, "loss": 0.345, "step": 7687 }, { "epoch": 2.76, "learning_rate": 3.13838402015476e-06, "loss": 0.3554, "step": 7688 }, { "epoch": 2.76, "learning_rate": 3.137484254093936e-06, "loss": 0.4242, "step": 7689 }, { "epoch": 2.76, "learning_rate": 3.1365844880331116e-06, "loss": 0.3147, "step": 7690 }, { "epoch": 2.76, "learning_rate": 3.1356847219722874e-06, "loss": 0.2775, "step": 7691 }, { "epoch": 2.76, "learning_rate": 3.1347849559114636e-06, "loss": 0.4191, "step": 7692 }, { "epoch": 2.76, "learning_rate": 3.133885189850639e-06, "loss": 0.3266, "step": 7693 }, { "epoch": 2.76, "learning_rate": 3.132985423789815e-06, "loss": 0.2785, "step": 7694 }, { "epoch": 2.76, "learning_rate": 3.132085657728991e-06, "loss": 0.3589, "step": 7695 }, { "epoch": 2.76, "learning_rate": 3.1311858916681666e-06, "loss": 0.3751, "step": 7696 }, { "epoch": 2.76, "learning_rate": 3.1302861256073424e-06, "loss": 0.3915, "step": 7697 }, { "epoch": 2.76, "learning_rate": 3.1293863595465177e-06, "loss": 0.3352, "step": 7698 }, { "epoch": 2.76, "learning_rate": 3.128486593485694e-06, "loss": 0.2489, "step": 7699 }, { "epoch": 2.76, "learning_rate": 3.12758682742487e-06, "loss": 0.326, "step": 7700 }, { "epoch": 2.76, "learning_rate": 3.1266870613640455e-06, "loss": 0.2964, "step": 7701 }, { "epoch": 2.76, "learning_rate": 3.1257872953032212e-06, "loss": 0.3606, "step": 7702 }, { "epoch": 2.76, "learning_rate": 3.1248875292423974e-06, "loss": 0.3475, "step": 7703 }, { "epoch": 2.77, "learning_rate": 3.1239877631815728e-06, "loss": 0.3242, "step": 7704 }, { "epoch": 2.77, "learning_rate": 3.123087997120749e-06, "loss": 0.3612, "step": 7705 }, { "epoch": 2.77, "learning_rate": 3.1221882310599243e-06, "loss": 0.2948, "step": 7706 }, { "epoch": 2.77, "learning_rate": 3.1212884649991005e-06, "loss": 0.306, "step": 7707 }, { "epoch": 2.77, "learning_rate": 3.1203886989382763e-06, "loss": 0.2972, "step": 7708 }, { "epoch": 2.77, "learning_rate": 3.119488932877452e-06, "loss": 0.3252, "step": 7709 }, { "epoch": 2.77, "learning_rate": 3.118589166816628e-06, "loss": 0.267, "step": 7710 }, { "epoch": 2.77, "learning_rate": 3.117689400755804e-06, "loss": 0.3504, "step": 7711 }, { "epoch": 2.77, "learning_rate": 3.1167896346949794e-06, "loss": 0.304, "step": 7712 }, { "epoch": 2.77, "learning_rate": 3.1158898686341555e-06, "loss": 0.3865, "step": 7713 }, { "epoch": 2.77, "learning_rate": 3.1149901025733313e-06, "loss": 0.3388, "step": 7714 }, { "epoch": 2.77, "learning_rate": 3.114090336512507e-06, "loss": 0.3576, "step": 7715 }, { "epoch": 2.77, "learning_rate": 3.113190570451683e-06, "loss": 0.3529, "step": 7716 }, { "epoch": 2.77, "learning_rate": 3.112290804390858e-06, "loss": 0.3054, "step": 7717 }, { "epoch": 2.77, "learning_rate": 3.1113910383300344e-06, "loss": 0.299, "step": 7718 }, { "epoch": 2.77, "learning_rate": 3.1104912722692106e-06, "loss": 0.2981, "step": 7719 }, { "epoch": 2.77, "learning_rate": 3.109591506208386e-06, "loss": 0.3344, "step": 7720 }, { "epoch": 2.77, "learning_rate": 3.108691740147562e-06, "loss": 0.2992, "step": 7721 }, { "epoch": 2.77, "learning_rate": 3.107791974086738e-06, "loss": 0.3213, "step": 7722 }, { "epoch": 2.77, "learning_rate": 3.1068922080259132e-06, "loss": 0.3176, "step": 7723 }, { "epoch": 2.77, "learning_rate": 3.1059924419650894e-06, "loss": 0.3537, "step": 7724 }, { "epoch": 2.77, "learning_rate": 3.1050926759042648e-06, "loss": 0.3306, "step": 7725 }, { "epoch": 2.77, "learning_rate": 3.104192909843441e-06, "loss": 0.3476, "step": 7726 }, { "epoch": 2.77, "learning_rate": 3.1032931437826167e-06, "loss": 0.2789, "step": 7727 }, { "epoch": 2.77, "learning_rate": 3.1023933777217925e-06, "loss": 0.3887, "step": 7728 }, { "epoch": 2.77, "learning_rate": 3.1014936116609683e-06, "loss": 0.3746, "step": 7729 }, { "epoch": 2.77, "learning_rate": 3.1005938456001445e-06, "loss": 0.2999, "step": 7730 }, { "epoch": 2.77, "learning_rate": 3.09969407953932e-06, "loss": 0.3482, "step": 7731 }, { "epoch": 2.78, "learning_rate": 3.098794313478496e-06, "loss": 0.2805, "step": 7732 }, { "epoch": 2.78, "learning_rate": 3.0978945474176718e-06, "loss": 0.3051, "step": 7733 }, { "epoch": 2.78, "learning_rate": 3.0969947813568475e-06, "loss": 0.3198, "step": 7734 }, { "epoch": 2.78, "learning_rate": 3.0960950152960233e-06, "loss": 0.3417, "step": 7735 }, { "epoch": 2.78, "learning_rate": 3.095195249235199e-06, "loss": 0.3707, "step": 7736 }, { "epoch": 2.78, "learning_rate": 3.094295483174375e-06, "loss": 0.3079, "step": 7737 }, { "epoch": 2.78, "learning_rate": 3.093395717113551e-06, "loss": 0.3214, "step": 7738 }, { "epoch": 2.78, "learning_rate": 3.0924959510527264e-06, "loss": 0.2904, "step": 7739 }, { "epoch": 2.78, "learning_rate": 3.0915961849919026e-06, "loss": 0.3052, "step": 7740 }, { "epoch": 2.78, "learning_rate": 3.0906964189310783e-06, "loss": 0.3136, "step": 7741 }, { "epoch": 2.78, "learning_rate": 3.0897966528702537e-06, "loss": 0.3404, "step": 7742 }, { "epoch": 2.78, "learning_rate": 3.08889688680943e-06, "loss": 0.3066, "step": 7743 }, { "epoch": 2.78, "learning_rate": 3.0879971207486052e-06, "loss": 0.3782, "step": 7744 }, { "epoch": 2.78, "learning_rate": 3.0870973546877814e-06, "loss": 0.3336, "step": 7745 }, { "epoch": 2.78, "learning_rate": 3.086197588626957e-06, "loss": 0.3542, "step": 7746 }, { "epoch": 2.78, "learning_rate": 3.085297822566133e-06, "loss": 0.3352, "step": 7747 }, { "epoch": 2.78, "learning_rate": 3.0843980565053087e-06, "loss": 0.3432, "step": 7748 }, { "epoch": 2.78, "learning_rate": 3.083498290444485e-06, "loss": 0.3089, "step": 7749 }, { "epoch": 2.78, "learning_rate": 3.0825985243836603e-06, "loss": 0.4074, "step": 7750 }, { "epoch": 2.78, "learning_rate": 3.0816987583228365e-06, "loss": 0.2969, "step": 7751 }, { "epoch": 2.78, "learning_rate": 3.0807989922620122e-06, "loss": 0.3428, "step": 7752 }, { "epoch": 2.78, "learning_rate": 3.079899226201188e-06, "loss": 0.3165, "step": 7753 }, { "epoch": 2.78, "learning_rate": 3.0789994601403638e-06, "loss": 0.3272, "step": 7754 }, { "epoch": 2.78, "learning_rate": 3.0780996940795395e-06, "loss": 0.3476, "step": 7755 }, { "epoch": 2.78, "learning_rate": 3.0771999280187153e-06, "loss": 0.4379, "step": 7756 }, { "epoch": 2.78, "learning_rate": 3.0763001619578915e-06, "loss": 0.3342, "step": 7757 }, { "epoch": 2.78, "learning_rate": 3.075400395897067e-06, "loss": 0.355, "step": 7758 }, { "epoch": 2.78, "learning_rate": 3.074500629836243e-06, "loss": 0.2843, "step": 7759 }, { "epoch": 2.79, "learning_rate": 3.073600863775419e-06, "loss": 0.3458, "step": 7760 }, { "epoch": 2.79, "learning_rate": 3.072701097714594e-06, "loss": 0.2894, "step": 7761 }, { "epoch": 2.79, "learning_rate": 3.0718013316537703e-06, "loss": 0.355, "step": 7762 }, { "epoch": 2.79, "learning_rate": 3.0709015655929457e-06, "loss": 0.3684, "step": 7763 }, { "epoch": 2.79, "learning_rate": 3.070001799532122e-06, "loss": 0.3165, "step": 7764 }, { "epoch": 2.79, "learning_rate": 3.069102033471298e-06, "loss": 0.3934, "step": 7765 }, { "epoch": 2.79, "learning_rate": 3.0682022674104734e-06, "loss": 0.3382, "step": 7766 }, { "epoch": 2.79, "learning_rate": 3.067302501349649e-06, "loss": 0.3075, "step": 7767 }, { "epoch": 2.79, "learning_rate": 3.0664027352888254e-06, "loss": 0.4043, "step": 7768 }, { "epoch": 2.79, "learning_rate": 3.0655029692280007e-06, "loss": 0.3393, "step": 7769 }, { "epoch": 2.79, "learning_rate": 3.064603203167177e-06, "loss": 0.432, "step": 7770 }, { "epoch": 2.79, "learning_rate": 3.0637034371063527e-06, "loss": 0.3447, "step": 7771 }, { "epoch": 2.79, "learning_rate": 3.0628036710455284e-06, "loss": 0.3855, "step": 7772 }, { "epoch": 2.79, "learning_rate": 3.0619039049847042e-06, "loss": 0.2953, "step": 7773 }, { "epoch": 2.79, "learning_rate": 3.06100413892388e-06, "loss": 0.3837, "step": 7774 }, { "epoch": 2.79, "learning_rate": 3.0601043728630558e-06, "loss": 0.3216, "step": 7775 }, { "epoch": 2.79, "learning_rate": 3.059204606802232e-06, "loss": 0.2772, "step": 7776 }, { "epoch": 2.79, "learning_rate": 3.0583048407414073e-06, "loss": 0.3016, "step": 7777 }, { "epoch": 2.79, "learning_rate": 3.0574050746805835e-06, "loss": 0.3515, "step": 7778 }, { "epoch": 2.79, "learning_rate": 3.0565053086197593e-06, "loss": 0.4035, "step": 7779 }, { "epoch": 2.79, "learning_rate": 3.055605542558935e-06, "loss": 0.2956, "step": 7780 }, { "epoch": 2.79, "learning_rate": 3.054705776498111e-06, "loss": 0.3885, "step": 7781 }, { "epoch": 2.79, "learning_rate": 3.053806010437286e-06, "loss": 0.3001, "step": 7782 }, { "epoch": 2.79, "learning_rate": 3.0529062443764623e-06, "loss": 0.3006, "step": 7783 }, { "epoch": 2.79, "learning_rate": 3.0520064783156385e-06, "loss": 0.4027, "step": 7784 }, { "epoch": 2.79, "learning_rate": 3.051106712254814e-06, "loss": 0.4013, "step": 7785 }, { "epoch": 2.79, "learning_rate": 3.0502069461939896e-06, "loss": 0.4083, "step": 7786 }, { "epoch": 2.8, "learning_rate": 3.049307180133166e-06, "loss": 0.3458, "step": 7787 }, { "epoch": 2.8, "learning_rate": 3.048407414072341e-06, "loss": 0.3881, "step": 7788 }, { "epoch": 2.8, "learning_rate": 3.0475076480115174e-06, "loss": 0.3718, "step": 7789 }, { "epoch": 2.8, "learning_rate": 3.0466078819506927e-06, "loss": 0.2821, "step": 7790 }, { "epoch": 2.8, "learning_rate": 3.045708115889869e-06, "loss": 0.2818, "step": 7791 }, { "epoch": 2.8, "learning_rate": 3.0448083498290447e-06, "loss": 0.3154, "step": 7792 }, { "epoch": 2.8, "learning_rate": 3.0439085837682204e-06, "loss": 0.3883, "step": 7793 }, { "epoch": 2.8, "learning_rate": 3.043008817707396e-06, "loss": 0.339, "step": 7794 }, { "epoch": 2.8, "learning_rate": 3.0421090516465724e-06, "loss": 0.3723, "step": 7795 }, { "epoch": 2.8, "learning_rate": 3.0412092855857477e-06, "loss": 0.2647, "step": 7796 }, { "epoch": 2.8, "learning_rate": 3.040309519524924e-06, "loss": 0.338, "step": 7797 }, { "epoch": 2.8, "learning_rate": 3.0394097534640997e-06, "loss": 0.3091, "step": 7798 }, { "epoch": 2.8, "learning_rate": 3.0385099874032755e-06, "loss": 0.3672, "step": 7799 }, { "epoch": 2.8, "learning_rate": 3.0376102213424512e-06, "loss": 0.3864, "step": 7800 }, { "epoch": 2.8, "learning_rate": 3.0367104552816266e-06, "loss": 0.4383, "step": 7801 }, { "epoch": 2.8, "learning_rate": 3.0358106892208028e-06, "loss": 0.3251, "step": 7802 }, { "epoch": 2.8, "learning_rate": 3.034910923159979e-06, "loss": 0.3001, "step": 7803 }, { "epoch": 2.8, "learning_rate": 3.0340111570991543e-06, "loss": 0.3529, "step": 7804 }, { "epoch": 2.8, "learning_rate": 3.03311139103833e-06, "loss": 0.2545, "step": 7805 }, { "epoch": 2.8, "learning_rate": 3.0322116249775063e-06, "loss": 0.3744, "step": 7806 }, { "epoch": 2.8, "learning_rate": 3.0313118589166816e-06, "loss": 0.3361, "step": 7807 }, { "epoch": 2.8, "learning_rate": 3.030412092855858e-06, "loss": 0.3178, "step": 7808 }, { "epoch": 2.8, "learning_rate": 3.029512326795033e-06, "loss": 0.3435, "step": 7809 }, { "epoch": 2.8, "learning_rate": 3.0286125607342094e-06, "loss": 0.3756, "step": 7810 }, { "epoch": 2.8, "learning_rate": 3.027712794673385e-06, "loss": 0.4034, "step": 7811 }, { "epoch": 2.8, "learning_rate": 3.026813028612561e-06, "loss": 0.3587, "step": 7812 }, { "epoch": 2.8, "learning_rate": 3.0259132625517367e-06, "loss": 0.3045, "step": 7813 }, { "epoch": 2.8, "learning_rate": 3.025013496490913e-06, "loss": 0.3947, "step": 7814 }, { "epoch": 2.81, "learning_rate": 3.024113730430088e-06, "loss": 0.3492, "step": 7815 }, { "epoch": 2.81, "learning_rate": 3.0232139643692644e-06, "loss": 0.3321, "step": 7816 }, { "epoch": 2.81, "learning_rate": 3.02231419830844e-06, "loss": 0.3537, "step": 7817 }, { "epoch": 2.81, "learning_rate": 3.021414432247616e-06, "loss": 0.3458, "step": 7818 }, { "epoch": 2.81, "learning_rate": 3.0205146661867917e-06, "loss": 0.3293, "step": 7819 }, { "epoch": 2.81, "learning_rate": 3.019614900125967e-06, "loss": 0.3117, "step": 7820 }, { "epoch": 2.81, "learning_rate": 3.0187151340651432e-06, "loss": 0.3248, "step": 7821 }, { "epoch": 2.81, "learning_rate": 3.0178153680043194e-06, "loss": 0.358, "step": 7822 }, { "epoch": 2.81, "learning_rate": 3.0169156019434948e-06, "loss": 0.3257, "step": 7823 }, { "epoch": 2.81, "learning_rate": 3.016015835882671e-06, "loss": 0.3664, "step": 7824 }, { "epoch": 2.81, "learning_rate": 3.0151160698218467e-06, "loss": 0.2756, "step": 7825 }, { "epoch": 2.81, "learning_rate": 3.014216303761022e-06, "loss": 0.3671, "step": 7826 }, { "epoch": 2.81, "learning_rate": 3.0133165377001983e-06, "loss": 0.3608, "step": 7827 }, { "epoch": 2.81, "learning_rate": 3.0124167716393736e-06, "loss": 0.3283, "step": 7828 }, { "epoch": 2.81, "learning_rate": 3.01151700557855e-06, "loss": 0.3198, "step": 7829 }, { "epoch": 2.81, "learning_rate": 3.0106172395177256e-06, "loss": 0.35, "step": 7830 }, { "epoch": 2.81, "learning_rate": 3.0097174734569013e-06, "loss": 0.3832, "step": 7831 }, { "epoch": 2.81, "learning_rate": 3.008817707396077e-06, "loss": 0.3924, "step": 7832 }, { "epoch": 2.81, "learning_rate": 3.0079179413352533e-06, "loss": 0.304, "step": 7833 }, { "epoch": 2.81, "learning_rate": 3.0070181752744287e-06, "loss": 0.3054, "step": 7834 }, { "epoch": 2.81, "learning_rate": 3.006118409213605e-06, "loss": 0.3752, "step": 7835 }, { "epoch": 2.81, "learning_rate": 3.0052186431527806e-06, "loss": 0.3079, "step": 7836 }, { "epoch": 2.81, "learning_rate": 3.0043188770919564e-06, "loss": 0.2758, "step": 7837 }, { "epoch": 2.81, "learning_rate": 3.003419111031132e-06, "loss": 0.3433, "step": 7838 }, { "epoch": 2.81, "learning_rate": 3.002519344970308e-06, "loss": 0.3571, "step": 7839 }, { "epoch": 2.81, "learning_rate": 3.0016195789094837e-06, "loss": 0.3792, "step": 7840 }, { "epoch": 2.81, "learning_rate": 3.00071981284866e-06, "loss": 0.3146, "step": 7841 }, { "epoch": 2.81, "learning_rate": 2.9998200467878352e-06, "loss": 0.3577, "step": 7842 }, { "epoch": 2.82, "learning_rate": 2.9989202807270114e-06, "loss": 0.2995, "step": 7843 }, { "epoch": 2.82, "learning_rate": 2.998020514666187e-06, "loss": 0.3084, "step": 7844 }, { "epoch": 2.82, "learning_rate": 2.9971207486053625e-06, "loss": 0.3113, "step": 7845 }, { "epoch": 2.82, "learning_rate": 2.9962209825445387e-06, "loss": 0.3319, "step": 7846 }, { "epoch": 2.82, "learning_rate": 2.995321216483714e-06, "loss": 0.3265, "step": 7847 }, { "epoch": 2.82, "learning_rate": 2.9944214504228903e-06, "loss": 0.3382, "step": 7848 }, { "epoch": 2.82, "learning_rate": 2.9935216843620665e-06, "loss": 0.3629, "step": 7849 }, { "epoch": 2.82, "learning_rate": 2.992621918301242e-06, "loss": 0.3141, "step": 7850 }, { "epoch": 2.82, "learning_rate": 2.9917221522404176e-06, "loss": 0.3156, "step": 7851 }, { "epoch": 2.82, "learning_rate": 2.9908223861795938e-06, "loss": 0.2826, "step": 7852 }, { "epoch": 2.82, "learning_rate": 2.989922620118769e-06, "loss": 0.3665, "step": 7853 }, { "epoch": 2.82, "learning_rate": 2.9890228540579453e-06, "loss": 0.3824, "step": 7854 }, { "epoch": 2.82, "learning_rate": 2.988123087997121e-06, "loss": 0.2715, "step": 7855 }, { "epoch": 2.82, "learning_rate": 2.987223321936297e-06, "loss": 0.3099, "step": 7856 }, { "epoch": 2.82, "learning_rate": 2.9863235558754726e-06, "loss": 0.3476, "step": 7857 }, { "epoch": 2.82, "learning_rate": 2.9854237898146484e-06, "loss": 0.3553, "step": 7858 }, { "epoch": 2.82, "learning_rate": 2.984524023753824e-06, "loss": 0.3488, "step": 7859 }, { "epoch": 2.82, "learning_rate": 2.9836242576930003e-06, "loss": 0.2942, "step": 7860 }, { "epoch": 2.82, "learning_rate": 2.9827244916321757e-06, "loss": 0.3499, "step": 7861 }, { "epoch": 2.82, "learning_rate": 2.981824725571352e-06, "loss": 0.2892, "step": 7862 }, { "epoch": 2.82, "learning_rate": 2.9809249595105276e-06, "loss": 0.3331, "step": 7863 }, { "epoch": 2.82, "learning_rate": 2.9800251934497034e-06, "loss": 0.3902, "step": 7864 }, { "epoch": 2.82, "learning_rate": 2.979125427388879e-06, "loss": 0.3747, "step": 7865 }, { "epoch": 2.82, "learning_rate": 2.9782256613280545e-06, "loss": 0.3202, "step": 7866 }, { "epoch": 2.82, "learning_rate": 2.9773258952672307e-06, "loss": 0.2726, "step": 7867 }, { "epoch": 2.82, "learning_rate": 2.976426129206407e-06, "loss": 0.3076, "step": 7868 }, { "epoch": 2.82, "learning_rate": 2.9755263631455823e-06, "loss": 0.2842, "step": 7869 }, { "epoch": 2.82, "learning_rate": 2.974626597084758e-06, "loss": 0.3024, "step": 7870 }, { "epoch": 2.83, "learning_rate": 2.9737268310239342e-06, "loss": 0.3205, "step": 7871 }, { "epoch": 2.83, "learning_rate": 2.9728270649631096e-06, "loss": 0.3585, "step": 7872 }, { "epoch": 2.83, "learning_rate": 2.9719272989022858e-06, "loss": 0.3497, "step": 7873 }, { "epoch": 2.83, "learning_rate": 2.9710275328414615e-06, "loss": 0.3461, "step": 7874 }, { "epoch": 2.83, "learning_rate": 2.9701277667806373e-06, "loss": 0.3256, "step": 7875 }, { "epoch": 2.83, "learning_rate": 2.969228000719813e-06, "loss": 0.3082, "step": 7876 }, { "epoch": 2.83, "learning_rate": 2.968328234658989e-06, "loss": 0.2905, "step": 7877 }, { "epoch": 2.83, "learning_rate": 2.9674284685981646e-06, "loss": 0.3168, "step": 7878 }, { "epoch": 2.83, "learning_rate": 2.966528702537341e-06, "loss": 0.324, "step": 7879 }, { "epoch": 2.83, "learning_rate": 2.965628936476516e-06, "loss": 0.3282, "step": 7880 }, { "epoch": 2.83, "learning_rate": 2.9647291704156923e-06, "loss": 0.3955, "step": 7881 }, { "epoch": 2.83, "learning_rate": 2.963829404354868e-06, "loss": 0.281, "step": 7882 }, { "epoch": 2.83, "learning_rate": 2.962929638294044e-06, "loss": 0.3755, "step": 7883 }, { "epoch": 2.83, "learning_rate": 2.9620298722332196e-06, "loss": 0.3612, "step": 7884 }, { "epoch": 2.83, "learning_rate": 2.961130106172395e-06, "loss": 0.3279, "step": 7885 }, { "epoch": 2.83, "learning_rate": 2.960230340111571e-06, "loss": 0.3573, "step": 7886 }, { "epoch": 2.83, "learning_rate": 2.9593305740507474e-06, "loss": 0.3731, "step": 7887 }, { "epoch": 2.83, "learning_rate": 2.9584308079899227e-06, "loss": 0.2973, "step": 7888 }, { "epoch": 2.83, "learning_rate": 2.9575310419290985e-06, "loss": 0.3486, "step": 7889 }, { "epoch": 2.83, "learning_rate": 2.9566312758682747e-06, "loss": 0.3463, "step": 7890 }, { "epoch": 2.83, "learning_rate": 2.95573150980745e-06, "loss": 0.3876, "step": 7891 }, { "epoch": 2.83, "learning_rate": 2.954831743746626e-06, "loss": 0.3048, "step": 7892 }, { "epoch": 2.83, "learning_rate": 2.9539319776858024e-06, "loss": 0.3391, "step": 7893 }, { "epoch": 2.83, "learning_rate": 2.9530322116249777e-06, "loss": 0.3708, "step": 7894 }, { "epoch": 2.83, "learning_rate": 2.9521324455641535e-06, "loss": 0.3365, "step": 7895 }, { "epoch": 2.83, "learning_rate": 2.9512326795033293e-06, "loss": 0.3191, "step": 7896 }, { "epoch": 2.83, "learning_rate": 2.950332913442505e-06, "loss": 0.3177, "step": 7897 }, { "epoch": 2.83, "learning_rate": 2.9494331473816812e-06, "loss": 0.3403, "step": 7898 }, { "epoch": 2.84, "learning_rate": 2.9485333813208566e-06, "loss": 0.3167, "step": 7899 }, { "epoch": 2.84, "learning_rate": 2.9476336152600328e-06, "loss": 0.3688, "step": 7900 }, { "epoch": 2.84, "learning_rate": 2.9467338491992086e-06, "loss": 0.3143, "step": 7901 }, { "epoch": 2.84, "learning_rate": 2.9458340831383843e-06, "loss": 0.3308, "step": 7902 }, { "epoch": 2.84, "learning_rate": 2.94493431707756e-06, "loss": 0.3472, "step": 7903 }, { "epoch": 2.84, "learning_rate": 2.9440345510167354e-06, "loss": 0.372, "step": 7904 }, { "epoch": 2.84, "learning_rate": 2.9431347849559116e-06, "loss": 0.3343, "step": 7905 }, { "epoch": 2.84, "learning_rate": 2.942235018895088e-06, "loss": 0.3187, "step": 7906 }, { "epoch": 2.84, "learning_rate": 2.941335252834263e-06, "loss": 0.3567, "step": 7907 }, { "epoch": 2.84, "learning_rate": 2.9404354867734394e-06, "loss": 0.2927, "step": 7908 }, { "epoch": 2.84, "learning_rate": 2.939535720712615e-06, "loss": 0.3828, "step": 7909 }, { "epoch": 2.84, "learning_rate": 2.9386359546517905e-06, "loss": 0.2831, "step": 7910 }, { "epoch": 2.84, "learning_rate": 2.9377361885909667e-06, "loss": 0.3474, "step": 7911 }, { "epoch": 2.84, "learning_rate": 2.936836422530143e-06, "loss": 0.361, "step": 7912 }, { "epoch": 2.84, "learning_rate": 2.935936656469318e-06, "loss": 0.278, "step": 7913 }, { "epoch": 2.84, "learning_rate": 2.935036890408494e-06, "loss": 0.3581, "step": 7914 }, { "epoch": 2.84, "learning_rate": 2.9341371243476697e-06, "loss": 0.3534, "step": 7915 }, { "epoch": 2.84, "learning_rate": 2.9332373582868455e-06, "loss": 0.3595, "step": 7916 }, { "epoch": 2.84, "learning_rate": 2.9323375922260217e-06, "loss": 0.3398, "step": 7917 }, { "epoch": 2.84, "learning_rate": 2.931437826165197e-06, "loss": 0.3445, "step": 7918 }, { "epoch": 2.84, "learning_rate": 2.9305380601043732e-06, "loss": 0.2802, "step": 7919 }, { "epoch": 2.84, "learning_rate": 2.929638294043549e-06, "loss": 0.3881, "step": 7920 }, { "epoch": 2.84, "learning_rate": 2.9287385279827248e-06, "loss": 0.3235, "step": 7921 }, { "epoch": 2.84, "learning_rate": 2.9278387619219005e-06, "loss": 0.3524, "step": 7922 }, { "epoch": 2.84, "learning_rate": 2.9269389958610763e-06, "loss": 0.3402, "step": 7923 }, { "epoch": 2.84, "learning_rate": 2.926039229800252e-06, "loss": 0.3347, "step": 7924 }, { "epoch": 2.84, "learning_rate": 2.9251394637394283e-06, "loss": 0.4659, "step": 7925 }, { "epoch": 2.84, "learning_rate": 2.9242396976786036e-06, "loss": 0.3453, "step": 7926 }, { "epoch": 2.85, "learning_rate": 2.92333993161778e-06, "loss": 0.3654, "step": 7927 }, { "epoch": 2.85, "learning_rate": 2.9224401655569556e-06, "loss": 0.3228, "step": 7928 }, { "epoch": 2.85, "learning_rate": 2.921540399496131e-06, "loss": 0.3589, "step": 7929 }, { "epoch": 2.85, "learning_rate": 2.920640633435307e-06, "loss": 0.3821, "step": 7930 }, { "epoch": 2.85, "learning_rate": 2.9197408673744833e-06, "loss": 0.3162, "step": 7931 }, { "epoch": 2.85, "learning_rate": 2.9188411013136587e-06, "loss": 0.3316, "step": 7932 }, { "epoch": 2.85, "learning_rate": 2.9179413352528344e-06, "loss": 0.356, "step": 7933 }, { "epoch": 2.85, "learning_rate": 2.91704156919201e-06, "loss": 0.2752, "step": 7934 }, { "epoch": 2.85, "learning_rate": 2.916141803131186e-06, "loss": 0.2778, "step": 7935 }, { "epoch": 2.85, "learning_rate": 2.915242037070362e-06, "loss": 0.3602, "step": 7936 }, { "epoch": 2.85, "learning_rate": 2.9143422710095375e-06, "loss": 0.3187, "step": 7937 }, { "epoch": 2.85, "learning_rate": 2.9134425049487137e-06, "loss": 0.3554, "step": 7938 }, { "epoch": 2.85, "learning_rate": 2.9125427388878895e-06, "loss": 0.3668, "step": 7939 }, { "epoch": 2.85, "learning_rate": 2.9116429728270652e-06, "loss": 0.4101, "step": 7940 }, { "epoch": 2.85, "learning_rate": 2.910743206766241e-06, "loss": 0.2684, "step": 7941 }, { "epoch": 2.85, "learning_rate": 2.9098434407054168e-06, "loss": 0.3763, "step": 7942 }, { "epoch": 2.85, "learning_rate": 2.9089436746445925e-06, "loss": 0.3091, "step": 7943 }, { "epoch": 2.85, "learning_rate": 2.9080439085837687e-06, "loss": 0.3754, "step": 7944 }, { "epoch": 2.85, "learning_rate": 2.907144142522944e-06, "loss": 0.2728, "step": 7945 }, { "epoch": 2.85, "learning_rate": 2.9062443764621203e-06, "loss": 0.2791, "step": 7946 }, { "epoch": 2.85, "learning_rate": 2.905344610401296e-06, "loss": 0.2959, "step": 7947 }, { "epoch": 2.85, "learning_rate": 2.9044448443404714e-06, "loss": 0.3542, "step": 7948 }, { "epoch": 2.85, "learning_rate": 2.9035450782796476e-06, "loss": 0.254, "step": 7949 }, { "epoch": 2.85, "learning_rate": 2.9026453122188238e-06, "loss": 0.2723, "step": 7950 }, { "epoch": 2.85, "learning_rate": 2.901745546157999e-06, "loss": 0.3607, "step": 7951 }, { "epoch": 2.85, "learning_rate": 2.9008457800971753e-06, "loss": 0.3101, "step": 7952 }, { "epoch": 2.85, "learning_rate": 2.8999460140363506e-06, "loss": 0.3669, "step": 7953 }, { "epoch": 2.85, "learning_rate": 2.8990462479755264e-06, "loss": 0.3614, "step": 7954 }, { "epoch": 2.86, "learning_rate": 2.8981464819147026e-06, "loss": 0.3413, "step": 7955 }, { "epoch": 2.86, "learning_rate": 2.897246715853878e-06, "loss": 0.4135, "step": 7956 }, { "epoch": 2.86, "learning_rate": 2.896346949793054e-06, "loss": 0.3601, "step": 7957 }, { "epoch": 2.86, "learning_rate": 2.89544718373223e-06, "loss": 0.3753, "step": 7958 }, { "epoch": 2.86, "learning_rate": 2.8945474176714057e-06, "loss": 0.2933, "step": 7959 }, { "epoch": 2.86, "learning_rate": 2.8936476516105815e-06, "loss": 0.3956, "step": 7960 }, { "epoch": 2.86, "learning_rate": 2.8927478855497572e-06, "loss": 0.3036, "step": 7961 }, { "epoch": 2.86, "learning_rate": 2.891848119488933e-06, "loss": 0.3761, "step": 7962 }, { "epoch": 2.86, "learning_rate": 2.890948353428109e-06, "loss": 0.3427, "step": 7963 }, { "epoch": 2.86, "learning_rate": 2.8900485873672845e-06, "loss": 0.3063, "step": 7964 }, { "epoch": 2.86, "learning_rate": 2.8891488213064607e-06, "loss": 0.2934, "step": 7965 }, { "epoch": 2.86, "learning_rate": 2.8882490552456365e-06, "loss": 0.3299, "step": 7966 }, { "epoch": 2.86, "learning_rate": 2.8873492891848123e-06, "loss": 0.3214, "step": 7967 }, { "epoch": 2.86, "learning_rate": 2.886449523123988e-06, "loss": 0.3596, "step": 7968 }, { "epoch": 2.86, "learning_rate": 2.8855497570631634e-06, "loss": 0.3285, "step": 7969 }, { "epoch": 2.86, "learning_rate": 2.8846499910023396e-06, "loss": 0.2863, "step": 7970 }, { "epoch": 2.86, "learning_rate": 2.8837502249415158e-06, "loss": 0.3829, "step": 7971 }, { "epoch": 2.86, "learning_rate": 2.882850458880691e-06, "loss": 0.3512, "step": 7972 }, { "epoch": 2.86, "learning_rate": 2.881950692819867e-06, "loss": 0.2978, "step": 7973 }, { "epoch": 2.86, "learning_rate": 2.881050926759043e-06, "loss": 0.4061, "step": 7974 }, { "epoch": 2.86, "learning_rate": 2.8801511606982184e-06, "loss": 0.3422, "step": 7975 }, { "epoch": 2.86, "learning_rate": 2.8792513946373946e-06, "loss": 0.312, "step": 7976 }, { "epoch": 2.86, "learning_rate": 2.8783516285765704e-06, "loss": 0.3703, "step": 7977 }, { "epoch": 2.86, "learning_rate": 2.877451862515746e-06, "loss": 0.3039, "step": 7978 }, { "epoch": 2.86, "learning_rate": 2.876552096454922e-06, "loss": 0.341, "step": 7979 }, { "epoch": 2.86, "learning_rate": 2.8756523303940977e-06, "loss": 0.4285, "step": 7980 }, { "epoch": 2.86, "learning_rate": 2.8747525643332734e-06, "loss": 0.3841, "step": 7981 }, { "epoch": 2.87, "learning_rate": 2.8738527982724496e-06, "loss": 0.319, "step": 7982 }, { "epoch": 2.87, "learning_rate": 2.872953032211625e-06, "loss": 0.2943, "step": 7983 }, { "epoch": 2.87, "learning_rate": 2.872053266150801e-06, "loss": 0.2977, "step": 7984 }, { "epoch": 2.87, "learning_rate": 2.871153500089977e-06, "loss": 0.32, "step": 7985 }, { "epoch": 2.87, "learning_rate": 2.8702537340291527e-06, "loss": 0.3551, "step": 7986 }, { "epoch": 2.87, "learning_rate": 2.8693539679683285e-06, "loss": 0.424, "step": 7987 }, { "epoch": 2.87, "learning_rate": 2.868454201907504e-06, "loss": 0.3099, "step": 7988 }, { "epoch": 2.87, "learning_rate": 2.86755443584668e-06, "loss": 0.3155, "step": 7989 }, { "epoch": 2.87, "learning_rate": 2.866654669785856e-06, "loss": 0.3153, "step": 7990 }, { "epoch": 2.87, "learning_rate": 2.8657549037250316e-06, "loss": 0.3184, "step": 7991 }, { "epoch": 2.87, "learning_rate": 2.8648551376642073e-06, "loss": 0.3209, "step": 7992 }, { "epoch": 2.87, "learning_rate": 2.8639553716033835e-06, "loss": 0.3233, "step": 7993 }, { "epoch": 2.87, "learning_rate": 2.863055605542559e-06, "loss": 0.3058, "step": 7994 }, { "epoch": 2.87, "learning_rate": 2.862155839481735e-06, "loss": 0.3719, "step": 7995 }, { "epoch": 2.87, "learning_rate": 2.8612560734209112e-06, "loss": 0.2858, "step": 7996 }, { "epoch": 2.87, "learning_rate": 2.8603563073600866e-06, "loss": 0.3241, "step": 7997 }, { "epoch": 2.87, "learning_rate": 2.8594565412992624e-06, "loss": 0.3911, "step": 7998 }, { "epoch": 2.87, "learning_rate": 2.858556775238438e-06, "loss": 0.305, "step": 7999 }, { "epoch": 2.87, "learning_rate": 2.857657009177614e-06, "loss": 0.32, "step": 8000 } ], "max_steps": 11144, "num_train_epochs": 4, "total_flos": 7.709989705663644e+18, "trial_name": null, "trial_params": null }